AI Safety

Improving Consistency in LLM Inference using Probabilistic Tokenization NAACL 2025 Improving Deep Detector Robustness via Detection-Related Discriminant Maximization and Reorganization WACV 2025 Improving Factuality with Explicit Working Memory ACL 2025 Improving Large Language Model Safety with Contrastive Representation Learning EMNLP 2025 Improving LLM Safety Alignment with Dual-Objective Optimization ICML 2025 Improving Out-of-Distribution Detection via Dynamic Covariance Calibration ICML 2025 Incentivizing Safer Actions in Policy Optimization for Constrained Reinforcement Learning IJCAI 2025 Increased Compute Efficiency and the Diffusion of AI Capabilities AAAI 2025 InfAlign: Inference-aware language model alignment ICML 2025 Influence-Based Fair Selection for Sample-Discriminative Backdoor Attack AAAI 2025 Influences on LLM Calibration: A Study of Response Agreement, Loss Functions, and Prompt Styles ACL 2025 Information-theoretic Distinctions Between Deception and Confusion IJCNLP 2025 Inherent and emergent liability issues in LLM-based agentic systems: a principal-agent perspective ACL 2025 Injecting Universal Jailbreak Backdoors into LLMs in Minutes ICLR 2025 Insights into Alignment: Evaluating DPO and its Variants Across Multiple Tasks ACL 2025 Inspection and Control of Self-Generated-Text Recognition Ability in Llama3-8b-Instruct ICLR 2025 Instantiation-based Formalization of Logical Reasoning Tasks Using Language Models and Logical Solvers IJCAI 2025 Instructional Segment Embedding: Improving LLM Safety with Instruction Hierarchy ICLR 2025 Intention Analysis Makes LLMs A Good Jailbreak Defender COLING 2025 Internal Activation Revision: Safeguarding Vision Language Models Without Parameter Update AAAI 2025 Internal Value Alignment in Large Language Models through Controlled Value Vector Activation ACL 2025 Interpreting Language Reward Models via Contrastive Explanations ICLR 2025 Intrinsic Model Weaknesses: How Priming Attacks Unveil Vulnerabilities in Large Language Models NAACL 2025 Intrinsic Test of Unlearning Using Parametric Knowledge Traces EMNLP 2025 Invariance Makes LLM Unlearning Resilient Even to Unanticipated Downstream Fine-Tuning ICML 2025