advantage

Definition

In reinforcement learning, how much better an action is than the state's average value: A = Q − V. Subtracting that baseline leaves the expected gradient unchanged while cutting its variance, which is why the advantage rather than the raw return is what policy-gradient methods such as PPO multiply into their objective. The plain English sense — a benefit — also occurs, so read the sentence.

Explainers using this term

• Paper Walkthrough: SA-MRPO — Stop Studying the Subject You've Already AcedLearn What's Left, Not What's Mastered: Saturation Aware Advantage Reweighting for Multi-Reward Policy Optimization