AI Paper Atlas

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

ICCV 2025 · Image generation › Text-to-image diffusion · approach: Compositional text-to-image guidance · 7 citations (OpenAlex)

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 changedWhereMeasuredEffect
replace vision Q and K (QK_replace) instead of KV_replace (ablation)QK-Edit (HunyuanVideo) · HiLo-BenchmarkCLIP_frames: 0.901 -> 0.908Improved
replace vision Q and K (QK_replace) instead of KV_replace (ablation)QK-Edit (HunyuanVideo) · HiLo-BenchmarkPickScore: 19.92 -> 20.08Improved
replace vision Q and K instead of V only (ablation)QK-Edit (HunyuanVideo) · HiLo-BenchmarkCLIP_frames: 0.900 -> 0.908Improved
replace vision Q and K instead of concatenating KV (ablation)QK-Edit (HunyuanVideo) · HiLo-BenchmarkCLIP_frames: 0.896 -> 0.908Improved
replace vision Q and K instead of concatenating KV (ablation)QK-Edit (HunyuanVideo) · HiLo-BenchmarkPickScore: 19.87 -> 20.08Improved
replace both Q and K rather than only Q or K or KV or V (ablation)QK-Edit · TGVE+ videostructure preservation vs prompt followingImproved
replace only KV or V (ablation)QK-Edit · TGVE+ videoprompt followingHurt
replace only Q or K (ablation)QK-Edit · TGVE+ videostructure preservationHurt
zero the text-to-vision attention quadrant (2nd) (ablation)FLUX · image generationgeneration qualityNo clear effect
zero the vision-to-text attention quadrant (3rd) (ablation)FLUX · image generationgeneration qualityHurt
QK-Edit vs RF-Solver on real image editingQK-Edit (FLUX.1-dev) · PIE-BenchCLIP-T: 0.257 -> 0.264Improved
QK-Edit vs DiTCtrl on generative video editingQK-Edit (HunyuanVideo) · TGVE+ViCLIP_dir: 0.141 -> 0.217Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy