benwtrent commented on PR #15564: URL: https://github.com/apache/lucene/pull/15564#issuecomment-3745711051
I am gonna rerun the 1M 1024 case to see how it fairs with "fake reranking" (e.g. just gathering some extra K). It is pretty neat to see how simply having query level refinement can improve recall. Makes me think we need to unlock allowing refinement with higher level bits on the query side generally. That could give nice recall improvements without having to rescore against the higher bits of the indexed vectors. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
