Unverified50% confidenceDecision RuleExact time
当检索到的文本块余弦相似度低于0.60时，图会完全绕过LLM合成环节，转而向用户请求澄清或回退到实时网络搜索
1
Sources
50%
Confidence
Long-term
Relevance
8/21/2026
First Seen
Sources
Related Claims
Unverified嵌入模型将文本转化为高维稠密向量（通常为768至4096维），通过近似最近邻算法（ANN，如HNSW或IVF-PQ）快速检索语义相似内容65% similarUnverifiedSemantic search converts text into high-dimensional vectors and finds the most relevant conclusions by computing cosine similarity, allowing queries phrased in different ways to match semantically equivalent results.64% similarUnverifiedHybrid retrieval combining keyword search with vector search can compensate for the coverage blind spots of pure semantic retrieval.64% similarUnverifiedSemantic caching converts queries into vector embeddings and calculates cosine similarity to identify semantically equivalent queries, which can reduce actual API call volume by 30%–70%.63% similarUnverified通过歌词文本进行语义搜索可利用余弦相似度进行模糊匹配，弥合记忆模糊与精确检索之间的鸿沟63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/782559API
curl https://kongchang.com/api/v1/knowledge/claims/782559MCP
get_claim(id=782559)