Mathematical Reasoning on IMO-Bench 400 (pass@1, pass@5)
4.88Pass@1DeepSeek-R1-Qwen-1.5B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DeepSeek-R1-Qwen-1.5BGroup Size (G)=162026.06 | 4.88 | 14.38 | |
| DeepSeek-R1-Qwen-1.5BGroup Size (G)=42026.06 | 4.67 | 13.38 | |
| DeepSeek-R1-Qwen-1.5BGroup Size (G)=Base2026.06 | 3.75 | 10.75 | |
| Nemotron-4BGroup Size (G)=42026.06 | 2.6 | 8.88 | |
| Nemotron-4BGroup Size (G)=Base2026.06 | 2.35 | 9 | |
| DeepSeek-7BGroup Size (G)=162026.06 | 2.23 | 7.88 | |
| DeepSeek-7BGroup Size (G)=42026.06 | 2.2 | 8.25 | |
| Nemotron-4BGroup Size (G)=162026.06 | 2.17 | 9 | |
| DeepSeek-7BGroup Size (G)=Base2026.06 | 2 | 7.5 |