Multi-discipline Multi-modal Understanding on MMMU (Robustness Evaluation)
50Accuracy (Noise)Latent Denoising
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Latent DenoisingArchitecture=Qwen-2.5-VL, Severity=52026.04 | 50 | 46.7 | 51 | 48.7 | |
| BaselineArchitecture=Qwen-2.5-VL, Severity=52026.04 | 45.5 | 43.7 | 47.1 | 46.4 |