Multi-modal Model Evaluation on MMBench (Robustness Metrics)
57.7Noise Robustness ScoreLatent Denoising
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Latent DenoisingArchitecture=Qwen-2.5-VL, Severity=5, Method Variant=MMB2026.04 | 57.7 | 62.5 | 71.1 | 67.4 | |
| BaselineArchitecture=Qwen-2.5-VL, Severity=5, Method Variant=MMB2026.04 | 54.7 | 60.2 | 69.1 | 63.5 |