arXiv cs.CL· Mustafa Omer Gul, Philippe Laban, Tanya Goyal, Jennifer Neville·· 13 小时前AI 评分31
UserLM-2.0：兼顾任务遵循与对话自然度的用户模拟器
Towards User Simulators That Stay On Task And Still Sound Natural
AI 导读
研究者提出一种强化学习方案，同时奖励任务意图遵循与行为自然度，并用轻量判别器与用户模拟器交替对抗训练来量化自然度，据此发布 UserLM-2.0。该模型生成同时满足任务忠实与自然度标准的有效模拟对话比例为 46.2%，是最强先前用户模拟器（14.1%）的三倍以上。
正文 · 原文
Abstract:Users increasingly rely on AI systems for extended, multi-turn tasks, yet evidence indicates that LLM-based system performance degrades over prolonged interaction. To evaluate and improve AI assistants in such settings, we need user simulators that faithfully follow complex user intents while exhibiting realistic conversational behavior. Existing approaches manifest a trade-off: prompted assistants can follow complex user intents but lack behavioral fidelity, while specialized user simulators better approximate conversational behavior but struggle to remain faithful to complex intents. In this paper, we propose a reinforcement learning (RL) recipe that jointly rewards task intent adherence and behavioral naturalness. A crucial contribution of the method is operationalizing naturalness with a lightweight discriminator trained in an alternating, adversarial loop with the user simulator. We release the resulting model, UserLM-2.0, which produces valid simulated conversations (i.e., meeting both task-faithfulness and naturalness criteria) 46.2% of the time, over three times the rate of the strongest prior user simulator (14.1%).
| Comments: | 30 pages, 16 figures, 6 tables. In submission |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.31512 [cs.CL] |
| (or arXiv:2610.31512v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.31512 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Mustafa Omer Gul [view email]
[v1]
Thu, 22 Oct 2026 16:26:48 UTC (935 KB)
来源：arXiv cs.CL · arxiv.org