{"configuration_count":243,"release_id":"gpqa-clean-results-v0.1.0","rows":[{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_792c41959f3f82ccceec08b9342f234829872842cf3fcc4036e28481504fa2e3","canonical_evidence_variant_id":"gpqa_evd_v1_d9e541024a37f5842559fb4df9b1b4f420f0361a976926db9a55a2692529c67d","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_792c41959f3f82ccceec08b9342f234829872842cf3fcc4036e28481504fa2e3","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d9e541024a37f5842559fb4df9b1b4f420f0361a976926db9a55a2692529c67d"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_00b750e02386f4541fe325efb0a8449df2b72a52b8dd40a582a7a752470e3a3d","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"openai/gpt-5-mini-2025-08-07","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-5-mini-2025-08-07_medium","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"71.654040%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.023192741878083397","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"CFNzodoTSVVVgp9PUXftUP"},"evidence_class":"direct_compatibility","model_configuration":"gpt-5-mini-2025-08-07_medium","organization":"OpenAI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_ca45a83a12af45522007a22520a874ec87c40ccee3a08cb3975b2f21710f3022","canonical_evidence_variant_id":"gpqa_evd_v1_b1fbd626729590cba9a7306c14eab42a7a0746d386cef5fb1bc3ed7c5ff05c7e","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_ca45a83a12af45522007a22520a874ec87c40ccee3a08cb3975b2f21710f3022","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_b1fbd626729590cba9a7306c14eab42a7a0746d386cef5fb1bc3ed7c5ff05c7e"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_00e9ae21c6c0f764c862027370ab0a348eb54d424a8aff0cb6a305a38618706b","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"mistral/open-mistral-7b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"mistral","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"open-mistral-7b","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"13.226010%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.009888086213040367","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"5XuVuHeUDekHqjUy2P8gww"},"evidence_class":"direct_compatibility","model_configuration":"open-mistral-7b","organization":"Mistral AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_aa060b8f563068bef04d82a383a85c2b89b4a1e4df2ea9536d763f3048f49a7e","canonical_evidence_variant_id":"gpqa_evd_v1_24dac4fd7128a97a3ed42af765c34b8cfe6575cf7653b97c34a6784dcb7f10e8","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_aa060b8f563068bef04d82a383a85c2b89b4a1e4df2ea9536d763f3048f49a7e","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_24dac4fd7128a97a3ed42af765c34b8cfe6575cf7653b97c34a6784dcb7f10e8"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-12-13","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_0101b0f415adc0308a5aeb576bd938ac45457f4a2b41b2f0ec765daf4da164d7","configuration_identity":{"context_policy":"source_default","effort":"xhigh","model_snapshot":"openai/gpt-5.2-2025-12-11","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.3","source_model_version":"gpt-5.2-2025-12-11_xhigh","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"+0.014141 pp","published_source_score_percent":"91.400000%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.018","status":"accepted_within_0_05pp"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Ejo6Fy7XYxUDcCwbejfcYX"},"evidence_class":"direct_strict","model_configuration":"gpt-5.2-2025-12-11_xhigh","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_cb8f2ea222449667ac21c1d4c90b2900c2ef60423d279cbdf95dad2b69a10da9","canonical_evidence_variant_id":"gpqa_evd_v1_7e75c86ea6628a8b743dede137652fad934dfc638af5c3b3141848607ff5cb29","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_cb8f2ea222449667ac21c1d4c90b2900c2ef60423d279cbdf95dad2b69a10da9","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_7e75c86ea6628a8b743dede137652fad934dfc638af5c3b3141848607ff5cb29"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-17","configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","configuration_id":"gpqa_cfg_v1_02806112b3fe0147c940db998df739ae579b592620fa54621d718d9ceef20010","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"grok/grok-4.6","output_policy":"no_output_ceiling","prompt_contract":"gpqa_native_answer_choice_v1","protocol":"mean@1","provider_route":"grok","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","generation_config_sha256":"4db33ee3aca837b8beb02b9e0714860bd517fe9ef65c086a876468bb7b9f3928","profile":"grok-4-6-high","task_sha256":"2501e59954a9e53479bc5e44bf9923db905866019f16c385ea3f27483b3697d0"}},"details":{"configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","evidence_label":"Independent native run","execution":{"automatic_retries":0,"fresh_calls":true,"output_ceiling":"none"},"reconciliation":{"paired_original_role":"Reconstructed from the same accepted all-198 native attempt universe as the Clean score.","published_source_score_percent":null,"status":"not_applicable_independent_run"},"recovery":{"kind":"none"},"score_source":"Independent native answer-safe projection","source_kind_label":"Independent project-native run","source_run_id":"gpqa-native-grok-4-6-high-all198-20260816-r3"},"evidence_class":"project_native","model_configuration":"Grok 4.6 (high)","organization":"xAI","protocol":"mean@1","validation_profile":"strict_all_normal"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_39f293bd3e07a520963865e115317a17f07e383f5d5743a1e7d98917d7cfb91d","canonical_evidence_variant_id":"gpqa_evd_v1_dfca09330d2c1339c6a87d2d4f2f9a6a498a98ef89616bbe63fbeece5147e5b4","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_39f293bd3e07a520963865e115317a17f07e383f5d5743a1e7d98917d7cfb91d","evidence_class":"targeted_recovery","evidence_variant_id":"gpqa_evd_v1_dfca09330d2c1339c6a87d2d4f2f9a6a498a98ef89616bbe63fbeece5147e5b4"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-16","configuration_date_basis":"validated composite generated_at metadata","configuration_id":"gpqa_cfg_v1_028330d0cbeadddca05767705c635f8f34aa6f91322bce88dd9bd97968f4a05a","configuration_identity":{"context_policy":"32k","effort":"unspecified","model_snapshot":"anthropic/claude-sonnet-4-5-20250929","output_policy":"source_artifact_policy_plus_uncapped_recovery","prompt_contract":"epoch_gpqa_diamond_source_plus_recovery_v1","protocol":"mean@1","provider_route":"anthropic_plus_openrouter","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","configuration_match":"exact_source_header_generate_config","execution_match":"cross_carrier","model_identity_basis":"declared_cross_carrier_substitution","source_model_version":"claude-sonnet-4-5-20250929_32K"}},"details":{"configuration_date_basis":"validated composite generated_at metadata","evidence_label":"Epoch base plus targeted transport recovery","reconciliation":{"difference_percentage_points":"0.000000 pp","paired_original_role":"Reconstructed from the completed source-plus-recovery attempt universe; not relabelled as the original Epoch aggregate.","published_source_score_percent":"81.725888%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027603867020557334","status":"exact"},"recovery":{"configuration_match":"exact_source_header_generate_config","execution_match":"cross_carrier","kind":"exact_targeted_transport_recovery","model_identity_basis":"declared_cross_carrier_substitution","output_ceiling":"none","recovered_pair_count":1},"score_source":"Accepted source choice scores plus exact targeted repair pairs","source_kind_label":"Epoch base plus targeted transport-recovery composite","source_run_id":"ii6JE9j57eSbuswEsJArTa"},"evidence_class":"targeted_recovery","model_configuration":"claude-sonnet-4-5-20250929_32K","organization":"Anthropic","protocol":"mean@1","validation_profile":"complete_with_epoch_targeted_transport_recovery"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_e8970f2c7575f713b293970e1f5428551b3b95a4954105e5ed533652f454d32c","canonical_evidence_variant_id":"gpqa_evd_v1_ce73963bfa92d1e064a11add6056c83c4b69e6383db4a60c7b869d6de29398af","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e8970f2c7575f713b293970e1f5428551b3b95a4954105e5ed533652f454d32c","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_ce73963bfa92d1e064a11add6056c83c4b69e6383db4a60c7b869d6de29398af"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-10-16","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_066a71969e005dcbca3c374019b97a0455b621a9da160724d16808d4ee7bdd9e","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-haiku-4-5-20251001","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-haiku-4-5-20251001","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"60.479798%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.028154535023020826","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"cgnopvpB6mzsfb9yGnYK2z"},"evidence_class":"direct_compatibility","model_configuration":"claude-haiku-4-5-20251001","organization":"Anthropic","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_9a16ed69ee959ae436a9aa2b1e45b6ca842413696d7ec819e54ae9ee8b5d7042","canonical_evidence_variant_id":"gpqa_evd_v1_931218e5efb2449dd5bab087c3c9e1a70dbb16add0778a884e14a740f9553cdc","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_9a16ed69ee959ae436a9aa2b1e45b6ca842413696d7ec819e54ae9ee8b5d7042","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_931218e5efb2449dd5bab087c3c9e1a70dbb16add0778a884e14a740f9553cdc"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-11","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_06fc60637355bfe507b394a67bd3cf0abc7c538efdf32fa6908642999f53ce6d","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/o4-mini-2025-04-16","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"o4-mini-2025-04-16_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"75.252525%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.030746300742124484","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"XwRFKnuHv5qdUp9m5uTAEb"},"evidence_class":"direct_strict","model_configuration":"o4-mini-2025-04-16_low","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_47a174a410566d1d53939cd53d53bb0a9d35d6e888f166d50d2f8be2551da5ec","canonical_evidence_variant_id":"gpqa_evd_v1_7e4b216cd43665ee130c572b462caad73d67ffa01fa6214de04878c9e9f3ff52","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_47a174a410566d1d53939cd53d53bb0a9d35d6e888f166d50d2f8be2551da5ec","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_7e4b216cd43665ee130c572b462caad73d67ffa01fa6214de04878c9e9f3ff52"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_0780c6b4ba641120d8293909ee3b41d284184448300fe82845ccd82ed21a22a8","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/claude-fable-5","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-fable-5_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"78.787879%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.029126522834586777","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"U5PsUdX2ak4kJ2hgnQkWnX"},"evidence_class":"direct_strict","model_configuration":"claude-fable-5_low","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_178cd776ee64853cf54697ca6eae803398dbded170cc215ba350fbb0affce958","canonical_evidence_variant_id":"gpqa_evd_v1_8325f704751ccd8d53ca9996d72cb19b61a6f9d27e9243b9b4a2a474023c2c2c","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_178cd776ee64853cf54697ca6eae803398dbded170cc215ba350fbb0affce958","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_8325f704751ccd8d53ca9996d72cb19b61a6f9d27e9243b9b4a2a474023c2c2c"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-12-11","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_08b0ee7e2e105ac8e5ed96cd2f47a42dff3bba9cd0b1c886aac128321d37784f","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"openai/gpt-5.2-2025-12-11","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.3","source_model_version":"gpt-5.2-2025-12-11_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"88.194444%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.01913858277633924","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"3Q6osT8EYezayG7yfxkSpr"},"evidence_class":"direct_strict","model_configuration":"gpt-5.2-2025-12-11_high","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_5e39fd30f2fea819992cae499e5283ef32d1247ded1e8b67c327c11717274ea5","canonical_evidence_variant_id":"gpqa_evd_v1_19777c5b3118044034b1d9244e71cd858bbdb22012be1097744b6064e7d9b84e","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_5e39fd30f2fea819992cae499e5283ef32d1247ded1e8b67c327c11717274ea5","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_19777c5b3118044034b1d9244e71cd858bbdb22012be1097744b6064e7d9b84e"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_08bd8befe0cade55c7342c9f96b69d027677412b414b670247ac45b3f919f540","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-3.5-turbo-1106","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-3.5-turbo-1106","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"28.030303%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.018720372561004013","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"enhRcQmfN6RFB2zu7R3Uxr"},"evidence_class":"direct_compatibility","model_configuration":"gpt-3.5-turbo-1106","organization":"OpenAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_dbb9f2125cc09e996ef57fb0c3a86c5039790a3328752ace68ac73c52fd858b8","canonical_evidence_variant_id":"gpqa_evd_v1_4c3fc342c1502e5abf9c4a73c86e03d99ee9af7721e0cac0d97df58e9b65a1fc","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_dbb9f2125cc09e996ef57fb0c3a86c5039790a3328752ace68ac73c52fd858b8","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_4c3fc342c1502e5abf9c4a73c86e03d99ee9af7721e0cac0d97df58e9b65a1fc"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_08f66b9ca9530d2a229621c663831d6ac6cfa888b11230239eb991363750e754","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"mistral/ministral-8b-2410","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"mistral","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"ministral-8b-2410","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"27.146465%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.020863706671553732","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"DGEzZA6nvTne4MEGSuo3Tm"},"evidence_class":"direct_compatibility","model_configuration":"ministral-8b-2410","organization":"Mistral AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_17a8008accd99ed44e41a2068583eb613fb01e9f0327b7fd4f50db4a94c9ea26","canonical_evidence_variant_id":"gpqa_evd_v1_0abc2af9bad5ea0ddf358b460cf0e986e41f057ec160ae64237b3bdf3c239496","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_17a8008accd99ed44e41a2068583eb613fb01e9f0327b7fd4f50db4a94c9ea26","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_0abc2af9bad5ea0ddf358b460cf0e986e41f057ec160ae64237b3bdf3c239496"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-17","configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","configuration_id":"gpqa_cfg_v1_0acf0885980bdc183084c5deb8f6341eda2074c91344c8d89bebce44801f4394","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"openai/gpt-5.6-terra","output_policy":"no_output_ceiling","prompt_contract":"gpqa_native_answer_choice_v1","protocol":"mean@1","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","generation_config_sha256":"7a9193a3f2ff640e27d2e75b0bc5d50ce1cf053bbac7e7465825822222e18080","profile":"gpt-5-6-terra-max","task_sha256":"2501e59954a9e53479bc5e44bf9923db905866019f16c385ea3f27483b3697d0"}},"details":{"configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","evidence_label":"Independent native run","execution":{"automatic_retries":0,"fresh_calls":true,"output_ceiling":"none"},"reconciliation":{"paired_original_role":"Reconstructed from the same accepted all-198 native attempt universe as the Clean score.","published_source_score_percent":null,"status":"not_applicable_independent_run"},"recovery":{"kind":"none"},"score_source":"Independent native answer-safe projection","source_kind_label":"Independent project-native run","source_run_id":"gpqa-native-gpt-5-6-terra-max-all198-20260816-r1"},"evidence_class":"project_native","model_configuration":"GPT-5.6 Terra (max)","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_all_normal"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_46b41b2c72fc6db332d75c94d8c8133134f7350e46ceb3cfd161d0cb03b21859","canonical_evidence_variant_id":"gpqa_evd_v1_bb1e2d5bf137420f1d4aca14a95963c00a74060207679f087ecccfed1bb2f326","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_46b41b2c72fc6db332d75c94d8c8133134f7350e46ceb3cfd161d0cb03b21859","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_bb1e2d5bf137420f1d4aca14a95963c00a74060207679f087ecccfed1bb2f326"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-30","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_0aecd21ee0bc485147dfb311f84fdcd13ab5be37a49413da6200f0e3a3f8a9e4","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/Qwen/Qwen2.5-32B-Instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"qwen2.5-32b-instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"46.085859%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02595783004579065","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"d6JihAFVqUy3u9suNt4P3L"},"evidence_class":"direct_compatibility","model_configuration":"qwen2.5-32b-instruct","organization":"Alibaba","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_88e47969d46da216fc880a0887465e0ae54b02d16fbd32933b18dd75f3d8ab1a","canonical_evidence_variant_id":"gpqa_evd_v1_f51c42739dfd105b28ddb7ab432843e2f89aabfec69eda041292bdd0a1e2af8a","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_88e47969d46da216fc880a0887465e0ae54b02d16fbd32933b18dd75f3d8ab1a","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f51c42739dfd105b28ddb7ab432843e2f89aabfec69eda041292bdd0a1e2af8a"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-10-28","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_0bdf7182d5f04c31013cbf0810c5d35c04d75be94de6c6030b1439ce8d8f0278","configuration_identity":{"context_policy":"59k","effort":"unspecified","model_snapshot":"anthropic/claude-sonnet-4-5-20250929","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.0","source_model_version":"claude-sonnet-4-5-20250929_59K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"82.323232%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027178752639044908","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"XaGuWDxPtqVvyxqdMncKFj"},"evidence_class":"direct_compatibility","model_configuration":"claude-sonnet-4-5-20250929_59K","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_8643462e411041d61c09b3c93ef9495747941781878ee9de507ccce76e9782d8","canonical_evidence_variant_id":"gpqa_evd_v1_36f4bde2c0391d4f5f492f677c315435eb53db6d72b8a92e2d19b35ec273d51f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_8643462e411041d61c09b3c93ef9495747941781878ee9de507ccce76e9782d8","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_36f4bde2c0391d4f5f492f677c315435eb53db6d72b8a92e2d19b35ec273d51f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-11-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_0ec60fbcfbd0a5dc9ad3b1d16e08bf7c3678cacdd15194446fb742b017dc743b","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"openai/gpt-5.1","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.1","source_model_version":"gpt-5.1-2025-11-13_medium","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"85.037879%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.021250422082249816","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"V3YQtSEnQ3TPZEPPgkCKQM"},"evidence_class":"direct_strict","model_configuration":"gpt-5.1-2025-11-13_medium","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_f8b55dfa9d69764aed33a4fa19c62e976ae46b2f92ece4e21abdbf78bbdfdcc0","canonical_evidence_variant_id":"gpqa_evd_v1_b264bec2407ede2a026598c2260682c73d6cfc7412a6e7a9596aa89c6b092fd9","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_f8b55dfa9d69764aed33a4fa19c62e976ae46b2f92ece4e21abdbf78bbdfdcc0","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_b264bec2407ede2a026598c2260682c73d6cfc7412a6e7a9596aa89c6b092fd9"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_114b61ba4b46f7fbd32882dba76813956cbdfdc2782903c93f510f22a3b1fc54","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"epoch/claude-opus-4-7","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-opus-4-7_max","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"86.363636%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02445015597318985","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Ju7pUkkwYVj6ugi9Ri5Jjp"},"evidence_class":"direct_strict","model_configuration":"claude-opus-4-7_max","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_8feb80681a8b5c93af2d6f169a81b92d5277ab2284b7c0eff8bf8507f840f244","canonical_evidence_variant_id":"gpqa_evd_v1_eefbb4d726b36d38e57429d4802a9358be6e10e1c00ab79f8cbae4ccdf8e81ea","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_8feb80681a8b5c93af2d6f169a81b92d5277ab2284b7c0eff8bf8507f840f244","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_eefbb4d726b36d38e57429d4802a9358be6e10e1c00ab79f8cbae4ccdf8e81ea"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_139cbaca23c931c65e3d28182449928faf23acb7ce5481a3b9bb7577edf379de","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"mistral/ministral-3b-2410","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"mistral","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"ministral-3b-2410","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"25.252525%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.019305614722797293","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"bJV4D8SEo9zMJPEYTTXJNZ"},"evidence_class":"direct_compatibility","model_configuration":"ministral-3b-2410","organization":"Mistral AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_ca6fd8d7a3c517eaf6b033d194ee471bc1f4d5157a213767a2f4ccbe777baab6","canonical_evidence_variant_id":"gpqa_evd_v1_936530850bcd6fd6b0da2509916a80b6b158953fefbe7488b118ccb7d763982d","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_ca6fd8d7a3c517eaf6b033d194ee471bc1f4d5157a213767a2f4ccbe777baab6","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_936530850bcd6fd6b0da2509916a80b6b158953fefbe7488b118ccb7d763982d"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-10-30","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_153bcc443e2c2d97f408c94b5455fe8eded12a0a5d244c70e8fe4d7e9700b5e0","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"openai/gpt-5-nano-2025-08-07","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.1","source_model_version":"gpt-5-nano-2025-08-07_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"69.444444%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027594508998251045","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"2pSDjMSGDh4HsAxkqUuNQP"},"evidence_class":"direct_strict","model_configuration":"gpt-5-nano-2025-08-07_high","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":2,"canonical_evaluation_run_id":"gpqa_run_v1_4ac367431f74dbfcd34e9f8b7882af29e08138f13020d3729e1d3fae75aa95a5","canonical_evidence_variant_id":"gpqa_evd_v1_805e3318ae823eaf6a2c7bfeaae0d35cc1194cff3f81ab3b1a84975eb384cc33","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_4ac367431f74dbfcd34e9f8b7882af29e08138f13020d3729e1d3fae75aa95a5","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_805e3318ae823eaf6a2c7bfeaae0d35cc1194cff3f81ab3b1a84975eb384cc33"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-12-11","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_16b158ba0a71b9966a8c0e86927eb1da15d0ee18a198e905d951edda2884f53d","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"alibaba/qwen3-235b-a22b-thinking-2507","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@2","provider_route":"alibaba","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"2","benchmark_version":"1.0.3","source_model_version":"Qwen3-235B-A22B-Thinking-2507","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"80.050505%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026000933515631548","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"gu39ndv9QxWP3NRxSUJRaX"},"evidence_class":"direct_strict","model_configuration":"Qwen3-235B-A22B-Thinking-2507","organization":"Alibaba","protocol":"mean@2","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_251c6f643f2b6c39f02c07ee62c2ad525cd70fbd0586de0d985613a4b9e99080","canonical_evidence_variant_id":"gpqa_evd_v1_666c33b98fbae48e7c49136578b6c602e384ae49094b2d2c4f0afa3756b38def","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_251c6f643f2b6c39f02c07ee62c2ad525cd70fbd0586de0d985613a4b9e99080","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_666c33b98fbae48e7c49136578b6c602e384ae49094b2d2c4f0afa3756b38def"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-09-29","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_172a5bf1e81d7ea7354117b64e82e5a863ddd3138848c51d12934571427fcfd5","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-sonnet-4-5-20250929","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-sonnet-4-5-20250929","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"73.737374%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.031353050095330834","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"fsZmzw8jjyUgMQGwxodBwN"},"evidence_class":"direct_compatibility","model_configuration":"claude-sonnet-4-5-20250929","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_570b0a07b609fc7ca3c59be7f25e500bef27dce875b4a2e540fc1d41e286c20c","canonical_evidence_variant_id":"gpqa_evd_v1_fa1e942e227b0d36cfcc50b1a9f58790f11c13327aa920e4cf1e6580d6cd1a34","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_570b0a07b609fc7ca3c59be7f25e500bef27dce875b4a2e540fc1d41e286c20c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_fa1e942e227b0d36cfcc50b1a9f58790f11c13327aa920e4cf1e6580d6cd1a34"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_1752768af2b13dfd2f0a9df6802e29b14ab9387775bdd045c8ea1c437f8607f9","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"epoch/claude-opus-4-6","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-opus-4-6_max","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"88.383838%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02282888177524936","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"JcQMgXdZKU54u52B8TGQKp"},"evidence_class":"direct_strict","model_configuration":"claude-opus-4-6_max","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_eed4e81138caaf681f745732a4045a78eddc7f7c37bc6f04fad9417e2024140e","canonical_evidence_variant_id":"gpqa_evd_v1_564911d072eaf17aab93846ce51f5c9aa06b468e0e1dc1a8e86bdeb3b5db0502","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_eed4e81138caaf681f745732a4045a78eddc7f7c37bc6f04fad9417e2024140e","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_564911d072eaf17aab93846ce51f5c9aa06b468e0e1dc1a8e86bdeb3b5db0502"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-10","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_17d1971555900db9f02b08fe2063bb81c1f406eb3f99ded57d6366a966344171","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/glm-5.1","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"glm-5.1","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"89.898990%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.021469735576055356","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"CBbeMKRRSeRsXnvSfn7Png"},"evidence_class":"direct_strict","model_configuration":"glm-5.1","organization":"Z.ai (Zhipu AI)","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_45aea34803bca828427c03f1a9fd3840200b233b1ee16c224e8d982f1c636674","canonical_evidence_variant_id":"gpqa_evd_v1_9560f7cb9c978b6e8f919d7675abd559fa0708481f617c01f7e4dc2cf26536da","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_45aea34803bca828427c03f1a9fd3840200b233b1ee16c224e8d982f1c636674","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_9560f7cb9c978b6e8f919d7675abd559fa0708481f617c01f7e4dc2cf26536da"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_18c8f9728ba3bccb250dcce8c3a5d52ac4be8aae99733f291cecb08e35b283db","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/claude-opus-5","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-opus-5_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"87.878788%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02325315795194205","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"WfaJS4L7Xdz5ZBXgJMbRGb"},"evidence_class":"direct_strict","model_configuration":"claude-opus-5_low","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_eec81bd37fd42fc5713cf1fcb43fa19910c58f9ff22360b4b92859007e4be14b","canonical_evidence_variant_id":"gpqa_evd_v1_03c820e2b115ed14cf9d3a0b8a3ad2a1fa0e09efdfda594c4c7028fbb7d8b484","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_eec81bd37fd42fc5713cf1fcb43fa19910c58f9ff22360b4b92859007e4be14b","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_03c820e2b115ed14cf9d3a0b8a3ad2a1fa0e09efdfda594c4c7028fbb7d8b484"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-21","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_18fba063eae1be42dd8389b74872396a60463f5d291bf70c9532cbf3f5cd9e92","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/mistralai/Mixtral-8x7B-Instruct-v0.1","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Mixtral-8x7B-Instruct-v0.1","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"30.587121%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.01937059603941351","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Sn2qbB2HiySe7NcG28jnyM"},"evidence_class":"direct_compatibility","model_configuration":"Mixtral-8x7B-Instruct-v0.1","organization":"Mistral AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_fa8ef9d620dc65f782ec32bf95577bb258e8db52c4da6295af0ab70a7ee2f4d3","canonical_evidence_variant_id":"gpqa_evd_v1_da000f42edd17dda4f211a367384b858c2cee397de1b48a51af535f720ab6df0","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_fa8ef9d620dc65f782ec32bf95577bb258e8db52c4da6295af0ab70a7ee2f4d3","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_da000f42edd17dda4f211a367384b858c2cee397de1b48a51af535f720ab6df0"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_1d5d4f300a0ef5bd1a8578e997798bae60cb1a0710e2c0c8a37c01ddcaec92c5","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-1.5-pro-002","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemini-1.5-pro-002","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"57.228535%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027959709636504244","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"9TrLEEw8TvsyugyCjwvUvD"},"evidence_class":"direct_compatibility","model_configuration":"gemini-1.5-pro-002","organization":"Google DeepMind","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_5d709ffdeedc728b74e4f1e90b1d5e6f71ee0d50aff14b612c3554a41b38afbe","canonical_evidence_variant_id":"gpqa_evd_v1_f35e032ebccb00ab309725ac6ba4027a25988d59ec638705940d71460ce3c2e2","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_5d709ffdeedc728b74e4f1e90b1d5e6f71ee0d50aff14b612c3554a41b38afbe","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f35e032ebccb00ab309725ac6ba4027a25988d59ec638705940d71460ce3c2e2"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-27","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_1d957091393fa4b70f270be10a21999490fcaeae1de9aa2a1d5f9a5196c80ea0","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-1.5-flash-001","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemini-1.5-flash-001","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"40.372475%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.023852475604024478","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Hc2ycMP6DyCZxCPjKbZFQU"},"evidence_class":"direct_compatibility","model_configuration":"gemini-1.5-flash-001","organization":"Google DeepMind","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_814a3fcb9d9f82fcc6444bd32c8d97992010cbcaf704280c1f4b682f692cb48a","canonical_evidence_variant_id":"gpqa_evd_v1_40b347dd19f8820810d1e9ab840304435b17392d485ef2f88a5dbc93537370cb","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_814a3fcb9d9f82fcc6444bd32c8d97992010cbcaf704280c1f4b682f692cb48a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_40b347dd19f8820810d1e9ab840304435b17392d485ef2f88a5dbc93537370cb"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_1df69af1ab0ce9fcc3c822f72011312d0cb297c6ba8c857f0c29740914d904fd","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/qwen3.6-35b-a3b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.6-35b-a3b","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"83.838384%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02622591986362927","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"EWAyxFsRUvuu4NpmmLqWkn"},"evidence_class":"direct_strict","model_configuration":"qwen3.6-35b-a3b","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_a969aad50a7ffba1c59209fd68db8a0d8ef59fb0d2a1724ee6e52c351c4d87eb","canonical_evidence_variant_id":"gpqa_evd_v1_f5d2a74fcae956da52dad90eff78f31de476b4545ec1e0fe6e3bc93288edd280","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_a969aad50a7ffba1c59209fd68db8a0d8ef59fb0d2a1724ee6e52c351c4d87eb","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_f5d2a74fcae956da52dad90eff78f31de476b4545ec1e0fe6e3bc93288edd280"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-15","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_1e7553c1137d1a896e5336d0f9e36f600eb3f8c1981c9496e39923a6dceac58a","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"epoch/gpt-5.4-2026-03-05","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.4-2026-03-05_medium","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"88.888889%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02239078763821682","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"ManJk43Kf9VuXowNqoDrov"},"evidence_class":"direct_strict","model_configuration":"gpt-5.4-2026-03-05_medium","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_75401b34ec4d25283ce1395b7ece31ff4524cd468cdeba3429407a8da6e8cd8a","canonical_evidence_variant_id":"gpqa_evd_v1_078a842bcb3e1914b31eb65db36bff3347e6d5fd2e782038b8737f54e5506c57","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_75401b34ec4d25283ce1395b7ece31ff4524cd468cdeba3429407a8da6e8cd8a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_078a842bcb3e1914b31eb65db36bff3347e6d5fd2e782038b8737f54e5506c57"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-15","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_1ebdf14944b16525b20d369d7989c36385e707ef96468f681ac093d3ef05807e","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/o3-2025-04-16","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"o3-2025-04-16_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"79.797980%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.028606204289229834","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"AZf6N4kygJGHPsSUkYtCba"},"evidence_class":"direct_strict","model_configuration":"o3-2025-04-16_low","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_95c1a3eb132c968f38d952bf90086d0976307651cb68b1ff6001e60b40d893e1","canonical_evidence_variant_id":"gpqa_evd_v1_b545ee33cdf7dee96b37c36a674516728f9bcc9057aa22a83543d7ae75160dfe","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_95c1a3eb132c968f38d952bf90086d0976307651cb68b1ff6001e60b40d893e1","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_b545ee33cdf7dee96b37c36a674516728f9bcc9057aa22a83543d7ae75160dfe"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-05-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_202023fa5943ac9a7c62a2c1d63c955492f2ac82a4850bb837b96a4adefe8dbe","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"mistral/mistral-medium-2505","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"mistral","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"mistral-medium-2505","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"59.532828%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02824697229654645","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"FNMbfvEEm9PRnrJnrqk2PH"},"evidence_class":"direct_compatibility","model_configuration":"mistral-medium-2505","organization":"Mistral