Student Data Paradox and Curious Case of Single Student-Tutor Model: Regressive Side Effects of Training LLMs for Personalized Learning

This paper investigates the 'Student Data Paradox' where training LLMs on student-tutor dialogue data to model student misconceptions inadvertently degrades the model's own factual knowledge and reasoning abilities. The authors propose hallucination tokens as a mitigation strategy to help LLMs distinguish between simulating student errors and providing accurate tutoring.

The pursuit of personalized education has led to the integration of Large Language Models (LLMs) in developing intelligent tutoring systems. To better understand and adapt to individual student needs, including their misconceptions, LLMs need to be trained on extensive datasets of student-tutor dialogues. Our research uncovers a fundamental challenge in this approach: the ``Student Data Paradox.'' This paradox emerges when LLMs, trained on student data to understand learner behavior, inadvertent

Source

Framework Categories

Tool Types

Tags