Safety at One Shot: Patching Fine-Tuned LLMs with A Single Instance

Documenti analoghi

• Understanding and Preserving Safety in Fine-Tuned LLMs
  di: Zhang, Jiawen, et al.
  Pubblicazione: (2026)
• Q-resafe: Assessing Safety Risks and Quantization-aware Safety Patching for Quantized Large Language Models
  di: Chen, Kejia, et al.
  Pubblicazione: (2025)
• Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense
  di: Zhang, Jiawen, et al.
  Pubblicazione: (2025)
• Mitigating Many-shot Jailbreak Attacks with One Single Demonstration
  di: Chen, Kejia, et al.
  Pubblicazione: (2026)
• Confidence-Aware Alignment Makes Reasoning LLMs More Reliable
  di: Chen, Kejia, et al.
  Pubblicazione: (2026)