Image Captioning on SARVLM-1M (test)
26.1BLEU-1SARCap*
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| SARCap*Image Encoder=ViT-L-14, Pretrain Data=LAION+HQRS, Tune On=SARVLM-1M2025.10 | 26.1 | 17.13 | 11.93 | 8.47 | 22.18 | 26.15 | 29.58 | 88.73 | |
| SARCapImage Encoder=ViT-L-14, Pretrain Data=–, Tune On=SARVLM-1M2025.10 | 22.42 | 14.44 | 10.45 | 7.79 | 19.23 | 21.76 | 24.32 | 88.11 | |
| CoCa-HQRSImage Encoder=ViT-L-14, Pretrain Data=LAION, Tune On=HQRS2025.10 | 10.79 | 4.02 | 1.14 | 0.39 | 8.21 | 13.11 | 7.47 | 84 | |
| CoCa-VanillaImage Encoder=ViT-L-14, Pretrain Data=LAION, Tune On=–2025.10 | 8.75 | 3.21 | 1.14 | 0.32 | 10.54 | 10.67 | 4.17 | 82.23 |