Host: Jin-Peng Liu 刘锦鹏

Abstract: In this talk, we will introduce AI Mathematician (AIM), a multi‑agent framework that leverages Large Reasoning Models (LRMs) to assist research‑level mathematics. We focus on two core challenges, the intrinsic complexity of research problems and the requirement of procedural rigor. To address these challenges, AIM incorporates two core strategies: an exploration mechanism to foster longer solution paths, and the pessimistic reasonable verification method to ensure reliability. In preliminary experiments, AIM produced substantial portions of proofs and non‑trivial insights, while still requiring expert review and corrections. We will discuss the workflow, experimental results and limitations.

Bio: Yanqiao Wang is a Ph.D. student at Qiuzhen College, Tsinghua University, supervised by Prof. Jin-Peng Liu (YMSC) and Prof. Yang Liu (Institute for AI Industry Research).