The Trojan Example: Jailbreaking LLMs through Template Filling and Unsafety Reasoning

Documenti analoghi

• RedVisor: Reasoning-Aware Prompt Injection Defense via Zero-Copy KV Cache Reuse
  di: Liu, Mingrui, et al.
  Pubblicazione: (2026)
• Mask-based Membership Inference Attacks for Retrieval-Augmented Generation
  di: Liu, Mingrui, et al.
  Pubblicazione: (2024)
• Wukong Framework for Not Safe For Work Detection in Text-to-Image systems
  di: Liu, Mingrui, et al.
  Pubblicazione: (2025)
• TASO: Jailbreak LLMs via Alternative Template and Suffix Optimization
  di: Wang, Yanting, et al.
  Pubblicazione: (2025)
• TrojanPraise: Jailbreak LLMs via Benign Fine-Tuning
  di: Xie, Zhixin, et al.
  Pubblicazione: (2026)