AI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_b9b889eeac2e0aa50f5bbb745478dcc5364150d9463dbd3c0d87da422abca143","canonical_evidence_variant_id":"gpqa_evd_v1_8c7c047ce4c5007a73283a6af356345a69e2810479df69b799ca9741cf9d627b","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_b9b889eeac2e0aa50f5bbb745478dcc5364150d9463dbd3c0d87da422abca143","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_8c7c047ce4c5007a73283a6af356345a69e2810479df69b799ca9741cf9d627b"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-17","configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","configuration_id":"gpqa_cfg_v1_211a09311a1e6c34a2d98ce45ee46c8aab95d0380c75c39c262f0d5dd6392e90","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"openai/gpt-5.6-luna","output_policy":"no_output_ceiling","prompt_contract":"gpqa_native_answer_choice_v1","protocol":"mean@1","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","generation_config_sha256":"7a9193a3f2ff640e27d2e75b0bc5d50ce1cf053bbac7e7465825822222e18080","profile":"gpt-5-6-luna-max","task_sha256":"2501e59954a9e53479bc5e44bf9923db905866019f16c385ea3f27483b3697d0"}},"details":{"configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","evidence_label":"Independent native run","execution":{"automatic_retries":0,"fresh_calls":true,"output_ceiling":"none"},"reconciliation":{"paired_original_role":"Reconstructed from the same accepted all-198 native attempt universe as the Clean score.","published_source_score_percent":null,"status":"not_applicable_independent_run"},"recovery":{"kind":"none"},"score_source":"Independent native answer-safe projection","source_kind_label":"Independent project-native run","source_run_id":"gpqa-native-gpt-5-6-luna-max-all198-20260816-r1"},"evidence_class":"project_native","model_configuration":"GPT-5.6 Luna (max)","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_all_normal"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_6abd58183594e22fae1475af34c1ec0277cf6a0009f12888db73e7a9f1435196","canonical_evidence_variant_id":"gpqa_evd_v1_dd25655b484a411f612db25e334d0f23ff76fa998151734f0b04f678aefce98c","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_6abd58183594e22fae1475af34c1ec0277cf6a0009f12888db73e7a9f1435196","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_dd25655b484a411f612db25e334d0f23ff76fa998151734f0b04f678aefce98c"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-02-13","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_211e3b3ef5c20703200d0b2a23fabe754e645ec0860ac579ec797da9ffd88edb","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"openai/o1-2024-12-17","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"o1-2024-12-17_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"76.767677%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.030088629490217445","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"NAkCFiFiDSN3NXMj7HMvMS"},"evidence_class":"direct_compatibility","model_configuration":"o1-2024-12-17_high","organization":"OpenAI","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_ef8f65ab4b75285fd76a2d42d67149ff08799bcd26e49c56b343c8354f4a32e0","canonical_evidence_variant_id":"gpqa_evd_v1_e842a50332b2f1acbf3c9ef5d7c9b255e8f2d0481ad5ca5137145ce0d9e4c36b","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_ef8f65ab4b75285fd76a2d42d67149ff08799bcd26e49c56b343c8354f4a32e0","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_e842a50332b2f1acbf3c9ef5d7c9b255e8f2d0481ad5ca5137145ce0d9e4c36b"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-05-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_24e612fff1c0d93cfbb8e6ac4c6cd800e066c2b55d58c67b0615a42278896a49","configuration_identity":{"context_policy":"32k","effort":"unspecified","model_snapshot":"anthropic/claude-sonnet-4-20250514","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-sonnet-4-20250514_32K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"78.282828%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02937661648494561","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"8hGbWtbpUjAQpebkKQfSi7"},"evidence_class":"direct_compatibility","model_configuration":"claude-sonnet-4-20250514_32K","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_b0f8ef5dae0e538e7eac0a1deed2e9f6384bc2db831c3b6b3a7f74c6158b0299","canonical_evidence_variant_id":"gpqa_evd_v1_1c5b515b9968d1975e565bea3e248fed8c94496529cf7af2ac2988b0bf61e97f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_b0f8ef5dae0e538e7eac0a1deed2e9f6384bc2db831c3b6b3a7f74c6158b0299","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_1c5b515b9968d1975e565bea3e248fed8c94496529cf7af2ac2988b0bf61e97f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-06-05","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_2512d5e4965dc519e0cfe48a64b1f7f33961e8ebe6df554e0815a1b4c98f9bd3","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-2.5-pro-preview-06-05","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemini-2.5-pro-preview-06-05","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"84.848485%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02554565042660364","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Q4qTzHnjLyiCCRvNpXioUq"},"evidence_class":"direct_compatibility","model_configuration":"gemini-2.5-pro-preview-06-05","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_88e1f487b6103967cbd712dc8da2e698590c83ea10f5e78720c5eb14f83de23c","canonical_evidence_variant_id":"gpqa_evd_v1_482634699135b907fc082cde0565498864c3156e5ad7d1ba04c678c58830a443","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_88e1f487b6103967cbd712dc8da2e698590c83ea10f5e78720c5eb14f83de23c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_482634699135b907fc082cde0565498864c3156e5ad7d1ba04c678c58830a443"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-01-28","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_25b6ff3864c66bd8fca985a9bbb8460bb4e531eb06f3d77a0f494f3b913eb510","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"fireworks/fireworks/kimi-k2p5","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"fireworks","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.4","source_model_version":"fireworks/kimi-k2p5","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"-0.036869 pp","published_source_score_percent":"87.600000%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.019","status":"accepted_within_0_05pp"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"FmqGUq87N4JcuSbo6Uu3U5"},"evidence_class":"direct_strict","model_configuration":"fireworks/kimi-k2p5","organization":"Moonshot","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_62e3bc8326e2e952542df4ddbbbef50524d07ab02ee9b73d21c72f2262cb5854","canonical_evidence_variant_id":"gpqa_evd_v1_d9bfe08a0212ed0f56f5bb6a95e50f5f4ed05ba6dcdfbbcdaede4306a327a80e","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_62e3bc8326e2e952542df4ddbbbef50524d07ab02ee9b73d21c72f2262cb5854","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_d9bfe08a0212ed0f56f5bb6a95e50f5f4ed05ba6dcdfbbcdaede4306a327a80e"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-10-30","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_25fa96ba13ac4c3502661f5ee3958b5fead6b8ed9a5a200ed41263fcfa970d6c","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"openai/gpt-5-mini-2025-08-07","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.1","source_model_version":"gpt-5-mini-2025-08-07_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"75.000000%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.022767295427113803","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"jP8EyLtPFsakgnsHe2oJM6"},"evidence_class":"direct_strict","model_configuration":"gpt-5-mini-2025-08-07_high","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_f340a350d20d1656f653d6ccc29e95473ba7ed10f7be6c073ed7e5162f3b724c","canonical_evidence_variant_id":"gpqa_evd_v1_98be177c252ba543c42683d01cf8c78b74a9f9260b1fc9f67ef99db0501a0fb0","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_f340a350d20d1656f653d6ccc29e95473ba7ed10f7be6c073ed7e5162f3b724c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_98be177c252ba543c42683d01cf8c78b74a9f9260b1fc9f67ef99db0501a0fb0"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_283043bba1e8117ffaa1f9aace7567853f92188831fd41efb8f9da8fb29f2f3f","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"epoch/gemini-3.1-pro-preview","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemini-3.1-pro-preview_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"94.444444%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.016319950700767385","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"BgmrBFsbsaD8rNpebv49Da"},"evidence_class":"direct_strict","model_configuration":"gemini-3.1-pro-preview_high","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_fb34c3033b6936e64041a9176019cecb087b661a1a9213f71544afacf98da583","canonical_evidence_variant_id":"gpqa_evd_v1_d7db75f0655adbf2456287bb2f2328e08012b49ca8e9325ed31716ad712ae1fc","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_fb34c3033b6936e64041a9176019cecb087b661a1a9213f71544afacf98da583","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d7db75f0655adbf2456287bb2f2328e08012b49ca8e9325ed31716ad712ae1fc"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-03-12","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_28ad4b722e7c76693ee321fb599eaa8bcce2126afeffed3cb5a53f356a229340","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-3-5-haiku-20241022","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-3-5-haiku-20241022","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"38.131313%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026351515390511556","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"JpLxrpvthx7A75sbaHZ9Et"},"evidence_class":"direct_compatibility","model_configuration":"claude-3-5-haiku-20241022","organization":"Anthropic","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_e5477e5a40319b819a2129e0271fad39918c8b13f8ef58231014c64fcde2340e","canonical_evidence_variant_id":"gpqa_evd_v1_436ff58a85ef500c61572711dda6fbcb0b8acae285c087098a98f0597eec8949","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e5477e5a40319b819a2129e0271fad39918c8b13f8ef58231014c64fcde2340e","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_436ff58a85ef500c61572711dda6fbcb0b8acae285c087098a98f0597eec8949"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_29e5aee9510b318690e575f5c2a1337b657fda114d776a3b4bbd40440424339b","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"mistral/open-mixtral-8x22b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"mistral","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"open-mixtral-8x22b","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"34.059343%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02068134779158647","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"ifAM3DJBdm5eU9Y2DjaEvW"},"evidence_class":"direct_compatibility","model_configuration":"open-mixtral-8x22b","organization":"Mistral AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_71d59e96915ee1c9f1e065c8a90b2578853a6ba822896791a753d1cd666ace6d","canonical_evidence_variant_id":"gpqa_evd_v1_8a451064c46ccfaf597b3db38e4a4e6a6ba9ef8e48e8606e2246b5d9d7929ab0","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_71d59e96915ee1c9f1e065c8a90b2578853a6ba822896791a753d1cd666ace6d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_8a451064c46ccfaf597b3db38e4a4e6a6ba9ef8e48e8606e2246b5d9d7929ab0"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-12-11","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_2c3d5349208f65704ed39c3181dd248e34283f155f2c5b6d9238c3eebf0c47b3","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"openai/gpt-5.2-2025-12-11","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.3","source_model_version":"gpt-5.2-2025-12-11_medium","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"87.878788%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.019090689343231743","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"MuhtxxvALeZZ37cSG5CVcc"},"evidence_class":"direct_strict","model_configuration":"gpt-5.2-2025-12-11_medium","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_14c867832a9a692bdf91006f3f7d4a8fba0a09e8d39e1af1a6b2d5abc320479b","canonical_evidence_variant_id":"gpqa_evd_v1_b81787e3541ebab349fe2a001a09bce4c911e8c9b8cd5af4ab1d06e2a7912d3f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_14c867832a9a692bdf91006f3f7d4a8fba0a09e8d39e1af1a6b2d5abc320479b","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_b81787e3541ebab349fe2a001a09bce4c911e8c9b8cd5af4ab1d06e2a7912d3f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-20","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_2d16e5d29fe93a718329e4d401a08f5d17e483de216571aa6a74b0aa9f945c11","configuration_identity":{"context_policy":"source_default","effort":"minimal","model_snapshot":"epoch/gpt-5-2025-08-07","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5-2025-08-07_minimal","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"71.717172%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.032087795587867514","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"n4wyt7wHjhDdjrGacE9smk"},"evidence_class":"direct_strict","model_configuration":"gpt-5-2025-08-07_minimal","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_13bf58c15c2cf11b057018e787247df5f9c2c70eb136c2c8b36b6ae8390dc068","canonical_evidence_variant_id":"gpqa_evd_v1_f9b5365b1625ed578c5bae2c8f17551d5fc0d98b23a29cc0614830bd68b68090","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_13bf58c15c2cf11b057018e787247df5f9c2c70eb136c2c8b36b6ae8390dc068","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_f9b5365b1625ed578c5bae2c8f17551d5fc0d98b23a29cc0614830bd68b68090"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-20","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_2e20ff4f87d82928ed38f791e0d054c52b489cc802e53dcdfc82bf4d51980434","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/o1-2024-12-17","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"o1-2024-12-17_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"74.242424%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.031156269519646826","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"hP6P5e9zuHPToaQTTQC9A2"},"evidence_class":"direct_strict","model_configuration":"o1-2024-12-17_low","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_b7cc1b2197543abe6f081a69f7b4b32fb658d1845a8a3af69f8d202cfdf1237a","canonical_evidence_variant_id":"gpqa_evd_v1_8d236b9fb51cb9b3b3c96cd1cca3a77a846671644b720ba0ea833a30af54db86","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_b7cc1b2197543abe6f081a69f7b4b32fb658d1845a8a3af69f8d202cfdf1237a","evidence_class":"targeted_recovery","evidence_variant_id":"gpqa_evd_v1_8d236b9fb51cb9b3b3c96cd1cca3a77a846671644b720ba0ea833a30af54db86"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-16","configuration_date_basis":"validated composite generated_at metadata","configuration_id":"gpqa_cfg_v1_2f4d76e7d0e2db97f02626aaf55d382c339a5f505c3d382cb8492f9239d5894c","configuration_identity":{"context_policy":"source_default","effort":"xhigh","model_snapshot":"openai/gpt-5.4-2026-03-05","output_policy":"source_artifact_policy_plus_uncapped_recovery","prompt_contract":"epoch_gpqa_diamond_source_plus_recovery_v1","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","configuration_match":"source_header_config_differs_csv_attested_recovery_profile","execution_match":"exact_source_route","model_identity_basis":"exact_historic_provider_route","source_model_version":"gpt-5.4-2026-03-05_xhigh"}},"details":{"configuration_date_basis":"validated composite generated_at metadata","evidence_label":"Epoch base plus targeted transport recovery","reconciliation":{"difference_percentage_points":"+0.339715 pp","paired_original_role":"Reconstructed from the completed source-plus-recovery attempt universe; not relabelled as the original Epoch aggregate.","published_source_score_percent":"93.300000%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.018","status":"material_conflict_explicit_inspect_source_of_truth"},"recovery":{"configuration_match":"source_header_config_differs_csv_attested_recovery_profile","execution_match":"exact_source_route","kind":"exact_targeted_transport_recovery","model_identity_basis":"exact_historic_provider_route","output_ceiling":"none","recovered_pair_count":43},"score_source":"Accepted source choice scores plus exact targeted repair pairs","source_kind_label":"Epoch base plus targeted transport-recovery composite","source_run_id":"fFatyce8UvpN7ZivdmrhAy"},"evidence_class":"targeted_recovery","model_configuration":"gpt-5.4-2026-03-05_xhigh","organization":"OpenAI","protocol":"mean@8","validation_profile":"complete_with_epoch_targeted_transport_recovery"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_64c1c9c06c465c18a63b94f520428d5764238b73e11f5ae1981125db3b9fe397","canonical_evidence_variant_id":"gpqa_evd_v1_c22e678b8f28d3c280fc6054b478e2a823af2cb92e1c790975684db0294cdbaa","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_64c1c9c06c465c18a63b94f520428d5764238b73e11f5ae1981125db3b9fe397","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_c22e678b8f28d3c280fc6054b478e2a823af2cb92e1c790975684db0294cdbaa"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-24","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_2fd227a793f3f537502469f4573a2c42c5e81c9dddbb46cd8400afad108ce3ec","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"together/deepseek-ai/DeepSeek-V3","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"together","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"DeepSeek-V3","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"56.534091%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02763990029950893","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Vwo3nMA8g2gGBCyTdKFMQJ"},"evidence_class":"direct_compatibility","model_configuration":"DeepSeek-V3","organization":"DeepSeek","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_1c256229a72afdf84b79498fc65ea39b9a088b233fc9fdd90ee8fb93496d40e8","canonical_evidence_variant_id":"gpqa_evd_v1_5982441681aa590d4c027f2ae0243505f6c56f206f82a611e9efe4c9202c1614","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_1c256229a72afdf84b79498fc65ea39b9a088b233fc9fdd90ee8fb93496d40e8","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_5982441681aa590d4c027f2ae0243505f6c56f206f82a611e9efe4c9202c1614"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-12-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_30fd0582710efe31443fff4d850caba23883e98a0a3df61e9dcbb85e5fb3ad3e","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-3-flash-preview","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.3","source_model_version":"gemini-3-flash-preview","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"83.207071%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.019036353018587762","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"oAu9CZSNPZgwiwKnoNpM8Q"},"evidence_class":"direct_strict","model_configuration":"gemini-3-flash-preview","organization":"Google DeepMind","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_a3f20303482242d26463b4d6d08959c23efb1301c1a7e6f91bd759ea7f96aae5","canonical_evidence_variant_id":"gpqa_evd_v1_27e29c8b2a4642558e73c768888c044653ecc7516793e83a566599713822dc17","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_a3f20303482242d26463b4d6d08959c23efb1301c1a7e6f91bd759ea7f96aae5","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_27e29c8b2a4642558e73c768888c044653ecc7516793e83a566599713822dc17"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-21","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_3166c2961eb47a41245ec120bbd6aa00af424fb8d90af97fb2168f58e13fb285","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/meta-llama/Meta-Llama-3-8B-Instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Meta-Llama-3-8B-Instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"26.073232%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.01703043361773576","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"iT3BBFWjsZk2rKHKeJDsM8"},"evidence_class":"direct_compatibility","model_configuration":"Meta-Llama-3-8B-Instruct","organization":"Meta AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_3ff6adeba5ec1314c7481630677123f787237af0ef3d9c0b7172929cc481810e","canonical_evidence_variant_id":"gpqa_evd_v1_f53d712a85d9dccef1d05e28777cb8ad3de3ff8e247d7446ec4170734ed60ef5","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_3ff6adeba5ec1314c7481630677123f787237af0ef3d9c0b7172929cc481810e","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_f53d712a85d9dccef1d05e28777cb8ad3de3ff8e247d7446ec4170734ed60ef5"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_319d02deeba5eff1fe5e7b0068499b112d9a54efc9456189fc1777dbacdee7a6","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/gemini-3.5-flash","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemini-3.5-flash_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"88.888889%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02239078763821682","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"eiTYduCDNq3HgwTg4ZhEw6"},"evidence_class":"direct_strict","model_configuration":"gemini-3.5-flash_low","organization":"Google","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_a130b7979f29bc1b9d879224b42a693a1044834fe43610fa05909a86b0808385","canonical_evidence_variant_id":"gpqa_evd_v1_e16b10917c76b9a9f0fe61840b40054fd59d27d4a8f1b0ec979013a0a244091d","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_a130b7979f29bc1b9d879224b42a693a1044834fe43610fa05909a86b0808385","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_e16b10917c76b9a9f0fe61840b40054fd59d27d4a8f1b0ec979013a0a244091d"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-10","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_3209319a9e249f3a03cbe28de9fcd90b9c603d24c23e54f4d19b3acf0eaa894d","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/MiniMaxAI/MiniMax-M3","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"MiniMax-M3","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"90.909091%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.020482086775424225","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"hDESmyGFrbZDomjff3rqzD"},"evidence_class":"direct_strict","model_configuration":"MiniMax-M3","organization":"MiniMax","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_39c4ce7df9b3ade8c5062cd6b5b7991617fd956fe43c7080d17910d6b97a05eb","canonical_evidence_variant_id":"gpqa_evd_v1_224469e684cd3105f7693029831040d119a347ce316175ad3a01c2e87bd2dd3f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_39c4ce7df9b3ade8c5062cd6b5b7991617fd956fe43c7080d17910d6b97a05eb","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_224469e684cd3105f7693029831040d119a347ce316175ad3a01c2e87bd2dd3f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-21","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_322dfa16da9ed38170bad2a7f7bdc94967cda750fa07cddee0c558f1bd8d506e","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/NousResearch/Hermes-2-Theta-Llama-3-70B","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Hermes-2-Theta-Llama-3-70B","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"37.468434%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024803009993948306","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"DHdeZNynu4VtRp2VGkCGvd"},"evidence_class":"direct_compatibility","model_configuration":"Hermes-2-Theta-Llama-3-70B","organization":"Nous Research,Arcee AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_dcb21a23564f01cdc1c264ade404b71aa500e53a3e399d757f8c497308719619","canonical_evidence_variant_id":"gpqa_evd_v1_18d8b74684579952ebe9cf782aff1b9a24f464f0d5095291c2fad65ab2642db6","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_dcb21a23564f01cdc1c264ade404b71aa500e53a3e399d757f8c497308719619","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_18d8b74684579952ebe9cf782aff1b9a24f464f0d5095291c2fad65ab2642db6"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_32831c12727d118cdd4d29d933e064c6f5efeb1e888228cb3ccaef5ca5e5a6c5","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/qwen3.7-flash","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.7-flash_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"80.808081%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.028057791672989062","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Yhds8GYkCXWgRQhuLiURua"},"evidence_class":"direct_strict","model_configuration":"qwen3.7-flash_none","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_92ba7d15e4ceb81888a46ecd1f69395d19555d0b6c84d5c2d9b3f5c1f9a87953","canonical_evidence_variant_id":"gpqa_evd_v1_687c53d044f8ad07f537ddf50a7fc464172bac7ac538c99799fdd98def9b9ab4","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_92ba7d15e4ceb81888a46ecd1f69395d19555d0b6c84d5c2d9b3f5c1f9a87953","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_687c53d044f8ad07f537ddf50a7fc464172bac7ac538c99799fdd98def9b9ab4"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-14","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_340eb8d0df205fcfca78e8fe1bef9408f9a08baa2fbe98a058f748c1dce19c09","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4.1-nano-2025-04-14","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4.1-nano-2025-04-14","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"48.926768%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024849734950435368","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"azWfeFPP6zSocuQX3JYBU5"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4.1-nano-2025-04-14","organization":"OpenAI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_eacdbe9bc45ea9592ffc838201b2238e93db76ab19475c10e241d5443ae49f0a","canonical_evidence_variant_id":"gpqa_evd_v1_50eebed0c14408666088d581f43d6c5da73019ea2ec8b715b708b1f9e9f97afc","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_eacdbe9bc45ea9592ffc838201b2238e93db76ab19475c10e241d5443ae49f0a","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_50eebed0c14408666088d581f43d6c5da73019ea2ec8b715b708b1f9e9f97afc"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_346701427f0d44ced8ca606e9855b570bb07e6ef55855f0a6332eed79ea25189","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"fireworks/accounts/fireworks/models/llama-v3p3-70b-instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"fireworks","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Llama-3.3-70B-Instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"47.443182%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.028523027562061026","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"TxXS78Wg2pSCDbpmeJuzQn"},"evidence_class":"direct_compatibility","model_configuration":"Llama-3.3-70B-Instruct","organization":"Meta AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_fc5488cb3b2dd15c9ca55eaa3ca5e1d9a9dd45a04f478f2b5e3ee33cfdf7d6e4","canonical_evidence_variant_id":"gpqa_evd_v1_8fbb01bae2958144e807395932fd73caf9656f6bc2e72fb5df5233603ad21814","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_fc5488cb3b2dd15c9ca55eaa3ca5e1d9a9dd45a04f478f2b5e3ee33cfdf7d6e4","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_8fbb01bae2958144e807395932fd73caf9656f6bc2e72fb5df5233603ad21814"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_3706e715c29d8cc087e5eea8fcdcfc37f9f912420fe1550b3d630f10b709d0cc","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/google/gemma-2-27b-it","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemma-2-27b-it","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"36.489899%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.023897884428691632","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"ZUJxsJDpHnLXxNiDcvj22W"},"evidence_class":"direct_compatibility","model_configuration":"gemma-2-27b-it","organization":"Google DeepMind","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_b489e4513748213af359d6e74207b6fedf629d9f4bdfd8c5dd19fe6e5f7b9519","canonical_evidence_variant_id":"gpqa_evd_v1_b326456b6f83bfbccf9cde7fc02a1fe96f750e01bd69cf4ed11bba768de8ba12","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_b489e4513748213af359d6e74207b6fedf629d9f4bdfd8c5dd19fe6e5f7b9519","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_b326456b6f83bfbccf9cde7fc02a1fe96f750e01bd69cf4ed11bba768de8ba12"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-10","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_372f86421b339917a5ea87ce041519de28881fa7f9f39ab618a6079c1668516e","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"grok/grok-3-mini-beta","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"grok","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"grok-3-mini-beta_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"76.262626%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03031371053819892","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"McxfQdC5aypnD3STwZUBjH"},"evidence_class":"direct_compatibility","model_configuration":"grok-3-mini-beta_low","organization":"xAI","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_cde3565a56b1a6656dafbec075f1438586402733203d291442929aa5a0fe8e98","canonical_evidence_variant_id":"gpqa_evd_v1_d90fff655dcfbdf0cfe93c973d7a1ee242c06aee473853daa7a5b1fa26756c4e","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_cde3565a56b1a6656dafbec075f1438586402733203d291442929aa5a0fe8e98","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d90fff655dcfbdf0cfe93c973d7a1ee242c06aee473853daa7a5b1fa26756c4e"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_39ebb391cd30cb3a52b963dd160804c308e182b68c53d4420593c1241547bbb7","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"mistral/open-mistral-nemo-2407","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"mistral","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"open-mistral-nemo-2407","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"29.892677%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02096191098761666","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"XCSWtbDquA7FUiWfd7kh9q"},"evidence_class":"direct_compatibility","model_configuration":"open-mistral-nemo-2407","organization":"Mistral AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_a574fc8089b5af19df1ac0afceca408ade5cf7e9158fe63bbd5f7983b27a3adc","canonical_evidence_variant_id":"gpqa_evd_v1_5e7669ec1acb94109cd873958540cade2ad14d07e0554103679a173f8777a6cb","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_a574fc8089b5af19df1ac0afceca408ade5cf7e9158fe63bbd5f7983b27a3adc","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_5e7669ec1acb94109cd873958540cade2ad14d07e0554103679a173f8777a6cb"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-05-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_3b4449cd7456be9a421061071fd80145c7806c116ccfedd101c8f06b7ea16dc5","configuration_identity":{"context_policy":"16k","effort":"unspecified","model_snapshot":"anthropic/claude-sonnet-4-20250514","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-sonnet-4-20250514_16K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"75.757576%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.030532892233932022","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"PPLoBnbKQ9CEYZNSk3E7LE"},"evidence_class":"direct_compatibility","model_configuration":"claude-sonnet-4-20250514_16K","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_0233e4a8f66947e14f6135b5a7ee0d5de8ae20eee33963f722ab2c71ba784588","canonical_evidence_variant_id":"gpqa_evd_v1_6b7ac484013d63c135c641e8c2a1e72fc0c3feae635631c5ded9e95e58e539a4","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_0233e4a8f66947e14f6135b5a7ee0d5de8ae20eee33963f722ab2c71ba784588","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_6b7ac484013d63c135c641e8c2a1e72fc0c3feae635631c5ded9e95e58e539a4"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-10-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_3c5870b608f9aaf878456e51a25d8aadb6e0d8e9b1e902670d6b72045777d494","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"alibaba/qwen3-max-2025-09-23","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"alibaba","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"qwen3-max-2025-09-23","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"72.601010%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02734528103171235","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"46RCssznDADYeg8MuSgvsj"},"evidence_class":"direct_compatibility","model_configuration":"qwen3-max-2025-09-23","organization":"Alibaba","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_272608934e92117c602d64514128c1867a07c1b57115700655d642e1fb197f2b","canonical_evidence_variant_id":"gpqa_evd_v1_56ada95164ff407147cf7b22059b1b336410612ab62b0e4d18ba32d836c47c5d","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_272608934e92117c602d64514128c1867a07c1b57115700655d642e1fb197f2b","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_56ada95164ff407147cf7b22059b1b336410612ab62b0e4d18ba32d836c47c5d"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-30","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_3d7b3a0e3e36b74bdde08e2c386e942a53b522bd7c61c1151e6dc007e6e9f001","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"mistral/mistral-small-2501","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"mistral","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"mistral-small-2501","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"45.296717%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024794688677057853","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"oDq358GfyV5Au65scERdVo"},"evidence_class":"direct_compatibility","model_configuration":"mistral-small-2501","organization":"Mistral AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_93b507d8115061a7314678682e910977a0c92d617f44bceb8b13a732b5c82465","canonical_evidence_variant_id":"gpqa_evd_v1_6870f284c264186f6933c0e05d3113748d58496f71560d088c424c042296fc32","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_93b507d8115061a7314678682e910977a0c92d617f44bceb8b13a732b5c82465","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_6870f284c264186f6933c0e05d3113748d58496f71560d088c424c042296fc32"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-02-25","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_41455c97d103f4b57508177da6055ed522be7154caac525ae1d06daba079907f","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"mistral/mistral-large-2411","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"mistral","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"mistral-large-2411","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"51.325758%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026887818295509087","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"n9gpXfUzo8SbWqvrG2q4xM"},"evidence_class":"direct_compatibility","model_configuration":"mistral-large-2411","organization":"Mistral AI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_8118b09fe4e06e98040342560692004b81a072d57e690453ac4b706d0cb591b3","canonical_evidence_variant_id":"gpqa_evd_v1_993a52cea93d40f3ea6090f1ccfc20c3fd1b37331197eda696eff23dfbcc76f4","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_8118b09fe4e06e98040342560692004b81a072d57e690453ac4b706d0cb591b3","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_993a52cea93d40f3ea6090f1ccfc20c3fd1b37331197eda696eff23dfbcc76f4"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-15","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_439fcb4bf6820e10b1024c2303f377da9ff1a1106a368e9fbe08acd16c8f32f9","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"epoch/gpt-5.4-2026-03-05","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.4-2026-03-05_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"89.898990%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.021469735576055356","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"FbhcT8NnJUntcyNGubJXY6"},"evidence_class":"direct_strict","model_configuration":"gpt-5.4-2026-03-05_high","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_f24d6a067fc4d63a07a09e02dd0a7c8c5977b7e161958929a780337da209cc6d","canonical_evidence_variant_id":"gpqa_evd_v1_99ceec1a96535e7c294d08862aad18ad46ed34e353d04fc1190c089cae51276c","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_f24d6a067fc4d63a07a09e02dd0a7c8c5977b7e161958929a780337da209cc6d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_99ceec1a96535e7c294d08862aad18ad46ed34e353d04fc1190c089cae51276c"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-11-16","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_457d4d9de46a09e8dda7707cbea4e03840a5686a817ef7dbd62b104c41ed8b9e","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-2.5-pro","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.1","source_model_version":"gemini-2.5-pro","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"85.290404%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02117010842348584","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"cqUZ6gcq2HBnShFRWvKjGr"},"evidence_class":"direct_strict","model_configuration":"gemini-2.5-pro","organization":"Google