Saved in:
Bibliographic Details
Main Authors: Taraday, Mitchell Keren, Wagner, Shahaf, Baskin, Chaim
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.06820
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908845250445312
author Taraday, Mitchell Keren
Wagner, Shahaf
Baskin, Chaim
author_facet Taraday, Mitchell Keren
Wagner, Shahaf
Baskin, Chaim
contents Multimodal retrieval still leans on embedding-based models like CLIP for fast vector search over pre-computed image embeddings. Yet, unlike text retrieval, where joint-encoder rerankers are standard, comparable vision-language rerankers are largely absent. We find that seminal joint encoders such as BLIP are severely bottlenecked by an expensive visual feature-extraction stage, preventing practical deployment at scale. Motivated by this bottleneck, we introduce EDJE, an Efficient Discriminative Joint Encoder that precomputes vision tokens offline and compresses them via a lightweight attention-based adapter, so online inference runs only a compact joint encoder over a small set of visual tokens plus the text. EDJE preserves strong retrieval performance while drastically reducing storage and online compute, enabling high-throughput inference. Specifically, EDJE processes 50k image--text pairs/second while requiring 49kB of disk storage per image, matching prior art on Flickr (zero-shot) and COCO (fine-tuned) retrieval.
format Preprint
id arxiv_https___arxiv_org_abs_2510_06820
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Efficient Discriminative Joint Encoders for Large Scale Vision-Language Reranking
Taraday, Mitchell Keren
Wagner, Shahaf
Baskin, Chaim
Computer Vision and Pattern Recognition
Machine Learning
Multimodal retrieval still leans on embedding-based models like CLIP for fast vector search over pre-computed image embeddings. Yet, unlike text retrieval, where joint-encoder rerankers are standard, comparable vision-language rerankers are largely absent. We find that seminal joint encoders such as BLIP are severely bottlenecked by an expensive visual feature-extraction stage, preventing practical deployment at scale. Motivated by this bottleneck, we introduce EDJE, an Efficient Discriminative Joint Encoder that precomputes vision tokens offline and compresses them via a lightweight attention-based adapter, so online inference runs only a compact joint encoder over a small set of visual tokens plus the text. EDJE preserves strong retrieval performance while drastically reducing storage and online compute, enabling high-throughput inference. Specifically, EDJE processes 50k image--text pairs/second while requiring 49kB of disk storage per image, matching prior art on Flickr (zero-shot) and COCO (fine-tuned) retrieval.
title Efficient Discriminative Joint Encoders for Large Scale Vision-Language Reranking
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2510.06820