{"total":231,"rows":[{"id":"2406.11717","title":"Refusal in Language Models Is Mediated by a Single Direction","created":"2024-06-17","cited_by_count":1032,"area":"safety"},{"id":"2406.05946","title":"Safety Alignment Should Be Made More Than Just a Few Tokens Deep","created":"2024-06-10","cited_by_count":472,"area":"alignment"},{"id":"2407.06460","title":"MUSE: Machine Unlearning Six-Way Evaluation for Language Models","created":"2024-07-08","cited_by_count":261,"area":"safety"},{"id":"2407.15549","title":"Latent Adversarial Training Improves Robustness to Persistent Harmful Behaviors in LLMs","created":"2024-07-22","cited_by_count":173,"area":"safety"},{"id":"2402.16835","title":"Eight Methods to Evaluate Robust Unlearning in LLMs","created":"2024-02-26","cited_by_count":163,"area":"safety"},{"id":"2408.17003","title":"Safety Layers in Aligned Large Language Models: The Key to LLM Security","created":"2024-08-30","cited_by_count":134,"area":"alignment"},{"id":"2504.01990","title":"Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems","created":"2025-03-31","cited_by_count":126,"area":"agents"},{"id":"2409.18025","title":"An Adversarial Perspective on Machine Unlearning for AI Safety","created":"2024-09-26","cited_by_count":123,"area":"safety"},{"id":"2410.23123","title":"On Memorization of Large Language Models in Logical Reasoning","created":"2024-10-30","cited_by_count":117,"area":"lr"},{"id":"2502.17420","title":"The Geometry of Refusal in Large Language Models: Concept Cones and Representational Independence","created":"2025-02-24","cited_by_count":115,"area":"alignment"},{"id":"2405.14577","title":"Representation Noising: A Defence Mechanism Against Harmful Finetuning","created":"2024-05-23","cited_by_count":96,"area":"safety"},{"id":"2405.18641","title":"Lisa: Lazy Safety Alignment for Large Language Models against Harmful Fine-tuning Attack","created":"2024-05-28","cited_by_count":92,"area":"alignment"},{"id":"2409.01586","title":"Booster: Tackling Harmful Fine-tuning for Large Language Models via Attenuating Harmful Perturbation","created":"2024-09-03","cited_by_count":92,"area":null},{"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":"2403.05030","title":"Defending Against Unforeseen Failure Modes with Latent Adversarial Training","created":"2024-03-08","cited_by_count":88,"area":"safety"},{"id":"2410.16454","title":"Catastrophic Failure of LLM Unlearning via Quantization","created":"2024-10-21","cited_by_count":81,"area":"safety"},{"id":"2502.02384","title":"STAIR: Improving Safety Alignment with Introspective Reasoning","created":"2025-02-04","cited_by_count":77,"area":"lr"},{"id":"2408.09600","title":"Antidote: Post-fine-tuning Safety Alignment for Large Language Models against Harmful Fine-tuning","created":"2024-08-18","cited_by_count":72,"area":"alignment"},{"id":"2410.13708","title":"On the Role of Attention Heads in Large Language Model Safety","created":"2024-10-17","cited_by_count":72,"area":"safety"},{"id":"2406.05954","title":"Aligning Large Language Models with Representation Editing: A Control Perspective","created":"2024-06-10","cited_by_count":64,"area":"alignment"},{"id":"2410.07471","title":"SEAL: Safety-enhanced Aligned LLM Fine-tuning via Bilevel Data Selection","created":"2024-10-09","cited_by_count":64,"area":null},{"id":"2403.15447","title":"Decoding Compressed Trust: Scrutinizing the Trustworthiness of Efficient LLMs Under Compression","created":"2024-03-18","cited_by_count":57,"area":null},{"id":"2405.19237","title":"ConceptPrune: Concept Editing in Diffusion Models via