While an unlearned language model may no longer recall a fact directly, the fact often remains recoverable through multi-hop reasoning over related knowledge. Most existing unlearning techniques overlook this vulnerability, targeting facts in isolation while leaving their supporting knowledge intact. To achieve true fo...
Jia-Lu Wang, Pei-Zhi Niu, Hao-Teng Yin et al.· 0 citations
World models simulate the consequences of action candidates, but good planning need not preserve every physical distinction required for accurate prediction. We formalize this gap through a hierarchy of mechanism, response, and decision sufficiency. Given a candidate set, the planning query determines which physical va...
Rong-Zhe Wei, Hans Hao-Hsun Hsu, Pei-Zhi Niu et al.· 1 citation
Experiments show that turn-level boundary supervision improves intervention localization, while reinforcement learning further improves the safety--utility trade-off.
Xin-Jie Shen, Rong-Zhe Wei, Pei-Zhi Niu et al.· 0 citations
SafeClawArena is developed, a benchmark of 406 adversarial tasks executed in containerized replicas of real agent platforms with canary-marked credentials and evaluated via automated taint tracking across nine output channels, exposing the inadequacy of current defenses and suggesting directions for future hardening.
