Reliability prediction on POPE Adversarial (test)
97.1AUROCQwen2-VL-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2-VL-7BReliability source=Probe2026.06 | 97.1 | |
| LLaVA-1.5-7BReliability source=Probe2026.06 | 95.6 | |
| PaliGemma-3BReliability source=Probe2026.06 | 73.8 |
| Method | Links | |
|---|---|---|
| Qwen2-VL-7BReliability source=Probe2026.06 | 97.1 | |
| LLaVA-1.5-7BReliability source=Probe2026.06 | 95.6 | |
| PaliGemma-3BReliability source=Probe2026.06 | 73.8 |