Abstract
A unified framework, World Snapshot Models, is proposed to train and perform nearly all visual tasks by reformulating them as generation tasks, validated across over 1,000 visual tasks.
In this paper, we propose a Unified Law for Visual Tasks, which could be further divided into Four Laws of Computer Vision, with the goal of enabling a single framework or method to uniformly train on and perform nearly all visual tasks. Our starting point is an inspiration and hypothesis: if the world information is encoded into image, which we call World Snapshots, then image processing is world processing, and image generation is World Snapshot generation. Furthermore, if all visual and visualization tasks can be reformulated as generation tasks, then a unified method, World Snapshot Models (WSMs), can perform the vast majority of visual tasks, much as large language models have largely unified natural language processing. This paper validates these laws across more than 1,000 trained and evaluated visual tasks (defined by distinct input-output mappings) and presents a unified training framework for future visual-task research and applications.
Community
I am very pleased to share our new work, the Visual Task Unification Law, which proposes a visual unification law. Following this law, a model called the Unified World Model can be created, unifying thousands of visual tasks. Specifically, we unified over 1148 visual tasks, of which hundreds were new visual tasks. Therefore, we also expanded the traditional field of computer vision by approximately 70%, which will to some extent promote the development of CV.
This work began in December 2024. We attempted to develop over 60 visual architectures, over 198 model variants/improvements, and trained them sequentially for 2293 times before summarizing the current version. This is a real improvement. We have for the first time achieved ultra-large-scale transformations from any visual modality to any visual modality, and from any visual modality to any non-visual modality; arbitrary visual perception; exponential visual expansion; and exponentially scalable visual chains. The first thought chain appeared in a pure visual model.
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- OmniTaskonomy: When Does Visual Generation Improve Visual Understanding? (2026)
- Not All Error Yields to Scale: Where Scaling Stops in Vision-Language Inference (2026)
- JUICE: Unified Multimodal Modeling in Pretrained Representation Spaces (2026)
- Unbundling Multimodal Vision Encoding (2026)
- Pixel is All You Need (2026)
- WOVEN: Weaving Visual World Modeling into Multimodal LLMs (2026)
- Where Do Vision Language Models Fail at 3D Reasoning? (2026)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: @librarian-bot recommend
Get this paper in your agent:
hf papers read 2611.24917 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper
Collections including this paper 0
No Collection including this paper