对话中的基础误解：一种面向MapTask的视角主义标注方案
Grounded Misunderstandings in Asymmetric Dialogue: A Perspectivist Annotation Scheme for MapTask
AI总结：
本文提出一种视角主义标注方案，用于分析MapTask中对话参与者对参照表达的 grounded 解释差异，揭示理解如何演变并修复，评估LLM对视角依赖性 grounding 的建模能力。
摘要：
协作对话依赖参与者逐步建立共同基础，但在不对称情境中，参与者可能认为自己达成一致，但实际指代不同实体。我们引入了一种针对HCRC MapTask语料库（Anderson等人，1991）的视角主义标注方案，分别捕捉每个参照表达的说话人和收信人的 grounded 解释，使我们能够追踪理解如何随时间产生、分歧和修复。使用受方案约束的LLM标注流程，我们获得了13,000个标注的参照表达，包含可靠性估计，并分析了由此产生的理解状态。结果表明，一旦词形变体被统一，完全误解很少见，但多重性差异系统性地引起分歧，揭示了看似稳固的基础可能掩盖指代不一致。我们的框架为研究 grounded 误解以及评估（V）LLMs对协作对话中视角依赖性 grounding 的建模能力提供了资源和分析视角。
英文摘要：
Collaborative dialogue relies on participants incrementally establishing common ground, yet in asymmetric settings they may believe they agree while referring to different entities. We introduce a perspectivist annotation scheme for the HCRC MapTask corpus (Anderson et al., 1991) that separately captures speaker and addressee grounded interpretations for each reference expression, enabling us to trace how understanding emerges, diverges, and repairs over time. Using a scheme-constrained LLM annotation pipeline, we obtain 13k annotated reference expressions with reliability estimates and analyze the resulting understanding states. The results show that full misunderstandings are rare once lexical variants are unified, but multiplicity discrepancies systematically induce divergences, revealing how apparent grounding can mask referential misalignment. Our framework provides both a resource and an analytic lens for studying grounded misunderstanding and for evaluating (V)LLMs' capacity to model perspective-dependent grounding in collaborative dialogue.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
