Hello,
I'm Niels and work as part of the open-source team at Hugging Face. I discovered your work on Arxiv and was wondering whether you would like to submit it to hf.co/papers to improve its discoverability. If you are one of the authors, you can submit it at https://huggingface.co/papers/submit.
The paper page lets people discuss your paper and find artifacts linked to it. I saw in your GitHub README for CLT-Forge that you are planning to release open-source CLTs (up to 8B parameters) soon. Would you like to host these checkpoints on https://huggingface.co/models when they are ready?
Hosting on Hugging Face will give your work more visibility and enable better discoverability within the interpretability community. We can add metadata tags so that people find the models easier, and link them directly to the paper page (read here) so readers can jump straight from the abstract to the weights.
If you're down, I'm leaving a guide here. For custom interpretability models, you can use the PyTorchModelHubMixin class which adds from_pretrained and push_to_hub to the model, or simply use hf_hub_download for easy access.
You can also build a demo for your visual interface on Spaces, and we can provide you a ZeroGPU grant if needed.
Let me know if you're interested or need any guidance!
Kind regards,
Niels
Hello,
I'm Niels and work as part of the open-source team at Hugging Face. I discovered your work on Arxiv and was wondering whether you would like to submit it to hf.co/papers to improve its discoverability. If you are one of the authors, you can submit it at https://huggingface.co/papers/submit.
The paper page lets people discuss your paper and find artifacts linked to it. I saw in your GitHub README for CLT-Forge that you are planning to release open-source CLTs (up to 8B parameters) soon. Would you like to host these checkpoints on https://huggingface.co/models when they are ready?
Hosting on Hugging Face will give your work more visibility and enable better discoverability within the interpretability community. We can add metadata tags so that people find the models easier, and link them directly to the paper page (read here) so readers can jump straight from the abstract to the weights.
If you're down, I'm leaving a guide here. For custom interpretability models, you can use the PyTorchModelHubMixin class which adds
from_pretrainedandpush_to_hubto the model, or simply use hf_hub_download for easy access.You can also build a demo for your visual interface on Spaces, and we can provide you a ZeroGPU grant if needed.
Let me know if you're interested or need any guidance!
Kind regards,
Niels