Scaling Offline RL via Efficient and Expressive Shortcut Models | Read Paper on Bytez