AI Safety

Inference-Time Intervention: Eliciting Truthful Answers from a Language Model NeurIPS 2023 InstructSafety: A Unified Framework for Building Multidimensional and Explainable Safety Detector through Instruction Tuning EMNLP 2023 Invariance in Policy Optimisation and Partial Identifiability in Reward Learning ICML 2023 INVITE: a Testbed of Automatically Generated Invalid Questions to Evaluate Large Language Models for Hallucinations EMNLP 2023 Iterative Reachability Estimation for Safe Reinforcement Learning NeurIPS 2023 Jailbroken: How Does LLM Safety Training Fail? NeurIPS 2023 Language Generation Models Can Cause Harm: So What Can We Do About It? An Actionable Survey EACL 2023 Language Model Alignment with Elastic Reset NeurIPS 2023 Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting NeurIPS 2023 Large Language Models as SocioTechnical Systems EMNLP 2023 Learning Shared Safety Constraints from Multi-task Demonstrations NeurIPS 2023 Learning to love diligent trolls: Accounting for rater effects in the dialogue safety task EMNLP 2023 Leveraging Unlabeled Data to Track Memorization ICLR 2023 LLMDet: A Third Party Large Language Models Generated Text Detection Tool EMNLP 2023 LM-Polygraph: Uncertainty Estimation for Language Models EMNLP 2023 Local Justice and Machine Learning: Modeling and Inferring Dynamic Ethical Preferences toward Allocations AAAI 2023 LogicAttack: Adversarial Attacks for Evaluating Logical Consistency of Natural Language Inference EMNLP 2023 Logic-driven Indirect Supervision: An Application to Crisis Counseling ACL 2023 Maestro: A Gamified Platform for Teaching AI Robustness AAAI 2023 [MASK] Insertion: a robust method for anti-adversarial attacks EACL 2023 Maximum Entropy Loss, the Silver Bullet Targeting Backdoor Attacks in Pre-trained Language Models ACL 2023 MEDIC: Remove Model Backdoors via Importance Driven Cloning CVPR 2023 Min-Max Multi-objective Bilevel Optimization with Applications in Robust Machine Learning ICLR 2023 Mitigating Adversarial Norm Training with Moral Axioms AAAI 2023 Mitigating Adversarial Vulnerability through Causal Parameter Estimation by Adversarial Double Machine Learning ICCV 2023