Search papers, labs, and topics across Lattice.
This paper introduces history-conditioned student simulation for LLM-powered tutoring tools, addressing the limitations of existing within-dialogue simulation methods. They propose a two-component framework consisting of a profile generator that summarizes a student's learning history and a simulator that predicts student turns based on the generated profile. Both components are trained using reinforcement learning to optimize for faithful student simulation, and are evaluated on a new real-world dataset of student dialogues and question responses.
Forget generic student models – this work shows you can get significantly more realistic student simulations by conditioning on a student's full learning history, unlocking better tutor model training and evaluation.
A key part of developing large language model (LLM)-powered, automated tutoring tools is student simulation, i.e., using LLMs to role-play as students, which can facilitate tutor model evaluation and training. Existing work mostly focuses on within-dialogue simulation, which lacks context on student knowledge and behavior, partly due to not grounding in past student question-answering or dialogue interactions. In this work, we introduce the task of history-conditioned student simulation, where the goal is to accurately predict student dialogue turns by leveraging information in the student's learning history. We propose a two-component framework in which a profile generator summarizes a student's history and a simulator predicts student turns conditioned on the resulting profile. We train both components with reinforcement learning (RL), yielding profiles optimized for faithful student simulation. We evaluate our method and baselines on the first-of-its-kind real-world dataset of student dialogues and question responses that we collect from a math learning platform. Extensive experiments show that our method significantly outperforms baselines, and demonstrate the importance of history, profiles, and RL training.
