MATH
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
MATH 500
84Average Score @16
4
MATH 500
95.8Average @ 4
4
Math Frustration v3 (held-out)
5.91AUC
4
MATH
30.62Average Latency (MATH)
4
MATH 500
76.82S-Format Score
4
MATH 500
802.3Average Output Length
4
MATH
86.5Pass@1 Accuracy
4
Math
33.92TPS
4
Math
36.89Throughput (Tokens/s)
4
MATH-500 sympy-verified (test)
23.6Accuracy (Config 64)
4
MATH 500
90.2MATH 500 Score
4
MATH
79Pass@1
4
MATH 500
78.8Majority@2048
4
Math500
97.6Accuracy (%)
4
MATH 500
3.04Success Rate (SR)
4
Math
0.94AUC
4
MATH500
13.7Full Length
4
MATH 500
33.8Accuracy (128 tokens)
4
MATH-500 (held-out)
4.79Accuracy Delta (pp)
4
MATH 500
41.2REST Score
4
MATH 500
5.12SGLang Acceptance Length
4
MATH 500
5.89Acceptance Length
4
MATH-500 n=200 (pilot)
69.5Accuracy (%)
4
MATH 500
97.2Accuracy
4
MATH-500
82.6Pass@8
4