Skilled Neuron Pruning","created":"2024-05-29","cited_by_count":55,"area":"diffusion"},{"id":"2407.10264","title":"What Makes and Breaks Safety Fine-tuning? A Mechanistic Study","created":"2024-07-14","cited_by_count":55,"area":"safety"},{"id":"2502.07557","title":"JBShield: Defending Large Language Models from Jailbreak Attacks through Activated Concept Analysis and Manipulation","created":"2025-02-11","cited_by_count":54,"area":"safety"},{"id":"2406.06144","title":"Language Models Resist Alignment: Evidence From Data Compression","created":"2024-06-10","cited_by_count":53,"area":null},{"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":"2407.10223","title":"On Large Language Model Continual Unlearning","created":"2024-07-14","cited_by_count":52,"area":"safety"},{"id":"2412.07097","title":"On Evaluating the Durability of Safeguards for Open-Weight LLMs","created":"2024-12-10","cited_by_count":52,"area":null},{"id":"2406.18664","title":"Evaluating Copyright Takedown Methods for Language Models","created":"2024-06-26","cited_by_count":43,"area":null},{"id":"2501.01765","title":"SaLoRA: Safety-Alignment Preserved Low-Rank Adaptation","created":"2025-01-03","cited_by_count":42,"area":"safety"},{"id":"2405.09055","title":"A safety realignment framework via subspace-oriented model fusion for large language models","created":"2024-05-15","cited_by_count":41,"area":"safety"},{"id":"2502.09674","title":"The Hidden Dimensions of LLM Alignment: A Multi-Dimensional Analysis of Orthogonal Safety Directions","created":"2025-02-13","cited_by_count":41,"area":"lr"},{"id":"2502.05209","title":"Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilities","created":"2025-02-03","cited_by_count":39,"area":null},{"id":"2410.09760","title":"Targeted Vaccine: Safety Alignment for Large Language Models against Harmful Fine-Tuning via Layer-wise Perturbation","created":"2024-10-13","cited_by_count":37,"area":"alignment"},{"id":"2412.12497","title":"NLSR: Neuron-Level Safety Realignment of Large Language Models Against Harmful Fine-Tuning","created":"2024-12-17","cited_by_count":37,"area":"alignment"},{"id":"2402.16382","title":"Immunization against harmful fine-tuning attacks","created":"2024-02-26","cited_by_count":35,"area":null},{"id":"2405.13967","title":"Model Editing as a Robust and Denoised variant of DPO: A Case Study on Toxicity","created":"2024-05-22","cited_by_count":35,"area":"alignment"},{"id":"2409.07253","title":"Alignment of Diffusion Models: Fundamentals, Challenges, and Future","created":"2024-09-11","cited_by_count":35,"area":"diffusion"},{"id":"2410.18210","title":"Towards Understanding the Fragility of Multilingual LLMs against Fine-Tuning Attacks","created":"2024-10-23","cited_by_count":33,"area":"safety"},{"id":"2406.16235","title":"Preference Tuning For Toxicity Mitigation Generalizes Across Languages","created":"2024-06-23","cited_by_count":31,"area":"alignment"},{"id":"2504.01550","title":"Representation Bending for Large Language Model Safety","created":"2025-04-02","cited_by_count":30,"area":"safety"},{"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":"2404.14461","title":"Competition Report: Finding Universal Jailbreak Backdoors in Aligned LLMs","created":"2024-04-22","cited_by_count":28,"area":"safety"},{"id":"2504.10694","title":"The Jailbreak Tax: How Useful are Your Jailbreak Outputs?","created":"2025-04-14","cited_by_count":28,"area":"safety"},{"id":"2503.17239","title":"SafeMERGE: Preserving Safety Alignment in Fine-Tuned Large Language Models via Selective Layer-Wise Model