SortingHat: AI teaching assistant for OS courses with RAG and MARL
RAG, MARL, and a 3D avatar combine to personalize OS education at scale.
SortingHat is a research prototype from Zhejiang University that tackles the complexity of Operating Systems education using a tailored AI teaching assistant. The system combines retrieval augmented generation (RAG) with multi-agent reinforcement learning (MARL) to adapt to each student’s learning history and performance. A 3D digital human interface, powered by large language models, provides empathetic, context-aware guidance and generates personalized exercises that reinforce weak areas and challenge advanced students. An automated evaluation pipeline ensures fair, consistent grading and delivers actionable feedback.
Beyond just content delivery, SortingHat transforms the traditional OS course into an engaging, scalable experience. By offloading adaptive tutoring and grading to AI, instructors can focus on higher-level mentorship. The paper, accepted at WWW '25 Companion, shows how AI can address the diversity of student backgrounds and learning speeds in one of CS’s toughest subjects. SortingHat represents a concrete step toward AI-powered education that is both personalized and practical.
- Combines retrieval augmented generation (RAG) and multi-agent reinforcement learning (MARL) to adapt to each student’s learning history and performance.
- Features a 3D digital human interface powered by LLMs for empathetic, context-aware guidance and personalized exercise generation.
- Includes an automated evaluation pipeline that delivers fair grading and actionable feedback, scaling OS education without losing personalization.
Why It Matters
SortingHat shows how AI can make notoriously hard OS courses adaptive and engaging, transforming education at scale.