Post-Training Steering in Offline Reinforcement Learning
Explore unified closed-form methods for adapting frozen policies in offline reinforcement learning without retraining, ensuring robust deployment.
AutoRISE: Advanced Agent-Driven Red-Teaming for LLM Security
Discover AutoRISE, an agent-driven strategy evolving tool that boosts red-teaming success rates for large language models without fine-tuning or GPUs.
Understanding GNNs’ Expressive Power with Global Readout
Explore how global readout enhances the expressive power of graph neural networks (GNNs) beyond traditional logic limits in this key research study.
Avionic Fuel Pump Simulation for Fault Diagnosis Benchmark
High-fidelity avionic main fuel pump simulation generates rich data for fault diagnosis and anomaly detection in aviation systems.
80% of US Government Agencies Use AI Agents Today
Over 80% of US government agencies use AI agents, boosting efficiency and service delivery. Discover the future of AI in public sector operations.