We have 3 of 7 papers
We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.
Not the right person? Other researchers publish under this name.
A reusable annotation skill is released that enables trajectories generated by new agent models to be standardized, annotated, and evaluated under the same framework for trajectory attribution.
It is argued that this gap leaves an irreducible residual of undetected attacks wherever inputs and actions are not restricted in advance to an enumerated set, and systematize defenses across runtime monitoring, architectural separation, and authorization.
Experiments show that Cunning training improves robustness to out-of-distribution jailbreak attacks and strengthens subsequent safety fine-tuning, and suggest that cunning data can strengthen model vigilance and complement conventional safety alignment.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
