[andrewlu0]

Will check this out - it seems like its doing two separate searches where you specify two separate top K values. Curious what the trade offs are between this approach and weighting sparse/dense vectors