Module maxsim 

MaxSim scoring for ColBERT-style late interaction models.

MaxSim computes similarity between multi-vector embeddings by taking the maximum similarity for each query token across all document tokens:

MaxSim(Q, D) = sum_{q_i in Q} max_{d_j in D} (q_i · d_j)

This module provides:

• maxsim - Compute MaxSim score between two multi-vectors
• maxsim_batch - Compute MaxSim scores against multiple documents
• MaxSimScorer - Reusable scorer with pre-computed query data
• [MaxSimScan] - Operator for multi-vector similarity search

Structs

MaxSimScorer

A reusable MaxSim scorer with pre-computed query data.

Functions

distance_to_maxsim

Convert a distance back to MaxSim score.

maxsim

Compute the MaxSim score between a query and document multi-vector.

maxsim_batch

Compute MaxSim scores between a query and multiple documents.

maxsim_cosine

Compute MaxSim with normalized vectors (for cosine-like similarity).

maxsim_to_distance

Convert MaxSim score to a distance (lower is more similar).