Safe Delta: Consistently Preserving Safety when Fine-Tuning LLMs on Diverse Datasets

Documenti analoghi

• Less is More: Understanding Word-level Textual Adversarial Attack via n-gram Frequency Descend
  di: Lu, Ning, et al.
  Pubblicazione: (2023)
• Benign Fine-Tuning Breaks Safety Alignment in Audio LLMs
  di: Roh, Jaechul, et al.
  Pubblicazione: (2026)
• Evolving Jailbreaks: Automated Multi-Objective Long-Tail Attacks on Large Language Models
  di: Hong, Wenjing, et al.
  Pubblicazione: (2026)
• Backdoor Graph Condensation
  di: Wu, Jiahao, et al.
  Pubblicazione: (2024)
• FedRLHF: A Convergence-Guaranteed Federated Framework for Privacy-Preserving and Personalized RLHF
  di: Fan, Flint Xiaofeng, et al.
  Pubblicazione: (2024)