DeepMind","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_203f5558a45d7350437aae82e52a8cd851089e49ce2a62abe6f2a1598ad7df4a","canonical_evidence_variant_id":"gpqa_evd_v1_b70737936b1bfc0ef3b3501ad10683d41623c62dc2fb664c7eef286ecbd01c26","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_203f5558a45d7350437aae82e52a8cd851089e49ce2a62abe6f2a1598ad7df4a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_b70737936b1bfc0ef3b3501ad10683d41623c62dc2fb664c7eef286ecbd01c26"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_46e58f3c8d85f65c9905d98e961570ffe8b923db2111db25bcf593e5a586ba5b","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"epoch/claude-sonnet-5","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-sonnet-5_max","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"80.303030%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.028335609732463334","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"FNXAt76ZUS3sWhQDFMga7U"},"evidence_class":"direct_strict","model_configuration":"claude-sonnet-5_max","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_711746500282f901b2cd94eb120ad984ae3a3f95786e881c2153bbfa1889a4ae","canonical_evidence_variant_id":"gpqa_evd_v1_31aec9986e3f2401cfd7f1131e8c104af9961fb562e2fef154db8a1e512e2f92","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_711746500282f901b2cd94eb120ad984ae3a3f95786e881c2153bbfa1889a4ae","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_31aec9986e3f2401cfd7f1131e8c104af9961fb562e2fef154db8a1e512e2f92"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-21","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_48c8d1776d8586ba5fce19ed21b2148cc8f42ab5c360ce1f53012f8973abf633","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/meta-llama/Llama-2-70b-chat-hf","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Llama-2-70b-chat-hf","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"26.325758%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.019901808558284103","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"evPjJw9SKqq3TieSdxrSQn"},"evidence_class":"direct_compatibility","model_configuration":"Llama-2-70b-chat-hf","organization":"Meta AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_0478d6f4b4226e20b7677c937472da09020d428b98df9793e6ce5489bade9b93","canonical_evidence_variant_id":"gpqa_evd_v1_20183cd6af48b75174c64572e9bc9e915b327f8aad814e80cbdd72db7405be62","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_0478d6f4b4226e20b7677c937472da09020d428b98df9793e6ce5489bade9b93","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_20183cd6af48b75174c64572e9bc9e915b327f8aad814e80cbdd72db7405be62"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-16","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_4c3a34cfb53d1bdeb2350a0a8d43d7dd34e2878b1834dd8a615a4f6100909a4c","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/deepseek-chat","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"deepseek-chat","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"71.212121%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03225883512300997","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"6osoAGM4euiAAksriYyLtj"},"evidence_class":"direct_strict","model_configuration":"deepseek-chat","organization":"DeepSeek","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_e2029b0e278266d4c4aad2243a7526b13fae5e7358fcf87c16fa509cc40daf35","canonical_evidence_variant_id":"gpqa_evd_v1_f50d63ba07fedab826a9217dd7cdea0d41bc467f08e51da4d9ab10df4d5ece40","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e2029b0e278266d4c4aad2243a7526b13fae5e7358fcf87c16fa509cc40daf35","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f50d63ba07fedab826a9217dd7cdea0d41bc467f08e51da4d9ab10df4d5ece40"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_4c7af7da68de4f30e4d7cddcc9db1fbbf77833670cae85770afbb589903f3414","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"mistral/mistral-large-2402","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"mistral","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"mistral-large-2402","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"38.762626%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024776598535417284","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"dDeREdDJ5bVyci3APZxzWb"},"evidence_class":"direct_compatibility","model_configuration":"mistral-large-2402","organization":"Mistral AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_634bf9f5a3db64f728934b2401ccc0d115ad69c9e938dd5eaff1ffedb1d45841","canonical_evidence_variant_id":"gpqa_evd_v1_198d46a19024fb71e67c9f4f7c71fad94752573965683dde8f8351f8d8fda854","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_634bf9f5a3db64f728934b2401ccc0d115ad69c9e938dd5eaff1ffedb1d45841","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_198d46a19024fb71e67c9f4f7c71fad94752573965683dde8f8351f8d8fda854"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_4cdcf5e7b5a7d772f20148bb6cb03e8e5e57acbe5490e9963e0df1ff0a13f315","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-3-5-sonnet-20240620","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-3-5-sonnet-20240620","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"54.040404%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027528555453348653","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"78Wpvp8WDByAcQRBas72cs"},"evidence_class":"direct_compatibility","model_configuration":"claude-3-5-sonnet-20240620","organization":"Anthropic","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":7,"canonical_evaluation_run_id":"gpqa_run_v1_4a49b8bc4ef67cb52c4630ef9d37fa289e44c5df0580843e17425b64798c4739","canonical_evidence_variant_id":"gpqa_evd_v1_53194c97190cec47bf239eed235fa952bb1155e0d6a78b10ff7465d767b9ba83","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_4a49b8bc4ef67cb52c4630ef9d37fa289e44c5df0580843e17425b64798c4739","evidence_class":"derived_uniform","evidence_variant_id":"gpqa_evd_v1_53194c97190cec47bf239eed235fa952bb1155e0d6a78b10ff7465d767b9ba83"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-02-06","configuration_date_basis":"validated Epoch CSV Started at metadata","configuration_id":"gpqa_cfg_v1_4cff4744741c24d1db4b97e733fde765e5f1c2056705a760369c260e9001e3d0","configuration_identity":{"context_policy":"64k","effort":"unspecified","model_snapshot":"epoch/claude-opus-4-6_64K","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_source_task_v1","protocol":"mean@7","provider_route":"epoch_import","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"7","declared_attempts_per_question":"8","excluded_epoch_numbers":"1","source_model_version":"claude-opus-4-6_64K"}},"details":{"configuration_date_basis":"validated Epoch CSV Started at metadata","evidence_label":"Derived uniform-epoch subset","reconciliation":{"paired_original_role":"Reconstructed from the selected uniform subset; not the native Epoch aggregate.","published_source_score_percent":"88.762626%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.019363218328924313","status":"exact"},"recovery":{"declared_attempts_per_question":8,"kind":"uniform_epoch_subset","omitted_attempt_count":1,"selected_attempts_per_question":7},"score_source":"Accepted binary choice scores from the complete uniform attempt subset","source_kind_label":"Epoch source with globally uniform attempt subset","source_run_id":"eB9e2mtd5nRwqdFyLEsz6k"},"evidence_class":"derived_uniform","model_configuration":"claude-opus-4-6_64K","organization":"Anthropic","protocol":"mean@7","validation_profile":"derived_uniform_epoch_subset"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_cc64f3f4ccbcb82bbbd085eff0af6626c4e907bc32ac462a1c97952a7b9f05fe","canonical_evidence_variant_id":"gpqa_evd_v1_631b865d16ea8a60f9a73143182597786f66c30e4e4cd9ba44a27a003085179e","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_cc64f3f4ccbcb82bbbd085eff0af6626c4e907bc32ac462a1c97952a7b9f05fe","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_631b865d16ea8a60f9a73143182597786f66c30e4e4cd9ba44a27a003085179e"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-11","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_4e08dd7d1b1197d17bb458f074103b518735398f7272662a3a46c37f381e18c3","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"alibaba/qwq-plus","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"alibaba","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"qwq-plus","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"65.404040%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.028469238770829935","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"PtxHF8BH8WuNdNfDnutGBb"},"evidence_class":"direct_compatibility","model_configuration":"qwq-plus","organization":"Alibaba","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_d987f92c42189d8e928c6ec2a35ab3b520867c90143588803e8bbe84b7306481","canonical_evidence_variant_id":"gpqa_evd_v1_d65c0c339d9fb28eb80cdd3c8012c18af1eaf202129b8a347da9486b187ac48b","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_d987f92c42189d8e928c6ec2a35ab3b520867c90143588803e8bbe84b7306481","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_d65c0c339d9fb28eb80cdd3c8012c18af1eaf202129b8a347da9486b187ac48b"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-10-29","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_4f3803305451e61c9246b82a829be2d36ae588a015920201efed023d91dabfaa","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"openai/gpt-5-2025-08-07","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.1","source_model_version":"gpt-5-2025-08-07_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"86.174242%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.020787836099851995","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"9nPyNrZxwtoT7eS4DZbQF6"},"evidence_class":"direct_strict","model_configuration":"gpt-5-2025-08-07_high","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_6f145229001723016030c8489246da79111c7919d38127a356e8a0ad74a3ca45","canonical_evidence_variant_id":"gpqa_evd_v1_c3994ed498ac294c3fb7ad1c0b4d15590978b729fabd6bf692874c979d1ecccb","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_6f145229001723016030c8489246da79111c7919d38127a356e8a0ad74a3ca45","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_c3994ed498ac294c3fb7ad1c0b4d15590978b729fabd6bf692874c979d1ecccb"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_4ff2a56599222d2a455b2c7961cc7f544af144e423b64cfd6a02e8ed9c88e446","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-3.5-turbo-0125","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-3.5-turbo-0125","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"27.178030%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.01953579968788737","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"eZaQhpbhbzbo3sbKGS5Efa"},"evidence_class":"direct_compatibility","model_configuration":"gpt-3.5-turbo-0125","organization":"OpenAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_039b9a8b33fc8e9dcbec6b6f37a8a39e30393f5d34d30ec6a9bf90e0eec55d8c","canonical_evidence_variant_id":"gpqa_evd_v1_243d5ed85082636e1f65f5ffc32b913ef50ce08428fcff34532ea024f39a76ae","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_039b9a8b33fc8e9dcbec6b6f37a8a39e30393f5d34d30ec6a9bf90e0eec55d8c","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_243d5ed85082636e1f65f5ffc32b913ef50ce08428fcff34532ea024f39a76ae"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_5169bc0c733d68b48affc9e28c896dd620d7ff8a7a428662b2f13213317c8364","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4o-2024-05-13","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4o-2024-05-13","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"48.895202%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026254109873371994","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Y59q8WbEDp2HH8oovyJCj9"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4o-2024-05-13","organization":"OpenAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_797dcb2ae3c120447661d4ccc2860d516d7f0155413c03816d86c43ba4f17ae0","canonical_evidence_variant_id":"gpqa_evd_v1_248b338d575cfa178b0122eaa86cad889196454926526426ee594ad104499afd","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_797dcb2ae3c120447661d4ccc2860d516d7f0155413c03816d86c43ba4f17ae0","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_248b338d575cfa178b0122eaa86cad889196454926526426ee594ad104499afd"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-04-24","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_524a33f2416077d9ee7c8a56c4e245d2737cfc16dc1505dcc72a7b1994afc88a","configuration_identity":{"context_policy":"source_default","effort":"xhigh","model_snapshot":"openai/gpt-5.5-pre-release","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.8","source_model_version":"gpt-5.5-pre-release_xhigh","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"94.002525%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.015469771177277509","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"FUv4EUwR78nAsdGbiwiLEr"},"evidence_class":"direct_strict","model_configuration":"gpt-5.5-pre-release_xhigh","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_7e2996f3d21e6291bdc7c3aeae14990995271215f5dccf59f770daa5482351a3","canonical_evidence_variant_id":"gpqa_evd_v1_9d41e3a2eb0616d21676ee40f289501786e33fe43dbba87eb35ffda47da57ef6","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_7e2996f3d21e6291bdc7c3aeae14990995271215f5dccf59f770daa5482351a3","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_9d41e3a2eb0616d21676ee40f289501786e33fe43dbba87eb35ffda47da57ef6"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_53004f5f1c902a577afd76e4f9016de7963f5d869c85bb3d7906463afa273244","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/gpt-5.4-nano-2026-03-17","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.4-nano-2026-03-17_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"72.222222%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03191178226713548","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Nq7w5iKALDSuT5oQdn5m8f"},"evidence_class":"direct_strict","model_configuration":"gpt-5.4-nano-2026-03-17_low","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_22da1b03bd0e0eccc48c7d5f0281927fb694a76fbf80ce2072ee1ebc84aa19ed","canonical_evidence_variant_id":"gpqa_evd_v1_d9359bd7a357e0356b20feeb043da151e0a7ffa86c18d08eb536b4b8a960d9c7","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_22da1b03bd0e0eccc48c7d5f0281927fb694a76fbf80ce2072ee1ebc84aa19ed","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d9359bd7a357e0356b20feeb043da151e0a7ffa86c18d08eb536b4b8a960d9c7"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-08","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_5308e01f281df009ce28185a83659d0ad19a59cdd3e55b5038c6f0b95a53d8b5","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"together/meta-llama/Llama-4-Scout-17B-16E-Instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"together","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Llama-4-Scout-17B-16E-Instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"51.830808%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.031549020124994755","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"MsXAXDEKP3xBKtiA7aTBBG"},"evidence_class":"direct_compatibility","model_configuration":"Llama-4-Scout-17B-16E-Instruct","organization":"Meta AI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_b7988379e8ab98297f032f9b27c27ce6660a7843e128165e84d42f0a6c76d710","canonical_evidence_variant_id":"gpqa_evd_v1_b16503dd683d625d37a83787926aecad20fd1027038dd1d52b14d7274d1e0d65","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_b7988379e8ab98297f032f9b27c27ce6660a7843e128165e84d42f0a6c76d710","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_b16503dd683d625d37a83787926aecad20fd1027038dd1d52b14d7274d1e0d65"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-20","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_536a3d475a86a0bde0d95ab2fea86e8ab597b125c21175652a1b26a307d66765","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"hyperbolic/meta-llama/Meta-Llama-3.1-70B-Instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"hyperbolic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Llama-3.1-70B-Instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"44.191919%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024522863472149384","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"MEwMcmSV32JHrsNoAq4BFL"},"evidence_class":"direct_compatibility","model_configuration":"Llama-3.1-70B-Instruct","organization":"Meta AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_3d4e5a3f8219c5d1cdcd32800a4256c4732de5820b869d7f2124f79b702a64f8","canonical_evidence_variant_id":"gpqa_evd_v1_fbdd43a5165165c2b670ee70b7ce930b5f7520a8cb06bb17327db3b8256b3bbe","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_3d4e5a3f8219c5d1cdcd32800a4256c4732de5820b869d7f2124f79b702a64f8","evidence_class":"targeted_recovery","evidence_variant_id":"gpqa_evd_v1_fbdd43a5165165c2b670ee70b7ce930b5f7520a8cb06bb17327db3b8256b3bbe"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-16","configuration_date_basis":"validated composite generated_at metadata","configuration_id":"gpqa_cfg_v1_538540e64da939fe97f8685591c8f9e03cd907c322ec906eea6256ce46fc72df","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"zhipu/glm-5","output_policy":"source_artifact_policy_plus_uncapped_recovery","prompt_contract":"epoch_gpqa_diamond_source_plus_recovery_v1","protocol":"mean@1","provider_route":"zhipu_plus_zai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","configuration_match":"exact_source_header_generate_config","execution_match":"provider_route_substitution","model_identity_basis":"declared_provider_route_substitution","source_model_version":"glm-5"}},"details":{"configuration_date_basis":"validated composite generated_at metadata","evidence_label":"Epoch base plus targeted transport recovery","reconciliation":{"difference_percentage_points":"0.000000 pp","paired_original_role":"Reconstructed from the completed source-plus-recovery attempt universe; not relabelled as the original Epoch aggregate.","published_source_score_percent":"87.817259%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02336331210092387","status":"exact"},"recovery":{"configuration_match":"exact_source_header_generate_config","execution_match":"provider_route_substitution","kind":"exact_targeted_transport_recovery","model_identity_basis":"declared_provider_route_substitution","output_ceiling":"none","recovered_pair_count":1},"score_source":"Accepted source choice scores plus exact targeted repair pairs","source_kind_label":"Epoch base plus targeted transport-recovery composite","source_run_id":"n4hyz75SVpwPEisKVVjvNg"},"evidence_class":"targeted_recovery","model_configuration":"glm-5","organization":"Z.ai (Zhipu AI)","protocol":"mean@1","validation_profile":"complete_with_epoch_targeted_transport_recovery"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_b4d9e290a09db97ad3f91d3eecc69fe970a85410a4d691945769b9d02d1f9683","canonical_evidence_variant_id":"gpqa_evd_v1_f43181f3ca5b2ed334391a207f511dbf7a7f52c71f2c9895aab453ee29721dc5","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_b4d9e290a09db97ad3f91d3eecc69fe970a85410a4d691945769b9d02d1f9683","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f43181f3ca5b2ed334391a207f511dbf7a7f52c71f2c9895aab453ee29721dc5"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-05-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_54aba346a03524e84a582d320e68f5d6e7e1ca6e1ff7a164ae903b0be1d435a6","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-sonnet-4-20250514","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-sonnet-4-20250514","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"66.666667%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.033586181457325226","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"ZF4eejw7J2KrcoNWd6hUbU"},"evidence_class":"direct_compatibility","model_configuration":"claude-sonnet-4-20250514","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_46fcc9979861645dd49c5dfbf566c799658d885aba05a023e8ffd845d4aec810","canonical_evidence_variant_id":"gpqa_evd_v1_522dfca6257a33cbb37330d9aa151d82132de83b96f985761e1702e486673f26","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_46fcc9979861645dd49c5dfbf566c799658d885aba05a023e8ffd845d4aec810","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_522dfca6257a33cbb37330d9aa151d82132de83b96f985761e1702e486673f26"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-02-13","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_54e29f3aa006e93a5e11a4220d4ffe0398120b3ceb40328c11661e50953226b2","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"openai/o1-mini-2024-09-12","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"o1-mini-2024-09-12_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"62.373737%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027427723511230753","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"X2gKewiGPZx5DLFEddUUqW"},"evidence_class":"direct_compatibility","model_configuration":"o1-mini-2024-09-12_high","organization":"OpenAI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":6,"canonical_evaluation_run_id":"gpqa_run_v1_8a99bf54406f9c83dff96771b325f108b7d26ac67c34b6fbe29422becea78312","canonical_evidence_variant_id":"gpqa_evd_v1_ad73cf533c8248b403ff3591bd404a3bc2f9a571e2750b18265e87582ac8cf2a","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_8a99bf54406f9c83dff96771b325f108b7d26ac67c34b6fbe29422becea78312","evidence_class":"derived_uniform","evidence_variant_id":"gpqa_evd_v1_ad73cf533c8248b403ff3591bd404a3bc2f9a571e2750b18265e87582ac8cf2a"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-05-01","configuration_date_basis":"validated Epoch CSV Started at metadata","configuration_id":"gpqa_cfg_v1_555ebc09735857529c6e51f2975f27f05313997eae5072969000efc2baf8faa7","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/kimi-k2.6","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_source_task_v1","protocol":"mean@6","provider_route":"epoch_import","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"6","declared_attempts_per_question":"8","excluded_epoch_numbers":"1,8","source_model_version":"kimi-k2.6"}},"details":{"configuration_date_basis":"validated Epoch CSV Started at metadata","evidence_label":"Derived uniform-epoch subset","reconciliation":{"paired_original_role":"Reconstructed from the selected uniform subset; not the native Epoch aggregate.","published_source_score_percent":"90.782828%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.0171745162866666","status":"exact"},"recovery":{"declared_attempts_per_question":8,"kind":"uniform_epoch_subset","omitted_attempt_count":2,"selected_attempts_per_question":6},"score_source":"Accepted binary choice scores from the complete uniform attempt subset","source_kind_label":"Epoch source with globally uniform attempt subset","source_run_id":"LjgJoDosTLkiGVT59smA7q"},"evidence_class":"derived_uniform","model_configuration":"kimi-k2.6","organization":"Moonshot","protocol":"mean@6","validation_profile":"derived_uniform_epoch_subset"},{"attempts_per_question":2,"canonical_evaluation_run_id":"gpqa_run_v1_6425cfc911fd497c8b8f7eea4d2f8695e7903b9e5ff309140b4c6619b014d12b","canonical_evidence_variant_id":"gpqa_evd_v1_79c1de5bc1713fd3813a81e8b6531d7fe800568a2b9664a42e0c8cbda907a5ba","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_6425cfc911fd497c8b8f7eea4d2f8695e7903b9e5ff309140b4c6619b014d12b","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_79c1de5bc1713fd3813a81e8b6531d7fe800568a2b9664a42e0c8cbda907a5ba"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-12-11","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_569cf92af26ce0effa0d7c6f3499ee8a4b59832cf43499b195482e8caf68b0e0","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"together/openai/gpt-oss-120b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@2","provider_route":"together","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"2","benchmark_version":"1.0.3","source_model_version":"openai/gpt-oss-120b_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"75.757576%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027437163443883767","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"LC5Rhog6wt2PT84CtJ483h"},"evidence_class":"direct_strict","model_configuration":"openai/gpt-oss-120b_high","organization":"OpenAI","protocol":"mean@2","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_441d4e54e4f5b3ef3a0baca49e9b9ed90456b39bb8196a7eb27954c07c7559fd","canonical_evidence_variant_id":"gpqa_evd_v1_c0532c6558b99168a64b9955a9315f8b6e21f0ba2d7b2f3af63334af7b628c99","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_441d4e54e4f5b3ef3a0baca49e9b9ed90456b39bb8196a7eb27954c07c7559fd","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_c0532c6558b99168a64b9955a9315f8b6e21f0ba2d7b2f3af63334af7b628c99"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_56d66d76f7fd4532bd5045af003c34d1f07f98b76657161a697fa41101524e25","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-3-sonnet-20240229","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-3-sonnet-20240229","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"40.593434%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.023920223491748827","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"QLejPZXhjX6SQCQgzjNvCG"},"evidence_class":"direct_compatibility","model_configuration":"claude-3-sonnet-20240229","organization":"Anthropic","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_9fbf8985c5ca2f5198c04f4e81c7b70e8964a76522f3abd5cbe7ca7da5d89927","canonical_evidence_variant_id":"gpqa_evd_v1_8e4c9d6bc023349b56f42fd7442941038c1397f038dc1e688b19e7bf00982974","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_9fbf8985c5ca2f5198c04f4e81c7b70e8964a76522f3abd5cbe7ca7da5d89927","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_8e4c9d6bc023349b56f42fd7442941038c1397f038dc1e688b19e7bf00982974"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-02-26","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_582401c4d6cff8722feac8c0b9c6982da4d7180861f4671dfc33dd6a447680d5","configuration_identity":{"context_policy":"16k","effort":"unspecified","model_snapshot":"anthropic/claude-3-7-sonnet-20250219","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-3-7-sonnet-20250219_16K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"76.767677%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.030088629490217445","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Vn3Ug88CRbxEaJhLbRzAWD"},"evidence_class":"direct_compatibility","model_configuration":"claude-3-7-sonnet-20250219_16K","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_e5b06d6b599075a92a772a091e296bfa01fd2e9190c7e4ec2e8f2ab799c5b407","canonical_evidence_variant_id":"gpqa_evd_v1_16fb7c049832fb20f09582abe93d651ad739e000bbce6986696c5f2a37bbe2ec","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e5b06d6b599075a92a772a091e296bfa01fd2e9190c7e4ec2e8f2ab799c5b407","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_16fb7c049832fb20f09582abe93d651ad739e000bbce6986696c5f2a37bbe2ec"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-21","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_58b6db074aeb963d09941f6ead70727311ab03f1d3e3eb559008d9b988dccb25","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"grok/grok-2-1212","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"grok","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"grok-2-1212","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"53.787879%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027066342316607196","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"83DRJD4FvjwqU9eStRaNiU"},"evidence_class":"direct_compatibility","model_configuration":"grok-2-1212","organization":"xAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_926aed6b613d9dd91e10f3a53004f8390136ded8ff7ae56eaa721656e8d8b8aa","canonical_evidence_variant_id":"gpqa_evd_v1_d4b7ed220a632326d3c67ac1fa54c4e79d8dc6a73dabaa9877b432abf48f3eb6","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_926aed6b613d9dd91e10f3a53004f8390136ded8ff7ae56eaa721656e8d8b8aa","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d4b7ed220a632326d3c67ac1fa54c4e79d8dc6a73dabaa9877b432abf48f3eb6"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-10-23","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_58d9552106e8f5dfbcb2f79f29d67a6ee305a8dc9f8492f63ddf624493d7530a","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4-0314","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4-0314","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"35.732323%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02363099375599328","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"FkbLkW5ed8rnZGFJMevDFn"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4-0314","organization":"OpenAI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_adae307af1764f3d1c10f241782323fb273b930e20aebda1a4c6c1c192aa7cd3","canonical_evidence_variant_id":"gpqa_evd_v1_248a16f13c340919d1ee009aa68164ebe4278b515dfce8051ba5c3d9d1c6585d","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_adae307af1764f3d1c10f241782323fb273b930e20aebda1a4c6c1c192aa7cd3","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_248a16f13c340919d1ee009aa68164ebe4278b515dfce8051ba5c3d9d1c6585d"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-21","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_58e2ae82542f35fedd02586ab2a506dc9dd9a92ad9f1913f2bcb0393aeaf603b","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/meta-llama/Meta-Llama-3-70B-Instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Meta-Llama-3-70B-Instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"40.561869%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026224849492112068","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"jrvutCoFFG4KxYtU7REdpe"},"evidence_class":"direct_compatibility","model_configuration":"Meta-Llama-3-70B-Instruct","organization":"Meta AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_2877769eccb658339c928b8169c2b496940c7a5cf8b027fa9d9c9dad4f96b44e","canonical_evidence_variant_id":"gpqa_evd_v1_cf36144f7e48a23a6a2309dfceda87c46953316341453d3108263f37f31d00e3","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_2877769eccb658339c928b8169c2b496940c7a5cf8b027fa9d9c9dad4f96b44e","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_cf36144f7e48a23a6a2309dfceda87c46953316341453d3108263f37f31d00e3"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_59886bda1c8504af950b9962ec90464beff5b80e797a4d747fee5b15546d0674","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-3-opus-20240229","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-3-opus-20240229","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"47.159091%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026395393956533515","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"hTnDa4gn8RYduCEYqoNywe"},"evidence_class":"direct_compatibility","model_configuration":"claude-3-opus-20240229","organization":"Anthropic","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_de2a82fcc1678230a76f2d08226dab25e981b5e3b13ed5a4d704ea96b0f13c93","canonical_evidence_variant_id":"gpqa_evd_v1_24901ab213808f65c6265c08123405c0d8fa99e9ee3f0cec372e5734be70cd37","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_de2a82fcc1678230a76f2d08226dab25e981b5e3b13ed5a4d704ea96b0f13c93","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_24901ab213808f65c6265c08123405c0d8fa99e9ee3f0cec372e5734be70cd37"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-17","configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","configuration_id":"gpqa_cfg_v1_5af397530c003e894fc8b0e8c6dfa9efeb6966457911dae825b08a1198fbb480","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"deepseek/deepseek-v4-flash","output_policy":"no_output_ceiling","prompt_contract":"gpqa_native_answer_choice_v1","protocol":"mean@1","provider_route":"deepseek","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","generation_config_sha256":"7582915618a182a752f56b4448601136297c15ad1acbbed527f93360a93c0fb4","profile":"deepseek-v4-flash-high","task_sha256":"2501e59954a9e53479bc5e44bf9923db905866019f16c385ea3f27483b3697d0"}},"details":{"configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","evidence_label":"Independent native run","execution":{"automatic_retries":0,"fresh_calls":true,"output_ceiling":"none"},"reconciliation":{"paired_original_role":"Reconstructed from the same accepted all-198 native attempt universe as the Clean score.","published_source_score_percent":null,"status":"not_applicable_independent_run"},"recovery":{"kind":"none","terminal_no_answer_policy":"No parseable answer is scored incorrect under the predeclared policy.","terminal_stop_count":3},"score_source":"Independent native answer-safe projection","source_kind_label":"Independent project-native run","source_run_id":"gpqa-native-deepseek-v4-flash-high-all198-20260816-r3"},"evidence_class":"project_native","model_configuration":"DeepSeek V4 Flash (high)","organization":"DeepSeek","protocol":"mean@1","validation_profile":"complete_with_provider_terminal_incorrect"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_5eb40b4a1fc21a66c876f059bb93f6f5a07dc7420d4b52b8bbf2bab9f19f06b2","canonical_evidence_variant_id":"gpqa_evd_v1_0e53e19644c87dc0e84c9e84ee9e3527e1f1f97f4dcd02d8d939efb7eb0401e0","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_5eb40b4a1fc21a66c876f059bb93f6f5a07dc7420d4b52b8bbf2bab9f19f06b2","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_0e53e19644c87dc0e84c9e84ee9e3527e1f1f97f4dcd02d8d939efb7eb0401e0"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_5b80556138a24f63b8f2cb80dccbea6656b707f83aedcd63a5cfe0c1d618c3d1","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"alibaba/qwen-plus-2025-01-25","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"alibaba","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"qwen-plus-2025-01-25","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"48.106061%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027144936812277986","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"QtraMoKiQRRpZc9txpFAYF"},"evidence_class":"direct_compatibility","model_configuration":"qwen-plus-2025-01-25","organization":"Alibaba","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_6e14e47b5b7b1a91e8f040222b2da203dc2c7cd6d19251e70929a30fefa0e2db","canonical_evidence_variant_id":"gpqa_evd_v1_d1dfeb319d5426dc160b9dea8728578465aa8b7a881e50f991c36d232ae4fcae","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_6e14e47b5b7b1a91e8f040222b2da203dc2c7cd6d19251e70929a30fefa0e2db","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d1dfeb319d5426dc160b9dea8728578465aa8b7a881e50f991c36d232ae4fcae"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-02-13","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_5b98063364ead570d569428619f6b536cda9944d79c01cd3d1325c9b59abb6bb","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"openai/o3-mini-2025-01-31","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"o3-mini-2025-01-31_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"77.020202%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.025855332698878485","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"MzsTvCEEfEpzn4WvxYdJkX"},"evidence_class":"direct_compatibility","model_configuration":"o3-mini-2025-01-31_high","organization":"OpenAI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_fc2703c5c2f9f81e611279817ffd0814e39c89a8abf47d038fd48d1da9ca2df1","canonical_evidence_variant_id":"gpqa_evd_v1_49e20b164dbe1f4b3cbb7e38bf2c70d4296f9044040795b06f2964900bb91ead","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_fc2703c5c2f9f81e611279817ffd0814e39c89a8abf47d038fd48d1da9ca2df1","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_49e20b164dbe1f4b3cbb7e38bf2c70d4296f9044040795b06f2964900bb91ead"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_5ba80b136d60d34017baf239819ca37835512b9eb474be2e9b599485c1de4c57","configuration_identity":{"context_policy":"source_default","effort":"minimal","model_snapshot":"epoch/gemma-4-31b-it","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemma-4-31b-it_minimal","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"75.757576%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03053289223393202","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"P5L8qbXo42vTvbU3UvTNjk"},"evidence_class":"direct_strict","model_configuration":"gemma-4-31b-it_minimal","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_cd28725ecd192d79320824311013bd267aaac131969bbe81362007435ea9e3f5","canonical_evidence_variant_id":"gpqa_evd_v1_ffdd72fff3ccfc6be88e91097c15ef87e0be38b4780e9f1ce68aae731188d9c0","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_cd28725ecd192d79320824311013bd267aaac131969bbe81362007435ea9e3f5","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_ffdd72fff3ccfc6be88e91097c15ef87e0be38b4780e9f1ce68aae731188d9c0"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_5c21ac281f3717fbc022623ee94cf1b26302c0a4bedd910769f1e8033717239a","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/qwen3.7-plus","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.7-plus","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"87.878788%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02325315795194205","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"URoWjw9Htc6QPwABVcrybw"},"evidence_class":"direct_strict","model_configuration":"qwen3.7-plus","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_5515cdbfedd9d940487b09a63e43e466478b0beb2320b9e0f21fc42720664d6d","canonical_evidence_variant_id":"gpqa_evd_v1_391cc973ca75513e6dc11c93ac32a504b912b4a2b07bb145ed77f3bb6584a8f7","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_5515cdbfedd9d940487b09a63e43e466478b0beb2320b9e0f21fc42720664d6d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_391cc973ca75513e6dc11c93ac32a504b912b4a2b07bb145ed77f3bb6584a8f7"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_5d0b6fa9beb00bd723082609126a60e2124f2f089224e8935b7733ee6207c2d7","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"epoch/gemini-3.1-flash-lite","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemini-3.1-flash-lite_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"81.818182%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027479603010538825","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"nE4hmJo3QV6X5jeGWAeAze"},"evidence_class":"direct_strict","model_configuration":"gemini-3.1-flash-lite_high","organization":"Google","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_7abfe4b2b794eb665fef12fc1a1e137207e333bcaf6baca00b3dd270ef65275f","canonical_evidence_variant_id":"gpqa_evd_v1_dfd90981850dbbd7283dcd1e760b706e7f423fb94a3ca31f10920df8e474d5d8","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_7abfe4b2b794eb665fef12fc1a1e137207e333bcaf6baca00b3dd270ef65275f","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_dfd90981850dbbd7283dcd1e760b706e7f423fb94a3ca31f10920df8e474d5d8"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_5deb548701e55fbdff283a419bd403f4d1d5b178a1239bc33ecc4508382d6b91","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/gemini-3.6-flash","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemini-3.6-flash_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"86.363636%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02445015597318985","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"JoraLb9y55QXpMg4Rrmr5x"},"evidence_class":"direct_strict","model_configuration":"gemini-3.6-flash_low","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_9f54b987dc70091682cf9a98d5d242268c0a722fcbce4ad077b2c539c710abb7","canonical_evidence_variant_id":"gpqa_evd_v1_57cbfb79bddc645ebc4a73a8fe89cafca7357cdaf497dfddd69160ac963e4318","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_9f54b987dc70091682cf9a98d5d242268c0a722fcbce4ad077b2c539c710abb7","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_57cbfb79bddc645ebc4a73a8fe89cafca7357cdaf497dfddd69160ac963e4318"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-13","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_5e269fbf7b862f78bdfb58b4e2d6a616de3f24a13172a171b7911c19e5882ee3","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"epoch/claude-sonnet-4-6","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-sonnet-4-6_medium","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"83.333333%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026552207828215258","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"e8MXnAM8HMF8eEzYtYA7xv"},"evidence_class":"direct_strict","model_configuration":"claude-sonnet-4-6_medium","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_cc3f60648c996eaaa52ecbef11d0122f18e87ca4999935be5a754209ac741c6d","canonical_evidence_variant_id":"gpqa_evd_v1_fae0271f6a0c6911f2f54f368c94835199df8d329517cae7cc978051576c7b30","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_cc3f60648c996eaaa52ecbef11d0122f18e87ca4999935be5a754209ac741c6d","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_fae0271f6a0c6911f2f54f368c94835199df8d329517cae7cc978051576c7b30"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-20","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_5e3089906c8c2437abc826f155bc422a429bb1076e8eaaa770f17c305e3498ab","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/01-ai/Yi-34B-Chat","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Yi-34B-Chat","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"14.741162%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.011372823721448395","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"WxMpEBcTf8qPFPz4BrF86B"},"evidence_class":"direct_compatibility","model_configuration":"Yi-34B-Chat","organization":"01.AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_f5a5438f7b8c161627cb8de0b374ccbefa3ba38c75404f423d0fadf85bfc5f2d","canonical_evidence_variant_id":"gpqa_evd_v1_4a61556655d70c802cbc91fb6fc7184525bb823a524da04da3736a5fd36b9f30","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_f5a5438f7b8c161627cb8de0b374ccbefa3ba38c75404f423d0fadf85bfc5f2d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_4a61556655d70c802cbc91fb6fc7184525bb823a524da04da3736a5fd36b9f30"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_636869f8e677b100d9390e305daa004e4d3cb6ad815d50d572ee8f9ea78baf1f","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/gpt-5.6-luna","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.6-luna_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"82.323232%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027178752639044908","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"nW39ZzuDQ3xTAGVoEfFG95"},"evidence_class":"direct_strict","model_configuration":"gpt-5.6-luna_low","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_a694f7431f860f256c8ff49ce93b03646e6ec24bbc0e969087c3ccf44a5ddd98","canonical_evidence_variant_id":"gpqa_evd_v1_cfab3223935926a78a54266517b83c9ac0cf667c36e40b727410e634cd0c77b9","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_a694f7431f860f256c8ff49ce93b03646e6ec24bbc0e969087c3ccf44a5ddd98","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_cfab3223935926a78a54266517b83c9ac0cf667c36e40b727410e634cd0c77b9"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-08-05","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_67756590ba7c7aa34e9e5076f7e90d0b4251a570d95d5dc77fcb42a6b854f2ee","configuration_identity":{"context_policy":"16k","effort":"unspecified","model_snapshot":"anthropic/claude-opus-4-1-20250805","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-opus-4-1-20250805_16K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"77.272727%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.029857515673386438","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"erLgyKQs2mg2memE9SJvK5"},"evidence_class":"direct_compatibility","model_configuration":"claude-opus-4-1-20250805_16K","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_b193aa6a2e3321b690b8feedd221a05f415cf95482f9316833c1ed630f527370","canonical_evidence_variant_id":"gpqa_evd_v1_d5e29f69d6f79e471f4d87e9c440a4096fd01c3d5fc73d7bf637b9b643fefbee","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_b193aa6a2e3321b690b8feedd221a05f415cf95482f9316833c1ed630f527370","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d5e29f69d6f79e471f4d87e9c440a4096fd01c3d5fc73d7bf637b9b643fefbee"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-05-29","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_678fc1ebbe85018672c472fa4980a0f10c18c89f0dccfe222b93697861ddf099","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"fireworks/accounts/fireworks/models/deepseek-r1-0528","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"fireworks","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"DeepSeek-R1-0528","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"76.325758%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024023818843083005","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"LiRiryJsGb3h5UuxX8bmJv"},"evidence_class":"direct_compatibility","model_configuration":"DeepSeek-R1-0528","organization":"DeepSeek","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_feac116dd3feb7570ec416738dea26a2b6c9d330ac69ac2286da20bc363fda19","canonical_evidence_variant_id":"gpqa_evd_v1_f6ff488a8cf0932cbd19204899fc2efb3e377f74078010108166cab31c2f80cf","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_feac116dd3feb7570ec416738dea26a2b6c9d330ac69ac2286da20bc363fda19","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f6ff488a8cf0932cbd19204899fc2efb3e377f74078010108166cab31c2f80cf"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-21","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_68be4e024b354a452f5a19040775c2d5aecb864abe7845e7832d52a990885029","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"together/databricks/dbrx-instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"together","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"dbrx-instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"32.891414%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03256944333564935","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"TTa3hu7PgRxHgizVreZsdJ"},"evidence_class":"direct_compatibility","model_configuration":"dbrx-instruct","organization":"Databricks","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_e77f3be95d397616315d404450f01d9a96c1a76b6ff583dca4f11237b17a8514","canonical_evidence_variant_id":"gpqa_evd_v1_4605c22099a0980ed8cd2bfc1998083dc4dcfca3c3318b39b1f3cbf743da9771","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e77f3be95d397616315d404450f01d9a96c1a76b6ff583dca4f11237b17a8514","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_4605c22099a0980ed8cd2bfc1998083dc4dcfca3c3318b39b1f3cbf743da9771"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-20","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_68f4f26b76cc464a36f5d412acf484bac54ad4002ceef770bb5a16e84938c442","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/o3-mini-2025-01-31","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"o3-mini-2025-01-31_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"68.181818%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.0331847733384533","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"hJf7Hxg4tgdftQ2QDNHMJA"},"evidence_class":"direct_strict","model_configuration":"o3-mini-2025-01-31_low","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_cbaf44835cd40df0459141c77d8487c65d1c16f08df2868ebd3051fef26f126b","canonical_evidence_variant_id":"gpqa_evd_v1_c257c84e036a375f8ed1ef5dadbd25a685b4c726883640acb3d45e7485443089","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_cbaf44835cd40df0459141c77d8487c65d1c16f08df2868ebd3051fef26f126b","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_c257c84e036a375f8ed1ef5dadbd25a685b4c726883640acb3d45e7485443089"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-20","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_6970bd17ac70725a425ec6d86eb63a2d2c878ee8acefa59ba35a8efaa4180714","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/Qwen/Qwen2-72B-Instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"qwen2-72b-instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"40.782828%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02574288051254005","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Pex6SpYJX8ZPySncfNNy7H"},"evidence_class":"direct_compatibility","model_configuration":"qwen2-72b-instruct","organization":"Alibaba","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_25e1ddd10af1aeefe1602f80e3c4b14035c6b3fd9fcda221ce2345f79001961e","canonical_evidence_variant_id":"gpqa_evd_v1_7d9a6073854f79a6b73296b8bc93de50c344e9bcef48c5aeef95b308f5294e95","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_25e1ddd10af1aeefe1602f80e3c4b14035c6b3fd9fcda221ce2345f79001961e","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_7d9a6073854f79a6b73296b8bc93de50c344e9bcef48c5aeef95b308f5294e95"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-13","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_69e0a76e213450c4c99586311e06c32b0188a7d080456a8a8e66f8c8bfd84fdb","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/gpt-5-nano-2025-08-07","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5-nano-2025-08-07_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"57.575758%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03521224908841589","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"mNZ34pXP8JN779JoHnjNQc"},"evidence_class":"direct_strict","model_configuration":"gpt-5-nano-2025-08-07_low","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_cb255829d8ee31b7dfe578019dec2f345f71f58b4e39ada5063aba69a50b7382","canonical_evidence_variant_id":"gpqa_evd_v1_f4bf55465a437307ef0725d6ccd555d935204df37560205c5a3e6d7efd41e927","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_cb255829d8ee31b7dfe578019dec2f345f71f58b4e39ada5063aba69a50b7382","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_f4bf55465a437307ef0725d6ccd555d935204df37560205c5a3e6d7efd41e927"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_6ad1e1d106980a4cd18c39c471f9ddbfbff5865f9045d01824a081231d140126","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/qwen3.6-plus","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.6-plus","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"88.383838%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.022828881775249357","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"f6ZdMtRpdfskGbrBCzrKSn"},"evidence_class":"direct_strict","model_configuration":"qwen3.6-plus","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_34dd6089244c5b77a8dfb5e171ebefb78748cafc2c60349bddeaacca71a11897","canonical_evidence_variant_id":"gpqa_evd_v1_410175173ea89dce5f3728c07a41e8a63e7d66f1aa76fa80cf8034ca583645b2","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_34dd6089244c5b77a8dfb5e171ebefb78748cafc2c60349bddeaacca71a11897","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_410175173ea89dce5f3728c07a41e8a63e7d66f1aa76fa80cf8034ca583645b2"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_6c8c1dd0bd4400f5eb27a320a475692af2d68a39af98d4f0b552fc0419ecb544","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/gpt-5.4-mini-2026-03-17","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.4-mini-2026-03-17_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"64.141414%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.034169036403915276","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Qyz2Qt5x4P2XwB3w4h4jFR"},"evidence_class":"direct_strict","model_configuration":"gpt-5.4-mini-2026-03-17_none","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_e2b45798d741d675bbda1d8e81f0a3375c43f07e431ec8c8cedede4b33a6b8db","canonical_evidence_variant_id":"gpqa_evd_v1_5f719566029f9039695afd7fa3022bab71168d50ea7d1f7e7bce08d69e1e2143","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e2b45798d741d675bbda1d8e81f0a3375c43f07e431ec8c8cedede4b33a6b8db","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_5f719566029f9039695afd7fa3022bab71168d50ea7d1f7e7bce08d69e1e2143"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-04-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_6fb295e570562ea20a65bcb23229741b5939aa34be0bd746777e7ba4407013f1","configuration_identity":{"context_policy":"source_default","effort":"xhigh","model_snapshot":"anthropic/claude-opus-4-7","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.7","source_model_version":"claude-opus-4-7_xhigh","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"90.151515%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.018464275887293338","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"VJezHrJ5YPCbpui8pZgcJg"},"evidence_class":"direct_strict","model_configuration":"claude-opus-4-7_xhigh","organization":"Anthropic","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_dd89d76ce81d26ca7c7c580858bb2a1c66866deffa835531f58e4dea0c4a0072","canonical_evidence_variant_id":"gpqa_evd_v1_6d2891acb2ca045878b887e2da219434ac2fbbd18a3aac6c6f3d14871f555fdc","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_dd89d76ce81d26ca7c7c580858bb2a1c66866deffa835531f58e4dea0c4a0072","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_6d2891acb2ca045878b887e2da219434ac2fbbd18a3aac6c6f3d14871f555fdc"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-14","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_71cf41df6359bc5fad4222b789a003c9cc8bf989e0345e4db04c1caeaf870496","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4.1-mini-2025-04-14","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4.1-mini-2025-04-14","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"65.845960%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027051500965295915","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"dXHddeGRX7JRVa5UNfad5C"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4.1-mini-2025-04-14","organization":"OpenAI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_2343c3fb0c777d1fae499e1d557cb8ed7cb87826dc5191bd8f6a21c53208ffee","canonical_evidence_variant_id":"gpqa_evd_v1_c5ec14d09aa3a7b4aca91e27c2c73d542f79b32fe4b7c3db3150fe93a8c51d42","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_2343c3fb0c777d1fae499e1d557cb8ed7cb87826dc5191bd8f6a21c53208ffee","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_c5ec14d09aa3a7b4aca91e27c2c73d542f79b32fe4b7c3db3150fe93a8c51d42"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_72074cf78c83a95bdcd075539dbc4c33faf276be230e0adcaf85fd54c8a051ed","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"epoch/claude-sonnet-4-6","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-sonnet-4-6_max","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"78.787879%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.029126522834586777","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"SEePRKRcv8oqn5HpD2KFxB"},"evidence_class":"direct_strict","model_configuration":"claude-sonnet-4-6_max","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_626b81dd77f63dd21786d87d428c9e61aed4851ee7dae2c85490a0fbb875d735","canonical_evidence_variant_id":"gpqa_evd_v1_4720948ed9461d98b96792c144585b799f71e0395f59f66fa2bc14271dda35af","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_626b81dd77f63dd21786d87d428c9e61aed4851ee7dae2c85490a0fbb875d735","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_4720948ed9461d98b96792c144585b799f71e0395f59f66fa2bc14271dda35af"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-23","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_724ee2f69b618a701e0d80941d264c74cb5c2349b4b958a51c605beca4d829c3","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/PRIME-RL/Eurus-2-7B-PRIME","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Eurus-2-7B-PRIME","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"33.901515%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02107317439394748","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"RJm4u2vXF6PUmThJGzAfHE"},"evidence_class":"direct_compatibility","model_configuration":"Eurus-2-7B-PRIME","organization":"Tsinghua University,University of Illinois Urbana-Champaign (UIUC),Shanghai AI Lab,Peking University,Shanghai Jiao Tong University,CUHK Shenzhen Research Institute","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_ecfbe5ae897e0b755a15fc6c08ccdf595d1570019e11de6a5441cde295325a78","canonical_evidence_variant_id":"gpqa_evd_v1_dac8aeaa9e559eae6dec72c411c74f09e7d7cc4345802beffe1c711904671881","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_ecfbe5ae897e0b755a15fc6c08ccdf595d1570019e11de6a5441cde295325a78","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_dac8aeaa9e559eae6dec72c411c74f09e7d7cc4345802beffe1c711904671881"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_73606fc2f28e306f4091aca42290cd5715fed6ba0d38ff358cc6a8ec8c2cc294","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/qwen3.5-flash","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.5-flash","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"82.323232%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027178752639044908","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"S7wbot7JR53BAjM8F9gaws"},"evidence_class":"direct_strict","model_configuration":"qwen3.5-flash","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_007966d785ed20dd83217409daa7620a16ff82f40c2b1dd8da0f34590b53b297","canonical_evidence_variant_id":"gpqa_evd_v1_39247dc711ff5ce6f23d19de68453a2aef23f1f158b08e99cf733dc15655f3f4","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_007966d785ed20dd83217409daa7620a16ff82f40c2b1dd8da0f34590b53b297","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_39247dc711ff5ce6f23d19de68453a2aef23f1f158b08e99cf733dc15655f3f4"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_7422854d7e27c6743941108a0a9a0f278c02b92c202c12726abb5c9bc544adf5","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/claude-opus-4-8","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-opus-4-8_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"85.353535%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.025190921114603887","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"ckbZdkrXPwbFdjn6bmUQsi"},"evidence_class":"direct_strict","model_configuration":"claude-opus-4-8_none","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_6c1185f91a27e0422602ee91305797592531d5e0ea144e8c4368cf375c360fdb","canonical_evidence_variant_id":"gpqa_evd_v1_0de7b89e2b44977dab0e6ffc0914fa006f3ae8b20f8113c181a6fa44f9281409","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_6c1185f91a27e0422602ee91305797592531d5e0ea144e8c4368cf375c360fdb","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_0de7b89e2b44977dab0e6ffc0914fa006f3ae8b20f8113c181a6fa44f9281409"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-05-05","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_74454bd1c2bf40c9cfef8af6222ded9c2c33796a97bb3e76f4227ab403ab23f6","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"openai/gpt-5.5","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.8","source_model_version":"gpt-5.5_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"90.656566%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.017532658377721708","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"2CdomqtBVz2it7a2Zx2G2t"},"evidence_class":"direct_strict","model_configuration":"gpt-5.5_low","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_ed7957e830968d6f66dac5ed25a1dde55aa7b4fc6bdfb2958e0565fcb1fca766","canonical_evidence_variant_id":"gpqa_evd_v1_f96ec6c10dcee032be7c4c6b94888a4b94c8768f45e35c94427de90290818efd","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_ed7957e830968d6f66dac5ed25a1dde55aa7b4fc6bdfb2958e0565fcb1fca766","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f96ec6c10dcee032be7c4c6b94888a4b94c8768f45e35c94427de90290818efd"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-08-05","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_758fdeb186e7e3aa8d76746a779e63abb3f3d53282b88caf7802f648df61985a","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-opus-4-1-20250805","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-opus-4-1-20250805","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"73.232323%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.031544498882702825","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"j62bkFDjBSeFyVpnGzuzg4"},"evidence_class":"direct_compatibility","model_configuration":"claude-opus-4-1-20250805","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_061c2f393ff622d4e751ae0e08d01b67d4af4b313eff6a92d9781a5c11104f22","canonical_evidence_variant_id":"gpqa_evd_v1_8400f0f288172f70235d3c08757a5ec0e10f3de0859cdf05109f4832d65145bd","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_061c2f393ff622d4e751ae0e08d01b67d4af4b313eff6a92d9781a5c11104f22","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_8400f0f288172f70235d3c08757a5ec0e10f3de0859cdf05109f4832d65145bd"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-11-19","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_75b5c82ad8fa3f7520d6c30258334fd2da432a999b7c23be7a10a04ab6bfe1de","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-3-pro-preview","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.2","source_model_version":"gemini-3-pro-preview","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"92.613636%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.016526963574741","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Cg54udQGcTCnfDGX2qePLZ"},"evidence_class":"direct_strict","model_configuration":"gemini-3-pro-preview","organization":"Google DeepMind","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_1d7073c1b5ec278ebd0b70e087c6c9cdcc58096a06be900979e39c7b90a2a4da","canonical_evidence_variant_id":"gpqa_evd_v1_d67b926cc255d512d885fef894902007d4e6404ec6870907d3fa016f3e806a1e","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_1d7073c1b5ec278ebd0b70e087c6c9cdcc58096a06be900979e39c7b90a2a4da","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_d67b926cc255d512d885fef894902007d4e6404ec6870907d3fa016f3e806a1e"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_75c4a32c0454fc968aa549dba870cfd4e1bd5a1f6c823d4e434b9e24f6f77f18","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/gemini-3.1-flash-lite","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemini-3.1-flash-lite_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"74.242424%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03115626951964683","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"99ys2jgQiPQ5UwrqnazfqA"},"evidence_class":"direct_strict","model_configuration":"gemini-3.1-flash-lite_low","organization":"Google","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_0b974c884dabb85580f0492f3b4acbedb0255440c18b9267d10c8ce520d381ba","canonical_evidence_variant_id":"gpqa_evd_v1_41e33403f86e686fd1e2a0dadcd443df22ea63ec8bc3bf8d57b05cb92dba1af2","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_0b974c884dabb85580f0492f3b4acbedb0255440c18b9267d10c8ce520d381ba","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_41e33403f86e686fd1e2a0dadcd443df22ea63ec8bc3bf8d57b05cb92dba1af2"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-02-05","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_766ddf5c78cac9398614df89b5abd827a064cc4b8b5645bfcd7cbdf12dbc9cc8","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4o-2024-11-20","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4o-2024-11-20","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"47.885101%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026336972178726107","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"gULKGqAe9nG9syxmYMtkhi"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4o-2024-11-20","organization":"OpenAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_6d26a2eb1f68b5c6272c41202e6e75991396d01079402fb6242dd14dad3d1390","canonical_evidence_variant_id":"gpqa_evd_v1_7289d0bfa90f78c9d37e3c289ff16a5e5d4335b72db4b1f0ba5fd894f405989c","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_6d26a2eb1f68b5c6272c41202e6e75991396d01079402fb6242dd14dad3d1390","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_7289d0bfa90f78c9d37e3c289ff16a5e5d4335b72db4b1f0ba5fd894f405989c"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_7671edcdc19c9d32865c60525cb98415cb4da59e1b67ca91aa8f1c27e9a823bf","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4-0613","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4-0613","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"30.650253%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02337322948270623","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"fu7pWqbSC4MUBdwv3GFNa5"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4-0613","organization":"OpenAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_3ba386640515238cce493e6e6214e420a8059589db6fb0812ded064d8cc542bc","canonical_evidence_variant_id":"gpqa_evd_v1_e191866acc8afba0d0bfd1e173f45f1af06507777c8d67c91ad414558d64d708","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_3ba386640515238cce493e6e6214e420a8059589db6fb0812ded064d8cc542bc","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_e191866acc8afba0d0bfd1e173f45f1af06507777c8d67c91ad414558d64d708"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-10-28","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_7855e86e24fde2ea81d498e4c025c7acf2b0fe6d817cfe394ef7306ce637d247","configuration_identity":{"context_policy":"16k","effort":"unspecified","model_snapshot":"anthropic/claude-sonnet-4-5-20250929","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.0","source_model_version":"claude-sonnet-4-5-20250929_16K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"78.787879%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02912652283458678","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"LkVuUxvsHNzpcg3sUiH7f3"},"evidence_class":"direct_compatibility","model_configuration":"claude-sonnet-4-5-20250929_16K","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_f3fb6e2596a70a9bbb7d4d4214fdf65dad510be6e18542ee1f1911f59e8ea9e8","canonical_evidence_variant_id":"gpqa_evd_v1_297403a07270c32fd998ca518f6cdf5668c956259038562f2be763279256da59","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_f3fb6e2596a70a9bbb7d4d4214fdf65dad510be6e18542ee1f1911f59e8ea9e8","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_297403a07270c32fd998ca518f6cdf5668c956259038562f2be763279256da59"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-02-28","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_788a0f0a26e95b58f7bad9146781e148246e45068dfba74be09ffb9e967a5f8a","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4.5-preview-2025-02-27","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4.5-preview-2025-02-27","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"68.686869%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.033042050878136546","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"dodibSv7pk8GtqeMy4YLQg"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4.5-preview-2025-02-27","organization":"OpenAI","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":7,"canonical_evaluation_run_id":"gpqa_run_v1_d0053cf20fd93eaa5868c7700e33fef962ac314e29c59e42d96b68879014d4d7","canonical_evidence_variant_id":"gpqa_evd_v1_a94afa1168cdc0773d4dda8f253740979e77cf3cf54f55f8aa8ef4a601d5ad75","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_d0053cf20fd93eaa5868c7700e33fef962ac314e29c59e42d96b68879014d4d7","evidence_class":"derived_uniform","evidence_variant_id":"gpqa_evd_v1_a94afa1168cdc0773d4dda8f253740979e77cf3cf54f55f8aa8ef4a601d5ad75"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-04-24","configuration_date_basis":"validated Epoch CSV Started at metadata","configuration_id":"gpqa_cfg_v1_7934de713f82e39e9592c19190f2c0c4b7a5afc288e8b146a975a0b4243d3076","configuration_identity":{"context_policy":"source_default","effort":"xhigh","model_snapshot":"epoch/gpt-5.5-pro-pre-release_xhigh","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_source_task_v1","protocol":"mean@7","provider_route":"epoch_import","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"7","declared_attempts_per_question":"8","excluded_epoch_numbers":"5","source_model_version":"gpt-5.5-pro-pre-release_xhigh"}},"details":{"configuration_date_basis":"validated Epoch CSV Started at metadata","evidence_label":"Derived uniform-epoch subset","reconciliation":{"paired_original_role":"Reconstructed from the selected uniform subset; not the native Epoch aggregate.","published_source_score_percent":"93.921356%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.015832823810775415","status":"exact"},"recovery":{"declared_attempts_per_question":8,"kind":"uniform_epoch_subset","omitted_attempt_count":1,"selected_attempts_per_question":7},"score_source":"Accepted binary choice scores from the complete uniform attempt subset","source_kind_label":"Epoch source with globally uniform attempt subset","source_run_id":"cBmhJzrC3SkfZ7uVgGVYZX"},"evidence_class":"derived_uniform","model_configuration":"gpt-5.5-pro-pre-release_xhigh","organization":"OpenAI","protocol":"mean@7","validation_profile":"derived_uniform_epoch_subset"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_e5ae10b7fa25360b7eab7f9fbf1ab5109372ab3b9e4f3b062ad7eead22f752dd","canonical_evidence_variant_id":"gpqa_evd_v1_99902969eaf1717bb1a5557a5e692c80987156860a36229e2c160b0f37eb0d17","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e5ae10b7fa25360b7eab7f9fbf1ab5109372ab3b9e4f3b062ad7eead22f752dd","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_99902969eaf1717bb1a5557a5e692c80987156860a36229e2c160b0f37eb0d17"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-31","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_79ea3c485cc923af3af9258fe5fa39ad7844f3c46260f2a10709c1ff9e1b57e6","configuration_identity":{"context_policy":"128k","effort":"medium","model_snapshot":"sagemaker/microsoft/Phi-3-medium-128k-instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Phi-3-medium-128k-instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"27.588384%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.016117685500292727","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"bDS2AGEf3AywbVcdDyAKZK"},"evidence_class":"direct_compatibility","model_configuration":"Phi-3-medium-128k-instruct","organization":"Microsoft","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":7,"canonical_evaluation_run_id":"gpqa_run_v1_4064f9f302dd20b904afc43b14493d0a8bb8a5fcf6febc63c3c6b4412c7c1118","canonical_evidence_variant_id":"gpqa_evd_v1_7d210d7ef8e66f427900f281667fcd9ca91b814e5ac9d1218fd88c7f8a6f44db","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_4064f9f302dd20b904afc43b14493d0a8bb8a5fcf6febc63c3c6b4412c7c1118","evidence_class":"derived_uniform","evidence_variant_id":"gpqa_evd_v1_7d210d7ef8e66f427900f281667fcd9ca91b814e5ac9d1218fd88c7f8a6f44db"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-08-07","configuration_date_basis":"validated Epoch CSV Started at metadata","configuration_id":"gpqa_cfg_v1_7c3ec07ae3ad1a5f35ccd1d892d17caa2ff48036699a386811c33885de02853a","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"epoch/gpt-5-2025-08-07_medium","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_source_task_v1","protocol":"mean@7","provider_route":"epoch_import","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"7","declared_attempts_per_question":"8","excluded_epoch_numbers":"7","source_model_version":"gpt-5-2025-08-07_medium"}},"details":{"configuration_date_basis":"validated