Mathematical Reasoning on ADDSUB

93.1Solve Rate

Qwen2-7B-SFT + Full-Step-DPO

Updated 1mo ago

Evaluation Results

Method	Links
Qwen2-7B-SFT + Full-Step-DPO 2025.02		93.1
Qwen2-7B-SFT 2025.02		92.7
PAL 2022.11		92.5
Qwen2-7B-SFT + Step-DPO 2025.02		92.2
COT PaLM-540B 2022.11		91.9
DIRECT Codex 2022.11		90.9
Zero-shot-CP 2024.03		90.46
Llama-3-8B-SFT + Full-Step-DPO 2025.02		88.8
MetaMath-Mistral-7B + Full-Step-DPO 2025.02		88.6
Llama-3-8B-SFT + Step-DPO 2025.02		88.5
Llama-3-8B-SFT 2025.02		88.4
DeepSeekMath-Base-SFT 2025.02		87.6
MetaMath-Mistral-7B + Step-DPO 2025.02		86.9
MetaMath-Mistral-7B 2025.02		86.6
DeepSeekMath-Base-SFT + DPO 2025.02		86.6
DeepSeekMath-Base-SFT + Full-Step-DPO 2025.02		86.6
Zero-shot-CoT 2024.03		86.16
COT Codex 2022.11		86
DeepSeekMath-Base-SFT + Step-DPO 2025.02		85.3
Zero-shot 2024.03		82.78
Llama-3-8B-SFT + DPO 2025.02		60
MetaMath-Mistral-7B + DPO 2025.02		53.4
COT LAMDA-137B 2022.11		52.9
Qwen2-7B-SFT + DPO 2025.02		25.6
COT UL2-20B 2022.11		18.2