Unverified50% confidenceSolutionExact time
Pre-storing or quickly computing Top-K results at each prefix node enables constant query cost — returning a fixed Top-K regardless of how many entries the prefix matches — and is the key to flattening P99 latency.
1
Sources
50%
Confidence
Long-term
Relevance
9/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTop-K检索是从向量数据库中取相似度最高的前K个片段作为候选内容65% similarUnverified向量数据库使用近似最近邻搜索（ANN）算法如HNSW、IVF-PQ，其结果本质上是概率性的，相同查询在索引重建后可能返回不同的Top-K结果64% similarUnverifiedThe Top-K pruning strategy is essentially a branch-and-bound algorithm based on a priority queue (min-heap).64% similarUnverifiedReranking is typically implemented using cross-encoders, which concatenate the query and candidate documents before feeding them in together; it usually reranks only the Top-50 to Top-100 candidates and finally selects Top-3 to Top-5 to pass to the generative model.63% similarUnverifiedMulti-query retrieval uses an LLM to automatically generate 3–5 semantically equivalent but differently phrased variant queries for the same question, retrieves results for each separately, and takes the union to improve recall.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/911708API
curl https://kongchang.com/api/v1/knowledge/claims/911708MCP
get_claim(id=911708)