Microsoft Research unveils StudentSim for AI tutor training

Microsoft Research introduced StudentSim, a system that builds a personalized simulator for each student and captures how their answers change after tutor guidance. It also released StudentSimEval, a benchmark covering chess, second-language English writing and math, with the paper and training code publicly available.

• StudentSim builds a per-student simulator that models responses to tutor guidance
• StudentSimEval covers 60 students across chess, English writing and math
• In chess StudentSim scores 0.5150 fidelity and 0.9067 responsiveness vs 0.2316 and 0.7186 for GPT-5.4
• Training uses a two-stage pipeline: a domain-level LoRA adapter, then per-student specialization