Unverified50% confidenceFactExact time
The memory stream paradigm of Stanford Generative Agents retrieves memories through three dimensions—Recency, Importance, and Relevance—where Importance is self-rated by the model on a scale of 1–10 and Relevance is calculated via embedding vector cosine similarity.
1
Sources
50%
Confidence
Long-term
Relevance
8/7/2026
First Seen
Sources
五个本地大模型建了座小镇，涌现出Facebook和鸭子征信局
Related Claims
UnverifiedResearchers from KAST University of Tübingen and Never AI Lab proposed the MEME (Multi-Entity Evolving Memory Evaluation) benchmark for systematically evaluating the performance of LLM memory systems on dependency reasoning.71% similarUnverified上下文图谱为智能体提供了介于临时上下文和模型权重微调之间的中间层记忆机制，可在不改变底层模型的前提下持续积累经验69% similarUnverifiedCognitive science research shows that the effectiveness of the Feynman Technique stems from the 'Generation Effect'—active output forms stronger memory encoding than passive input.68% similarUnverifiedThe conditional retrieval mechanism of the N-gram memory architecture can precisely recall specific details from a context of more than 10 million tokens.67% similarUnverifiedThe FSRS algorithm uses machine learning to fit users' actual memory curves and is approximately 30%–40% more accurate than SM-2 in predicting forgetting probability.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/703767API
curl https://kongchang.com/api/v1/knowledge/claims/703767MCP
get_claim(id=703767)