What characteristic distinguishes reinforcement learning from supervised learning?

Reinforcement learning is distinguished from supervised learning primarily because supervised learning uses labeled data to train models, where input-output pairs guide the learning process. In supervised learning, the model learns from a dataset consisting of features paired with corresponding labels that denote the correct output. This enables the model to make predictions or classifications based on the patterns it identifies during training.

On the other hand, reinforcement learning operates differently. In reinforcement learning, an agent learns by interacting with an environment and receiving feedback in the form of rewards or penalties based on its actions. This means that the learning process is driven by the agent's experiences rather than from explicit labeled data. The agent explores various actions to maximize cumulative rewards, without direct indications of what constitutes 'correct' or 'incorrect' actions, differentiating it fundamentally from supervised learning.

The other options mention characteristics that do not accurately describe the nature of reinforcement learning versus supervised learning, further emphasizing that it's the reliance on labeled data that sets supervised learning apart.