{"total":117,"rows":[{"id":"2204.05862","title":"Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback","created":"2022-04-12","cited_by_count":3720,"area":"alignment"},{"id":"2310.03693","title":"Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!","created":"2023-10-05","cited_by_count":1154,"area":"alignment"},{"id":"2307.04657","title":"BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset","created":"2023-07-10","cited_by_count":860,"area":"alignment"},{"id":"2212.09251","title":"Discovering Language Model Behaviors with Model-Written Evaluations","created":"2022-12-19","cited_by_count":836,"area":"alignment"},{"id":"2310.04451","title":"AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models","created":"2023-10-03","cited_by_count":785,"area":"alignment"},{"id":"2307.15217","title":"Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback","created":"2023-07-27","cited_by_count":759,"area":"alignment"},{"id":"2209.07858","title":"Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned","created":"2022-08-23","cited_by_count":729,"area":"alignment"},{"id":"2308.08998","title":"Reinforced Self-Training (ReST) for Language Modeling","created":"2023-08-17","cited_by_count":500,"area":"alignment"},{"id":"2312.09390","title":"Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision","created":"2023-12-14","cited_by_count":467,"area":"alignment"},{"id":"2310.06387","title":"Jailbreak and Guard Aligned Language Models with Only Few In-Context Demonstrations","created":"2023-10-10","cited_by_count":431,"area":"alignment"},{"id":"2308.06463","title":"GPT-4 Is Too Smart To Be Safe: Stealthy Chat with LLMs via Cipher","created":"2023-08-12","cited_by_count":423,"area":"alignment"},{"id":"2310.01377","title":"UltraFeedback: Boosting Language Models with Scaled AI Feedback","created":"2023-10-02","cited_by_count":406,"area":"alignment"},{"id":"2310.06452","title":"Understanding the Effects of RLHF on LLM Generalisation and Diversity","created":"2023-10-10","cited_by_count":396,"area":"alignment"},{"id":"2305.03047","title":"Principle-Driven Self-Alignment of Language Models from Scratch with Minimal Human Supervision","created":"2023-05-04","cited_by_count":318,"area":"alignment"},{"id":"2302.08582","title":"Pretraining Language Models with Human Preferences","created":"2023-02-16","cited_by_count":247,"area":"alignment"},{"id":"2310.20624","title":"LoRA Fine-tuning Efficiently Undoes Safety Training in Llama 2-Chat 70B","created":"2023-10-31","cited_by_count":195,"area":"alignment"},{"id":"2309.16240","title":"Beyond Reverse KL: Generalizing Direct Preference Optimization with Diverse Divergence Constraints","created":"2023-09-28","cited_by_count":174,"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":"2404.08555","title":"RLHF Deciphered: A Critical Analysis of Reinforcement Learning from Human Feedback for LLMs","created":"2024-04-12","cited_by_count":90,"area":"alignment"},{"id":"2404.17546","title":"Probabilistic Inference in Language Models via Twisted Sequential Monte Carlo","created":"2024-04-26","cited_by_count":90,"area":"alignment"},{"id":"2402.06627","title":"Feedback Loops With Language Models Drive In-Context Reward Hacking","created":"2024-02-09","cited_by_count":77,"area":"alignment"},{"id":"2310.00212","title":"Pairwise Proximal Policy Optimization: Harnessing Relative Feedback for LLM Alignment","created":"2023-09-30","cited_by_count":57,"area":"alignment"},{"id":"2401.16335","title":"Iterative Data Smoothing: Mitigating Reward Overfitting and Overoptimization in RLHF","created":"2024-01-29","cited_by_count":57,"area":"alignment"},{"id":"2401.00243","title":"Uncertainty-Penalized Reinforcement Learning from Human Feedback with Diverse Reward LoRA Ensembles","created":"2023-12-30","cited_by_count":56,"area":"alignment"},{"id":"2305.00955","title":"Bridging the Gap: A Survey on Integrating (Human) Feedback for Natural Language Generation","created":"2023-05-01","cited_by_count":55,"area":"alignment"},{"id":"2310.13798","title":"Specific versus General Principles for Constitutional AI","created":"2023-10-20","cited_by_count":47,"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":"2404.05530","title":"Best-of-Venom: Attacking RLHF by Injecting Poisoned Preference Data","created":"2024-04-08","cited_by_count":35,"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":"2409.14586","title":"Backtracking Improves Generation Safety","created":"2024-09-22","cited_by_count":32,"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":"2506.17871","title":"LLM Probability Concentration: How Alignment Shrinks the Generative Horizon","created":"2025-06-22","cited_by_count":26,"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":"2406.08725","title":"RL-JACK: Reinforcement Learning-powered Black-box Jailbreaking Attack against LLMs","created":"2024-06-13","cited_by_count":23,"area":"alignment"},{"id":"2407.08770","title":"Model Surgery: Modulating LLM's