Request for pretrained object-interaction / loco-manipulation policies

[yiweizuo]

Hi Holosoma team,

Thank you for releasing this impressive framework. Also, congratulations on OmniRetarget winning the ICRA 2026 Best Conference Paper Award and the Best Paper Award on Robot Manipulation and Locomotion！

I am currently exploring humanoid whole-body tracking and human-object interaction policy learning, and Holosoma is very relevant to my research.I noticed that the retargeting pipeline supports object-interaction data, such as OMOMO-style human-object motion retargeting. However, I could not find any released pretrained policy for object-interaction or loco-manipulation tasks.

I would like to ask:

1. Do you plan to release any pretrained ONNX policies for object-interaction / loco-manipulation tasks?
2. If ONNX policies cannot be released, would it be possible to release .pt checkpoints, W&B artifact links, or training configs for reproducing these policies?
3. For training object-interaction policies from the retargeted data, is there a recommended demo task, config, or baseline setting?

Thank you very much for your work and for any guidance you can provide.