List your company

Visual Question Answering on TextVQA (Corruption Robustness Metrics)

36.5Accuracy (Noise Corruption)Latent Denoising
34.00434.65235.335.948Apr 23, 2026

Evaluation results

MethodLinks
Latent DenoisingArchitecture=Qwen-2.5-VL, Severity=52026.0436.535.97360.5
Latent DenoisingBackbone=LLaVA+CLIP, Corruption Severity=52026.0436.526.540.338.7
BaselineBackbone=LLaVA+CLIP, Corruption Severity=52026.043524.838.137.5
BaselineArchitecture=Qwen-2.5-VL, Severity=52026.0434.133.970.958.9