Dzung Bui

Thanks @vigyasharma for the comment! I updated the comment to make it less confusing. I'll think about generalization, but the idea is that as long as the field can expose...

No worry, I think generalization makes sense to me too. And I like the idea of moving this forward. The use of `KnnFloatVectorQuery` in this implementation only adds a small...

I changed the query to be generic but keep other the same. (Didn't have JDK24 yet so build would fail.)

> This gets handled by the k you pass to searcher.search(), and the top docs collector you use. It's same as any other query. I guess the main difference is...

Let me rephrase it better. I'm not worrying about the computation cost, as you mentioned they are the same. In Amazon (where both Vigya and I are working at), we...

I'm thinking that instead of using `vectorValues(...)` we can probably use DoubleValueSource as per @msokolov suggestion for a generic rescorer. Then we can use scorer from another (vector) field for...

I published another revision to support a custom DoubleValueSource for re-ranking, and we can reuse the `FloatVectorSimilarityValuesSource` in case we want to rescore with other field, or the `FullPrecisionFloatVectorSimilarityValuesSource` in...

> is when users want to change the matchset of a KnnVectorQuery using full-precision or other reranking Yes that's correct, @vigyasharma. We are using a hybrid search where KnnFloatVectorQuery and...

Sorry for spamming the replies! I should have gone to the Files changed tab, which allow sending all replies in the same message.

I'm wondering if ACORN would work for this use case: https://arxiv.org/abs/2403.04871