# Licenses and attribution

Release: `gpqa-clean-results-v0.1.0`
GPQA-Diamond-Clean: `1.2.2`

## Result data and documentation

The aggregate result data and accompanying documentation in this bundle are made available under the Creative Commons Attribution 4.0 International license: https://creativecommons.org/licenses/by/4.0/

## Pipeline code

The release pipeline code is licensed under Apache-2.0 in its source repository. This data bundle does not include that code.

## Upstream benchmark

GPQA is by David Rein, Betty Li Hou, Asa Cooper Stickland, Jackson Petty, Richard Yuanzhe Pang, Julien Dirani, Julian Michael, and Samuel R. Bowman. Cite the original paper: https://arxiv.org/abs/2311.12022

GPQA-Diamond-Clean 1.2.2 is a corrected, ID-only overlay. Its manifest SHA-256 is `65edff43aa69d39a7e97fe6fc2a78433c0c77a62af7fd0be8bac20833e770a96`; its exclusion ledger SHA-256 is `d51f84fe1df413b062b3ce1b8c7f93ecd1f29484711e9019bdd1d019c25292a5`. See https://github.com/adamallcock/gpqa-diamond-clean

## Evaluation evidence

Epoch AI is attributed as the source of the public GPQA Diamond roster and imported evaluation evidence where labelled. Independent native runs and targeted recovery composites are explicitly marked and are not presented as original Epoch aggregates.

This is an independent research project and is not affiliated with or endorsed by GPQA's authors, Epoch AI, or any model provider.