Research map / Image generation / Text-to-image diffusion
Parallel Multimodal Diffusion Language Models for Thinking-Aware Editing and Generation
Ye Tian, Ling Yang, JiongFan Yang, Anran Wang, Yu Tian, Jiani zheng, Haochen Wang, Zhiyang Teng, Zhuochen Wang, Yinjie Wang, Yunhai Tong, Mengdi Wang and 1 more
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 parallel instead of sequential text-then-image denoising (ablation) | MMaDA-Parallel · ParaBench | Output Alignment: 48.9 -> 51.5 | Improved |
| use parallel instead of sequential decoding (ablation) | MMaDA-Parallel · ParaBench | Text Alignment: 70.6 -> 70.4 | No clear effect |
| add ParaRL post-training (ablation) | MMaDA-Parallel · ParaBench | Output Alignment: 51.5 -> 59.8 | Improved |
| use trajectory-level rewards instead of output-level RL (ablation) | ParaRL · ParaBench | Output Alignment: 53.6 -> 59.8 | Improved |
| use output-level RL over no RL (ablation) | MMaDA-Parallel · ParaBench | Output Alignment: 51.5 -> 53.6 | Improved |
| increase ParaRL sampled steps from 2 to 3 (ablation) | ParaRL · ParaBench | Output Alignment: 53.6 -> 59.8 | Improved |
| increase ParaRL sampled steps from 3 to 4 (ablation) | ParaRL · ParaBench | Output Alignment: 59.8 -> 58.7 | No clear effect |
| enable thinking in Bagel on spatial edits | Bagel · ParaBench editing (Spatial) | editing score: 69.8 -> 65.0 | Hurt |
| enable thinking in Bagel on temporal edits | Bagel · ParaBench editing (Temporal) | editing score: 72.3 -> 75.6 | Improved |
| apply parallel framework and ParaRL to Lumina-DiMOO (ablation) | Lumina-DiMOO · ParaBench | Output Alignment: 63.3 -> 68.8 | Improved |
| use MMaDA-Parallel instead of sequential MMaDA on GenEval | MMaDA · GenEval | Overall: 0.68 -> 0.71 | Improved |
| use MMaDA-Parallel instead of sequential MMaDA on RISEBench | MMaDA · RISEBench | Overall accuracy: 5.5% -> 5.75% | Improved |
Closest papers by meaning
- DiffThinker: Towards Generative Multimodal Reasoning with Diffusion Models (ICML 2026 · similarity 0.88)
- MMaDA: Multimodal Large Diffusion Language Models (NeurIPS 2025 · similarity 0.88)
- I Think, Therefore I Diffuse: Enabling Multimodal In-Context Reasoning in Diffusion Models (ICML 2025 · similarity 0.88)
- Dual Diffusion for Unified Image Generation and Understanding (CVPR 2025 · similarity 0.88)
- Thinking Diffusion: Penalize and Guide Visual-Grounded Reasoning in Diffusion Multimodal Language Models (CVPR 2026 · similarity 0.87)
- Exploring the Deep Fusion of Large Language Models and Diffusion Transformers for 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.
