Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Commonsense Reasoning on HellaSwag, PIQA, OBQA, COPA, and WinoGrande

32.8HellaSwag Accuracy

D3

31.13631.5683232.432May 29, 2026
Updated 2d ago

Evaluation Results

MethodLinks
2026.05
32.868.933.171.253.451.88
2026.05
31.869.229.665.552.549.72
2026.05
31.768.929.26450.648.88
2026.05
31.668.729.86651.449.5
2026.05
31.468.829.466.553.649.94
2026.05
31.368.430.265.951.349.42
2026.05
31.268.528.46451.148.64