DEFINITION

Reflection

Agent Reflection

Definition

A mechanism where AI agents analyze their own past actions, decisions, and outcomes to improve future behavior. Reflection typically involves reviewing memory, identifying patterns, and updating internal models or strategies based on experience.

Examples in the Wild

  • Example 1:Agent reviewing failed tool calls to adjust approach
  • Example 2:Summarizing conversation history to extract lessons learned
  • Example 3:Evaluating decision quality and adjusting confidence scores

See it in action

View agent-with-reflection Template →View self-improving-agent Template →