List your company

Object Hallucination Probing on A-OKVQA (Adversarial split)

79.1AccuracySchroMind
67.6670.6373.676.57Dec 22, 2025Dec 30, 2025Jan 7, 2026Jan 16, 2026Jan 24, 2026Feb 1, 2026Feb 10, 2026

Evaluation results

MethodLinks
SchroMindModel=LLaVA-1.5-7B2026.0279.180.15
HDDModel=LLaVA-NeXT2025.1279.179.5
VCDModel=LLaVA-NeXT2025.1278.778.9
HALCModel=LLaVA-NeXT2025.1278.779.2
BeamModel=LLaVA-NeXT2025.1278.678.3
ScalpelBase Model=LLaVA-1.5-7B2026.0278.479.71
OPERAModel=LLaVA-NeXT2025.1278.479.1
HDDModel=LLaVA-1.52025.127879.9
HALCModel=LLaVA-1.52025.1277.579.6
VCDModel=LLaVA-1.52025.1275.878.6
HDDModel=InstructBLIP2025.1275.779
ICTBase Model=LLaVA-1.5-7B2026.0275.5678.68
ICTModel=LLaVA-1.5-7B2026.0275.5678.68
VCDBase Model=LLaVA-1.5-7B2026.0274.9777.73
VCDModel=LLaVA-1.5-7B2026.0274.9777.73
VCDModel=InstructBLIP2025.1274.977.6
HALCModel=InstructBLIP2025.1274.977.2
BeamModel=InstructBLIP2025.1274.777.9
VanillaBase Model=LLaVA-1.5-7B2026.0274.0475.15
RegularModel=LLaVA-1.5-7B2026.0274.0475.15
OPERABase Model=LLaVA-1.5-7B2026.0273.8277.91
OPERAModel=LLaVA-1.5-7B2026.0273.8277.91
OPERAModel=InstructBLIP2025.1271.576.4
OPERAModel=LLaVA-1.52025.1269.175.4
AVISCModel=LLaVA-1.5-7B2026.0268.9775.11
M3IDModel=LLaVA-1.5-7B2026.0268.174.58
BeamModel=LLaVA-1.52025.1268.174.9