Object Hallucination Evaluation on CHAIR
57CHAIRi ScoreGreedy
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| GreedyArchitecture=InstructBLIP2026.03 | 57 | — | — | 70.4 | — | — | — | 16.7 | — | |
| Beam SearchArchitecture=InstructBLIP2026.03 | 56.2 | — | — | 71.4 | — | — | — | 14.7 | — | |
| Beam SearchArchitecture=LLaVA-1.52026.03 | 55.6 | — | — | 76.7 | — | — | — | 16 | — | |
| IMCCDArchitecture=InstructBLIP2026.03 | 55.3 | — | — | 73 | — | — | — | 17 | — | |
| GreedyArchitecture=LLaVA-1.52026.03 | 53 | — | — | 76.7 | — | — | — | 15.2 | — | |
| IMCCDArchitecture=LLaVA-1.52026.03 | 52.2 | — | — | 76.6 | — | — | — | 13.5 | — | |
| CLVAArchitecture=InstructBLIP2026.03 | 50.2 | — | — | 74 | — | — | — | 14 | — | |
| ClearSightArchitecture=LLaVA-1.52026.03 | 50 | — | — | 78.8 | — | — | — | 13.6 | — | |
| MemVRArchitecture=LLaVA-1.52026.03 | 49.2 | — | — | 78.4 | — | — | — | 12.9 | — | |
| CLVAArchitecture=LLaVA-1.52026.03 | 48 | — | — | 80.2 | — | — | — | 12.7 | — | |
| VanillaModel=mPLUG-Owl2026.04 | 32.1 | — | — | — | — | 60.1 | — | 88 | — | |
| LogicCheckGPTModel=mPLUG-Owl2026.04 | 25.9 | — | — | — | — | 61.4 | — | 70 | — | |
| BaseBase Model=InstructBLIP2026.04 | 24.2 | — | — | — | — | — | — | 55.6 | — | |
| VanillaBackbone=LLaVA v1.5 7B2026.05 | 23.5 | — | — | — | — | 65 | — | 47.4 | 70.8 | |
| HalTrapperBackbone=LLaVA v1.5 7B2026.05 | 23.5 | — | — | — | — | 64.9 | — | 47.2 | 70.6 | |
| DOP-OBCBase Model=InstructBLIP2026.04 | 22.5 | — | — | — | — | — | — | 50.2 | — | |
| R-CoVModel=mPLUG-Owl2026.04 | 21.7 | — | — | — | — | 64.3 | — | 64 | — | |
| OPERABackbone=LLaVA-v1.5-7B2026.04 | 19.8 | — | — | — | — | 70.7 | — | 54 | — | |
| TinyLLaVA-1.5B2026.03 | 17.5 | — | — | — | — | — | — | 58.6 | — | |
| Orig.Model=Shikra-7b, Decoding Strategy=VCD2026.04 | 16.8 | — | — | — | — | 70.9 | — | 54.6 | — | |
| BaseBase Model=Chat-UniVi2026.04 | 16.7 | — | — | — | — | — | — | 52.3 | — | |
| ICDBackbone=LLaVA-1.5, Max output token length=1024, Len=103.42026.05 | 16.3 | — | — | 16.3 | 36.3 | — | — | 56.2 | — | |
| AUEModel=Shikra-7b, Decoding Strategy=VCD2026.04 | 16.2 | — | — | — | — | 70.7 | — | 48.8 | — | |
| VanillaModel=LLaVA-1.52026.04 | 16.2 | — | — | — | — | 73.2 | — | 52 | — | |
| VanillaBackbone=LLaVA-v1.5-7B2026.04 | 16.2 | — | — | — | — | 73.2 | — | 52 | — | |
| Orig.Model=Shikra-7b, Decoding Strategy=Greedy2026.04 | 16.1 | — | — | — | — | 71.5 | — | 52.4 | — | |
| OursModel=Shikra-7b, Decoding Strategy=VCD2026.04 | 15.8 | — | — | — | — | 68.2 | — | 48.2 | — | |
| VCDBackbone=LLaVA v1.5 7B2026.05 | 15.8 | — | — | — | — | 76.5 | — | 52.8 | 72.6 | |
| BaseBase Model=Video-LLaVA2026.04 | 15.6 | — | — | — | — | — | — | 50.2 | — | |
| Video-LLaVA (RoPE)Base Model=Video-LLaVA2026.04 | 15.6 | — | — | — | — | — | — | 50.2 | — | |
| LLaVA-1.5Backbone=LLaVA-1.5, Max output token length=1024, Len=100.62026.05 | 15.4 | — | — | 77.1 | 32.7 | — | — | 50 | — | |
| Orig.Model=LLaVA-1.5-7b, Decoding Strategy=VCD2026.04 | 14.9 | — | — | — | — | 71.3 | — | 46.2 | — | |
| FixVPEBase Model=Video-LLaVA2026.04 | 14.9 | — | — | — | — | — | — | 48.5 | — | |
