A long paper at the 27th International Conference on Artificial Intelligence in Education, part of the Festival of Learning in Seoul. The work asks a question every researcher using LLMs for coding eventually faces: when the model says it is confident, can we believe it?
