Autoregressive Image Generation on MS-COCO 2017 (val)
0.333CLIP ScoreTarget Model
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Target ModelTarget Model=Lumina-mGPT2026.01 | 0.333 | 28.99 | 0.6855 | 1 | 170.14 | 1 | |
| Eagle-1Target Model=Lumina-mGPT2026.01 | 0.333 | 29.05 | 0.6883 | 2.76 | 71.66 | 2.37 | |
| LANTERN++Target Model=Lumina-mGPT, lambda=2, k=102026.01 | 0.3328 | 30.31 | 0.6697 | 2.99 | 68.64 | 2.48 | |
| COOL-SDTarget Model=Lumina-mGPT, delta=1.12026.01 | 0.3325 | 30.3 | 0.6699 | 3.11 | 63.24 | 2.69 | |
| COOL-SDTarget Model=LlamaGen-XL, delta=1.12026.01 | 0.3167 | 21.02 | -0.0997 | 2.73 | 4.46 | 2.27 | |
| Target ModelTarget Model=LlamaGen-XL2026.01 | 0.3162 | 21.08 | -0.0763 | 1 | 10.11 | 1 | |
| Eagle-1Target Model=LlamaGen-XL2026.01 | 0.3157 | 20.97 | -0.0859 | 2.42 | 4.99 | 2.03 | |
| LANTERN++Target Model=LlamaGen-XL, lambda=2, k=102026.01 | 0.3157 | 21.17 | -0.1155 | 2.67 | 4.7 | 2.15 | |
| COOL-SDTarget Model=LlamaGen-XL, delta=22026.01 | 0.3154 | 21.2 | -0.1353 | 3.34 | 3.72 | 2.72 |