Hallucination Evaluation on A-OKVQA (POPE Accuracy Metrics)
93.76Accuracy (Random)OPERA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OPERABackbone=Qwen3-VL2026.03 | 93.76 | 89.5 | 83.86 | |
| WoodpeckerBackbone=InternVL3.52026.03 | 93.73 | 89.8 | 84 | |
| RITUALBackbone=InternVL3.52026.03 | 93.71 | 89.75 | 83.9 | |
| KestrelBackbone=InternVL3.52026.03 | 93.57 | 91.8 | 87.13 | |
| OPERABackbone=InternVL3.52026.03 | 93.55 | 89.79 | 83.81 | |
| VCDBackbone=Qwen3-VL2026.03 | 93.53 | 87.86 | 82 | |
| KestrelBackbone=Qwen3-VL2026.03 | 93.46 | 91.73 | 86.76 | |
| DeGFBackbone=InternVL3.52026.03 | 93.39 | 89.68 | 84.11 | |
| WoodpeckerBackbone=Qwen3-VL2026.03 | 93.23 | 88.9 | 83.33 | |
| DeGFBackbone=Qwen3-VL2026.03 | 92.96 | 87.7 | 82.61 | |
| Qwen3-VL agentBackbone=Qwen3-VL2026.03 | 92.87 | 85.2 | 78.03 | |
| VCDBackbone=InternVL3.52026.03 | 92.87 | 89.73 | 83.73 | |
| BaselineBackbone=InternVL3.52026.03 | 92.67 | 87.83 | 81.53 | |
| BaselineBackbone=Qwen3-VL2026.03 | 92.36 | 86.67 | 81.87 | |
| RITUALBackbone=Qwen3-VL2026.03 | 87.67 | 83.5 | 77.76 |