Object Hallucination Evaluation on POPE Random
94AccuracyR-CoV
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| R-CoVBase Model=LLaVA-1.52026.04 | 94 | — | 93.67 | — | |
| VanillaBase Model=LLaVA-1.52026.04 | 93 | — | 93.02 | — | |
| LogicCheckGPTBase Model=LLaVA-1.52026.04 | 92.67 | — | 92.25 | — | |
| R-CoVBase Model=mPLUG-Owl2026.04 | 91.67 | — | 91.35 | — | |
| SpecEyesBase Model=Thyme, Aggregation Strategy=log2026.03 | 91.47 | 1.59 | — | — | |
| CRG(ours)Model=Qwen2.5-VL-7B-Instruct2026.05 | 91.45 | — | 89.23 | — | |
| ACEBase Model=LLaVA-NeXT2026.05 | 91.28 | — | 90.63 | — | |
| SpecEyesBase Model=Thyme, Aggregation Strategy=mean2026.03 | 91.27 | 1.65 | — | — | |
| SpecEyesBase Model=Thyme, Aggregation Strategy=bottom2026.03 | 91.27 | 1.73 | — | — | |
| SpecEyesBase Model=Thyme, Aggregation Strategy=min2026.03 | 91.27 | 1.7 | — | — | |
| ACEBase Model=LLaVA-1.52026.05 | 90.87 | — | 88.93 | — | |
| SpecEyesBase Model=DeepEyes, Aggregation Strategy=log2026.03 | 90.5 | 2.05 | — | — | |
| VTIModel=Qwen2.5-VL-7B-Instruct2026.05 | 90.43 | — | 89.44 | — | |
| SelfCheckBase Model=LLaVA-1.52026.04 | 90.33 | — | 89.53 | — | |
| CRG(ours)Model=LLaVA-1.5-7B2026.05 | 90.3 | — | 89.51 | — | |
| SpecReasonBase Model=Thyme2026.03 | 90.27 | 0.46 | — | — | |
| ThymeBase Model=Thyme2026.03 | 90.17 | 1 | — | — | |
| OPERABase Model=LLaVA-NeXT2026.05 | 90.17 | — | 89.74 | — | |
| SpecEyesBase Model=DeepEyes, Aggregation Strategy=mean2026.03 | 90.13 | 2.14 | — | — | |
| SpecEyesBase Model=DeepEyes, Aggregation Strategy=bottom2026.03 | 90.13 | 2.11 | — | — | |
| SpecEyesBase Model=DeepEyes, Aggregation Strategy=min2026.03 | 90.13 | 2.19 | — | — | |
| LogicCheckGPTBase Model=mPLUG-Owl2026.04 | 90 | — | 89.44 | — | |
| SIDBase Model=LLaVA-NeXT2026.05 | 89.96 | — | 90.1 | — | |
| GreedyBase Model=LLaVA-NeXT2026.05 | 89.74 | — | 89.03 | — | |
| OPERAModel=Qwen2.5-VL-7B-Instruct2026.05 | 89.72 | — | 90.02 | — | |
| LUREBase Model=LLaVA-1.52026.04 | 89.67 | — | 89.7 | — | |
| GreedyBase Model=LLaVA-1.52026.05 | 89.62 | — | 89.14 | — | |
| VCDBase Model=LLaVA-1.52026.05 | 89.55 | — | 89.31 | — | |
| PAIModel=Qwen2.5-VL-7B-Instruct2026.05 | 89.54 | — | 88.72 | — | |
| VTIModel=LLaVA-1.5-7B2026.05 | 89.5 | — | 88.89 | — | |
| CRG(ours)Model=Qwen-VL-Chat2026.05 | 89.46 | — | 88.33 | — | |
| OursDecoding Strategy=Greedy, Base Model=LLaVA-1.52025.01 | 89.33 | — | 88.89 | — | |
