🧠 Improve 3D multimodal model reasoning using causal continuous rotary positional encoding for enhanced spatial understanding and performance.
