Developing Visual Augmented Q&A System using Scalable Vision Embedding Retrieval & Late Interaction Re-ranker

TL;DR

This paper presents a scalable and efficient multimodal retrieval system for vision-augmented Q&A, combining hybrid search and late interaction re-ranking to improve speed and stability without sacrificing accuracy.

Contribution

It introduces a multi-step hybrid search approach with late interaction re-ranking to enhance large-scale multimodal retrieval for enterprise-grade Q&A systems.

Findings

Significant speed-up in retrieval process.

Stable performance without quality degradation.

Suitable for production deployment in enterprises.

Abstract

Traditional information extraction systems face challenges with text only language models as it does not consider infographics (visual elements of information) such as tables, charts, images etc. often used to convey complex information to readers. Multimodal LLM (MLLM) face challenges of finding needle in the haystack problem i.e., either longer context length or substantial number of documents as search space. Late interaction mechanism over visual language models has shown state of the art performance in retrieval-based vision augmented Q&A tasks. There are yet few challenges using it for RAG based multi-modal Q&A. Firstly, many popular and widely adopted vector databases do not support native multi-vector retrieval. Secondly, late interaction requires computation which inflates space footprint and can hinder enterprise adoption. Lastly, the current state of late interaction mechanism does not leverage the approximate neighbor search indexing methods for large speed ups in retrieval process. This paper explores a pragmatic approach to make vision retrieval process scalable and efficient without compromising on performance quality. We propose multi-step custom implementation utilizing widely adopted hybrid search (metadata & embedding) and state of the art late interaction re-ranker to retrieve best matching pages. Finally, MLLM are prompted as reader to generate answers from contextualized best matching pages. Through experiments, we observe that the proposed design is scalable (significant speed up) and stable (without degrading performance quality), hence can be used as production systems at enterprises.