通过语义嵌入空间实现大模型知识删除，更彻底且抗提示重述。
Align-then-Unlearn: Embedding Alignment for LLM Unlearning
- 在语义嵌入空间中进行知识删除，而非直接修改输出词元。
- 微调模型使预测嵌入与目标概念嵌入相似度最低，实现有效遗忘。
- 适合关注模型隐私保护与可控知识管理的研究者与开发者。
大型语言模型（LLMs）在海量数据上训练，可能意外保留敏感信息，引发隐私与伦理问题。无学习旨在选择性移除特定数据，如个人信息或受版权保护内容。现有方法针对输出词元层面的特定序列，常无法完全遗忘，且易受提示重述影响。本文提出Align-then-Unlearn框架，将无学习过程置于语义嵌入空间，而非直接作用于输出词元。该方法首先为LLM增加一个嵌入预测模块，用于预测未来上下文表示；随后通过微调模型，使其预测嵌入与代表待删除概念的目标嵌入的相似度最小化，从而实现无学习。初步结果表明，该方法能有效移除目标知识，同时对模型整体性能影响较小。研究结果表明，基于嵌入的无学习为概念性知识删除提供了有前景且鲁棒的新路径。代码已开源：https://github.com/ExplainableML/align-then-unlearn。
原文摘要 · Abstract (English)
As large language models (LLMs) are trained on massive datasets, they have raised significant privacy and ethical concerns due to their potential to inadvertently retain sensitive information. Unlearning seeks to selectively remove specific data from trained models, such as personal information or copyrighted content. Current approaches targeting specific output sequences at the token level often fail to achieve complete forgetting and remain susceptible to prompt rephrasing. We propose Align-then-Unlearn, a novel framework that performs unlearning in the semantic embedding space rather than directly on output tokens. Align-then-Unlearn first augments the LLM with an embedding prediction module trained to anticipate future context representations. Unlearning is then achieved by fine-tuning the model to minimize the similarity between these predicted embeddings and a target embedding that represents the concept to be removed. Initial results show that Align-then-Unlearn effectively removes targeted knowledge with minimal degradation in overall model utility. These findings suggest that embedding-based unlearning offers a promising and robust approach to removing conceptual knowledge. Our code is available at https://github.com/ExplainableML/align-then-unlearn.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。