Unverified50% confidenceTradeoffExact time
In scenarios with small context sizes, BM25 term-frequency retrieval outperforms semantic embedding solutions in terms of latency and cost
1
Sources
50%
Confidence
Long-term
Relevance
9/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedBM25 is a probabilistic improvement over TF-IDF that comprehensively considers term frequency, inverse document frequency, and document length normalization, but is unable to capture semantic similarity.72% similarVerifiedHybrid retrieval (combining dense vector retrieval with BM25 sparse retrieval) performs better than either method alone on most benchmarks.66% similarUnverifiedLLM-based structured extraction has strong generalization capability but incurs additional API call costs and latency; rule-based extraction is fast and low-cost but requires writing site-specific rules and is prone to breaking after redesigns—the two involve a significant trade-off.66% similarUnverified描述越短、关键词越精准，路由决策往往越准，上下文成本也越低65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/920798API
curl https://kongchang.com/api/v1/knowledge/claims/920798MCP
get_claim(id=920798)