exploitation

Definition

Choosing the action currently believed to be best in order to actually collect reward, as opposed to gathering information. Too much of it and the agent never finds the better option it has not tried; too little and it forgoes reward it could have banked. Despite the everyday connotation, the term is entirely neutral here — it just means using what you already know, and the exploration-exploitation trade-off names the tension between the two.

Explainers using this term

• Paper Walkthrough — WarpSAC: When RL's Safety Rails Become HandcuffsWarpSAC: Towards the Pinnacle of Scalable Off-policy RL by Rethinking Exploration and Exploitation