MultiVectorAggregation

Enums · AiDotNet.Enums

How a multi-vector retriever combines the query similarities of one document’s vectors into a single score.

For Beginners

A document can be stored as several vectors (for example a summary followed by its sections). This setting decides how their individual match scores become one document score.

Fields

Field Summary
MultiVectorAggregation Max The best-matching vector’s score (recommended for precise queries).
MultiVectorAggregation Mean The average score of the matched vectors (recommended for general queries).
MultiVectorAggregation Weighted A weighted average where vector N (from the “docId_vector_N” id) has weight 1 / (N + 1), so earlier vectors such as summaries count more than later details.
int value__