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

Mathematical Reasoning on GSM8K (test) (Accuracy, Tokens, Training Time, Speedup)

76.85Accuracy

CPPO

21.501235.870650.2464.6094May 27, 2026
Updated 7d ago

Evaluation Results

MethodLinks
2026.05
76.85265.7-1.22,5414.65
2026.05
76.57262.64-11,8161
2026.05
76.55253.173.611,1971.06
2026.05
76.48161.0338.72,2715.2
2026.05
76.24177.2131.62,0083.58
2026.05
75.92147.2543.92,0025.9
2026.05
75.84232.0110.46,0021.2
2026.05
75.65259.05-7,1861
2026.05
75.5285.72-10.32,3613.04
2026.05
75.34155.0640.11,8123.97
2026.05
23.63----