Epoch CSV Started at metadata","evidence_label":"Derived uniform-epoch subset","reconciliation":{"paired_original_role":"Reconstructed from the selected uniform subset; not the native Epoch aggregate.","published_source_score_percent":"85.353535%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.021314825297984473","status":"exact"},"recovery":{"declared_attempts_per_question":8,"kind":"uniform_epoch_subset","omitted_attempt_count":1,"selected_attempts_per_question":7},"score_source":"Accepted binary choice scores from the complete uniform attempt subset","source_kind_label":"Epoch source with globally uniform attempt subset","source_run_id":"YQCwJRoGbGAqNCQoaL6cjk"},"evidence_class":"derived_uniform","model_configuration":"gpt-5-2025-08-07_medium","organization":"OpenAI","protocol":"mean@7","validation_profile":"derived_uniform_epoch_subset"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_9c0f340b80401c41bb60c544ab1d8468f1fdf56e4acbb2f32b9020be0ba25cba","canonical_evidence_variant_id":"gpqa_evd_v1_3accbd60ebe0b166a74b437856fea71c18e59901ce28a1cc35f3c26f7135e4b4","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_9c0f340b80401c41bb60c544ab1d8468f1fdf56e4acbb2f32b9020be0ba25cba","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_3accbd60ebe0b166a74b437856fea71c18e59901ce28a1cc35f3c26f7135e4b4"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_7db9c0fa233c93399222a3bfbac6056222f2bcacd6282d5bb91bc0f71c37ae84","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/gpt-5.6-terra","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.6-terra_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"77.272727%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.029857515673386438","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"F7v3JtKFb27j2YVmsndLM5"},"evidence_class":"direct_strict","model_configuration":"gpt-5.6-terra_none","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_6fdcfcbc939cac11581b5cd167f87ce8ad1a8c643783737ff83f0d36485b9eec","canonical_evidence_variant_id":"gpqa_evd_v1_908d2fb385d65caf4e9235b3a9d0cadd7f65a37115bc964222f5247ad188dadc","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_6fdcfcbc939cac11581b5cd167f87ce8ad1a8c643783737ff83f0d36485b9eec","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_908d2fb385d65caf4e9235b3a9d0cadd7f65a37115bc964222f5247ad188dadc"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-21","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_7f2ac119b9c9f420c06ad2e6f93c6e2e82828c27cec46dd73558c7faf28e5573","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/mistralai/Mistral-7B-Instruct-v0.3","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Mistral-7B-Instruct-v0.3","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"15.183081%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.010148891037227303","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"YbQkMPC2GtNLhxqTEhhVsa"},"evidence_class":"direct_compatibility","model_configuration":"Mistral-7B-Instruct-v0.3","organization":"Mistral AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_d0ba371b076810f91175ff17795e3a89abf2616e381bc55dff6bd330861faaf8","canonical_evidence_variant_id":"gpqa_evd_v1_42dfda83b547f30bf90fc46faf232f79087f3a861423aae0c5efee1c7b6dc543","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_d0ba371b076810f91175ff17795e3a89abf2616e381bc55dff6bd330861faaf8","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_42dfda83b547f30bf90fc46faf232f79087f3a861423aae0c5efee1c7b6dc543"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-11-13","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_7f8e8f18de3290be2ccfe3701c2d2a4c8a6512f34dc46b7db05c5915a94d40e5","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"openai/gpt-5.1","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.1","source_model_version":"gpt-5.1-2025-11-13_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"87.626263%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.01875368281077595","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"CSVtMksEVjBNsGJGjW4GcQ"},"evidence_class":"direct_strict","model_configuration":"gpt-5.1-2025-11-13_high","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_b95c5f6b24ce76ade25a8510d7dfa4dcad0b3f60d6ff9a5419bd7149873c5c88","canonical_evidence_variant_id":"gpqa_evd_v1_9c937b91510ffd935a34c98172ef6e8050865a7c13658a0e93aa458d42f33b6f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_b95c5f6b24ce76ade25a8510d7dfa4dcad0b3f60d6ff9a5419bd7149873c5c88","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_9c937b91510ffd935a34c98172ef6e8050865a7c13658a0e93aa458d42f33b6f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-01","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_8377771cdc0ba4e2788fa70543be3075e24b56931201a00dc7de01ffaebc6583","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"fireworks/accounts/fireworks/models/deepseek-v3-0324","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"fireworks","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"DeepSeek-V3-0324","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"67.613636%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026836302525030095","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"VjfLwKbT6kTprYsMGcz2YT"},"evidence_class":"direct_compatibility","model_configuration":"DeepSeek-V3-0324","organization":"DeepSeek","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_bdbf986b2d3259bd5c00ee776060fe30b54bacb6b4aff74149a800bf87be7cb9","canonical_evidence_variant_id":"gpqa_evd_v1_0fd65c0260532a12b47d3506b0ab22b48168dba2e1e02c07a0a9deb2ec82de1f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_bdbf986b2d3259bd5c00ee776060fe30b54bacb6b4aff74149a800bf87be7cb9","evidence_class":"derived_uniform","evidence_variant_id":"gpqa_evd_v1_0fd65c0260532a12b47d3506b0ab22b48168dba2e1e02c07a0a9deb2ec82de1f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-01-29","configuration_date_basis":"validated Epoch CSV Started at metadata","configuration_id":"gpqa_cfg_v1_8481fd6a05a2aa64f8278e43afe1cb7914779e0b788942a039de1af847ce744b","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/glm-4.7","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_source_task_v1","protocol":"mean@1","provider_route":"epoch_import","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","declared_attempts_per_question":"2","excluded_epoch_numbers":"1","source_model_version":"glm-4.7"}},"details":{"configuration_date_basis":"validated Epoch CSV Started at metadata","evidence_label":"Derived uniform-epoch subset","reconciliation":{"paired_original_role":"Reconstructed from the selected uniform subset; not the native Epoch aggregate.","published_source_score_percent":"83.333333%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02374901666263655","status":"exact"},"recovery":{"declared_attempts_per_question":2,"kind":"uniform_epoch_subset","omitted_attempt_count":1,"selected_attempts_per_question":1},"score_source":"Accepted binary choice scores from the complete uniform attempt subset","source_kind_label":"Epoch source with globally uniform attempt subset","source_run_id":"6phXzd4aCbnBUSQmBsaywa"},"evidence_class":"derived_uniform","model_configuration":"glm-4.7","organization":"Z.ai (Zhipu AI)","protocol":"mean@1","validation_profile":"derived_uniform_epoch_subset"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_46f3f78481720a365d57af638878d0ec26af8758ab8886c8eb51449c568cbe01","canonical_evidence_variant_id":"gpqa_evd_v1_2ddbb950ca1860d699684d4213cdacf2330dcf95cbf97b94493b14e198e477dd","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_46f3f78481720a365d57af638878d0ec26af8758ab8886c8eb51449c568cbe01","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_2ddbb950ca1860d699684d4213cdacf2330dcf95cbf97b94493b14e198e477dd"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-03-10","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_84be1c2e599b07f80232212023c305f33061879967f9ca1cd6e6a2aa92d3430f","configuration_identity":{"context_policy":"32k","effort":"unspecified","model_snapshot":"anthropic/claude-3-7-sonnet-20250219","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-3-7-sonnet-20250219_32K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"76.767677%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.030088629490217445","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"iv9yfwZ9vTznxQfvJF6J7p"},"evidence_class":"direct_compatibility","model_configuration":"claude-3-7-sonnet-20250219_32K","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_694d453f03f600041e46e00ce0428ffb49531b0c37a69594aaa5288ba588fb6c","canonical_evidence_variant_id":"gpqa_evd_v1_f0076ff82b55cc6aa5ba6b90b2b2bb73c109e21b23b3b83961abf972c178d0c9","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_694d453f03f600041e46e00ce0428ffb49531b0c37a69594aaa5288ba588fb6c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_f0076ff82b55cc6aa5ba6b90b2b2bb73c109e21b23b3b83961abf972c178d0c9"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_84d05ad544dcd1793dbdc24f941a24829266c8784cb6aeac8d0b08838b5b63f4","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/gpt-5.6-sol","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.6-sol_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"89.898990%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.021469735576055356","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"6cmgkNAQZekBr6e6rVdLe9"},"evidence_class":"direct_strict","model_configuration":"gpt-5.6-sol_low","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_f0268b3f6b184e0701abb740ed28efa27bd588b9292ad027ebeacd5787b305f4","canonical_evidence_variant_id":"gpqa_evd_v1_c93b6e72660dcbb86bcca25a2fe0a3780b2f0711ceefa8edb838cd88fe811031","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_f0268b3f6b184e0701abb740ed28efa27bd588b9292ad027ebeacd5787b305f4","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_c93b6e72660dcbb86bcca25a2fe0a3780b2f0711ceefa8edb838cd88fe811031"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-04-14","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_850dd081ff0af5e05ae06d189189973570eabc763dcdf7b9c49aeb6b97112d76","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"openai/gpt-5.4-nano-2026-03-17","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.6","source_model_version":"gpt-5.4-nano-2026-03-17_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"+0.002222 pp","published_source_score_percent":"78.470000%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.0244","status":"accepted_within_0_05pp"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"fZB3qGSohAVYzN97mniuUp"},"evidence_class":"direct_strict","model_configuration":"gpt-5.4-nano-2026-03-17_high","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_ea1888a35618a235b72c73014b2b8a255e891f289fe2e9f58127eab6738b782c","canonical_evidence_variant_id":"gpqa_evd_v1_6bc2ac5f0c651c7dcfe59063ccdfdaa19bc85d3a1340e90e42815b766a7820ed","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_ea1888a35618a235b72c73014b2b8a255e891f289fe2e9f58127eab6738b782c","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_6bc2ac5f0c651c7dcfe59063ccdfdaa19bc85d3a1340e90e42815b766a7820ed"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-17","configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","configuration_id":"gpqa_cfg_v1_86c15018f283c868987b63ef745efbca663fee38b52cc60a94981c8efd381fab","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"deepseek/deepseek-v4-flash","output_policy":"no_output_ceiling","prompt_contract":"gpqa_native_answer_choice_v1","protocol":"mean@1","provider_route":"deepseek","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","generation_config_sha256":"4c15636f00f8fc64ee7cd1c33970689264e28de9e36a6c725a08a6724f44ce40","profile":"deepseek-v4-flash-max","task_sha256":"2501e59954a9e53479bc5e44bf9923db905866019f16c385ea3f27483b3697d0"}},"details":{"configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","evidence_label":"Independent native run","execution":{"automatic_retries":0,"fresh_calls":true,"output_ceiling":"none"},"reconciliation":{"paired_original_role":"Reconstructed from the same accepted all-198 native attempt universe as the Clean score.","published_source_score_percent":null,"status":"not_applicable_independent_run"},"recovery":{"kind":"none"},"score_source":"Independent native answer-safe projection","source_kind_label":"Independent project-native run","source_run_id":"gpqa-native-deepseek-v4-flash-max-all198-20260816-r1"},"evidence_class":"project_native","model_configuration":"DeepSeek V4 Flash (max)","organization":"DeepSeek","protocol":"mean@1","validation_profile":"strict_all_normal"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_fd1b02832b3a497ea2ddebabaf67d89d1969262cab76cc266f44c5f7261faefa","canonical_evidence_variant_id":"gpqa_evd_v1_70611083db50c82b95c428288e25680fc6e62ad8850949f6b3c45bb650df36d3","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_fd1b02832b3a497ea2ddebabaf67d89d1969262cab76cc266f44c5f7261faefa","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_70611083db50c82b95c428288e25680fc6e62ad8850949f6b3c45bb650df36d3"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_8c7fa3c8345975553ad175f719559715fdda562d71eb9822dec2cf33e6dfeb1e","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4-0125-preview","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4-0125-preview","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"42.266414%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026795949166711065","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"4j3WRhhruXrfaq9Jki3NDc"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4-0125-preview","organization":"OpenAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_dfc5597ff3f4619ab2c684c61d8977155d5597fbcb6df734467f38c9f4af5c60","canonical_evidence_variant_id":"gpqa_evd_v1_a4dbad6bcf403b3842bdf7f7d57b60ba54618170886586f899fc4cc7beab88c3","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_dfc5597ff3f4619ab2c684c61d8977155d5597fbcb6df734467f38c9f4af5c60","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_a4dbad6bcf403b3842bdf7f7d57b60ba54618170886586f899fc4cc7beab88c3"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_8ce85e5a084acc7fe3dcf3183d33f605314c4189cf5aec161b8576be34f00bd1","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"epoch/o3-2025-04-16","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"o3-2025-04-16_medium","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"80.808081%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.028057791672989062","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"TBYA7JWneLUCy9kyktFBbt"},"evidence_class":"direct_strict","model_configuration":"o3-2025-04-16_medium","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_d87f5e0e4fe7e7954e6a342801578c69a887560bf781fa8f5764c150865c4845","canonical_evidence_variant_id":"gpqa_evd_v1_15af887c85b934814a755440ae3f0d603f5fdbd1cc2a87c1c78969242d02497f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_d87f5e0e4fe7e7954e6a342801578c69a887560bf781fa8f5764c150865c4845","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_15af887c85b934814a755440ae3f0d603f5fdbd1cc2a87c1c78969242d02497f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-10","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_8d361a9663df4edd4bbe91b30f4d62bbd4be4a7a778484c5a867483645d19a6c","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/glm-5.2","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"glm-5.2_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"71.212121%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03225883512300997","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"YDQJAMgMjLaikPkeVDbAjr"},"evidence_class":"direct_strict","model_configuration":"glm-5.2_none","organization":"Z.ai (Zhipu AI)","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_b0602ac739ac95eb15086eed34884e50a032e0ce63a66984b9d8bda38e7b6c30","canonical_evidence_variant_id":"gpqa_evd_v1_d32231861767d662305da1d39e78d304940ef6d207e9b52b3f170616bd8a2721","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_b0602ac739ac95eb15086eed34884e50a032e0ce63a66984b9d8bda38e7b6c30","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d32231861767d662305da1d39e78d304940ef6d207e9b52b3f170616bd8a2721"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-02-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_8d5c9aff7847b4d80b03824f778e301bb2f71d82a7d54bc8752a32b2d93880b9","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-2.0-pro-exp-02-05","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemini-2.0-pro-exp-02-05","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"65.656566%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03383201223244441","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"g5YE7G6DXb45EH4HxEdiht"},"evidence_class":"direct_compatibility","model_configuration":"gemini-2.0-pro-exp-02-05","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_dc613f0a761c2857feb508ffcd540d632d23ed7bbab5e6c1c2acff6df1cd66f7","canonical_evidence_variant_id":"gpqa_evd_v1_184f378cffcfabf6ab4fa02c4adfd83638aee252429f0ba058653c6ecb3e41fb","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_dc613f0a761c2857feb508ffcd540d632d23ed7bbab5e6c1c2acff6df1cd66f7","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_184f378cffcfabf6ab4fa02c4adfd83638aee252429f0ba058653c6ecb3e41fb"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_8db4d324508b283804b4f687b846605f27d9446a2eb2c449127b198717926f5a","configuration_identity":{"context_policy":"source_default","effort":"xhigh","model_snapshot":"epoch/gpt-5.4-mini-2026-03-17","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.4-mini-2026-03-17_xhigh","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"86.868687%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02406315641682249","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"gKB2rCD7mHzFN2MCt2YCni"},"evidence_class":"direct_strict","model_configuration":"gpt-5.4-mini-2026-03-17_xhigh","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_95795a18eae1390656174f941596019018183dc96fa60299a2ea18a566d4d000","canonical_evidence_variant_id":"gpqa_evd_v1_3f0bb4738ba6e2ff2de8408943325693e40959a162efdddf5977f4860aa470c3","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_95795a18eae1390656174f941596019018183dc96fa60299a2ea18a566d4d000","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_3f0bb4738ba6e2ff2de8408943325693e40959a162efdddf5977f4860aa470c3"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_8dc9a927af2441ac3a54b771539215cedc19d0bdaf6864c0faff20a39f9e062f","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-3-haiku-20240307","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-3-haiku-20240307","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"36.300505%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024180910672074975","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"T3FWdCCWBryVjxwTcoY5bK"},"evidence_class":"direct_compatibility","model_configuration":"claude-3-haiku-20240307","organization":"Anthropic","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_88dda16a219b4deefb03b1b8c48485c39684eba5099d7bb57271704b446d2532","canonical_evidence_variant_id":"gpqa_evd_v1_665b4a95100cff19984ec1fe10ff9d2230434d77d3c7a09e39ee3e3cbb8a4ff7","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_88dda16a219b4deefb03b1b8c48485c39684eba5099d7bb57271704b446d2532","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_665b4a95100cff19984ec1fe10ff9d2230434d77d3c7a09e39ee3e3cbb8a4ff7"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_8f361ae1265b1142ff8c109e0b1829cdd625d41f3ee9506a6cdcdbab3641ee73","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/qwen3.7-plus","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.7-plus_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"81.818182%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027479603010538825","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"BWz8uig4QyqPtke59pM88z"},"evidence_class":"direct_strict","model_configuration":"qwen3.7-plus_none","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_7801ae2fa3603319120be85a5b8e63b4cdf0acd885082a516aca085cf8832ada","canonical_evidence_variant_id":"gpqa_evd_v1_a773a87326a0a16be70727919d9511d8f0eca09dd21227358031ac9cbfafa55e","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_7801ae2fa3603319120be85a5b8e63b4cdf0acd885082a516aca085cf8832ada","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_a773a87326a0a16be70727919d9511d8f0eca09dd21227358031ac9cbfafa55e"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_902ec55e0bbd019f00882dc1d375e59dbbf9bd5154a35f0daad1099813288565","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/o1-preview-2024-09-12","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"o1-preview-2024-09-12","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"50.315657%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.028257821072909393","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"7CkADTsoNbDd88uqYbqfwj"},"evidence_class":"direct_compatibility","model_configuration":"o1-preview-2024-09-12","organization":"OpenAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_f1ed3cb9a6241703507b26c9843c16274c913bbf9fa5e1aca5256344444bdf58","canonical_evidence_variant_id":"gpqa_evd_v1_87cc644071a83e463191e8a4df404665af7a8397aec7455b408e4b91a3ba015c","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_f1ed3cb9a6241703507b26c9843c16274c913bbf9fa5e1aca5256344444bdf58","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_87cc644071a83e463191e8a4df404665af7a8397aec7455b408e4b91a3ba015c"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-05-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_90806d19c82368364d7c0c55ac1dcd465817154c590e19fccb69699999696009","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-opus-4-20250514","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-opus-4-20250514","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"69.191919%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03289477330098615","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"EnjJ8bGi9vRNNQFzfkLCkQ"},"evidence_class":"direct_compatibility","model_configuration":"claude-opus-4-20250514","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_da0ad14552640843bcc7826c443d6dcbeff2f28482f5d200205d2ace274e80e6","canonical_evidence_variant_id":"gpqa_evd_v1_cadf02c95b51f6a80c92e672acd6a4d0545c03ae524cb69f503648a922ce4da3","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_da0ad14552640843bcc7826c443d6dcbeff2f28482f5d200205d2ace274e80e6","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_cadf02c95b51f6a80c92e672acd6a4d0545c03ae524cb69f503648a922ce4da3"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-31","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_90e5ce6b4a67252218334c529e05589fdb776c6a026b382161690d249ff27e18","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"openai/o3-mini-2025-01-31","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"o3-mini-2025-01-31_medium","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"74.273990%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.025411943362857046","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"ehvA5nisC7GMbgbyy3Z4Et"},"evidence_class":"direct_compatibility","model_configuration":"o3-mini-2025-01-31_medium","organization":"OpenAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_d6eefa5846e1079c4320fb7305fd2ac524ddefaad2af3a84562582a79a040e9f","canonical_evidence_variant_id":"gpqa_evd_v1_0f572fc59b5afdca713ce1314b8c70bde32f8a7f0ca91a9354cecc032867983e","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_d6eefa5846e1079c4320fb7305fd2ac524ddefaad2af3a84562582a79a040e9f","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_0f572fc59b5afdca713ce1314b8c70bde32f8a7f0ca91a9354cecc032867983e"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-23","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_9176a8be353931cb51348dc081ae7761a85c8a92f75a1939ae96e66f66b67305","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"openai/o1-2024-12-17","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"o1-2024-12-17_medium","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"75.757576%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.030532892233932022","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Uczhz7MKLstjLkSbbGuwdN"},"evidence_class":"direct_compatibility","model_configuration":"o1-2024-12-17_medium","organization":"OpenAI","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_171cf451c16bc9a1bf1bc32e261cf4616708b5f4be30c11a4fd6733b290ab59d","canonical_evidence_variant_id":"gpqa_evd_v1_76b04aaf9ed0e8d643348ea46e48e52224ed5df3fd7fe47410106a956c73ce5d","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_171cf451c16bc9a1bf1bc32e261cf4616708b5f4be30c11a4fd6733b290ab59d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_76b04aaf9ed0e8d643348ea46e48e52224ed5df3fd7fe47410106a956c73ce5d"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_91d52572fee638e3cdcba8d3d9db001f9c9ce30df32420006ed56c7f4c01fca1","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/qwen3.6-flash","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.6-flash","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"83.333333%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026552207828215258","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"H2FvRXagQRQmiNa7WrNMU2"},"evidence_class":"direct_strict","model_configuration":"qwen3.6-flash","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_8742f5d043b230a117ab97f3722ca16013881b722b9406f2d2891e8c4eb57b28","canonical_evidence_variant_id":"gpqa_evd_v1_623f1f7d3a7483eb695cbbed60452bcbf7f61b8493ba36f0ba7f480c33e2f8a3","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_8742f5d043b230a117ab97f3722ca16013881b722b9406f2d2891e8c4eb57b28","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_623f1f7d3a7483eb695cbbed60452bcbf7f61b8493ba36f0ba7f480c33e2f8a3"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-14","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_934610fb54d8c86a7bea8d514dee077ee2eb5884977abbe065c70009f334e3b7","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4.1-2025-04-14","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4.1-2025-04-14","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"66.919192%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02821440093855908","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Gj8pd5gU9r4LqwDK8XZ4mL"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4.1-2025-04-14","organization":"OpenAI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_c6bd22f4f5247b1f051af18d19da4bdc9c1f2d209d93da1064e0b851bfd1fd34","canonical_evidence_variant_id":"gpqa_evd_v1_a824ee6311547ec13d7bd99dbe7f19fad0f95c1046e582fa09887b8f15ef5e6b","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_c6bd22f4f5247b1f051af18d19da4bdc9c1f2d209d93da1064e0b851bfd1fd34","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_a824ee6311547ec13d7bd99dbe7f19fad0f95c1046e582fa09887b8f15ef5e6b"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-21","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_937900a08828620bb405325f43551909f02e0fa51fe3dcfd74cfe0f34f2404f2","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/deepseek-ai/deepseek-llm-67b-chat","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"deepseek-llm-67b-chat","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"24.621212%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.013623351950761947","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"82RTgqgiC2UuM5M37L3WWM"},"evidence_class":"direct_compatibility","model_configuration":"deepseek-llm-67b-chat","organization":"DeepSeek","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_ac9984b33f79389262288d99038c56128b471ee5a1e2476e05b98dd232e3d21a","canonical_evidence_variant_id":"gpqa_evd_v1_ffff552e9fd10ae3b515b91b12b881f737cf27f85b92163d1d0a7143e73a9bc0","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_ac9984b33f79389262288d99038c56128b471ee5a1e2476e05b98dd232e3d21a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_ffff552e9fd10ae3b515b91b12b881f737cf27f85b92163d1d0a7143e73a9bc0"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_9b2222a171dd46e01575f8e6e673c3b8b1659464ddb17ad4146c5e1da3a26a81","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"epoch/claude-fable-5","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-fable-5_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"83.333333%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026552207828215258","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"d23hYvSNGFo7rniLFyBWiu"},"evidence_class":"direct_strict","model_configuration":"claude-fable-5_high","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_35ca9ef5130ca50321fbfd90738cb342d545fa463d231948898b65ec476cab64","canonical_evidence_variant_id":"gpqa_evd_v1_d5547fadfac67fbb8f45b56678573e2f1a140b03887f0cf581f3cc0228953155","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_35ca9ef5130ca50321fbfd90738cb342d545fa463d231948898b65ec476cab64","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_d5547fadfac67fbb8f45b56678573e2f1a140b03887f0cf581f3cc0228953155"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-11-24","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_9d62b338e275defb1013265362d0e2e54512238ef149f471ef7b2873e01a0b0d","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-opus-4-5-20251101","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.2","source_model_version":"claude-opus-4-5-20251101","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"80.681818%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.023909479653766753","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"mEkZaYpiurQ6TEMAALK3su"},"evidence_class":"direct_strict","model_configuration":"claude-opus-4-5-20251101","organization":"Anthropic","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_323b108018e7aba020f9e7afa951cf32e57c3e77d59f0942c05a48d002299c99","canonical_evidence_variant_id":"gpqa_evd_v1_bcfd0eede3b0cddcd79ced89e10a322dfd89a5df91f9b2c3cfc0e6ad6197a9b5","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_323b108018e7aba020f9e7afa951cf32e57c3e77d59f0942c05a48d002299c99","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_bcfd0eede3b0cddcd79ced89e10a322dfd89a5df91f9b2c3cfc0e6ad6197a9b5"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-05-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_9d9fcd13faf4c34aa3dfac137ba313739953d4e6cb6b34f7004fe8f8a6bc0016","configuration_identity":{"context_policy":"16k","effort":"unspecified","model_snapshot":"anthropic/claude-opus-4-20250514","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-opus-4-20250514_16K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"76.262626%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03031371053819892","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"CtxFusPxCraL3EeRJu2c5k"},"evidence_class":"direct_compatibility","model_configuration":"claude-opus-4-20250514_16K","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_fedafd74cf86b8ed517d88a3f649bef8521dcb14017cc9efd5531eb98a06b277","canonical_evidence_variant_id":"gpqa_evd_v1_7935ec345c8c47d45a8fc1278e7c2b591bfbc3ef956b43da9fc62ebd321319fb","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_fedafd74cf86b8ed517d88a3f649bef8521dcb14017cc9efd5531eb98a06b277","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_7935ec345c8c47d45a8fc1278e7c2b591bfbc3ef956b43da9fc62ebd321319fb"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-08-05","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_9e05b4b262c02d63871fd36e5097241f13ddb790d81aed45c3cdaa5f0066fe16","configuration_identity":{"context_policy":"27k","effort":"unspecified","model_snapshot":"anthropic/claude-opus-4-1-20250805","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-opus-4-1-20250805_27K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"76.767677%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.030088629490217445","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"3obcsGCfuwYeEm9wRyqzPg"},"evidence_class":"direct_compatibility","model_configuration":"claude-opus-4-1-20250805_27K","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_f43b71725dbadfc796496d181aa4ea130e427290e208c538c7f33456aaff2d61","canonical_evidence_variant_id":"gpqa_evd_v1_e96691995d2919032340f96d1f25fa80b649eebb0736dbf4e4d03b3b6589517c","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_f43b71725dbadfc796496d181aa4ea130e427290e208c538c7f33456aaff2d61","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_e96691995d2919032340f96d1f25fa80b649eebb0736dbf4e4d03b3b6589517c"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_9ea8f6cc0477f44d90b6a00049e5e0d5848d4514109e686a2ebe2c8c74ade197","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"mistral/open-mixtral-8x7b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"mistral","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"open-mixtral-8x7b","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"29.829545%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.019146998037923697","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Y6KPSujFepdwqinKLf9mDF"},"evidence_class":"direct_compatibility","model_configuration":"open-mixtral-8x7b","organization":"Mistral AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_d0173bf897b84effeca3bbf845afe3773e13a3ebe3a577039d76cf9a084a67bd","canonical_evidence_variant_id":"gpqa_evd_v1_672bab37c4befe1323a7c63e129290e70301302a02097a114b31eefe261601aa","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_d0173bf897b84effeca3bbf845afe3773e13a3ebe3a577039d76cf9a084a67bd","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_672bab37c4befe1323a7c63e129290e70301302a02097a114b31eefe261601aa"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-16","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_9eb89da0bad3d09e7ddb8a304b1163ece8e34a5dc839f463fab250c970819ca6","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"openai/o3-2025-04-16","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"o3-2025-04-16_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"81.818182%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.021267123846387202","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"evANZQ9oQTdGYbDERw7VrD"},"evidence_class":"direct_compatibility","model_configuration":"o3-2025-04-16_high","organization":"OpenAI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_c4c9571f3a66edb9d81d78b50b0833b458ea5c0cff5ffdd1fa5ecadbcf484b3b","canonical_evidence_variant_id":"gpqa_evd_v1_4e869a342ecda24ebbf73895bbe205af9ad2a2ddd416b459ca09df248261173b","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_c4c9571f3a66edb9d81d78b50b0833b458ea5c0cff5ffdd1fa5ecadbcf484b3b","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_4e869a342ecda24ebbf73895bbe205af9ad2a2ddd416b459ca09df248261173b"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_9fdc200cc70fb609af6cd2ea5c6ec5818e339138000d351fc142b07d30c01378","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"epoch/claude-fable-5","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-fable-5_max","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"85.858586%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024825909793343353","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"S5pm7ctk9zgPfbyHrbD9AV"},"evidence_class":"direct_strict","model_configuration":"claude-fable-5_max","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_505ce6fbf3263877df0b03176c3eb6e48ab75c5465bd8e93f26b1d3c1492602c","canonical_evidence_variant_id":"gpqa_evd_v1_0d87f39869e0bcbe38244e446fc4e00bcd437aa68bae612afa74825b598e395c","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_505ce6fbf3263877df0b03176c3eb6e48ab75c5465bd8e93f26b1d3c1492602c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_0d87f39869e0bcbe38244e446fc4e00bcd437aa68bae612afa74825b598e395c"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_a0a4273f3e80a8760824125f9fe0ff0835270ab6ca17ddf3d6af4f4400e5ecd9","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/qwen3.5-397b-a17b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.5-397b-a17b","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"85.858586%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024825909793343353","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"gZMFRbVKnBmcnZU4YqqKMN"},"evidence_class":"direct_strict","model_configuration":"qwen3.5-397b-a17b","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_a712406e21d87161483d38e2436f8659ea8517a92ddd1b5e0d62cd6c7041350f","canonical_evidence_variant_id":"gpqa_evd_v1_00f055aabb5434ab877a0514d4d53738943a2425d8671c732d841e26ebe63241","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_a712406e21d87161483d38e2436f8659ea8517a92ddd1b5e0d62cd6c7041350f","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_00f055aabb5434ab877a0514d4d53738943a2425d8671c732d841e26ebe63241"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_a0e72767f9478bbe4b04b9a3bea317ca374e27b8e4bbf00393cac2b87d1a2775","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"epoch/gpt-oss-20b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-oss-20b_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"45.959596%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03550702465131341","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"hYATFPeUnMq7Std2nLUVAJ"},"evidence_class":"direct_strict","model_configuration":"gpt-oss-20b_high","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_53d27296521152fec647a546e2bae59f8d0b62d0db3fc9dfe258ea0fb3a9a762","canonical_evidence_variant_id":"gpqa_evd_v1_ed610c61828b46c64fc27e91d164993b3f17894dbe48858e76abf3048a53c522","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_53d27296521152fec647a546e2bae59f8d0b62d0db3fc9dfe258ea0fb3a9a762","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_ed610c61828b46c64fc27e91d164993b3f17894dbe48858e76abf3048a53c522"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-17","configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","configuration_id":"gpqa_cfg_v1_a30430189e9844037ecd5dd092ea5d76f41c830802306d85f70843d9e087f3cd","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"deepseek/deepseek-v4-pro","output_policy":"no_output_ceiling","prompt_contract":"gpqa_native_answer_choice_v1","protocol":"mean@1","provider_route":"deepseek","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","generation_config_sha256":"4c15636f00f8fc64ee7cd1c33970689264e28de9e36a6c725a08a6724f44ce40","profile":"deepseek-v4-pro-max","task_sha256":"2501e59954a9e53479bc5e44bf9923db905866019f16c385ea3f27483b3697d0"}},"details":{"configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","evidence_label":"Independent native run","execution":{"automatic_retries":0,"fresh_calls":true,"output_ceiling":"none"},"reconciliation":{"paired_original_role":"Reconstructed from the same accepted all-198 native attempt universe as the Clean score.","published_source_score_percent":null,"status":"not_applicable_independent_run"},"recovery":{"kind":"none"},"score_source":"Independent native answer-safe projection","source_kind_label":"Independent project-native run","source_run_id":"gpqa-native-deepseek-v4-pro-max-all198-20260816-r3"},"evidence_class":"project_native","model_configuration":"DeepSeek V4 Pro (max)","organization":"DeepSeek","protocol":"mean@1","validation_profile":"strict_all_normal"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_b042b219137e076e93f11b0f62773434bba47e2808e866a349483859c342625a","canonical_evidence_variant_id":"gpqa_evd_v1_8a50e6711db4f0b886818807f15f02136ae878f80e5d68f6e52c0a25a9c66392","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_b042b219137e076e93f11b0f62773434bba47e2808e866a349483859c342625a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_8a50e6711db4f0b886818807f15f02136ae878f80e5d68f6e52c0a25a9c66392"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_a385ae2ba85bd7d7184db8aaf5c23614c6f6b5d4da48acbb3e476fa738e7f196","configuration_identity":{"context_policy":"source_default","effort":"minimal","model_snapshot":"epoch/gemini-3.1-flash-lite","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemini-3.1-flash-lite_minimal","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"73.737374%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.031353050095330834","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"CmHy8cVECGkvQTwakfjRo2"},"evidence_class":"direct_strict","model_configuration":"gemini-3.1-flash-lite_minimal","organization":"Google","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_4a3ce3ae867ebf1571567fbd514c98dabb03030370b51bf2cda912beee66f06d","canonical_evidence_variant_id":"gpqa_evd_v1_fbdcb2eda572d6c2000e7447375499cee1edb211e4db200e6c985e09f29cab8d","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_4a3ce3ae867ebf1571567fbd514c98dabb03030370b51bf2cda912beee66f06d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_fbdcb2eda572d6c2000e7447375499cee1edb211e4db200e6c985e09f29cab8d"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_a4204d1bf0943d754d0da9cb0991f909d6a74aecfad020560a30706bf493ea5d","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/claude-opus-5","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-opus-5","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"92.929293%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.018263105420199485","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"mLmUnBuX6DYKtb6x46GteM"},"evidence_class":"direct_strict","model_configuration":"claude-opus-5","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_cc2237da439b936769ccc5f28f02a2e51f625ff41cabd8e9752ff05e864a806c","canonical_evidence_variant_id":"gpqa_evd_v1_046386ed49e6001d37de1c91c358d44e7aa2d07114feaf85a7344d39d3391e62","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_cc2237da439b936769ccc5f28f02a2e51f625ff41cabd8e9752ff05e864a806c","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_046386ed49e6001d37de1c91c358d44e7aa2d07114feaf85a7344d39d3391e62"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_a655fc2ed8e0041b02a2ea22c580e50c6da8f701a9667563da361120fddcdab2","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-1.5-flash-002","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemini-1.5-flash-002","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"47.316919%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027737408277040958","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"PCLvrs4TEBtqQqcYeU8pAG"},"evidence_class":"direct_compatibility","model_configuration":"gemini-1.5-flash-002","organization":"Google DeepMind","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_e0c490303a55d69319dec22bff959acb555b73644e6b2c388e9043bace59ab6d","canonical_evidence_variant_id":"gpqa_evd_v1_9a537c87ec56513fbcdf9abf92d98029fa6ae035190d848d8e01ced5f6603ced","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e0c490303a55d69319dec22bff959acb555b73644e6b2c388e9043bace59ab6d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_9a537c87ec56513fbcdf9abf92d98029fa6ae035190d848d8e01ced5f6603ced"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_a7d0e1353fd8be2dc557c8c94f0706588043b35fd3151e4c1a3ca31801ecb166","configuration_identity":{"context_policy":"source_default","effort":"minimal","model_snapshot":"epoch/gemini-3.6-flash","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemini-3.6-flash_minimal","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"85.858586%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024825909793343353","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"aPmR8fV49N98a6NVKpASAf"},"evidence_class":"direct_strict","model_configuration":"gemini-3.6-flash_minimal","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_623be4cb923c894d0ba39c8b55b80262820038b25fbf64edc1e79d9b68d4eb4e","canonical_evidence_variant_id":"gpqa_evd_v1_af128ee2d2ac65a058ec23e7de20431228f2a59971664236ea6299d00df1d105","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_623be4cb923c894d0ba39c8b55b80262820038b25fbf64edc1e79d9b68d4eb4e","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_af128ee2d2ac65a058ec23e7de20431228f2a59971664236ea6299d00df1d105"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_a8f7b12d9b5ebb9fdec3e12f1320eef478aff93bb92584ed2185df0ebaf28be5","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"alibaba/qwen-turbo-2024-11-01","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"alibaba","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"qwen-turbo-2024-11-01","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"41.792929%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02617727508870251","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"9Ke5At9bYuf56SmKJJ7Sz7"},"evidence_class":"direct_compatibility","model_configuration":"qwen-turbo-2024-11-01","organization":"Alibaba","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_e2b70488b4085155fa2bcf52be361866ebc99758c5289dd751324b11e04b97af","canonical_evidence_variant_id":"gpqa_evd_v1_5bca14b62dec3a5a55fd9c92b9e4aa5ca99ada14db4bd88a6918d8b55f1003c2","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e2b70488b4085155fa2bcf52be361866ebc99758c5289dd751324b11e04b97af","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_5bca14b62dec3a5a55fd9c92b9e4aa5ca99ada14db4bd88a6918d8b55f1003c2"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-31","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_a95dfdcebc4badfda7cb23fb421a952fccefaeb6b649462fb1643f6e2ab604fc","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/microsoft/phi-4","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"phi-4","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"56.060606%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02590274664132679","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"RAwuPRwJedy7P2KeQhEmr8"},"evidence_class":"direct_compatibility","model_configuration":"phi-4","organization":"Microsoft Research","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_1d8842cd707e4be369af820729ad0b74fff0b1ffbd484d9486044b330154ca40","canonical_evidence_variant_id":"gpqa_evd_v1_2367b1c3f49024155d12edadedb0d216eb5c09be7d49d6a5f5e2b971ac96d54d","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_1d8842cd707e4be369af820729ad0b74fff0b1ffbd484d9486044b330154ca40","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_2367b1c3f49024155d12edadedb0d216eb5c09be7d49d6a5f5e2b971ac96d54d"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-15","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_a9fa8925e31e2718da34308ac52093f3a67b1df4c9d64b4dd42be93e3f710655","configuration_identity":{"context_policy":"source_default","effort":"minimal","model_snapshot":"epoch/gemini-3.5-flash","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemini-3.5-flash_minimal","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"86.363636%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02445015597318985","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"QDZRoAVaEDtdS6nECEeE35"},"evidence_class":"direct_strict","model_configuration":"gemini-3.5-flash_minimal","organization":"Google","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_10801929840f40ed71000744d2c8be3015711ba3e564ea85b8dcd2e6cc8698d6","canonical_evidence_variant_id":"gpqa_evd_v1_0b5102859a1d1695bf2db4b37b7c23c96ea393213bfd8639459f5bd44de2d768","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_10801929840f40ed71000744d2c8be3015711ba3e564ea85b8dcd2e6cc8698d6","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_0b5102859a1d1695bf2db4b37b7c23c96ea393213bfd8639459f5bd44de2d768"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-12-11","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_aa564e65d8869f7ebd98b3ce39ff8e2e4e55cefa67b9e71802509a553137b9b3","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"openai/gpt-5.2-2025-12-11","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.3","source_model_version":"gpt-5.2-2025-12-11_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"82.702020%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02265112408975899","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"oD4Jpoi75KWChqTPDaN9aB"},"evidence_class":"direct_strict","model_configuration":"gpt-5.2-2025-12-11_low","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_a74a3808884cde5a67b2377868f7515759fac615213833cf01c5285fea943c45","canonical_evidence_variant_id":"gpqa_evd_v1_3ad649a9917c67d4ce049eecbed02a91dda453e5810ad37aa65605cac50fcbb9","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_a74a3808884cde5a67b2377868f7515759fac615213833cf01c5285fea943c45","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_3ad649a9917c67d4ce049eecbed02a91dda453e5810ad37aa65605cac50fcbb9"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-01","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_af55e40ea3129f1d80b5438a797e0c4771653bc22e47f0d34ca14dc79968ea23","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"alibaba/qwen-max-2025-01-25","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"alibaba","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"qwen-max-2025-01-25","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"56.123737%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027858831475709448","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"GBf7rqHkGGzfX87V37F9mH"},"evidence_class":"direct_compatibility","model_configuration":"qwen-max-2025-01-25","organization":"Alibaba","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_791730173c974e57d98c0f611b7ce7a170f14ebdf7b1b83475cf29a12763b566","canonical_evidence_variant_id":"gpqa_evd_v1_d79c274235ba29911b048b6b6fb91e533d94958108f9da79557a211dfcf16887","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_791730173c974e57d98c0f611b7ce7a170f14ebdf7b1b83475cf29a12763b566","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_d79c274235ba29911b048b6b6fb91e533d94958108f9da79557a211dfcf16887"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_af5c54e09d41d150ba0a812d4c0d2dd9ded66334836f64da4a4bb102999428cd","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/gpt-5.6-luna","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.6-luna_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"63.636364%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03427308652999934","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"bDKKx7zTdW5WwMCfwqsLi4"},"evidence_class":"direct_strict","model_configuration":"gpt-5.6-luna_none","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_2cca30309b582a4e3555ef0b255f847265203ef08dcbace764efec23e3893f55","canonical_evidence_variant_id":"gpqa_evd_v1_2c05fd618e478c4fdbab86926c2327f20e857a8e3fb36b7ce506264f05559539","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_2cca30309b582a4e3555ef0b255f847265203ef08dcbace764efec23e3893f55","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_2c05fd618e478c4fdbab86926c2327f20e857a8e3fb36b7ce506264f05559539"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-13","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_afeeefbcbb8acb87e92f6e3238d36a6aaacc8dc9149121f1521a3cb957fcba15","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/gpt-5.2-2025-12-11","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.2-2025-12-11_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"73.232323%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.031544498882702825","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"4sPs53wmM9bvN8JXzsiNCk"},"evidence_class":"direct_strict","model_configuration":"gpt-5.2-2025-12-11_none","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_d5884032b0fe6618caf332eeae77e09d3029497ccb09e7d6f3d330d10b5b02ee","canonical_evidence_variant_id":"gpqa_evd_v1_3ee2714591cbebb53903dbb1325cf6f31790b046542c5d8ba8e6e8e9465e5e00","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_d5884032b0fe6618caf332eeae77e09d3029497ccb09e7d6f3d330d10b5b02ee","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_3ee2714591cbebb53903dbb1325cf6f31790b046542c5d8ba8e6e8e9465e5e00"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-10-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_b25fbe60dede1fb3425e3063a2463d869b9c4c00d4b6b0aed6903c5818c3f024","configuration_identity":{"context_policy":"32k","effort":"unspecified","model_snapshot":"anthropic/claude-haiku-4-5-20251001","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-haiku-4-5-20251001_32K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"71.212121%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03225883512300997","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"GEaAdb9UmoUFbEtBx7WzAR"},"evidence_class":"direct_compatibility","model_configuration":"claude-haiku-4-5-20251001_32K","organization":"Anthropic","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_76be72ecb2d1012730f713f0346f63f3c49fd02d87119f948d4de67172a1ea2a","canonical_evidence_variant_id":"gpqa_evd_v1_e8faa7b7198a65cad9ded7ad71a744e652bb936bfd5b100a03bf8c7a13b9558f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_76be72ecb2d1012730f713f0346f63f3c49fd02d87119f948d4de67172a1ea2a","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_e8faa7b7198a65cad9ded7ad71a744e652bb936bfd5b100a03bf8c7a13b9558f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_b36681f056f84dece01834a86b30fd4e56b779b993f6548b69eb449624c0d284","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"hyperbolic/Qwen/Qwen2.5-72B-Instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"hyperbolic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"qwen2.5-72b-instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"49.147727%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026947767404178667","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"WtALgW6VSFdihDZWXRm2AZ"},"evidence_class":"direct_compatibility","model_configuration":"qwen2.5-72b-instruct","organization":"Alibaba","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_18ede0f30376d27589270d01b893d1a9e6ab9daa2988644267e40983cfaacd54","canonical_evidence_variant_id":"gpqa_evd_v1_8e628442c92f56910bae1767681a204217285335eb2d8ed4366f7857d08a5496","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_18ede0f30376d27589270d01b893d1a9e6ab9daa2988644267e40983cfaacd54","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_8e628442c92f56910bae1767681a204217285335eb2d8ed4366f7857d08a5496"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_b469234b3adc87ce38b386bcaf40c26f1aa236300b162eae3f193b325d11ed95","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"fireworks/accounts/fireworks/models/llama-v3p2-90b-vision-instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"fireworks","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Llama-3.2-90B-Vision-Instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"41.035354%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.022948341091025694","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"9SSHEP95FYVZvHDdoHSRy3"},"evidence_class":"direct_compatibility","model_configuration":"Llama-3.2-90B-Vision-Instruct","organization":"Meta AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_fe427bef732f5f1a516b2590fa00be675554c2e40a20fa0ead1cc7602785c293","canonical_evidence_variant_id":"gpqa_evd_v1_56b4fdcabe7d5a6922b99b83db483de7ec02e83ebc8cffacfb60a3b0382b154e","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_fe427bef732f5f1a516b2590fa00be675554c2e40a20fa0ead1cc7602785c293","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_56b4fdcabe7d5a6922b99b83db483de7ec02e83ebc8cffacfb60a3b0382b154e"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-20","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_b6ca95c2ac417673b46a71c78ad3dd16d60626c176c639d56b530e84b8805233","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"hyperbolic/meta-llama/Meta-Llama-3.1-8B-Instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"hyperbolic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Llama-3.1-8B-Instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"25.946970%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.01674966399258744","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"WWS3aotkrymLmBBDhhtvWQ"},"evidence_class":"direct_compatibility","model_configuration":"Llama-3.1-8B-Instruct","organization":"Meta AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_9e118dbe167fd2b665d992d648605798ee009bec98127fdb1c034f3eca1a59f5","canonical_evidence_variant_id":"gpqa_evd_v1_a132b7a5bd8da5abbe730a453743f03692e12b95b6a03c9a398fe10f2073501f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_9e118dbe167fd2b665d992d648605798ee009bec98127fdb1c034f3eca1a59f5","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_a132b7a5bd8da5abbe730a453743f03692e12b95b6a03c9a398fe10f2073501f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-06-03","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_b6e973f9d6f6139a95d79c29885fb8d73070eadc9565d29c3978bdf06333826d","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"alibaba/qwen3-235b-a22b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"alibaba","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"qwen3-235b-a22b","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"70.707071%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027124209393506977","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"agebmxGseTvmvN4AHN8i4m"},"evidence_class":"direct_compatibility","model_configuration":"qwen3-235b-a22b","organization":"Alibaba","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":7,"canonical_evaluation_run_id":"gpqa_run_v1_3ced4fbc7a914e379f2ddb1f5d74538c090671f9fd3ed597cc9004240a7b4384","canonical_evidence_variant_id":"gpqa_evd_v1_6de8ed01eeaadd5ec84d2baa19feac6cc625ae736e7979d3469c085b7137091c","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_3ced4fbc7a914e379f2ddb1f5d74538c090671f9fd3ed597cc9004240a7b4384","evidence_class":"derived_uniform","evidence_variant_id":"gpqa_evd_v1_6de8ed01eeaadd5ec84d2baa19feac6cc625ae736e7979d3469c085b7137091c"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-12-16","configuration_date_basis":"validated Epoch CSV Started at metadata","configuration_id":"gpqa_cfg_v1_b79f40ee05c16a27a3fae7d506fb2e654418d9d40244460a93ad4d4fc6b45468","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/deepseek-reasoner","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_source_task_v1","protocol":"mean@7","provider_route":"epoch_import","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"7","declared_attempts_per_question":"8","excluded_epoch_numbers":"1","source_model_version":"deepseek-reasoner"}},"details":{"configuration_date_basis":"validated Epoch CSV Started at metadata","evidence_label":"Derived uniform-epoch subset","reconciliation":{"paired_original_role":"Reconstructed from the selected uniform subset; not the native Epoch aggregate.","published_source_score_percent":"83.423521%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.0201537212608744","status":"exact"},"recovery":{"declared_attempts_per_question":8,"kind":"uniform_epoch_subset","omitted_attempt_count":1,"selected_attempts_per_question":7},"score_source":"Accepted binary choice scores from the complete uniform attempt subset","source_kind_label":"Epoch source with globally uniform attempt subset","source_run_id":"CDdjbqiYYPtEBGoZLMPM3n"},"evidence_class":"derived_uniform","model_configuration":"deepseek-reasoner","organization":"DeepSeek","protocol":"mean@7","validation_profile":"derived_uniform_epoch_subset"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_1c4918f08653f9b887f57bafc02a28a088ec4633d22db99727ad4d3740d00e21","canonical_evidence_variant_id":"gpqa_evd_v1_5a1006a45145e6d1f13ee773ce37108bd2b070ba59b8c252e336a15f9987b3a5","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_1c4918f08653f9b887f57bafc02a28a088ec4633d22db99727ad4d3740d00e21","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_5a1006a45145e6d1f13ee773ce37108bd2b070ba59b8c252e336a15f9987b3a5"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-21","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_b8507ecd7db462de6e723c0dd6cefa055697adc5afe4eaa772a1d3f43c291371","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"together/microsoft/WizardLM-2-8x22B","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"together","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"WizardLM-2-8x22B","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"43.434343%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.023232093333444247","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"MQJA3RUjsQzSubGJiukZ7A"},"evidence_class":"direct_compatibility","model_configuration":"WizardLM-2-8x22B","organization":"Microsoft","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_9d512d9f0cf52e51fefcc86a1269f13b4e27fffc906225428130f2ff13d7a1ec","canonical_evidence_variant_id":"gpqa_evd_v1_2ab2a1cf44a4b3cb002d148f563c852e367cd00c47aa6fed450290047551b0b1","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_9d512d9f0cf52e51fefcc86a1269f13b4e27fffc906225428130f2ff13d7a1ec","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_2ab2a1cf44a4b3cb002d148f563c852e367cd00c47aa6fed450290047551b0b1"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-16","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_b8ff63f0f960b4e2f35e0a38f75a361a14b4ce8d403568ee967f98dc1ff60ed7","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"openai/o4-mini-2025-04-16","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"o4-mini-2025-04-16_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"79.608586%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024034764096121958","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"9Sgsz5XCjscuZkexf7MX9c"},"evidence_class":"direct_compatibility","model_configuration":"o4-mini-2025-04-16_high","organization":"OpenAI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_ff9ed21ccc61d209594a1e016635d3088037d63ddac362a0e43bef48729d5892","canonical_evidence_variant_id":"gpqa_evd_v1_759f3f6f4b5a2cf6f8c0f0854f626fd30c94a331e3abf1b238813400e9af6b72","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_ff9ed21ccc61d209594a1e016635d3088037d63ddac362a0e43bef48729d5892","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_759f3f6f4b5a2cf6f8c0f0854f626fd30c94a331e3abf1b238813400e9af6b72"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_bb7b9a95d5cff8aa62bb04540cc6a3c48427cc5b28e23bde5a188484ba9cda5d","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-2.0","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-2.0","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"34.659091%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.025076255983551475","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"mTa4f52zsEuxdrPy7WfQvX"},"evidence_class":"direct_compatibility","model_configuration":"claude-2.0","organization":"Anthropic","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_ad363578f65b2e9b208a3f95847b49869248d41bd9700e746dfd783640e4ac7c","canonical_evidence_variant_id":"gpqa_evd_v1_202f68076109b3a6baa05665805091c8c81bcdae256914acbf9aab584d424f67","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_ad363578f65b2e9b208a3f95847b49869248d41bd9700e746dfd783640e4ac7c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_202f68076109b3a6baa05665805091c8c81bcdae256914acbf9aab584d424f67"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_bc3e18e0e130722c5f1e133ef0ec99c49ebc62c0cf1d9a48f4491777870f83b6","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"epoch/qwen3.7-max","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.7-max","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"90.909091%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.020482086775424225","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"W6mTanyjvXhasFBzHukv4S"},"evidence_class":"direct_strict","model_configuration":"qwen3.7-max","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_5f85db9f6e9ea1d09922a756afb331f3156ee27cd024227e1bfa68f46e145174","canonical_evidence_variant_id":"gpqa_evd_v1_98d4be1393e2c39c14ec82adf828f8420be9fc3d887c76cb23379312fa4b18a2","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_5f85db9f6e9ea1d09922a756afb331f3156ee27cd024227e1bfa68f46e145174","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_98d4be1393e2c39c14ec82adf828f8420be9fc3d887c76cb23379312fa4b18a2"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-20","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_bcfd3a2eaf724a8a23244b602b5694e0a00e463c8d82eb1c2c4ab4333abe3a1e","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/Qwen/Qwen1.5-72B-Chat","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"qwen1.5-72b-chat","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"28.819444%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.01993691228122493","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"NeqeQSxHmdUNLk9Ytpo6qo"},"evidence_class":"direct_compatibility","model_configuration":"qwen1.5-72b-chat","organization":"Alibaba","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_556475bf264b636c3d54229a7e82a9bba4f3848974f830dff8ccdb7a312b4647","canonical_evidence_variant_id":"gpqa_evd_v1_946bc4c5c62e8e724c128e71783e863c64910c7f628307b210ad614581401316","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_556475bf264b636c3d54229a7e82a9bba4f3848974f830dff8ccdb7a312b4647","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_946bc4c5c62e8e724c128e71783e863c64910c7f628307b210ad614581401316"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_bd2829d168e72420d4610fdbbbcdd955033646500d77a719befb520b54bcc796","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4-1106-preview","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4-1106-preview","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"42.361111%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024076142618301676","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"AWewipyfSQb8hXtS8d4fNL"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4-1106-preview","organization":"OpenAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_441ffd9ecb95e12894b5ccd9cf32b7c3bb475f7721657ba3739f78bd1e1b3a9c","canonical_evidence_variant_id":"gpqa_evd_v1_2294e88cba5d4836f3b78fa9fb13e3fd9138c00ce5d2e64a8f0b9f270bb58069","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_441ffd9ecb95e12894b5ccd9cf32b7c3bb475f7721657ba3739f78bd1e1b3a9c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_2294e88cba5d4836f3b78fa9fb13e3fd9138c00ce5d2e64a8f0b9f270bb58069"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_bdda016ac92da795727c89d7eb197d32b9654ce78d9d202bf7a7aae394898d1b","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/gpt-5.5-2026-04-23","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.5_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"77.272727%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.029857515673386438","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"KesKa5EXtFG7H4UMBFzgCh"},"evidence_class":"direct_strict","model_configuration":"gpt-5.5_none","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_ee0798a79b64398cd44f82da072f043234162e67c9f3cf60c3121943b3cfda24","canonical_evidence_variant_id":"gpqa_evd_v1_c2c9632865b7bbf8e73f7a9a723e1db9e87ac3c12347c97b537b653d6412bc9a","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_ee0798a79b64398cd44f82da072f043234162e67c9f3cf60c3121943b3cfda24","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_c2c9632865b7bbf8e73f7a9a723e1db9e87ac3c12347c97b537b653d6412bc9a"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_be00ebd659f5798739f50eb4ddb03eec87164876f7bc5f04ab7cadb78d6ba39c","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/magistral-small-2506","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"magistral-small-2506","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"56.060606%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.0353608594752948","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"dwGLob8Vhrwx4Wn376jVmt"},"evidence_class":"direct_strict","model_configuration":"magistral-small-2506","organization":"Mistral AI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_223fdbc5d16a5cb48e4f360d845c8ae5407a47374e8f7513e71d9a1ce2ac0836","canonical_evidence_variant_id":"gpqa_evd_v1_26554676aa4c49359991cc9f9b408328e46c02a13be963472c8e02f194af8d26","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_223fdbc5d16a5cb48e4f360d845c8ae5407a47374e8f7513e71d9a1ce2ac0836","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_26554676aa4c49359991cc9f9b408328e46c02a13be963472c8e02f194af8d26"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-17","configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","configuration_id":"gpqa_cfg_v1_bfdb9cba805be9f524447f939b0d26c6c2a84b2eb93aa007c9f5c6d024b1fd8a","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"gemini-flex/gemini-3.6-flash","output_policy":"no_output_ceiling","prompt_contract":"gpqa_native_answer_choice_v1","protocol":"mean@1","provider_route":"gemini-flex","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","generation_config_sha256":"4db33ee3aca837b8beb02b9e0714860bd517fe9ef65c086a876468bb7b9f3928","profile":"gemini-3-6-flash-high","task_sha256":"2501e59954a9e53479bc5e44bf9923db905866019f16c385ea3f27483b3697d0"}},"details":{"configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","evidence_label":"Independent native run","execution":{"automatic_retries":0,"fresh_calls":true,"output_ceiling":"none"},"reconciliation":{"paired_original_role":"Reconstructed from the same accepted all-198 native attempt universe as the Clean score.","published_source_score_percent":null,"status":"not_applicable_independent_run"},"recovery":{"kind":"exact_source_plus_transport_recovery","output_ceiling":"none","recovered_source_record_count":140,"source_explicit_failure_record_count":9,"source_missing_record_count":131},"score_source":"Independent native answer-safe projection","source_kind_label":"Independent project-native run","source_run_id":"gpqa-native-gemini-3-6-flash-high-all198-20260816-r3"},"evidence_class":"project_native","model_configuration":"Gemini 3.6 Flash (high)","organization":"Google","protocol":"mean@1","validation_profile":"complete_with_transport_recovery"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_0f66c1cf64f2d84e13fca098cb05d8b0757e36a0c3961a71d6fe2ac91425bee6","canonical_evidence_variant_id":"gpqa_evd_v1_3dbd547caea1d744783d04ac307a3e69692aefe4377b0878abe297c58c1cb12f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_0f66c1cf64f2d84e13fca098cb05d8b0757e36a0c3961a71d6fe2ac91425bee6","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_3dbd547caea1d744783d04ac307a3e69692aefe4377b0878abe297c58c1cb12f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_bfdeb68407b430f0101f3ac3ac9ff34e4e1bcbc7034bb257d6a50df851148d7e","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/deepseek-v4-pro","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"deepseek-v4-pro_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"73.232323%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.031544498882702825","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"FtGueFbnFnKBAuLmEEufFg"},"evidence_class":"direct_strict","model_configuration":"deepseek-v4-pro_none","organization":"DeepSeek","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_9a3d010c3218fc3c7862a01c0ba72a02508f075c4f5bf3d85d2f71765f4cdcce","canonical_evidence_variant_id":"gpqa_evd_v1_49152866a50817f6fe994c5543da3fb80dd1e8349a147f155d18b9c761ac27e7","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_9a3d010c3218fc3c7862a01c0ba72a02508f075c4f5bf3d85d2f71765f4cdcce","evidence_class":"targeted_recovery","evidence_variant_id":"gpqa_evd_v1_49152866a50817f6fe994c5543da3fb80dd1e8349a147f155d18b9c761ac27e7"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-16","configuration_date_basis":"validated composite generated_at metadata","configuration_id":"gpqa_cfg_v1_c0291f9a5fc3b09556c001225b7f3d6b3b1de8db4895ffa3e136dbc321bb3809","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-3.1-pro-preview","output_policy":"source_artifact_policy_plus_uncapped_recovery","prompt_contract":"epoch_gpqa_diamond_source_plus_recovery_v1","protocol":"mean@8","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","configuration_match":"exact_source_header_generate_config","execution_match":"exact_source_route","model_identity_basis":"exact_historic_provider_route","source_model_version":"gemini-3.1-pro-preview"}},"details":{"configuration_date_basis":"validated