Mathematical Reasoning on Math Domain In-Domain benchmark leaderboard · SOTA2 Research

Mathematical Reasoning on Math Domain In-Domain

91MATH500DFPO

Evaluation Results

Method	Links
DFPO2026.02		91	55	31.99	85.83	65.96	39.34
Dr. GRPO2026.02		90.8	46.67	30.88	88.33	64.17	38.23
Reinforce++2026.02		90.2	50	30.88	84.17	63.81	38.1
FlowRL2026.02		90.2	55	29.41	86.67	65.32	38.78
lambda-GRPO2026.02		90	50	31.62	80	62.91	37.51
Robust Bellman2026.02		89.4	36.67	31.99	84.17	60.56	35.98
GRPO2026.02		89.2	35	28.68	84.17	59.26	35.39
Base2026.02		88.4	50	31.25	80.83	62.62	37.18
BAPO2026.02		87	38.33	30.51	75	57.71	34.48
PPO2026.02		86.4	36.67	30.51	80	58.4	34.8
KTAE2026.02		86	38.33	29.41	78.33	58.02	34.62