AI Models Can Provably Hide Arbitrary Capabilities
Capability evaluations aim to surface dangerous model behaviors before deployment, but their reliability depends on the assumption that hidden capabilities can be elicited. We show this assumption does not hold under adversarial conditions by constructing backdoor attacks that embed encrypted neural networks within host model weights, decrypting and executing them only upon receiving a secret trigger. Using digital lockers, these hidden circuits remain provably hard to elicit and interpret…
arXivLabs is a framework for developing and sharing arXiv features, emphasizing openness and user privacy.
arXivLabs allows collaborators to develop and share arXiv features.
arXiv values openness, community, and user data privacy.
Partners must adhere to arXiv's values of excellence and privacy.
Summarised automatically by AI from the original article by Hacker News. AI can make mistakes, so check the original for details.
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Read the full story on Hacker NewsThat's the opening of the story. The full piece is published by Hacker News.