解决大模型遗忘中难易样本不平衡问题，实现更均衡的删除效果。
BalDRO: A Distributionally Robust Optimization based Framework for Large Language Model Unlearning
- 采用分布鲁棒优化框架，动态识别最难遗忘的样本并强化训练。
- 在TOFU和MUSE数据集上显著提升遗忘质量与模型可用性。
- 提供两种高效变体，兼容主流训练流程，适合实际部署使用。
随着大语言模型（LLMs）日益影响网络内容生成，从已训练好的模型中移除特定信息（即模型遗忘）成为网络治理的关键。核心挑战在于遗忘集合中样本级别的不平衡：不同样本遗忘难度差异大，导致遗忘进程不一致，部分知识未充分清除而另一些则被过度删除。为此，我们提出BalDRO，一种新颖高效的平衡式大模型遗忘框架。BalDRO将遗忘建模为一个最小-最大过程：内层步骤识别出强调难遗忘样本的最差情况数据分布，外层步骤在此分布下更新模型参数。我们通过两种高效变体实现该框架：BalDRO-G，基于离散GroupDRO近似，聚焦高损失子集；BalDRO-DV，基于连续Donsker-Varadhan对偶方法，支持在标准训练流水线中平滑自适应加权。在TOFU和MUSE数据集上的实验表明，BalDRO显著优于现有方法，在遗忘质量与模型实用性方面均有提升，并已开源代码以供复现。
原文摘要 · Abstract (English)
As Large Language Models (LLMs) increasingly shape online content, removing targeted information from well-trained LLMs (also known as LLM unlearning) has become critical for web governance. A key challenge lies in sample-wise imbalance within the forget set: different samples exhibit widely varying unlearning difficulty, leading to asynchronous forgetting where some knowledge remains insufficiently erased while others become over-forgotten. To address this, we propose BalDRO, a novel and efficient framework for balanced LLM unlearning. BalDRO formulates unlearning as a min-sup process: an inner step identifies a worst-case data distribution that emphasizes hard-to-unlearn samples, while an outer step updates model parameters under this distribution. We instantiate BalDRO via two efficient variants: BalDRO-G, a discrete GroupDRO-based approximation focusing on high-loss subsets, and BalDRO-DV, a continuous Donsker-Varadhan dual method enabling smooth adaptive weighting within standard training pipelines. Experiments on TOFU and MUSE show that BalDRO significantly improves both forgetting quality and model utility over existing methods, and we release code for reproducibility.
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。