pipette
ESEspañol

CU in RAG: Compositional Concept Unlearning in Retrieval-Augmented Generation

Yiyang Duan, Fan Wu, Tiantong Wu, Che Wang, Zhiren Gong, Wei Guo, Yang Cao, Wei Yang Bryan Lim

Preprint

In the authors' words

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 (CU) 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.

Main resultThe abstract does not state a limitation.

Appeared: Thursday, October 22. arXiv. Preprint, not yet peer-reviewed.

Authors' comment: 37 pages, 8 figures, including appendices