Mathematical Reasoning on MATH (test) (Model-Specific Performance) benchmark leaderboard · SOTA2 Research

Mathematical Reasoning on MATH (test) (Model-Specific Performance)

58.7Mistral-7B AccuracyOptimal

Evaluation Results

Method	Links
Optimal2025.10		58.7	65.8	74.4	65.7	61.9	65.3
Optimal2025.10		58.5	67.9	75.4	67.6	63.1	66.5
Optimal2025.10		58.1	65.9	73.8	65.6	61.6	65
Optimal2025.10		57.7	73.1	75.1	67.3	68.9	68.4
Linear WV2025.10		55.8	70.3	72.2	66.2	65.5	66
Optimal2025.10		55.6	64.5	71.7	59.3	57.2	61.7
Logit WV2025.10		55.6	62.9	70.7	62.8	60.7	62.6
Logit WV2025.10		55.6	70.2	72	66	65.5	65.9
Logit WV2025.10		54.9	61.6	69.9	61.9	59.2	61.5
Vanilla WV2025.10		54.9	69.4	71.6	60.3	61.5	63.5
Linear WV2025.10		54.7	62.2	70.1	62.3	60.2	61.9
Linear WV2025.10		54.7	61.5	69.9	60.7	58.9	61.1
Vanilla WV2025.10		54.3	62.9	70.2	59.9	56.8	60.8
Logit WV2025.10		54.3	63.5	70.5	65.1	62.9	63.3
Vanilla WV2025.10		54.3	61.7	69.6	57.1	55.7	59.7
BoN2025.10		54.2	58	68.7	56.9	53.5	58.3
BoN2025.10		54.2	70.3	72.5	63.5	66.1	65.3
Linear WV2025.10		54.1	63	70.3	64.7	62.6	62.9
Vanilla WV2025.10		53.9	62.2	69.4	56.9	54.8	59.4
KDE WV2025.10		53.9	62	68.8	58	58.1	60.2
KDE WV2025.10		53.8	68.4	70.6	60.6	62.8	63.2
BoN2025.10		53.7	62.6	70.4	62.5	59.8	61.8
KDE WV2025.10		53.3	61.4	68.7	58.3	56.9	59.7
Linear WV2025.10		53	61.5	67.9	53.8	51.8	57.6
Logit WV2025.10		53	61.5	67.7	53.9	51.9	57.6
KDE WV2025.10		52.8	61.9	68.8	59.1	58.2	60.2
Vanilla WV2025.10		52.5	61.4	67.7	53.7	51.7	57.4
KDE WV2025.10		52.3	61.2	68	54	52	57.5
MV2025.10		51.3	60.6	67.8	54.1	51.9	57.1
MV2025.10		51.3	60.6	67.8	54.1	51.9	57.1
MV2025.10		51.3	60.6	67.8	54.1	51.9	57.1
BoN2025.10		51.3	59.5	68.8	59	55.9	58.9
MV2025.10		51.3	60.6	67.8	54.1	51.9	57.1
MV2025.10		51.3	65.8	67.8	54.1	57.1	59.2
BoN2025.10		49.5	57.3	65.3	45	46.1	52.7