UserLM-8b

microsoft/UserLM-8b model card Model description Unlike typical LLMs that are trained to play the role of the "assistant" in conversation, we trained UserLM-8b to simulate the “user” role in conversation (by training it to predict user turns in a large corpus of conversations called WildChat). This model is useful in simulating more realistic conversations, which is in turn useful in the development of more robust assistants. The model takes a single input, which is the “task intent”, which defines the high-level objective that the user simulator should pursue. The user can then be used to generate: (1) a first-turn user utterance, (2) generate follow-up user utterances based on a conversation state (one or several user-assistant turn exchanges), and (3) generate a token when the user simulator expects that the conversation has run its course. Developed by: Tarek Naous (intern at MSR Summer 2025), Philippe Laban (MSR), Wei Xu, Jennifer Neville (MSR) Paper: https://arxiv.org/abs/2510.06552 How to get started with the model Here’s a simple snippet to use the model: Uses Direct intended uses The UserLM-8b is released for use by researchers involved in the evaluation of assistant LLMs. In such scenarios, UserLM-8b can be used to simulate multi-turn conversations, with our analyses (see Section 3 of the paper) giving evidence that UserLM-8b provides more realistic simulation of user behavior than other methods (such as prompting an assistant model). UserLM-8b offers a user simulation environment that can better estimate the performance of an assistant LLM with real users. See Section 4 of the paper for an initial implementation of such an evaluation. Downstream uses We envision several potential uses for UserLM-8b that we did not implement yet in our presented work but describe in our Discussion section as potential research directions for UserLMs. These potential applications include: (1) user modeling (i.e., predicting user responses to a given set of questions), (2) f