Published research on arXiv studying retrying and resampling patterns in AI control, examining how blocked risky actions affect model behavior and safety.
Referenced in coverage
“Published research on arXiv studying retrying and resampling patterns in AI control, examining how blocked risky actions affect model behavior and safety.”