Object Hallucination Probing on A-OKVQA (Adversarial split)
79.1AccuracySchroMind
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SchroMindModel=LLaVA-1.5-7B2026.02 | 79.1 | 80.15 | |
| HDDModel=LLaVA-NeXT2025.12 | 79.1 | 79.5 | |
| VCDModel=LLaVA-NeXT2025.12 | 78.7 | 78.9 | |
| HALCModel=LLaVA-NeXT2025.12 | 78.7 | 79.2 | |
| BeamModel=LLaVA-NeXT2025.12 | 78.6 | 78.3 | |
| ScalpelBase Model=LLaVA-1.5-7B2026.02 | 78.4 | 79.71 | |
| OPERAModel=LLaVA-NeXT2025.12 | 78.4 | 79.1 | |
| HDDModel=LLaVA-1.52025.12 | 78 | 79.9 | |
| HALCModel=LLaVA-1.52025.12 | 77.5 | 79.6 | |
| VCDModel=LLaVA-1.52025.12 | 75.8 | 78.6 | |
| HDDModel=InstructBLIP2025.12 | 75.7 | 79 | |
| ICTBase Model=LLaVA-1.5-7B2026.02 | 75.56 | 78.68 | |
| ICTModel=LLaVA-1.5-7B2026.02 | 75.56 | 78.68 | |
| VCDBase Model=LLaVA-1.5-7B2026.02 | 74.97 | 77.73 | |
| VCDModel=LLaVA-1.5-7B2026.02 | 74.97 | 77.73 | |
| VCDModel=InstructBLIP2025.12 | 74.9 | 77.6 | |
| HALCModel=InstructBLIP2025.12 | 74.9 | 77.2 | |
| BeamModel=InstructBLIP2025.12 | 74.7 | 77.9 | |
| VanillaBase Model=LLaVA-1.5-7B2026.02 | 74.04 | 75.15 | |
| RegularModel=LLaVA-1.5-7B2026.02 | 74.04 | 75.15 | |
| OPERABase Model=LLaVA-1.5-7B2026.02 | 73.82 | 77.91 | |
| OPERAModel=LLaVA-1.5-7B2026.02 | 73.82 | 77.91 | |
| OPERAModel=InstructBLIP2025.12 | 71.5 | 76.4 | |
| OPERAModel=LLaVA-1.52025.12 | 69.1 | 75.4 | |
| AVISCModel=LLaVA-1.5-7B2026.02 | 68.97 | 75.11 | |
| M3IDModel=LLaVA-1.5-7B2026.02 | 68.1 | 74.58 | |
| BeamModel=LLaVA-1.52025.12 | 68.1 | 74.9 |