dungba88 commented on PR #14009: URL: https://github.com/apache/lucene/pull/14009#issuecomment-2903029304
> 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 when the results got trimmed down. For the alternative implementation we leave that up to the final hit collection to return the top-k based on a ranking mechanism (as this Query is usually just part of the final Query we send to IndexSearcher) while this implementation of the Query would output top-k by its own. If the final score is determined independently from the score returned by this Query, then the re-ranking we did would be a waste (and we can just over-sample and let the final scorer to do the work)? With that being said, I have no strong opinion. Wondering what would make sense for the actual use case that it will use (quantization and late-interaction model). We can also use special value (like negative number) to tell us not to trim down the results if we do need both way. -- 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: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
