OpenAI Baselines is a collection of reference implementations of reinforcement learning algorithms designed to match published paper performance. These implementations serve as research baselines, making it easier for the community to replicate results, refine approaches, and compare new methods against established ones.