Build and post-train AI agents for long-horizon scientific research. Turn research workflows into environments, evaluations and reward signals, and test reproducibility and evidence quality. Develop reinforcement-learning approaches with research par