Reflexion
Reflexion is an AI agent capability level that enables a model to perform self-evaluation and reflective correction of its own outputs. Through an internal review mechanism, it improves the quality of task completion.
Core Facts
Timeline (last 90 days)
Reflexion-type agents have the model generate a self-evaluation report after each task completion and write the experience into memory for future reference.
Architectures such as Reflexion and Tree-of-Thought are designed to introduce systematic backtracking and verification capabilities in multi-step reasoning.
RACER's reflective loop design draws on self-examination mechanisms from frameworks such as ReAct and Reflexion.
Research from MemGPT, Reflexion, and others found that having agents explicitly generate experience summaries after task completion and write them back to memory can significantly improve cross-task transfer capability.
Current LLM-driven RSI practice relies on models' code generation and self-critique capabilities, with typical representatives including AlphaCode's iterative repair process and the reflection mechanism described in the Reflexion paper
The concept of the Reflection mechanism originates from the 2023 paper "Reflexion" by Shinn et al.
The Reflexion architecture is most sensitive to token consumption, as the self-evaluation loop may bring several times the cost increase; it is recommended to enable it only in scenarios with strict requirements on result accuracy and to set a maximum number of reflection rounds.
The limitation of the Reflexion mechanism is that the reflection content disappears as the context window scrolls, making it essentially still short-term.
The industry typically divides Agent capabilities into several levels: single-step instruction execution (Tool Use), multi-step planning and execution (ReAct/Plan-and-Execute), autonomous reflection and correction (Reflexion), and multi-Agent collaboration.
Currently, mainstream LLM agent architectures are primarily divided into two categories: prompt-engineering-based methods (such as ReAct and Reflexion) and reinforcement-learning-based methods.
7 more timeline events
All Facts (20)
The Reflexion or Self-Refine mechanism was systematically proposed by Shinn et al. in a 2023 paper.
65%UnverifiedLLM-based Agent可以通过自然语言反思在极少次尝试中实现策略优化，这被学术界称为'反思型Agent'（Reflexion）架构
85%UnverifiedAI自我反思机制最早由Shinn等人在2023年提出的Reflexion框架中系统化阐述
85%Unverified2023年Noah Shinn等人提出的Reflexion框架首次系统化地论证了让LLM反思自身输出并迭代改进的有效性
85%Unverified斯坦福大学2023年发布了Reflexion框架，通过语言反馈而非梯度更新来实现Agent的学习
85%Unverified在Reflexion框架的编程任务中，加入自我反思循环后代码通过率从67%提升到了91%
80%UnverifiedReflexion机制已被证明能显著提升Agent在复杂任务中的成功率，部分实验显示成功率提升可达20%-30%
75%UnverifiedThe ReAct paradigm has spawned variants such as Reflexion, LATS, and Plan-and-Execute.
60%UnverifiedThe Reflexion mechanism allows an Agent to generate natural-language experience summaries after failures and reference them in subsequent decision-making.
60%UnverifiedReflexion-type agents have the model generate a self-evaluation report after each task completion and write the experience into memory for future reference.
50%UnverifiedArchitectures such as Reflexion and Tree-of-Thought are designed to introduce systematic backtracking and verification capabilities in multi-step reasoning.
50%UnverifiedRACER's reflective loop design draws on self-examination mechanisms from frameworks such as ReAct and Reflexion.
50%UnverifiedResearch from MemGPT, Reflexion, and others found that having agents explicitly generate experience summaries after task completion and write them back to memory can significantly improve cross-task transfer capability.
50%UnverifiedCurrent LLM-driven RSI practice relies on models' code generation and self-critique capabilities, with typical representatives including AlphaCode's iterative repair process and the reflection mechanism described in the Reflexion paper
50%UnverifiedThe concept of the Reflection mechanism originates from the 2023 paper "Reflexion" by Shinn et al.
50%UnverifiedThe Reflexion architecture is most sensitive to token consumption, as the self-evaluation loop may bring several times the cost increase; it is recommended to enable it only in scenarios with strict requirements on result accuracy and to set a maximum number of reflection rounds.
50%UnverifiedThe limitation of the Reflexion mechanism is that the reflection content disappears as the context window scrolls, making it essentially still short-term.
50%UnverifiedThe industry typically divides Agent capabilities into several levels: single-step instruction execution (Tool Use), multi-step planning and execution (ReAct/Plan-and-Execute), autonomous reflection and correction (Reflexion), and multi-Agent collaboration.
50%UnverifiedCurrently, mainstream LLM agent architectures are primarily divided into two categories: prompt-engineering-based methods (such as ReAct and Reflexion) and reinforcement-learning-based methods.
50%UnverifiedThe core mechanism of reflective prompt evolution is to have an LLM generate natural language reflections based on task execution results and feed them back into the prompt to form an iterative optimization loop; representative works include APE, EvoPrompt, and Reflexion.
50%