Unverified50% confidenceConceptExact time
Retention is one of the core metrics for evaluating language model state updates, referring to whether a model, after completing target field updates, can avoid modifying or forgetting fields that the conversation did not ask to change.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMemory机制在大语言模型工程中对应系统提示词工程与外部记忆存储的结合，用于模拟无状态模型的持久化状态71% similarUnverifiedThe append-only subtitle strategy (Stable Hypothesis) means that once the model outputs a word it no longer modifies it, avoiding frequent subtitle flickering and rewrites.69% similarUnverifiedWhether a model can accurately locate key information in an ultra-long context and avoid the lost-in-the-middle phenomenon is a decisive factor in determining coding performance.68% similarUnverifiedREPL架构的核心特征是持久化状态，每次执行结果保留在内存中供后续调用引用，这与无状态的函数调用形成根本区别68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1007757API
curl https://kongchang.com/api/v1/knowledge/claims/1007757MCP
get_claim(id=1007757)