🔍Basic reflection involves generating an initial response, reflecting on it, and generating a critique for improvement.
🔄Reflexion agents include executing suggested tool queries to gather additional context for a revised response.
🤝Language agent research uses large language models as agents for generation, value functions, and optimization.
🔎Reflexion language agents combine language models, reinforcement learning, and reflection for more advanced processing.
⏰The reflexion actor architecture reduces processing time by separating execution and revision stages.