List your company

VQA Hallucination Detection on POPE Average (Random, Popular, Adversarial) 2023

89.4AccuracyCRoPS
72.86477.15781.4585.743Jan 2, 2026

Evaluation results

MethodLinks
CRoPSModel Architecture=LLaVA-NeXT2026.0189.489.4
SIDModel Architecture=LLaVA-NeXT2026.0188.688.4
M3IDModel Architecture=LLaVA-NeXT2026.018888.1
VCDModel Architecture=LLaVA-NeXT2026.0187.788.1
OPERAModel Architecture=LLaVA-NeXT2026.0187.488.2
SamplingModel Architecture=LLaVA-NeXT2026.0187.387.6
ClearSightModel Architecture=LLaVA-NeXT2026.0187.387.7
ICDModel Architecture=LLaVA-NeXT2026.0186.687.4
CRoPSModel Architecture=Qwen2-VL2026.0186.185.3
SIDModel Architecture=Qwen2-VL2026.0185.683.9
OPERAModel Architecture=Qwen2-VL2026.0185.584.5
ICDModel Architecture=Qwen2-VL2026.0185.284.3
M3IDModel Architecture=Qwen2-VL2026.0184.983.5
VCDModel Architecture=Qwen2-VL2026.0184.783.2
SamplingModel Architecture=Qwen2-VL2026.0184.282.8
CRoPSModel Architecture=LLaVA-1.52026.0183.984.6
SIDModel Architecture=LLaVA-1.52026.0182.783.3
VCDModel Architecture=LLaVA-1.52026.0182.683.3
M3IDModel Architecture=LLaVA-1.52026.0182.483.4
OPERAModel Architecture=LLaVA-1.52026.0182.382.9
ICDModel Architecture=LLaVA-1.52026.0182.282.9
ClearSightModel Architecture=LLaVA-1.52026.018283
SamplingModel Architecture=LLaVA-1.52026.0181.782.8
ClearSightModel Architecture=Qwen2-VL2026.0173.574.5