Hallucination Evaluation on GQA
91.8Accuracy (Random)OPERA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OPERABackbone=Qwen3-VL2026.03 | 91.8 | 87.11 | 83.3 | |
| KestrelBackbone=Qwen3-VL2026.03 | 91.67 | 90.33 | 86.27 | |
| VCDBackbone=InternVL3.52026.03 | 91.6 | 85.07 | 83.37 | |
| KestrelBackbone=InternVL3.52026.03 | 91.57 | 89.87 | 86.53 | |
| VCDBackbone=Qwen3-VL2026.03 | 91.56 | 85.76 | 81.93 | |
| DeGFBackbone=InternVL3.52026.03 | 91.48 | 85.06 | 83.2 | |
| OPERABackbone=InternVL3.52026.03 | 91.45 | 85.2 | 83.31 | |
| WoodpeckerBackbone=InternVL3.52026.03 | 91.43 | 85.16 | 83.26 | |
| Qwen3-VL agentBackbone=Qwen3-VL2026.03 | 91.41 | 81.81 | 78.3 | |
| RITUALBackbone=InternVL3.52026.03 | 91.39 | 85.18 | 83.29 | |
| WoodpeckerBackbone=Qwen3-VL2026.03 | 91.27 | 86.27 | 82.77 | |
| DeGFBackbone=Qwen3-VL2026.03 | 91.13 | 83.79 | 82 | |
| BaselineBackbone=Qwen3-VL2026.03 | 90.7 | 83.63 | 81.5 | |
| BaselineBackbone=InternVL3.52026.03 | 89.77 | 84.1 | 81.31 | |
| RITUALBackbone=Qwen3-VL2026.03 | 86.86 | 82.3 | 78.23 |