First, congratulations on the excellent work! The sim-to-real results demonstrated of the paper are particularly impressive, and I'm keen to reproduce these experiments using your real-world benchmark.

To ensure accurate implementation, may I ask for clarification on two key aspects:

1. Details of the real-world benchmark dataset
   1.1 What specific data were used for the real-world tests (e.g., what was the test clothing)?
   1.2 Is the code for real-world deployment and evaluation publicly available?

2. Sim-to-Real implementation
   The paper mentions the SimToReal approach, but I cannot find the corresponding code in the repositories. Are there key components that have not yet been released (e.g. Noisy Observation, Point Cloud Alignment, Keypoint Embedding alignment)?

If these components are not yet released, could you consider sharing a minimal example (e.g. real_world_interface.py stub or sample dataset) as a temporary reference? This would greatly help the community validate and learn from your approach.

Thank you for your valuable time and openness!