1 paper
Xiaoyu Xu, Yulan Pan, Xiaosong Yuan +4
Large reasoning models (LRMs) that generate long chains of thought now perform well on multi-step math, science, and coding tasks. However, their behavior is still unstable and har…