{"total":217,"rows":[{"id":"2412.14093","title":"Alignment faking in large language models","created":"2024-12-18","cited_by_count":360,"area":"alignment"},{"id":"2406.14598","title":"SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal","created":"2024-06-20","cited_by_count":240,"area":"alignment"},{"id":"2511.18397","title":"Natural Emergent Misalignment from Reward Hacking in Production RL","created":"2025-11-23","cited_by_count":136,"area":"alignment"},{"id":"2402.09154","title":"Attacking Large Language Models with Projected Gradient Descent","created":"2024-02-14","cited_by_count":94,"area":"alignment"},{"id":"2402.09063","title":"Soft Prompt Threats: Attacking Safety Alignment and Unlearning in Open-Source LLMs through the Embedding Space","created":"2024-02-14","cited_by_count":91,"area":"alignment"},{"id":"2409.18169","title":"Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey","created":"2024-09-26","cited_by_count":90,"area":"alignment"},{"id":"2410.03642","title":"Aligning LLMs with Individual Preferences via Interaction","created":"2024-10-04","cited_by_count":63,"area":"alignment"},{"id":"2404.16369","title":"Don't Say No: Jailbreaking LLM by Suppressing Refusal","created":"2024-04-25","cited_by_count":56,"area":"alignment"},{"id":"2406.14144","title":"Towards Understanding Safety Alignment: A Mechanistic Perspective from Safety Neurons","created":"2024-06-20","cited_by_count":53,"area":"alignment"},{"id":"2503.03710","title":"Improving LLM Safety Alignment with Dual-Objective Optimization","created":"2025-03-05","cited_by_count":43,"area":"alignment"},{"id":"2506.07468","title":"Chasing Moving Targets with Online Self-Play Reinforcement Learning for Safer Language Models","created":"2025-06-09","cited_by_count":39,"area":"alignment"},{"id":"2504.09466","title":"AdaSteer: Your Aligned LLM is Inherently an Adaptive Jailbreak Defender","created":"2025-04-13","cited_by_count":38,"area":"alignment"},{"id":"2409.00598","title":"Automatic Pseudo-Harmful Prompt Generation for Evaluating False Refusals in Large Language Models","created":"2024-09-01","cited_by_count":35,"area":"alignment"},{"id":"2406.11801","title":"Safety Arithmetic: A Framework for Test-time Safety Alignment of Language Models by Steering Parameters and Activations","created":"2024-06-17","cited_by_count":32,"area":"alignment"},{"id":"2409.14586","title":"Backtracking Improves Generation Safety","created":"2024-09-22","cited_by_count":32,"area":"alignment"},{"id":"2505.17147","title":"MTSA: Multi-turn Safety Alignment for LLMs through Multi-round Red-teaming","created":"2025-05-22","cited_by_count":32,"area":"alignment"},{"id":"2405.16229","title":"No Two Devils Alike: Unveiling Distinct Mechanisms of Fine-tuning Attacks","created":"2024-05-25","cited_by_count":29,"area":"alignment"},{"id":"2506.08123","title":"QA-LIGN: Aligning LLMs through Constitutionally Decomposed QA","created":"2025-06-09","cited_by_count":28,"area":"alignment"},{"id":"2408.15313","title":"Bi-Factorial Preference Optimization: Balancing Safety-Helpfulness in Language Models","created":"2024-08-27","cited_by_count":27,"area":"alignment"},{"id":"2410.15362","title":"Faster-GCG: Efficient Discrete Optimization Jailbreak Attacks against Aligned Large Language Models","created":"2024-10-20","cited_by_count":25,"area":"alignment"},{"id":"2408.01420","title":"Mission Impossible: A Statistical Perspective on Jailbreaking LLMs","created":"2024-08-02","cited_by_count":24,"area":"alignment"},{"id":"2507.18631","title":"Layer-Aware Representation Filtering: Purifying Finetuning Data to Preserve LLM Safety Alignment","created":"2025-07-24","cited_by_count":24,"area":"alignment"},{"id":"2509.11864","title":"NeuroStrike: Neuron-Level Attacks on Aligned LLMs","created":"2025-09-15","cited_by_count":22,"area":"alignment"},{"id":"2410.03523","title":"A Probabilistic Perspective on Unlearning and Alignment for Large Language Models","created":"2024-10-04","cited_by_count":21,"area":"alignment"},{"id":"2602.01539","title":"MAGIC: A Co-Evolving Attacker-Defender Adversarial Game for Robust LLM Safety","created":"2026-02-02","cited_by_count":21,"area":"alignment"},{"id":"2501.14940","title":"CASE-Bench: Context-Aware SafEty Benchmark for Large Language