| VCDBackbone=LLaVA-1.5, Max output token length=1024, Len=100.42026.05 | 14.9 | — | — | 77.3 | 31.8 | — | — | 48.6 | — | |
| ICDBackbone=LLaVA v1.5 7B2026.05 | 14.7 | — | — | — | — | 77.4 | — | 51.8 | 73.7 | |
| OPERABackbone=LLaVA-1.5, Max output token length=1024, Len=98.62026.05 | 14.6 | — | — | 76.8 | 31.2 | — | — | 47.8 | — | |
| AUEModel=Shikra-7b, Decoding Strategy=Greedy2026.04 | 14.5 | — | — | — | — | 72.2 | — | 47.2 | — | |
| EDVTBase Model=LLaVA-1.52026.04 | 14.5 | — | — | — | — | — | — | 46.8 | — | |
| VCDBackbone=LLaVA-v1.5-7B2026.04 | 14.4 | — | — | — | — | 74.8 | — | 46 | — | |
| CAIGLA-gatedMax Token=1282026.06 | 14.4 | — | — | — | — | — | — | 35 | — | |
| DOP-OBCBase Model=Chat-UniVi2026.04 | 14.3 | — | — | — | — | — | — | 46.2 | — | |
| ICDBackbone=Qwen-VL-Chat, Max output token length=1024, Len=48.12026.05 | 14.3 | — | — | 37.6 | 16.3 | — | — | 18.4 | — | |
| VARBase Model=LLaVA-1.5-7B2026.05 | 14.17 | — | — | — | — | — | 88.72 | 52.5 | — | |
| OursModel=Shikra-7b, Decoding Strategy=Greedy2026.04 | 14.1 | — | — | — | — | 70.7 | — | 44.6 | — | |
| BaseBase Model=LLaVA-1.52026.04 | 13.9 | — | — | — | — | — | — | 48 | — | |
| LLaVA-1.5 (RoPE)Base Model=LLaVA-1.52026.04 | 13.9 | — | — | — | — | — | — | 48 | — | |
| LLaVA-1.5-7BBackbone=LLaVA-1.5-7B2026.04 | 13.9 | — | — | — | — | — | — | 53 | — | |
| VCDBackbone=LLaVA-1.5-7B2026.04 | 13.9 | — | — | — | — | — | — | 53 | — | |
| MemVRBackbone=LLaVA-1.5, Max output token length=1024, Len=101.52026.05 | 13.8 | — | — | 79.6 | 30.7 | — | — | 47.6 | — | |
| CODEBackbone=LLaVA v1.5 7B2026.05 | 13.8 | — | — | — | — | 75.7 | — | 49.8 | 76 | |
| EDVTBase Model=Video-LLaVA2026.04 | 13.7 | — | — | — | — | — | — | 46.8 | — | |
| BeamBase Model=LLaVA-1.5-7B2026.05 | 13.7 | — | — | — | — | — | 93.28 | 52.1 | — | |
| ICDBase Model=LLaVA-1.52026.04 | 13.6 | — | — | — | — | — | — | 47.7 | — | |
| GreedyBase Model=LLaVA-1.5-7B2026.05 | 13.52 | — | — | — | — | — | 82.66 | 48.6 | — | |
| Orig.Model=LLaVA-1.5-7b, Decoding Strategy=OPERA2026.04 | 13.5 | — | — | — | — | 74.4 | — | 44.2 | — | |
| Orig.Model=LLaVA-1.5-13b, Decoding Strategy=VCD2026.04 | 13.5 | — | — | — | — | 73.1 | — | 48.8 | — | |
| VCDBase Model=LLaVA-1.5-7B2026.05 | 13.47 | — | — | — | — | — | 85.93 | 48.4 | — | |
| SPARCBase Model=LLaVA-1.5-7B2026.05 | 13.42 | — | — | — | — | — | 89.44 | 48.9 | — | |
| FixVPEBase Model=LLaVA-1.52026.04 | 13.4 | — | — | — | — | — | — | 47.3 | — | |
| CODEBase Model=LLaVA-1.5-7B2026.05 | 13.34 | — | — | — | — | — | 78.34 | 48.2 | — | |
| Orig.Model=LLaVA-1.5-7b, Decoding Strategy=Greedy2026.04 | 13.3 | — | — | — | — | 74.2 | — | 45 | — | |
| VAFBase Model=LLaVA-1.5-7B2026.05 | 13.28 | — | — | — | — | — | 89.38 | 49.8 | — | |
| AGLABase Model=LLaVA-1.5-7B2026.05 | 13.27 | — | — | — | — | — | 89.74 | 46.4 | — | |
| ONLYBase Model=LLaVA-1.5-7B2026.05 | 13.23 | — | — | — | — | — | 95.85 | 47.8 | — | |
| VCDBase Model=LLaVA-1.52026.04 | 13.2 | — | — | — | — | — | — | 46.8 | — | |
| FarSightBase Model=LLaVA-1.52026.04 | 13.2 | — | — | — | — | — | — | 41.6 | — | |
| FarSightBase Model=LLaVA-1.52026.04 | 13.2 | — | — | — | — | — | — | 41.6 | — | |
| VanillaModel=MiniGPT-42026.04 | 13 | — | — | — | — | 66 | — | 46 | — | |
