arXiv：cs.LG· Vedant Palit, Florent Draye, Nicolas Zucchet, Zhijing Jin, Bernhard Sch\"olkopf·· 7 小时前AI 评分47
当遗忘并非灾难性：论虚假遗忘的机制
When Forgetting is not Catastrophic: On the Mechanics of Spurious Forgetting
AI 导读
语言模型微调时看似遗忘的知识常仍被存储且可恢复，这一现象被称为虚假遗忘。研究者用一个最小联想记忆模型复现了该动态：共享结构的键、集中的新值和网络归一化三者共同作用，微调会将旧表征沿同一方向移动从而掩盖旧事实但保留其相对几何关系，归一化在新事实学会后撤回该偏移，而事实特异的变化则累积造成侵蚀。
正文
Abstract:Knowledge that a language model appears to forget during finetuning often remains stored and can be recovered, a phenomenon called spurious forgetting. Finetuning on new facts can even produce forgetting that undoes itself: recall of the old facts collapses, recovers as training continues on new facts alone, and only then erodes for good. We seek to understand when such forgetting is not catastrophic. A minimal associative memory reproduces these dynamics with three ingredients: keys with shared structure, concentrated new values, and normalization in the network. Finetuning moves all old representations along a common direction, hiding the old facts while preserving their relative geometry; normalization withdraws this shift once the new facts are learned, whereas fact-specific changes accumulate and cause the erosion. Moreover, subtracting the common shift eliminates the collapse in a Transformer trained on synthetic data, and removing a single direction from each weight update restores old facts in a pretrained language model. Forgetting thus combines a shared, reversible loss of access with a slow erosion of individual facts, and only the second is catastrophic. Which one dominates depends on whether the new data move old memories together or apart.
| Subjects: | Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.08718 [cs.CL] |
| (or arXiv:2610.08718v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.08718 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Nicolas Zucchet [view email]
[v1]
Tue, 6 Oct 2026 17:25:48 UTC (741 KB)
来源：arXiv：cs.LG · arxiv.org