Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LeafTutor: An AI Agent for Programming Assignment Tutoring

Madison Bochard, Tim Conser, Alyssa Duran, Lazaro Martull, Pu Tian, Yalong Wu

TL;DR

LeafTutor addresses the shortage of instructors in programming education by delivering guided, step-by-step tutoring powered by LLMs and grounded in assignment-specific materials and code-execution feedback. The system architecture includes instructor and student interfaces with an integrated code editor and backend compiler, enabling context-aware responses derived from retrieved course content. Experimental testing across four Java assignments demonstrates competence in diagnosing conceptual, structural, and implementation issues, while runtime-error tracing reveals current limitations in precise fault localization. Overall, LeafTutor shows potential for classroom deployment as a scalable, personalized tutoring solution and sets the stage for deeper integration with learning management systems and adaptive student models.

Abstract

High enrollment in STEM-related degree programs has created increasing demand for scalable tutoring support, as universities experience a shortage of qualified instructors and teaching assistants (TAs). To address this challenge, LeafTutor, an AI tutoring agent powered by large language models (LLMs), was developed to provide step-by-step guidance for students. LeafTutor was evaluated through real programming assignments. The results indicate that the system can deliver step-by-step programming guidance comparable to human tutors. This work demonstrates the potential of LLM-driven tutoring solutions to enhance and personalize learning in STEM education. If any reader is interested in collaboration with our team to improve or test LeafTutor, please contact Pu Tian (pu.tian@stockton.edu) or Yalong Wu (wuy@uhcl.edu).

LeafTutor: An AI Agent for Programming Assignment Tutoring

TL;DR

LeafTutor addresses the shortage of instructors in programming education by delivering guided, step-by-step tutoring powered by LLMs and grounded in assignment-specific materials and code-execution feedback. The system architecture includes instructor and student interfaces with an integrated code editor and backend compiler, enabling context-aware responses derived from retrieved course content. Experimental testing across four Java assignments demonstrates competence in diagnosing conceptual, structural, and implementation issues, while runtime-error tracing reveals current limitations in precise fault localization. Overall, LeafTutor shows potential for classroom deployment as a scalable, personalized tutoring solution and sets the stage for deeper integration with learning management systems and adaptive student models.

Abstract

High enrollment in STEM-related degree programs has created increasing demand for scalable tutoring support, as universities experience a shortage of qualified instructors and teaching assistants (TAs). To address this challenge, LeafTutor, an AI tutoring agent powered by large language models (LLMs), was developed to provide step-by-step guidance for students. LeafTutor was evaluated through real programming assignments. The results indicate that the system can deliver step-by-step programming guidance comparable to human tutors. This work demonstrates the potential of LLM-driven tutoring solutions to enhance and personalize learning in STEM education. If any reader is interested in collaboration with our team to improve or test LeafTutor, please contact Pu Tian (pu.tian@stockton.edu) or Yalong Wu (wuy@uhcl.edu).
Paper Structure (11 sections, 3 figures)

This paper contains 11 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Related Works
  3. System Architecture
  4. Experimental Testing and Validation
  5. Code Not As Expected Error
  6. Code-Detail Error
  7. Logical Error
  8. Runtime Error
  9. Results
  10. Future Work
  11. Conclusion

Figures (3)

  • Figure 1: Overall workflow of the LeafTutor system, from instructor account creation and material ingestion to student assignment interaction and AI-assisted feedback.
  • Figure 2: Code Not As Expected Error. Interaction demonstrating LeafTutor diagnosing a non-error output issue. After detecting that the student’s linked-list implementation prints a memory reference rather than its contents, the tutor explains the cause, recommends implementing a toString() method, and confirms the corrected behavior once added.
  • Figure 3: Logical Error. LeafTutor identifying and correcting a logical error in BST minimum/maximum value computation.