- Notifications
You must be signed in to change notification settings - Fork 25.6k
Closed
Labels
:Search Relevance/VectorsVector searchVector searchenhancement"" muted="" aria-describedby="MDU6TGFiZWwyMzE3NA==-tooltip :R5b96b:">>enhancementTeam:Search RelevanceMeta label for the Search Relevance team in ElasticsearchMeta label for the Search Relevance team in Elasticsearch
Description
Description
We know to achieve adequate recall in a distributed vector index with bbq
oversampling and rescoring is required. Let's have that on by default for users.
Benchmarking shows that a default around 3
gives a nice balance of query latency and recall.
Of course, this should be overridable at query time, and if the user wants to provide their own oversampling defaults, they should be able to do that at the mapping level.
Related: #124581
Metadata
Metadata
Assignees
Labels
:Search Relevance/VectorsVector searchVector searchenhancement"" muted="" aria-describedby="MDU6TGFiZWwyMzE3NA==-tooltip :Ra5pmb:">>enhancementTeam:Search RelevanceMeta label for the Search Relevance team in ElasticsearchMeta label for the Search Relevance team in Elasticsearch