Seeing Red, Thinking Bad: Color Bias in Vision Language Models
Vision language models (VLMs) are increasingly used in industrial decision-making systems, such as recruitment support and recommendation. This motivates careful analysis of how VLMs process visual and textual information. In this work, we study how VLMs interpret text rendered as an image, and investigate the influence of visual styling biases. To this end, we introduce Stealth Visual Prompts, which subtly change visual styling of text, such as color and contrast, while preserving semantic content. Using these prompts, we systematically control the visual styling of words in text and measure
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- FuzzySimilar title/name (fuzzy) · 59%VioletVision-3B →
“Fuzzy title match (0.73): “Seeing Red, Thinking Bad: Color Bias in Vision Language Mode” ≈ “VioletVision-3B””
- PossiblePossibly related (embedding) · 47%Understanding large language models demands distinguishing human projection from machine cognition - Nature →
- PossiblePossibly related (embedding) · 47%By modeling visual saliency, AI improves ratings of artistic product designs - Tech Xplore →
- FuzzySimilar title/name (fuzzy) · 84%pytorch/vision →
“Fuzzy title match (0.92): “Seeing Red, Thinking Bad: Color Bias in Vision Language Mode” ≈ “pytorch/vision””
- LinkedLinked via arxiv author · 85%Kohsuke Ide →
“Seeing Red, Thinking Bad: Color Bias in Vision Language Models”
- LinkedLinked via arxiv author · 85%Ryousuke Yamada →
“Seeing Red, Thinking Bad: Color Bias in Vision Language Models”
- LinkedLinked via arxiv author · 85%Yoshihiro Fukuhara →
“Seeing Red, Thinking Bad: Color Bias in Vision Language Models”
- LinkedLinked via arxiv author · 85%Hirokatsu Kataoka →
“Seeing Red, Thinking Bad: Color Bias in Vision Language Models”