ACE-TA: An Agentic Teaching Assistant for Grounded Q&A, Quiz Generation, and Code Tutoring
Researchers' new agentic teaching assistant uses LLMs to provide stepwise coding guidance and adaptive assessments.
A research team from Mississippi State University and other institutions has published a paper on arXiv introducing ACE-TA (Agentic Coding and Explanations Teaching Assistant), a novel framework designed to automate and enhance programming education. The system leverages pre-trained Large Language Models (LLMs) to autonomously route student queries through three specialized, coordinated modules. This represents a shift from simple chatbots to a structured, multi-faceted agent capable of handling different educational tasks.
The first module is a retrieval-grounded conceptual Q&A system. It doesn't just generate answers from the LLM's general knowledge; it retrieves relevant context from specific course materials to provide precise, context-aligned explanations, reducing hallucinations. The second is an adaptive quiz generator that constructs multi-topic assessments designed to test higher-order understanding, not just rote memorization. The third and most interactive component is a code tutor that guides students through step-by-step reasoning, offering iterative feedback and utilizing sandboxed code execution for safe practice.
By combining these three functions—grounded explanations, assessment creation, and interactive tutoring—ACE-TA aims to create a more comprehensive and reliable automated teaching assistant. The framework's agentic design allows it to decide which tool is best for a given student query, moving beyond single-purpose AI tools. This coordinated approach addresses multiple pain points in scalable programming education simultaneously.
- Framework uses three LLM-powered modules: grounded Q&A, adaptive quiz generation, and interactive code tutoring.
- Code tutor provides step-by-step guidance with sandboxed execution for safe, iterative learning.
- System autonomously routes student queries to the appropriate module based on the task required.
Why It Matters
It automates scalable, personalized programming education, providing reliable, context-aware support that reduces instructor workload.