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

Few-shot Commonsense Reasoning on Commonsense Reasoning Suite (HellaSwag, PIQA, OBQA, COPA, WinoGrande)

27.3HellaSwag Accuracy

26.05226.37626.727.024May 29, 2026
Updated 2d ago

Evaluation Results

MethodLinks
2026.05
27.355.715.263.551.142.56
2026.05
26.656.813.85950.141.26
2026.05
26.554.5136249.141.02
2026.05
26.455.712.25949.940.64
2026.05
26.255.811.56250.441.18
2026.05
26.155.511.75849.940.24
2026.05
26.155.613.2605040.98