1 paper
Nikita Sharma, Pranav Sara, Karan Singla
Frontier multimodal models can guess whether a person is lying from a testimony video. To do so, they stream that raw face and voice to a third-party model. We ask whether the heav…