MATH
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
MATH-500
73Accuracy (Raw Consensus)
6
Math AIME26, HMMT (test)
69.3AIME 26 Score
6
MATH 500
86.8Pass@1
6
MATH random architecture
6ASR
6
MATH500
64Accuracy
6
MATH 500
64Accuracy
6
MATH 500
4.5Acceptance Length
6
MATH500 Level 4
56.2IAA
6
MATH 500
82.8Accuracy
6
MATH standard (test)
77.2Pass@1
6
MATH mini
36.4EM Accuracy
6
MATH (test)
45.5Accuracy
6
MATH 500
86.8Accuracy
6
MATH-500
73Token Reduction (%)
6
MATH
96.04Algebra Accuracy
6
MATH
9.42Score
6
MATH
67.6PR-AUC
6
MATH 128
31.41Accuracy
6
MATH
48.1Accuracy
6
MATH
27Accuracy
6
MATH 500
93.4Reasoning Performance
6
MATH
37.3Score
6
MATH500
79Accuracy (MATH500)
6
MATH
43.8Clean Error
6
MATH (test)
96.9TDR
6