Abstract: Text-to-3D generation enables the creation of 3D content with infinite possibilities. Existing methods typically involve training 3D generative models, which suffer from poor semantic ...