Request for VulnLLM-R Model Weights on Hugging Face

[Explorer1092]

Dear VulnLLM-R Team,

First of all, thank you for open-sourcing the code and methodology for VulnLLM-R. This is impressive work, especially regarding the specialized reasoning capabilities for vulnerability detection.

I noticed that while the training recipes and agent scaffold are available in this repository, the trained model weights seem to be missing. Since training a 7B parameter model from scratch requires significant computational resources, this makes it difficult for many researchers and developers to reproduce the results or test the model directly.

I would like to kindly ask if there are plans to release the trained VulnLLM-R model weights?

If so, I highly suggest publishing them on the Hugging Face Hub. This would greatly lower the barrier for the community to access, evaluate, and build upon your work. Adding a link to the model and a quick inference example in the README would be incredibly helpful.

Thank you for your contribution to the open-source security community!

Best regards

[rucnyz]

Hi @Explorer1092,

Thank you for your interest in this paper! I completely agree with you. We actually released it a while ago here: https://huggingface.co/UCSB-SURFI/VulnLLM-R-7B

I've now added this link to the README as well. Feel free to use it!