List your company
ResearchFormerly WizWand

Mathematical Reasoning on MATH in-domain (test)

80.8AccuracyGEAR
-0.94420.27841.562.722Oct 5, 2023Mar 11, 2024Aug 16, 2024Jan 22, 2025Jun 29, 2025Dec 4, 2025May 12, 2026
Updated 5mo ago

Evaluation results

MethodLinks
GEARBackbone=Qwen3-8B, Evaluation Context=Standard2026.0580.8
MT-GRPOBackbone=Qwen3-8B, Evaluation Context=Standard2026.0576.8
ARPOBackbone=Qwen3-8B, Evaluation Context=Standard2026.0576.3
GRPOBackbone=Qwen3-8B, Evaluation Context=Standard2026.0575.5
OPSDBackbone=Qwen3-8B, Evaluation Context=Standard2026.0574.1
GEARBackbone=Qwen3-4B, Evaluation Context=Standard2026.0573.8
GEARBackbone=Qwen3-4B, Evaluation Context=Multi-Domain Generalization2026.0572.4
OPSD+RLBackbone=Qwen3-8B, Evaluation Context=Standard2026.0570.1
OPSDBackbone=Qwen3-4B, Evaluation Context=Standard2026.0569.8
GPT-4 Code2023.1069.7
ARPOBackbone=Qwen3-4B, Evaluation Context=Standard2026.0569.1
MT-GRPOBackbone=Qwen3-4B, Evaluation Context=Standard2026.0568
GRPOBackbone=Qwen3-4B, Evaluation Context=Standard2026.0567.6
OPSD+RLBackbone=Qwen3-4B, Evaluation Context=Standard2026.0566.8
GRPOBackbone=Qwen3-4B, Evaluation Context=Multi-Domain Generalization2026.0564.3
BaseBackbone=Qwen3-8B, Evaluation Context=Standard2026.0557.2
BaseBackbone=Qwen3-4B, Evaluation Context=Standard2026.0556.5
MathCoder-CLBase=CodeLlama, Size=34B2023.1045.2
MathCoder-LBase=Llama-2, Size=70B2023.1045.1
GPT-42023.1042.5
MathCoder-CLBase=CodeLlama, Size=13B2023.1035.9
PaLM-22023.1034.3
ChatGPT-3.52023.1034.1
MathCoder-CLBase=CodeLlama, Size=7B2023.1030.2
MathCoder-LBase=Llama-2, Size=13B2023.1029.9
MathCoder-LBase=Llama-2, Size=7B2023.1023.3
WizardMathBase=Llama-2, Size=70B2023.1022.7
WizardMathBase=Llama-2, Size=13B2023.1014
GalacticaSize=30B2023.1012.7
WizardMathBase=Llama-2, Size=7B2023.1010.7
Llama-1 RFTBase=Llama-1, Size=34B2023.107.4
Llama-1 RFTBase=Llama-1, Size=7B2023.105.2
Llama-1 RFTBase=Llama-1, Size=13B2023.105.1
GalacticaSize=6.7B2023.102.2