VLMShield: Efficient and Robust Defense of Vision-Language Models against Malicious Prompts

Documenti analoghi

• SafeGuider: Robust and Practical Content Safety Control for Text-to-Image Models
  di: Qi, Peigui, et al.
  Pubblicazione: (2025)
• A Validated Prompt Bank for Malicious Code Generation: Separating Executable Weapons from Security Knowledge in 1,554 Consensus-Labeled Prompts
  di: Young, Richard J., et al.
  Pubblicazione: (2026)
• Refusal Evaluation in Coding LLMs and Code Agents: A Systematic Review of Thirteen Malicious-Code Prompt Corpora (2023-2025)
  di: Young, Richard J., et al.
  Pubblicazione: (2026)
• Adaptive Defense Orchestration for RAG: A Sentinel-Strategist Architecture against Multi-Vector Attacks
  di: Pallerla, Pranav, et al.
  Pubblicazione: (2026)
• Whose wife is it anyway? Assessing bias against same-gender relationships in machine translation
  di: Stewart, Ian, et al.
  Pubblicazione: (2024)