Unverified85% confidenceFactTime unknown
斯坦福大学的生成式智能体（Generative Agents）研究和Letta（原MemGPT）等项目也在探索跨会话持久记忆和系统级学习方向
1
Sources
85%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
Unverified知识蒸馏思路的理论依据来自斯坦福Alpaca项目及后续ToolBench、AgentInstruct等工作68% similarUnverified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66% similarUnverifiedKnowly's active learning feature draws on the learning science principle of Spaced Repetition and combines it with the contextual understanding capabilities of LLMs.62% similarUnverified任务路由的思想根源可追溯到早期专家系统和集成学习中的模型选择与混合专家（MoE）机制62% similarUnverified斯坦福的ACT和CMU的Diffusion Policy验证了端到端感知-动作学习在灵巧操作任务中的有效性，Diffusion Policy将DDPM引入动作生成62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/28736API
curl https://kongchang.com/api/v1/knowledge/claims/28736MCP
get_claim(id=28736)