Visual Question Answering on TextVQA Far OOD
43.56AccuracyAdam-SPD(LORA)
Evaluation Results
| Method | Links | |
|---|---|---|
| Adam-SPD(LORA)Backbone=PaliGemma-3B2024.11 | 43.56 | |
| Vanilla FT(LORA)Backbone=PaliGemma-3B2024.11 | 42.08 | |
| WiSE-FT(LORA)Backbone=PaliGemma-3B2024.11 | 36.98 | |
| LP-FT(LORA)Backbone=PaliGemma-3B2024.11 | 31.41 | |
| Linear Prob.Backbone=PaliGemma-3B2024.11 | 29.61 | |
| Zero-ShotBackbone=PaliGemma-3B2024.11 | 14.86 |