model safety

Erasing More Than Intended? How Concept Erasure Degrades the Generation of Non-Target Concepts ICCV 2025 Extracting and Understanding the Superficial Knowledge in Alignment NAACL 2025 Fine-Tuning Lowers Safety and Disrupts Evaluation Consistency ACL 2025 FLEX: A Benchmark for Evaluating Robustness of Fairness in Large Language Models NAACL 2025 Functional Homotopy: Smoothing Discrete Optimization via Continuous Parameters for LLM Jailbreak Attacks ICLR 2025 Generative Psycho-Lexical Approach for Constructing Value Systems in Large Language Models ACL 2025 Hatevolution: What Static Benchmarks Don’t Tell Us ACL 2025 Hierarchical Safety Realignment: Lightweight Restoration of Safety in Pruned Large Vision-Language Models ACL 2025 How Jailbreak Defenses Work and Ensemble? A Mechanistic Investigation EMNLP 2025 How Well Can Reasoning Models Identify and Recover from Unhelpful Thoughts? EMNLP 2025 ImgTrojan: Jailbreaking Vision-Language Models with ONE Image NAACL 2025 Improved Techniques for Optimization-Based Jailbreaking on Large Language Models ICLR 2025 Improving Large Language Model Safety with Contrastive Representation Learning EMNLP 2025 Internal Activation Revision: Safeguarding Vision Language Models Without Parameter Update AAAI 2025 Knowledge Graph Guided Evaluation of Abstention Techniques NAACL 2025 Large Language Models as Detectors or Instigators of Hate Speech in Low-resource Ethiopian Languages EMNLP 2025 Layer-Aware Representation Filtering: Purifying Finetuning Data to Preserve LLM Safety Alignment EMNLP 2025 Libra-Leaderboard: Towards Responsible AI through a Balanced Leaderboard of Safety and Capability NAACL 2025 LSSF: Safety Alignment for Large Language Models through Low-Rank Safety Subspace Fusion ACL 2025 MCIP: Protecting MCP Safety via Model Contextual Integrity Protocol EMNLP 2025 Mixture of insighTful Experts (MoTE): The Synergy of Reasoning Chains and Expert Mixtures in Self-Alignment ACL 2025 MULBERE: Multilingual Jailbreak Robustness Using Targeted Latent Adversarial Training EMNLP 2025 Multitask-Bench: Unveiling and Mitigating Safety Gaps in LLMs Fine-tuning COLING 2025 NLPART at SemEval-2025 Task 4: Forgetting is harder than Learning SemEval 2025 Phi: Preference Hijacking in Multi-modal Large Language Models at Inference Time EMNLP 2025