Object Hallucination Evaluation on MSCOCO (Random)
91.5AccuracyHDD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HDDModel=LLaVA-NeXT, Base Decoding=Greedy2025.12 | 91.5 | 91.3 | |
| MESAModel=LLaVA-v1.52026.04 | 90.27 | 90.2 | |
| HDDModel=LLaVA-1.5, Base Decoding=Greedy2025.12 | 90.1 | 90 | |
| SIDModel=LLaVA-NeXT, Base Decoding=Greedy2025.12 | 90 | 90 | |
| CAIBackbone=Qwen-VL2026.06 | 89.9 | 89.49 | |
| VCDBackbone=Qwen-VL2026.06 | 89.59 | 89.18 | |
| CAIBackbone=InstructBLIP2026.06 | 89.55 | 89.35 | |
| SIDModel=LLaVA-1.5, Base Decoding=Greedy2025.12 | 89.5 | 89.6 | |
| GreedyModel=LLaVA-NeXT, Base Decoding=Greedy2025.12 | 89.3 | 88.8 | |
| CAIBackbone=LLaVA-1.52026.06 | 89.24 | 89.14 | |
| HDDModel=InstructBLIP, Base Decoding=Greedy2025.12 | 89.2 | 88.7 | |
| NulluModel=LLaVA-v1.52026.04 | 89.03 | 88.52 | |
| ONLYBackbone=Qwen-VL2026.06 | 88.97 | 88.37 | |
| ICTModel=LLaVA-v1.52026.04 | 88.96 | 87.85 | |
| MESAModel=Qwen-VL2026.04 | 88.83 | 87.11 | |
| GreedyModel=LLaVA-1.5, Base Decoding=Greedy2025.12 | 88.7 | 87.3 | |
| VCDModel=LLaVA-1.5, Base Decoding=Greedy2025.12 | 88.7 | 88.3 | |
| VCDModel=Qwen-VL2026.04 | 88.63 | 87.81 | |
| ONLYBackbone=LLaVA-1.52026.06 | 88.38 | 88.41 | |
| NulluModel=Qwen-VL2026.04 | 88.26 | 86.13 | |
| ICTModel=Qwen-VL2026.04 | 88.06 | 86.55 | |
| VTIModel=Qwen-VL2026.04 | 87.97 | 86.79 | |
| VCDModel=LLaVA-NeXT, Base Decoding=Greedy2025.12 | 87.8 | 87 | |
| VCDModel=LLaVA-v1.52026.04 | 87.73 | 87.16 | |
| VTIModel=LLaVA-v1.52026.04 | 87.7 | 86.45 | |
| VAFModel=LLaVA-v1.52026.04 | 87.36 | 85.94 | |
| ONLYBackbone=InstructBLIP2026.06 | 87.32 | 87.77 | |
| VanillaModel=Qwen-VL2026.04 | 87.23 | 85.5 | |
| SIDModel=InstructBLIP, Base Decoding=Greedy2025.12 | 87.2 | 87 | |
| GreedyModel=InstructBLIP, Base Decoding=Greedy2025.12 | 87 | 86.3 | |
| VAFModel=Qwen-VL2026.04 | 86.8 | 84.98 | |
| PAIBackbone=LLaVA-1.52026.06 | 86.67 | 87.26 | |
| VCDModel=InstructBLIP, Base Decoding=Greedy2025.12 | 86.5 | 85.9 | |
| VCDBackbone=LLaVA-1.52026.06 | 85.74 | 86.7 | |
| PAIBackbone=Qwen-VL2026.06 | 85.74 | 84.2 | |
| VanillaModel=LLaVA-v1.52026.04 | 85.53 | 88.8 | |
| VCDBackbone=InstructBLIP2026.06 | 85.53 | 85.32 | |
| RegularBackbone=LLaVA-1.52026.06 | 85.46 | 86.07 | |
| PAIBackbone=InstructBLIP2026.06 | 85.43 | 83.6 | |
| ICDModel=LLaVA-v1.52026.04 | 85.23 | 83.52 | |
| RegularBackbone=Qwen-VL2026.06 | 85.09 | 83.46 | |
| RegularBackbone=InstructBLIP2026.06 | 82.82 | 83.56 | |
| ICDModel=Qwen-VL2026.04 | 82.43 | 78.89 |