Visual Prompt Generators (VPGs): Encoding Images to LLM Tokens

Written by instancing | Published 2025/11/14
Tech Story Tags: visual-prompt-generator | mllm-architecture | q-former | perceiver-resampler | cross-attention | deep-learning | deep-learning-adapters | llm-tokens

TLDR

Explains how MLLMs use VPGs and cross-attention with learnable query embeddings to extract essential visual tokens from image patches for LLM inputvia the TL;DR App

Table of Links

Abstract and 1 Introduction

Related Work

2.1. Multimodal Learning

2.2. Multiple Instance Learning
Methodology

3.1. Preliminaries and Notations

3.2. Relations between Attention-based VPG and MIL

3.3. MIVPG for Multiple Visual Inputs

3.4. Unveiling Instance Correlation in MIVPG for Enhanced Multi-instance Scenarios
Experiments and 4.1. General Setup

4.2. Scenario 1: Samples with Single Image

4.3. Scenario 2: Samples with Multiple Images, with Each Image as a General Embedding

4.4. Scenario 3: Samples with Multiple Images, with Each Image Having Multiple Patches to be Considered and 4.5. Case Study
Conclusion and References

Supplementary Material

A. Detailed Architecture of QFormer

B. Proof of Proposition

C. More Experiments

3. Methodology

3.1. Preliminaries and Notations

Authors:

(1) Wenliang Zhong, The University of Texas at Arlington ([email protected]);

(2) Wenyi Wu, Amazon ([email protected]);

(3) Qi Li, Amazon ([email protected]);

(4) Rob Barton, Amazon ([email protected]);

(5) Boxin Du, Amazon ([email protected]);

(6) Shioulin Sam, Amazon ([email protected]);

(7) Karim Bouyarmane, Amazon ([email protected]);

(8) Ismail Tutar, Amazon ([email protected]);

(9) Junzhou Huang, The University of Texas at Arlington ([email protected]).

This paper is available on arxiv under CC by 4.0 Deed (Attribution 4.0 International) license.

Written by instancing | Pioneering instance management, driving innovative solutions for efficient resource utilization, and enabling a more sus

Published by HackerNoon on 2025/11/14