Is poisoning a real threat to LLM alignment? Maybe more so than you think

Documenti analoghi

• PoisonedParrot: Subtle Data Poisoning Attacks to Elicit Copyright-Infringing Content from Large Language Models
  di: Panaitescu-Liess, Michael-Andrei, et al.
  Pubblicazione: (2025)
• The 2020 US Decennial Census is more private than you (might) think
  di: Su, Buxin, et al.
  Pubblicazione: (2024)
• Reward Models Can Improve Themselves: Reward-Guided Adversarial Failure Mode Discovery for Robust Reward Modeling
  di: Pathmanathan, Pankayaraj, et al.
  Pubblicazione: (2025)
• LLM Dataset Inference: Did you train on my dataset?
  di: Maini, Pratyush, et al.
  Pubblicazione: (2024)
• Cross-Modal Safety Alignment: Is textual unlearning all you need?
  di: Chakraborty, Trishna, et al.
  Pubblicazione: (2024)