Release SMTL models and datasets on Hugging Face

[NielsRogge]

Hi @qianben 🤗

I'm Niels and work as part of the open-source team at Hugging Face. I discovered your work through Hugging Face's daily papers as yours got featured: https://huggingface.co/papers/2602.22675.

The paper page lets people discuss your paper and find related artifacts (like your models and datasets). I noticed on the first page of your PDF that you plan to open-source the SMTL code, models, and datasets, even featuring the Hugging Face logo next to them!

I see the GitHub link is currently leading to a 404, so I assume you're still preparing the release. Would you like to host the agent checkpoints (like SMTL-300) and the synthesized search datasets on the Hugging Face Hub?

Hosting on Hugging Face will give your work better visibility and discoverability. We can add metadata tags so people can find the models easily, and the Dataset Viewer will allow people to explore your synthesized data directly in the browser.

Uploading models

You can host the checkpoints at https://huggingface.co/models. If these are custom PyTorch models, you can use the PyTorchModelHubMixin class to add from_pretrained and push_to_hub support.

Uploading datasets

Making your datasets available at https://huggingface.co/datasets would allow users to load them instantly:

from datasets import load_dataset
dataset = load_dataset("your-hf-org-or-username/your-dataset")

Once uploaded, we can link them to the paper page so researchers can easily find and use your artifacts.

Let me know if you're interested or if you need any guidance!

Kind regards,

Niels
ML Engineer @ Hugging Face 🤗