Image Captioning on ToS-COCO Caption (BLEU-4, CIDEr, SPICE variations)
43.94BLEU-4 AvgUpper-bound (PA)
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Upper-bound (PA)# Trainable Params=12.29 M2026.06 | 43.94 | — | — | 126.91 | — | — | 24.18 | — | — | |
| Upper-bound (Q-Former)# Trainable Params=107.13 M2026.06 | 43.82 | — | — | 126.19 | — | — | 24.1 | — | — | |
| ECA# Trainable Params=12.29 M2026.06 | 43.42 | -0.64 | 7.39 | 125.56 | -1.86 | 20.58 | 23.8 | -0.35 | 3 | |
| CODA-Prompt# Trainable Params=15.41 M2026.06 | 43.1 | -0.67 | 6.9 | 124.2 | -1.19 | 19.44 | 23.52 | -0.38 | 2.59 | |
| Dual-Prompt# Trainable Params=14.30 M2026.06 | 43.03 | -0.62 | 6.77 | 123.59 | -1.6 | 19.04 | 23.47 | -0.52 | 2.65 | |
| LwF# Trainable Params=12.29 M2026.06 | 42.91 | -1.07 | 6.56 | 123.88 | -3.78 | 19.2 | 23.51 | -0.69 | 2.71 | |
| EWC# Trainable Params=12.29 M2026.06 | 42.86 | -1.45 | 6.62 | 123.66 | -4.03 | 18.73 | 23.55 | -0.57 | 2.55 | |
| Vanilla (PA)# Trainable Params=12.29 M2026.06 | 42.7 | -1.49 | 6.48 | 123 | -4.5 | 19.15 | 23.39 | -0.78 | 2.64 | |
| Vanilla (Q-Former)# Trainable Params=107.13 M2026.06 | 42.21 | -2.02 | 6.6 | 123.66 | -4.58 | 19.23 | 23.42 | -0.85 | 2.68 | |
| MoE-LoRA# Trainable Params=98.84 M2026.06 | 42.2 | -1.56 | 6.25 | 122.77 | -3.53 | 17.76 | 23.39 | -0.68 | 2.51 | |
| ZeroShot# Trainable Params=0 M2026.06 | 36 | — | — | 104.65 | — | — | 21.12 | — | — |