One Perturbation, Two Failure Modes: Probing VLM Safety via Embedding-Guided Typographic Perturbations

Documenti analoghi

• Reading Between the Pixels: Linking Text-Image Embedding Alignment to Typographic Attack Success on Vision-Language Models
  di: Balakrishnan, Ravikumar, et al.
  Pubblicazione: (2026)
• VISOR++: Universal Visual Inputs based Steering for Large Vision Language Models
  di: Balakrishnan, Ravikumar, et al.
  Pubblicazione: (2025)
• VISOR: Visual Input-based Steering for Output Redirection in Vision-Language Models
  di: Phute, Mansi, et al.
  Pubblicazione: (2025)
• Unveiling Typographic Deceptions: Insights of the Typographic Vulnerability in Large Vision-Language Model
  di: Cheng, Hao, et al.
  Pubblicazione: (2024)
• More Images, More Problems? A Controlled Analysis of VLM Failure Modes
  di: Das, Anurag, et al.
  Pubblicazione: (2026)