Hallucination Evaluation on MS-COCO
91.6Accuracy (Rand.)RITUAL
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| RITUALBackbone=InternVL3.52026.03 | 91.6 | 89.03 | 87.48 | |
| KestrelBackbone=Qwen3-VL2026.03 | 91.53 | 89.3 | 87.53 | |
| OPERABackbone=InternVL3.52026.03 | 91.53 | 89.18 | 87.55 | |
| DeGFBackbone=InternVL3.52026.03 | 91.43 | 89.12 | 87.37 | |
| VCDBackbone=InternVL3.52026.03 | 91.35 | 89.22 | 87.6 | |
| KestrelBackbone=InternVL3.52026.03 | 91.27 | 89.27 | 88.1 | |
| WoodpeckerBackbone=InternVL3.52026.03 | 91.2 | 89.11 | 87.5 | |
| Qwen3-VL agentBackbone=Qwen3-VL2026.03 | 91.03 | 88.06 | 86.13 | |
| BaselineBackbone=InternVL3.52026.03 | 90.77 | 88.1 | 85.73 | |
| OPERABackbone=Qwen3-VL2026.03 | 90.5 | 88.83 | 87.5 | |
| VCDBackbone=Qwen3-VL2026.03 | 90.4 | 88.8 | 87.41 | |
| DeGFBackbone=Qwen3-VL2026.03 | 90.33 | 88.16 | 86.9 | |
| WoodpeckerBackbone=Qwen3-VL2026.03 | 89.97 | 88.03 | 87.1 | |
| BaselineBackbone=Qwen3-VL2026.03 | 89 | 86.92 | 86.2 | |
| RITUALBackbone=Qwen3-VL2026.03 | 86.2 | 83.67 | 82.27 |