AI Safety

Neurosymbolic Reinforcement Learning with Formally Verified Exploration NeurIPS 2020 On Adaptive Attacks to Adversarial Example Defenses NeurIPS 2020 Once-for-All Adversarial Training: In-Situ Tradeoff between Robustness and Accuracy for Free NeurIPS 2020 On Isometry Robustness of Deep 3D Point Cloud Models Under Adversarial Attacks CVPR 2020 On the Need for Topology-Aware Generative Models for Manifold-Based Defenses ICLR 2020 On the Robustness of Data-Driven Controllers for Linear Systems L4DC 2020 On the Trade-off between Adversarial and Backdoor Robustness NeurIPS 2020 Partial Adversarial Behavior Deception in Security Games IJCAI 2020 Pessimism About Unknown Unknowns Inspires Conservatism COLT 2020 Poison Attacks against Text Datasets with Conditional Adversarially Regularized Autoencoder EMNLP 2020 Policy Optimization for H_2 Linear Control with H_∞ Robustness Guarantee: Implicit Regularization and Global Convergence L4DC 2020 Prediction Poisoning: Towards Defenses Against DNN Model Stealing Attacks ICLR 2020 Projection-Based Constrained Policy Optimization ICLR 2020 Querying to Find a Safe Policy under Uncertain Safety Constraints in Markov Decision Processes AAAI 2020 Randomized Smoothing of All Shapes and Sizes ICML 2020 Reactive motion planning with probabilisticsafety guarantees CoRL 2020 RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models EMNLP 2020 Recovery of sparse linear classifiers from mixture of responses NeurIPS 2020 Reevaluating Adversarial Examples in Natural Language EMNLP 2020 Reinforcement Learning for Safety-Critical Control under Model Uncertainty, using Control Lyapunov Functions and Control Barrier Functions RSS 2020 Responsive Safety in Reinforcement Learning by PID Lagrangian Methods ICML 2020 Robust anomaly detection and backdoor attack detection via differential privacy ICLR 2020 Robust Deep Learning as Optimal Control: Insights and Convergence Guarantees L4DC 2020 Robust Deep Reinforcement Learning against Adversarial Perturbations on State Observations NeurIPS 2020 Robustness Certificates for Sparse Adversarial Attacks by Randomized Ablation AAAI 2020