Visual Question Answering on DocHaystack 100
0.8624AccuracyQwen2-VL-f.t. + V-RAG
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen2-VL-f.t. + V-RAGFramework=V-RAG, Resolution=Low-resolution, Adaptation=LoRA2024.11 | 0.8624 | |
| Qwen2-VL + V-RAGFramework=V-RAG, Resolution=Low-resolution2024.11 | 0.8257 | |
| GPT-4o + V-RAGFramework=V-RAG, Resolution=Low-resolution2024.11 | 0.8165 | |
| Gemini + V-RAGFramework=V-RAG2024.11 | 0.7339 | |
| LLaVA-OV + V-RAGFramework=V-RAG2024.11 | 0.6972 | |
| Gemini2024.11 | 0.5046 | |
| Qwen2-VLResolution=Low-resolution2024.11 | 0.4128 | |
| GPT-4oResolution=Low-resolution2024.11 | 0.2752 | |
| MIRAGE2024.11 | 0.0367 |