1
Divergence Decoding: Inference-Time Unlearning via Auxiliary Models
提出推理时遗忘新方法，用辅助模型实现无需重新训练的知识擦除，效率与安全性兼顾。
arXiv:2605.31293v1 Announce Type: new Abstract: Large Language Models (LLMs) frequently memorize sensitive training data thereby creating significant …