difficulty-estimation

This paper investigates how LLM-generated difficulty ratings for math items align with actual student performance, finding that LLMs systematically underestimate difficulty for items driven by learner misconceptions, a phenomenon termed the 'Easy Trap'.