Less Is More -- Until It Breaks: Security Pitfalls of Vision Token Compression in Large Vision-Language Models

Documenti analoghi

• Large Language Model Watermark Stealing With Mixed Integer Programming
  di: Zhang, Zhaoxi, et al.
  Pubblicazione: (2024)
• Character-Level Perturbations Disrupt LLM Watermarks
  di: Zhang, Zhaoxi, et al.
  Pubblicazione: (2025)
• Masked Language Model Based Textual Adversarial Example Detection
  di: Zhang, Xiaomei, et al.
  Pubblicazione: (2023)
• The Security Threat of Compressed Projectors in Large Vision-Language Models
  di: Zhang, Yudong, et al.
  Pubblicazione: (2025)
• On the Adversarial Robustness of Large Vision-Language Models under Visual Token Compression
  di: Zhang, Xinwei, et al.
  Pubblicazione: (2026)