利用跨语言不一致性定位 MoE 大语言模型中的知识
Knowledge Localization in Mixture-of-Experts LLMs Using Cross-Lingual Inconsistency
AI 导读
研究者提出一种利用跨语言回答不一致性的知识定位框架,用于分析 MoE 大语言模型:先让模型在多语言下回答难题,划分“成功/失败”激活组,再对 MoE router logits 做对比统计分析,识别关键专家。关闭约 20 个(共 6000 个)专家后,模型在超过 40% 的情况下答错。该方法为复杂 LLM 提供了可扩展的知识定位手段。
正文 · 原文
Abstract:Modern LLMs continue to exhibit significant variance in behavior across languages, such as being able to recall factual information in some languages but not others. While typically studied as a problem to be mitigated, in this work, we propose leveraging this cross-lingual inconsistency as a tool for interpretability in mixture-of-experts (MoE) LLMs. Our knowledge localization framework contrasts routing for sets of languages where the model correctly recalls information from languages where it fails. This allows us to isolate model components that play a functional role in answering about a piece of knowledge. Our method proceeds in two stages: (1) querying the model with difficult factual questions across a diverse set of languages to generate "success" and "failure" activation buckets and then (2) applying a statistical contrastive analysis to the MoE router logits to identify experts important for knowledge. To validate the necessity of this small number of experts for answering a knowledge question, we deactivate them and re-ask the question. We find that despite only deactivating about 20 out of 6000 experts, the model no longer answers correctly in over 40% of cases. Generally, this method provides a realistic and scalable knowledge localization approach to address increasingly complex LLMs.
| Comments: | NeurIPS 2026, this https URL |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2603.17102 [cs.CL] |
| (or arXiv:2603.17102v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2603.17102 arXiv-issued DOI via DataCite |
Submission history
From: Lucas Bandarkar [view email]
[v1]
Tue, 17 Mar 2026 19:48:44 UTC (1,216 KB)
[v2]
Fri, 6 Nov 2026 16:33:23 UTC (351 KB)
[v3]
Wed, 18 Nov 2026 15:00:10 UTC (351 KB)
来源:arXiv cs.AI · arxiv.org