arXiv cs.CL· Yiyang Duan, Fan Wu, Tiantong Wu, Che Wang, Zhiren Gong, Wei Guo, Yang Cao, Wei Yang Bryan Lim·· 13 小时前AI 评分31
C²U：在 RAG 中实现组合式概念遗忘
C$^2$U in RAG: Compositional Concept Unlearning in Retrieval-Augmented Generation
AI 导读
研究提出 Compositional Concept Unlearning（C²U），在编码器与文档索引固定的前提下做查询侧概念抑制，通过增强原子库路由出 profile 方向，仅向原始查询加入有界的“编辑减原始”增量。
正文 · 原文
Abstract:Retrieval-augmented generation can expose unwanted concepts through retrieved evidence even when the generator is moderated. We study query-side concept suppression with a fixed encoder and document index. Compositional Concept Unlearning (C$^2$U) routes a profile-derived direction through an augmented atom bank, adding only a bounded edited-minus-original delta to the raw query. Across five datasets, four encoders, and three seeds, the validation-selected configuration reduces pooled binary target Hit@5 by 4.25 percentage points with non-positive control degradation and target-exposure changes. Under the evaluated query transforms, it suppresses target density without increasing budgeted control-query exposure. Separately, ArXiv four-concept composition improves over an averaged-direction comparator after independent validation gates and first-use test evaluation.
| Comments: | 37 pages, 8 figures, including appendices |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.29223 [cs.CL] |
| (or arXiv:2610.29223v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.29223 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yiyang Duan [view email]
[v1]
Wed, 21 Oct 2026 05:30:06 UTC (6,379 KB)
来源：arXiv cs.CL · arxiv.org