Research map / Image generation / Editing and style transfer
ICE-Bench: A Unified and Comprehensive Benchmark for Image Creating and Editing
Yulin Pan, Xiangteng He, Chaojie Mao, Zhen Han, Zeyinzi Jiang, Jingfeng Zhang, Yu Liu
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 OmniGen (3.8B params, up to 2280x2280 training images) instead of ACE (0.6B, 512x512) | text-to-image creating · ICE-Bench Task 1 | imaging quality (IMG): 0.534 -> 0.726 | Improved |
| use ACE instead of OmniGen for style reference | ICE-Bench Style Reference Creating · ICE-Bench Task 3 | prompt following (PF): 0.431 -> 0.232 | Hurt |
| use ACE instead of OmniGen for style reference | ICE-Bench Style Reference Creating · ICE-Bench Task 3 | reference consistency (REF): 0.432 -> 0.802 | Improved |
| use IP-Adapter instead of ACE for style reference | ICE-Bench Style Reference Creating · ICE-Bench Task 3 | prompt following (PF): 0.232 -> 0.288 | Improved |
| train on more pixel-aligned editing data (ACE instead of OmniGen) | ICE-Bench Global Editing · ICE-Bench Tasks 5-16 | source consistency (SRC): 0.844 -> 0.899 | Improved |
| train on more pixel-aligned editing data (UltraEdit instead of OmniGen) | ICE-Bench Local Editing · ICE-Bench Tasks 17-22 | source consistency (SRC): 0.853 -> 0.953 | Improved |
| use limited, low-quality training data (InstructPix2Pix instead of UltraEdit) | ICE-Bench Global Editing · ICE-Bench Tasks 5-16 | prompt following (PF): 0.470 -> 0.355 | Hurt |
| use the unified model OmniGen instead of ACE on face swap | ICE-Bench Face Swap · ICE-Bench Task 31 | reference consistency (REF): 0.250 -> 0.477 | Improved |
| use FLUX-Control instead of OminiControl (lacks pose control) | ICE-Bench Controllable Generation · ICE-Bench Tasks 23-27 | controllability (CTRL): 0.687 -> 0.846 | Improved |
Closest papers by meaning
- ImagenHub: Standardizing the evaluation of conditional image generation models (ICLR 2024 · similarity 0.89)
- GenScale: A Benchmark for Relative Object Scale in Image Generation and Editing (NeurIPS 2026 · similarity 0.89)
- ImgEdit: A Unified Image Editing Dataset and Benchmark (NeurIPS 2025 · similarity 0.89)
- MMIG-Bench: Towards Comprehensive and Explainable Evaluation of Multi-Modal Image Generation Models (NeurIPS 2025 · similarity 0.89)
- IDEA-Bench: How Far are Generative Models from Professional Designing? (CVPR 2025 · similarity 0.88)
- EditAR: Unified Conditional Generation with Autoregressive Models (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.
