Image-text alignment on MS-COCO 5K (test)
32.5BLEUGround Truth (upper bound)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Ground Truth (upper bound)Backbone=VL-T52022.06 | 32.5 | 27.5 | 108.3 | 20.4 | — | |
| Partisamples_per_prompt=16, ranking_model=CoCa2022.06 | 26.4 | 23.9 | 83.9 | 16.5 | — | |
| Retrieval BaselineSource=DALL-Eval2022.06 | 24.7 | 23.9 | 84.1 | 16.6 | — | |
| X-LXMERT2022.06 | 18.5 | 19.1 | 55.8 | 12.1 | — | |
| minDALL-E2022.06 | 16.6 | 17.6 | 48 | 10.5 | — | |
| ruDALL-E-XLSource=DALL-Eval2022.06 | 13.9 | 16 | 38.7 | 8.7 | — | |
| DALL-E SmallSource=DALL-Eval2022.06 | 9.3 | 12.9 | 20.2 | 5.6 | — | |
| Random Train ImagesSource=DALL-Eval2022.06 | 4.4 | 9.2 | 4.8 | 2 | — | |
| DiT-ST MediumParameters=2B2025.05 | — | — | — | — | 34.09 | |
| Flux.1 DevParameters=12B2025.05 | — | — | — | — | 31.52 | |
| PixArt-αParameters=0.6B2025.05 | — | — | — | — | 32.58 | |
| SDv3 MediumParameters=2B2025.05 | — | — | — | — | 31.31 | |
| SDv3.5 LargeParameters=8B2025.05 | — | — | — | — | 32.74 |