A Survey on LLM-based Conversational User Simulation
Bo Ni, Leyao Wang, Yu Wang, Branislav Kveton, Franck Dernoncourt, Yu Xia, Hongjie Chen, Reuben Leura, Samyadeep Basu, Subhojyoti Mukherjee, Puneet Mathur, Nesreen Ahmed, Junda Wu, Li Li, Huixin Zhang, Ruiyi Zhang, Tong Yu, Sungchul Kim, Jiuxiang Gu, Zhengzhong Tu, Alexa Siu, Zichao Wang
2026-04-30
Summary
This paper is about how we're getting better at creating computer programs that can convincingly act like people in conversations, thanks to new advances in artificial intelligence.
What's the problem?
Building realistic conversational AI requires being able to predict what a user might say or do next. Traditionally, this was hard to do well. While user simulation exists, it hasn't fully kept pace with the capabilities of modern AI like large language models, which are really good at generating human-like text. There was a need to understand how these new AI tools could be used to create better, more realistic simulations of users in conversations.
What's the solution?
The researchers looked at all the recent work using large language models to simulate users. They created a way to categorize these different approaches based on how detailed the user simulation is and what the goal of the simulation is. They also analyzed the common techniques used and how people are evaluating how well these simulations work, providing a clear overview of the field.
Why it matters?
This work is important because it helps AI developers build better chatbots, virtual assistants, and other conversational AI systems. By understanding the latest techniques and challenges in user simulation, researchers can create more effective and engaging AI that can interact with people in a more natural and helpful way. It also provides a framework for future research in this rapidly evolving area.
Abstract
User simulation has long played a vital role in computer science due to its potential to support a wide range of applications. Language, as the primary medium of human communication, forms the foundation of social interaction and behavior. Consequently, simulating conversational behavior has become a key area of study. Recent advancements in large language models (LLMs) have significantly catalyzed progress in this domain by enabling high-fidelity generation of synthetic user conversation. In this paper, we survey recent advancements in LLM-based conversational user simulation. We introduce a novel taxonomy covering user granularity and simulation objectives. Additionally, we systematically analyze core techniques and evaluation methodologies. We aim to keep the research community informed of the latest advancements in conversational user simulation and to further facilitate future research by identifying open challenges and organizing existing work under a unified framework.