Publications
Preprint · arXiv · February 24, 2026
When can we trust untrusted monitoring? A safety case sketch across collusion strategies
Safety Systems Engineering in AutonomyEthics and Social Impacts of AIAdversarial Robustness in Machine LearningAI safetysafety casefocal points
Related publications
- Screwtape: Adversarial Transcript Editing for Monitor Red-TeamingarXiv · 2026 · Same author: Георгий Кожевников
- Beyond Attention: Module-Level Refusal Directions in Mamba-Based Hybrid Language ModelsLecture notes in computer science · 2026 · Shared topic: Adversarial Robustness in Machine Learning
- PhysioTRACE: Provenance-Aware Stress Tests for Physiological Foundation ModelsarXiv · 2026 · Shared topic: Adversarial Robustness in Machine Learning
- Evaluating Robustness of AI-Generated Text Detectors Under Simulated Iterative Obfuscation2026 · Shared topic: Adversarial Robustness in Machine Learning