1
Optimus: A Robust Defense Framework for Mitigating Toxicity while Fine-Tuning Conversational AI
一项针对LLM微调毒性风险的新型防御框架Optimus，在保持对话效用的同时有效减轻有害行为。
arXiv:2507.05660v3 Announce Type: replace-cross Abstract: Customizing Large Language Models (LLMs) on untrusted datasets poses severe risks of injecti…