Reciprocal rank fusion

RrfQuery merges several result lists by rank, not by score, so the children do not need score normalisation.

Score formula

score(d) = Σ 1 / (k + rank_i(d))

Where rank_i(d) is the 1-based position of document d in child query i's results. k defaults to 60.

Combining text and vector results

using Rowles.LeanLucene.Search.Queries;

var rrf = new RrfQuery(k: 60)
    .Add(new TermQuery("body", "machine"))
    .Add(new VectorQuery("embedding", queryVector, topK: 50));

var hits = searcher.Search(rrf, topN: 10);

Tuning k

Higher k flattens the contribution from top-ranked documents. The default of 60 is the value used in the original RRF paper.

Combining pre-computed TopDocs

For ad-hoc fusion (e.g., results from external systems), use the static helper:

var fused = RrfQuery.Combine(new[] { topDocsA, topDocsB }, topN: 10, k: 60);

See also

• RrfQuery