Unverified50% confidenceSolutionExact time
The REMORY paper proposes using a very small number of residual memory soft tokens to let a frozen LLM reconstruct long conversation history with nearly no loss.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedShort-term memory generally retains the raw content of the most recent 5–10 conversation turns and triggers a compression summary when the threshold is exceeded.77% similarUnverified短期记忆通过在提示词中拼接历史对话实现，受限于模型上下文长度，通常4K到200K tokens不等75% similarUnverifiedThe paper proposes the concept of minimum memory length, defined as the minimum history length that must be retained to achieve the same predictive performance as the complete history.74% similarVerifiedLangGraph supports two storage modes: short-term memory (single-session context) and long-term memory (cross-session persistent storage).70% similarUnverifiedThe three most common strategies for persisting conversation history are Window Buffer Memory, Summary Memory, and Vector Store Memory.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/998477API
curl https://kongchang.com/api/v1/knowledge/claims/998477MCP
get_claim(id=998477)