List your company

Visual Hallucination Evaluation on POPE MS-COCO Adversarial sampling (val)

85.48AccuracyRVCD
48.268857.929467.5977.2506Sep 20, 2023Dec 31, 2023Apr 11, 2024Jul 23, 2024Nov 2, 2024Feb 12, 2025May 26, 2025

Evaluation results

MethodLinks
RVCDModel=mPLUG-Owl2, Setting=Adversarial2025.0585.4884.4686.9785.7—
RVCDModel=LLaVA-1.5, Setting=Adversarial2025.0585.3684.1787.0985.61—
ICT w/o objectModel=Qwen-VL2024.1185.23——84.62—
ICT w/o imageModel=Qwen-VL2024.1185.13——84.35—
ICTModel=Qwen-VL2024.1184.96——84.42—
OPERAModel=Qwen-VL2024.1184.78——83.45—
ICTModel=LLaVA-v1.52024.1184.43——83.74—
VCDModel=Qwen-VL2024.1184.26——83.9—
ICT w/o imageModel=LLaVA-v1.52024.1184.13——83.31—
ICT w/o objectModel=LLaVA-v1.52024.1183.63——82.13—
RVCDModel=MiniGPT-4, Setting=Adversarial2025.0583.0682.9283.3783.13—
RegularModel=Qwen-VL2024.1182.26——80.37—
OPERAModel=LLaVA-v1.52024.1181.24——81.38—
VCDModel=LLaVA-v1.52024.1180.88——81.33—
RegularModel=LLaVA-v1.52024.1178.96——77.57—
Beam SearchModel=mPLUG-Owl2, Setting=Adversarial2025.0574.2867.5693.4278.41—
Beam SearchModel=LLaVA-1.5, Setting=Adversarial2025.0573.2366.593.6477.77—
OPERAModel=mPLUG-Owl2, Setting=Adversarial2025.0572.7565.994.3477.59—
DREAMLLM-7Bzero-shot=true, backbone=7B2023.0972.6367.0788.9376.4766.3
InstructBLIP-14Bzero-shot=true, backbone=14B2023.0972.165.1395.1377.3273.03
OPERAModel=LLaVA-1.5, Setting=Adversarial2025.0569.9863.2695.3676.06—
VCDModel=mPLUG-Owl2, Setting=Adversarial2025.0569.6563.4692.6675.33—
DoLaModel=MiniGPT-4, Setting=Adversarial2025.056969.7667.168.39—
HALCModel=MiniGPT-4, Setting=Adversarial2025.056969.7667.168.39—
DoLaModel=mPLUG-Owl2, Setting=Adversarial2025.0568.4261.7896.7275.39—
HALCModel=mPLUG-Owl2, Setting=Adversarial2025.0568.461.7796.7275.38—
GreedyModel=mPLUG-Owl2, Setting=Adversarial2025.0568.261.696.875.28—
VCDModel=LLaVA-1.5, Setting=Adversarial2025.0566.3160.4294.5673.73—
DoLaModel=LLaVA-1.5, Setting=Adversarial2025.0566.260.0596.8174.12—
HALCModel=LLaVA-1.5, Setting=Adversarial2025.0566.260.0596.8174.12—
GreedyModel=LLaVA-1.5, Setting=Adversarial2025.0565.9259.8496.8273.97—
MiniGPT-4-14Bzero-shot=true, backbone=14B2023.0965.1761.1982.9370.4267.77
OPERAModel=MiniGPT-4, Setting=Adversarial2025.0561.9557.5791.0270.53—
Beam SearchModel=MiniGPT-4, Setting=Adversarial2025.0561.757.4590.470.25—
VCDModel=MiniGPT-4, Setting=Adversarial2025.0559.3256.7378.6165.9—
GreedyModel=MiniGPT-4, Setting=Adversarial2025.0556.7253.8294.6868.63—
mPLUG-Owl-7Bzero-shot=true, backbone=7B2023.0950.6750.3499.3366.8298.67
MMGPT-7Bzero-shot=true, backbone=7B2023.09505010066.67100
LLaVA-13Bzero-shot=true, backbone=13B2023.0949.749.8599.0766.3299.37