Enum FusionStrategy 

pub enum FusionStrategy {
    RRF {
        k: f32,
    },
    Linear {
        dense_weight: f32,
    },
    Convex {
        alpha: f32,
    },
    DBSF,
    Union,
    Intersection,
    ThreeWay {
        dense_weight: f32,
        sparse_weight: f32,
        multivector_weight: f32,
    },
}

Strategy for fusing dense and sparse retrieval results

Variants

RRF

Reciprocal Rank Fusion

Fields

k: f32

Constant k for RRF (typically 60)

Linear

Linear combination of normalized scores

Fields

dense_weight: f32

Weight for dense results (sparse weight = 1 - dense_weight)

Convex

Convex combination after min-max normalization

Fields

alpha: f32

Alpha parameter for convex combination

DBSF

Distribution-Based Score Fusion

Union

Take union of results, prefer dense ranking for ties

Intersection

Intersection: only return results in both

ThreeWay

Three-way weighted combination of dense, sparse, and multi-vector results

Fields

dense_weight: f32

Weight for dense results

sparse_weight: f32

Weight for sparse results

multivector_weight: f32

Weight for multi-vector results

Implementations

impl FusionStrategy

pub fn fuse( &self, dense_results: &[(ChunkId, f32)], sparse_results: &[(ChunkId, f32)], ) -> Vec<(ChunkId, f32)>

Fuse dense and sparse retrieval results

pub fn fuse_three( &self, dense: &[(ChunkId, f32)], sparse: &[(ChunkId, f32)], multivector: &[(ChunkId, f32)], ) -> Vec<(ChunkId, f32)>

Fuse three result sets: dense, sparse, and multi-vector.

This method combines results from all three retrieval methods using a weighted linear combination after min-max normalization.

Arguments

• dense - Dense (single-vector) retrieval results
• sparse - Sparse (BM25) retrieval results
• multivector - Multi-vector (ColBERT-style) retrieval results

Returns

Fused results sorted by combined score descending.

pub fn three_way( dense_weight: f32, sparse_weight: f32, multivector_weight: f32, ) -> Self

Create a three-way fusion strategy with the given weights.

Weights should ideally sum to 1.0 but this is not enforced.

Trait Implementations

impl Clone for FusionStrategy

fn clone(&self) -> FusionStrategy

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for FusionStrategy

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for FusionStrategy

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for FusionStrategy

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for FusionStrategy

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.