# Microsoft describes StudentSim research for evaluating AI tutors

> StudentSim creates personalized simulators from de-identified learner records; the research system is not a classroom product.

- Canonical URL: https://newsverum.com/en/news/microsoft-studentsim-research-tests-ai-tutor-feedback-2026
- Publisher: Newsverum
- Section: Technology
- Country: United States
- Published: 2026-10-12T23:33:38+00:00
- Updated: 2026-10-12T23:33:38+00:00
- Event / data period: 2026-10-12
- Other editions: [Türkçe](https://newsverum.com/tr/news/microsoft-studentsim-research-tests-ai-tutor-feedback-2026.md), [العربية](https://newsverum.com/ar/news/microsoft-studentsim-research-tests-ai-tutor-feedback-2026.md), [Français](https://newsverum.com/fr/news/microsoft-studentsim-research-tests-ai-tutor-feedback-2026.md), [Español](https://newsverum.com/es/news/microsoft-studentsim-research-tests-ai-tutor-feedback-2026.md), [Deutsch](https://newsverum.com/de/news/microsoft-studentsim-research-tests-ai-tutor-feedback-2026.md), [Português](https://newsverum.com/pt/news/microsoft-studentsim-research-tests-ai-tutor-feedback-2026.md)

Microsoft Research described StudentSim, a student-simulation system, in a research announcement dated October 12. It learns from de-identified learner records publicly released for research and creates a simulator for each student. The evaluation covers chess, second-language English writing and mathematics. The paper was previously released in September 2026; the October 12 blog explains its method and results.

The study separately measures how well a simulator matches a particular learner’s responses and how it changes its response after tutor guidance. The announcement reports StudentSim chess scores of 0.5150 for behavioral fidelity and 0.9067 for guidance responsiveness. These are research evaluations; they do not show classroom deployment or improvements of those magnitudes in real students’ learning.

## Data

### StudentSim chess research scores

Model comparison reported by Microsoft for the same research test.

Methodology: Scores retain the source’s decimal values; both evaluations use a 0–1 scale. Values concern the student simulators’ chess test and do not measure classroom learning outcomes or general model performance. No percentage conversion is applied.

| Model | Behavioral fidelity score (0–1) | Guidance responsiveness score (0–1) |
|---|---|---|
| StudentSim | 0.515 | 0.9067 |
| GPT-5.4 | 0.2316 | 0.7186 |
| Maia2 | 0.4535 | 0.2721 |

Sources: <https://www.microsoft.com/en-us/research/blog/simulating-student-learning-for-better-ai-tutoring/> · Download data: [JSON](https://newsverum.com/api/v1/articles/microsoft-studentsim-research-tests-ai-tutor-feedback-2026/data/studentsim-chess-research-scores?lang=en&format=json) · [CSV](https://newsverum.com/api/v1/articles/microsoft-studentsim-research-tests-ai-tutor-feedback-2026/data/studentsim-chess-research-scores?lang=en&format=csv)

## Sources

- [Microsoft Research](<https://www.microsoft.com/en-us/research/blog/simulating-student-learning-for-better-ai-tutoring/>) (Source checked: 2026-10-13; Published: 2026-10-12)
- [Microsoft Research publication record](<https://www.microsoft.com/en-us/research/publication/studentsim-training-llm-based-student-simulators/>) (Source checked: 2026-10-13)

---

Cite the canonical URL above. [Editorial standards and corrections](https://newsverum.com/en/editorial)