Generative AI in Education

Designing and developing generative AI-powered tools and systems that enhance teaching and learning. This includes AI tutors, simulation-based learning environments, and automated educational content generation.

LLM Chatbot Content Generation EdTech

Representative: Can ChatGPT be a debate partner? Developing ChatGPT-based application "DEBO" for debate education (ETS, 2024)

Pedagogical Alignment

Aligning large language models with pedagogical principles to ensure AI systems provide educationally sound responses. We explore reward modeling, RLHF, and instruction tuning techniques grounded in learning science theories.

RLHF Instruction Tuning Learning Science

Representative: Rewarding How Models Think Pedagogically: Integrating Pedagogical Reasoning and Thinking Rewards for LLMs in Education (arXiv, 2026)

Large Language Models

Developing, fine-tuning, and evaluating large language models for educational applications. Research includes building benchmarks, domain-specific model training, and reasoning capabilities for educational tasks.

Fine-tuning Benchmark Reasoning Evaluation

Representative: OpenLearnLM Benchmark: A Unified Framework for Evaluating Knowledge, Skill, and Attitude in Educational LLMs (arXiv, 2026)

Knowledge Tracing

Modeling and predicting student knowledge states using deep learning and language model-based approaches. We develop intelligent systems that can diagnose learning gaps and prescribe personalized learning paths.

Deep Learning Student Modeling Learning Analytics

Representative: A Training-Free Large Reasoning Model-based Knowledge Tracing Framework for Unified Prediction and Prescription (arXiv, 2026)

Multimodal AI for Education

Building vision-language models and multimodal systems for educational contexts including art appreciation, teacher analytics, and educational game-based assessments.

Vision-Language Multimodal VQA

Representative: LLaVA-docent: Instruction tuning with multimodal large language model to support art appreciation education (CAEAI, 2024)

Robotics (VLA)

Developing Vision-Language-Action models for educational robotics. We explore embodied AI agents that can interact with students in physical learning environments, including social robots for special education and voice-controlled educational robots.

VLA Embodied AI Social Robot Jetbot

Representative: Pedagogical Alignment for Vision-Language-Action Models: A Comprehensive Framework for Data, Architecture, and Evaluation in Education (arXiv, 2026)