GroupContrast: Semantic-aware Self-supervised Representation Learning for 3D Understanding | Read Paper on Bytez