Behavior Via Simple Parameter Editing","created":"2024-07-11","cited_by_count":23,"area":"alignment"},{"id":"2506.10597","title":"SoK: Evaluating Jailbreak Guardrails for Large Language Models","created":"2025-06-12","cited_by_count":21,"area":"alignment"},{"id":"2402.17358","title":"SoFA: Shielded On-the-fly Alignment via Priority Rule Following","created":"2024-02-27","cited_by_count":18,"area":"alignment"},{"id":"2502.13775","title":"VITAL: A New Dataset for Benchmarking Pluralistic Alignment in Healthcare","created":"2025-02-19","cited_by_count":17,"area":"alignment"},{"id":"2409.09774","title":"Generalizing Alignment Paradigm of Text-to-Image Generation with Preferences through $f$-divergence Minimization","created":"2024-09-15","cited_by_count":15,"area":"alignment"},{"id":"2501.11463","title":"Curiosity-Driven Reinforcement Learning from Human Feedback","created":"2025-01-20","cited_by_count":15,"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":"2410.10862","title":"Superficial Safety Alignment Hypothesis","created":"2024-10-07","cited_by_count":14,"area":"alignment"},{"id":"2506.06298","title":"Pairwise Calibrated Rewards for Pluralistic Alignment","created":"2025-05-17","cited_by_count":14,"area":"alignment"},{"id":"2406.10977","title":"Toward Optimal LLM Alignments Using Two-Player Games","created":"2024-06-16","cited_by_count":13,"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":"2312.07876","title":"Causality Analysis for Evaluating the Security of Large Language Models","created":"2023-12-13","cited_by_count":11,"area":"alignment"},{"id":"2405.11647","title":"Hummer: Towards Limited Competitive Preference Dataset","created":"2024-05-19","cited_by_count":10,"area":"alignment"},{"id":"2406.07168","title":"Teaching Language Models to Self-Improve by Learning from Language Feedback","created":"2024-06-11","cited_by_count":9,"area":"alignment"},{"id":"2412.13862","title":"Energy-Based Preference Model Offers Better Offline Alignment than the Bradley-Terry Preference Model","created":"2024-12-18","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":"2408.10701","title":"Ferret: Faster and Effective Automated Red Teaming with Reward-Based Scoring Technique","created":"2024-08-20","cited_by_count":8,"area":"alignment"},{"id":"2602.11180","title":"Mechanistic Interpretability for Large Language Model Alignment: Progress, Challenges, and Future Directions","created":"2026-01-21","cited_by_count":8,"area":"alignment"},{"id":"2407.15399","title":"Imposter.AI: Adversarial Attacks with Hidden Intentions towards Aligned Large Language Models","created":"2024-07-22","cited_by_count":7,"area":"alignment"},{"id":"2410.11677","title":"Understanding Likelihood Over-optimisation in Direct Alignment Algorithms","created":"2024-10-15","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":"2502.12485","title":"Safe at the Margins: A General Approach to Safety Alignment in Low-Resource English Languages -- A Singlish Case Study","created":"2025-02-18","cited_by_count":6,"area":"alignment"},{"id":"2510.01167","title":"Simultaneous Multi-objective Alignment Across Verifiable and Non-verifiable Rewards","created":"2025-10-01","cited_by_count":6,"area":"alignment"},{"id":"2507.16329","title":"DREAM: Scalable Red Teaming for Text-to-Image Generative Systems via Distribution Modeling","created":"2025-07-22","cited_by_count":5,"area":"alignment"},{"id":"2508.17184","title":"Towards Alignment-Centric Paradigm: A Survey of Instruction Tuning in Large Language Models","created":"2025-08-24","cited_by_count":5,"area":"alignment"},{"id":"2408.12163","title":"Preference-Guided Reflective Sampling for Aligning Language Models","created":"2024-08-22","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":"2504.16628","title":"ParetoHqD: Fast Offline Multiobjective Alignment of Large Language Models using Pareto High-quality Data","created":"2025-04-23","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":"2311.07587","title":"Frontier Language Models are not Robust to Adversarial Arithmetic, or \"What do I need to say so you agree 2+2=5?","created":"2023-11-08","cited_by_count":3,"area":"alignment"},{"id":"2311.08487","title":"Alignment is not sufficient to prevent large language models from generating harmful information: A psychoanalytic perspective","created":"2023-11-14","cited_by_count":3,"area":"alignment"},{"id":"2406.11039","title":"Dynamic Normativity: Necessary and Sufficient Conditions for Value Alignment","created":"2024-06-16","cited_by_count":3,"area":"alignment"},{"id":"2407.09447","title":"ASTPrompter: Preference-Aligned Automated Language Model Red-Teaming to Generate Low-Perplexity Unsafe Prompts","created":"2024-07-12","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":"2510.21190","title":"The Trojan Example: Jailbreaking LLMs through Template Filling and Unsafety Reasoning","created":"2025-10-24","cited_by_count":3,"area":"alignment"},{"id":"2604.02686","title":"Beyond