Google

An action is the set of all possible moves the agent can make. Agents choose from a list of possible actions. In video games, for example, the list might include: running right or left, jumping high or low, crouching or standing still.

Action in Reinforcement Learning

When applying reinforcement learning to Large Language Models, an action, denoted as $$a$$, corresponds to a possible decision the agent can make. Specifically, an action represents a predicted token chosen from the model's vocabulary.

Action in the Context of LLMs

An agent is being trained to play tic-tac-toe on a 3x3 grid. The agent's set of possible moves is defined as {'place X on square 1', 'place X on square 2', ..., 'place X on square 9'}. Critically evaluate this set of moves. Identify one significant flaw and explain why it is a problem for the agent's learning process.

Evaluating an Agent's Action Set

Read the following scenario and identify the complete set of possible moves the agent can make at each decision point.

Smart Thermostat Agent Actions

A simple robotic arm is being trained to sort objects on a conveyor belt. The arm can perform only three distinct movements from its resting position: it can pick up an object, it can place an object in a bin, or it can do nothing and wait. In this learning scenario, what does the set {pick up, place, wait} represent?

Learn Before

Related