List your company
ResearchFormerly WizWand

Object Hallucination Evaluation on POPE Concurrence

92.33AccuracyORCA
48.306859.735971.16582.5941Sep 18, 2025
Updated 6mo ago

Evaluation results

MethodLinks
ORCABase LVLM=Qwen-VL-Chat (13B), Agent Backbone=LLaMA-3.2-Vision-Instruct (11B), Visual Backbone=OpenCLIP2025.0992.3391.93
ORCABase LVLM=MiniGPT-4 (13B), Agent Backbone=LLaMA-3.2-Vision-Instruct (11B), Visual Backbone=EVA2025.099291.84
ORCABase LVLM=mPlug-Owl (7B), Agent Backbone=LLaMA-3.2-Vision-Instruct (11B), Visual Backbone=CLIP2025.0990.3390.24
mPlug-Owl + ORCALVLM=mPlug-Owl, Variant=+ ORCA2025.098685.21
Qwen-VL-ChatParameters=13B, Backbone=OpenCLIP, Setup=Standalone2025.0985.6785.71
MiniGPT-4 + ORCALVLM=MiniGPT-4, Variant=+ ORCA2025.0984.6783.57
Qwen-VL-Chat + ORCALVLM=Qwen-VL-Chat, Variant=+ ORCA2025.0983.6780.93
Qwen-VL-Chat (Standalone)LVLM=Qwen-VL-Chat, Variant=Standalone2025.0982.3379.21
mPlug-Owl + ORCALVLM=mPlug-Owl, Method=+ ORCA, Attack Scenario=Adversarial Illusions2025.0981.3380
MiniGPT-4 + ORCALVLM=MiniGPT-4, Method=+ ORCA, Attack Scenario=Adversarial Illusions2025.0980.6779.58
Qwen-VL-Chat (Standalone)LVLM=Qwen-VL-Chat, Method=Standalone, Attack Scenario=Adversarial Illusions2025.0979.6776.62
Qwen-VL-Chat + ORCALVLM=Qwen-VL-Chat, Method=+ ORCA, Attack Scenario=Adversarial Illusions2025.097772.06
MiniGPT-4Parameters=13B, Backbone=EVA, Setup=Standalone2025.097074.29
MiniGPT-4 (Standalone)LVLM=MiniGPT-4, Variant=Standalone2025.0964.3373.58
MiniGPT-4 (Standalone)LVLM=MiniGPT-4, Method=Standalone, Attack Scenario=Adversarial Illusions2025.0961.6771.18
mPlug-OwlParameters=7B, Backbone=CLIP, Setup=Standalone2025.095167.11
mPlug-Owl (Standalone)LVLM=mPlug-Owl, Variant=Standalone2025.095066.67
mPlug-Owl (Standalone)LVLM=mPlug-Owl, Method=Standalone, Attack Scenario=Adversarial Illusions2025.095066.67