Research map / Image generation / Text-to-image diffusion
Discrete Diffusion Bridges for Spatiotemporally Aligned Image Translation and Generation
Xing Xie, Jiawei Liu, Shijun Zhou, Huijie Fan, Zhi Han, Yandong Tang, Liangqiong Qu
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 |
|---|---|---|---|
| inject source tokens into the absorbing state (hybrid absorption, ratio 0.5 random-ratio) instead of pure mask (ablation) | DDB on Lumina-DiMOO · OmniEdit image editing | Edit Score: 0.743 -> 0.779 | Improved |
| train with random injection ratio instead of a fixed ratio (ablation) | DDB · OmniEdit image editing | Edit Score: 0.757 -> 0.779 | Improved |
| raise source injection ratio (ablation) | DDB · OmniEdit image editing | DINO vs CLIP-T: DINO 0.765 (ratio 0.0) -> 0.788 (ratio 1.0) | Mixed |
| use stochastic proportional mixing (M3) rather than deterministic allocation (M1) (ablation) | DDB source injection · OmniEdit image editing | Edit Score: 0.695 -> 0.779 | Improved |
| deterministic injection of mask into high-information regions (M1) versus no injection (ablation) | DDB · SynthRAD modality translation | PSNR: 23.15 -> 21.96 | Hurt |
| add information-guided masking schedule (image difference) (ablation) | DDB · OmniEdit image editing | Edit Score: 0.755 -> 0.779 | Improved |
| use gradient as information metric instead of image difference (ablation) | DDB · OmniEdit image editing | Edit Score: 0.779 -> 0.686 | Hurt |
| use target variance as information metric for unanchored generation (ablation) | DDB · MIMIC-CXR report-to-image | FID: 14.88 -> 9.81 | Improved |
| apply DDB at 5 sampling steps versus baseline Lumina-DiMOO | Lumina-DiMOO · OmniEdit image editing | Edit Score: 0.541 -> 0.606 | Improved |
| apply DDB over Lumina-DiMOO backbone on all-in-one restoration | Lumina-DiMOO · CDD-11 | PSNR: 21.96 -> 22.61 | Improved |
| DDB versus Bagel on background preservation | DDB · OmniEdit image editing | DINO: 0.836 (Bagel) vs 0.790 (DDB) | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Lavida-O — Lavida-O: Elastic Large Masked Diffusion Models for Unified Multimodal Understanding and Generation (ICLR 2026) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Dual Diffusion Implicit Bridges for Image-to-Image Translation (ICLR 2023 · similarity 0.90)
- Anchoring and Steering Diffusion: Enhancing the Faithfulness of Text-to-Image Generation at Inference Time (ECCV 2026 · similarity 0.90)
- Adaptive Domain Shift in Diffusion Models for Cross-Modality Image Translation (ICLR 2026 · similarity 0.90)
- Noise Diffusion for Enhancing Semantic Faithfulness in Text-to-Image Synthesis (CVPR 2025 · similarity 0.88)
- Scale Where It Matters: Training-Free Localized Scaling for Diffusion Models (NeurIPS 2026 · similarity 0.87)
- Spatial Transport Optimization by Repositioning Attention Map for Training-Free Text-to-Image Synthesis (CVPR 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.
