computing education

LLM-curated lecture clips offered source-grounded computing help, but the pilot measured use rather than learning

Owen Tang, Alexandra Vassar, Jake Renzella

arXiv preprint

Tang, Vassar and Renzella tested whether language models could answer computing questions by retrieving short segments from instructor-recorded C-programming lectures. GPT-5.4 Pro retrieved more of a lecturer's reference selections than Gemini 3.1 Pro, while Qwen3.5 performed poorly. A seven-week pilot logged 343 prompts from 903 enrolled students, but sparse feedback and no learning-outcome comparison mean the study supports feasibility, not effectiveness.