[Alibaba Tongyi Laboratory Intelligent Computing Team Launches New Algorithm FIPO] Jin10 News, April 7 – On April 7, the Alibaba Tongyi Laboratory Intelligent Computing Team announced the launch of a new algorithm, FIPO (Future-KL Influenced Policy Optimization). The algorithm introduces the Future-KL mechanism, rewarding key tokens to address the issue of 'inference length stagnation' in pure reinforcement learning (Pure RL) training. According to the team, under the 32B-scale pure RL setting, it has achieved a performance surpassing o1-mini and the same-scale DeepSeek-Zero-MATH.