AI Paper Atlas

Research map / Image generation / Editing and style transfer

I2I-Bench: A Comprehensive Benchmark Suite for Image-to-Image Editing Models

Juntong Wang, Jiarui Wang, Huiyu Duan, Jiaxiang Kang, Guangtao Zhai, Xiongkuo Min

CVPR 2026 · Image generation › Editing and style transfer · approach: Instruction-based diffusion editing · 9 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use specialist model (ArtiMuse) instead of pure LMM for aesthetic quality (ablation)I2I-Bench evaluation pipeline · SE Aesthetic-QualityPearson rho with human preference: -0.0643 -> 0.9889Improved
use hybrid LMM-specialist feature matching instead of pure LMM for subject identity (ablation)I2I-Bench evaluation pipeline · SE Subject-IdentityPearson rho with human preference: -0.3494 -> 0.9133Improved
use specialist Q-Insight instead of pure LMM for image quality in multi-image editing (ablation)I2I-Bench evaluation pipeline · ME Image-QualityPearson rho with human preference: -0.9202 -> 0.6579Improved
use hybrid pipeline instead of pure LMM overall (ablation)I2I-Bench evaluation pipeline · SE overallPearson rho with human preference: 0.7277 -> 0.9425Improved
use hybrid pipeline instead of pure LMM overall (ablation)I2I-Bench evaluation pipeline · ME overallPearson rho with human preference: 0.6829 -> 0.8688Improved
use LMM VQA-5level versus specialist for Blending-Naturalness and Instruction-Following (same method used) (ablation)I2I-Bench evaluation pipeline · SE Blending-NaturalnessPearson rho with human preference: 0.8866 -> 0.8866No clear effect
I2I-Bench hybrid evaluation instead of LMM4Editevaluation metric · SE 15 dimensions averagePearson rho with human preference: 0.5968 -> 0.9425Improved
LMM4Edit versus I2I-Bench on Spatial-Accuracyevaluation metric · SE Spatial-AccuracyPearson rho with human preference: 0.9655 -> 0.9303Hurt
switch from single-image to multi-image editingQwen-Image-Edit-2509 · I2I-Bench Blending-Naturalnessscore: 0.807 -> 0.556Hurt
switch from single-image to multi-image editingQwen-Image-Edit-2509 · I2I-Bench Instruction-Followingscore: 0.930 -> 0.744Hurt
increase task cognitive complexity from Local Editing to World KnowledgeQwen-Image-Edit-2509 · I2I-Bench SE Instruction-Followingscore: 0.992 -> 0.805Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy