Object Hallucination Evaluation on POPE MSCOCO
93.97F1 ScoreMHSA
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MHSALVLM Architecture=LLaVA-v1.5-7B, Training Dataset=Obj3652026.05 | 93.97 | — | — | — | — | |
| MHSALVLM Architecture=LLaVA-v1.5-7B, Training Dataset=COCO2026.05 | 92.27 | — | — | — | — | |
| MHSAModel=LLaVA-v1.5-7B2026.05 | 92.27 | 92.1 | 90.7 | 93.89 | 52 | |
| VLIModel=Qwen3-VL2026.01 | 92.19 | 92.58 | — | — | — | |
| VCDModel=Qwen3-VL2026.01 | 91.37 | 91.92 | — | — | — | |
| CICDModel=Qwen3-VL2026.01 | 91.08 | 91.43 | — | — | — | |
| OriginModel=Qwen3-VL2026.01 | 90.53 | 91.14 | — | — | — | |
| OPERAModel=Qwen3-VL2026.01 | 90.28 | 90.87 | — | — | — | |
| VCDBase Model=LLaVA-1.5-7B, Sampling Strategy=Random2026.06 | 90 | 89.9 | — | — | — | |
| ICLABackbone=LLaVA1.5-7B, Setting=Random2026.02 | 89.95 | 89.93 | — | — | — | |
| VTIModel=Qwen3-VL2026.01 | 89.94 | 90.62 | — | — | — | |
| DoLABackbone=LLaVA1.5-7B, Setting=Random2026.02 | 89.77 | 89.7 | — | — | — | |
| VanillaBackbone=LLaVA1.5-7B, Setting=Random2026.02 | 89.71 | 89.6 | — | — | — | |
| DAMOBackbone=LLaVA1.5-7B, Setting=Random2026.02 | 89.69 | 89.53 | — | — | — | |
| POVIDBackbone=LLaVA1.5-7B, Setting=Random2026.02 | 89.56 | 89.5 | — | — | — | |
| DeCoBackbone=LLaVA1.5-7B, Setting=Random2026.02 | 89.29 | 88.83 | — | — | — | |
| VLIModel=LLaVA-1.52026.01 | 89.27 | 89.61 | — | — | — | |
| VCDBase Model=InstructBLIP-7B, Sampling Strategy=Random2026.06 | 88.6 | 89.1 | — | — | — | |
| ClearSightModel=LLaVA-1.52026.01 | 88.41 | 88.74 | — | — | — | |
| FADEBase Model=LLaVA-1.5-7B, Sampling Strategy=Random2026.06 | 88.3 | 89.2 | — | — | — | |
| VCDBase Model=mPLUG-Owl2-7B, Sampling Strategy=Random2026.06 | 88.1 | 87.9 | — | — | — | |
| DAMOBase Model=InstructBLIP-7B, Sampling Strategy=Random2026.06 | 88.1 | 88.2 | — | — | — | |
| OPERAModel=LLaVA-1.52026.01 | 87.96 | 88.34 | — | — | — | |
| VISTABase Model=LLaVA-1.5-7B, Sampling Strategy=Random2026.06 | 87.8 | 88.8 | — | — | — | |
| PAIBase Model=mPLUG-Owl2-7B, Sampling Strategy=Random2026.06 | 87.8 | 88.5 | — | — | — | |
| FADEBase Model=mPLUG-Owl2-7B, Sampling Strategy=Random2026.06 | 87.8 | 88.5 | — | — | — | |
| VTIModel=LLaVA-1.52026.01 | 87.69 | 87.95 | — | — | — | |
| NulluModel=Qwen3-VL2026.01 | 87.62 | 88.76 | — | — | — | |
| VCDBackbone=LLaVA1.5-7B, Setting=Random2026.02 | 87.6 | 87.37 | — | — | — | |
| PAIBase Model=LLaVA-1.5-7B, Sampling Strategy=Random2026.06 | 87.4 | 88.5 | — | — | — | |
| GreedyBase Model=mPLUG-Owl2-7B, Sampling Strategy=Random2026.06 | 87.4 | 88.2 | — | — | — | |
| VISTABase Model=mPLUG-Owl2-7B, Sampling Strategy=Random2026.06 | 87.4 | 88.2 | — | — | — | |
| GreedyBase Model=LLaVA-1.5-7B, Sampling Strategy=Random2026.06 | 87.3 | 88.5 | — | — | — | |
| CICDModel=LLaVA-1.52026.01 | 87.13 | 86.46 | — | — | — | |
| VCDBase Model=LLaVA-1.5-7B, Sampling Strategy=Popular2026.06 | 87.1 | 86.6 | — | — | — | |
