Unverified60% confidenceFactExact time
The Reflexion mechanism allows an Agent to generate natural-language experience summaries after failures and reference them in subsequent decision-making.
2
Sources
60%
Confidence
Long-term
Relevance
8/26/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedError feedback refers to failure information flowing into the current task context so the Agent makes a different choice, but this does not mean the Agent learns permanently unless the system is deliberately designed to store and leverage this experience.72% similarUnverifiedLarge language model-driven Agents perform implicit confidence assessments when making decisions, and when confidence is high they tend to shorten the action chain and skip verification steps.69% similarUnverified策略漂移指目标和状态认知正确但Agent解决问题的方法逐渐退化，反映模型缺乏有效的元认知和自我纠偏机制68% similarUnverifiedLetta leaves the memory write decision to the Agent itself; its reflection prompt includes an already-captured check: if the memory or skill already sufficiently contains this information, skip it, which is an instruction to the model rather than a threshold in code.67% similarUnverifiedReasoning-type models exhibit a typical failure mode in Agent scenarios: reasoning loops and action starvation, where the model continuously generates text about execution but never truly triggers tool calls.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/802156API
curl https://kongchang.com/api/v1/knowledge/claims/802156MCP
get_claim(id=802156)