Object Hallucination Evaluation on POPE GQA (Adversarial)
84.27AccuracySIRA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SIRABackbone=Qwen2.5-VL2026.05 | 84.27 | — | 85.18 | |
| ICTBackbone=Qwen2.5-VL2026.05 | 84.1 | — | 83.53 | |
| MaskCDBackbone=Qwen2.5-VL2026.05 | 83.25 | — | 84.3 | |
| DoLaBackbone=Qwen2.5-VL2026.05 | 82 | — | 81.51 | |
| OPERABackbone=Qwen2.5-VL2026.05 | 80.24 | — | 80.64 | |
| VCDBackbone=Qwen2.5-VL2026.05 | 80.01 | — | 80.75 | |
| SIRABackbone=LLaVA-v1.52026.05 | 79.6 | — | 81.17 | |
| SIDBackbone=Qwen2.5-VL2026.05 | 79.58 | — | 79.88 | |
| BaseBackbone=Qwen2.5-VL2026.05 | 78.77 | — | 78.56 | |
| ICTBackbone=LLaVA-v1.52026.05 | 77.4 | — | 80.11 | |
| MaskCDBackbone=LLaVA-v1.52026.05 | 77.1 | — | 79.3 | |
| SIDBackbone=LLaVA-v1.52026.05 | 76.62 | — | 78.99 | |
| VCDBackbone=LLaVA-v1.52026.05 | 76.09 | — | 78.78 | |
| BaseBackbone=LLaVA-v1.52026.05 | 75.08 | — | 76.06 | |
| OPERABackbone=LLaVA-v1.52026.05 | 75 | — | 78.71 | |
| DoLaBackbone=LLaVA-v1.52026.05 | 68.73 | — | 74.78 | |
| CoFi-DecBackbone=Qwen-VL2025.12 | 0.8169 | 0.8496 | 0.8167 | |
| DeGFBackbone=Qwen-VL2025.12 | 0.8113 | 0.8418 | 0.8057 | |
| RITUALBackbone=Qwen-VL2025.12 | 0.8107 | 0.8329 | 0.8041 | |
| M3IDBackbone=Qwen-VL2025.12 | 0.8103 | 0.8293 | 0.8094 | |
| VCDBackbone=Qwen-VL2025.12 | 0.8087 | 0.8107 | 0.808 | |
| RegularBackbone=Qwen-VL2025.12 | 0.7903 | 0.8043 | 0.7854 | |
| CoFi-DecBackbone=LLaVA-1.52025.12 | 0.7521 | 0.6834 | 0.7976 | |
| DeGFBackbone=LLaVA-1.52025.12 | 0.7407 | 0.6742 | 0.7822 | |
| CoFi-DecBackbone=InstructBLIP2025.12 | 0.7325 | 0.6968 | 0.7687 | |
| DeGFBackbone=InstructBLIP2025.12 | 0.7245 | 0.6852 | 0.7532 | |
| M3IDBackbone=InstructBLIP2025.12 | 0.7117 | 0.6579 | 0.7536 | |
| VCDBackbone=InstructBLIP2025.12 | 0.71 | 0.6575 | 0.7514 | |
| RITUALBackbone=InstructBLIP2025.12 | 0.7017 | 0.6476 | 0.7478 | |
| VCDBackbone=LLaVA-1.52025.12 | 0.6883 | 0.6226 | 0.7543 | |
| M3IDBackbone=LLaVA-1.52025.12 | 0.6867 | 0.6216 | 0.7528 | |
| RegularBackbone=LLaVA-1.52025.12 | 0.686 | 0.6243 | 0.7484 | |
| RegularBackbone=InstructBLIP2025.12 | 0.686 | 0.6394 | 0.731 | |
| RITUALBackbone=LLaVA-1.52025.12 | 0.6823 | 0.6175 | 0.751 |