For years, most AI systems creating 3D content have relied on techniques adapted from large language models or image generation. These approaches typically convert geometric data into sequences before ...