ULIP-2: Towards Scalable Multimodal Pre-Training for 3D Understanding | IEEE Conference Publication | IEEE Xplore