🔥 SPHINX: A Mixer of Tasks, Domains, and Embeddings

Official implementation of 'SPHINX: A Mixer of Tasks, Domains, and Embeddings Advances Multi-modal Large Language Models'.

Try out our web demo 🚀 here!

Github link: Github • 👋 join our WeChat

Introduction

We present SPHINX, a versatile multi-modal large language model (MLLM) with a mixer of training tasks, data domains, and visual embeddings.

• Task Mix. For all-purpose capabilities, we mix a variety of vision-language tasks for mutual improvement: VQA, REC, REG, OCR, DET, POSE, REL DET, T2I, etc.

• Embedding Mix. We capture robust visual representations by fusing distinct visual architectures, pre-training, and granularity.

• Domain Mix. For data from real-world and synthetic domains, we mix the weights of two domain-specific models for complementarity.

On top of SPHINX, we propose to further mix visual scales and sub-images for better capture fine-grained semantics on high-resolution images.

Installation

SPHINX is built upon LLaMA2-Accessory, please follow the instructions here for environment setup.

Inference

This section provides a step-by-step guide for hosting a local SPHINX demo. If you're already familiar with the LLAMA2-Accessory toolkit, note that hosting a SPHINX demo follows the same pipeline as hosting demos for the other models supported by LLAMA2-Accessory.

Weights

We provide the beta-version checkpoints on HuggingFace🤗. Please download them to your own machine. The file structure should appear as follows:

ckpt_path/
├── consolidated.00-of-02.model.pth
└── consolidated.01-of-02.model.pth

Host Local Demo

Please follow the instructions here to see the instruction and complete the use of the model.

Result

We provide a comprehensive evaluation of SPHINX and showcase results across multiple benchmarks.

Our evaluation encompasses both quantitative metrics and qualitative assessments, providing a holistic understanding of our VLM model's performance.

Evaluation Prompt Design

• In evaluation, we prioritize aligning with each benchmark's desired output format.
• We employ distinct prompts tailored to benchmarks that necessitate long answers, short answers, and multiple-choice responses.
• For tasks involving visual grounding, we directly utilize the prompts during training to enhance the model's performance on these particular challenges.

Benchmarks on Multimodal Large Language Models

We test our model on recently proposed MLLM benchmarks which is based on VQA to comprehensive evaluation of the model's characteristic such as MME, Seedbench, POPE, LLaVA-Bench (In-the-Wild), MM-Vet, MathVista, MMbench, CCbench.

The Long-SPHINX achieve new stat of arts result on 5 out of 9 benchmarks

Visual Question Answering

• We evaluate general VQA benchmarks, such as VQAV2, OKVQA, GQA, vizwiz, scienceQA, visual spatial reasoning (VSR), IconQA.
• Additionally, we conduct experiments on Text-oriented VQA such as TextVQA,OCR-VQA.
• Long-Sphinx achieve comparative results across all benchmarks. We observe that Long-Sphinx outperforms Sphinx in VQA datasets that demand fine-grained visual information, showcasing the effectiveness of our visual mixed-up approach for achieving high resolution without relying on a visual encoder trained specifically on high-resolution images.

Visual Grounding

• The SPHINX model and baseline models on REC benchmarks results on table4.
• SPHINX exhibits robust performance in visual grounding tasks such as RefCOCO, RefCOCO+, and RefCOCOg, surpassing other vision-language generalist models.
• Notably, SPHINX outperforms specialist models G-DINO-L by more than 1.54% in accuracy across all tasks within RefCOCO/RefCOCO+/RefCOCOg.

Frequently Asked Questions (FAQ)

❓ Encountering issues or have further questions? Find answers to common inquiries here. We're here to assist you!

License

Llama 2 is licensed under the LLAMA 2 Community License, Copyright (c) Meta Platforms, Inc. All Rights Reserved.