Reliability prediction on VQA v2
89.2AUROCPaliGemma-3B
Evaluation Results
| Method | Links | |
|---|---|---|
| PaliGemma-3BReliability source=Output2026.06 | 89.2 | |
| Qwen2-VL-7BReliability source=Output2026.06 | 89.2 | |
| PaliGemma-3BReliability source=Probe2026.06 | 79.5 | |
| Qwen2-VL-7BReliability source=Probe2026.06 | 77.8 | |
| LLaVA-1.5-7BReliability source=Probe2026.06 | 74.5 | |
| LLaVA-1.5-7BReliability source=Output2026.06 | 55.9 |