Our paper Agent World Model: Infinity Synthetic Environments for Agentic Reinforcement Learning is accepted to ICML 2026. [arXiv] [code]