可区分删除：统一知识擦除与拒绝用于大语言模型去学习
Distinguishable Deletion: Unifying Knowledge Erasure and Refusal for Large Language Model Unlearning
AI总结：
本文提出D^2方法，通过限制潜在表示中的响应分布来擦除不受欢迎的知识，同时区分保留知识，从而实现安全且一致的拒绝机制，以提高大语言模型去学习的效果。
摘要：
减轻敏感和有害输出对于确保大型语言模型（LLM）的安全部署至关重要。现有方法通常遵循两种范式：知识删除（KD），在训练期间擦除不受欢迎的信息，以及可区分拒绝（DR），在推理期间引导模型远离使用敏感知识。尽管进展迅速，基于KD的去学习在抑制特定令牌序列作为完整知识移除替代物时面临偏见删除的问题，而基于DR的去学习则因底层知识仍然完整而有重新出现有害知识的风险。为了解决这些问题，我们提出了可区分删除（D^2），一种通过限制潜在表示中的响应分布来擦除不受欢迎知识，同时区分保留知识的范式，从而能够安全且一致地处理去学习的输入。为了实现D^2，我们引入了一个能量指数，该指数量化了知识的存在以及去学习内容与保留内容之间的分离。数学和实证分析表明，能量既准确又高效，使能量基于去学习对齐（EUA）能够在训练期间强制执行能量边界去学习，并在推理时应用基于能量的拒绝机制。广泛的实验表明，EUA显著优于先前方法，表明D^2的优越性。我们的代码可在https://github.com/Puning97/EUA-for-LLM-Unlearning获取。
英文摘要：
Mitigating sensitive and harmful outputs is fundamental to ensuring safe deployment of LLMs. Existing approaches typically follow two paradigms: Knowledge Deletion (KD), which erases undesirable information during training, and Distinguishable Refusal (DR), which steers models away from using sensitive knowledge during inference. Despite rapid progress, KD-based unlearning struggles with biased deletion due to suppressing specific token sequences as a substitute for complete knowledge removal, whereas DR-based unlearning risks the re-emergence of harmful knowledge because the underlying knowledge remains intact. To address these issues, we propose Distinguishable Deletion ($\mathrm{D^2}$), a paradigm that restricts the response distribution in the latent representation rather than specific tokens to erase undesirable knowledge, while distinguishing it from retained knowledge, enabling a refusal mechanism to handle unlearned inputs safely and coherently. To implement $\mathrm{D^2}$, we introduce an energy index that quantifies the presence of knowledge and the separation between unlearned and retained content. Mathematical and empirical analyses show that energy is both accurate and efficient, enabling Energy-based Unlearning Alignment (EUA) to enforce energy-boundary unlearning during training and apply an energy-based refusal mechanism at inference. Extensive experiments demonstrate that EUA significantly outperforms previous methods, indicating the superiority of $\mathrm{D^2}$. Our code is available at https://github.com/Puning97/EUA-for-LLM-Unlearning.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
