Provides modular, self-contained environments for training and evaluating LLMs with reinforcement learning—each environment bundles datasets, a model harness (tools/sandbox), and rubrics for automated scoring. Integrates with Prime RL and the Environments Hub.
