ACE：面向多跳事实回忆的属性控制知识编辑

摘要

大型語言模型（LLMs）需要高效的知識編輯（KE）來更新事實信息，然而現有方法在多跳事實回憶中表現出顯著的性能衰退。這種失敗在編輯涉及推理鏈中的中間隱含主體時尤為嚴重。通過因果分析，我們揭示了這一限制源於對鏈式知識在神經元層面如何動態表示和利用的忽視。我們發現，在多跳推理過程中，隱含主體作為查詢神經元發揮作用，它們依次激活跨變壓器層的相應值神經元，以累積信息直至得出最終答案，這一動態過程是先前KE工作所忽視的。基於這一洞察，我們提出了ACE：面向多跳事實回憶的屬性控制知識編輯框架，該框架利用神經元層面的屬性來識別和編輯這些關鍵的查詢-值（Q-V）路徑。ACE為多跳KE提供了一種基於機制的解決方案，在GPT-J和Qwen3-8B上分別比最先進的方法高出9.44%和37.46%。我們的分析進一步揭示了Qwen3中更細粒度的激活模式，並表明值神經元的語義可解釋性是由查詢驅動的累積所協調的。這些發現基於對內部推理機制的原則性理解，為提升KE能力開辟了一條新途徑。

English

Large Language Models (LLMs) require efficient knowledge editing (KE) to update factual information, yet existing methods exhibit significant performance decay in multi-hop factual recall. This failure is particularly acute when edits involve intermediate implicit subjects within reasoning chains. Through causal analysis, we reveal that this limitation stems from an oversight of how chained knowledge is dynamically represented and utilized at the neuron level. We discover that during multi hop reasoning, implicit subjects function as query neurons, which sequentially activate corresponding value neurons across transformer layers to accumulate information toward the final answer, a dynamic prior KE work has overlooked. Guided by this insight, we propose ACE: Attribution-Controlled Knowledge Editing for Multi-hop Factual Recall, a framework that leverages neuron-level attribution to identify and edit these critical query-value (Q-V) pathways. ACE provides a mechanistically grounded solution for multi-hop KE, empirically outperforming state-of-the-art methods by 9.44% on GPT-J and 37.46% on Qwen3-8B. Our analysis further reveals more fine-grained activation patterns in Qwen3 and demonstrates that the semantic interpretability of value neurons is orchestrated by query-driven accumulation. These findings establish a new pathway for advancing KE capabilities based on the principled understanding of internal reasoning mechanisms.