AI Paper Atlas

Research map / Benchmarks / MagicBrush

MagicBrush leaderboard

16 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on MagicBrush. Best DINO: 0.971 (DesignEdit, AAAI 2025).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

MagicBrush · DINO

13 papers, each with its best reported number for its own method; higher is better.

#DINO ↑MethodPaperTopic
10.971DesignEdit
SDXL-1.0
DesignEdit: Unify Spatial-Aware Image Editing via Training-free Inpainting with a Multi-Layered Latent Diffusion Framework
AAAI 2025 · Table 2
Editing and style transfer
20.932CoLoGenCoLoGen: Progressive Learning of Concept-Localization Duality for Unified Image Generation
CVPR 2026 · Table 2
Text-to-image diffusion
30.897IEAP
FLUX.1-dev
Image Editing As Programs with Diffusion Models
NeurIPS 2025 · Table 1
Editing and style transfer
40.882DIM-4.6B-Edit
1.6B generator
Draw-In-Mind: Rebalancing Designer-Painter Roles in Unified Multimodal Models Benefits Image Editing
ICLR 2026 · Table 4
Text-to-image diffusion
50.881EditMGT (Ours)EditMGT: Unleashing Potentials of Masked Generative Transformers in Image Editing
CVPR 2026 · Table 1
Editing and style transfer
60.881AnySD w/ AnyEdit
Stable Diffusion 1.5
AnyEdit: Mastering Unified High-Quality Image Editing for Any Idea
CVPR 2025 · Table 4
Editing and style transfer
70.877MURE
BAGEL
Unlocking Complex Image Editing via Natively Interleaved Visual Textual CoT with Deep Confidence Reasoning
ECCV 2026 · Table 1
Multimodal LLMs
80.8768LLMGALLMGA: Multimodal Large Language Model based Generation Assistant
ECCV 2024 · Table 2
Multimodal LLMs
90.876PixWizard
Lumina-Next-T2I
PixWizard: Versatile Image-to-Image Visual Assistant with Open-Language Instructions
ICLR 2025 · Image editing table
Text-to-image diffusion
100.8541LayerCraft
FLUX.1-dev
LayerCraft: Enhancing Text-to-Image Generation with CoT Reasoning and Layered Object Integration
NeurIPS 2025 · Table: Multi-turn editing resu
Text-to-image diffusion
3 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

MagicBrush · L1

11 papers, each with its best reported number for its own method; lower is better.

#L1 ↓MethodPaperTopic
10.047Vitron
Vicuna 7B v1.5
Vitron: A Unified Pixel-level Vision LLM for Understanding, Generating, Segmenting, Editing
NeurIPS 2024 · Table: Image editing results
Multimodal LLMs
20.049MUREUnlocking Complex Image Editing via Natively Interleaved Visual Textual CoT with Deep Confidence Reasoning
ECCV 2026 · Table 1
Multimodal LLMs
30.051AnySD w/ AnyEdit
Stable Diffusion 1.5
AnyEdit: Mastering Unified High-Quality Image Editing for Any Idea
CVPR 2025 · Table 4
Editing and style transfer
40.0516Kontext w/ Attention InjectionRePlan: Reasoning-Guided Region Planning for Complex Instruction-Based Image Editing
ECCV 2026 · Table A3
Multimodal LLMs
50.052Emu Edit
Emu
Emu Edit: Precise Image Editing via Recognition and Generation Tasks
CVPR 2024 · Table 2
Editing and style transfer
60.058EditMGT (Ours)EditMGT: Unleashing Potentials of Masked Generative Transformers in Image Editing
CVPR 2026 · Table 1
Editing and style transfer
70.06IEAP
FLUX.1-dev
Image Editing As Programs with Diffusion Models
NeurIPS 2025 · Table 1
Editing and style transfer
80.06ICEdit
FLUX.1 Fill
Enabling Instructional Image Editing with In-Context Generation in Large Scale Diffusion Transformer
NeurIPS 2025 · Table 2
Editing and style transfer
90.065DIM-4.6B-Edit
1.6B generator
Draw-In-Mind: Rebalancing Designer-Painter Roles in Unified Multimodal Models Benefits Image Editing
ICLR 2026 · Table 4
Text-to-image diffusion
100.081UniRealUniReal: Universal Image Generation and Editing via Learning Real-world Dynamics
CVPR 2025 · Table 2
Video generation
1 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.