environment

Definition

In reinforcement learning, everything outside the agent: it receives an action, updates its state, and returns an observation and a reward. Defining the environment is defining the task, which is why benchmark environments matter as much as algorithms and why reward specification bugs show up as bizarre agent behaviour. Not the software sense of a runtime or dependency environment, though papers on agentic coding use both meanings within a page of each other.

Explainers using this term

• CogEvol: What the Reward Cannot Measure, RL Will Quietly DestroyCogEvol: Towards Efficient and Reliable Learning Environment Generation
• Paper Walkthrough: EnvACE — Agents That Rehearse the World Instead of Calling ItEnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning