Alleviating the Fear of Losing Alignment in LLM Fine-tuning
2025 IEEE Symposium on Security and Privacy (SP)（2025）
关键词
Large Language Models,Task Performance,Gradient Descent,Model Weights,Direct Alignment,Fine-tuned Model,Hidden Layer,Time Cost,Hidden State,Number Of Weights,Human Values,Target Model,Training Subsets,Local Layer,Human Preferences,Fine-tuning Process,Structured Query Language,Fast Gradient Sign Method
AI 理解论文
溯源树
样例
生成溯源树，研究论文发展脉络
Chat Paper
正在生成论文摘要