作者位次 #3/7ICML 2026Spotlight· Deep Learning / Large Language Models
Surgery: Mitigating Harmful Fine-Tuning for Large Language Models via Attention Sink
Guozhi Liu · Weiwei Lin · Tiansheng Huang · Ruichao Mo · Qi Mu · Xiumin Wang · Li Shen
首次获取该候选人的完整档案及 StartLux 评估共 20 积分；每位用户单独解锁，之后查看免费。主动刷新 20 积分。
实时查找电子邮箱，费用不超过20积分；未找到邮箱不扣费。
Guozhi Liu · Weiwei Lin · Tiansheng Huang · Ruichao Mo · Qi Mu · Xiumin Wang · Li Shen
Biao Yi · Tiansheng Huang · Baolei Zhang · Tong Li · Lihai Nie · Zheli Liu · Li Shen
Fine-tuning-as-a-service, while commercially successful for Large Language Model (LLM) providers, exposes models to harmful fine-tuning attacks. As a widely explored defense paradigm against such attacks, unlearning attempts to remove malicious knowledge from LLMs, thereby essentially preventing...