0

MMLU Redux 2

MMLU-Redux is a subset of 5,700 manually re-annotated questions across 57 MMLU subjects. Implementation of: https://huggingface.co/datasets/edinburgh-dawg/mmlu-redux-2.0.

Domain
rl-env
License
unknown
Published
Jan 2026

Cite

Notes

Only stored in your browser.

Attribution

Leaderboard scores
OpenReward
Attribution policy →

Top score 95 by Qwen3.5 397B A17B - 3 models reporting

Score history

2
0255075100Feb 26Mar 26Apr 26Qwen3.5 397B A17B

Top models

3
MMLU Redux 2Bar chart with 3 bars. Highest value: Qwen3.5 397B A17B at 95.
3 models

FAQ

What is MMLU Redux 2?
MMLU-Redux is a subset of 5,700 manually re-annotated questions across 57 MMLU subjects. Implementation of: https://huggingface.co/datasets/edinburgh-dawg/mmlu-redux-2.0.
What is the current top score on MMLU Redux 2?
The top reported score is 95 by Qwen3.5 397B A17B, across 3 models reporting.
What license is MMLU Redux 2 under?
MMLU Redux 2 is available under unknown.