---
license: apache-2.0
tags:
- code
size_categories:
- 1K<n<10K
dataset_info:
  features:
  - name: id
    dtype: string
  - name: source_prompt
    dtype: string
  - name: target_prompt
    dtype: string
  - name: image
    dtype: image
  - name: image_width
    dtype: int32
  - name: image_height
    dtype: int32
  splits:
  - name: test
    num_bytes: 29728028
    num_examples: 80
  download_size: 29723807
  dataset_size: 29728028
configs:
- config_name: default
  data_files:
  - split: test
    path: data/test-*
---

<div align="center">

<h1>SNR-Edit: Structure-Aware Noise Rectification for Inversion-Free Flow-Based Editing</h1>

<div>
  <strong>Lifan Jiang</strong> &nbsp;
  <strong>Boxi Wu</strong> &nbsp;
  <strong>Yuhang Pei</strong> &nbsp;
  <strong>Tianrun Wu</strong> &nbsp;
  <br>
  <strong>Yongyuan Chen</strong> &nbsp;
  <strong>Yan Zhao</strong> &nbsp;
  <strong>Shiyu Yu</strong> &nbsp;
  <strong>Deng Cai</strong>
</div>

<div>
  State Key Lab of CAD&CG, Zhejiang University
</div>

<div>
  <em>Submitted to ICML 2026</em>
</div>

<br>

<p>
  <a href="https://arxiv.org/abs/2601.19180">
    <img src="https://img.shields.io/badge/Paper-arXiv-b31b1b?style=flat&logo=arxiv&logoColor=white" alt="Paper">
  </a>
  &nbsp;&nbsp;
  <a href="http://47.94.209.197:5008/">
    <img src="https://img.shields.io/badge/Project-Page-20BEFF?style=flat&logo=google-chrome&logoColor=white" alt="Project Page">
  </a>
  &nbsp;&nbsp;
  <a href="https://github.com/Tankowa/SNR-Edit">
    <img src="https://img.shields.io/badge/GitHub-Code-181717?style=flat&logo=github&logoColor=white" alt="GitHub">
  </a>
</p>

</div>

---

## 🚧 Coming Soon

SNR-Bench comprises 80 high-quality image-editing cases. Approximately 50% are sampled from PIE-Bench to ensure continuity with standard benchmarks, and the remaining 50% are collected from the web to introduce richer textures and more complex real-world scenes. We cover four editing operations: adjust, change, remove, and add. To minimize ambiguity and improve instruction consistency, all editing instructions for the non--PIE-Bench subset are manually written, refined, and verified through human annotation.

The dataset is currently being prepared and will be released soon. 

This repository contains the **SNR-Bench** dataset used in the paper *SNR-Edit: Structure-Aware Noise Rectification for Inversion-Free Flow-Based Editing*.

## 📌 Citation

If you find this dataset helpful, please consider citing:

```bibtex
@misc{jiang2026snreditstructureawarenoiserectification,
      title={SNR-Edit: Structure-Aware Noise Rectification for Inversion-Free Flow-Based Editing}, 
      author={Lifan Jiang and Boxi Wu and Yuhang Pei and Tianrun Wu and Yongyuan Chen and Yan Zhao and Shiyu Yu and Deng Cai},
      year={2026},
      eprint={2601.19180},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2601.19180}, 
}