Object Hallucination Evaluation on MSCOCO
93.87AccuracyMHSA
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MHSAModel=InternVL2-8B2026.05 | 93.87 | 90.16 | 94.16 | 98.54 | 54.9 | |
| VACoDeModel=Qwen3-VL-8B2025.10 | 88.97 | — | 88.47 | — | — | |
| Self-AugModel=Qwen3-VL-8B2025.10 | 88.79 | — | 88.2 | — | — | |
| VCDModel=Qwen3-VL-8B2025.10 | 88.76 | — | 88.22 | — | — | |
| MultinomialModel=Qwen3-VL-8B2025.10 | 88.59 | — | 87.93 | — | — | |
| RVCDModel=LLaVA-1.5, Decoding Strategy=RVCD2025.05 | 88.54 | 89.92 | 88.43 | — | — | |
| RVCDModel=mPLUG-Owl2, Decoding Strategy=RVCD2025.05 | 87.45 | 87.91 | 87.41 | — | — | |
| BaselineModel=InternVL2-8B2026.05 | 87.07 | 90.86 | 86.5 | 82.54 | 45.6 | |
| RVCDModel=MiniGPT-4, Decoding Strategy=RVCD2025.05 | 85.96 | 88.14 | 85.63 | — | — | |
| Self-AugModel=LLaVA-1.5-13B2025.10 | 85.37 | — | 83.96 | — | — | |
| VACoDeModel=LLaVA-1.5-13B2025.10 | 84.86 | — | 84.17 | — | — | |
| VACoDeModel=LLaVA-1.5-7B2025.10 | 84.29 | — | 83.59 | — | — | |
| MultinomialModel=LLaVA-1.5-13B2025.10 | 83.86 | — | 81.02 | — | — | |
| VCDModel=LLaVA-1.5-13B2025.10 | 83.86 | — | 82.68 | — | — | |
| VCDModel=LLaVA-1.5-7B2025.10 | 83.66 | — | 82.55 | — | — | |
| Self-AugModel=LLaVA-1.5-7B2025.10 | 82.93 | — | 83.57 | — | — | |
| Self-AugModel=InstructBLIP2025.10 | 82.86 | — | 82.34 | — | — | |
| MultinomialModel=LLaVA-1.5-7B2025.10 | 82.07 | — | 80.48 | — | — | |
| Beam SearchModel=mPLUG-Owl2, Decoding Strategy=Beam Search2025.05 | 80.17 | 74.3 | 82.64 | — | — | |
| OPERAModel=mPLUG-Owl2, Decoding Strategy=OPERA2025.05 | 79.01 | 72.71 | 81.98 | — | — | |
| VACoDeModel=Qwen-VL2025.10 | 78.35 | — | 76.08 | — | — | |
| Beam SearchModel=LLaVA-1.5, Decoding Strategy=Beam Search2025.05 | 78.27 | 71.94 | 81.28 | — | — | |
| VCDModel=Qwen-VL2025.10 | 77.98 | — | 75.42 | — | — | |
| Self-AugModel=Qwen-VL2025.10 | 77.58 | — | 74.71 | — | — | |
| MultinomialModel=Qwen-VL2025.10 | 75.72 | — | 72.07 | — | — | |
| DoLAModel=mPLUG-Owl2, Decoding Strategy=DoLA2025.05 | 74.56 | 67.43 | 79.34 | — | — | |
| HALCModel=mPLUG-Owl2, Decoding Strategy=HALC2025.05 | 74.54 | 67.42 | 79.33 | — | — | |
| OPERAModel=LLaVA-1.5, Decoding Strategy=OPERA2025.05 | 74.43 | 67.43 | 78.94 | — | — | |
| GreedyModel=mPLUG-Owl2, Decoding Strategy=Greedy2025.05 | 74.36 | 67.23 | 79.23 | — | — | |
| VCDModel=mPLUG-Owl2, Decoding Strategy=VCD2025.05 | 73.66 | 67.33 | 77.93 | — | — | |
| VACoDeModel=InstructBLIP2025.10 | 73.29 | — | 74.26 | — | — | |
| DoLAModel=LLaVA-1.5, Decoding Strategy=DoLA2025.05 | 72.48 | 65.54 | 78.04 | — | — | |
| HALCModel=LLaVA-1.5, Decoding Strategy=HALC2025.05 | 72.48 | 65.54 | 78.04 | — | — | |
| DoLAModel=MiniGPT-4, Decoding Strategy=DoLA2025.05 | 72.26 | 75.41 | 70.86 | — | — | |
| HALCModel=MiniGPT-4, Decoding Strategy=HALC2025.05 | 72.26 | 75.41 | 70.86 | — | — | |
| GreedyModel=LLaVA-1.5, Decoding Strategy=Greedy2025.05 | 72.19 | 65.28 | 77.86 | — | — | |
| VCDModel=InstructBLIP2025.10 | 71.99 | — | 72.77 | — | — | |
| VCDModel=LLaVA-1.5, Decoding Strategy=VCD2025.05 | 69.86 | 63.44 | 75.87 | — | — | |
| MultinomialModel=InstructBLIP2025.10 | 68.7 | — | 69.34 | — | — | |
| Beam SearchModel=MiniGPT-4, Decoding Strategy=Beam Search2025.05 | 67.51 | 62.67 | 73.82 | — | — | |
| OPERAModel=MiniGPT-4, Decoding Strategy=OPERA2025.05 | 67.49 | 62.41 | 73.89 | — | — | |
| GreedyModel=MiniGPT-4, Decoding Strategy=Greedy2025.05 | 62.98 | 58.72 | 72.24 | — | — | |
| VCDModel=MiniGPT-4, Decoding Strategy=VCD2025.05 | 61.79 | 58.97 | 67.34 | — | — |