Unverified50% confidenceFactExact time
Reflexion框架通过让Agent生成自然语言经验总结，以言语强化学习替代传统奖励信号
1
Sources
50%
Confidence
Long-term
Relevance
7/23/2026
First Seen
Sources
Related Claims
UnverifiedReflexion (2023) proposed replacing numerical rewards with language feedback, allowing agents to summarize failure experiences in natural language and store them in memory.83% similarUnverified强化学习的核心机制源自行为心理学中的操作性条件反射理论，智能体通过接收奖励或惩罚信号来调整行为策略69% similarVerifiedAfter training on massive conversational datasets, modern large language models have acquired a certain degree of emotional reasoning capability.62% similarUnverifiedAI情感陪伴通过大模型自然语言生成能力，结合RLHF（基于人类反馈的强化学习）训练出的对话风格来模拟具有同理心的表达方式62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/598289API
curl https://kongchang.com/api/v1/knowledge/claims/598289MCP
get_claim(id=598289)