Assessing Programming Proficiency Through Eye Gaze Analysis Using Fine-Tuned Large Language Model

The exploration of human eye movement data offers a groundbreaking approach to assessing programming expertise, marking a significant stride in advancing STEM education. This research explores the capabilities of fine-tuned large language models (LLMs) to assess the programming proficiency of readers by analyzing their eye movements during the reading of program code. To achieve this, we fine-tuned the LLMs using the largest eye movement in programming dataset that is publicly available. Subsequently, we executed a series of experiments to evaluate the efficacy of our fine-tuned models in assessing programming skills based on eye movement traces.