Humanity’s Last Exam
CAIS / Scale AI / Knowledge · Difficult academic knowledge and reasoning across disciplines.
What it measures and how
Fixes tasks, tools, and judges. HLE, HLE-Rolling, and tool-enabled results remain separate. AA v4.3 already includes HLE; future specialist results may inform categories without adding overall weight.
How this evidence is used
Under review: some new models are now covered, but raw results lack verifiable evaluation dates. Epoch's self-evaluation license is not substituted; AA evidence is not counted again.
Evaluation limits & attribution
Under review: some new models are now covered, but raw results lack verifiable evaluation dates. Epoch's self-evaluation license is not substituted; AA evidence is not counted again.
Data license: Evaluation code is MIT. Official-result redisplay terms remain unconfirmed.
Results are published by CAIS / Scale AI. Original scores and the NexGateHub consensus index use different scales and cannot be added directly.