Research map / Image generation / Text-to-image diffusion
Illuminating Unified Multimodal Model for Free-form Interleaved Text-Image Generation
Chonghuinan Wang, Zhikai Chen, Chunwei Wang, Yecong Wan, Junwei Yang, Zhixin Wang, Wei Zhang, Jiaqi Xu, Renjing Pei, Xiaohe Wu, FAN LI, Wangmeng Zuo
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 |
|---|---|---|---|
| add CoT to interleaved generation (text CFG 8.0, image CFG 1.0) (ablation) | ILLUME-X · Visual Storytelling | score: 7.12 -> 7.28 | Improved |
| add CoT to interleaved generation (text CFG 8.0, image CFG 1.0) (ablation) | ILLUME-X · Image-Text Complementation | score: 5.69 -> 5.91 | Improved |
| increase text guidance scale from 2.0 to 4.0 (image scale 1.0) (ablation) | ILLUME-X · Image-Text Complementation | score: 5.35 -> 5.59 | Improved |
| increase text guidance scale from 4.0 to 8.0 (image scale 1.0) (ablation) | ILLUME-X · Visual Storytelling | score: 7.13 -> 7.12 | No clear effect |
| increase image guidance scale from 1.0 to 1.5 (text scale 8.0) (ablation) | ILLUME-X · Visual Storytelling | score: 7.12 -> 6.85 | Hurt |
| increase image guidance scale from 1.0 to 1.25 (text scale 8.0) (ablation) | ILLUME-X · Image-Text Complementation | score: 5.69 -> 5.52 | Hurt |
| use ILLUME-X instead of Emu 3.5 (34B) for inference time | ILLUME-X · Inference cost | time per image (s): 409.50 -> 81.33 | Improved |
| fine-tune BAGEL into ILLUME-X on interleaved data | BAGEL · GenEval | overall: 0.82 -> 0.85 | Improved |
| fine-tune BAGEL into ILLUME-X on interleaved data | BAGEL · DPG-Bench | overall: 85.07 -> 86.38 | Improved |
| fine-tune BAGEL into ILLUME-X on interleaved data | BAGEL · DPG-Bench Relation | score: 90.82 -> 82.37 | Hurt |
| use ILLUME-X instead of Anole (unified models) | ILLUME-X · ISG-Bench | average score: 2.81 -> 6.26 | Improved |
| use ILLUME-X unified model instead of ISG-Agent | ILLUME-X · ISG-Bench | average score: 6.26 -> 6.26 | No clear effect |
Closest papers by meaning
- Towards Unified Multimodal Interleaved Generation via Group Relative Policy Optimization (NeurIPS 2025 · similarity 0.90)
- CoMM: A Coherent Interleaved Image-Text Dataset for Multimodal Understanding and Generation (CVPR 2025 · similarity 0.89)
- A High Quality Dataset and Reliable Evaluation for Interleaved Image-Text Generation (ICLR 2026 · similarity 0.89)
- OpenING: A Comprehensive Benchmark for Judging Open-ended Interleaved Image-Text Generation (CVPR 2025 · similarity 0.89)
- SILMM: Self-Improving Large Multimodal Models for Compositional Text-to-Image Generation (CVPR 2025 · similarity 0.88)
- DuoGen: Towards Autonomous Interleaved Multimodal Generation (CVPR 2026 · similarity 0.88)
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.
