List your company

Multimodal Understanding on MMMU (Corruption Metrics: Noise, Blur, Weather, Digital)

37.3Accuracy (Noise)Latent Denoising
35.42835.91436.436.886Apr 23, 2026

Evaluation results

MethodLinks
Latent DenoisingBackbone=LLaVA+CLIP, Corruption Severity=52026.0437.335.536.736.9
BaselineBackbone=LLaVA+CLIP, Corruption Severity=52026.0435.534.633.635.2