Object Hallucination Evaluation on POPE GQA (random)
92.23AccuracySIRA
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SIRABackbone=Qwen2.5-VL2026.05 | 92.23 | 92.09 | — | — | |
| CMPCBase Model=LLaVA 1.52025.01 | 89.53 | 88.8 | 90.62 | 88.2 | |
| SIRABackbone=LLaVA-v1.52026.05 | 89.47 | 89.53 | — | — | |
| MaskCDBackbone=Qwen2.5-VL2026.05 | 89.25 | 89.1 | — | — | |
| ICTBackbone=LLaVA-v1.52026.05 | 89.23 | 89.34 | — | — | |
| ICTBackbone=Qwen2.5-VL2026.05 | 88.96 | 87.89 | — | — | |
| OPERABackbone=LLaVA-v1.52026.05 | 88.13 | 88.91 | — | — | |
| MaskCDBackbone=LLaVA-v1.52026.05 | 87.85 | 88.25 | — | — | |
| DoLaBackbone=Qwen2.5-VL2026.05 | 87.13 | 86.05 | — | — | |
| SIDBackbone=LLaVA-v1.52026.05 | 86.97 | 87.17 | — | — | |
| VCDBackbone=LLaVA-v1.52026.05 | 86.65 | 86.99 | — | — | |
| VCDBase Model=LLaVA 1.52025.01 | 86.65 | 86.99 | 84.85 | 89.24 | |
| CMPCBase Model=InstructBLIP2025.01 | 86.03 | 85.72 | 87.67 | 83.87 | |
| OPERABackbone=Qwen2.5-VL2026.05 | 86.02 | 85.29 | — | — | |
| ICDBase Model=LLaVA 1.52025.01 | 85.73 | 84.79 | 90.79 | 79.53 | |
| VCDBackbone=Qwen2.5-VL2026.05 | 85.59 | 85.33 | — | — | |
| BaseBackbone=Qwen2.5-VL2026.05 | 85.1 | 83.87 | — | — | |
| ICDBase Model=InstructBLIP2025.01 | 85.1 | 85.29 | 84.21 | 86.4 | |
| SIDBackbone=Qwen2.5-VL2026.05 | 84.86 | 84.25 | — | — | |
| BaseBackbone=LLaVA-v1.52026.05 | 83.73 | 82.95 | — | — | |
| BaselineBase Model=LLaVA 1.52025.01 | 83.73 | 82.95 | 87.16 | 79.12 | |
| DoLaBackbone=LLaVA-v1.52026.05 | 83.7 | 85.29 | — | — | |
| VCDBase Model=InstructBLIP2025.01 | 83.69 | 84.16 | 81.84 | 86.61 | |
| BaselineBase Model=InstructBLIP2025.01 | 79.65 | 80.56 | 77.14 | 84.29 |