ResearchBenchmarksHallucination Evaluation on AMBER AttackVLM (adversarially perturbed)Follow24.4CHAIRmPlug-Owl3.7049.07714.4519.823Sep 18, 2025Evaluation ResultsMethodMethodLinksCHAIRHallucination RateCognitive ScoremPlug-OwlMode=StandaloneMode=Standalone2025.0924.48411.8ORCABase LVLM=mPlug-OwlBase LVLM=mPlug-Owl2025.0920.7544.9MiniGPT-4Mode=StandaloneMode=Standalone2025.0916.47813.4ORCABase LVLM=MiniGPT-4Base LVLM=MiniGPT-42025.0913.5403.7Qwen-VL-ChatMode=StandaloneMode=Standalone2025.097303.3ORCABase LVLM=Qwen-VL-ChatBase LVLM=Qwen-VL-Chat2025.094.5100