CoDA: Collaborative Novel Box Discovery and Cross-modal Alignment for Open-vocabulary 3D Object Detection Yang Cao 1 Hang Xu2 Dan Xu1