Models","created":"2025-01-24","cited_by_count":20,"area":"alignment"},{"id":"2502.09990","title":"X-Boundary: Establishing Exact Safety Boundary to Shield LLMs from Multi-Turn Jailbreaks without Compromising Usability","created":"2025-02-14","cited_by_count":20,"area":"alignment"},{"id":"2405.13068","title":"Uncovering Logit Suppression Vulnerabilities in LLM Safety Alignment","created":"2024-05-20","cited_by_count":18,"area":"alignment"},{"id":"2503.02574","title":"LLM-Safety Evaluations Lack Robustness","created":"2025-03-04","cited_by_count":18,"area":"alignment"},{"id":"2506.17368","title":"SAFEx: Analyzing Vulnerabilities of MoE-Based LLMs via Stable Safety-critical Expert Identification","created":"2025-06-20","cited_by_count":18,"area":"alignment"},{"id":"2406.10288","title":"Do as I do (Safely): Mitigating Task-Specific Fine-tuning Risks in Large Language Models","created":"2024-06-12","cited_by_count":17,"area":"alignment"},{"id":"2502.04322","title":"Speak Easy: Eliciting Harmful Jailbreaks from LLMs with Simple Interactions","created":"2025-02-06","cited_by_count":15,"area":"alignment"},{"id":"2502.17254","title":"REINFORCE Adversarial Attacks on Large Language Models: An Adaptive, Distributional, and Semantic Objective","created":"2025-02-24","cited_by_count":15,"area":"alignment"},{"id":"2505.20259","title":"Lifelong Safety Alignment for Language Models","created":"2025-05-26","cited_by_count":15,"area":"alignment"},{"id":"2407.16637","title":"Course-Correction: Safety Alignment Using Synthetic Preferences","created":"2024-07-23","cited_by_count":13,"area":"alignment"},{"id":"2502.11455","title":"Adversary-Aware DPO: Enhancing Safety Alignment in Vision Language Models via Adversarial Training","created":"2025-02-17","cited_by_count":12,"area":"alignment"},{"id":"2505.17598","title":"One Model Transfer to All: On Robust Jailbreak Prompts Generation against LLMs","created":"2025-05-23","cited_by_count":12,"area":"alignment"},{"id":"2505.21605","title":"SoSBench: Benchmarking Safety Alignment on Six Scientific Domains","created":"2025-05-27","cited_by_count":12,"area":"alignment"},{"id":"2406.14393","title":"Jailbreaking as a Reward Misspecification Problem","created":"2024-06-20","cited_by_count":11,"area":"alignment"},{"id":"2412.16974","title":"Cannot or Should Not? Automatic Analysis of Refusal Composition in IFT/RLHF Datasets and Refusal Behavior of Black-Box LLMs","created":"2024-12-22","cited_by_count":11,"area":"alignment"},{"id":"2501.16727","title":"xJailbreak: Representation Space Guided Reinforcement Learning for Interpretable LLM Jailbreaking","created":"2025-01-28","cited_by_count":11,"area":"alignment"},{"id":"2512.11391","title":"Mitigating the Safety Alignment Tax with Null-Space Constrained Policy Optimization","created":"2025-12-12","cited_by_count":11,"area":"alignment"},{"id":"2408.14866","title":"Advancing Adversarial Suffix Transfer Learning on Aligned Large Language Models","created":"2024-08-27","cited_by_count":10,"area":"alignment"},{"id":"2502.15435","title":"Single-pass Detection of Jailbreaking Input in Large Language Models","created":"2025-02-21","cited_by_count":10,"area":"alignment"},{"id":"2507.20333","title":"The Blessing and Curse of Dimensionality in Safety Alignment","created":"2025-07-27","cited_by_count":10,"area":"alignment"},{"id":"2510.02833","title":"Attack via Overfitting: 10-shot Benign Fine-tuning to Jailbreak LLMs","created":"2025-10-03","cited_by_count":10,"area":"alignment"},{"id":"2603.01246","title":"Defensive Refusal Bias: How Safety Alignment Fails Cyber Defenders","created":"2026-03-01","cited_by_count":10,"area":"alignment"},{"id":"2406.01514","title":"Decoupled Alignment for Robust Plug-and-Play Adaptation","created":"2024-06-03","cited_by_count":9,"area":"alignment"},{"id":"2505.17072","title":"Safety Alignment Can Be Not Superficial With Explicit Safety Signals","created":"2025-05-19","cited_by_count":9,"area":"alignment"},{"id":"2512.13655","title":"Comparative Analysis of LLM Abliteration Methods: A Cross-Architecture Evaluation","created":"2025-12-15","cited_by_count":9,"area":"alignment"},{"id":"2603.18507","title":"Expert Personas Improve LLM Alignment but Damage Accuracy: Bootstrapping Intent-Based Persona Routing with