Reasoning on MATH, GSM8K, AQUA, AIME, AMC, MMLU, GPQA benchmark leaderboard · SOTA2 Research

Reasoning on MATH, GSM8K, AQUA, AIME, AMC, MMLU, GPQA

82.15Average AccuracyMARSHAL

Evaluation Results

Method	Links
MARSHAL2025.10		82.15	95.2	94.54	86.61	66.67	92.5	89.53	50
MARSHAL2025.10		81.54	95.8	94.39	86.61	63.33	92.5	89.65	48.48
MARSHAL2025.10		81.54	94.4	94.54	86.22	66.67	95	88.98	44.95
MARSHAL2025.10		80.15	94.4	94.69	87.01	60	90	89.53	45.45
SPIRAL2025.10		80.05	94.2	94.47	86.61	60	87.5	91.6	45.96
Qwen3-4B2025.10		79.14	93.4	94.69	85.04	56.67	87.5	89.21	47.47
MARSHAL2025.10		75.96	92.8	95.6	83.86	46.67	80	87.36	45.45
MARSHAL2025.10		75.01	92.2	96.06	83.07	43.33	82.5	86.76	41.12
MARSHAL2025.10		74.54	91.6	96.21	82.68	40	82.5	87.39	41.41
MARSHAL2025.10		73.7	91.4	95.6	82.68	43.33	77.5	87.04	38.38
SPIRAL2025.10		73.41	91.6	95.45	81.89	40	77.5	87.01	40.4
Qwen3-4B2025.10		72.45	90.2	95.91	80.71	40	75	87.42	37.88
SPIRAL2025.10		63.75	87.5	94.8	51.2	36.7	80	58.7	37.37
MARSHAL2025.10		63.54	89.1	95.2	46.5	40	77.5	57.6	38.89
MARSHAL2025.10		62.79	89.9	94.6	52	33.3	75	59.9	34.85
MARSHAL2025.10		62.05	88.1	94.7	48	43.3	65	58.9	36.36
MARSHAL2025.10		61.38	87.8	94.5	48.4	33.3	72.5	59.3	33.84
Qwen3-4B2025.10		60.74	87.6	94.6	39.8	36.7	70	57.1	39.39