Research map / Video / Video generation
MotionEdit: Benchmarking and Learning Motion-Centric Image Editing
Yixin Wan, Lei Ke, Wenhao Yu, Kai-Wei Chang, Dong Yu
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 MotionNFT optical-flow motion reward post-training | FLUX.1 Kontext [Dev] · MotionEdit-Bench | Overall (MLLM, 0-5): 3.84 -> 4.25 | Improved |
| add MotionNFT optical-flow motion reward post-training | Qwen-Image-Edit · MotionEdit-Bench | Overall (MLLM, 0-5): 4.65 -> 4.72 | Improved |
| add MotionNFT post-training | FLUX.1 Kontext [Dev] · MotionEdit-Bench | Motion Alignment Score: 53.73 -> 55.45 | Improved |
| add MotionNFT post-training | Qwen-Image-Edit · MotionEdit-Bench | Motion Alignment Score: 56.46 -> 57.23 | Improved |
| add MotionNFT post-training | Qwen-Image-Edit · MotionEdit-Bench | Win Rate (%): 72.80 -> 73.67 | Improved |
| add MotionNFT post-training | FLUX.1 Kontext [Dev] · MotionEdit-Bench | Win Rate (%): 57.71 -> 64.95 | Improved |
| add MotionNFT post-training (ablation) | FLUX.1 Kontext [Dev] · ImgEdit-Bench | Overall: 3.26 -> 3.50 | Improved |
| add MotionNFT post-training (ablation) | Qwen-Image-Edit · ImgEdit-Bench | Overall: 4.08 -> 4.20 | Improved |
| add optical-flow motion reward on top of MLLM-only reward (ablation) | FLUX.1 Kontext [Dev] · MotionEdit-Bench | Motion Alignment Score: 54.58 -> 55.45 | Improved |
| add optical-flow motion reward on top of MLLM-only reward (ablation) | Qwen-Image-Edit · MotionEdit-Bench | Motion Alignment Score: 56.46 -> 57.23 | Improved |
| MLLM-only reward training (UniWorld-V2) (ablation) | Qwen-Image-Edit · MotionEdit-Bench | Win Rate (%): 73.01 -> 72.77 | Hurt |
| MLLM-only reward training (UniWorld-V2) (ablation) | FLUX.1 Kontext [Dev] · MotionEdit-Bench | Overall (MLLM, 0-5): 3.84 -> 4.20 | Improved |
Closest papers by meaning
- Learning Action and Reasoning-Centric Image Editing from Videos and Simulation (NeurIPS 2024 · similarity 0.91)
- Dynamic Motion Blending for Versatile Motion Editing (CVPR 2025 · similarity 0.90)
- MotionEditor: Editing Video Motion via Content-Aware Diffusion (CVPR 2024 · similarity 0.90)
- SimMotionEdit: Text-Based Human Motion Editing with Motion Similarity Prediction (CVPR 2025 · similarity 0.89)
- VINCIE-NExT: Unlocking Video Editing from Images via In-Context Modeling (NeurIPS 2026 · similarity 0.88)
- ViTeX-Bench: Benchmarking High-Fidelity Video Scene Text Editing (NeurIPS 2026 · 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.
