VisionLLM v2: An End-to-End Generalist Multimodal Large Language Model for Hundreds of Vision-Language Tasks

TL;DR

VisionLLM v2 is a comprehensive multimodal large language model that unifies visual perception, understanding, and generation, enabling it to perform a wide range of vision-language tasks with a single, end-to-end trained framework.

Contribution

It introduces the 'super link' mechanism for flexible task information transmission and demonstrates end-to-end training on hundreds of diverse vision-language tasks.

Findings

Achieves performance comparable to task-specific models across many tasks.

Supports diverse vision tasks including VQA, object localization, pose estimation, and image editing.

Effectively resolves multi-task training conflicts.

Abstract

We present VisionLLM v2, an end-to-end generalist multimodal large model (MLLM) that unifies visual perception, understanding, and generation within a single framework. Unlike traditional MLLMs limited to text output, VisionLLM v2 significantly broadens its application scope. It excels not only in conventional visual question answering (VQA) but also in open-ended, cross-domain vision tasks such as object localization, pose estimation, and image generation and editing. To this end, we propose a new information transmission mechanism termed "super link", as a medium to connect MLLM with task-specific decoders. It not only allows flexible transmission of task information and gradient feedback between the MLLM and multiple downstream decoders but also effectively resolves training conflicts in multi-tasking scenarios. In addition, to support the diverse range of tasks, we carefully collected and combed training data from hundreds of public vision and vision-language tasks. In this way, our model can be joint-trained end-to-end on hundreds of vision language tasks and generalize to these tasks using a set of shared parameters through different user prompts, achieving performance comparable to task-specific models. We believe VisionLLM v2 will offer a new perspective on the generalization of MLLMs.