揭示大模型如何识别并处理单复数指代关系
A Circuit for Plural Reference: How LLMs Represent and Retrieve Singular and Plural Entities
- 通过注意力头分析，发现模型能分别处理单复数实体表示
- 复数指代更倾向用复数代词，且在语义相似+用'和'连接时更明显
- 方法适用于理解模型推理机制，适合自然语言处理研究者
指代消解是上下文推理中的关键任务。本文研究大模型在处理复数指代时对单复数实体的表征与检索机制。结合机制可解释性与注意力模式分析，探究模型如何预测代词所指的前指实体。通过多种因果干预技术，发现一组注意力头负责：（1）在输入中表征指代信息，（2）识别构成复数指代的实体，（3）将信息传递至选择前指并预测代词的组件。此外，发现大模型在偏好上与人类一致，当实体语义相似且由‘和’连接时，更倾向于以复数形式指代。
原文摘要 · Abstract (English)
Coreference resolution is an important task in contextual reasoning. In this paper, we investigate the mechanism for representing and retrieving singular and plural entities for plural reference. We use a combination of mechanistic interpretability and attention pattern analysis to study the process in which LLMs predict a pronoun to refer back to previously mentioned entities. Using a range of causal intervention techniques, we find a set of attention heads that are responsible for (1) representing coreference information in the input, (2) identifying entities that form a plural reference, (3) transferring the information to the component that is responsible for selecting the antecedents and predicting the pronoun. We also find that LLMs align with humans in preference for plural pronoun. Specifically, entities in a plural construction are more likely to be referred to as a plural entity if they are ontologically similar and are linked by the conjunction "and".
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。