PRISM","created":"2026-03-19","cited_by_count":9,"area":"alignment"},{"id":"2604.19018","title":"Local Linearity of LLMs Enables Activation Steering via Model-Based Linear Optimal Control","created":"2026-04-21","cited_by_count":9,"area":"alignment"},{"id":"2507.11500","title":"ARMOR: Aligning Secure and Safe Large Language Models via Meticulous Reasoning","created":"2025-07-14","cited_by_count":8,"area":"alignment"},{"id":"2509.15202","title":"Beyond Surface Alignment: Rebuilding LLMs Safety Mechanism via Probabilistically Ablating Refusal Direction","created":"2025-09-18","cited_by_count":8,"area":"alignment"},{"id":"2512.20806","title":"Safety Alignment of LMs via Non-cooperative Games","created":"2025-12-23","cited_by_count":8,"area":"alignment"},{"id":"2601.01887","title":"Safety at One Shot: Patching Fine-Tuned LLMs with A Single Instance","created":"2026-01-05","cited_by_count":8,"area":"alignment"},{"id":"2602.04896","title":"Steering Externalities: Benign Activation Steering Unintentionally Increases Jailbreak Risk for Large Language Models","created":"2026-02-03","cited_by_count":8,"area":"alignment"},{"id":"2501.10639","title":"Latent-space adversarial training with post-aware calibration for defending large language models against jailbreak attacks","created":"2025-01-18","cited_by_count":7,"area":"alignment"},{"id":"2505.24369","title":"Adversarial Preference Learning for Robust LLM Alignment","created":"2025-05-30","cited_by_count":7,"area":"alignment"},{"id":"2508.14904","title":"Efficient Switchable Safety Control in LLMs via Magic-Token-Guided Co-Training","created":"2025-08-12","cited_by_count":7,"area":"alignment"},{"id":"2510.05052","title":"Proactive defense against LLM Jailbreak","created":"2025-10-06","cited_by_count":7,"area":"alignment"},{"id":"2511.19517","title":"Automating Deception: Scalable Multi-Turn LLM Jailbreaks","created":"2025-11-24","cited_by_count":7,"area":"alignment"},{"id":"2601.03868","title":"What Matters For Safety Alignment?","created":"2026-01-07","cited_by_count":7,"area":"alignment"},{"id":"2602.06258","title":"GRP-Obliteration: Unaligning LLMs With a Single Unlabeled Prompt","created":"2026-02-05","cited_by_count":7,"area":"alignment"},{"id":"2603.18736","title":"CausalRM: Causal-Theoretic Reward Modeling for RLHF from Observational User Feedbacks","created":"2026-03-19","cited_by_count":7,"area":"alignment"},{"id":"2604.08524","title":"What Drives Representation Steering? A Mechanistic Case Study on Steering Refusal","created":"2026-04-09","cited_by_count":7,"area":"alignment"},{"id":"2502.01925","title":"PANDAS: Improving Many-shot Jailbreaking via Positive Affirmation, Negative Demonstration, and Adaptive Sampling","created":"2025-02-04","cited_by_count":6,"area":"alignment"},{"id":"2502.13946","title":"Why Safeguarded Ships Run Aground? Aligned Large Language Models' Safety Mechanisms Tend to Be Anchored in The Template Region","created":"2025-02-19","cited_by_count":6,"area":"alignment"},{"id":"2505.23473","title":"EVOREFUSE: Evolutionary Prompt Optimization for Evaluation and Mitigation of LLM Over-Refusal to Pseudo-Malicious Instructions","created":"2025-05-29","cited_by_count":6,"area":"alignment"},{"id":"2404.16020","title":"Investigating Adversarial Trigger Transfer in Large Language Models","created":"2024-04-24","cited_by_count":5,"area":"alignment"},{"id":"2411.02785","title":"Stochastic Monkeys at Play: Random Augmentations Cheaply Break LLM Safety Alignment","created":"2024-11-05","cited_by_count":5,"area":"alignment"},{"id":"2412.07192","title":"PrisonBreak: Jailbreaking Large Language Models with at Most Twenty-Five Targeted Bit-flips","created":"2024-12-10","cited_by_count":5,"area":"alignment"},{"id":"2507.02332","title":"PII Jailbreaking in LLMs via Activation Steering Reveals Personal Information Leakage","created":"2025-07-03","cited_by_count":5,"area":"alignment"},{"id":"2508.20766","title":"Turning the Spell Around: Lightweight Alignment Amplification via Rank-One Safety Injection","created":"2025-08-28","cited_by_count":5,"area":"alignment"},{"id":"2511.06852","title":"Differentiated Directional Intervention A Framework for Evading LLM Safety Alignment","created":"2025-11-10","cited_by_count":5,"area":"alignment"},{"id":"2412.08201","title":"Model-Editing-Based