Yang Cao

SeClaw reveals that existing benchmarks fall short in capturing the complexities of agent behavior, enabling a more nuanced evaluation of security risks in autonomous systems.

Applying representation interventions adaptively based on input characteristics dramatically improves alignment without sacrificing general capabilities, a feat previously unmet by uniform intervention methods.

Forget training LLMs to understand privacy policies – a specialized, expert-annotated dataset and hybrid framework can do it better, achieving superior readability and reliability.