LightEval Benchmarks

Complete list of supported benchmarks.

Coming Soon

Detailed benchmark documentation is in progress.

Benchmark Categories

• Commonsense Reasoning: HellaSwag, WinoGrande, OpenBookQA
• Scientific Reasoning: ARC Easy, ARC Challenge
• Physical Commonsense: PIQA
• Truthfulness: TruthfulQA
• Mathematics: GSM8K, MATH
• Knowledge: MMLU, TriviaQA
• Language Understanding: GLUE benchmarks

For complete documentation, see the LightEval README.