Visual Question Answering on IV-VQA (Near OOD)
0.9525AccuracyAdam-SPD(LORA)
Evaluation Results
| Method | Links | |
|---|---|---|
| Adam-SPD(LORA)Backbone=PaliGemma-3B, Adaptation Method=LoRA2024.11 | 0.9525 | |
| Vanilla FT(LORA)Backbone=PaliGemma-3B, Adaptation Method=LoRA2024.11 | 0.9443 | |
| LP-FT(LORA)Backbone=PaliGemma-3B, Adaptation Method=LoRA2024.11 | 0.933 | |
| Linear Prob.Backbone=PaliGemma-3B, Adaptation Method=Linear Probing2024.11 | 0.8783 | |
| WiSE-FT(LORA)Backbone=PaliGemma-3B, Adaptation Method=LoRA2024.11 | 0.8506 | |
| Zero-ShotBackbone=PaliGemma-3B, Adaptation Method=Zero-Shot2024.11 | 0.6395 |