Unverified50% confidenceFactExact time
PagedAttention draws on the operating system paging model, managing the lifecycle of shared physical blocks through reference counting, and reallocates new blocks via Copy-on-Write when a shared block is modified.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPagedAttention splits the KV Cache into fixed-size non-contiguous physical blocks and dynamically manages them through a mapping table from logical blocks to physical blocks, allowing different requests to share physical blocks.65% similarUnverifiedThe Flush+Reload attack exploits the physical sharing characteristics of shared memory pages and the CPU last-level cache, inferring the victim's memory access patterns with nanosecond-level precision by measuring the time difference for cache line reloading.64% similarUnverifiedPagedAttention manages the KV Cache in a manner similar to OS-level memory paging, reducing video memory fragmentation during long-context inference.64% similarUnverifiedThe framework supports creating separate iteration versions, with each iteration bound to corresponding pages and requirements, and all historical iterations archived in a unified manner for retrieval at any time.64% similarUnverifiedThe PagedAttention mechanism enables higher throughput and the ability to handle more concurrent requests on the same hardware.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/908132API
curl https://kongchang.com/api/v1/knowledge/claims/908132MCP
get_claim(id=908132)