Jailbreak against Safety-aligned Large Language Models","created":"2024-12-11","cited_by_count":4,"area":"alignment"},{"id":"2502.13603","title":"Efficient Safety Retrofitting Against Jailbreaking for LLMs","created":"2025-02-19","cited_by_count":4,"area":"alignment"},{"id":"2505.13862","title":"PandaGuard: Systematic Evaluation of LLM Safety against Jailbreaking Attacks","created":"2025-05-20","cited_by_count":4,"area":"alignment"},{"id":"2511.19218","title":"Adversarial Attack-Defense Co-Evolution for LLM Safety Alignment via Tree-Group Dual-Aware Search and Optimization","created":"2025-11-24","cited_by_count":4,"area":"alignment"},{"id":"2512.09403","title":"Black-Box Behavioral Distillation Breaks Safety Alignment in Medical LLMs","created":"2025-12-10","cited_by_count":4,"area":"alignment"},{"id":"2512.12066","title":"The Instability of Safety: How Random Seeds and Temperature Expose Inconsistent LLM Refusal Behavior","created":"2025-12-12","cited_by_count":4,"area":"alignment"},{"id":"2601.17642","title":"Health-ORSC-Bench: A Benchmark for Measuring Over-Refusal and Safety Completion in Health Context","created":"2026-01-25","cited_by_count":4,"area":"alignment"},{"id":"2602.05228","title":"Surgery: Mitigating Harmful Fine-Tuning for Large Language Models via Attention Sink","created":"2026-02-05","cited_by_count":4,"area":"alignment"},{"id":"2602.07340","title":"Revisiting Robustness for LLM Safety Alignment via Selective Geometry Control","created":"2026-02-07","cited_by_count":4,"area":"alignment"},{"id":"2602.20042","title":"AI Alignment Breaks at the Edge","created":"2026-02-23","cited_by_count":4,"area":"alignment"},{"id":"2603.11331","title":"Jailbreak Scaling Laws for Large Language Models: Polynomial-Exponential Crossover","created":"2026-03-11","cited_by_count":4,"area":"alignment"},{"id":"2603.15684","title":"State-Dependent Safety Failures in Multi-Turn Language Model Interaction","created":"2026-03-15","cited_by_count":4,"area":"alignment"},{"id":"2604.18901","title":"Harmful Intent as a Geometrically Recoverable Feature of LLM Residual Streams","created":"2026-04-20","cited_by_count":4,"area":"alignment"},{"id":"2605.00195","title":"Diversity in Large Language Models under Supervised Fine-Tuning","created":"2026-04-30","cited_by_count":4,"area":"alignment"},{"id":"2606.02530","title":"SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment","created":"2026-06-01","cited_by_count":4,"area":"alignment"},{"id":"2502.08657","title":"Refining Positive and Toxic Samples for Dual Safety Self-Alignment of LLMs with Minimal Human Interventions","created":"2025-02-08","cited_by_count":3,"area":"alignment"},{"id":"2509.19360","title":"Semantic Representation Attack against Aligned Large Language Models","created":"2025-09-18","cited_by_count":3,"area":"alignment"},{"id":"2510.08604","title":"LatentBreak: Jailbreaking Large Language Models through Latent Space Feedback","created":"2025-10-07","cited_by_count":3,"area":"alignment"},{"id":"2510.21190","title":"The Trojan Example: Jailbreaking LLMs through Template Filling and Unsafety Reasoning","created":"2025-10-24","cited_by_count":3,"area":"alignment"},{"id":"2511.17937","title":"Alignment Faking - the Train -> Deploy Asymmetry: Through a Game-Theoretic Lens with Bayesian-Stackelberg Equilibria","created":"2025-11-22","cited_by_count":3,"area":"alignment"},{"id":"2511.21050","title":"Breaking the Safety-Capability Tradeoff: Reinforcement Learning with Verifiable Rewards Maintains Safety Guardrails in LLMs","created":"2025-11-26","cited_by_count":3,"area":"alignment"},{"id":"2602.08813","title":"Robust Policy Optimization to Prevent Catastrophic Forgetting","created":"2026-02-09","cited_by_count":3,"area":"alignment"},{"id":"2603.16152","title":"HIPO: Instruction Hierarchy via Constrained Reinforcement Learning","created":"2026-03-17","cited_by_count":3,"area":"alignment"},{"id":"2603.23184","title":"ImplicitRM: Unbiased Reward Modeling from Implicit Preference Data for LLM alignment","created":"2026-03-24","cited_by_count":3,"area":"alignment"},{"id":"2604.12710","title":"LASA: Language-Agnostic Semantic Alignment at the Semantic Bottleneck for LLM Safety","created":"2026-04-13","cited_by_count":3,"area":"alignment"}],"next_cursor":"100"}