Reflexion Architecture
Domain: Agent Architectures
🎯 Motivation
After each attempt, the agent reflects on failures and adds self-critique to memory, improving on the next try.
📋 Concrete Example
Attempt 1: code fails. Reflection: ‘Forgot to handle empty input.’ Attempt 2: code with empty input handling. Reflection: ‘Now works but slow.’ Attempt 3: optimized. Each attempt is better.
🔗 Analogy
A student taking practice tests - after each test, they note which questions they missed and why. Before the next test, they study those areas. Each practice test improves their score.