How Do Inpainting Artifacts Propagate to Language?

TL;DR

This paper investigates how artifacts from diffusion-based image inpainting influence language generation in vision-language models, revealing that reconstruction quality impacts captioning performance and model behavior.

Contribution

It introduces a diagnostic framework to analyze the effect of inpainting artifacts on downstream language tasks in multimodal models.

Findings

Reconstruction fidelity correlates with caption quality.

Inpainting artifacts cause layer-dependent changes in model behavior.

Visual artifacts systematically influence language generation.

Abstract

We study how visual artifacts introduced by diffusion-based inpainting affect language generation in vision-language models. We use a two-stage diagnostic setup in which masked image regions are reconstructed and then provided to captioning models, enabling controlled comparisons between captions generated from original and reconstructed inputs. Across multiple datasets, we analyze the relationship between reconstruction fidelity and downstream caption quality. We observe consistent associations between pixel-level and perceptual reconstruction metrics and both lexical and semantic captioning performance. Additional analysis of intermediate visual representations and attention patterns shows that inpainting artifacts lead to systematic, layer-dependent changes in model behavior. Together, these results provide a practical diagnostic framework for examining how visual reconstruction quality influences language generation in multimodal systems.