Math Reasoning on Odyssey MATH
53.7AccuracyDeepSeek-Coder-V2-Instruct
Evaluation results
| Method | Links | |
|---|---|---|
| DeepSeek-Coder-V2-Instructsize=236B, open=true2024.06 | 53.7 | |
| GPT-4o-0513open=false2024.06 | 53.2 | |
| Qwen2-72B-Instruct + Step-DPOsize=72B, open=true2024.06 | 50.1 | |
| GPT-4-1106open=false2024.06 | 49.1 | |
| Qwen2-72B-SFT + Step-DPOsize=72B, open=true, Training Data=299K SFT2024.06 | 47 | |
| Qwen2-72B-Instructsize=72B, open=true2024.06 | 47 | |
| GPT-4-Turbo-0409open=false2024.06 | 46.8 | |
| Gemini-1.5-Proopen=false2024.06 | 45 | |
| DeepSeek-Coder-V2-Lite-Instructsize=16B, open=true2024.06 | 44.4 | |
| Qwen2-72B-SFTsize=72B, open=true, Training Data=299K SFT2024.06 | 44.2 | |
| Claude-3-Opusopen=false2024.06 | 40.6 | |
| Llama-3-70B-Instructsize=70B, open=true2024.06 | 27.9 |