Hi @JasperHaozhe 馃
I'm Niels and work as part of the open-source team at Hugging Face. I discovered your amazing work through Hugging Face Daily Papers: https://huggingface.co/papers/2607.05382
The paper page lets people discuss about your paper and lets them find artifacts about it, and you can also claim the paper as yours which will show up on your public profile at HF, and add Github and project page URLs.
First of all, congratulations on the paper and on hosting the SearchGen-20K, SearchGen-Bench, and SearchGen-Corpus-1M datasets directly on the Hub! That is fantastic for reproducibility and makes it incredibly straightforward to work with.
Would you also be open to hosting the pre-trained co-trained models you developed (such as the Gen-Adaptive RFT-8B reasoner and the Klein-4B-DPO-v2 generator) on https://huggingface.co/models?
Hosting the models on Hugging Face will give them much more visibility and enable better discoverability. We can add tags to the model cards so that people can filter and find the models easily, and we can link them directly to your paper page so that all of your artifacts (datasets and models) are discoverable in one place.
If you're down, leaving a guide here. If they are custom PyTorch models, you can use the PyTorchModelHubMixin class, which adds from_pretrained and push_to_hub to the model so people can download and run them right away. Alternatively, you can use hf_hub_download to fetch weights directly from your HF repositories.
Let me know if you are interested or need any guidance!
Kind regards,
Niels
Hi @JasperHaozhe 馃
I'm Niels and work as part of the open-source team at Hugging Face. I discovered your amazing work through Hugging Face Daily Papers: https://huggingface.co/papers/2607.05382
The paper page lets people discuss about your paper and lets them find artifacts about it, and you can also claim the paper as yours which will show up on your public profile at HF, and add Github and project page URLs.
First of all, congratulations on the paper and on hosting the SearchGen-20K, SearchGen-Bench, and SearchGen-Corpus-1M datasets directly on the Hub! That is fantastic for reproducibility and makes it incredibly straightforward to work with.
Would you also be open to hosting the pre-trained co-trained models you developed (such as the Gen-Adaptive RFT-8B reasoner and the Klein-4B-DPO-v2 generator) on https://huggingface.co/models?
Hosting the models on Hugging Face will give them much more visibility and enable better discoverability. We can add tags to the model cards so that people can filter and find the models easily, and we can link them directly to your paper page so that all of your artifacts (datasets and models) are discoverable in one place.
If you're down, leaving a guide here. If they are custom PyTorch models, you can use the PyTorchModelHubMixin class, which adds
from_pretrainedandpush_to_hubto the model so people can download and run them right away. Alternatively, you can use hf_hub_download to fetch weights directly from your HF repositories.Let me know if you are interested or need any guidance!
Kind regards,
Niels