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

Mathematical Reasoning on GSM8K (Accuracy@4 & GRPO Speedup)

93.9Accuracy (mean@4)

DUET

57.08466.64276.285.758May 8, 2026
Updated 22d ago

Evaluation Results

MethodLinks
2026.05
93.91.43
2026.05
92.32.38
2026.05
90.31.19
2026.05
90.11
2026.05
90.10.75
2026.05
89.90.63
2026.05
84.71.62
2026.05
842.51
2026.05
79.61.3
2026.05
78.10.95
2026.05
781
2026.05
77.40.81
2026.05
75.91.26
2026.05
75.71
2026.05
75.50.82
2026.05
75.20.95
2026.05
74.81.13
2026.05
72.82.04
2026.05
60.4-
2026.05
60.3-
2026.05
58.5-