Image Captioning on ROCO (test)
16.96BLEU@1LCLgap
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| LCLgapLoss=LCLgap (Align True Pairs + Centroid Uniformity), Method Label=Ours2026.02 | 16.96 | 10.07 | 6.09 | 3.64 | 19.9 | 25.22 | |
| CLIP (FT)Temperature Strategy=Fixed (0.07), Loss=Conventional CLIP loss2026.02 | 16.71 | 10.07 | 6.05 | 3.59 | 19.82 | 26.05 | |
| CLIP (LT)Temperature Strategy=Learnable, Loss=Conventional CLIP loss2026.02 | 16.51 | 9.82 | 5.91 | 3.56 | 19.61 | 25.24 |