SESKStuff Everyone Should Know
SESK: Stuff Everyone Should Know · sesk.info/ai-reinforcement-reward

In reinforcement learning, what tells the agent how well an action went?

A numerical reward from the environment.

Start playing free

  1. A label written by a human for each row
  2. The clustering score
  3. A numerical reward from the environment ✔

Why: A reinforcement learning agent takes actions in an environment and gets a number back, the reward. It learns a policy that earns the most total reward over time.

Huh, didn't know that: Game-playing AIs often learn this way: points or wins are the reward.

The dad joke
My dog uses reinforcement learning. Sit equals treat. Bark at mailman equals no treat but worth it.
Like this card?
Play the game

References

Answer: A numerical reward from the environment. A reinforcement learning agent takes actions in an environment and gets a number back, the reward. It learns a policy that earns the most total reward over time.

  1. Google Machine Learning Glossary: reward developers.google.com/machine-learning/glossary
    In reinforcement learning, the numerical result of taking an action in a state, as defined by the environment.
    Checked 2026-10-10.
  2. Google Machine Learning Glossary: reinforcement learning (RL) developers.google.com/machine-learning/glossary
    A family of algorithms that learn an optimal policy, whose goal is to maximize return when interacting with an environment.
    Checked 2026-10-10.

Think this answer is wrong? Tap "Challenge this answer" on the card in the game and tell us why.

Test yourself in the game

Artificial IntelligenceToolformer taught a language model to call tools. Which weak spots of LMs did that target?Artificial IntelligenceWith function calling, who actually runs a tool that the developer defined?Artificial IntelligenceWhat does the ReAct method have a language model do?

That's stuff everyone should know.