Counterfactual Reasoning on C-VQA
80.6Numerical Direct AccuracyViperGPT
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| ViperGPT2023.03 | 80.6 | 19.4 | 31.6 | 68.4 | 21.6 | 72.4 | |
| LLaMA-AdapterModel Size=7B2023.03 | 30.1 | 5.8 | 34.3 | 5.6 | 45.8 | 14.5 | |
| LLaVAModel Size=7B2023.03 | 27 | 9.9 | 25 | 15.2 | 58.5 | 4.8 | |
| LLaVAModel Size=13B2023.03 | 24.8 | 11.9 | 20.8 | 21.2 | 56.3 | 4.7 |