| DAMOBase Model=mPLUG-Owl2-7B, Sampling Strategy=Random2026.06 | 87 | 87.9 | — | — | — | |
| ICLABackbone=LLaVA1.5-7B, Setting=Popular2026.02 | 86.91 | 86.43 | — | — | — | |
| VISTABase Model=InstructBLIP-7B, Sampling Strategy=Random2026.06 | 86.9 | 88 | — | — | — | |
| FADEBase Model=LLaVA-1.5-7B, Sampling Strategy=Popular2026.06 | 86.9 | 87.7 | — | — | — | |
| NulluModel=LLaVA-1.52026.01 | 86.84 | 87.18 | — | — | — | |
| DoLABackbone=LLaVA1.5-7B, Setting=Popular2026.02 | 86.84 | 86.3 | — | — | — | |
| DAMOBackbone=LLaVA1.5-7B, Setting=Popular2026.02 | 86.84 | 86.2 | — | — | — | |
| VanillaBackbone=LLaVA1.5-7B, Setting=Popular2026.02 | 86.82 | 86.23 | — | — | — | |
| POVIDBackbone=LLaVA1.5-7B, Setting=Popular2026.02 | 86.78 | 86.23 | — | — | — | |
| VISTABase Model=LLaVA-1.5-7B, Sampling Strategy=Popular2026.06 | 86.5 | 87.4 | — | — | — | |
| PAIBase Model=LLaVA-1.5-7B, Sampling Strategy=Popular2026.06 | 86.3 | 87.4 | — | — | — | |
| GreedyBase Model=LLaVA-1.5-7B, Sampling Strategy=Popular2026.06 | 86.1 | 87.2 | — | — | — | |
| DeCoBackbone=LLaVA1.5-7B, Setting=Popular2026.02 | 85.9 | 84.73 | — | — | — | |
| PAIBase Model=mPLUG-Owl2-7B, Sampling Strategy=Popular2026.06 | 85.9 | 86.4 | — | — | — | |
| FADEBase Model=mPLUG-Owl2-7B, Sampling Strategy=Popular2026.06 | 85.9 | 86.4 | — | — | — | |
| GreedyBase Model=InstructBLIP-7B, Sampling Strategy=Random2026.06 | 85.8 | 87.2 | — | — | — | |
| DAMOBase Model=mPLUG-Owl2-7B, Sampling Strategy=Popular2026.06 | 85.6 | 86.4 | — | — | — | |
| VISTABase Model=mPLUG-Owl2-7B, Sampling Strategy=Popular2026.06 | 85.6 | 86.2 | — | — | — | |
| FADEBase Model=InstructBLIP-7B, Sampling Strategy=Random2026.06 | 85.5 | 86.9 | — | — | — | |
| GreedyBase Model=mPLUG-Owl2-7B, Sampling Strategy=Popular2026.06 | 85.5 | 86.2 | — | — | — | |
| VCDBase Model=InstructBLIP-7B, Sampling Strategy=Popular2026.06 | 85.2 | 85.2 | — | — | — | |
| VCDModel=LLaVA-1.52026.01 | 85.14 | 84.67 | — | — | — | |
| VCDBackbone=LLaVA1.5-7B, Setting=Popular2026.02 | 84.8 | 83.97 | — | — | — | |
| DAMOBase Model=LLaVA-1.5-7B, Sampling Strategy=Random2026.06 | 84.6 | 86.4 | — | — | — | |
| VISTABase Model=InstructBLIP-7B, Sampling Strategy=Popular2026.06 | 84.6 | 85.5 | — | — | — | |
| FADEBase Model=LLaVA-1.5-7B, Sampling Strategy=Adversarial2026.06 | 84.6 | 85.2 | — | — | — | |
| BaselineLVLM Architecture=LLaVA-v1.5-7B2026.05 | 84.53 | — | — | — | — | |
| BaselineModel=LLaVA-v1.5-7B2026.05 | 84.53 | 85.57 | 91.49 | 78.55 | 43.1 | |
| VISTABase Model=LLaVA-1.5-7B, Sampling Strategy=Adversarial2026.06 | 84.5 | 85.2 | — | — | — | |
| VCDBase Model=mPLUG-Owl2-7B, Sampling Strategy=Popular2026.06 | 84.4 | 83.6 | — | — | — | |
| PAIBase Model=LLaVA-1.5-7B, Sampling Strategy=Adversarial2026.06 | 84.3 | 85.3 | — | — | — | |
| GreedyBase Model=LLaVA-1.5-7B, Sampling Strategy=Adversarial2026.06 | 84.2 | 85.1 | — | — | — | |
| OriginModel=LLaVA-1.52026.01 | 84.18 | 83.82 | — | — | — | |
