Visual Question Answering on VQARAD (AUROC, AUARC, ECE)
0.752AUROCSGPU
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SGPUEvaluation=Averaged across LVLM architectures2025.12 | 0.752 | 0.74 | 0.21 | |
| UMPIREEvaluation=Averaged across LVLM architectures2025.12 | 0.724 | 0.622 | 0.26 | |
| PEEvaluation=Averaged across LVLM architectures2025.12 | 0.705 | 0.624 | 0.132 | |
| Cos EigenscoreEvaluation=Averaged across LVLM architectures2025.12 | 0.702 | 0.618 | 0.201 | |
| Cov EigenscoreEvaluation=Averaged across LVLM architectures2025.12 | 0.695 | 0.616 | 0.234 | |
| SEEvaluation=Averaged across LVLM architectures2025.12 | 0.691 | 0.621 | 0.282 | |
| DSEEvaluation=Averaged across LVLM architectures2025.12 | 0.691 | 0.599 | 0.292 | |
| KLE-MaternEvaluation=Averaged across LVLM architectures2025.12 | 0.684 | 0.634 | 0.159 | |
| KLE-HeatEvaluation=Averaged across LVLM architectures2025.12 | 0.644 | 0.64 | 0.23 |