Unverified50% confidenceFactExact time
For models that support long conversations, the context may reach tens of thousands of tokens, often carrying personal information, project codenames, or internal terminology that users inadvertently reveal.
1
Sources
50%
Confidence
Long-term
Relevance
10/27/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedDetails inadvertently disclosed in user conversations, such as writing style, geographic information, and personal experiences, may serve as clues for re-identification.71% similarVerified大语言模型在长时间对话中会因上下文窗口限制而丢失早期信息71% similarUnverifiedLarge language models have a maximum length limit, measured in tokens, for the historical interaction content they can read within a single conversation.69% similarUnverifiedLarge language models show insight on cryptography problems often because they have encountered large amounts of public literature, code, and discussion during training, allowing them to quickly connect scattered clues.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1104060API
curl https://kongchang.com/api/v1/knowledge/claims/1104060MCP
get_claim(id=1104060)