Sparse Feature Coactivation Reveals Causal Semantic Modules in Large Language Models
Ruixuan Deng, Xiaoyang Hu Brown University, Miles Gilberti University of Michigan - Ann Arbor, Shane Storks University of Michigan - Ann Arbor, Aman Taxali University of Michigan - Ann Arbor, Mike Angstadt University of Michigan - Ann Arbor, Chandra Sripada University of Michigan, Joyce Chai
We identify semantically coherent, context-consistent network components in large language models (LLMs) using coactivation of sparse autoencoder (SAE) features collected from just a handful of prompts. Focusing on concept-relation prediction tasks, we show that ablating these components for concepts (e.g., countries and words) and relations (e.g., capital city and translation language) changes model outputs in predictable ways, while amplifying these components induces counterfactual responses. Notably, composing relation and concept components yields compound counterfactual outputs. Further analysis reveals that while most concept components emerge from the very first layer, more abstract relation components are concentrated in later layers. Lastly, we show that extracted components more comprehensively capture concepts and relations than individual features while maintaining specificity. Overall, our findings suggest a modular organization of knowledge and advance methods for efficient, targeted LLM manipulation.
尚无 AI 解读，展示原文摘要
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表