Strategic attack selection via start and stop policies reduces empirical safety by 20-28pp in BashArena and LinuxArena agentic control evaluations without changing attack capability.
arXiv preprint arXiv:2510.09462 , year =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Attack Selection in Agentic AI Control Evaluations Meaningfully Decreases Safety
Strategic attack selection via start and stop policies reduces empirical safety by 20-28pp in BashArena and LinuxArena agentic control evaluations without changing attack capability.