[2603.29142] REFINE: Real-world Exploration of Interactive Feedback and Student Behaviour

Computer Science > Artificial Intelligence arXiv:2603.29142 (cs) [Submitted on 31 Mar 2026] Title:REFINE: Real-world Exploration of Interactive Feedback and Student Behaviour Authors:Fares Fawzi, Seyed Parsa Neshaei, Marta Knezevic, Tanya Nazaretsky, Tanja Käser View a PDF of the paper titled REFINE: Real-world Exploration of Interactive Feedback and Student Behaviour, by Fares Fawzi and 4 other authors View PDF HTML (experimental) Abstract:Formative feedback is central to effective learning, yet providing timely, individualised feedback at scale remains a persistent challenge. While recent work has explored the use of large language models (LLMs) to automate feedback, most existing systems still conceptualise feedback as a static, one-way artifact, offering limited support for interpretation, clarification, or follow-up. In this work, we introduce REFINE, a locally deployable, multi-agent feedback system built on small, open-source LLMs that treats feedback as an interactive process. REFINE combines a pedagogically-grounded feedback generation agent with an LLM-as-a-judge-guided regeneration loop using a human-aligned judge, and a self-reflective tool-calling interactive agent that supports student follow-up questions with context-aware, actionable responses. We evaluate REFINE through controlled experiments and an authentic classroom deployment in an undergraduate computer science course. Automatic evaluations show that judge-guided regeneration significantly improves fe...