KITE: A RAG-Based Tutoring System for Algorithmic Problem-Solving
The Knowledge-Informed Tutoring Engine (KITE) is an intelligent tutoring system that leverages Retrieval-Augmented Generation (RAG) to support students in mastering algorithms. It implements an intent-aware Socratic approach, offering customized hints, guiding questions, and adaptive scaffolding based on each learner's requirements. Utilizing a multimodal RAG pipeline, KITE retrieves pertinent information from course materials to ensure that its responses are consistent with the curriculum. The system's effectiveness was assessed through RAGAs-based metrics focusing on response quality and grounding, expert evaluations of pedagogical effectiveness, and a simulation study. The research paper can be found on arXiv with the ID 2605.12988.
Key facts
- KITE stands for Knowledge-Informed Tutoring Engine
- It is a Retrieval-Augmented Generation (RAG)-based intelligent tutoring system
- Designed for algorithmic reasoning and problem-solving tasks
- Uses an intent-aware Socratic response strategy
- Employs a multimodal RAG pipeline to retrieve from course materials
- Evaluated using RAGAs-based metrics, expert evaluation, and simulation
- Paper published on arXiv with ID 2605.12988
- Focuses on algorithm tracing and debugging reasoning errors
Entities
Institutions
- arXiv