Naturally Supervised 3D Visual Grounding with Language-Regularized Concept Learners | Read Paper on Bytez