通过动态路由机制，让稀疏自编码器跨层提取更丰富、可解释的特征。
Route Sparse Autoencoder to Interpret Large Language Models
- 引入路由机制，共享稀疏自编码器跨多层提取特征。
- 在64稀疏度下，比基线多提取22.5%特征，可解释性提升22.3%。
- 适合做模型内部机制分析与可控干预，代码已开源。
大语言模型的机械可解释性旨在揭示信息传播与推理的内部过程。稀疏自编码器（SAEs）在提取可解释的单义特征方面展现出潜力。然而，以往研究主要聚焦于单层特征提取，难以有效捕捉跨层激活。本文提出路由稀疏自编码器（RouteSAE），将路由机制与共享SAE结合，动态分配不同层激活的权重，在参数开销极小的情况下实现高可解释性与灵活性，支持目标特征操控。我们在Llama-3.2-1B-Instruct上进行了大量实验，结果表明，在相同64稀疏度约束下，RouteSAE比基线SAE多提取22.5%特征，可解释性得分提升22.3%。这些结果证明了RouteSAE在大语言模型可解释性方面的可扩展性与有效性，适用于特征发现与模型干预。代码已公开于https://github.com/swei2001/RouteSAEs。
原文摘要 · Abstract (English)
Mechanistic interpretability of large language models (LLMs) aims to uncover the internal processes of information propagation and reasoning. Sparse autoencoders (SAEs) have demonstrated promise in this domain by extracting interpretable and monosemantic features. However, prior works primarily focus on feature extraction from a single layer, failing to effectively capture activations that span multiple layers. In this paper, we introduce Route Sparse Autoencoder (RouteSAE), a new framework that integrates a routing mechanism with a shared SAE to efficiently extract features from multiple layers. It dynamically assigns weights to activations from different layers, incurring minimal parameter overhead while achieving high interpretability and flexibility for targeted feature manipulation. We evaluate RouteSAE through extensive experiments on Llama-3.2-1B-Instruct. Specifically, under the same sparsity constraint of 64, RouteSAE extracts 22.5% more features than baseline SAEs while achieving a 22.3% higher interpretability score. These results underscore the potential of RouteSAE as a scalable and effective method for LLM interpretability, with applications in feature discovery and model intervention. Our codes are available at https://github.com/swei2001/RouteSAEs.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。