Unverified50% confidenceFactExact time
Large models themselves are stateless; the model does not retain historical information between API calls and can only see the context provided in the current request.
1
Sources
50%
Confidence
Long-term
Relevance
9/9/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedEven if a model supports a context window of 128K or longer, stuffing all historical information into the prompt does not guarantee the model will correctly utilize that information, as attention dilution can cause critical information to be overlooked.67% similarUnverifiedInformation outside the context window cannot be accessed by the model.67% similarUnverifiedEven models with a 128K or larger context window cannot automatically perceive project directory structure, existing database schemas, team coding standards, or historical decisions.64% similarUnverified将推理链隐藏在API后端并不等于真正保护它，只要模型行为可被外部观测和大量查询就存在被逆向工程的空间63% similarUnverifiedNative large language models have shortcomings including no persistent memory, knowledge cutoff dates, inability to access the internet, and inability to read local files or call external tools.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/882652API
curl https://kongchang.com/api/v1/knowledge/claims/882652MCP
get_claim(id=882652)