Unverified50% confidenceFactExact time
Multi-hop retrieval means first retrieving a first batch of documents, constructing a new query based on the results and then retrieving the next batch, gradually assembling the complete evidence chain; it applies to the situation where the all-evidence score is low while any-hit is normal.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe system first rapidly obtains preliminary results in parallel, and only initiates sequential search mode when the quality of the preliminary results is insufficient.71% similarUnverifiedThe reranking module typically uses a cross-encoder model to compute relevance scores between queries and documents, performing a secondary ranking of the candidate documents from the initial retrieval.69% similarUnverifiedFunnel-style layered evaluation architecture: the first layer uses cheap and fast semantic verifiers to pre-screen large volumes of traces, and the second layer passes ambiguous or high-value samples to a more powerful judge model for detailed evaluation.69% similarUnverified在可验证的优化任务中，搜索过程会被评估信号直接影响——哪些候选解被保留、哪些分支被剪枝，都依赖于此前评估过的结果；因此档案中记录的历史评估信息不是可有可无的元数据，而是决定下一步搜索行为的输入。68% similarUnverifiedThis solution employs multi-step verified entity extraction, extracting entities in a stepwise and verifiable manner to reduce error accumulation caused by one-shot extraction.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1048186API
curl https://kongchang.com/api/v1/knowledge/claims/1048186MCP
get_claim(id=1048186)