Unverified50% confidenceBenchmarkExact time
SGLang's RadixAttention organizes the KV of prompts and generated results into a RadixTree, with common prefixes at the upper levels; paper data shows that the additional overhead is below 0.3% for workloads without prefix reuse.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified在RAG场景下RadixAttention可将共享上下文的KV Cache计算从O(n×请求数)降低至O(n)54% similarUnverifiedUsing retrieval-augmented generation (RAG) instead of the full codebase enables intelligent context selection to reduce token costs.54% similarUnverified文本分块策略、检索召回数量（Top-K）、重排序模型（Reranker）是影响RAG系统最终效果的三个关键变量51% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/908135API
curl https://kongchang.com/api/v1/knowledge/claims/908135MCP
get_claim(id=908135)