Research map / Image generation / Editing and style transfer
Towards Robust Sequential Decomposition for Complex Image Editing
Zilai Zeng, Mingdeng Cao, Zijie Li, Xiaochen Lian, Yichun Shi, Peihao Zhu, Chen Sun, Peng Wang
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 |
|---|---|---|---|
| naive sequential decomposition instead of single-turn | BAGEL zero-shot · Independent chains (500 unseen) | GPT-5 score: 2.43 -> 2.24 | Hurt |
| naive sequential decomposition instead of single-turn | GPT-4o · Independent chains (500 unseen) | GPT-5 score: 5.35 -> 4.61 | Hurt |
| sequential editing (CGSE K=3) instead of single-turn after finetuning | finetuned BAGEL · Independent chains (500 unseen) | GPT-5 score: 4.0 -> 4.15 | Improved |
| use context guidance (CGSE) instead of full-context sequential editing, K=3 | finetuned BAGEL · Independent chains (500 unseen) | DINO-I: 0.738 -> 0.749 | Improved |
| clarify the dependent instruction before single-turn editing | finetuned BAGEL · Dependent chains | GPT-5 score: 3.57 -> 3.91 | Improved |
| guide with last editing result only instead of all previous results (ablation) | CGSE K=3 · Independent chains (100) | GPT-5 score: 4.18 -> 4.35 | Improved |
| guide with last editing result only instead of all previous results (ablation) | CGSE K=3 · Dependent chains (100) | GPT-5 score: 4.13 -> 4.23 | Improved |
| raise gamma_ctx from 0.5 to 2.5 (ablation) | CGSE K=3 · Dependent chains (100) | GPT-5 score: 3.96 -> 4.23 | Improved |
| raise gamma_ctx from 0.5 to 2.5 (ablation) | CGSE K=3 · Independent chains (100) | GPT-5 score: 4.41 -> 4.35 | No clear effect |
| increase decomposition steps from 1 to 4 (ablation) | CGSE · Independent chains | GPT-5 score: 4.0 -> 4.27 | Improved |
| increase decomposition steps from 1 to 5 (ablation) | CGSE · Dependent chains | GPT-5 score: 3.91 -> 4.14 | Improved |
| add synthetic sequence data to real data for finetuning (ablation) | BAGEL single-turn · Complex-Edit | IP: 6.00 -> 6.26 | Improved |
Closest papers by meaning
- ImgEdit: A Unified Image Editing Dataset and Benchmark (NeurIPS 2025 · similarity 0.89)
- Meta-CoT: Enhancing Granularity and Generalization in Image Editing (CVPR 2026 · similarity 0.89)
- InsightEdit: Towards Better Instruction Following for Image Editing (CVPR 2025 · similarity 0.89)
- AnyEdit: Mastering Unified High-Quality Image Editing for Any Idea (CVPR 2025 · similarity 0.89)
- Multi-Reward as Condition for Instruction-based Image Editing (ICLR 2025 · similarity 0.89)
- Learning Action and Reasoning-Centric Image Editing from Videos and Simulation (NeurIPS 2024 · similarity 0.89)
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.
