超越原始转录：基于LLM的数字孪生的结构化人物提取

AI 摘要

本研究探讨了基于LLM的数字孪生中人物信息结构化的影响，提出固定BDE结构在同类任务上优于原始转录，但在异构任务上效果不佳。为此，作者提出自动结构发现流程，在13个多样化子研究中恢复性能，表明关键限制在于信息组织方式而非信息量。

正文节选

Beyond Raw Transcripts: Structured Persona Extraction for LLM-Based Digital Twins Abstract LLM-based "digital twins" aim to simulate how an individual would behave in new environments or respond to novel questions, given some representation of that individual’s prior responses. A common approach constructs this representation from survey transcripts or summaries derived from them, and evaluates performance by predicting holdout responses. Prior work shows that compressing long transcripts into s

发布时间：2026-08-24 12:00

抓取时间：2026-08-24 12:04

来源机构：arXiv