Multimodal Image Generation on DreamBench++ (test)
95CP (Animal)Lumina-mGPT
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Lumina-mGPTT2I Model=Chameleon, Train Data=10M, Model Size=7.00B, Evaluation Protocol=Test-Time Tuning-Free2025.07 | 95 | 97 | 89 | 85 | 91 | 31 | 25 | 15 | 25 | 23 | 3.63 | |
| IP-A-Plus ViT-HT2I Model=SDXL v1.0, Train Data=10M, Model Size=3.00B, Evaluation Protocol=Test-Time Tuning-Free2025.07 | 90 | 84 | 76 | 91 | 84 | 50 | 39 | 28 | 41 | 35 | 2.02 | |
| Unified-IO2*T2I Model=Unified-IO2, Train Data=8.5B, Model Size=7.00B, Evaluation Protocol=Test-Time Tuning-Free, trained from scratch=true2025.07 | 77 | 80 | 64 | 82 | 72 | 24 | 18 | 11 | 19 | 14 | 3.74 | |
| DreamEngineT2I Model=SD3.5, Train Data=21M, Model Size=10.50B, Evaluation Protocol=Test-Time Tuning-Free2025.07 | 76 | 72 | 61 | 73 | 68 | 44 | 37 | 25 | 37 | 26 | 1.84 | |
| DreamBooth-LT2I Model=SDXL v1.0, Model Size=2.60B, Evaluation Protocol=Finetuned on Test Set2025.07 | 75 | 31 | 54 | 71 | 60 | 90 | 90 | 75 | 87 | 52 | 0.69 | |
| BLIP-DiffusionT2I Model=SD v1.5, Train Data=130M, Model Size=1.56B, Evaluation Protocol=Test-Time Tuning-Free2025.07 | 67 | 56 | 47 | 51 | 55 | 58 | 52 | 30 | 50 | 27 | 1.11 | |
| Emu2T2I Model=SDXL v1.0, Train Data=16M, Model Size=37.00B, Evaluation Protocol=Test-Time Tuning-Free2025.07 | 67 | 55 | 45 | 44 | 53 | 73 | 73 | 56 | 69 | 36 | 0.76 | |
| IP-A ViT-GT2I Model=SDXL v1.0, Train Data=10M, Model Size=2.50B, Evaluation Protocol=Test-Time Tuning-Free2025.07 | 67 | 56 | 50 | 75 | 59 | 74 | 63 | 45 | 64 | 38 | 0.92 | |
| MENTORT2I Model=LlamaGen, Train Data=3M, Model Size=2.31B, Evaluation Protocol=Test-Time Tuning-Free2025.07 | 65 | 36 | 57 | 47 | 56 | 86 | 85 | 80 | 84 | 47 | 0.66 | |
| DreamBoothT2I Model=SD v1.5, Model Size=860M, Evaluation Protocol=Finetuned on Test Set2025.07 | 64 | 19 | 49 | 48 | 49 | 79 | 78 | 51 | 72 | 36 | 0.68 | |
| Kosmos-GT2I Model=SD v1.5, Train Data=200M, Model Size=3.00B, Evaluation Protocol=Test-Time Tuning-Free2025.07 | 62 | 64 | 46 | 56 | 55 | 48 | 62 | 39 | 51 | 28 | 1.07 | |
| Textual Inv.T2I Model=SD v1.5, Model Size=860M, Evaluation Protocol=Finetuned on Test Set2025.07 | 50 | 35 | 30 | 36 | 38 | 68 | 70 | 44 | 63 | 24 | 0.6 | |
| OmniGen*T2I Model=OmniGen, Train Data=700M, Model Size=3.80B, Evaluation Protocol=Test-Time Tuning-Free, trained from scratch=true2025.07 | 39 | 35 | 25 | 22 | 30 | 70 | 71 | 67 | 70 | 21 | 0.43 | |
| VARGPT-v1.1T2I Model=Infinity (VAR), Train Data=8.3M, Model Size=9.00B, Evaluation Protocol=Test-Time Tuning-Free2025.07 | 25 | 13 | 14 | 27 | 19 | 31 | 48 | 33 | 37 | 7 | 0.51 |