Visual Question Answering on OK-VQA v1.0 (val)
59.56VQA AccuracyHiFICL
Evaluation Results
| Method | Links | |
|---|---|---|
| HiFICLModel=Idefics2, # Params (M)=2.2 (×1.0)2026.03 | 59.56 | |
| MimICModel=Idefics2, # Params (M)=0.26 (×0.12)2026.03 | 58.74 | |
| 8-shot ICLModel=Idefics2, # Params (M)=-2026.03 | 57.68 | |
| LoRAModel=Idefics2, # Params (M)=17.6 (×8.0)2026.03 | 55.05 | |
| LIVEModel=Idefics2, # Params (M)=0.13 (×0.06)2026.03 | 54.86 | |
| HiFICLModel=LLaVA, # Params (M)=2.2 (×1.0)2026.03 | 54.19 | |
| MimICModel=LLaVA, # Params (M)=17.0 (×7.7)2026.03 | 52.29 | |
| LIVEModel=LLaVA, # Params (M)=0.13 (×0.05)2026.03 | 51.77 | |
| LoRAModel=LLaVA, # Params (M)=19.7 (×8.95)2026.03 | 48.19 | |
| 8-shot ICLModel=LLaVA, # Params (M)=-2026.03 | 43.84 | |
| Zero-shotModel=Idefics2, # Params (M)=-2026.03 | 43.08 | |
| Zero-shotModel=LLaVA, # Params (M)=-2026.03 | 5.1 |