SchwinnL / LLM_Embedding_Attack

Code to conduct an embedding attack on LLMs
18 stars 3 forks source link

When will more code be released? #1

Closed pkulium closed 5 months ago

pkulium commented 5 months ago

Great work! When will more code be released for the paper(Soft Prompt Threats: Attacking Safety Alignment and Unlearning in Open-Source LLMs through the Embedding Space)?

SchwinnL commented 5 months ago

Hi,

I appreciate your interest! I'm planning to release the code in the following two weeks.

SchwinnL commented 5 months ago

Hi, unfortunately, I got sick and have not been able to upload the code yet. If you want you can send me an email: l.schwinn@tum.de, and I will send you the original non-refactored code.