Math generation on AIME
33.07Pass@10Qwen2.5-VL-72B-Ins
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen2.5-VL-72B-Ins2025.09 | 33.07 | — | 41.39 | — | |
| GRPO-MATN=4, AN=4, S/S=132.872025.09 | 14.7 | 3.4 | 27.6 | 5 | |
| GRPOTN=16, AN=1, S/S=225.432025.09 | 12.89 | 2.93 | 21.72 | 15 | |
| GRPOTN=4, AN=1, S/S=111.242025.09 | 11.78 | 2.07 | 20.32 | 5 | |
| GRPOTN=8, AN=1, S/S=140.052025.09 | 11.16 | 2.2 | 21.3 | 13 | |
| SFT2025.09 | 11.07 | — | 18.11 | — | |
| Qwen2.5-VL-7B-Ins2025.09 | 9.97 | — | 18.39 | — | |
| Qwen2.5-VL-3B-Ins2025.09 | 9.27 | — | 16.25 | — |