SimpleBench
SimpleBench / Reasoning · Everyday common-sense and logic questions assess recognition of real constraints.
What it measures and how
Separates multiple-choice and open-ended answers. AVG@5, temperature, and prompting are fixed. A leaderboard-entry date is not an evaluation date.
How this evidence is used
Candidate under review: new-model coverage is timely. Official reuse terms and a fixed data version are pending. Code licensing is not substituted; no automatic collection or scoring has begun.
Evaluation limits & attribution
Everyday common-sense questions add evidence, but task representativeness and answer format affect interpretation.
Data license: Example tasks and code use MIT. Redisplay terms for the latest independently published official results remain unconfirmed.
Results are published by SimpleBench. Original scores and the NexGateHub consensus index use different scales and cannot be added directly.