composite generated_at metadata","evidence_label":"Epoch base plus targeted transport recovery","reconciliation":{"difference_percentage_points":"+0.343437 pp","paired_original_role":"Reconstructed from the completed source-plus-recovery attempt universe; not relabelled as the original Epoch aggregate.","published_source_score_percent":"94.100000%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.017","status":"material_conflict_explicit_inspect_source_of_truth"},"recovery":{"configuration_match":"exact_source_header_generate_config","execution_match":"exact_source_route","kind":"exact_targeted_transport_recovery","model_identity_basis":"exact_historic_provider_route","output_ceiling":"none","recovered_pair_count":44},"score_source":"Accepted source choice scores plus exact targeted repair pairs","source_kind_label":"Epoch base plus targeted transport-recovery composite","source_run_id":"kR9F3oHBCnwQW4AiTPKwpd"},"evidence_class":"targeted_recovery","model_configuration":"gemini-3.1-pro-preview","organization":"Google DeepMind","protocol":"mean@8","validation_profile":"complete_with_epoch_targeted_transport_recovery"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_4d4f8e3135650879f10070cd2680e14f6a49e663a66f52da7d43541d4a221050","canonical_evidence_variant_id":"gpqa_evd_v1_71700c48dc2e1abbfec61082f8ae34ba9d3037412e9b9ae8348008be452dbf27","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_4d4f8e3135650879f10070cd2680e14f6a49e663a66f52da7d43541d4a221050","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_71700c48dc2e1abbfec61082f8ae34ba9d3037412e9b9ae8348008be452dbf27"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-21","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_c030a6e64a8b74f6e73831a3aae768dc618173ab7994e9387b5f3a60cd882de0","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-1.0-pro-001","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemini-1.0-pro-001","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"33.964646%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.020110933405176737","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"ibpS8bs4z3wD3Hb6F3kpdV"},"evidence_class":"direct_compatibility","model_configuration":"gemini-1.0-pro-001","organization":"Google DeepMind","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_7c5f5aedd5059e41d6225ac7422c2245780ae7ff7ff62dc0027bbae0edb76c98","canonical_evidence_variant_id":"gpqa_evd_v1_43fd53c4f6f37ac29010b1ec216659d5796e7d85c5a296a0cfde43c8f9e6932b","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_7c5f5aedd5059e41d6225ac7422c2245780ae7ff7ff62dc0027bbae0edb76c98","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_43fd53c4f6f37ac29010b1ec216659d5796e7d85c5a296a0cfde43c8f9e6932b"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_c09343b2de7a3b30deceaaba8fcf1b1e2930f59ccdd6ba4efcd34c200b253701","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"epoch/gemini-3-flash-preview","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemini-3-flash-preview_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"89.393939%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.021938047738853106","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"gVdraDT3pXmaExP9VbqME4"},"evidence_class":"direct_strict","model_configuration":"gemini-3-flash-preview_high","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_6893083f0ef2fba44be6d48e4ab8e77f95e774692d113149cb4b0fdc5cc57a68","canonical_evidence_variant_id":"gpqa_evd_v1_7b24b1da63ad62151bdc7d36a8f372843d4664e35829971354750ad1a70b5b94","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_6893083f0ef2fba44be6d48e4ab8e77f95e774692d113149cb4b0fdc5cc57a68","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_7b24b1da63ad62151bdc7d36a8f372843d4664e35829971354750ad1a70b5b94"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_c11ad0dd40116f32651cfd124cde890429cc3ff2cb1a8b1b529c9d58661f9850","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/qwen3.6-27b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.6-27b_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"84.848485%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02554565042660364","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"WeP7Q27N5PzxSgg4ybgijX"},"evidence_class":"direct_strict","model_configuration":"qwen3.6-27b_none","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_6dbe971b027e9b7ff9f5a2ac758872199004bc23eba046bb852d8762b12a2fbb","canonical_evidence_variant_id":"gpqa_evd_v1_b2979d2667e6d1608bb2130cf65b9073f482e281753982f55d3545e6c324a4d2","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_6dbe971b027e9b7ff9f5a2ac758872199004bc23eba046bb852d8762b12a2fbb","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_b2979d2667e6d1608bb2130cf65b9073f482e281753982f55d3545e6c324a4d2"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_c1662dd832d707580c960ccdf5da02a7a38f399834c024fd516e6015ff35eaa2","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"epoch/kimi-k3","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"kimi-k3_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"91.919192%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.019417681889724515","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"AoAD5uVtgVDnXmwA7STuno"},"evidence_class":"direct_strict","model_configuration":"kimi-k3_high","organization":"Moonshot","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_c0da6efa4e9f19cddbdea469776df805e3444c48cefd686086017bff4f5bc1af","canonical_evidence_variant_id":"gpqa_evd_v1_9c8a0e08aa9dbc17aab6da7a789f24d40afda3b4ac40259a5f8cd7bb4a47d695","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_c0da6efa4e9f19cddbdea469776df805e3444c48cefd686086017bff4f5bc1af","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_9c8a0e08aa9dbc17aab6da7a789f24d40afda3b4ac40259a5f8cd7bb4a47d695"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_c1890a410e08444aa2e81859f26860a80431a4a5028c50bc7f3514cbc2e50a82","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/qwen3.5-397b-a17b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.5-397b-a17b_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"86.363636%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02445015597318985","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"bXrmQqzF9sXbuZfRrSc7NP"},"evidence_class":"direct_strict","model_configuration":"qwen3.5-397b-a17b_none","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_c4291c1149beaaeedd87a1e987d06d3384f9ba33ddba4e80416ddc38771a396d","canonical_evidence_variant_id":"gpqa_evd_v1_8887a5d926c0f619f97426e65b45cf84d7ecc002ecc510f8a9df014e9bdd355b","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_c4291c1149beaaeedd87a1e987d06d3384f9ba33ddba4e80416ddc38771a396d","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_8887a5d926c0f619f97426e65b45cf84d7ecc002ecc510f8a9df014e9bdd355b"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-03-31","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_c254d074bf7d4d562416e11363147be58eab58afb4c70c33aee01ccb278b50a8","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-2.5-pro-exp-03-25","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemini-2.5-pro-exp-03-25","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"83.838384%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02622591986362927","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"WKjzEQdfnBHnYeZoKeCWnB"},"evidence_class":"direct_compatibility","model_configuration":"gemini-2.5-pro-exp-03-25","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_59d0bd8a33443ddf2b37cfc2c208107de7dea73e825054b67d1074f55f0a4a4e","canonical_evidence_variant_id":"gpqa_evd_v1_6b961fd55ef7315eab24f3b486b42c7cec1b605f80a795374b8fa76e31439cf5","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_59d0bd8a33443ddf2b37cfc2c208107de7dea73e825054b67d1074f55f0a4a4e","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_6b961fd55ef7315eab24f3b486b42c7cec1b605f80a795374b8fa76e31439cf5"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_c413680e8fd715c2789e6cbb466901dd7cf9f221fe5a9a6f5aab61e02c530288","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/qwen3.5-plus-2026-04-20","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.5-plus","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"84.848485%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.025545650426603637","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"mGXYSUNCMisvaRzhQp6g2L"},"evidence_class":"direct_strict","model_configuration":"qwen3.5-plus","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_88afd62f888feaa64bfd19ce1b3107b7a12ceaf768680d041bc5cbf374a06b80","canonical_evidence_variant_id":"gpqa_evd_v1_52c8fadd084b1897a73631b39ea57ce0a8ffbec358a2975f7a95906190039081","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_88afd62f888feaa64bfd19ce1b3107b7a12ceaf768680d041bc5cbf374a06b80","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_52c8fadd084b1897a73631b39ea57ce0a8ffbec358a2975f7a95906190039081"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_c486088d5d896e439d15de33eede35061d90e78929121ccb49dcd5c09ad7b1cd","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/google/gemma-2-9b-it","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemma-2-9b-it","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"27.462121%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.023032467954778795","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"fQjoHKyUEPAuwdrBGfL8SG"},"evidence_class":"direct_compatibility","model_configuration":"gemma-2-9b-it","organization":"Google DeepMind","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_46db62f2e515b497fc8911933a8451454331ed63da7531543e40384debed32d7","canonical_evidence_variant_id":"gpqa_evd_v1_31e2d1edf2d19e00df42190b15321c5b234078a1744c36f57f56815a58d311b8","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_46db62f2e515b497fc8911933a8451454331ed63da7531543e40384debed32d7","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_31e2d1edf2d19e00df42190b15321c5b234078a1744c36f57f56815a58d311b8"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-05-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_c61eab9033f8b5854c5649db1e12ecba8df925dfd4b6c0455cbe678320b35d78","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"epoch/gemini-3.5-flash","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.9","source_model_version":"gemini-3.5-flash_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"92.803030%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.016387002835275905","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"dYYtcwENFRtm9jZTDA2p9C"},"evidence_class":"direct_strict","model_configuration":"gemini-3.5-flash_high","organization":"Google","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_c2255b330f903b8fcf755fc55cefaeacf5b6dfe0959dcbe1100ec824c787166f","canonical_evidence_variant_id":"gpqa_evd_v1_0210e91c44742be09ffb25b8c97f0a9c1eb9e1209a24a2fb3176e1c23c0ba1f7","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_c2255b330f903b8fcf755fc55cefaeacf5b6dfe0959dcbe1100ec824c787166f","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_0210e91c44742be09ffb25b8c97f0a9c1eb9e1209a24a2fb3176e1c23c0ba1f7"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-20","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_c6f1b461149c22be3ff762bed74ced5bc575c8324790a49a899f7f672ee32cfa","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"hyperbolic/meta-llama/Meta-Llama-3.1-405B-Instruct","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"hyperbolic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Llama-3.1-405B-Instruct","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"50.915404%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.025861992702770648","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"S5QYXSvQBRSbUbXSnAGbMm"},"evidence_class":"direct_compatibility","model_configuration":"Llama-3.1-405B-Instruct","organization":"Meta AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_6577834900dfbcb589f50fc3582bde113ebf93aee5f2d1aa5d81d4e59789e406","canonical_evidence_variant_id":"gpqa_evd_v1_cf6c49d92bc312e635cf3c7a8c96ff6a926ca11dca564059c163bc0ab00663c7","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_6577834900dfbcb589f50fc3582bde113ebf93aee5f2d1aa5d81d4e59789e406","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_cf6c49d92bc312e635cf3c7a8c96ff6a926ca11dca564059c163bc0ab00663c7"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-02-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_c7c57d8f54855969165427be088ee517a83a6182a8dfb2e215a3d8ce25b8e0ef","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-2.0-flash-thinking-exp-01-21","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemini-2.0-flash-thinking-exp-01-21","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"57.070707%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03526552724601199","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"ag9w2w4xa47ghZaWt5oZey"},"evidence_class":"direct_compatibility","model_configuration":"gemini-2.0-flash-thinking-exp-01-21","organization":"Google DeepMind,Google","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_61b4ef5388c418bf739fe689d906ab4b0be33f09b4dba70dc8cf4912c55119d6","canonical_evidence_variant_id":"gpqa_evd_v1_b180de987161bdbf94b237ff62e9826189aa1661c91bdfe0183489a9814f261a","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_61b4ef5388c418bf739fe689d906ab4b0be33f09b4dba70dc8cf4912c55119d6","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_b180de987161bdbf94b237ff62e9826189aa1661c91bdfe0183489a9814f261a"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_ca66e0d6fa724848f897b1528b7c9822969050042ffe53bc36c3bad551d8bbbe","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/gemini-3.5-flash-lite","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemini-3.5-flash-lite_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"75.757576%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.030532892233932022","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"dgSJTFQ4UtuQKEhB9ss6K7"},"evidence_class":"direct_strict","model_configuration":"gemini-3.5-flash-lite_low","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_8f0f1f40c4ae6fb2b37d51816719e6564c712c79f04cb6c95299b61460522874","canonical_evidence_variant_id":"gpqa_evd_v1_1b5cc11fffab55b07f6d9640a719846848a6c214ad4abd1a27a0364b071a2511","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_8f0f1f40c4ae6fb2b37d51816719e6564c712c79f04cb6c95299b61460522874","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_1b5cc11fffab55b07f6d9640a719846848a6c214ad4abd1a27a0364b071a2511"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-11-24","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_cad9ce1ea155bb333c5ebd7a19fec521f9db3f4590d2bf1ec498004a3a8f0362","configuration_identity":{"context_policy":"32k","effort":"unspecified","model_snapshot":"anthropic/claude-opus-4-5-20251101","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.2","source_model_version":"claude-opus-4-5-20251101_32K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"86.047980%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.020686959671393914","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Kwwo3UFcPDk6iLAeSeeRaa"},"evidence_class":"direct_strict","model_configuration":"claude-opus-4-5-20251101_32K","organization":"Anthropic","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_81d603eb7852d0c2381cccf6581f79b07c1a4587d624c00276deac0263a5404a","canonical_evidence_variant_id":"gpqa_evd_v1_c1d607e0d4fc89e9b91829beebbcad48937243ac67b3be6691b1ad2d8b468d05","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_81d603eb7852d0c2381cccf6581f79b07c1a4587d624c00276deac0263a5404a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_c1d607e0d4fc89e9b91829beebbcad48937243ac67b3be6691b1ad2d8b468d05"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_cc64a12d6bdca099340cd29a26afe83aeffe666f68bc05604bf7f4e7668632cf","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"epoch/o4-mini-2025-04-16","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"o4-mini-2025-04-16_medium","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"77.777778%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02962022787479049","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"noAGevebNp5W5yAnmbEYaz"},"evidence_class":"direct_strict","model_configuration":"o4-mini-2025-04-16_medium","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_0d0d113c8d7eabc4945c34d294cc03b1a939c9dc430c2bc2e47073fa5399a810","canonical_evidence_variant_id":"gpqa_evd_v1_0e7c910758a90f1beaa6eb551535379224127bfd2847151f0cf00a389e52e0b2","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_0d0d113c8d7eabc4945c34d294cc03b1a939c9dc430c2bc2e47073fa5399a810","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_0e7c910758a90f1beaa6eb551535379224127bfd2847151f0cf00a389e52e0b2"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_cc760f480aef1f4d05a51dc48fb94f9bd8123b5a19ed8a8b623a9aa5936069ff","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"openai/gpt-5-nano-2025-08-07","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-5-nano-2025-08-07_medium","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"67.424242%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026607011205052013","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"ELAxkvAigvEVdMmhxiaMiy"},"evidence_class":"direct_compatibility","model_configuration":"gpt-5-nano-2025-08-07_medium","organization":"OpenAI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_eb28c28f8efad78efe685bfae2b1850b87edca5dd86b1d99cb38de1108ccfd84","canonical_evidence_variant_id":"gpqa_evd_v1_86bdece77771e95b2bd5306d06515006efbf5c26bbb6cc432651c53d48bb77fb","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_eb28c28f8efad78efe685bfae2b1850b87edca5dd86b1d99cb38de1108ccfd84","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_86bdece77771e95b2bd5306d06515006efbf5c26bbb6cc432651c53d48bb77fb"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-02-24","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_cf79ba0cd7bfae3b0d2ce505ab46e7b0b3957b3840464c52248da9bfb033e3d1","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-3-7-sonnet-20250219","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-3-7-sonnet-20250219","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"66.035354%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026703858684060923","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"PyNEkFCfMEReodeE2oiPhh"},"evidence_class":"direct_compatibility","model_configuration":"claude-3-7-sonnet-20250219","organization":"Anthropic","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_539787a00fa51afcec3ccb65ccff97d320fb4694a812942a0833e0e8eb153471","canonical_evidence_variant_id":"gpqa_evd_v1_36660bc48bc7969c439dc1c3957e2acbc6c27c7402ae011671ba60bc49221447","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_539787a00fa51afcec3ccb65ccff97d320fb4694a812942a0833e0e8eb153471","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_36660bc48bc7969c439dc1c3957e2acbc6c27c7402ae011671ba60bc49221447"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-02-20","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_cf7a0ccdcb575260739a0107fbf38e7b46265e7cca9bb7f750093cf0cbf5cc6c","configuration_identity":{"context_policy":"32k","effort":"unspecified","model_snapshot":"anthropic/claude-sonnet-4-6","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.5","source_model_version":"claude-sonnet-4-6_32K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"87.373737%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.020034282226331888","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"58eQmyCfPa3FufhXJFvAL2"},"evidence_class":"direct_strict","model_configuration":"claude-sonnet-4-6_32K","organization":"Anthropic","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_cc195c4dda12a63110db2e3a9d13dc8d93244f1708be103f58ff08786f43648f","canonical_evidence_variant_id":"gpqa_evd_v1_5ebc7f414be92b8496480dd45cf9a6128627b0f00a0a0ec5025a55245d488e65","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_cc195c4dda12a63110db2e3a9d13dc8d93244f1708be103f58ff08786f43648f","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_5ebc7f414be92b8496480dd45cf9a6128627b0f00a0a0ec5025a55245d488e65"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-21","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_cf9f9998388c8eddb101a1c6d6fdf01748173bb032842e230ac263e40899daa5","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/01-ai/Yi-1.5-34B-Chat","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Yi-1.5-34B-Chat","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"31.976010%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.020005291661900883","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Jwi8JTeeTxJuJY8qidxngw"},"evidence_class":"direct_compatibility","model_configuration":"Yi-1.5-34B-Chat","organization":"01.AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_2e9ebbffb2f4446f3d4354b423cda97a3f1b88a383fc75157db4feff80d8b54f","canonical_evidence_variant_id":"gpqa_evd_v1_7e88c3c69bcc6af2eb0a12752dc66994a940cfbff475a81429f406c322692e3f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_2e9ebbffb2f4446f3d4354b423cda97a3f1b88a383fc75157db4feff80d8b54f","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_7e88c3c69bcc6af2eb0a12752dc66994a940cfbff475a81429f406c322692e3f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-15","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_d03c3f1680526d4f61dc66ec0cfd3f39d637ecad3f65e348f985bb848fc3c8e7","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/gpt-5.4-2026-03-05","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.4-2026-03-05_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"84.848485%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.025545650426603637","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"aYgDc2iSDpmHkQfu7YjCHc"},"evidence_class":"direct_strict","model_configuration":"gpt-5.4-2026-03-05_low","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_b105d021c8954641ef6d8b150852735703a54dc4d76622afa303271ae0433417","canonical_evidence_variant_id":"gpqa_evd_v1_3ec585c7244555667609b4cc8f50007cdf75bb8200ef0a2c0cd0622ce2032a60","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_b105d021c8954641ef6d8b150852735703a54dc4d76622afa303271ae0433417","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_3ec585c7244555667609b4cc8f50007cdf75bb8200ef0a2c0cd0622ce2032a60"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_d1221f2c8514da265ead360bb375cb64430d6cd83d164f77739373eae7f37879","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-3-5-sonnet-20241022","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-3-5-sonnet-20241022","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"55.303030%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.028208090147078847","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"JCCVcVLz3J4TQXSU67hMJZ"},"evidence_class":"direct_compatibility","model_configuration":"claude-3-5-sonnet-20241022","organization":"Anthropic","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_a1d0d683a699d85732a54ec4ce6fe93f60323ce13c14f2558ff6922042da7709","canonical_evidence_variant_id":"gpqa_evd_v1_38d18632501697af4ada9c62e750c8fd8b8cc0901b5dd7425f68c740790779b1","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_a1d0d683a699d85732a54ec4ce6fe93f60323ce13c14f2558ff6922042da7709","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_38d18632501697af4ada9c62e750c8fd8b8cc0901b5dd7425f68c740790779b1"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-02-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_d1854f026c5b433ca83c02a7d0582332f2eb4594f2a0f3223a075ac97db14769","configuration_identity":{"context_policy":"32k","effort":"unspecified","model_snapshot":"anthropic/claude-opus-4-6","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.4","source_model_version":"claude-opus-4-6_32K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"90.530303%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.016916245311450143","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"DiXCMhdfhcjaCUSzYWgACy"},"evidence_class":"direct_strict","model_configuration":"claude-opus-4-6_32K","organization":"Anthropic","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_8fc819f56f02502eeca9699434f3dd5b84bdb0616d07c0a012272c7a721786b1","canonical_evidence_variant_id":"gpqa_evd_v1_2e9f5bc60a43efb1e18bff823dd3eb43757ce1e13a698d8a5a9d24d3a785c77f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_8fc819f56f02502eeca9699434f3dd5b84bdb0616d07c0a012272c7a721786b1","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_2e9f5bc60a43efb1e18bff823dd3eb43757ce1e13a698d8a5a9d24d3a785c77f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_d212f3b7012127430e247ff410264d8f3ad2193551b8da2fd4265bff806d73fa","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4o-mini-2024-07-18","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4o-mini-2024-07-18","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"37.720960%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024068925424620136","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"L9yfwZC3snsKZshBXTaXjc"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4o-mini-2024-07-18","organization":"OpenAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_c6a9c4a5c08a066c77e81fe8228a1174a0376b9d8ffab4570806764d9d0063d0","canonical_evidence_variant_id":"gpqa_evd_v1_3785c6853487ce53ec1e2d0239ab56dc955496b1dd1e9813324d4c628900c02e","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_c6a9c4a5c08a066c77e81fe8228a1174a0376b9d8ffab4570806764d9d0063d0","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_3785c6853487ce53ec1e2d0239ab56dc955496b1dd1e9813324d4c628900c02e"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_d255d05df7d6452741ae49d1c8831a4ac1bdeb8644fc8d5d6b034ded7651c060","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/kimi-k3","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"kimi-k3_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"84.848485%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.025545650426603637","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"mvNb3hNgSZ2gS6jHMtvCdk"},"evidence_class":"direct_strict","model_configuration":"kimi-k3_low","organization":"Moonshot","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_74681dd1b2c1294340edec610a41dad935c775a6e62717aa6c6b11d2320be9b5","canonical_evidence_variant_id":"gpqa_evd_v1_9363b6a95d4c5080b049e9e51f01065d4e54cc46c7771bca9f19e02654853141","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_74681dd1b2c1294340edec610a41dad935c775a6e62717aa6c6b11d2320be9b5","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_9363b6a95d4c5080b049e9e51f01065d4e54cc46c7771bca9f19e02654853141"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_d6f20b74ffa7e4b76d5c03ebd92e8ebb6053e6c47efc1d83f30e9db73129cb4f","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/qwen3.6-27b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.6-27b","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"85.858586%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.024825909793343353","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"buLf8Ne29xmzMZBRQkvrQG"},"evidence_class":"direct_strict","model_configuration":"qwen3.6-27b","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_e24e64b1867767e63662a1a29ec36856570f8e7c17f2479d7d156f4ce0b5326a","canonical_evidence_variant_id":"gpqa_evd_v1_1445ad1e03ba2ff428c21e82355430097a475172456c0f79f66a9b2219e595a4","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e24e64b1867767e63662a1a29ec36856570f8e7c17f2479d7d156f4ce0b5326a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_1445ad1e03ba2ff428c21e82355430097a475172456c0f79f66a9b2219e595a4"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-11-25","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_d7b58af1465d20d5223ea9f8a5e54b764667d1cb51518b107c4dbfe5c939e1fd","configuration_identity":{"context_policy":"16k","effort":"unspecified","model_snapshot":"anthropic/claude-opus-4-5-20251101","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.2","source_model_version":"claude-opus-4-5-20251101_16K","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"85.479798%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.021508147720723254","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"cXkPywkhy3NnW992ri73yG"},"evidence_class":"direct_strict","model_configuration":"claude-opus-4-5-20251101_16K","organization":"Anthropic","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_1cf62a6fef04552758a0be4132c418f9d2fb9203aa472977263543983c086b56","canonical_evidence_variant_id":"gpqa_evd_v1_bf02fd015d98b1222021573eae11fae1dd14a77e3a0dd48c197fdce39565ba3f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_1cf62a6fef04552758a0be4132c418f9d2fb9203aa472977263543983c086b56","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_bf02fd015d98b1222021573eae11fae1dd14a77e3a0dd48c197fdce39565ba3f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-03-10","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_d83a57d702d18c236094a9c8c5928b5c1aee75f30eb5fa16b86d87b7c6f6c312","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"together/deepseek-ai/deepseek-r1-distill-llama-70b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"together","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"DeepSeek-R1-Distill-Llama-70B","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"55.744949%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02973465219179067","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"f6pnAATjxmrFJS353EjQKi"},"evidence_class":"direct_compatibility","model_configuration":"DeepSeek-R1-Distill-Llama-70B","organization":"DeepSeek","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_4b30364c841a91549b51f0cff62105edb2a9f5186db422d8648952e345968b8d","canonical_evidence_variant_id":"gpqa_evd_v1_620b9510b36a4677a28fe7760dd4a198397edcdce979248c60e734b8e117e9aa","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_4b30364c841a91549b51f0cff62105edb2a9f5186db422d8648952e345968b8d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_620b9510b36a4677a28fe7760dd4a198397edcdce979248c60e734b8e117e9aa"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_de907b3d30ffa90340ef7c1d7ab47a458ec5f37194fde893d5354bb08a8c14fc","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/gpt-5.4-nano-2026-03-17","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.4-nano-2026-03-17_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"55.555556%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.035402943770953675","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"gLewKMc7ttCD2moUrgAyYX"},"evidence_class":"direct_strict","model_configuration":"gpt-5.4-nano-2026-03-17_none","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_147dc8f413f53a986ea0be696787aa4a7bcc8f7c81f149e1b7e184866ca9c683","canonical_evidence_variant_id":"gpqa_evd_v1_fa4d7e040bc5714277737efe82f63c1f88d40ca91d4d26a189907fdb1fd9afee","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_147dc8f413f53a986ea0be696787aa4a7bcc8f7c81f149e1b7e184866ca9c683","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_fa4d7e040bc5714277737efe82f63c1f88d40ca91d4d26a189907fdb1fd9afee"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_dfcacd48f5945f1701b6202adec1a676dc8eb0e19b95f390dfe26319840ae1d5","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/gemma-3-27b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemma-3-27b-it","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"43.939394%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03536085947529479","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"AsVZCdJLQvNykhKjisihe2"},"evidence_class":"direct_strict","model_configuration":"gemma-3-27b-it","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_e2e8c76ceab18cf5dc14a36af0f055bed47763cb11426d5b0ac65a54a57055fb","canonical_evidence_variant_id":"gpqa_evd_v1_8607afe2336d204b79bed741f39ffa46a59fc049b7bc0649101afacead6b91e9","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e2e8c76ceab18cf5dc14a36af0f055bed47763cb11426d5b0ac65a54a57055fb","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_8607afe2336d204b79bed741f39ffa46a59fc049b7bc0649101afacead6b91e9"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e0339f47a0e11c791da4e131cf06ead62b4fe9b676d100b883e92ae30cec47ad","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"epoch/deepseek-v4-pro","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"deepseek-v4-pro_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"90.909091%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.020482086775424225","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"TPdRwHTJtaJYSu2Cxc5Czn"},"evidence_class":"direct_strict","model_configuration":"deepseek-v4-pro_high","organization":"DeepSeek","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_7190e4ed2bc64cec4c9dad819a2bf26b434da72593287bebdbaff37a24feeed2","canonical_evidence_variant_id":"gpqa_evd_v1_614f4386bf81f4cf532eb2cc0564908bc03e9f16ec017282de07dbe80fc46c4f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_7190e4ed2bc64cec4c9dad819a2bf26b434da72593287bebdbaff37a24feeed2","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_614f4386bf81f4cf532eb2cc0564908bc03e9f16ec017282de07dbe80fc46c4f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-02-27","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e0a3d196b9052321e7008c698e73f936a46f53cfdf612305492866420090cc21","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4-turbo-2024-04-09","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4-turbo-2024-04-09","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"46.590909%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026576883332580552","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"TaCoN7eyPQEfDWXqs2MpcF"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4-turbo-2024-04-09","organization":"OpenAI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_87b9b646a70b83e9dee4e484887cb3697e1645d3bafca76160ca769c766338ca","canonical_evidence_variant_id":"gpqa_evd_v1_9366d13835c1409504e6a6524428eaf482ae1c956f2dd8dd4d32a7fc49320512","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_87b9b646a70b83e9dee4e484887cb3697e1645d3bafca76160ca769c766338ca","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_9366d13835c1409504e6a6524428eaf482ae1c956f2dd8dd4d32a7fc49320512"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-20","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e20f7fe639d9555d87a238b746aca2117c1f00aca9069bf5ace08476f215d92f","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/Qwen/Qwen1.5-32B-Chat","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"qwen1.5-32b-chat","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"30.744949%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.020200517973424446","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"9K6c9b6v9zQrQYtw827BDn"},"evidence_class":"direct_compatibility","model_configuration":"qwen1.5-32b-chat","organization":"Alibaba","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_52b9ebb484e2f8e3269183cfeda70aa34c7ff9f273833761779c21b713a3d7a5","canonical_evidence_variant_id":"gpqa_evd_v1_43384103fa2c391d8438da506a04ac275685979836549369ee26c4e24809aa64","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_52b9ebb484e2f8e3269183cfeda70aa34c7ff9f273833761779c21b713a3d7a5","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_43384103fa2c391d8438da506a04ac275685979836549369ee26c4e24809aa64"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-13","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e303a40eed6943f0cc681f7368cff1f8202b2615ff9e972a0f6b2ec5c323afbc","configuration_identity":{"context_policy":"source_default","effort":"minimal","model_snapshot":"epoch/gpt-5-nano-2025-08-07","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5-nano-2025-08-07_minimal","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"48.484848%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.03560716516531066","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"hp4x7bhBYhkQqZoZFf4dTZ"},"evidence_class":"direct_strict","model_configuration":"gpt-5-nano-2025-08-07_minimal","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_a8a12678ac767ad0ce614c0a52fb3cb05212df7e6fae29008ff1a1c853d066e7","canonical_evidence_variant_id":"gpqa_evd_v1_475900f63a4c5d423cc5855258c5df8fe2c5b17f4f743308257e87deb22ca1e5","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_a8a12678ac767ad0ce614c0a52fb3cb05212df7e6fae29008ff1a1c853d066e7","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_475900f63a4c5d423cc5855258c5df8fe2c5b17f4f743308257e87deb22ca1e5"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-02-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e306757163f97ebb8a7e53e35a5f2c8b98039f375eba0fff85b09abd20803643","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-2.0-flash-001","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemini-2.0-flash-001","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"64.141414%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027395618179027574","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"fbszwTApm3f28z5VA6oXBK"},"evidence_class":"direct_compatibility","model_configuration":"gemini-2.0-flash-001","organization":"Google