UniRAG: Universal Retrieval Augmentation for Large Vision Language Models

TL;DR

UniRAG is a versatile retrieval augmentation method that improves large vision-language models' output quality by incorporating relevant retrieved information into prompts during inference, benefiting both common and uncommon entity understanding.

Contribution

UniRAG introduces a plug-and-play retrieval augmentation technique that enhances the performance of various large vision-language models across multiple tasks.

Findings

Significant quality improvements in image captioning and visual question answering.

Effective augmentation benefits both proprietary and open-source models.

Retrieval augmentation enhances understanding of common entities in vision-language tasks.

Abstract

Recently, Large Vision Language Models (LVLMs) have unlocked many complex use cases that require Multi-Modal (MM) understanding (e.g., image captioning or visual question answering) and MM generation (e.g., text-guided image generation or editing) capabilities. To further improve the output fidelityof LVLMs we introduce UniRAG, a plug-and-play technique that adds relevant retrieved information to prompts as few-shot examples during inference. Unlike the common belief that Retrieval Augmentation (RA) mainly improves generation or understanding of uncommon entities, our evaluation results on the MSCOCO dataset with common entities show that both proprietary models like GPT-4o and Gemini-Pro and smaller open-source models like LLaVA, LaVIT, and Emu2 significantly enhance their generation quality when their input prompts are augmented with relevant information retrieved by Vision-Language (VL) retrievers like UniIR models. All the necessary code to reproduce our results is available at https://github.com/castorini/UniRAG