Shortcuts and Boundaries: Verifiable RL Misalignment and Contextual Safety Learning
Verifiable reward–context conflicts reveal how agentic RL learns shortcuts and how execution feedback supports useful, transferable contextual safety.
Xinjie Shen
AI Safety
Distributed Hidden Malice: Attacks and Defenses
How malicious intent can be distributed across individually harmless turns, and how to attack and defend at the point where capability becomes complete.