← Search

Andrew Gritsevskiy

1 accepted papers

2024

Unelicitable Backdoors via Cryptographic Transformer Circuits

NeurIPS 2024poster

The rapid proliferation of open-source language models significantly increases the risks of downstream backdoor attacks. These backdoors can introduce dangerous behaviours during model deployment and can evade detection by conventional cybersecurity monitoring systems. In this paper, we introduce a…

Cited by 5SourcePDFScholar