ScaleEdit-12M: Scaling Open-Source Image Editing Data Generation via Multi-Agent Framework
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866908908482723840 |
|---|---|
| author | Chen, Guanzhou Cui, Erfei Tian, Changyao Yang, Danni Yang, Ganlin Qiao, Yu Li, Hongsheng Luo, Gen Zhang, Hongjie |
| author_facet | Chen, Guanzhou Cui, Erfei Tian, Changyao Yang, Danni Yang, Ganlin Qiao, Yu Li, Hongsheng Luo, Gen Zhang, Hongjie |
| contents | Instruction-based image editing has emerged as a key capability for unified multimodal models (UMMs), yet constructing large-scale, diverse, and high-quality editing datasets without costly proprietary APIs remains challenging. Previous image editing datasets either rely on closed-source models for annotation, which prevents cost-effective scaling, or employ fixed synthetic editing pipelines, which suffer from limited quality and generalizability. To address these challenges, we propose ScaleEditor, a fully open-source hierarchical multi-agent framework for end-to-end construction of large-scale, high-quality image editing datasets. Our pipeline consists of three key components: source image expansion with world-knowledge infusion, adaptive multi-agent editing instruction-image synthesis, and a task-aware data quality verification mechanism. Using ScaleEditor, we curate ScaleEdit-12M, the largest open-source image editing dataset to date, spanning 23 task families across diverse real and synthetic domains. Fine-tuning UniWorld-V1 and Bagel on ScaleEdit yields consistent gains, improving performance by up to 10.4% on ImgEdit and 35.1% on GEdit for general editing benchmarks and by up to 150.0% on RISE and 26.5% on KRIS-Bench for knowledge-infused benchmarks. These results demonstrate that open-source, agentic pipelines can approach commercial-grade data quality while retaining cost-effectiveness and scalability. Both the framework and dataset will be open-sourced. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_20644 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | ScaleEdit-12M: Scaling Open-Source Image Editing Data Generation via Multi-Agent Framework Chen, Guanzhou Cui, Erfei Tian, Changyao Yang, Danni Yang, Ganlin Qiao, Yu Li, Hongsheng Luo, Gen Zhang, Hongjie Computer Vision and Pattern Recognition Instruction-based image editing has emerged as a key capability for unified multimodal models (UMMs), yet constructing large-scale, diverse, and high-quality editing datasets without costly proprietary APIs remains challenging. Previous image editing datasets either rely on closed-source models for annotation, which prevents cost-effective scaling, or employ fixed synthetic editing pipelines, which suffer from limited quality and generalizability. To address these challenges, we propose ScaleEditor, a fully open-source hierarchical multi-agent framework for end-to-end construction of large-scale, high-quality image editing datasets. Our pipeline consists of three key components: source image expansion with world-knowledge infusion, adaptive multi-agent editing instruction-image synthesis, and a task-aware data quality verification mechanism. Using ScaleEditor, we curate ScaleEdit-12M, the largest open-source image editing dataset to date, spanning 23 task families across diverse real and synthetic domains. Fine-tuning UniWorld-V1 and Bagel on ScaleEdit yields consistent gains, improving performance by up to 10.4% on ImgEdit and 35.1% on GEdit for general editing benchmarks and by up to 150.0% on RISE and 26.5% on KRIS-Bench for knowledge-infused benchmarks. These results demonstrate that open-source, agentic pipelines can approach commercial-grade data quality while retaining cost-effectiveness and scalability. Both the framework and dataset will be open-sourced. |
| title | ScaleEdit-12M: Scaling Open-Source Image Editing Data Generation via Multi-Agent Framework |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.20644 |