Research map / Image generation / Autoregressive and latent generation
InstaEdit: Instant Image Editing via Optimized Noise Prediction
Ning Ma, Yangrui Shao
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 reparameterization module (ablation) | InstaEdit · GEdit-Bench (English) | overall: 5.35 -> 5.67 | Improved |
| add singular value prediction module (ablation) | InstaEdit · GEdit-Bench (English) | overall: 5.27 -> 5.67 | Improved |
| add noise regularization loss L_reg (ablation) | InstaEdit · GEdit-Bench (English) | overall: 5.48 -> 5.67 | Improved |
| add Stage II LoRA fine-tuning of foundation model (ablation) | InstaEdit · GEdit-Bench (English) | overall: 5.51 -> 5.67 | Improved |
| add InstaEdit with 4 steps vs 8-step baseline | FLUX.1-Kontext · GEdit-Bench (English) | overall: 5.21 -> 5.67 | Improved |
| add InstaEdit with 4 steps vs 8-step baseline | FLUX.2-klein-base-9B · ImgEdit-Bench | overall: 6.81 -> 7.47 | Improved |
| InstaEdit 4 steps vs FlowFast 10 steps | FLUX.1-Kontext · GEdit-Bench (English) | overall: 5.45 -> 5.67 | Improved |
| InstaEdit vs LCM at 4 steps | FLUX.1-Kontext · GEdit-Bench (English) (4-step) | overall: 5.23 -> 5.67 | Improved |
| InstaEdit 4 steps vs 30-step Step1X-Edit | Step1X-Edit · GEdit-Bench (English) | overall: 6.79 -> 6.84 | No clear effect |
| InstaEdit 4 steps vs 30-step FLUX.2-klein-base-4B | FLUX.2-klein-base-4B · GEdit-Bench (English) | speedup: 2.57× | Mixed |
| add InstaEdit with 3 steps on distilled model | FLUX.2-klein-9B · GEdit-Bench (distilled) | overall: 7.36 -> 7.35 | No clear effect |
| InstaEdit 8 steps vs LCM 8 steps | FLUX.1-Kontext · GEdit-Bench (English) (8-step) | overall: 5.60 -> 5.75 | 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.
- + RegionE — RegionE: Adaptive Region-Aware Generation for Efficient Image Editing (ICLR 2026) · better on 2 of 2 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
- DNAEdit: Direct Noise Alignment for Text-Guided Rectified Flow Editing (NeurIPS 2025 · similarity 0.88)
- FeedEdit: Text-Based Image Editing with Dynamic Feedback Regulation (CVPR 2025 · similarity 0.87)
- ImgEdit: A Unified Image Editing Dataset and Benchmark (NeurIPS 2025 · similarity 0.87)
- Enabling Instructional Image Editing with In-Context Generation in Large Scale Diffusion Transformer (NeurIPS 2025 · similarity 0.87)
- InstaFlow: One Step is Enough for High-Quality Diffusion-Based Text-to-Image Generation (ICLR 2024 · similarity 0.87)
- AutoEdit: Automatic Hyperparameter Tuning for Image Editing (NeurIPS 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.
