AI Safety

On the Limitations of Stochastic Pre-processing Defenses NeurIPS 2022 On the Machine Learning of Ethical Judgments from Natural Language NAACL 2022 On the Safety of Conversational Models: Taxonomy, Dataset, and Benchmark ACL 2022 Out of Distribution Detection via Neural Network Anchoring ACML 2022 OVERT: An Algorithm for Safety Verification of Neural Network Control Policies for Nonlinear Systems JMLR 2022 Parametrically Retargetable Decision-Makers Tend To Seek Power NeurIPS 2022 PARSE: An Efficient Search Method for Black-box Adversarial Text Attacks COLING 2022 Patch-Fool: Are Vision Transformers Always Robust Against Adversarial Perturbations? ICLR 2022 Path-Specific Objectives for Safer Agent Incentives AAAI 2022 Pipelines for Social Bias Testing of Large Language Models ACL 2022 Planning to Avoid Side Effects AAAI 2022 Poisoning and Backdooring Contrastive Learning ICLR 2022 Policy Smoothing for Provably Robust Reinforcement Learning ICLR 2022 Post-Training Detection of Backdoor Attacks for Two-Class and Multi-Attack Scenarios ICLR 2022 PPT: Backdoor Attacks on Pre-trained Models via Poisoned Prompt Tuning IJCAI 2022 Practical Evaluation of Adversarial Robustness via Adaptive Auto Attack CVPR 2022 Pre-activation Distributions Expose Backdoor Neurons NeurIPS 2022 Predicting Out-of-Distribution Error with the Projection Norm ICML 2022 Predicting the Influence of Fake and Real News Spreaders (Student Abstract) AAAI 2022 Preparing High School Teachers to Integrate AI Methods into STEM Classrooms AAAI 2022 Protecting Intellectual Property of Language Generation APIs with Lexical Watermark AAAI 2022 Provable Defense against Backdoor Policies in Reinforcement Learning NeurIPS 2022 Provable Guarantees for Understanding Out-of-Distribution Detection AAAI 2022 Provably Adversarially Robust Detection of Out-of-Distribution Data (Almost) for Free NeurIPS 2022 Quarantine: Sparsity Can Uncover the Trojan Attack Trigger for Free CVPR 2022