Research map / Image generation / Text-to-image diffusion
Dynamic Prompt Learning: Addressing Cross-Attention Leakage for Text-Based Image Editing
Kai Wang, Fei Yang, Shiqi Yang, Muhammad Atif Butt, Joost van de Weijer
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 NTI token conditioning with per-timestamp dynamic noun tokens and leakage-repair losses | P2P image editing with Stable Diffusion v1.4 · Cat → leopard; dog → tiger | CLIP-Score: 31.50% -> 32.23% | Improved |
| Replace NTI token conditioning with per-timestamp dynamic noun tokens and leakage-repair losses | P2P image editing with Stable Diffusion v1.4 · Cat → leopard; dog → tiger | Structure Dist: 0.0800 -> 0.0312 | Improved |
| Replace NTI token conditioning with per-timestamp dynamic noun tokens and leakage-repair losses | P2P image editing with Stable Diffusion v1.4 · Book → box; clock → sunflower | CLIP-Score: 12.91% -> 14.29% | Improved |
| Replace NTI token conditioning with per-timestamp dynamic noun tokens and leakage-repair losses | P2P image editing with Stable Diffusion v1.4 · Book → box; clock → sunflower | Structure Dist: 0.0191 -> 0.0163 | Improved |
| Replace NTI token conditioning with per-timestamp dynamic noun tokens and leakage-repair losses | P2P image editing with Stable Diffusion v1.4 · Multi-object image translations | User Study preference: 12.8% -> 87.2% | Improved |
| Use the disjoint object attention loss alone (ablation) | Dynamic Prompt Learning · MO-Set validation localization | IoU | No clear effect |
| Use the background leakage loss alone (ablation) | Dynamic Prompt Learning · MO-Set validation localization | IoU | No clear effect |
| Add the attention balancing loss (ablation) | Dynamic Prompt Learning · MO-Set validation localization | IoU | Improved |
| Add the disjoint object and background leakage losses alongside attention balancing (ablation) | Dynamic Prompt Learning · MO-Set validation localization | IoU | Improved |
| Optimize noun tokens separately at each timestamp rather than using a single token embedding across timestamps (ablation) | Dynamic Prompt Learning · Cross-attention map quality | Cross-attention quality | Improved |
Closest papers by meaning
- Textualize Visual Prompt for Image Editing via Diffusion Bridge (AAAI 2025 · similarity 0.91)
- An Item Is Worth a Prompt: Versatile Image Editing with Disentangled Control (AAAI 2025 · similarity 0.91)
- Dynamic Prompt Optimizing for Text-to-Image Generation (CVPR 2024 · similarity 0.90)
- Prompt Augmentation for Self-supervised Text-guided Image Manipulation (CVPR 2024 · similarity 0.89)
- Towards Understanding Cross and Self-Attention in Stable Diffusion for Text-Guided Image Editing (CVPR 2024 · similarity 0.89)
- AP-Adapter: Improving Generalization of Automatic Prompts on Unseen Text-to-Image Diffusion Models (NeurIPS 2024 · similarity 0.89)
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.
