Object Hallucination Assessment on A-OKVQA POPE (Adversarial)
0.8363AccuracySIRA
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SIRABackbone=Qwen2.5-VL2026.05 | 0.8363 | — | 0.8461 | — | |
| ICTBackbone=Qwen2.5-VL2026.05 | 0.836 | — | 0.8302 | — | |
| MaskCDBackbone=Qwen2.5-VL2026.05 | 0.8275 | — | 0.8395 | — | |
| CAIBackbone=Qwen-VL2026.06 | 0.826 | — | 0.832 | — | |
| ONLYBackbone=Qwen-VL2026.06 | 0.822 | — | 0.8274 | — | |
| VCDBackbone=Qwen-VL2026.06 | 0.8213 | — | 0.8292 | — | |
| PAIBackbone=Qwen-VL2026.06 | 0.821 | — | 0.8158 | — | |
| DoLaBackbone=Qwen2.5-VL2026.05 | 0.8183 | — | 0.8133 | — | |
| VCDBackbone=Qwen2.5-VL2026.05 | 0.8127 | — | 0.8238 | — | |
| CoFi-DecModel=Qwen-VL2025.12 | 0.8126 | 0.8097 | 0.8104 | — | |
| RegularBackbone=Qwen-VL2026.06 | 0.8103 | — | 0.8036 | — | |
| SIDBackbone=Qwen2.5-VL2026.05 | 0.8084 | — | 0.8151 | — | |
| OPERABackbone=Qwen2.5-VL2026.05 | 0.8082 | — | 0.8154 | — | |
| DeGFModel=Qwen-VL2025.12 | 0.8075 | 0.8037 | 0.8046 | — | |
| BaseBackbone=Qwen2.5-VL2026.05 | 0.8037 | — | 0.7968 | — | |
| RITUALModel=Qwen-VL2025.12 | 0.802 | 0.7908 | 0.8058 | — | |
| M3IDModel=Qwen-VL2025.12 | 0.795 | 0.7754 | 0.8021 | — | |
| VCDModel=Qwen-VL2025.12 | 0.7913 | 0.7604 | 0.803 | — | |
| CMPCBase Model=LLaVA 1.52025.01 | 0.7913 | 0.7645 | 0.8014 | 0.842 | |
| CAIBackbone=LLaVA-1.52026.06 | 0.774 | — | 0.7979 | — | |
| SIRABackbone=LLaVA-v1.52026.05 | 0.773 | — | 0.7943 | — | |
| ICDBase Model=LLaVA 1.52025.01 | 0.772 | 0.7547 | 0.7795 | 0.806 | |
| RegularModel=Qwen-VL2025.12 | 0.769 | 0.7559 | 0.7748 | — | |
| MaskCDBackbone=LLaVA-v1.52026.05 | 0.759 | — | 0.784 | — | |
| CMPCBase Model=InstructBLIP2025.01 | 0.757 | 0.7138 | 0.7793 | 0.858 | |
| ICTBackbone=LLaVA-v1.52026.05 | 0.7556 | — | 0.7868 | — | |
| SIDBackbone=LLaVA-v1.52026.05 | 0.755 | — | 0.7794 | — | |
| CAIBackbone=InstructBLIP2026.06 | 0.7523 | — | 0.7804 | — | |
| VCDBackbone=LLaVA-v1.52026.05 | 0.7497 | — | 0.7773 | — | |
| VCDBase Model=LLaVA 1.52025.01 | 0.7497 | 0.7001 | 0.7773 | 0.8736 | |
| ICDBase Model=InstructBLIP2025.01 | 0.7442 | 0.7024 | 0.7848 | 0.8893 | |
| VCDBase Model=InstructBLIP2025.01 | 0.7433 | 0.6946 | 0.7719 | 0.8687 | |
| BaseBackbone=LLaVA-v1.52026.05 | 0.7404 | — | 0.7515 | — | |
| BaselineBase Model=LLaVA 1.52025.01 | 0.7404 | 0.7208 | 0.7515 | 0.7849 | |
| OPERABackbone=LLaVA-v1.52026.05 | 0.7382 | — | 0.7791 | — | |
| CoFi-DecModel=InstructBLIP2025.12 | 0.7344 | 0.6697 | 0.7716 | — | |
| VCDBackbone=InstructBLIP2026.06 | 0.7333 | — | 0.77 | — | |
| DeGFModel=InstructBLIP2025.12 | 0.7187 | 0.6565 | 0.7596 | — | |
| CoFi-DecModel=LLaVA-1.52025.12 | 0.713 | 0.681 | 0.7826 | — | |
| VCDModel=InstructBLIP2025.12 | 0.71 | 0.6541 | 0.7545 | — | |
| BaselineBase Model=InstructBLIP2025.01 | 0.7071 | 0.6591 | 0.7556 | 0.8583 | |
| DeGFModel=LLaVA-1.52025.12 | 0.707 | 0.667 | 0.7686 | — | |
| ONLYBackbone=LLaVA-1.52026.06 | 0.7047 | — | 0.7586 | — | |
| RITUALModel=InstructBLIP2025.12 | 0.7027 | 0.6415 | 0.7555 | — | |
| M3IDModel=InstructBLIP2025.12 | 0.701 | 0.6428 | 0.7516 | — | |
| RegularModel=InstructBLIP2025.12 | 0.6987 | 0.6454 | 0.7454 | — | |
| ONLYBackbone=InstructBLIP2026.06 | 0.6877 | — | 0.7495 | — | |
| PAIBackbone=InstructBLIP2026.06 | 0.6867 | — | 0.7455 | — | |
| M3IDModel=LLaVA-1.52025.12 | 0.686 | 0.6222 | 0.7511 | — | |
| RITUALModel=LLaVA-1.52025.12 | 0.6857 | 0.6226 | 0.7499 | — | |
| RegularBackbone=InstructBLIP2026.06 | 0.6833 | — | 0.7389 | — | |
| VCDBackbone=LLaVA-1.52026.06 | 0.683 | — | 0.7486 | — | |
| DoLaBackbone=LLaVA-v1.52026.05 | 0.6803 | — | 0.7449 | — | |
| PAIBackbone=LLaVA-1.52026.06 | 0.676 | — | 0.7423 | — | |
| VCDModel=LLaVA-1.52025.12 | 0.674 | 0.6139 | 0.7421 | — | |
| RegularModel=LLaVA-1.52025.12 | 0.6723 | 0.6156 | 0.737 | — | |
| RegularBackbone=LLaVA-1.52026.06 | 0.6707 | — | 0.737 | — |