Research map / Image generation / Autoregressive and latent generation
MICE-Bench: A Challenging and Comprehensive Benchmark for Multi-Reference Image Creation and Editing
Siqi Luo, Huayu Zheng, Jianghan Shen, Yi Xin, Luxin Xu, Jiyao Liu, Xinyu Zhang, Hang Zhou, Pengyu Xie, Xiaohui Li, Shuo Cao, Yuandong Pu and 7 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 |
|---|---|---|---|
| upgrade Seedream 4.0 to 4.5 | Seedream · MICE-Bench creation | overall: 7.02 -> 7.47 | Improved |
| upgrade NanoBanana to NanoBanana 2 | NanoBanana · MICE-Bench creation | overall: 7.19 -> 7.48 | Improved |
| upgrade GPT-Image 1.0 to 1.5 | GPT-Image · MICE-Bench creation | overall: 7.37 -> 7.59 | Improved |
| upgrade Qwen-Edit-2509 to 2511 | Qwen-Edit · MICE-Bench creation | overall: 5.43 -> 6.50 | Improved |
| upgrade FLUX.2 dev to klein | FLUX.2 · MICE-Bench creation | overall: 7.10 -> 7.26 | Improved |
| upgrade NanoBanana to NanoBanana 2 | NanoBanana · MICE-Bench local editing | overall: 7.30 -> 8.55 | Improved |
| upgrade GPT-Image 1.0 to 1.5 | GPT-Image · MICE-Bench global editing | overall: 7.82 -> 8.50 | Improved |
| upgrade Qwen-Edit-2509 to 2511 | Qwen-Edit · MICE-Bench local editing | overall: 5.67 -> 7.16 | Improved |
| increase concept count from K=2 to K=7 | GPT-Image 1.5 · MICE-Bench creation | concept retention rate (%): 95.56% -> 76.36% | Hurt |
| increase concept count from K=5 to K=6 for Control concept | GPT-Image 1.5 · MICE-Bench creation | Control concept retention (%): 72.60 -> 45.16 | Hurt |
| increase concept count from K=6 to K=7 for Material concept | GPT-Image 1.5 · MICE-Bench creation | Material concept retention (%): 94.59 -> 77.78 | Hurt |
| switch from creation to editing tasks | BAGEL · MICE-Bench | overall: 6.25 -> 4.39 | Hurt |
Closest papers by meaning
- MMIG-Bench: Towards Comprehensive and Explainable Evaluation of Multi-Modal Image Generation Models (NeurIPS 2025 · similarity 0.88)
- Factuality Matters: When Image Generation and Editing Meet Structured Visuals (ICLR 2026 · similarity 0.88)
- Imagine Before You Draw: Visual Prompt Engineering for Image Generation (NeurIPS 2026 · similarity 0.88)
- EdiVal-Agent: An Object-Centric Framework for Automated, Fine-Grained Evaluation of Multi-Turn Editing (ICLR 2026 · similarity 0.88)
- IDEA-Bench: How Far are Generative Models from Professional Designing? (CVPR 2025 · similarity 0.88)
- GIR-Bench: Versatile Benchmark for Generating Images with Reasoning (ICLR 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.
