.png)
GRENZE International Journal of Engineering and Technology
Vol. 12
(2026), Issue 2
AI-Powered Multi-Agent Debugging Platform for Computer Science Education
Authors
Sai Venkata Aditya Vardhan Mandava, Y Padma Geetha RajaRajeswari Kuncha, Prabhu sharanya Makkapati, Eswar Teja Kosuru
Abstract
Programmers struggle with debugging due to inadequate feedback from conventional tools. DebugMentor proposes a closed-loop multi-agent pipeline—Analyze, Fix, Verify—leveraging LLMs with dynamic few-shot learning, a five-strategy SmartComparator, and Docker-sandboxed execution. Unlike static AI tools, it delivers verified fixes with root cause analysis, achieving 87% fix accuracy and 34% superior explanation quality across four programming languages.
Pages:
6183 - 6197