OmniEdu improves K-12 AI teaching and learning models
OmniEdu: Open Foundation Models for Learning and Teaching
Computation and Language
Summary
Many educational AI tools either focus on solving problems or tutoring students, but don't handle all tasks well together. The authors created OmniEdu, a set of open AI models trained on diverse educational materials to better understand school subjects, curricula, student difficulties, and teaching methods. These models perform well on tests that check math skills, curriculum knowledge, and tutoring ability. This shows that carefully balancing different teaching skills during training can make AI better at helping both students and teachers.
What this means in practice
- •For educational software developers: Build AI-powered tutoring apps that understand curriculum structure and diagnose student difficulties more effectively using OmniEdu's open foundation models.$Commercial implications: OmniEdu enables creation of advanced personalized tutoring products for K-12 students by combining curriculum knowledge with teaching strategies.
- •For education content creators: Develop intelligent tools that generate or assess K-12 educational problems and instructional content aligned with curricula and learning goals.
Authors
Hao Liang, Qihan Lin, Meiyi Qiang, Linzhuang Sun, Hengyi Feng, Mingrui Chen, Sizhe Qiu, Wentao Zhang
Abstract
Educational foundation models must solve problems, understand curriculum structure, diagnose learner difficulties, and provide appropriate instructional support. Existing educational language models often focus on either problem solving or tutoring, with training mixtures organized by source or task rather than capability. We present OmniEdu, an open family of foundation models for K-12 learning and teaching. Its instruction-tuning corpus combines over 100 educational resources and general instruction sources, organized around four capabilities: subject competence, curriculum grounding, diagnostic reasoning, and pedagogical action and scaffolding. Our pipeline integrates deterministic cleaning, semantic auditing and rewriting, task-specific quality scoring, token-budgeted diversity selection, and pedagogical instruction assignment. It yields 69,999 examples and 15.96M supervised response tokens, including 60,951 education-specific examples. We fine-tune 4B, 9B, and 27B models and evaluate curriculum grounding, K-12 problem solving, and pedagogical tutoring, alongside general capability. Education-oriented tuning consistently improves all three educational benchmark groups across model scales. OmniEdu-27B achieves 63.12% EM and 76.69% F1 on K12-Bench, 85.89% on MathFish, 86.95% on EDUMATH, and 78.74% in MathTutorBench's Scaffold setting. It also achieves the highest Teaching average on LongTutor among the evaluated models, at 3.02. These results demonstrate the value of curated, capability-balanced supervision for adapting general language models to educational tasks spanning problem solving, curriculum understanding, and instructional support.