We have 3 of 5 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.
This work exposes the fragility of chain-of-thought monitors with a controlled attack: it rewrite only the agent's reasoning to read as good-faith engineering, leaving every command and output byte-identical, so the exploit is unchanged.
This work shows that chain-of-thought monitoring is meant to catch the reward hacks that look clean in the actions and betray themselves only in the reasoning, and it is shown that this is exactly where an adversary who controls the reasoning can de-feat it.
This work formalizes protection against a strong-adversary threat model as per-report $(0, \delta)-differential privacy on the transcript of audit selections, and proves that a natural approach can never outperform uniform random auditing by more than $\delta$ at any horizon.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
