AI Safety

Large Language Models as Detectors or Instigators of Hate Speech in Low-resource Ethiopian Languages EMNLP 2025 Large Language Models can Become Strong Self-Detoxifiers ICLR 2025 Large Language Models Have Intrinsic Meta-Cognition, but Need a Good Lens EMNLP 2025 Large Language Models Often Say One Thing and Do Another ICLR 2025 Large Language Models Threaten Language’s Epistemic and Communicative Foundations EMNLP 2025 Layer-Aware Representation Filtering: Purifying Finetuning Data to Preserve LLM Safety Alignment EMNLP 2025 Layer-Level Self-Exposure and Patch: Affirmative Token Mitigation for Jailbreak Attack Defense NAACL 2025 Leaky Thoughts: Large Reasoning Models Are Not Private Thinkers EMNLP 2025 Learn and Unlearn: Addressing Misinformation in Multilingual LLMs EMNLP 2025 Learning Auxiliary Tasks Improves Reference-Free Hallucination Detection in Open-Domain Long-Form Generation ACL 2025 Learning Diverse Attacks on Large Language Models for Robust Red-Teaming and Safety Tuning ICLR 2025 Learning Randomized Algorithms with Transformers ICLR 2025 Learning Safety Constraints for Large Language Models ICML 2025 Learning to Rewrite: Generalized LLM-Generated Text Detection ACL 2025 LED-Merging: Mitigating Safety-Utility Conflicts in Model Merging with Location-Election-Disjoint ACL 2025 LEGEND: Leveraging Representation Engineering to Annotate Safety Margin for Preference Datasets AAAI 2025 Lessons for Editors of AI Incidents from the AI Incident Database AAAI 2025 Let Them Down Easy! Contextual Effects of LLM Guardrails on User Perceptions and Preferences EMNLP 2025 Leveraging Constraint Violation Signals for Action Constrained Reinforcement Learning AAAI 2025 Leveraging Human Input to Enable Robust, Interactive, and Aligned AI Systems AAAI 2025 LEVIS: Large Exact Verifiable Input Spaces for Neural Networks ICML 2025 LibVulnWatch: A Deep Assessment Agent System and Leaderboard for Uncovering Hidden Vulnerabilities in Open-Source AI Libraries ACL 2025 Lightweight Safety Guardrails Using Fine-tuned BERT Embeddings COLING 2025 Limits to scalable evaluation at the frontier: LLM as judge won’t beat twice the data ICLR 2025 Line of Duty: Evaluating LLM Self-Knowledge via Consistency in Feasibility Boundaries NAACL 2025