Research map / Image generation / Text-to-image diffusion
Rethinking UMM Visual Generation: Masked Modeling for Efficient Image-Only Pre-training
Peng Sun, Jun Xie, Tao Lin
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 |
|---|---|---|---|
| pre-train on image-only instead of text-image pairs (ablation) | IOMM-B · GenEval | overall score: 0.88 -> 0.89 | Improved |
| pre-train on image-only data instead of pairs (zero-shot editing) (ablation) | IOMM-B · ImgEdit-Bench | overall: 2.61 -> 2.82 | Improved |
| fine-tune with mixed data instead of pair data (ablation) | OpenUni-L · GenEval | overall score: 0.88 -> 0.88 | No clear effect |
| fine-tune with mixed data (ablation) | OpenUni-L · GenEval | overall score: 0.85 -> 0.88 | Improved |
| fine-tune with mixed data (ablation) | Qwen-Image 512 · GenEval | overall score: 0.85 -> 0.89 | Improved |
| fine-tune with mixed data (ablation) | Qwen-Image 1024 · GenEval | overall score: 0.87 -> 0.89 | Improved |
| fine-tune on image-only data (ablation) | OpenUni-L · GenEval | overall score: 0.85 -> 0.74 | Hurt |
| fine-tune on image-only data (ablation) | Qwen-Image 512 · GenEval | overall score: 0.85 -> 0.42 | Hurt |
| fine-tune on image-only data (ablation) | Qwen-Image 1024 · GenEval | overall score: 0.87 -> 0.47 | Hurt |
| fine-tune with pair data (ablation) | OpenUni-L · WISE | overall score: 0.52 -> 0.62 | Improved |
| increase image token mask ratio to 0.95 from 0.45 (ablation) | IOMM-XL · GenEval | overall score: 0.88 -> 0.77 | Hurt |
| increase image token mask ratio to 0.95 from 0.45 (ablation) | IOMM-XL · DPGBench | score: 79.79 -> 69.41 | Hurt |
Closest papers by meaning
- DuoGen: Towards Autonomous Interleaved Multimodal Generation (CVPR 2026 · similarity 0.90)
- Learning Visual Generative Priors without Text (CVPR 2025 · similarity 0.89)
- MMIG-Bench: Towards Comprehensive and Explainable Evaluation of Multi-Modal Image Generation Models (NeurIPS 2025 · similarity 0.89)
- OmniGen2: Towards Instruction-Aligned Multimodal Generation (CVPR 2026 · similarity 0.89)
- UniGen: Enhanced Training & Test-Time Strategies for Unified Multimodal Understanding and Generation (NeurIPS 2025 · similarity 0.89)
- VIGC: Visual Instruction Generation and Correction (AAAI 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.
