Research map / Image generation / Text-to-image diffusion
LazyDrag: Enabling Stable Drag-Based Editing on Multi-Modal Diffusion Transformers via Explicit Correspondence
Zixin Yin, Xili Dai, Duomin Wang, Xianfang Zeng, Lionel Ni, Gang YU, Heung-Yeung Shum
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 |
|---|---|---|---|
| use LazyDrag with explicit correspondence instead of TTO-based DragText | LazyDrag · DragBench | MD: 21.51 -> 21.49 | Improved |
| use LazyDrag instead of TTO-free Inpaint4Drag | LazyDrag · DragBench | VIEScore O: 7.615 -> 8.210 | Improved |
| use LazyDrag instead of TTO-free FastDrag | LazyDrag · DragBench | MD: 31.84 -> 21.49 | Improved |
| add WTA and Latent Init (ablation) | LazyDrag · DragBench | MD | Improved |
| add background preservation (ablation) | LazyDrag · DragBench | SC and O | Improved |
| add correspondence-driven preservation instead of attention-similarity control (ablation) | LazyDrag · DragBench | all metrics | Improved |
| increase activation timestep from 20 to 50 (ablation) | LazyDrag · DragBench | drag accuracy vs artifacts | Mixed |
| add WTA blending and Gaussian-noise init to U-Net FastDrag (ablation) | FastDrag · DragBench | MD and SC | Improved |
| use Optimized setting (20 steps, inversion strength 0.7) instead of Default | LazyDrag · DragBench runtime | total time (s): 14.10 -> 4.31 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 5 of 5 on most shared results.
- FastDrag — FastDrag: Manipulate Anything in One Step (NeurIPS 2024) · better on 4 of 5 shared results
- GoodDrag — GoodDrag: Towards Good Practices for Drag Editing with Diffusion Models (ICLR 2025) · better on 4 of 4 shared results
- Inpaint4Drag — Inpaint4Drag: Repurposing Inpainting Models for Drag-Based Image Editing via Bidirectional Warping (ICCV 2025) · better on 4 of 4 shared results
- DiffEditor — DiffEditor: Boosting Accuracy and Flexibility on Diffusion-based Image Editing (CVPR 2024) · better on 4 of 4 shared results
- FreeDrag — FreeDrag: Feature Dragging for Reliable Point-based Image Editing (CVPR 2024) · 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
- FastDrag: Manipulate Anything in One Step (NeurIPS 2024 · similarity 0.91)
- DragDiffusion: Harnessing Diffusion Models for Interactive Point-based Image Editing (CVPR 2024 · similarity 0.90)
- GoodDrag: Towards Good Practices for Drag Editing with Diffusion Models (ICLR 2025 · similarity 0.90)
- EasyDrag: Efficient Point-based Manipulation on Diffusion Models (CVPR 2024 · similarity 0.90)
- DragonDiffusion: Enabling Drag-style Manipulation on Diffusion Models (ICLR 2024 · similarity 0.89)
- FlowDrag: 3D-aware Drag-based Image Editing with Mesh-guided Deformation Vector Flow Fields (ICML 2025 · similarity 0.88)
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.
