14 Oct 2026
Screwtape: Adversarial Transcript Editing for Monitor Red-Teaming
14 Oct 2026 · Preprint · arXiv · Artificial Intelligence
@baraltiva
ML Research Intern · Redwood Research
I work at Redwood Research on mitigating catastrophic risks from frontier AI systems.
14 Oct 2026 · Preprint · arXiv · Artificial Intelligence
24 Feb 2026 · Preprint · arXiv · Safety, Risk, Reliability and Quality