New QR-inspired attack targets infrared vision-language models

Researchers have developed InfraQR, a novel attack method targeting infrared vision-language models. This attack places a QR-inspired structured patch on image boundaries, rather than directly on objects, to disrupt model performance. InfraQR significantly reduces accuracy on classification, captioning, and visual question answering tasks, demonstrating vulnerability even in infrared domains. AI

IMPACT Highlights potential security vulnerabilities in infrared vision-language models, necessitating further research into their robustness.

RANK_REASON The cluster contains a research paper detailing a novel attack method on AI models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• GPT-5.4
• InfraQR
• infrared vision-language models
• OpenAI CLIP

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →