SEAD: Self-Evolving Agent for Multi-Turn Service Dialogue

2026-02-03

AI Summary

This work addresses the limitations of existing task-oriented dialogue systems, which are often trained on noisy, low-quality human conversational data and struggle to simulate realistic user behaviors. To overcome these challenges, the authors propose SEAD, a novel framework that decouples user modeling into a Profile Controller—dynamically generating user states—and a User Role-play Model that emulates authentic interactions, thereby establishing an adaptive, non-adversarial training environment. Built upon large language models and integrating curriculum learning with role-playing mechanisms, SEAD enables unsupervised and efficient policy learning without requiring extensive human annotations. Experimental results demonstrate that SEAD achieves a 17.6% improvement in task completion rate and an 11.1% gain in dialogue efficiency, significantly outperforming both open-source and proprietary baselines.

Technology Category

• Multiagent Systems: Adversarial Agents
• Natural Language Processing: Conversational AI/Dialog Systems
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Large Language Models have demonstrated remarkable capabilities in open-domain dialogues. However, current methods exhibit suboptimal performance in service dialogues, as they rely on noisy, low-quality human conversation data. This limitation arises from data scarcity and the difficulty of simulating authentic, goal-oriented user behaviors. To address these issues, we propose SEAD (Self-Evolving Agent for Service Dialogue), a framework that enables agents to learn effective strategies without large-scale human annotations. SEAD decouples user modeling into two components: a Profile Controller that generates diverse user states to manage training curriculum, and a User Role-play Model that focuses on realistic role-playing. This design ensures the environment provides adaptive training scenarios rather than acting as an unfair adversary. Experiments demonstrate that SEAD significantly outperforms Open-source Foundation Models and Closed-source Commercial Models, improving task completion rate by 17.6% and dialogue efficiency by 11.1%. Code is available at: https://github.com/Da1yuqin/SEAD.

Problem

• service dialogue
• data scarcity
• user behavior simulation
• goal-oriented dialogue
• low-quality human data

Innovation

• Self-Evolving Agent
• Service Dialogue
• User Modeling
• Role-Play Simulation
• Curriculum Learning