Open-ended Medical Visual Question Answering on VinDr
54.2F1 ScoreLobA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| LobAModel Category=Adversarial Techniques2025.04 | 54.2 | 53.3 | 56 | |
| LLaVA-MedModel Category=Open-source Models2025.04 | 52.3 | 51.9 | 53 | |
| PH + LLaVA-MedModel Category=Adversarial Techniques2025.04 | 52.2 | 51 | 52.7 | |
| LISAModel Category=Open-source Models2025.04 | 52.1 | 50.9 | 53.4 | |
| LobA w/o self-promptModel Category=Adversarial Techniques, Self-prompt=false2025.04 | 51.1 | 52.6 | 50.7 | |
| CRG + LLaVA-MedModel Category=Adversarial Techniques2025.04 | 50.9 | 52.4 | 49.4 | |
| VCD + LLaVA-MedModel Category=Adversarial Techniques2025.04 | 50.1 | 54.7 | 49.7 | |
| CheXagentModel Category=Open-source Models2025.04 | 37.4 | 34.2 | 33 | |
| MedFlamingoModel Category=Open-source Models2025.04 | 30.1 | 28.5 | 32.2 | |
| BioMedGPTModel Category=Open-source Models2025.04 | 28.9 | 26.5 | 31 | |
| GPT-4-OModel Category=Proprietary Models2025.04 | 9.7 | 8.7 | 6.9 | |
| GPT-4-VisionModel Category=Proprietary Models2025.04 | 7.4 | 5.3 | 7.2 | |
| Gemini-1.5-Flash-8BModel Category=Proprietary Models2025.04 | 3.8 | 9.8 | 2.3 |