Article 164 autotrust/JEV-27B-VL: a decision model that learned to see without a single image of training
