Image Captioning on Nocaps (Clean CIDEr, Robust CIDEr)
105.7Clean CIDErOurs (2 trees)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Ours (2 trees)Architecture=CoCa, Zero-shot=true2026.04 | 105.7 | 64.6 | |
| OursArchitecture=CoCa, Zero-shot=true2026.04 | 105.4 | 63.8 | |
| PMG-FTArchitecture=CoCa, Zero-shot=true2026.04 | 102.1 | 60.2 | |
| FAREArchitecture=CoCa, Zero-shot=true2026.04 | 101.8 | 60.5 | |
| Ours (2 trees)Architecture=BLIP, Zero-shot=true2026.04 | 101.4 | 63 | |
| OursArchitecture=BLIP, Zero-shot=true2026.04 | 101 | 62.5 | |
| TeCoAArchitecture=CoCa, Zero-shot=true2026.04 | 100.5 | 59.2 | |
| FAREArchitecture=BLIP, Zero-shot=true2026.04 | 98.1 | 58.7 | |
| PMG-FTArchitecture=BLIP, Zero-shot=true2026.04 | 97.5 | 58.2 | |
| TeCoAArchitecture=BLIP, Zero-shot=true2026.04 | 96.9 | 57.8 |