Youtu-VL: Unleashing Visual Potential via Unified Vision-Language Supervision

TL;DR

Youtu-VL introduces a unified autoregressive supervision framework that enhances visual detail retention in vision-language models, enabling more comprehensive multimodal understanding and vision-centric task performance.

Contribution

It proposes VLUAS, a novel training paradigm shifting from vision-as-input to vision-as-target, improving visual detail modeling in VLMs without task-specific modifications.

Findings

Achieves competitive results on multimodal tasks

Enables vision-centric tasks without additional modules

Provides a robust foundation for generalist visual agents

Abstract

Despite the significant advancements represented by Vision-Language Models (VLMs), current architectures often exhibit limitations in retaining fine-grained visual information, leading to coarse-grained multimodal comprehension. We attribute this deficiency to a suboptimal training paradigm inherent in prevailing VLMs, which exhibits a text-dominant optimization bias by conceptualizing visual signals merely as passive conditional inputs rather than supervisory targets. To mitigate this, we introduce Youtu-VL, a framework leveraging the Vision-Language Unified Autoregressive Supervision (VLUAS) paradigm, which fundamentally shifts the optimization objective from ``vision-as-input'' to ``vision-as-target.'' By integrating visual tokens directly into the prediction stream, Youtu-VL applies unified autoregressive supervision to both visual details and linguistic content. Furthermore, we extend this paradigm to encompass vision-centric tasks, enabling a standard VLM to perform vision-centric tasks without task-specific additions. Extensive empirical evaluations demonstrate that Youtu-VL achieves competitive performance on both general multimodal tasks and vision-centric tasks, establishing a robust foundation for the development of comprehensive generalist visual agents.