Discovering Decoupled Functional Modules in Large Language Models
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (41) · 34503-34511
DOI: 10.1609/aaai.v40i41.40749 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Yanke Yu und 5 weiteren Autor:innen.
Abstract (Originalsprache)
Understanding the internal functional organization of Large Language Models (LLMs) is crucial for improving their trustworthiness and performance. However, how LLMs organize different functions into modules remains highly unexplored. To bridge this gap, we formulate a function module discovery problem and propose an Unsupervised LLM Cross-layer MOdule Discovery (ULCMOD) framework that simultaneously disentangles the large set of neurons in the entire LLM into modules while discovering the topics of input samples related to these modules. Our framework introduces a novel objective function and an efficient Iterative Decoupling (IterD) algorithm. Extensive experiments show that our method discovers high-quality, disentangled modules that capture more meaningful semantic information and achieve superior performance in various downstream tasks. Moreover, our qualitative analysis reveals that the discovered modules show function comprehensiveness, function hierarchy, and clear function spatial arrangement within LLMs. Our work provides a novel tool for interpreting LLMs' function modules, filling a critical gap in LLMs' interpretability research.
Zitieren
APA
Yu, Y., Li, J., Sun, Y., Li, P., Wang, Z., & Zheng, Y. (2026). Discovering Decoupled Functional Modules in Large Language Models. Proceedings of the AAAI Conference on Artificial Intelligence, 40(41), 34503-34511. https://doi.org/10.1609/aaai.v40i41.40749
Vancouver
Yu Y, Li J, Sun Y, Li P, Wang Z, Zheng Y. Discovering Decoupled Functional Modules in Large Language Models. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(41):34503-34511. doi:10.1609/aaai.v40i41.40749
BibTeX
@article{Yu2026,
author = {Yu, Yanke and Li, Jin and Sun, Ying and Li, Ping and Wang, Zhefeng and Zheng, Yi},
title = {Discovering Decoupled Functional Modules in Large Language Models},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {41},
pages = {34503--34511},
doi = {10.1609/aaai.v40i41.40749}
}
EUVIMED