Unverified50% confidenceFactExact time
更大的上下文窗口并不等于长期记忆，它只在单次推理中能处理更多内容，但不会在会话之间保留信息
1
Sources
50%
Confidence
Long-term
Relevance
8/18/2026
First Seen
Sources
Related Claims
Unverified压缩功能更适合处理冗余日志，而非承载关键推理链的对话历史73% similarUnverifiedThe context window refers to the length of text a model can remember and reference in a single conversation; information beyond this range is truncated and discarded.73% similarUnverifiedMemory Tags' automatic extraction feature is currently limited to word-level information extraction, which may be insufficient for deep learning scenarios that require understanding of conceptual relationships.71% similarUnverified实际开发中通常在存储长期记忆前让大模型对对话做总结摘要，提炼核心信息后再持久化，以节省空间并提高检索精确度70% similarVerifiedThe memory of large language models fundamentally relies on the attention mechanism of the Transformer architecture, and its effective information capacity is limited by the number of tokens in the context window.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/767523API
curl https://kongchang.com/api/v1/knowledge/claims/767523MCP
get_claim(id=767523)