option
Home
Flash News
Content
FredWhite
FredWhite
April 8, 2026

Alibaba's Tongyi Lab introduces FIPO algorithm, enabling reinforcement learning models to handle complex reasoning. The algorithm extends reasoning length beyond 10,000 tokens and achieved a key result: a 32B-scale model with FIPO, in a pure RL setup, surpassed competitors including OpenAI's o1-mini on some metrics.

Comments (0)
0/300
OR