Multimodal Robustness on MMStar (test)
69.8MMStar ScoreQwen2.5-VL
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen2.5-VLParams=32B2026.01 | 69.8 | |
| GPT-4oParams=-2026.01 | 63.9 | |
| LVRParams=7B2026.01 | 59.4 | |
| Qwen2.5-VLParams=7B2026.01 | 58.9 | |
| Naive SFTParams=3B2026.01 | 55.53 | |
| LaViTParams=3B2026.01 | 54.07 | |
| LVR_RLParams=3B2026.01 | 53.73 | |
| PAPOParams=3B2026.01 | 52.7 | |
| R1-OneVisionParams=7B2026.01 | 52.1 | |
| DMLRParams=3B2026.01 | 51.2 | |
| Qwen2.5-VL (Baseline)Params=3B2026.01 | 50.2 |