Unverified85% confidenceConceptTime unknown
上下文腐烂（Context Rot）包含两种不同的故障模式：Distraction（注意力稀释）和Poisoning（因果链污染）
1
Sources
85%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
Unverified尽管上下文窗口很大，上下文腐化（Context Rot）依然是巨大问题，模型对上下文中不同位置信息的注意力分布不均匀（Lost in the Middle效应）80% similarUnverifiedContext Rot refers to the phenomenon where, as the number of conversation turns increases, irrelevant information accumulates and the attention budget becomes filled with low-value content, causing reasoning quality to continuously decline.76% similarUnverifiedContext Rot（上下文腐烂）指随着对话变长、输入 Token 增多，模型准确度会逐步下降，该现象在 Claude、OpenAI、Gemini 等主流模型上普遍存在71% similarUnverifiedRAG存在两个漏洞：模型处理上下文时仍可能误读误用，以及检索到的上下文本身可能错误或无关63% similarUnverified尽管Opus 4.7已有百万级上下文窗口，但上下文腐化（Context Rot）依然是巨大问题59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/22384API
curl https://kongchang.com/api/v1/knowledge/claims/22384MCP
get_claim(id=22384)