AMES: Approximate Multi-modal Enterprise Search via Late Interaction Retrieval

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Joseph, Tony, Pareja, Carlos, Pegna, David Lopes, Singh, Abhishek
Format:	Preprint
Published:	2026
Subjects:	Information Retrieval Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

_version_	1866911514695303168
author	Joseph, Tony Pareja, Carlos Pegna, David Lopes Singh, Abhishek
author_facet	Joseph, Tony Pareja, Carlos Pegna, David Lopes Singh, Abhishek
contents	We present AMES (Approximate Multimodal Enterprise Search), a unified multimodal late interaction retrieval architecture which is backend agnostic. AMES demonstrates that fine-grained multimodal late interaction retrieval can be deployed within a production grade enterprise search engine without architectural redesign. Text tokens, image patches, and video frames are embedded into a shared representation space using multi-vector encoders, enabling cross-modal retrieval without modality specific retrieval logic. AMES employs a two-stage pipeline: parallel token level ANN search with per document Top-M MaxSim approximation, followed by accelerator optimized Exact MaxSim re-ranking. Experiments on the ViDoRe V3 benchmark show that AMES achieves competitive ranking performance within a scalable, production ready Solr based system.
format	Preprint
id	arxiv_https___arxiv_org_abs_2603_13537
institution	arXiv
publishDate	2026
record_format	arxiv
spellingShingle	AMES: Approximate Multi-modal Enterprise Search via Late Interaction Retrieval Joseph, Tony Pareja, Carlos Pegna, David Lopes Singh, Abhishek Information Retrieval Machine Learning We present AMES (Approximate Multimodal Enterprise Search), a unified multimodal late interaction retrieval architecture which is backend agnostic. AMES demonstrates that fine-grained multimodal late interaction retrieval can be deployed within a production grade enterprise search engine without architectural redesign. Text tokens, image patches, and video frames are embedded into a shared representation space using multi-vector encoders, enabling cross-modal retrieval without modality specific retrieval logic. AMES employs a two-stage pipeline: parallel token level ANN search with per document Top-M MaxSim approximation, followed by accelerator optimized Exact MaxSim re-ranking. Experiments on the ViDoRe V3 benchmark show that AMES achieves competitive ranking performance within a scalable, production ready Solr based system.
title	AMES: Approximate Multi-modal Enterprise Search via Late Interaction Retrieval
topic	Information Retrieval Machine Learning
url	https://arxiv.org/abs/2603.13537