Verified65% confidenceFactExact time
The Reflexion architecture was proposed in 2023, introducing a self-reflection mechanism in which an Agent reviews its action trajectory at the language level after completing a task and stores the experience in long-term memory, enabling continuous improvement of task performance without updating model weights.
3
Sources
65%
Confidence
Long-term
Relevance
7/8/2026
First Seen
Sources
Related Claims
UnverifiedReflexio intervenes and learns in three situations: when users actively correct agent behavior, when an execution path fails, and when a particular approach works especially well.74% similarUnverifiedReflexion (the self-reflection framework) adds error summarization and strategy correction steps outside the loop, enabling proactive strategy adjustment after multiple failures.72% similarUnverifiedCommon solutions to long-horizon task planning include task graph mechanisms, the ReAct framework with self-correction capabilities, and the introduction of external memory modules.70% similarUnverifiedThe memory component enables an Agent to remember previous conversations and task execution progress, ensuring continuity across multi-step tasks.69% similarUnverifiedResearch 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.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/235807API
curl https://kongchang.com/api/v1/knowledge/claims/235807MCP
get_claim(id=235807)