Unverified50% confidenceTradeoffExact time
Fine-tuning is not suitable as memory because retraining every time new user data is added is extremely costly, and a fine-tuned model makes it hard to precisely delete a piece of information or to guarantee per-user isolation.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedFor users who can fit the entire model into VRAM, this caching optimization is of limited significance, as the bottleneck is not in data transfer.75% similarUnverified现有向量数据库记忆方案缺乏时效性衰减机制，无法区分记忆是否被采纳并产生正向效果，也没有机制合并冗余或删除有害记忆74% similarUnverifiedThe problem of memory drift almost never exists in traditional software, because what is written to the database is explicit field values, yet it occurs very easily in architectures that allow models to autonomously generate and update memories.72% similarUnverifiedThe attention layer cannot benefit from batching because each sequence has its own unique KV cache.72% similarUnverifiedModels are generally lazy and will not proactively save information unless explicitly instructed to do so, requiring users to repeatedly remind them to save.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1060568API
curl https://kongchang.com/api/v1/knowledge/claims/1060568MCP
get_claim(id=1060568)