AI Safety

What Counts Underlying LLMs’ Moral Dilemma Judgments? ACL 2025 What Features in Prompts Jailbreak LLMs? Investigating the Mechanisms Behind Attacks EMNLP 2025 What Really Matters in Many-Shot Attacks? An Empirical Study of Long-Context Vulnerabilities in LLMs ACL 2025 What’s the most important value? INVP: INvestigating the Value Priorities of LLMs through Decision-making in Social Scenarios COLING 2025 “What’s Up, Doc?”: Analyzing How Users Seek Health Information in Large-Scale Conversational AI Datasets EMNLP 2025 When Allies Turn Foes: Exploring Group Characteristics of LLM-Based Multi-Agent Collaborative Systems Under Adversarial Attacks EMNLP 2025 When Backdoors Speak: Understanding LLM Backdoor Attacks Through Model-Generated Explanations ACL 2025 When Can Proxies Improve the Sample Complexity of Preference Learning? ICML 2025 When Data-Free Knowledge Distillation Meets Non-Transferable Teacher: Escaping Out-of-Distribution Trap is All You Need ICML 2025 When Lighting Deceives: Exposing Vision-Language Models' Illumination Vulnerability Through Illumination Transformation Attack ICCV 2025 When Truthful Representations Flip Under Deceptive Instructions? EMNLP 2025 Where Confabulation Lives: Latent Feature Discovery in LLMs EMNLP 2025 Where Does This Data Come From? Enhanced Source Inference Attacks in Federated Learning IJCAI 2025 Why LVLMs Are More Prone to Hallucinations in Longer Responses: The Role of Context ICCV 2025 Why Not Act on What You Know? Unleashing Safety Potential of LLMs via Self-Aware Guard Enhancement ACL 2025 Why Safeguarded Ships Run Aground? Aligned Large Language Models’ Safety Mechanisms Tend to Be Anchored in The Template Region ACL 2025 Wicked Oddities: Selectively Poisoning for Effective Clean-Label Backdoor Attacks ICLR 2025 WordGame: Efficient & Effective LLM Jailbreak via Simultaneous Obfuscation in Query and Response NAACL 2025 Words or Vision: Do Vision-Language Models Have Blind Faith in Text? CVPR 2025 X-Boundary: Establishing Exact Safety Boundary to Shield LLMs from Jailbreak Attacks without Compromising Usability EMNLP 2025 X-Guard: Multilingual Guard Agent for Content Moderation ACL 2025 YESciEval: Robust LLM-as-a-Judge for Scientific Question Answering ACL 2025 Your RAG is Unfair: Exposing Fairness Vulnerabilities in Retrieval-Augmented Generation via Backdoor Attacks EMNLP 2025 Your Semantic-Independent Watermark is Fragile: A Semantic Perturbation Attack against EaaS Watermark EMNLP 2025 Your Weak LLM is Secretly a Strong Teacher for Alignment ICLR 2025