4 papers
From Refusal Geometry to Safety Geometry: Harmfulness--Refusal Coupling under Dynamic Adversarial Fine-Tuning
Wenhao Lan, Shan Li, Xinhua Lai +3
Safety alignment requires language models to refuse harmful requests without losing the ability to answer benign ones. Existing robustness evaluations, however, do not reveal wheth…
Inference-Time Robot Behavior Steering through Physically-Aware Reconfiguration of Task-Structure
Yiyuan Pan, Hanjiang Hu, Shangtao Li +2
A central challenge in deploying learned robot policies is inference-time behavior steering: redirecting a policy at test time to satisfy user preferences not anticipated during tr…
-Reachability: Geometric-Horizon Safety Bellman Equations for Humanoid Safety
Rui Chen, Shangtao Li, Yifan Sun +1
We introduce -Reachability, a scalable approach to Hamilton--Jacobi safety analysis for high-dimensional robotic systems. Unlike prior discounted formulations that rely on fixe…
Learning Safe-Stoppability Monitors for Humanoid Robots
Yifan Sun, Yiyuan Pan, Shangtao Li +4
Emergency stop (E-stop) mechanisms are the de facto standard for robot safety. However, for humanoid robots, abruptly cutting power can itself cause catastrophic failures; instead,…