Actor-Critic

Fact

A reinforcement learning method with one part choosing actions and one part scoring them.

In Plain Words

One kid tries dance moves. The judge holds a scorecard: “Less robot arms.”

In AI, the actor picks the move. The critic scores it, so robots, games, and RLHF train with less random guessing.

Related Concepts

• Policy Gradient
• Q-Learning
• Reinforcement Learning
• Reinforcement Learning from Human Feedback

Policy Gradient
Actor-Critic adds a value score to Policy Gradient, so training is steadier.

Q-Learning
It borrows value learning, but it also learns the actions directly.

RL
Actor-Critic is a classic RL method family, often used for control.

RLHF
Many RLHF systems use it to improve the policy and keep learning stable.