A new research paper explores the potential of System-One Large Language Models (LLMs) for knowledge tracing (KT) in educational settings, particularly when limited learner data is available. The study introduces a model called Jev, which, even without specific training data from a new platform, achieved a higher AUC score than existing deep KT models trained on a small number of learners. Further enhancements with JevKT, incorporating limited logged data and similar-learner statistics, improved performance, outperforming deep KT models up to 16 learners. AI
IMPACT This research suggests LLMs could significantly reduce the data requirements for educational modeling, enabling personalized learning from the outset.
RANK_REASON Research paper published on arXiv detailing a new approach to knowledge tracing using LLMs. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
