StudentSim es una plataforma de investigación que simula la interacción de los estudiantes con tutores de IA, permitiendo evaluar y perfeccionar los sistemas de tutoría antes de su uso real. La simulación modela la participación en áreas como ajedrez, escritura y matemáticas, ofreciendo retroaliment

핵심 요약 • Microsoft Research의 StudentSim 소개 블로그와 StudentSim 논문은 학생 한 명의 기록으로 그 학생을 흉내 내는 시뮬레이터를 학습시킵니다. 평가는 학생과 닮은 정도(F)와 튜터 지도에 반응하는 정도(R)로 합니다. • 체스 30명 평가에서 StudentSim은 F 0.51, R 0.91로 GPT-5.4(0.23, 0.72)와 Maia2(0.45, 0.27)를 앞섰습니다. 다만 영작문과 수학에서 GPT-5.4와의 F 차이는 0.05와 0.03입니다. • 튜터 강화학습 실험의 보상에는 학생별 시뮬레이터가 아니라 100명의 기록으로 학습한 공용 시뮬레이터를 썼고,...

StudentSim is a research system for evaluating and improving AI tutoring tools before they are tested with learners. It models how individuals may engage with tutors, giving feedback to improve tutor performance across subjects like chess, writing, and mathematics:

StudentSim is a research system for evaluating and improving AI tutoring tools before they are tested with learners. It models how individuals may engage with tutors, giving feedback to improve tutor ...

Microsoft Research hat mit StudentSim ein neues Forschungssystem vorgestellt, das KI-Tutoren vor einem praktischen Einsatz mit echten Schülern trainieren und evaluieren soll. Wie Berichte vom 12.10.2026 darlegen, bildet das System individuelle Lernverhalten über computergestützte Modelle ab. Dadurch soll die pädagogische Interaktion zwischen virtuellen Lehrkräften und Lernenden gezielt erprobt werden, ohne dass unausgereifte Systeme direkt an […] The post StudentSim: Microsoft Research trainiert KI-Tutoren mit 60 simulierten Lernenden appeared first on BornCity.

AIチュータリングの精度向上を目指す学生シミュレーション技術。

Researchers from Microsoft Research and the University of Illinois Urbana-Champaign developed StudentSim, a framework for training individualized LLM-based student simulators that simultaneously...

Join the discussion on this paper page

AI tutors are most useful when they adapt to each student's strengths, weaknesses, and preferred guidance, but evidence about which guidance works for which student is sparse, slow, and costly to coll...

Microsoft and the University of Illinois built StudentSim, which creates a digital replica of each student from just a handful of past records so AI tutors can get fast, cheap feedback; the researchers report it beats GPT-5.4 at mimicking real students across chess, English, and math.

System StudentSim buduje spersonalizowane symulacje konkretnych uczniów na podstawie zaledwie kilku próbek ich pracy, odtwarzając ich unikalne błędy i reakcje na wskazówki. Pozwala to na skuteczne trenowanie korepetytorów AI bez udziału prawdziwych ludzi i zaporowych kosztów zbierania danych.

Microsoft and the University of Illinois built StudentSim to replicate individual students from limited data and give AI tutors fast, low-cost feedback. In tests covering 60 students across chess, English, and math, it outperformed GPT-5.4. A chess tutor trained with StudentSim also earned the highest expert ratings among three versions tested.

Join the discussion on this paper page

AI tutors are most useful when they adapt to each student's strengths, weaknesses, and preferred guidance, but evidence about which guidance works for which student is sparse, slow, and costly to coll...