Keynote – AI数学家能否不仅仅是一个黑箱？

Large language models are getting better and better at solving mathematics problems, but when one is learning mathematics one wants more than just a solution: one wants to understand the insights that led to the solution. These can often be hard to extract from the output of an LLM even when that output includes a "chain of thought". I shall discuss work my research group is undertaking that is aimed at improving this situation.