UniREditBench: A Unified Reasoning-based Image Editing Benchmark

Feng Han1,2*, Yibin Wang1,2*, Chenglin Li2,3, Zheming Liang2, Dianyi Wang1,2,
Yang Jiao1, Zhipeng Wei4, Chao Gong1, Cheng Jin1,2, Jingjing Chen1†, Jiaqi Wang2†
1Fudan University, 2Shanghai Innovation Intuition, 3Zhejiang University 4UC Berkeley
Paper 😊 UniREditBench Code
😊 UniREdit-Data-100K 😊 UniREdit-Bagel 😊 Leaderboard

UniREditBench Overview

pipeline
pipeline

Benchmark Comparison

data-overview

Reasoning-based image editing benchmark comparison.
UniREditBench excels in broader scenario and evaluation dimension coverage.

data-overview

Image editing evaluation comparison.
Current text-reference-only evaluation potentially leads to misjudging, while our dual-reference evaluation results in more reliable assessments.

Evaluation Dimensions

data-overview

Qualitative Cases.
We present qualitative examples for each dimension across both real-world and game-world scenarios.

Multi-scenario Data Synthesis Pipeline

pipeline

Benchmarking Results on UniREditBench

pipeline

UniREDit-Data-100K

pipeline

UniREdit-Bagel

pipeline

Out-of-distribution Results Comparison on RISEBench

pipeline

Out-of-distribution Results Comparison on KRISBench

pipeline

BibTeX