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

Mathematical Reasoning on GSM8K (Accuracy, Token count)

96.6Accuracy

Vanilla

92.4493.5294.695.68Apr 8, 2026
Updated 9d ago

Evaluation Results

MethodLinks
2026.04
96.61,707
2026.04
96.21,672
2026.04
96.2849
2026.04
962,208
2026.04
96849
2026.04
95.81,133
2026.04
95.81,212
2026.04
95.5979
2026.04
95.3896
2026.04
95.1963
2026.04
95.11,152
2026.04
95.1278
2026.04
94.91,307
2026.04
94.8283
2026.04
92.6303