AI Safety

Alignment Quality Index (AQI) : Beyond Refusals: AQI as an Intrinsic Alignment Diagnostic via Latent Geometry, Cluster Divergence, and Layer wise Pooled Representations EMNLP 2025 Align-Pro: A Principled Approach to Prompt Optimization for LLM Alignment AAAI 2025 A Likelihood Based Approach for Watermark Detection AISTATS 2025 ALIS: Aligned LLM Instruction Security Strategy for Unsafe Input Prompt COLING 2025 All You Need Is S P A C E: When Jailbreaking Meets Bias Audit and Reveals What Lies Beneath the Guardrails (Student Abstract) AAAI 2025 AlphaVerus: Bootstrapping Formally Verified Code Generation through Self-Improving Translation and Treefinement ICML 2025 ALRPHFS: Adversarially Learned Risk Patterns with Hierarchical Fast & Slow Reasoning for Robust Agent Defense EMNLP 2025 AMIA: Automatic Masking and Joint Intention Analysis Makes LVLMs Robust Jailbreak Defenders EMNLP 2025 A Mousetrap: Fooling Large Reasoning Models for Jailbreak with Chain of Iterative Chaos ACL 2025 An Auditing Test to Detect Behavioral Shift in Language Models ICLR 2025 Anecdoctoring: Automated Red-Teaming Across Language and Place EMNLP 2025 An Engorgio Prompt Makes Large Language Model Babble on ICLR 2025 An Evolutionary Perspective on AI Alignment (Student Abstract) AAAI 2025 An Interpretable N-gram Perplexity Threat Model for Large Language Model Jailbreaks ICML 2025 An Optimistic Algorithm for online CMDPS with Anytime Adversarial Constraints ICML 2025 An Optimizable Suffix Is Worth A Thousand Templates: Efficient Black-box Jailbreaking without Affirmative Phrases via LLM as Optimizer NAACL 2025 Answer When Needed, Forget When Not: Language Models Pretend to Forget via In-Context Knowledge Unlearning ACL 2025 Antidote: Post-fine-tuning Safety Alignment for Large Language Models against Harmful Fine-tuning Attack ICML 2025 AntiLeakBench: Preventing Data Contamination by Automatically Constructing Benchmarks with Updated Real-World Knowledge ACL 2025 A Practical Examination of AI-Generated Text Detectors for Large Language Models NAACL 2025 Arabic Dataset for LLM Safeguard Evaluation NAACL 2025 AraSafe: Benchmarking Safety in Arabic LLMs EMNLP 2025 AraTrust: An Evaluation of Trustworthiness for LLMs in Arabic COLING 2025 Arbiters of Ambivalence: Challenges of using LLMs in No-Consensus tasks ACL 2025 Are Bias Evaluation Methods Biased ? ACL 2025