解耦微调大语言模型中的内部表征变化与因果重要性

Computer Science > Artificial Intelligence arXiv:2609.21113 (cs) [Submitted on 17 Sep 2026 (v1), last revised 23 Sep 2026 (this version, v2)] Title:Decoupling Internal Representational Changes and Causal Importance in Fine-Tuned Large Language Models Authors:Lingfang Li, Procheta Sen, Shubham Das, Danushka Bollegala View a PDF of the paper titled Decoupling Internal Representational Changes and Causal Importance in Fine-Tuned Large Language Models, by Lingfang Li and 3 other authors View PDF HTML (experimental) Abstract:Fine-tuning has emerged as a widely adopted approach for adapting LLMs to a variety of downstream tasks. However, how it reshapes their internal mechanisms remains poorly understood. To address this, we investigate how fine-tuning alters internal representations in LLMs, including attention patterns and layer-wise activations, and examine whether these changes are linked to task-relevant components identified by EAP (e.g., attention heads and logit-level activations) that drive task performance. We find that EAP-identified components are concentrated within specific layers, indicating a degree of functional localisation in how models internalise task-specific behavior. Notably, the distribution of these components across layers is largely uncorrelated with the layers undergoing the most substantial representational changes during fine-tuning. Furthermore, we observe that overlap in EAP-identified components across tasks does not translate into cross-task performance transfer if the tasks are different in nature (e.g. classification vs. generative tasks). More specifically, fine-tuning on one task can lead to a degradation of performance on another when the two tasks exhibit a high degree of overlap in their EAP-identified components. Comments: Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.21113 [cs.AI] (or arXiv:2609.21113v2 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.21113 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Lingfang Li [view email] [v1] Thu, 17 Sep 2026 21:55:20 UTC (1,793 KB) [v2] Wed, 23 Sep 2026 08:52:42 UTC (1,733 KB) Full-text links: Access Paper: View a PDF of the paper titled Decoupling Internal Representational Changes and Causal Importance in Fine-Tuned Large Language Models, by Lingfang Li and 3 other authors View PDF HTML (experimental) TeX Source view license Current browse context: cs.AI new | recent | 2026-09 Change to browse by: cs References & Citations NASA ADS Google Scholar Semantic Scholar export BibTeX citation Loading... BibTeX formatted citation × loading... Data provided by: Bookmark Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)