Research map / Image generation / Editing and style transfer
MICo-150K: A Comprehensive Dataset Advancing Multi-Image Composition
Xinyu Wei, Kangrui Cen, Hongyang Wei, Zhen Guo, Bairui Li, Zeqing Wang, Jinrui Zhang, Lei Zhang
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 |
|---|---|---|---|
| fine-tune on MICo-150K | BLIP3-o · MICo-Bench Overall | Weighted-Ref-VIEScore: 2.2 -> 42.2 | Improved |
| fine-tune on MICo-150K | DiMOO · MICo-Bench Overall | Weighted-Ref-VIEScore: 4.3 -> 32.3 | Improved |
| fine-tune on MICo-150K | BAGEL · MICo-Bench Overall | Weighted-Ref-VIEScore: 33.3 -> 42.6 | Improved |
| fine-tune on MICo-150K | Qwen-Image-Edit · MICo-Bench Overall | Weighted-Ref-VIEScore: 38.5 -> 56.4 | Improved |
| fine-tune on MICo-150K | OmniGen2 · MICo-Bench Overall | Weighted-Ref-VIEScore: 41.0 -> 50.6 | Improved |
| add De&Re data with real targets (ablation) | Qwen-Image-Edit · MICo-Bench De&Re | Weighted-Ref-VIEScore: 58.2 -> 60.7 | Improved |
| add De&Re data with real targets (ablation) | BAGEL · MICo-Bench De&Re | Weighted-Ref-VIEScore: 48.1 -> 50.9 | Improved |
| add De&Re data with real targets (ablation) | DiMOO · MICo-Bench De&Re | Weighted-Ref-VIEScore: 23.9 -> 27.0 | Improved |
| replace real De&Re targets with synthetic targets (ablation) | Qwen-Image-Edit · MICo-Bench De&Re | Weighted-Ref-VIEScore: 60.7 -> 60.6 | No clear effect |
| replace real De&Re targets with synthetic targets (ablation) | BAGEL · MICo-Bench Overall | Weighted-Ref-VIEScore: 44.3 -> 44.1 | No clear effect |
| add De&Re data (ablation) | OmniGen2 · MICo-Bench Overall | Weighted-Ref-VIEScore: 50.6 -> 51.2 | No clear effect |
Closest papers by meaning
- MICON-Bench: Benchmarking and Enhancing Multi-Image Context Image Generation in Unified Multimodal Models (CVPR 2026 · similarity 0.86)
- MMIG-Bench: Towards Comprehensive and Explainable Evaluation of Multi-Modal Image Generation Models (NeurIPS 2025 · similarity 0.85)
- RealCompo: Balancing Realism and Compositionality Improves Text-to-Image Diffusion Models (NeurIPS 2024 · similarity 0.85)
- ImagenHub: Standardizing the evaluation of conditional image generation models (ICLR 2024 · similarity 0.85)
- MMIU: Multimodal Multi-image Understanding for Evaluating Large Vision-Language Models (ICLR 2025 · similarity 0.85)
- ImgEdit: A Unified Image Editing Dataset and Benchmark (NeurIPS 2025 · similarity 0.84)
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.
