Research map / Image generation / Text-to-image diffusion
QK-Edit: Revisiting Attention-based Injection in MM-DiT for Image and Video Editing
Tiancheng Shen, Zilong Huang, Xiangtai Li, Zhijie Lin, Jiyang Liu, Yitong Wang, Jiashi Feng, Ming-Hsuan Yang, Jun Hao Liew
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 |
|---|---|---|---|
| replace vision Q and K (QK_replace) instead of KV_replace (ablation) | QK-Edit (HunyuanVideo) · HiLo-Benchmark | CLIP_frames: 0.901 -> 0.908 | Improved |
| replace vision Q and K (QK_replace) instead of KV_replace (ablation) | QK-Edit (HunyuanVideo) · HiLo-Benchmark | PickScore: 19.92 -> 20.08 | Improved |
| replace vision Q and K instead of V only (ablation) | QK-Edit (HunyuanVideo) · HiLo-Benchmark | CLIP_frames: 0.900 -> 0.908 | Improved |
| replace vision Q and K instead of concatenating KV (ablation) | QK-Edit (HunyuanVideo) · HiLo-Benchmark | CLIP_frames: 0.896 -> 0.908 | Improved |
| replace vision Q and K instead of concatenating KV (ablation) | QK-Edit (HunyuanVideo) · HiLo-Benchmark | PickScore: 19.87 -> 20.08 | Improved |
| replace both Q and K rather than only Q or K or KV or V (ablation) | QK-Edit · TGVE+ video | structure preservation vs prompt following | Improved |
| replace only KV or V (ablation) | QK-Edit · TGVE+ video | prompt following | Hurt |
| replace only Q or K (ablation) | QK-Edit · TGVE+ video | structure preservation | Hurt |
| zero the text-to-vision attention quadrant (2nd) (ablation) | FLUX · image generation | generation quality | No clear effect |
| zero the vision-to-text attention quadrant (3rd) (ablation) | FLUX · image generation | generation quality | Hurt |
| QK-Edit vs RF-Solver on real image editing | QK-Edit (FLUX.1-dev) · PIE-Bench | CLIP-T: 0.257 -> 0.264 | Improved |
| QK-Edit vs DiTCtrl on generative video editing | QK-Edit (HunyuanVideo) · TGVE+ | ViCLIP_dir: 0.141 -> 0.217 | Improved |
Closest papers by meaning
- Towards Understanding Cross and Self-Attention in Stable Diffusion for Text-Guided Image Editing (CVPR 2024 · similarity 0.89)
- An Item Is Worth a Prompt: Versatile Image Editing with Disentangled Control (AAAI 2025 · similarity 0.87)
- Prompt Augmentation for Self-supervised Text-guided Image Manipulation (CVPR 2024 · similarity 0.87)
- Towards Efficient Diffusion-Based Image Editing with Instant Attention Masks (AAAI 2024 · similarity 0.87)
- Textualize Visual Prompt for Image Editing via Diffusion Bridge (AAAI 2025 · similarity 0.87)
- VideoGrain: Modulating Space-Time Attention for Multi-Grained Video Editing (ICLR 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.
