arXiv cs.CL· Tyler Lizzo, Larry Heck·· 2 天前AI 评分26
行为遗忘不等于隐私：机器遗忘评估的跨语言漏洞
Behavioral Forgetting Is Not Privacy in Machine Unlearning
AI 导读
研究指出，机器遗忘评估仅衡量指定基准分布上的行为遗忘，不足以代表隐私保护效果。跨语言查询案例显示，多个多语言模型家族在查询语义等价但语言和文字不同的信息时，遗忘程度差异显著。单一访问路径即可暴露行为遗忘与信息不可获取之间的真实差距，需将访问路径泛化与对抗性可恢复性纳入评估。
正文 · 原文
Abstract:Machine unlearning aims to remove targeted information from large language models, but existing evaluations often measure forgetting only on a designated benchmark distribution. We argue that such behavioral forgetting is an insufficient standard for privacy-oriented unlearning, and that evaluation should instead test whether forgetting persists across the access paths through which the same information may be recovered. We demonstrate this with a controlled case study of cross-lingual querying: across multiple multilingual model families, forgetting varies substantially when semantically equivalent information is queried in different languages and scripts. A single access path is therefore enough to expose a real gap between behavioral forgetting and information inaccessibility, motivating broader evaluation of access-path generalization and adversarial recoverability alongside in-distribution forgetting.
| Comments: | Accepted as a poster at the PriLOM Workshop at NeurIPS 2026 |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.21799 [cs.CL] |
| (or arXiv:2610.21799v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.21799 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Tyler Lizzo [view email]
[v1]
Thu, 15 Oct 2026 12:48:16 UTC (66 KB)
来源：arXiv cs.CL · arxiv.org