arXiv cs.CL· Ziye Mei, Haotian Ma·· 2 天前AI 评分24
稀疏自编码器中的表征重分配：RSS 分数如何识别可复现的特征变化
Representation Redistribution in Sparse Autoencoders
AI 导读
研究提出重分配敏感度分数（RSS），用于识别在提示词变换下方向一致且可复现的大幅特征响应。在覆盖 8 个语言模型、27 个预训练 SAE 上，RSS 支持集在未参与选择的变体上呈现可复现的重分配特征。在两个冻结安全读出中，RSS 支持侵蚀结合完整干净边际，以中位数 20 和 19 个特征分别取得 0.935 和 0.958 的复现失败 AUROC，低于完整侵蚀的 0.965 和 0.964。
正文 · 原文
Abstract:Sparse autoencoders (SAEs) decompose language model activations into sparse features, but prompt transformations designed to preserve the underlying request can redistribute activity across them. A feature change observed on one transformed prompt need not recur, and even a recurring change need not affect a downstream computation. We introduce the redistribution sensitivity score (RSS), which identifies large feature responses that recur in a consistent direction, and evaluate the resulting supports on variants excluded from selection. Across 27 pretrained SAEs spanning eight language models, the tested prompt transformations produce a reproducible redistribution profile on these excluded variants. We then separate recurring representation change from signed readout contribution and decision failure. For a fixed linear readout, an exact decomposition shows how response magnitude, readout coefficients, directional alignment, and the clean margin determine downstream consequences. In two frozen safety readouts, combining RSS support erosion with the full clean margin yields recurrent failure AUROCs of 0.935 and 0.958 using median supports of 20 and 19 features, compared with 0.965 and 0.964 using full erosion. RSS therefore provides a way to identify recurring sparse representation changes and quantify their contribution to a downstream readout without equating representation change with failure. Code is available at this https URL
| Comments: | 35 pages, 15 figures. Code: this https URL |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.21240 [cs.LG] |
| (or arXiv:2610.21240v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.21240 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Ziye Mei [view email]
[v1]
Thu, 15 Oct 2026 04:37:53 UTC (288 KB)
来源：arXiv cs.CL · arxiv.org