Microsoft and University of Illinois launch StudentSim to train AI tutors
A joint effort by Microsoft and the University of Illinois introduced StudentSim, a system that creates digital replicas of individual learners from very limited data. The replicas mimic each student’s answers and typical mistakes, then adjust when given tutor feedback. StudentSim first trains a base model on pooled data from all students, using Alibaba’s Qwen3-4B-Instruct, and then fine‑tunes…
Key points
- StudentSim builds a replica for each learner using as few as three to five recorded essays.
- Across chess, English, and math, StudentSim outperformed GPT‑5.4, predicting chess moves about twice as often.
- Tutors trained with StudentSim showed higher factual accuracy and explanation quality than those trained with GPT‑5.4 or no extra training.
The researchers evaluated StudentSim with 60 students across chess, English as a foreign language, and math, using public datasets. In all three subjects it outperformed the larger GPT‑5.4 model; in chess it predicted the next move about twice as often and followed corrective hints almost always. A follow‑up test showed that a chess tutor trained with StudentSim made fewer factual errors and received higher scores for explanation quality than tutors trained with GPT‑5.4 or with no extra training. The team stresses this is a proof of concept and not a claim of having built the best tutor, and plans to model longer‑term knowledge acquisition next.
Simulated students that make realistic mistakes help AI tutors learn faster
The Decoder · 20 September 2026
Loading the full article…
This text was published by The Decoder and written by Jonathan Kemper. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- OpenAI offers U.S. university labs free ChatGPT Business seats with Rosalind access · 2 src
- Anthropic reports small AI models may self-replicate on IoT devices within months · 29 src
- Researchers propose LLM-referenced C-test for evaluating supra-intelligence · 8 src
- Researchers study permutation diffusion LLMs for recommendation reranking · 1 src
- Researchers compare Non-Generative verifier to Astra LLM for RAG claim checks · 1 src
Comments
via GitHub Discussions