Unlocking True Entity-Level Retrieval: New Array-of-Structs and MAX_SIM Capabilities in Milvus
Dec 5, 2025Jeremy Zhu, Min Tian

Summary
Introduces and explains Milvus's new Array of Structs field type, which allows a single record to store an ordered list of Struct elements, each capable of holding multiple vectors and scalar fields. Details the new MAX_SIM scoring metric, which computes entity-level similarity by taking the maximum similarity score between query vectors and all vectors within an Array of Structs field for a given entity. Demonstrates how these features together enable true entity-level retrieval, solving issues like fragmented search results and duplicate entities in RAG, e-commerce, and other multi-vector applications.
More from the AI Data Storage Engine trail
- Milvus Snapshots: Point-in-Time Collection Views Without Copying Data
- Announcing Milvus 3.0: Lake-Native Vector Search and a More Powerful Retrieval Engine
- How to Recover Milvus Data After etcd Corruption Without a Backup
- Open Tag: An Open-Source Claude Tag for Claude Code and Codex
- How Force Merge Compaction Nearly Doubled Milvus Search QPS