3 papers
cs.LG2026
SteinGate: Tail-Sensitive Safe Reinforcement Learning via Stein Discrepancy
Yassine Chemingui, Chenhua Fan, Honghao Wei +1
SteinGate introduces a distributional safety certificate based on Kernelized Stein Discrepancy to detect rare, high-cost tail events in reinforcement learning and dynamically switc…
cs.LG2026
Towards Fast Safe Online Reinforcement Learning via Policy Finetuning
Keru Chen, Honghao Wei, Zhigang Deng +1
The high costs and risks involved in extensive environment interactions hinder the practical application of current online safe reinforcement learning (RL) methods. While offline s…
cs.LG2025
Constraint-Adaptive Policy Switching for Offline Safe Reinforcement Learning
Yassine Chemingui, Aryan Deshwal, Honghao Wei +2
Offline safe reinforcement learning (OSRL) involves learning a decision-making policy to maximize rewards from a fixed batch of training data to satisfy pre-defined safety constrai…