Unverified90% confidenceFactTime unknown
大语言模型的所有记忆都存储在当前的上下文窗口（Context Window）中，开启新Thread后这些信息全部清空
1
Sources
90%
Confidence
Long-term
Relevance
5/31/2026
First Seen
Sources
Codex使用指南：掌握Agent工作流的五个关键概念
bilibiliAli厂长
Related Entities
Related Claims
Unverified大模型的记忆仅限于当前上下文窗口内的token序列，对话结束后所有中间状态消失73% similarUnverified上下文窗口是指模型在单次对话中能够记住并参考的文本长度，超出范围的信息会被截断丢弃72% similarUnverifiedLLM本质上是无状态的推理引擎，每次调用模型时所有记忆都必须以文本形式塞入上下文窗口71% similarVerified上下文窗口（Context Window）是大语言模型能够同时处理的文本长度上限，当对话历史过长时，模型可能无法有效关注早期的重要信息，导致产生幻觉（Hallucination）71% similarUnverifiedLarge language models are limited by context window length, requiring external memory mechanisms to extend their memory capacity.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/5855API
curl https://kongchang.com/api/v1/knowledge/claims/5855MCP
get_claim(id=5855)