| OPERAModel=LLaVA-1.5-7B2026.05 | 89.2 | — | 88.81 | — | |
| RC-DPOModel=MM-Eureka2026.05 | 89.17 | — | 87.95 | — | |
| OursDecoding Strategy=Sampling, Base Model=LLaVA-1.52025.01 | 89.1 | — | 88.6 | — | |
| SIDBase Model=LLaVA-1.52026.05 | 89.09 | — | 88.83 | — | |
| R-CoVBase Model=Qwen2.5-VL2026.04 | 89 | — | 88 | — | |
| OursDecoding Strategy=Greedy, Base Model=Qwen-VL2025.01 | 88.92 | — | 88.48 | — | |
| OPERADecoding Strategy=Greedy, Base Model=LLaVA-1.52025.01 | 88.85 | — | 88.67 | — | |
| DeepEyesBase Model=DeepEyes2026.03 | 88.83 | 1 | — | — | |
| PRISM2025.02 | 88.7 | — | — | — | |
| mPlug-Owl + ORCALVLM=mPlug-Owl, Variant=+ ORCA2025.09 | 88.67 | — | 87.5 | — | |
| OPERABase Model=LLaVA-1.52026.05 | 88.64 | — | 88.43 | — | |
| OursDecoding Strategy=Sampling, Base Model=Qwen-VL2025.01 | 88.61 | — | 88.26 | — | |
| VCDBase Model=LLaVA-NeXT2026.05 | 88.6 | — | 87.89 | — | |
| PAIDecoding Strategy=Greedy, Base Model=LLaVA-1.52025.01 | 88.32 | — | 88.01 | — | |
| PAIDecoding Strategy=Sampling, Base Model=Qwen-VL2025.01 | 88.15 | — | 87.94 | — | |
| VCDDecoding Strategy=Greedy, Base Model=LLaVA-1.52025.01 | 88.11 | — | 88.15 | — | |
| PAIDecoding Strategy=Greedy, Base Model=Qwen-VL2025.01 | 87.86 | — | 87.63 | — | |
| VCDDecoding Strategy=Greedy, Base Model=Qwen-VL2025.01 | 87.83 | — | 87.39 | — | |
| VCDModel=Qwen2.5-VL-7B-Instruct2026.05 | 87.82 | — | 88.23 | — | |
| ICDDecoding Strategy=Greedy, Base Model=LLaVA-1.52025.01 | 87.82 | — | 86.73 | — | |
| VCDDecoding Strategy=Sampling, Base Model=Qwen-VL2025.01 | 87.82 | — | 87.68 | — | |
| VCDModel=LLaVA-1.5-7B2026.05 | 87.73 | — | 87.16 | — | |
| VCDModel=MM-Eureka2026.05 | 87.73 | — | 86.15 | — | |
| PAIDecoding Strategy=Sampling, Base Model=LLaVA-1.52025.01 | 87.73 | — | 87.65 | — | |
| PAIBase LVLM=Qwen-VL-Chat2026.04 | 87.69 | — | 86.94 | — | |
| OursDecoding Strategy=Greedy, Base Model=InstructBLIP2025.01 | 87.68 | — | 87.36 | — | |
| PTIBase LVLM=Qwen-VL-Chat2026.04 | 87.53 | — | 86.54 | — | |
| SIDModel=MM-Eureka2026.05 | 87.5 | — | 85.88 | — | |
| ICONS2025.02 | 87.5 | — | — | — | |
| BaselineDecoding Strategy=Greedy, Base Model=InstructBLIP2025.01 | 87.44 | — | 86.8 | — | |
| LogicCheckGPTBase Model=Qwen2.5-VL2026.04 | 87.33 | — | 85.71 | — | |
| OPA-DPOModel=MM-Eureka2026.05 | 87.33 | — | 85.63 | — | |
| Full-Finetune2025.02 | 87.3 | — | — | — | |
