提出MUDMAN方法，让大模型彻底遗忘危险知识，防止被恢复。
Robust LLM Unlearning with MUDMAN: Meta-Unlearning with Disruption Masking And Normalization
- 只更新梯度方向一致的权重，避免破坏性修改
- 通过梯度归一化和元学习提升遗忘效果，比前人方法高40%
- 适合需要安全可控的AI系统开发者使用
语言模型在经过大量安全微调后仍可能保留危险知识与能力，带来滥用和对齐风险。近期研究发现，即使采用专门的遗忘方法也容易被逆转。为此，我们系统评估了多种现有及新提出的遗忘组件，识别出实现不可逆遗忘的关键要素。提出干扰掩蔽（Disruption Masking）技术：仅允许在未遗忘梯度与保留梯度符号相同的权重上进行更新，确保所有更新均为非破坏性。同时，识别出对遗忘梯度进行归一化的必要性，并验证了元学习的有效性。将这些洞察整合为MUDMAN（Meta-Unlearning with Disruption Masking and Normalization），在防止危险能力恢复方面表现优异，相比之前最先进的TAR方法提升40%，达到当前最 robust 的遗忘水平。
原文摘要 · Abstract (English)
Language models can retain dangerous knowledge and skills even after extensive safety fine-tuning, posing both misuse and misalignment risks. Recent studies show that even specialized unlearning methods can be easily reversed. To address this, we systematically evaluate many existing and novel components of unlearning methods and identify ones crucial for irreversible unlearning. We introduce Disruption Masking, a technique in which we only allow updating weights, where the signs of the unlearning gradient and the retaining gradient are the same. This ensures all updates are non-disruptive. Additionally, we identify the need for normalizing the unlearning gradients, and also confirm the usefulness of meta-learning. We combine these insights into MUDMAN (Meta-Unlearning with Disruption Masking and Normalization) and validate its effectiveness at preventing the recovery of dangerous capabilities. MUDMAN outperforms the prior TAR method by 40%, setting a new state-of-the-art for robust unlearning.
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。