Frontier Curriculum and Adaptive Test-Time Compute for Efficient RLOO
It is asked whether a single success-probability signal can allocate compute more efficiently in both phases on Countdown with Qwen2.5-0.5B.
We have 2 of 2 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.
It is asked whether a single success-probability signal can allocate compute more efficiently in both phases on Countdown with Qwen2.5-0.5B.
No method reliably surfaces every backdoor, but a broadly jailbreakable model is itself anomalous, a useful signal even when the exact trigger is not recovered, allowing defenders to filter jailbreaks.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.