Merging","created":"2025-03-21","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":"2505.06843","title":"Benign Samples Matter! Fine-tuning On Outlier Benign Samples Severely Breaks Safety","created":"2025-05-11","cited_by_count":22,"area":"safety"},{"id":"2506.01339","title":"Invariance Makes LLM Unlearning Resilient Even to Unanticipated Downstream Fine-Tuning","created":"2025-06-02","cited_by_count":21,"area":"safety"},{"id":"2412.11041","title":"Separate the Wheat from the Chaff: A Post-Hoc Approach to Safety Re-Alignment for Fine-Tuned Language Models","created":"2024-12-15","cited_by_count":20,"area":"alignment"},{"id":"2501.18100","title":"Panacea: Mitigating Harmful Fine-tuning for Large Language Models via Post-fine-tuning Perturbation","created":"2025-01-30","cited_by_count":20,"area":null},{"id":"2506.05346","title":"Why LLM Safety Guardrails Collapse After Fine-tuning: A Similarity Analysis Between Alignment and Fine-tuning Datasets","created":"2025-06-05","cited_by_count":20,"area":"alignment"},{"id":"2410.01524","title":"HarmAug: Effective Data Augmentation for Knowledge Distillation of Safety Guard Models","created":"2024-10-02","cited_by_count":19,"area":"distill"},{"id":"2410.16531","title":"Bayesian scaling laws for in-context learning","created":"2024-10-21","cited_by_count":19,"area":"safety"},{"id":"2506.07596","title":"TwinBreak: Jailbreaking LLM Security Alignments based on Twin Prompts","created":"2025-06-09","cited_by_count":19,"area":"safety"},{"id":"2506.08473","title":"AsFT: Anchoring Safety During LLM Fine-Tuning Within Narrow Safety Basin","created":"2025-06-10","cited_by_count":19,"area":null},{"id":"2502.15799","title":"Investigating the Impact of Quantization Methods on the Safety and Reliability of Large Language Models","created":"2025-02-18","cited_by_count":18,"area":"lr"},{"id":"2405.13820","title":"Towards Comprehensive Post Safety Alignment of Large Language Models via Safety Patching","created":"2024-05-22","cited_by_count":17,"area":"alignment"},{"id":"2405.19524","title":"AI Risk Management Should Incorporate Both Safety and Security","created":"2024-05-29","cited_by_count":17,"area":null},{"id":"2406.02378","title":"On the Intrinsic Self-Correction Capability of LLMs: Uncertainty and Latent Concept","created":"2024-06-04","cited_by_count":16,"area":null},{"id":"2505.15710","title":"Advancing LLM Safe Alignment with Safety Representation Ranking","created":"2025-05-21","cited_by_count":16,"area":"safety"},{"id":"2502.02260","title":"Position: Adversarial ML for LLMs Is Not Making Any Progress","created":"2025-02-04","cited_by_count":15,"area":null},{"id":"2408.02205","title":"Swiss Cheese Model for AI Safety: A Taxonomy and Reference Architecture for Multi-Layered Guardrails of Foundation Model Based Agents","created":"2024-08-05","cited_by_count":14,"area":"agents"},{"id":"2410.10862","title":"Superficial Safety Alignment Hypothesis","created":"2024-10-07","cited_by_count":14,"area":"alignment"},{"id":"2411.18797","title":"SEUF: Is Unlearning One Expert Enough for Mixture-of-Experts LLMs?","created":"2024-11-27","cited_by_count":14,"area":"moe"},{"id":"2502.01208","title":"On Almost Surely Safe Alignment of Large Language Models at Inference-Time","created":"2025-02-03","cited_by_count":14,"area":"alignment"},{"id":"2505.18588","title":"Safety Alignment via Constrained Knowledge Unlearning","created":"2025-05-24","cited_by_count":14,"area":"alignment"},{"id":"2406.18566","title":"Memorized Images in Diffusion Models share a Subspace that can be Located and