Advance the reasoning and planning capabilities of large foundation models. Enhance model performance across the entire development lifecycle—including data acquisition, supervised fine-tuning (SFT), reward modelling, and reinforcement learning—while driving innovations in reasoning and decision-making.