Phi: Preference Hijacking in Multi-modal Large Language Models at Inference Time

AAMLRe-assign community