SLOW-THINKING REASONING MODEL WITH FASTCURL

Summary from the TLDR AI newsletter

FastCuRL-1.5B-Preview advances slow-thinking reasoning models with curriculum reinforcement learning, achieving state-of-the-art results in fewer training steps.