arXiv cs.CL· Shuhaib Mehri, Philippe Laban, Michel Galley, Dilek T\"ur, Jennifer Neville, Jianfeng Gao·· 12 小时前AI 评分34
用用户表征提升模拟保真度：UserRepGen-4B 通过事后蒸馏生成用户表征
Learning User Representations to Improve Simulation Fidelity
AI 导读
研究提出用自然语言用户表征条件化用户模拟器，并设计对话重建任务来评估表征质量，同时引入事后蒸馏（hindsight distillation）在线训练表征生成器。由此训练的 UserRepGen-4B 在 700 段对话、7 个数据集上生成的表征，使模拟保真度超过 GPT-5.5 生成的表征。学习到的表征还能提升未见任务的模拟保真度、更好保留真实用户的人群多样性，并更有效区分不同对话中的用户。
正文 · 原文
Abstract:User simulators enable interactive training and evaluation of AI assistants at scale. Yet current simulators struggle to reproduce the diverse behaviors of real users, leading to assistants that generalize poorly to underrepresented users. We identify user representations as an underexplored bottleneck in user simulation. User representations are natural-language descriptions that condition simulators to reproduce a particular user's behavior. A representation's quality becomes apparent in hindsight, through the behavior it induces in a simulator. To evaluate and optimize user representations, we formulate a conversation reconstruction task: given an original conversation, we generate a representation, condition a simulator on it, simulate a conversation, and then measure fidelity by comparing user behavior in the original and simulated conversations. This comparison also yields hindsight feedback, which identifies where the representation falls short and how it can be improved. We introduce hindsight distillation, an online training procedure that distills feedback-conditioned improvements into the representation generator, enabling it to generate improved representations directly from the original conversation. We present UserRepGen-4B, a representation generator trained with hindsight distillation. Across 700 conversations spanning seven datasets, hindsight feedback helps improve user representations at test time, and UserRepGen-4B generates representations that enable more faithful simulation than those generated by GPT-5.5. Beyond conversation reconstruction, learned representations improve simulation fidelity for unseen tasks, better preserve the population-level diversity of real users, and better distinguish users across conversations.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.22518 [cs.CL] |
| (or arXiv:2610.22518v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.22518 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Shuhaib Mehri [view email]
[v1]
Thu, 15 Oct 2026 18:13:21 UTC (596 KB)
来源：arXiv cs.CL · arxiv.org