Research map / Image generation / Text-to-image diffusion
Asynchronous Denoising Diffusion Models for Aligning Text-to-Image Generation
Zijing Hu, Yunze Tong, Fengda Zhang, Junkun Yuan, Jun Xiao, Kun Kuang
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 asynchronous denoising with mask-guided concave schedule | SD 2.1 DM · Animal Activity | ImageReward: 0.7543 -> 0.9219 | Improved |
| use asynchronous denoising | SD 2.1 DM · GenEval | QwenScore: 4.9390 -> 5.3390 | Improved |
| use asynchronous denoising | SD 2.1 DM · MSCOCO | QwenScore: 5.8507 -> 6.2601 | Improved |
| add prompt-region mask on top of concave scheduler (ablation) | DM_concave · Animal Activity | ImageReward: 0.8544 -> 0.9219 | Improved |
| use dynamic mask instead of fixed mask (ablation) | AsynDM · Animal Activity | QwenScore: 5.2593 -> 5.5218 | Improved |
| use piecewise linear instead of quadratic scheduler (ablation) | AsynDM · Animal Activity | ImageReward: 0.9219 -> 0.8472 | Hurt |
| use exponential instead of quadratic scheduler (ablation) | AsynDM · Animal Activity | ImageReward: 0.9219 -> 0.8686 | Hurt |
| apply asynchronous denoising | SD 2.1 · FID-30K | FID: 48.63 -> 49.38 | Hurt |
| apply asynchronous denoising | SD 2.1 · sampling 1,280 images | time (min): 78 -> 86 | Hurt |
| use CFG++ guidance | SD 2.1 DM · GenEval | ImageReward: 0.1541 -> -0.1344 | Hurt |
Closest papers by meaning
- InitNO: Boosting Text-to-Image Diffusion Models via Initial Noise Optimization (CVPR 2024 · similarity 0.89)
- AsyncDiff: Parallelizing Diffusion Models by Asynchronous Denoising (NeurIPS 2024 · similarity 0.89)
- Towards Better Alignment: Training Diffusion Models with Reinforcement Learning Against Sparse Rewards (CVPR 2025 · similarity 0.89)
- Denoising Autoregressive Transformers for Scalable Text-to-Image Generation (ICLR 2025 · similarity 0.89)
- On the Scalability of Diffusion-based Text-to-Image Generation (CVPR 2024 · similarity 0.89)
- One Stone with Two Birds: A Null-Text-Null Frequency-Aware Diffusion Models for Text-Guided Image Inpainting (NeurIPS 2025 · 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.
