arXiv：cs.CL· Salima Lamsiyah, Adelaide Danilov, Oleksandr Marchenko Breneur, Aria Nourbakhsh·· 10 小时前
稀疏自编码器在 NLP 机制可解释性中的批判性综述：可恢复性、可解释性与实用性
Sparse Autoencoders for Mechanistic Interpretability in NLP: A Critical Survey of Recoverability, Interpretability, and Utility
AI 导读
一篇涵盖 138 项研究的批判性综述系统梳理了稀疏自编码器（SAE）在 NLP 机制可解释性中的表现，并提出连接字典设计、特征操作与 NLP 功能的任务导向分类体系。综述考察了稀疏度分配、几何结构、数据覆盖与推理对特征可恢复性的影响，并评估了语义可解释性、机制忠实性与行为实用性的证据。研究还综合了多语言处理、检索、推理、分类与行为调控等方向的发现，指出待解问题并提出更严格评估的方向。
正文
Abstract:Sparse autoencoders (SAEs) have emerged as a promising approach to mechanistic interpretability by decomposing language-model activations into sparse, potentially interpretable features. However, differences in design assumptions, evaluation objectives, and experimental settings complicate conclusions about their effectiveness. We present a critical survey of 138 studies on SAEs in natural language processing. Our contributions are fourfold: (1) we propose a task-grounded taxonomy connecting dictionary design, feature operations, and NLP functions; (2) we examine how sparsity allocation, geometry, data coverage, and inference affect feature recoverability; (3) we assess evidence for semantic interpretability, mechanistic faithfulness, and behavioral utility; and (4) we synthesize findings across multilingual processing, retrieval, reasoning, classification, and behavioral regulation, highlighting the roles of objectives, baselines, supervision, and intervention pipelines. Through comparative analyses of published results, we reconcile apparent disagreements, identify unresolved questions, and propose directions for more rigorous evaluation. An anonymous research register this https URL accompanies the survey.
| Comments: | 28 pages, 5 figures |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.20139 [cs.CL] |
| (or arXiv:2610.20139v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.20139 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Adelaide Danilov [view email]
[v1]
Wed, 14 Oct 2026 12:56:00 UTC (67 KB)
来源：arXiv：cs.CL · arxiv.org