AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

EditHF-1M: A Million-Scale Rich Human Preference Feedback for Image Editing

Zitong Xu, Huiyu Duan, Zhongpeng Ji, Xinyun Zhang, Yutao Liu, Xiongkuo Min, Ke Gu, Jian Zhang, Shusong Xu, Jinwei Chen, Bo Li, Guangtao Zhai

ECCV 2026 · Image generation › Text-to-image diffusion · approach: Preference optimization for diffusion · 5 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
add MLP score head (ablation)EditHF (InternVL3.5, pointwise) · EditHF-1M Visual QualitySRCC (global): 0.7184 -> 0.7762Improved
add pairwise training on top of pointwise training (ablation)EditHF · EditHF-1M Visual QualitySRCC (global): 0.7762 -> 0.8548Improved
add pointwise training on top of pairwise training (ablation)EditHF · EditHF-1M Visual QualitySRCC (global): 0.8018 -> 0.8548Improved
pairwise-only vs pointwise-only training (ablation)EditHF · EditHF-1M Visual Qualitypairwise accuracy: 0.7826 (pointwise) vs 0.8663 (pairwise)Mixed
use InternVL3.5 instead of InternVL3 backbone (ablation)EditHF · EditHF-1M Visual QualitySRCC (global): 0.8316 -> 0.8548Improved
use InternVL3.5 instead of Qwen3-VL backbone (ablation)EditHF · EditHF-1M Editing AlignmentSRCC (global): 0.8231 -> 0.8488Improved
fine-tune MLLM on EditHF-1MInternVL3 (8B) · EditHF-1M Visual QualitySRCC (global): 0.2462 -> 0.8044Improved
use EditHF vs EditReward on unseen benchmarkEditHF · ImagenHubSRCC: 0.3520 -> 0.4526Improved
fine-tune Qwen-Image-Edit with Self-DPO from EditHFQwen-Image-Edit · EditHF-1M human overalloverall score: 75.39 -> 78.36Improved
fine-tune Qwen-Image-Edit with Global-DPO using multi-model pairsQwen-Image-Edit · EditHF-1M human overalloverall score: 75.39 -> 84.26Improved
Global-DPO instead of Self-DPO (ablation)Qwen-Image-Edit · EBench-18K human overalloverall score: 73.80 -> 74.18Improved
Self-DPO on AROURA-Bench judged by EditHFQwen-Image-Edit · AROURA-Bench EditHF scoreoverall score: 68.47 -> 69.26Improved

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