Arrays of candidates, each sorted by descending relevance. Each list represents results from one retrieval method.
Optionalk: numberSmoothing constant. Higher values reduce the influence of top-ranked positions. Default: 60 (the standard value from the RRF paper and used by Azure AI Search, OpenSearch, and Elasticsearch).
Optionalweights: number[]Optional per-list weights aligned by index with rankedLists.
Lists with no entry in the array fall back to a weight of 1.0.
A weight of 0 suppresses that list's contribution entirely.
Omitting the parameter is equivalent to all-ones (canonical RRF).
Fused candidates sorted by descending RRF score.
Compute Reciprocal Rank Fusion across multiple ranked result lists, with optional per-list weights.
Unweighted formula:
FusedScore(d) = Σ_i 1 / (k + rank_i(d))Weighted formula:FusedScore(d) = Σ_i w_i / (k + rank_i(d))where
rank_i(d)is the 1-based rank of documentdin listi,kis a smoothing constant (default 60, per the original paper), andw_iis the optional weight for listi(default 1.0).Documents not present in a list receive no contribution from that list (they are simply absent, not penalized).