DeepMind,Google","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_ffc346943686ed230a2d0f91fde947bf8cd434dd5bbc80188a285678a84ba876","canonical_evidence_variant_id":"gpqa_evd_v1_01baead65449a08b3c3cce21e89541808b31b6242f5a2e5463e9c35452333126","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_ffc346943686ed230a2d0f91fde947bf8cd434dd5bbc80188a285678a84ba876","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_01baead65449a08b3c3cce21e89541808b31b6242f5a2e5463e9c35452333126"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e3b6650e66d0e9441ad0ed99a22261238f2ac8b4e5ede02f20cde8b6b5dd76c7","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/gpt-5.6-terra","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.6-terra_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"87.373737%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.023664359402880177","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"JcuoAWhcnngPEZBPmiUyLF"},"evidence_class":"direct_strict","model_configuration":"gpt-5.6-terra_low","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_01da94af7bb743d58604ed51d08b94986a33217955dd249115f6180997b2da55","canonical_evidence_variant_id":"gpqa_evd_v1_f01bdbc29607daac82913d046b5635bde01b1397052d7a828aee5e16548c5eb6","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_01da94af7bb743d58604ed51d08b94986a33217955dd249115f6180997b2da55","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f01bdbc29607daac82913d046b5635bde01b1397052d7a828aee5e16548c5eb6"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e4ead69d7e02fd150d9663a883879015c0849b66ff61f05140cf8318e45c4f6d","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"mistral/mistral-large-2407","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"mistral","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"mistral-large-2407","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"49.021465%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.025918840988656967","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"ZQPNnicTuhuscJvni85PFY"},"evidence_class":"direct_compatibility","model_configuration":"mistral-large-2407","organization":"Mistral AI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_afd4e1102acfce0a7be890d674343db5836d1bd5e4dac9122614973fbf39a7a0","canonical_evidence_variant_id":"gpqa_evd_v1_3a596d2e4295830d1c2ac6fa59311afe7cd4ade34ea5e652f41fa63a15c9ca5f","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_afd4e1102acfce0a7be890d674343db5836d1bd5e4dac9122614973fbf39a7a0","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_3a596d2e4295830d1c2ac6fa59311afe7cd4ade34ea5e652f41fa63a15c9ca5f"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e4fc3d3f5f663e3176f37cb7a7a9414683ad660a66fd87026e1ba802553ac32a","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"epoch/gemini-3.5-flash-lite","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemini-3.5-flash-lite_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"83.333333%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026552207828215258","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"chbh6XstiqtUa5W9zuyfjS"},"evidence_class":"direct_strict","model_configuration":"gemini-3.5-flash-lite_high","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_d07caaae495eaf532cacb85535e3a37be795fc2d7ee2b915e44ae1b266f7e8e5","canonical_evidence_variant_id":"gpqa_evd_v1_20469bd2762a92a967dc6393d649831dc84f0402893151ed2c8d213121629037","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_d07caaae495eaf532cacb85535e3a37be795fc2d7ee2b915e44ae1b266f7e8e5","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_20469bd2762a92a967dc6393d649831dc84f0402893151ed2c8d213121629037"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-15","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e53c5a14f4aeac84c113c190e76b609ecf71fe6a8a08960e87cf07e538dfe203","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/gpt-5.4-2026-03-05","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.4-2026-03-05_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"74.747475%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.030954055470365872","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"LaBuzZExRqocD4FET6yx26"},"evidence_class":"direct_strict","model_configuration":"gpt-5.4-2026-03-05_none","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_e6638e326bb3918bfd4455ca78a1f45c7ce4c0bff2025c897f8537e38257012b","canonical_evidence_variant_id":"gpqa_evd_v1_203ab8170562b95e7b03b6991d2986c568fe66f69f7e984309afe47250368097","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e6638e326bb3918bfd4455ca78a1f45c7ce4c0bff2025c897f8537e38257012b","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_203ab8170562b95e7b03b6991d2986c568fe66f69f7e984309afe47250368097"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e7c76115b41d855efcf0f7ae19b2b6d7a23a44cfa36c8e7a39886451c18c0663","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/qwen3.7-flash","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.7-flash","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"82.323232%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027178752639044908","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"i6i3gQf3ng2aAipvowPpws"},"evidence_class":"direct_strict","model_configuration":"qwen3.7-flash","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_7328c8f7177f793e15d9d53f2624e436d745bbff10e6af514a4b2640a127bd81","canonical_evidence_variant_id":"gpqa_evd_v1_60cc48b4832f2b78d2284b10200cf111ff6b2a011c4185c570b5f9f8e2ac61eb","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_7328c8f7177f793e15d9d53f2624e436d745bbff10e6af514a4b2640a127bd81","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_60cc48b4832f2b78d2284b10200cf111ff6b2a011c4185c570b5f9f8e2ac61eb"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-22","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e8a07532d90482fd2050087865a6e3e30cf6a9ffb8a09595ff45e6be66586c51","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"sagemaker/allenai/Llama-3.1-Tulu-3-70B-DPO","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"sagemaker","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Llama-3.1-Tulu-3-70B-DPO","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"46.275253%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02405836362093961","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"5JgivNDZp2D7Lc7cpDhS28"},"evidence_class":"direct_compatibility","model_configuration":"Llama-3.1-Tulu-3-70B-DPO","organization":"Allen Institute for AI,University of Washington","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_abb18121ff6c108f21e683268963d7b9cbb876b8a1322df5154be358e44f1485","canonical_evidence_variant_id":"gpqa_evd_v1_0cdcc0925e07404e04451714b736e0bf13b28957076b77bb4d86f83f490e0e58","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_abb18121ff6c108f21e683268963d7b9cbb876b8a1322df5154be358e44f1485","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_0cdcc0925e07404e04451714b736e0bf13b28957076b77bb4d86f83f490e0e58"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e8b7f10a34ef4830b8a98b69d984ed96a9afc3e9e8a7858993bef4958cb0eecc","configuration_identity":{"context_policy":"source_default","effort":"minimal","model_snapshot":"epoch/gpt-5-mini-2025-08-07","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5-mini-2025-08-07_minimal","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"71.717172%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.032087795587867514","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"hYSstHWNUNwYoZdmZpYQqK"},"evidence_class":"direct_strict","model_configuration":"gpt-5-mini-2025-08-07_minimal","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_fab72f269f02cefca7966939c8e8f6a5bffa9549cee9e1b3090c403e466cbdc0","canonical_evidence_variant_id":"gpqa_evd_v1_416483ba5d12e8e8f379e458032a4991a17d4b2d5e06e79b3361ff05b803e6be","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_fab72f269f02cefca7966939c8e8f6a5bffa9549cee9e1b3090c403e466cbdc0","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_416483ba5d12e8e8f379e458032a4991a17d4b2d5e06e79b3361ff05b803e6be"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-06-03","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e965137fed2c696942a43e020de955f796ca37ab01b735427391d3454e309d79","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-2.5-pro-preview-05-06","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemini-2.5-pro-preview-05-06","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"66.666667%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.033586181457325226","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"TT66gGiVoeYLGQMCESPrEG"},"evidence_class":"direct_compatibility","model_configuration":"gemini-2.5-pro-preview-05-06","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_dc5be65670991d694c17df37c817319ec24167d10e9fbd23aa16c4fec6d522d6","canonical_evidence_variant_id":"gpqa_evd_v1_c05840e26a572eb38e857946454ba564f6a5b66e80cfd79471bbc81e648ae354","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_dc5be65670991d694c17df37c817319ec24167d10e9fbd23aa16c4fec6d522d6","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_c05840e26a572eb38e857946454ba564f6a5b66e80cfd79471bbc81e648ae354"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-17","configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","configuration_id":"gpqa_cfg_v1_e99e75b593bbdcaf766b1b732e2a1f92ac6b0acbdc8bd59190ce0972000f26e1","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"openai/gpt-5.6-sol","output_policy":"no_output_ceiling","prompt_contract":"gpqa_native_answer_choice_v1","protocol":"mean@1","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","generation_config_sha256":"7a9193a3f2ff640e27d2e75b0bc5d50ce1cf053bbac7e7465825822222e18080","profile":"gpt-5-6-sol-max","task_sha256":"2501e59954a9e53479bc5e44bf9923db905866019f16c385ea3f27483b3697d0"}},"details":{"configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","evidence_label":"Independent native run","execution":{"automatic_retries":0,"fresh_calls":true,"output_ceiling":"none"},"reconciliation":{"paired_original_role":"Reconstructed from the same accepted all-198 native attempt universe as the Clean score.","published_source_score_percent":null,"status":"not_applicable_independent_run"},"recovery":{"kind":"none"},"score_source":"Independent native answer-safe projection","source_kind_label":"Independent project-native run","source_run_id":"gpqa-native-gpt-5-6-sol-max-all198-20260816-r3"},"evidence_class":"project_native","model_configuration":"GPT-5.6 Sol (max)","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_all_normal"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_75308a9a4ce0f186eaeb19b265a31da6d64b5697df4ec64bfcbe3d350c15a145","canonical_evidence_variant_id":"gpqa_evd_v1_3f3dfa8d602ab38ffad53d167bf75d6e6d2bc864ef55a1621a34e1c62ea15b30","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_75308a9a4ce0f186eaeb19b265a31da6d64b5697df4ec64bfcbe3d350c15a145","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_3f3dfa8d602ab38ffad53d167bf75d6e6d2bc864ef55a1621a34e1c62ea15b30"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_e9fe429ba7834a25f5cf0d8f2deb496f1db6fe7b7552467a7b880fdd61f18cc6","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-1.5-pro-001","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemini-1.5-pro-001","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"45.864899%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02617295639558329","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"PKdqpKZ9tqDWyWJ7GksBWz"},"evidence_class":"direct_compatibility","model_configuration":"gemini-1.5-pro-001","organization":"Google DeepMind","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_47519735add1b438c17a478ad1993733210b87670a2f431a9c3025693dd8f5c8","canonical_evidence_variant_id":"gpqa_evd_v1_c53bfe125d212be0e5ca41fe06f7d38f12eef8451b7a84cb0ac79fc7ee4f1bf9","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_47519735add1b438c17a478ad1993733210b87670a2f431a9c3025693dd8f5c8","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_c53bfe125d212be0e5ca41fe06f7d38f12eef8451b7a84cb0ac79fc7ee4f1bf9"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_eb0badf81e252fdb63ecf80d40bb300712ba693f20e27eab39d39914053768f9","configuration_identity":{"context_policy":"source_default","effort":"minimal","model_snapshot":"epoch/gemini-3.5-flash-lite","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gemini-3.5-flash-lite_minimal","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"74.242424%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.031156269519646826","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"7aAKcnsCXoSumfAu8mr2un"},"evidence_class":"direct_strict","model_configuration":"gemini-3.5-flash-lite_minimal","organization":"Google DeepMind","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_2353d94ce816c022abcc725ec98763afab9beb630bafdf4d8f4b53929c8b82a9","canonical_evidence_variant_id":"gpqa_evd_v1_1c402b295c3e04ffba4e8f41d2514d5dd0a8c690e48240a464ee9e1f92943521","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_2353d94ce816c022abcc725ec98763afab9beb630bafdf4d8f4b53929c8b82a9","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_1c402b295c3e04ffba4e8f41d2514d5dd0a8c690e48240a464ee9e1f92943521"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-07-13","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_eb8810fec6811f5a48d750f48d865427a72cac5e58054882006e251bfa0c646c","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"epoch/claude-sonnet-4-6","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-sonnet-4-6_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"83.333333%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026552207828215258","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"DmdKPD9Z9RXwYyATAoG4Uw"},"evidence_class":"direct_strict","model_configuration":"claude-sonnet-4-6_high","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_419e23c5901499f84adb13a55e1fb49886f3825bac67ccffb0c8a95e5330c540","canonical_evidence_variant_id":"gpqa_evd_v1_eb64148b9ec84f0196e0e89910508d0d29ab7ba3969e9b7871880454d31aaca1","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_419e23c5901499f84adb13a55e1fb49886f3825bac67ccffb0c8a95e5330c540","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_eb64148b9ec84f0196e0e89910508d0d29ab7ba3969e9b7871880454d31aaca1"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_ebe93e140bced2f2c9f1711124393f7b59a700e61071b4652a4827cd3e713257","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/gpt-5.1-2025-11-13","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.1-2025-11-13_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"66.666667%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.033586181457325226","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"UT5fVL4EGvSTZBKMTPy9YN"},"evidence_class":"direct_strict","model_configuration":"gpt-5.1-2025-11-13_none","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_1c9e192f85b264d68234f5d769b6214dc0efe4da9362a5329f0bd6e5ff330295","canonical_evidence_variant_id":"gpqa_evd_v1_09e49c686372b8cb2f26f51a86830f6fa6da26694e7a114a4f24bce18f7e4848","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_1c9e192f85b264d68234f5d769b6214dc0efe4da9362a5329f0bd6e5ff330295","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_09e49c686372b8cb2f26f51a86830f6fa6da26694e7a114a4f24bce18f7e4848"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_ed668418bc3a7abc717016181a36d45c94836c4061801b735c989f31101fc953","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/gpt-5.6-sol","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"gpt-5.6-sol_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"82.828283%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.026869716187429872","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"DDfYPyKQW7vgDtGgkwcKSi"},"evidence_class":"direct_strict","model_configuration":"gpt-5.6-sol_none","organization":"OpenAI","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_67e0b61faf747c532ac6c52c154c923005bd0cbe2d60146d961de158b993737b","canonical_evidence_variant_id":"gpqa_evd_v1_10e84a7924c260673a9bef7488f3c9b63e0f6bb27776dd6e504700e6a4a4a53a","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_67e0b61faf747c532ac6c52c154c923005bd0cbe2d60146d961de158b993737b","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_10e84a7924c260673a9bef7488f3c9b63e0f6bb27776dd6e504700e6a4a4a53a"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_ee7683485330a0bb484846eff9b707715b9688adb7f1e75b90ec838b53083c08","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"openai/gpt-4o-2024-08-06","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gpt-4o-2024-08-06","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"49.210859%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.025999270277271707","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"fairJpL9VekDBKxYHG9ZoG"},"evidence_class":"direct_compatibility","model_configuration":"gpt-4o-2024-08-06","organization":"OpenAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_8adf4c17b0dd41a26412938738bf1bc4c241411f3bfbf92cd828b502e3e20b1d","canonical_evidence_variant_id":"gpqa_evd_v1_6b9efd62f14bc0f7a0b3a80a1eeaf7e87b621361f1dc52119f98cfbd275a0873","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_8adf4c17b0dd41a26412938738bf1bc4c241411f3bfbf92cd828b502e3e20b1d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_6b9efd62f14bc0f7a0b3a80a1eeaf7e87b621361f1dc52119f98cfbd275a0873"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_efb979697550c6b7b9d2f7fbf5e13c8cdcdc7f6b1ed66a691f8f08a9078593bc","configuration_identity":{"context_policy":"source_default","effort":"none","model_snapshot":"epoch/qwen3.6-35b-a3b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.6-35b-a3b_none","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"84.848485%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02554565042660364","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"9kErpmgQ42wTvgJaFn478v"},"evidence_class":"direct_strict","model_configuration":"qwen3.6-35b-a3b_none","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_805a74e8548e777ae3cd3cfbffd8506fe94858411c008499aa834454e55cd72f","canonical_evidence_variant_id":"gpqa_evd_v1_834c47908d3de9900df586ebe7ca68f03d829204b0ea23d8b4f496cc22bb5f13","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_805a74e8548e777ae3cd3cfbffd8506fe94858411c008499aa834454e55cd72f","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_834c47908d3de9900df586ebe7ca68f03d829204b0ea23d8b4f496cc22bb5f13"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-04-08","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_f1d46d247ebf82560a68e3bc5cf4e07c002603eaadfa51893bf7ed84564bbdbf","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"together/meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"together","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"Llama-4-Maverick-17B-128E-Instruct-FP8","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"66.982323%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.028244823529728173","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"LyHy99ubGCqBhoaksQjkiF"},"evidence_class":"direct_compatibility","model_configuration":"Llama-4-Maverick-17B-128E-Instruct-FP8","organization":"Meta AI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_6708f641f7b1c2453853d0c0f5b44fce92f7e1fe8bbaf674c8ac8b68e3ebd355","canonical_evidence_variant_id":"gpqa_evd_v1_18558b809255892beb305d1594a920bffd5ebadad5299e140cf625a0a4b32be5","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_6708f641f7b1c2453853d0c0f5b44fce92f7e1fe8bbaf674c8ac8b68e3ebd355","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_18558b809255892beb305d1594a920bffd5ebadad5299e140cf625a0a4b32be5"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_f2ec9c6f943036ef8ca6bef0edc2ea301077976070fe65d840676bbb44703a53","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"anthropic/claude-2.1","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"anthropic","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"claude-2.1","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"32.954545%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02324885083265289","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"eMAisrJd7fAUHPfcPWXeXo"},"evidence_class":"direct_compatibility","model_configuration":"claude-2.1","organization":"Anthropic","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_9e34e9e4b250171d57b71fa15eb1d53e2777c0e37c57f5e790702d34506d435b","canonical_evidence_variant_id":"gpqa_evd_v1_c47972f8303d2007196ba7f5312bc24202247bddab4ecdffe534d98886e18363","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_9e34e9e4b250171d57b71fa15eb1d53e2777c0e37c57f5e790702d34506d435b","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_c47972f8303d2007196ba7f5312bc24202247bddab4ecdffe534d98886e18363"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-03-12","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_f37332153c1ff0005df8b2ece84a9097d786a435e9046ba52befebedb2adfd3e","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"google/gemini-1.5-flash-8b-001","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"google","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"gemini-1.5-flash-8b-001","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"32.954545%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02416551116211757","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"gFT36W4KU3wggLQ9nNCctg"},"evidence_class":"direct_compatibility","model_configuration":"gemini-1.5-flash-8b-001","organization":"Google DeepMind","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_e53fe107892393c2cca4ab8b9572a08fae423556b3acb3a07e56eb39fafaf584","canonical_evidence_variant_id":"gpqa_evd_v1_a5dcd773dbe55fda5fc636b9f84c9bd0a59d39eaddee70cffcaad64103e05720","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_e53fe107892393c2cca4ab8b9572a08fae423556b3acb3a07e56eb39fafaf584","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_a5dcd773dbe55fda5fc636b9f84c9bd0a59d39eaddee70cffcaad64103e05720"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-17","configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","configuration_id":"gpqa_cfg_v1_f657ce567632b0af5c3025bd73318b5a4796ffe43af7c00f6a98baea08eea2b2","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"grok/grok-4.3","output_policy":"no_output_ceiling","prompt_contract":"gpqa_native_answer_choice_v1","protocol":"mean@1","provider_route":"grok","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","generation_config_sha256":"4db33ee3aca837b8beb02b9e0714860bd517fe9ef65c086a876468bb7b9f3928","profile":"grok-4-3-high","task_sha256":"2501e59954a9e53479bc5e44bf9923db905866019f16c385ea3f27483b3697d0"}},"details":{"configuration_date_basis":"frozen release date fallback because the answer-safe native import has no run timestamp","evidence_label":"Independent native run","execution":{"automatic_retries":0,"fresh_calls":true,"output_ceiling":"none"},"reconciliation":{"paired_original_role":"Reconstructed from the same accepted all-198 native attempt universe as the Clean score.","published_source_score_percent":null,"status":"not_applicable_independent_run"},"recovery":{"kind":"none"},"score_source":"Independent native answer-safe projection","source_kind_label":"Independent project-native run","source_run_id":"gpqa-native-grok-4-3-high-all198-20260816-r2"},"evidence_class":"project_native","model_configuration":"Grok 4.3 (high)","organization":"xAI","protocol":"mean@1","validation_profile":"strict_all_normal"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_11891105ed0f8462d7e8df193d42d17c0293f0ff9eb109ebb09098dc4d3afe7b","canonical_evidence_variant_id":"gpqa_evd_v1_77b79be9e51b33386fd343cef692dea70e6249188126a76ed404f6bbd83c0b6c","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_11891105ed0f8462d7e8df193d42d17c0293f0ff9eb109ebb09098dc4d3afe7b","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_77b79be9e51b33386fd343cef692dea70e6249188126a76ed404f6bbd83c0b6c"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-11-11","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_f76c85c895ef2ab274d28dcefa237ebea5b991f40fdb01ee93070b5634a7ca73","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"moonshot/kimi-k2-thinking-turbo","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"moonshot","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.1","source_model_version":"kimi-k2-thinking-turbo","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"84.217172%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.020879845731985747","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"muGecFpAg9wHd6gexWJ99L"},"evidence_class":"direct_strict","model_configuration":"kimi-k2-thinking-turbo","organization":"Moonshot","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":16,"canonical_evaluation_run_id":"gpqa_run_v1_89e7ac1d13ebb3e7f1326b2e0d714604e4b41c21a83e3c8ea7485423ff0ff2e7","canonical_evidence_variant_id":"gpqa_evd_v1_f202413019dd16e849f4fb809beb659c1a0a00d2be151135811bbfe77437672e","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_89e7ac1d13ebb3e7f1326b2e0d714604e4b41c21a83e3c8ea7485423ff0ff2e7","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f202413019dd16e849f4fb809beb659c1a0a00d2be151135811bbfe77437672e"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-01-17","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_f77deb38bca0cfa2e68170980f653b63a4bea919a4a2c2475cdc2a37c7b82e48","configuration_identity":{"context_policy":"source_default","effort":"medium","model_snapshot":"openai/o1-mini-2024-09-12","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@16","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"16","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"o1-mini-2024-09-12_medium","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"59.501263%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02795124238637276","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"4NiUpCrLfz3oFaerkrMeBt"},"evidence_class":"direct_compatibility","model_configuration":"o1-mini-2024-09-12_medium","organization":"OpenAI","protocol":"mean@16","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_d8361abeb8e6acd4751b11badb2b386a2411608213c72001ac4522da086c93f1","canonical_evidence_variant_id":"gpqa_evd_v1_1b43335efd685be40f7926a1a9b21834dfe490d6f861040f1f46e757dae3ef25","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_d8361abeb8e6acd4751b11badb2b386a2411608213c72001ac4522da086c93f1","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_1b43335efd685be40f7926a1a9b21834dfe490d6f861040f1f46e757dae3ef25"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_f9fede76cf79ca860917f59fb5ebf3126982f366fe21b16ef0e22d4d13f4038a","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"epoch/kimi-k2.7-code","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"kimi-k2.7-code","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"87.878788%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02325315795194205","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"XneEig6pACkAWeekvgCwNk"},"evidence_class":"direct_strict","model_configuration":"kimi-k2.7-code","organization":"Moonshot","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_f769a1bb47fe0a249f6a61005b7ad9fa5cb56532a7c6c2cde29b07fc99a78462","canonical_evidence_variant_id":"gpqa_evd_v1_deb5a2dc0e3d18d13b3f704e153198f623e64bbe63a62ea3e8bba380e8e6e30b","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_f769a1bb47fe0a249f6a61005b7ad9fa5cb56532a7c6c2cde29b07fc99a78462","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_deb5a2dc0e3d18d13b3f704e153198f623e64bbe63a62ea3e8bba380e8e6e30b"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-03-10","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_fb3188e98dcd04325c815261985bb5384ea7240ee9d0ef804de82473aede6bc9","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"together/deepseek-ai/deepseek-r1-distill-qwen-14b","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"together","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"DeepSeek-R1-Distill-Qwen-14B","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"44.696970%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.02863094678190378","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"8UqaLobhmNJLuPGTTpeaCw"},"evidence_class":"direct_compatibility","model_configuration":"DeepSeek-R1-Distill-Qwen-14B","organization":"DeepSeek","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_d88b3fde412bc3daef1d1dbaac2e7984bf68c0f93a18e75fcab47d9a1f1dc75f","canonical_evidence_variant_id":"gpqa_evd_v1_d3f22782b417834252665ddb1102c1fc18877445b8152cbacf1ca1140fe4664d","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_d88b3fde412bc3daef1d1dbaac2e7984bf68c0f93a18e75fcab47d9a1f1dc75f","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d3f22782b417834252665ddb1102c1fc18877445b8152cbacf1ca1140fe4664d"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2025-03-18","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_fe008500b306f51fcb6fd66c7d472ecf21625ed009686be7a81709ebd6aad0c7","configuration_identity":{"context_policy":"source_default","effort":"unspecified","model_snapshot":"mistral/mistral-small-2503","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"mistral","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"unavailable_in_accepted_legacy_header","source_model_version":"mistral-small-2503","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Compatibility-complete legacy-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"47.474747%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.027616495140012243","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"S8bBDAbucsFceirHxeEDvu"},"evidence_class":"direct_compatibility","model_configuration":"mistral-small-2503","organization":"Mistral AI","protocol":"mean@8","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_2a3998731aae032047bf6cdf10d6214e0452a01a333e66aa6a4563bf93377a96","canonical_evidence_variant_id":"gpqa_evd_v1_155d6aa5896c8d17cff48bb97ca705e21a1f690a4fd87441cb4ace09eb37a96a","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_2a3998731aae032047bf6cdf10d6214e0452a01a333e66aa6a4563bf93377a96","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_155d6aa5896c8d17cff48bb97ca705e21a1f690a4fd87441cb4ace09eb37a96a"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-06","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_fe902b8590238e8946b66aae41d7ee7406b4975a2264ad5dc8639eef08bc33ac","configuration_identity":{"context_policy":"source_default","effort":"low","model_snapshot":"epoch/claude-opus-4-8","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"claude-opus-4-8_low","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"88.383838%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.022828881775249357","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"GThtce6GAmRAS8srQyGQoA"},"evidence_class":"direct_strict","model_configuration":"claude-opus-4-8_low","organization":"Anthropic","protocol":"mean@1","validation_profile":"strict_complete"},{"attempts_per_question":8,"canonical_evaluation_run_id":"gpqa_run_v1_0e4b6e4c02d3232be69a1c3170e5e8ff805e13c1eb71d62ced537c556c817201","canonical_evidence_variant_id":"gpqa_evd_v1_f939de26b6886fbb10f4be0e85613b485726fae0db2aa4d5de86f1a5db887a63","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_0e4b6e4c02d3232be69a1c3170e5e8ff805e13c1eb71d62ced537c556c817201","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_f939de26b6886fbb10f4be0e85613b485726fae0db2aa4d5de86f1a5db887a63"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-04-15","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_fee207fd49a1b7d70751742c0f126d3d087b989d89849747726ee9468fdf26f8","configuration_identity":{"context_policy":"source_default","effort":"high","model_snapshot":"openai/gpt-5.4-mini-2026-03-17","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@8","provider_route":"openai","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"8","benchmark_version":"1.0.6","source_model_version":"gpt-5.4-mini-2026-03-17_high","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"83.585859%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.022274469189069883","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"AQF46S9hBhosKPEMy3fyB9"},"evidence_class":"direct_strict","model_configuration":"gpt-5.4-mini-2026-03-17_high","organization":"OpenAI","protocol":"mean@8","validation_profile":"strict_complete"},{"attempts_per_question":1,"canonical_evaluation_run_id":"gpqa_run_v1_88001e90b5523fd6dac83e0c41e3b7e835d362af3bca441242d12c32d47900cb","canonical_evidence_variant_id":"gpqa_evd_v1_95397797b4a12df58b6492f30b469f0caaf6693aa46c9ee2bb567093236faf38","canonical_selection":{"contract_version":"gpqa-results-identity-v1","evidence_history":[{"canonical":true,"evaluation_run_id":"gpqa_run_v1_88001e90b5523fd6dac83e0c41e3b7e835d362af3bca441242d12c32d47900cb","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_95397797b4a12df58b6492f30b469f0caaf6693aa46c9ee2bb567093236faf38"}],"precedence":["project_native","direct_strict","direct_compatibility","targeted_recovery","derived_uniform"],"score_blind":true,"selection_timestamp_policy":"All hash-pinned accepted inputs are assigned the frozen release timestamp only for deterministic canonical selection. This administrative timestamp is not evidence creation, run, first-seen, or real admission time."},"configuration_date":"2026-08-07","configuration_date_basis":"validated source created_at metadata","configuration_id":"gpqa_cfg_v1_fefffa8ea64c405943e964126ee6c767f5126e9879cdf3c2989670b3db7d61ba","configuration_identity":{"context_policy":"source_default","effort":"max","model_snapshot":"epoch/qwen3.6-max-preview","output_policy":"source_artifact_policy","prompt_contract":"epoch_gpqa_diamond_v1_0_8_task_v0","protocol":"mean@1","provider_route":"epoch","scorer_contract":"inspect_choice_scorer_v1","settings":{"aggregation":"mean","attempts_per_question":"1","benchmark_version":"1.0.11","source_model_version":"qwen3.6-max-preview","task_version":"0"}},"details":{"configuration_date_basis":"validated source created_at metadata","evidence_label":"Strict modern-header Epoch rescore","reconciliation":{"difference_percentage_points":"0.000000 pp","published_source_score_percent":"87.373737%","published_source_score_role":"Roster attribution only; paired deltas use reconstructed outcomes from one attempt universe.","source_stderr":"0.023664359402880177","status":"exact"},"recovery":{"kind":"none"},"score_source":"Accepted Inspect reductions","source_kind_label":"Epoch public evaluation artifact","source_run_id":"Jv2hJvjpvA8xJW3Bzv7bfw"},"evidence_class":"direct_strict","model_configuration":"qwen3.6-max-preview","organization":"Alibaba","protocol":"mean@1","validation_profile":"strict_complete"}],"schema_version":"gpqa-clean-results-configurations-v1"}
