List your company
ResearchFormerly WizWand

Mathematical Reasoning on CMATH (test)

89.7AccuracyQwen2.5-Math-72B
49.97260.28670.680.914Feb 5, 2024Jun 18, 2024Oct 30, 2024Mar 13, 2025Jul 25, 2025Dec 6, 2025Apr 20, 2026
Updated 6mo ago

Evaluation results

MethodLinks
Qwen2.5-Math-72Bshot=6-shot, prompting=few-shot chain-of-thought2024.0989.7
DeepSeekMath-RLSize=7B, Reasoning Mode=Chain-of-Thought, Source Type=Open-Source, Evaluation Protocol=Top12024.0288.8
T2MBackbone=LLaDA2.1-mini (16B MoE), Decoding Strategy=greedy decoding, Temperature=0, block length=32, LOWPROB parameters=τ=0.3, Cmax=1, ρmax=0.252026.0488.25
DeepSeekMath-RLSize=7B, Reasoning Mode=Tool-Integrated, Source Type=Open-Source, Evaluation Protocol=Top12024.0287.6
Qwen2-Math-72Bshot=6-shot, prompting=few-shot chain-of-thought2024.0986.4
GPT-4Size=-, Reasoning Mode=Chain-of-Thought, Source Type=Closed-Source, Evaluation Protocol=Top12024.0286
DeepSeek-LLM-ChatSize=67B, Reasoning Mode=Tool-Integrated, Source Type=Open-Source, Evaluation Protocol=Top12024.0285.4
Qwen2-72Bshot=6-shot, prompting=few-shot chain-of-thought2024.0985.4
Qwen2.5-Math-7Bshot=6-shot, prompting=few-shot chain-of-thought2024.0985
DeepSeekMath-InstructSize=7B, Reasoning Mode=Chain-of-Thought, Source Type=Open-Source, Evaluation Protocol=Top12024.0284.6
DeepSeekMath-InstructSize=7B, Reasoning Mode=Tool-Integrated, Source Type=Open-Source, Evaluation Protocol=Top12024.0284.3
Qwen2-Math-7Bshot=6-shot, prompting=few-shot chain-of-thought2024.0983.2
Qwen2.5-Math-1.5Bshot=6-shot, prompting=few-shot chain-of-thought2024.0983
T2TBackbone=LLaDA2.1-mini (16B MoE), Decoding Strategy=greedy decoding, Temperature=0, block length=322026.0482.33
DeepSeek-LLM-ChatSize=67B, Reasoning Mode=Chain-of-Thought, Source Type=Open-Source, Evaluation Protocol=Top12024.0280.3
Qwen2-Math-1.5Bshot=6-shot, prompting=few-shot chain-of-thought2024.0979.6
DeepSeek-Coder-V2-Lite-Baseshot=6-shot, prompting=few-shot chain-of-thought2024.0977.8
Qwen2-7Bshot=6-shot, prompting=few-shot chain-of-thought2024.0976.7
Llama-3.1-70Bshot=6-shot, prompting=few-shot chain-of-thought2024.0975.5
GPT-3.5Size=-, Reasoning Mode=Chain-of-Thought, Source Type=Closed-Source, Evaluation Protocol=Top12024.0273.8
DeepSeekMath-Base-7Bshot=6-shot, prompting=few-shot chain-of-thought2024.0971.7
MetaMathSize=70B, Reasoning Mode=Chain-of-Thought, Source Type=Open-Source, Evaluation Protocol=Majority Vote (32 candidates)2024.0270.9
InternLM2-Math-Base-20Bshot=6-shot, prompting=few-shot chain-of-thought2024.0965.9
WizardMath-v1.0Size=70B, Reasoning Mode=Chain-of-Thought, Source Type=Open-Source, Evaluation Protocol=Majority Vote (32 candidates)2024.0265.4
Qwen2-1.5Bshot=6-shot, prompting=few-shot chain-of-thought2024.0955.6
ToRASize=34B, Reasoning Mode=Tool-Integrated, Source Type=Open-Source, Evaluation Protocol=Top12024.0253.4
Llama-3.1-8Bshot=6-shot, prompting=few-shot chain-of-thought2024.0951.5