Papers › UniWorld-V1: High-Resolution Semantic Encoders for Unified Visual Understanding and Generation

UniWorld-V1: High-Resolution Semantic Encoders for Unified Visual Understanding and Generation

3 Jun 2025arXiv:2506.03147archive 2025-07-28

Bin Lin, Zongjian Li, Xinhua Cheng, Yuwei Niu, Yang Ye, Xianyi He, Shenghai Yuan, Wangbo Yu, Shaodong Wang, Yunyang Ge, Yatian Pang, Li Yuan

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.

PaperPDFCode

In Syntology View this paper on Syntology, its page in Syntology's graph. That page lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

PKU-YuanGroup/UniWorld-V1 mentioned on GitHubpytorchSyntology: not harvested report
pku-yuangroup/imgedit mentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to ImgEdit: A Unified Image Editing Dataset and Benchmark (arXiv:2505.20275; that paper's own run record: 2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Image EditingImage GenerationImage ManipulationText to Image GenerationText-to-Image Generation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
WISE UniWorld-V1 Biology 0.45 #1 of 11 Archive leaderboard report
WISE UniWorld-V1 Chemistry 0.41 #1 of 11 Archive leaderboard report
WISE UniWorld-V1 Cultural 0.53 #1 of 11 Archive leaderboard report
WISE UniWorld-V1 Overall 0.55 #1 of 11 Archive leaderboard report
WISE UniWorld-V1 Physics 0.59 #1 of 11 Archive leaderboard report
WISE UniWorld-V1 Space 0.73 #1 of 11 Archive leaderboard report
WISE UniWorld-V1 Time 0.55 #1 of 11 Archive leaderboard report
Image Editing ImgEdit-Data UniWorld-V1 Action 2.74 #2 of 9 Archive leaderboard report
Image Editing ImgEdit-Data UniWorld-V1 Add 3.82 #2 of 9 Archive leaderboard report
Image Editing ImgEdit-Data UniWorld-V1 Adjust 3.64 #2 of 9 Archive leaderboard report
Image Editing ImgEdit-Data UniWorld-V1 Background 2.99 #2 of 9 Archive leaderboard report
Image Editing ImgEdit-Data UniWorld-V1 Extract 2.27 #2 of 9 Archive leaderboard report
Image Editing ImgEdit-Data UniWorld-V1 Hybrid 2.96 #2 of 9 Archive leaderboard report
Image Editing ImgEdit-Data UniWorld-V1 Overall 3.26 #2 of 9 Archive leaderboard report
Image Editing ImgEdit-Data UniWorld-V1 Remove 3.24 #2 of 9 Archive leaderboard report
Image Editing ImgEdit-Data UniWorld-V1 Replace 3.47 #2 of 9 Archive leaderboard report
Image Editing ImgEdit-Data UniWorld-V1 Style 4.21 #2 of 9 Archive leaderboard report
Image Generation WISE UniWorld-V1 Biology 0.45 #4 of 14 Archive leaderboard report
Image Generation WISE UniWorld-V1 Chemistry 0.41 #4 of 14 Archive leaderboard report
Image Generation WISE UniWorld-V1 Cultural 0.53 #4 of 14 Archive leaderboard report
Image Generation WISE UniWorld-V1 Overall 0.55 #4 of 14 Archive leaderboard report
Image Generation WISE UniWorld-V1 Physics 0.59 #4 of 14 Archive leaderboard report
Image Generation WISE UniWorld-V1 Space 0.73 #4 of 14 Archive leaderboard report
Image Generation WISE UniWorld-V1 Time 0.55 #4 of 14 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 (Rewrite) Color Attri. 0.71 #2 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 (Rewrite) Colors 0.90 #2 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 (Rewrite) Counting 0.81 #2 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 (Rewrite) Overall 0.84 #2 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 (Rewrite) Position 0.74 #2 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 (Rewrite) Single Obj. 0.98 #2 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 (Rewrite) Two Obj. 0.93 #2 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 Color Attri. 0.70 #4 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 Colors 0.89 #4 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 Counting 0.79 #4 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 Overall 0.80 #4 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 Position 0.49 #4 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 Single Obj. 0.99 #4 of 20 Archive leaderboard report
Text-to-Image Generation GenEval UniWorld-V1 Two Obj. 0.93 #4 of 20 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections