arXiv cs.CL· Yibin Lou, Yi Hu, Jianhui Chen, Chenpeng Wang, Muhan Zhang, Liangming Pan·· 3 天前AI 评分29
LLM 该信任哪些知识？研究揭示跨知识冲突的通用引导方向
Which Knowledge Should LLMs Trust? Uncovering a Generalizable Steering Direction Across Knowledge Conflicts
AI 导读
研究发现 LLM 在内部参数知识与外部上下文知识之间的选择是线性可表示的，可通过残差流中的单一方向进行控制。该方向能跨上下文格式、知识类别和数据集泛化，并可从简单事实冲突迁移到 RAG、记忆和规则遵循三类任务，在四个模型家族中一致提升模型对提示所提供信息的利用。
正文 · 原文
Abstract:Large Language Models (LLMs) frequently encounter conflicts between internal parametric knowledge and external contextual knowledge, yet the mechanisms governing this selection remain unclear. We show that the choice between context and parametric knowledge is linearly represented and can be controlled along a single direction in the model's residual stream. We first demonstrate that this direction generalizes across factual-conflict settings, including context formats, knowledge categories, and datasets, while consistently shifting behavior between parametric and contextual answers. We then transfer a direction learned from simple factual conflicts to three more complex task families: RAG, memory, and rule following, where models must acquire information from retrieved documents, interaction histories, or explicit rules. Across four model families and multiple transfer settings, steering consistently improves the use of prompt-provided information over the unsteered model. These results suggest that the direction captures a transferable preference for external knowledge rather than a mechanism specific to factual conflicts.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.20067 [cs.CL] |
| (or arXiv:2610.20067v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.20067 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yibin Lou [view email]
[v1]
Wed, 14 Oct 2026 12:06:32 UTC (1,191 KB)
来源：arXiv cs.CL · arxiv.org