Image Captioning on TextCaps K = 3 (test)
78.6mBLEU-4LoRA-MLE
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| LoRA-MLEDecoding=BS, Beam=6, r=24, alpha=962025.07 | 78.6 | 1.557 | 24.6 | 89.5 | |
| LoRA-MLEDecoding=BS, Beam=3, r=24, alpha=962025.07 | 68.8 | 1.517 | 24.4 | 87.3 | |
| LoRA-MoEDecoding=DBS (λ = 0.8), Beam=62025.07 | 67.8 | 1.608 | 25.5 | 92.2 | |
| LoRA-MLEDecoding=DBS (λ = 0.8), Beam=6, r=24, alpha=962025.07 | 67.1 | 1.573 | 25.1 | 90.3 | |
| LoRA-MLEDecoding=DBS (λ = 0.8), Beam=3, r=8, alpha=32, note=marked †2025.07 | 53.1 | 1.589 | 25.5 | 91.2 | |
| LoRA-MCLDecoding=BS, Beam=1, r=8, alpha=32, epsilon=0.12025.07 | 52 | 1.674 | 25.5 | 95.5 | |
| LoRA-MCLDecoding=BS, Beam=2, r=8, alpha=32, epsilon=0.12025.07 | 49 | 1.627 | 25.8 | 93.2 | |
| LoRA-MoEDecoding=DBS (λ = 0.8), Beam=32025.07 | 44.1 | 1.616 | 25.4 | 92.4 | |
| LoRA-MLEDecoding=DBS (λ = 0.8), Beam=3, r=24, alpha=962025.07 | 43.7 | 1.59 | 25.1 | 90.9 | |
| LoRA-MLEDecoding=DBS (λ = 1.0), Beam=3, r=8, alpha=32, note=marked †2025.07 | 42.5 | 1.601 | 25.2 | 91.5 | |
| LoRA-MoEDecoding=DBS (λ = 1.0), Beam=32025.07 | 42.1 | 1.622 | 25.3 | 92.6 | |
| LoRA-MLEDecoding=DBS (λ = 1.0), Beam=3, r=24, alpha=962025.07 | 41.6 | 1.586 | 25 | 90.6 |