AI Safety

Query-Efficient and Scalable Black-Box Adversarial Attacks on Discrete Sequential Data via Bayesian Optimization ICML 2022 QUERY EFFICIENT DECISION BASED SPARSE ATTACKS AGAINST BLACK-BOX DEEP LEARNING MODELS ICLR 2022 Randomized Message-Interception Smoothing: Gray-box Certificates for Graph Neural Networks NeurIPS 2022 Reachability Constrained Reinforcement Learning ICML 2022 Red Teaming Language Models with Language Models EMNLP 2022 Reducing Offensive Replies in Open Domain Dialogue Systems Interspeech 2022 Regret Bounds for Risk-Sensitive Reinforcement Learning NeurIPS 2022 Resiliency of Perception-Based Controllers Against Attacks L4DC 2022 Responsible Disclosure of Generative Models Using Scalable Fingerprinting ICLR 2022 ResSFL: A Resistance Transfer Framework for Defending Model Inversion Attack in Split Federated Learning CVPR 2022 Rethinking Adversarial Transferability from a Data Distribution Perspective ICLR 2022 Rethinking Lipschitz Neural Networks and Certified Robustness: A Boolean Function Perspective NeurIPS 2022 Rethinking the Reverse-engineering of Trojan Triggers NeurIPS 2022 Robots-Dont-Cry: Understanding Falsely Anthropomorphic Utterances in Dialog Systems EMNLP 2022 Robust Hate Speech Detection via Mitigating Spurious Correlations AACL 2022 Robust Heterogeneous Graph Neural Networks against Adversarial Attacks AAAI 2022 Robust Learning Meets Generative Models: Can Proxy Distributions Improve Adversarial Robustness? ICLR 2022 ROSE: Robust Selective Fine-tuning for Pre-trained Language Models EMNLP 2022 Safe Control Under Input Limits with Neural Control Barrier Functions CoRL 2022 Safe Control with Neural Network Dynamic Models L4DC 2022 Safe Neurosymbolic Learning with Differentiable Symbolic Execution ICLR 2022 SaFeRDialogues: Taking Feedback Gracefully after Conversational Safety Failures ACL 2022 Safe Reinforcement Learning with Chance-constrained Model Predictive Control L4DC 2022 Safe Robot Learning in Assistive Devices through Neural Network Repair CoRL 2022 SafeText: A Benchmark for Exploring Physical Safety in Language Models EMNLP 2022