Latest version: v0.1.1
The information on this page was curated by experts in our Cybersecurity Intelligence Team.
Keeping language models honest by directly eliciting knowledge encoded in their activations
No known vulnerabilities found
Has known vulnerabilities