arXiv cs.AI· Hang Zheng, Nan Jiang, Zichen Zhu, Zixu Yang, Zhiyang Tang, Shuai Fan, Lu Chen, Xiaobao Wu, Kai Yu·· 2 天前AI 评分29
ARGUS：面向可靠用户模拟的议程驱动框架
ARGUS: An Agenda-Driven Framework for Reliable User Simulation
AI 导读
研究者提出 ARGUS（Agenda-dRiven Goal-oriented User Simulator），一种神经符号框架，通过符号操作维护动态议程栈来跟踪用户目标状态，将状态追踪与自然语言生成严格解耦，以实现严格的目标推进和约束对齐交互。
正文 · 原文
Abstract:High-fidelity user simulators are essential for evaluating and training agents in interactive environments. However, current LLM-based simulators struggle with reliability issues, such as goal omission and policy non-compliance. We propose ARGUS (Agenda-dRiven Goal-oriented User Simulator), a neuro-symbolic framework that maintains the user's goal state using a dynamic agenda stack manipulated via symbolic operations. By rigorously decoupling state tracking from natural language generation, ARGUS facilitates strict goal progression and constraint-aligned interactions. For cost-efficient deployment, we distill ARGUS into compact models using a unified pipeline: evaluator-guided SFT followed by online RL with dense step-level rewards. To rigorously evaluate behavioral fidelity, we introduce a rubric-based metric Composite Behavioral Score (CBS). Evaluations on $\tau^3$-Bench and MultiWOZ demonstrate the significant superiority of the ARGUS architecture and the efficacy of our training pipeline. Code and data will be released soon.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.21346 [cs.AI] |
| (or arXiv:2610.21346v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.21346 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Hang Zheng [view email]
[v1]
Thu, 15 Oct 2026 06:29:36 UTC (1,850 KB)
来源：arXiv cs.AI · arxiv.org