| PAIBase Model=mPLUG-Owl2-7B, Sampling Strategy=Adversarial2026.06 | 83.9 | 84.2 | — | — | — | |
| FADEBase Model=mPLUG-Owl2-7B, Sampling Strategy=Adversarial2026.06 | 83.9 | 84.2 | — | — | — | |
| DAMOBase Model=LLaVA-1.5-7B, Sampling Strategy=Popular2026.06 | 83.7 | 85.4 | — | — | — | |
| GreedyBase Model=InstructBLIP-7B, Sampling Strategy=Popular2026.06 | 83.6 | 84.8 | — | — | — | |
| GreedyBase Model=mPLUG-Owl2-7B, Sampling Strategy=Adversarial2026.06 | 83.6 | 84 | — | — | — | |
| DAMOBase Model=mPLUG-Owl2-7B, Sampling Strategy=Adversarial2026.06 | 83.6 | 84.2 | — | — | — | |
| VISTABase Model=mPLUG-Owl2-7B, Sampling Strategy=Adversarial2026.06 | 83.6 | 84 | — | — | — | |
| FADEBase Model=InstructBLIP-7B, Sampling Strategy=Popular2026.06 | 83.4 | 84.7 | — | — | — | |
| DAMOBase Model=InstructBLIP-7B, Sampling Strategy=Popular2026.06 | 83.3 | 82.4 | — | — | — | |
| ClearSightModel=Qwen3-VL2026.01 | 83.18 | 85.04 | — | — | — | |
| VCDBase Model=LLaVA-1.5-7B, Sampling Strategy=Adversarial2026.06 | 82.7 | 81.2 | — | — | — | |
| ICLABackbone=Qwen2.5-VL-7B, Setting=Random2026.02 | 82.59 | 85.1 | — | — | — | |
| VCDBase Model=InstructBLIP-7B, Sampling Strategy=Adversarial2026.06 | 82.5 | 82 | — | — | — | |
| VISTABase Model=InstructBLIP-7B, Sampling Strategy=Adversarial2026.06 | 82.5 | 83 | — | — | — | |
| DAMOBase Model=LLaVA-1.5-7B, Sampling Strategy=Adversarial2026.06 | 82.3 | 83.9 | — | — | — | |
| GreedyBase Model=InstructBLIP-7B, Sampling Strategy=Adversarial2026.06 | 82 | 83 | — | — | — | |
| ICLABackbone=Qwen2.5-VL-7B, Setting=Popular2026.02 | 81.92 | 84.37 | — | — | — | |
| ICLABackbone=LLaVA1.5-7B, Setting=Adversarial2026.02 | 81.9 | 80.13 | — | — | — | |
| FADEBase Model=InstructBLIP-7B, Sampling Strategy=Adversarial2026.06 | 81.9 | 82.9 | — | — | — | |
| POVIDBackbone=LLaVA1.5-7B, Setting=Adversarial2026.02 | 81.88 | 80.03 | — | — | — | |
| DAMOBackbone=Qwen2.5-VL-7B, Setting=Random2026.02 | 81.83 | 84.57 | — | — | — | |
| DeCoBackbone=Qwen2.5-VL-7B, Setting=Random2026.02 | 81.78 | 84.5 | — | — | — | |
| VanillaBackbone=LLaVA1.5-7B, Setting=Adversarial2026.02 | 81.76 | 79.77 | — | — | — | |
| DoLABackbone=LLaVA1.5-7B, Setting=Adversarial2026.02 | 81.69 | 79.73 | — | — | — | |
| DAMOBackbone=LLaVA1.5-7B, Setting=Adversarial2026.02 | 81.65 | 79.53 | — | — | — | |
| VCDBackbone=Qwen2.5-VL-7B, Setting=Random2026.02 | 81.64 | 79.17 | — | — | — | |
| VCDBase Model=mPLUG-Owl2-7B, Sampling Strategy=Adversarial2026.06 | 81.6 | 79.9 | — | — | — | |
| VanillaBackbone=Qwen2.5-VL-7B, Setting=Random2026.02 | 81.58 | 84.37 | — | — | — | |
| ICLABackbone=Qwen2.5-VL-7B, Setting=Adversarial2026.02 | 81.5 | 83.97 | — | — | — | |
| DAMOBackbone=Qwen2.5-VL-7B, Setting=Popular2026.02 | 81.25 | 83.9 | — | — | — | |
| DeCoBackbone=Qwen2.5-VL-7B, Setting=Popular2026.02 | 81.24 | 83.9 | — | — | — | |
| DeCoBackbone=LLaVA1.5-7B, Setting=Adversarial2026.02 | 81.14 | 78.37 | — | — | — |