Will Yeadon

LLM grading effectiveness hinges on task structure, not model power, study finds

A new study published on arXiv investigates the effectiveness of using large language models (LLMs) as automated graders for physics assessments. The research found that LLM performance is highly dependent on the specif…