Unverified50% confidenceFactExact time
Reflexion（Shinn等，2023）引入语言强化学习概念，将任务成败信号转化为自然语言反思存入记忆缓冲区供后续参考
1
Sources
50%
Confidence
Long-term
Relevance
7/20/2026
First Seen
Sources
Related Claims
Unverified反思（Reflection）机制源于Reflexion框架，通过任务后生成自然语言经验总结实现言语强化学习72% similarUnverifiedReflexion算法引入语言层面的自我反思机制，使Agent在每轮循环后生成自然语言评估并注入下一轮上下文72% similarUnverifiedMemory机制在大语言模型工程中对应系统提示词工程与外部记忆存储的结合，用于模拟无状态模型的持久化状态66% similarUnverified潜在缓存技术源于大语言模型中的KV Cache机制，Mistral将其从推理阶段迁移至训练阶段65% similarVerifiedAccording to situated learning theory, when knowledge is acquired in the specific context where it is applied, memory encoding is deeper and transfer ability is stronger.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/572479API
curl https://kongchang.com/api/v1/knowledge/claims/572479MCP
get_claim(id=572479)