Learning and Debugging Assistant for Python with Retrieval Augmented Generation Capabilities
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Large Language Models(LLMs) often lack domain-specific grounding and may produce generic, hallucinated, or context-insensitive responses, despite their remarkable capability in assisting programming tasks. Retrieval-Augmented Generation (RAG) systems aim to address these limitations by incorporating external knowledge sources; however, many existing implementations remain passive responders with limited interactivity and pedagogical depth. This work presents the Rubber Duck Assistant, a RAG-powered python learning and debugging system that combines interactive dialog, metacognitive learning, semantic retrieval via ChromaDB, and real-time code execution. Inspired by the Rubber Duck Debugging approach, the assistant guides students to articulate and reflect on their code through precise explanations, bug identification, and correction strategies. Key contributions include a hybrid keyword-based LLM query classifier, format-aware chunk embedding, and a dedicated Debug Sandbox for safe code exploration. The Evaluation results demonstrate the effectiveness of the assistant, achieving 92% accuracy and a 91% F1 score in Python query classification. By fostering deeper learning and independent debugging skills, this study highlights how interactive and grounded artificial intelligence tutors can enhance programming education.
Fields of Study & Subjects
Teaching and Learning Programming
Software Engineering Research
Intelligent Tutoring Systems and Adaptive Learning
Publication details
- Journal / source
- International Conference on Computing, Communication, Control and Cyber-Physical Systems (I5CPS)
- Pages
- 1–6
- Publication date
- 26 February 2026
- Language
- EN