We have developed https://github.com/tensorchord/qtext to explore how reranking improve the quality of hybrid search. The results can be visualized in the terminal powered by https://github.com/textualize/textual . Dense/sparse/text search provides different insight of the query. The reranking step has dramatically improved quality of the top-k candidates. You can try different queries to see which retrieval method suit your requirements better.