Semantic Manipulation: Token-Space Attacks on Reward Models","created":"2026-04-03","cited_by_count":3,"area":"alignment"},{"id":"2604.10585","title":"Calibration Collapse Under Sycophancy Fine-Tuning: How Reward Hacking Breaks Uncertainty Quantification in LLMs","created":"2026-04-12","cited_by_count":3,"area":"alignment"},{"id":"2310.15694","title":"COPR: Continual Learning Human Preference through Optimal Policy Regularization","created":"2023-10-24","cited_by_count":2,"area":"alignment"},{"id":"2502.11853","title":"StructTransform: A Scalable Attack Surface for Safety-Aligned Large Language Models","created":"2025-02-17","cited_by_count":2,"area":"alignment"},{"id":"2504.05050","title":"Revealing the Intrinsic Ethical Vulnerability of Aligned Large Language Models","created":"2025-04-07","cited_by_count":2,"area":"alignment"},{"id":"2512.01037","title":"How Semantically Stable Are LLM Refusals? Measuring Confusion in Local Safety Boundaries","created":"2025-11-30","cited_by_count":2,"area":"alignment"},{"id":"2512.11838","title":"D-STEER - Preference Alignment Techniques Learn to Behave, not to Believe -- Beneath the Surface, DPO as Steering Vector Perturbation in Activation Space","created":"2025-12-03","cited_by_count":2,"area":"alignment"},{"id":"2601.18061","title":"Expert Evaluation and the Limits of Human Feedback in Mental Health AI Safety Testing","created":"2026-01-26","cited_by_count":2,"area":"alignment"},{"id":"2602.00426","title":"LLMs as High-Dimensional Nonlinear Autoregressive Models with Attention: Training, Alignment and Inference","created":"2026-01-31","cited_by_count":2,"area":"alignment"},{"id":"2602.01685","title":"Semantic-aware Wasserstein Policy Regularization for Large Language Model Alignment","created":"2026-02-02","cited_by_count":2,"area":"alignment"},{"id":"2603.03081","title":"TAO-Attack: Toward Advanced Optimization-Based Jailbreak Attacks for Large Language Models","created":"2026-03-03","cited_by_count":2,"area":"alignment"},{"id":"2603.10243","title":"GR-SAP: Generative Replay for Safety Alignment Preservation during Fine-Tuning","created":"2026-03-10","cited_by_count":2,"area":"alignment"},{"id":"2604.16916","title":"When Choices Become Risks: Safety Failures of Large Language Models under Multiple-Choice Constraints","created":"2026-04-18","cited_by_count":2,"area":"alignment"},{"id":"2606.21550","title":"AI Alignment From Social Choice Perspectives","created":"2026-06-19","cited_by_count":2,"area":"alignment"},{"id":"2404.12150","title":"Aligning language models with human preferences","created":"2024-04-18","cited_by_count":1,"area":"alignment"},{"id":"2502.16691","title":"Responsible Federated LLMs via Safety Filtering and Constitutional AI","created":"2025-02-23","cited_by_count":1,"area":"alignment"},{"id":"2503.07660","title":"Research Superalignment Should Advance Now with Alternating Competence and Conformity Optimization","created":"2025-03-08","cited_by_count":1,"area":"alignment"},{"id":"2602.01137","title":"Self-Generative Adversarial Fine-Tuning for Large Language Models","created":"2026-02-01","cited_by_count":1,"area":"alignment"},{"id":"2602.07799","title":"Fairness Aware Reward Optimization","created":"2026-02-08","cited_by_count":1,"area":"alignment"},{"id":"2603.16210","title":"MOSAIC: Composable Safety Alignment with Modular Control Tokens","created":"2026-03-17","cited_by_count":1,"area":"alignment"},{"id":"2604.03238","title":"RLHF May Not Reflect Genuine Preferences","created":"2026-01-31","cited_by_count":1,"area":"alignment"},{"id":"2604.17769","title":"Reverse Constitutional AI: A Framework for Controllable Toxic Data Generation via Probability-Clamped RLAIF","created":"2026-04-20","cited_by_count":1,"area":"alignment"},{"id":"2605.17413","title":"Ablating Safety: Mechanisms for Removing Alignment in Language Models for Security Applications","created":"2026-05-17","cited_by_count":1,"area":"alignment"},{"id":"2406.10203","title":"A Probability--Quality Trade-off in Aligned Language Models and its Relation to Sampling Adaptors","created":"2024-06-14","cited_by_count":0,"area":"alignment"},{"id":"2410.13334","title":"BiasJailbreak:Analyzing Ethical Biases and Jailbreak Vulnerabilities in Large Language Models","created":"2024-10-17","cited_by_count":0,"area":"alignment"},{"id":"2506.08885","title":"AdversariaL attacK sAfety aLIgnment(ALKALI): Safeguarding LLMs through GRACE: Geometric Representation-Aware Contrastive Enhancement- Introducing Adversarial Vulnerability Quality Index (AVQI)","created":"2025-06-10","cited_by_count":0,"area":"alignment"},{"id":"2506.10020","title":"From Threat to Tool: Leveraging Refusal-Aware Injection Attacks for Safety Alignment","created":"2025-06-07","cited_by_count":0,"area":"alignment"},{"id":"2511.12487","title":"ToxSearch: Evolving Prompts for Toxicity Search in Large Language Models","created":"2025-11-16","cited_by_count":0,"area":"alignment"}],"next_cursor":"100"}