| RC-DPOModel=R1-Onevision2026.05 | 87.27 | — | 85.56 | — | |
| ACEBase Model=InstructBLIP2026.05 | 86.96 | — | 86.98 | — | |
| VCDBase LVLM=Qwen-VL-Chat2026.04 | 86.93 | — | 85.46 | — | |
| VCDModel=Qwen-VL-Chat2026.05 | 86.93 | — | 85.46 | — | |
| VCDDecoding Strategy=Sampling, Base Model=LLaVA-1.52025.01 | 86.84 | — | 86.83 | — | |
| PAIBase LVLM=DeepSeek-VL-Chat2026.04 | 86.83 | — | 86.29 | — | |
| ICDDecoding Strategy=Greedy, Base Model=InstructBLIP2025.01 | 86.76 | — | 86.83 | — | |
| VTIModel=Qwen-VL-Chat2026.05 | 86.73 | — | 85.59 | — | |
| VCDBase LVLM=DeepSeek-VL-Chat2026.04 | 86.7 | — | 85.97 | — | |
| VISTABase LVLM=DeepSeek-VL-Chat2026.04 | 86.7 | — | 85.54 | — | |
| VanillaBase Model=Qwen2.5-VL2026.04 | 86.67 | — | 84.62 | — | |
| OursDecoding Strategy=Sampling, Base Model=InstructBLIP2025.01 | 86.67 | — | 86.1 | — | |
| VCDDecoding Strategy=Greedy, Base Model=InstructBLIP2025.01 | 86.66 | — | 86.37 | — | |
| VanillaBase LVLM=Qwen-VL-Chat2026.04 | 86.52 | — | 85.59 | — | |
| PAIDecoding Strategy=Greedy, Base Model=InstructBLIP2025.01 | 86.51 | — | 86.19 | — | |
| ICDDecoding Strategy=Greedy, Base Model=Qwen-VL2025.01 | 86.49 | — | 85.86 | — | |
| Qwen3-VL-2BMode=draft only2026.03 | 86.47 | 4.07 | — | — | |
| ICDDecoding Strategy=Sampling, Base Model=Qwen-VL2025.01 | 86.45 | — | 85.8 | — | |
| VanillaModel=MM-Eureka2026.05 | 86.43 | — | 84.55 | — | |
| SIDBase Model=InstructBLIP2026.05 | 86.34 | — | 86.83 | — | |
| MiniGPT-4 + ORCALVLM=MiniGPT-4, Variant=+ ORCA2025.09 | 86.33 | — | 84.64 | — | |
| R-CoVBase Model=MiniGPT-42026.04 | 86.33 | — | 84.98 | — | |
| PAIModel=LLaVA-1.5-7B2026.05 | 86.33 | — | 84.56 | — | |
| RLAIF-VModel=MM-Eureka2026.05 | 86.3 | — | 84.28 | — | |
| PTIBase LVLM=DeepSeek-VL-Chat2026.04 | 86.22 | — | 86.25 | — | |
| VTIBase LVLM=DeepSeek-VL-Chat2026.04 | 86.15 | — | 86.01 | — | |
| COINCIDE2025.02 | 86.1 | — | — | — | |
| VISTABase LVLM=Qwen-VL-Chat2026.04 | 85.91 | — | 84.66 | — | |
| ICDDecoding Strategy=Sampling, Base Model=InstructBLIP2025.01 | 85.78 | — | 85.73 | — | |
| VTIBase LVLM=Qwen-VL-Chat2026.04 | 85.77 | — | 84.68 | — | |
| PAIBase LVLM=LLAVA-1.52026.04 | 85.73 | — | 86.32 | — | |
| OPERAModel=Qwen-VL-Chat2026.05 | 85.71 | — | 84.64 | — | |
| ICDDecoding Strategy=Sampling, Base Model=LLaVA-1.52025.01 | 85.7 | — | 84.43 | — | |
| EL2N2025.02 | 85.6 | — | — | — | |
| TIVE2025.02 | 85.6 | — | — | — |