Research map / Image generation / Editing and style transfer
A²-Edit: Precise Reference-Guided Image Editing of Arbitrary Objects and Ambiguous Masks
华渝 郑, Guangzhao Li, Baixuan Zhao, Siqi Luo, Hantao Jiang, Guangtao Zhai, Xiaohong Liu
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add Mixture of Transformers (routing over attention and FFN) instead of FFN-only MoE (ablation) | A2-Edit · UniEdit test (rough mask) | DINO-I: 56.11 -> 62.28 | Improved |
| add the full A2-Edit framework instead of baseline (ablation) | FLUX.1 Fill · UniEdit test (fine mask) | VLM score: 64.3 -> 82.3 | Improved |
| add FFN-only MoE over baseline (ablation) | FLUX.1 Fill · UniEdit test (fine mask) | VLM score: 64.3 -> 70.3 | Improved |
| add UniEdit-500K training data (ablation) | A2-Edit · UniEdit test (rough mask) | FID: 72.84 -> 54.62 | Improved |
| add UniEdit-500K training data (ablation) | A2-Edit · UniEdit test (fine mask) | DINO-I: 58.64 -> 61.71 | Improved |
| add augmented rough mask training after fine mask training (ablation) | A2-Edit (MATS) · UniEdit test (rough mask) | DINO-I: 55.65 -> 59.70 | Improved |
| add augmented rough mask training after fine mask training (ablation) | A2-Edit (MATS) · UniEdit test (rough mask) | FID: 74.77 -> 60.33 | Improved |
| add bounding box training stage (ablation) | A2-Edit (MATS) · UniEdit test (rough mask) | DINO-I: 59.70 -> 62.28 | Improved |
| add bounding box training stage (ablation) | A2-Edit (MATS) · UniEdit test (rough mask) | VLM score: 73.0 -> 79.3 | Improved |
| train with fine masks only (ablation) | A2-Edit · UniEdit test (rough mask) | DINO-I: 56.27 -> 55.65 | Hurt |
| train LiDiff-style baseline Insert Anything vs A2-Edit on rough mask | A2-Edit · VITON-HD (rough mask) | DINO-I: 60.87 -> 63.79 | Improved |
| use A2-Edit on UniEdit test (rough mask) vs ACE++ | A2-Edit · UniEdit test (rough mask) | FID: 63.19 -> 54.62 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Insert Anything — Insert Anything: Image Insertion via In-Context Editing in DiT (AAAI 2026) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Paint by Inpaint: Learning to Add Image Objects by Removing Them First (CVPR 2025 · similarity 0.88)
- Improving Editability in Image Generation with Layer-wise Memory (CVPR 2025 · similarity 0.88)
- ImgEdit: A Unified Image Editing Dataset and Benchmark (NeurIPS 2025 · similarity 0.87)
- Learning Where to Edit Vision Transformers (NeurIPS 2024 · similarity 0.87)
- Referring Image Editing: Object-level Image Editing via Referring Expressions (CVPR 2024 · similarity 0.87)
- Click2Mask: Local Editing with Dynamic Mask Generation (AAAI 2025 · similarity 0.87)
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.
