Unverified50% confidenceFactExact time
The recursive state-compression structure of linear attention brings constant memory usage and linear time, but all historical information must be squeezed into a fixed-size memory, and interference easily occurs when new information conflicts with old information.
1
Sources
50%
Confidence
Long-term
Relevance
10/20/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedEach time a memory is activated, it briefly enters an unstable state (the reconsolidation process), during which it is susceptible to being rewritten by new information.77% similarUnverifiedActive forgetting and summary compression are necessary measures to combat the unbounded growth of memory stores.74% similarUnverified在推理阶段内存带宽往往比纯算力更早成为瓶颈，注意力机制需在每步生成时读取全部KV缓存，上下文长度增大时内存读写开销呈平方级增长74% similarUnverifiedThe most common problem in on-device inference is memory, as context length silently inflates the KV cache until it collides with the weights.73% similarUnverifiedInjecting all chat history into the context leads to problems such as context explosion, attention dilution, and memory fragmentation.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1073420API
curl https://kongchang.com/api/v1/knowledge/claims/1073420MCP
get_claim(id=1073420)