EnterpriseGym Corecraft: Training Generalizable Agents on High-Fidelity RL Environments

arXiv:2602.16179v1 Announce Type: new Abstract: We show that training AI agents on high-fidelity reinforcement learning environments produces capabilities that generalize beyond the training distribution. We …