Search for a command to run...
Can One Domain Help Others? A Data-Centric Study on Multi-Domain Reasoning via Reinforcement Learning