RL Environment
A packaged task world for reinforcement learning - the world, rules, and feedback loop of state, action, and reward a model trains inside.
Plain English
A gym the model practices in; without one, RL is an algorithm with nothing to act on.
Example
A team wraps their issue tracker's API in an environment with a 'bug fixed, tests pass, no drive-by changes' reward and fine-tunes a coding model that is unusually good at their stack.
Why it matters
Model improvement is shifting from scraping data to building environments - the Environments Hub made them shareable like datasets, and teams that wrap their product can train and eval against it.