| FarSightBase Model=Video-LLaVA2026.04 | 12.9 | — | — | — | — | — | — | 44.8 | — | |
| LogicCheckGPTModel=MiniGPT-42026.04 | 12.9 | — | — | — | — | 66.4 | — | 44 | — | |
| Orig.Model=Shikra-7b, Decoding Strategy=OPERA2026.04 | 12.8 | — | — | — | — | 70 | — | 35.2 | — | |
| OPERABase Model=LLaVA-1.52026.04 | 12.7 | — | — | — | — | — | — | 45.2 | — | |
| OursModel=Shikra-7b, Decoding Strategy=OPERA2026.04 | 12.6 | — | — | — | — | 66.5 | — | 31.6 | — | |
| DeCoBackbone=LLaVA-1.5, Max output token length=1024, Len=103.42026.05 | 12.6 | — | — | 76.7 | 27.7 | — | — | 42.8 | — | |
| Qwen-3.5Max Token=1282026.06 | 12.6 | — | — | — | — | — | — | 36.2 | — | |
| OursModel=LLaVA-1.5-7b, Decoding Strategy=VCD2026.04 | 12.5 | — | — | — | — | 70.4 | — | 35.6 | — | |
| LLaVA-1.5-13B2026.03 | 12.3 | — | — | — | — | — | — | 43.8 | — | |
| VCDBackbone=Qwen-VL-Chat, Max output token length=1024, Len=115.42026.05 | 12.3 | — | — | 47.9 | 12.6 | — | — | 13 | — | |
| PAIMax Token=1282026.06 | 12.3 | — | — | — | — | — | — | 37.2 | — | |
| AUEModel=LLaVA-1.5-7b, Decoding Strategy=VCD2026.04 | 11.8 | — | — | — | — | 70.5 | — | 34.2 | — | |
| Orig.Model=LLaVA-1.5-13b, Decoding Strategy=Greedy2026.04 | 11.8 | — | — | — | — | 75.3 | — | 41 | — | |
| LaSCDBackbone=LLaVA-1.5, Max output token length=1024, Len=100.02026.05 | 11.8 | — | — | 77.6 | 26.2 | — | — | 40.6 | — | |
| HIREBase Model=TinyLLaVA-1.5B2026.03 | 11.7 | — | — | — | — | — | — | 36.6 | — | |
| OursModel=LLaVA-1.5-13b, Decoding Strategy=VCD2026.04 | 11.7 | — | — | — | — | 70.7 | — | 35.4 | — | |
| VCDBackbone=Qwen3-VL 8B2026.05 | 11.7 | — | — | — | — | 68.8 | — | 38.4 | 79.5 | |
| Orig.Model=LLaVA-1.5-13b, Decoding Strategy=OPERA2026.04 | 11.5 | — | — | — | — | 75 | — | 37.8 | — | |
| LogicCheckGPTModel=LLaVA-1.52026.04 | 11.5 | — | — | — | — | 74.8 | — | 38 | — | |
| CAIMax Token=1282026.06 | 11.4 | — | — | — | — | — | — | 29.8 | — | |
| AUEModel=LLaVA-1.5-13b, Decoding Strategy=VCD2026.04 | 11.3 | — | — | — | — | 71 | — | 32.2 | — | |
| AUEModel=LLaVA-1.5-7b, Decoding Strategy=OPERA2026.04 | 11.2 | — | — | — | — | 72.3 | — | 29.8 | — | |
| OursModel=LLaVA-1.5-7b, Decoding Strategy=OPERA2026.04 | 11.2 | — | — | — | — | 71.4 | — | 31.6 | — | |
| DOP-OBCBase Model=LLaVA-1.52026.04 | 11.2 | — | — | — | — | — | — | 39.7 | — | |
| DOP–OBCBase Model=LLaVA-1.52026.04 | 11.2 | — | — | — | — | — | — | 39.7 | — | |
| AUEModel=LLaVA-1.5-13b, Decoding Strategy=OPERA2026.04 | 10.7 | — | — | — | — | 71.5 | — | 27.8 | — | |
| Orig.Model=LLaVA-1.5-13b, Decoding Strategy=PAI2026.04 | 10.7 | — | — | — | — | 75.2 | — | 33.8 | — | |
| AUEModel=Shikra-7b, Decoding Strategy=OPERA2026.04 | 10.7 | — | — | — | — | 68.8 | — | 28.8 | — | |
| Orig.Model=Shikra-7b, Decoding Strategy=Devils2026.04 | 10.7 | — | — | — | — | 70 | — | 27.8 | — | |
| AUEModel=Shikra-7b, Decoding Strategy=Devils2026.04 | 10.5 | — | — | — | — | 66.7 | — | 20.8 | — | |
| AUEModel=LLaVA-1.5-7b, Decoding Strategy=Greedy2026.04 | 10.4 | — | — | — | — | 71.8 | — | 30.2 | — | |
| OursModel=Shikra-7b, Decoding Strategy=Devils2026.04 | 10.3 | — | — | — | — | 67 | — | 21.8 | — |