Visual Question Answering on S3VQA
29.3Model Score (GPT-4)ReflectiVA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ReflectiVALLM=LLaMA-3.1-8B, Zero-shot=true2024.11 | 29.3 | 16.8 | |
| ReflectiVA (w/o KB)LLM=LLaMA-3.1-8B, Zero-shot=true, Knowledge Base=None2024.11 | 26.9 | — | |
| LLaVA-v1.5LLM=LLaMA-3.1-8B, Zero-shot=true2024.11 | 24.4 | — | |
| LLaVA-v1.5LLM=Vicuna-7B, Zero-shot=true2024.11 | 23.9 | — | |
| Wiki-LLaVA (E-VQA)LLM=LLaMA-3.1-8B, Zero-shot=true2024.11 | 22.7 | — | |
| Wiki-LLaVA (InfoSeek)LLM=LLaMA-3.1-8B, Zero-shot=true2024.11 | 21.8 | — |