Saved in:
Bibliographic Details
Main Authors: Lin, Bin, Li, Zongjian, Cheng, Xinhua, Niu, Yuwei, Ye, Yang, He, Xianyi, Yuan, Shenghai, Yu, Wangbo, Wang, Shaodong, Ge, Yunyang, Pang, Yatian, Yuan, Li
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.03147
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908413095575552
author Lin, Bin
Li, Zongjian
Cheng, Xinhua
Niu, Yuwei
Ye, Yang
He, Xianyi
Yuan, Shenghai
Yu, Wangbo
Wang, Shaodong
Ge, Yunyang
Pang, Yatian
Yuan, Li
author_facet Lin, Bin
Li, Zongjian
Cheng, Xinhua
Niu, Yuwei
Ye, Yang
He, Xianyi
Yuan, Shenghai
Yu, Wangbo
Wang, Shaodong
Ge, Yunyang
Pang, Yatian
Yuan, Li
contents Although existing unified models achieve strong performance in vision-language understanding and text-to-image generation, they remain limited in addressing image perception and manipulation -- capabilities increasingly demanded in practical applications. Recently, OpenAI introduced the powerful GPT-4o-Image model, which showcases advanced capabilities in comprehensive image perception and manipulation, sparking widespread interest. Through carefully designed experiments, we observe that GPT-4o-Image likely relies on semantic encoders rather than VAEs for feature extraction, despite VAEs being commonly regarded as crucial for image manipulation tasks. Inspired by this insight, we propose UniWorld-V1, a unified generative framework built upon semantic features extracted from powerful multimodal large language models and contrastive semantic encoders. Using only 2.7M training data, UniWorld-V1 achieves impressive performance across diverse tasks, including image understanding, generation, manipulation, and perception. We fully open-source the UniWorld-V1 framework, including model weights, training and evaluation scripts, and datasets to promote reproducibility and further research.
format Preprint
id arxiv_https___arxiv_org_abs_2506_03147
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle UniWorld-V1: High-Resolution Semantic Encoders for Unified Visual Understanding and Generation
Lin, Bin
Li, Zongjian
Cheng, Xinhua
Niu, Yuwei
Ye, Yang
He, Xianyi
Yuan, Shenghai
Yu, Wangbo
Wang, Shaodong
Ge, Yunyang
Pang, Yatian
Yuan, Li
Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
Although existing unified models achieve strong performance in vision-language understanding and text-to-image generation, they remain limited in addressing image perception and manipulation -- capabilities increasingly demanded in practical applications. Recently, OpenAI introduced the powerful GPT-4o-Image model, which showcases advanced capabilities in comprehensive image perception and manipulation, sparking widespread interest. Through carefully designed experiments, we observe that GPT-4o-Image likely relies on semantic encoders rather than VAEs for feature extraction, despite VAEs being commonly regarded as crucial for image manipulation tasks. Inspired by this insight, we propose UniWorld-V1, a unified generative framework built upon semantic features extracted from powerful multimodal large language models and contrastive semantic encoders. Using only 2.7M training data, UniWorld-V1 achieves impressive performance across diverse tasks, including image understanding, generation, manipulation, and perception. We fully open-source the UniWorld-V1 framework, including model weights, training and evaluation scripts, and datasets to promote reproducibility and further research.
title UniWorld-V1: High-Resolution Semantic Encoders for Unified Visual Understanding and Generation
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2506.03147