Deleted","created":"2024-06-01","cited_by_count":13,"area":"diffusion"},{"id":"2507.21182","title":"SDD: Self-Degraded Defense against Malicious Fine-tuning","created":"2025-07-27","cited_by_count":13,"area":"alignment"},{"id":"2404.13752","title":"Adversarial Representation Engineering: A General Model Editing Framework for Large Language Models","created":"2024-04-21","cited_by_count":12,"area":"mechinterp"},{"id":"2410.21083","title":"Stealthy Jailbreak Attacks on Large Language Models via Benign Data Mirroring","created":"2024-10-28","cited_by_count":12,"area":"safety"},{"id":"2506.00772","title":"LIFT the Veil for the Truth: Principal Weights Emerge after Rank Reduction for Reasoning-Focused Supervised Fine-Tuning","created":"2025-06-01","cited_by_count":12,"area":"lr"},{"id":"2506.15606","title":"LoX: Low-Rank Extrapolation Robustifies LLM Safety Against Fine-tuning","created":"2025-06-18","cited_by_count":12,"area":"safety"},{"id":"2510.20956","title":"Self-Jailbreaking: Language Models Can Reason Themselves Out of Safety Alignment After Benign Reasoning Training","created":"2025-10-23","cited_by_count":12,"area":"lr"},{"id":"2501.19180","title":"Enhancing Model Defense Against Jailbreaks with Proactive Safety Reasoning","created":"2025-01-31","cited_by_count":11,"area":"lr"},{"id":"2505.16559","title":"CTRAP: Embedding Collapse Trap to Safeguard Large Language Models from Harmful Fine-Tuning","created":"2025-05-22","cited_by_count":11,"area":"safety"},{"id":"2506.09227","title":"SoK: Machine Unlearning for Large Language Models","created":"2025-06-10","cited_by_count":11,"area":"safety"},{"id":"2508.07172","title":"Gradient Surgery for Safe LLM Fine-Tuning","created":"2025-08-10","cited_by_count":11,"area":"alignment"},{"id":"2508.17158","title":"Towards Safeguarding LLM Fine-tuning APIs against Cipher Attacks","created":"2025-08-23","cited_by_count":11,"area":null},{"id":"2512.12069","title":"Rethinking Jailbreak Detection of Large Vision Language Models with Representational Contrastive Scoring","created":"2025-12-12","cited_by_count":11,"area":"lr"},{"id":"2512.21008","title":"GateBreaker: Gate-Guided Attacks on Mixture-of-Expert LLMs","created":"2025-12-24","cited_by_count":11,"area":"safety"},{"id":"2403.03942","title":"The Heuristic Core: Understanding Subnetwork Generalization in Pretrained Language Models","created":"2024-03-06","cited_by_count":10,"area":null},{"id":"2505.12186","title":"Self-Destructive Language Model","created":"2025-05-18","cited_by_count":10,"area":"safety"},{"id":"2506.05451","title":"Interpretation Meets Safety: A Survey on Interpretation Methods and Tools for Improving LLM Safety","created":"2025-06-05","cited_by_count":10,"area":"safety"},{"id":"2506.18931","title":"Safe Pruning LoRA: Robust Distance-Guided Pruning for Safety Alignment in Adaptation of LLMs","created":"2025-06-21","cited_by_count":10,"area":"alignment"},{"id":"2508.09853","title":"STREAM (ChemBio): A Standard for Transparently Reporting Evaluations in AI Model Reports","created":"2025-08-13","cited_by_count":10,"area":null},{"id":"2510.01243","title":"Detoxifying Large Language Models via Autoregressive Reward Guided Representation Editing","created":"2025-09-24","cited_by_count":10,"area":"alignment"},{"id":"2603.24543","title":"Analysing the Safety Pitfalls of Steering Vectors","created":"2026-03-25","cited_by_count":10,"area":"safety"},{"id":"2501.19202","title":"Improving LLM Unlearning Robustness via Random Perturbations","created":"2025-01-31","cited_by_count":9,"area":"safety"},{"id":"2505.14536","title":"Breaking Bad Tokens: Detoxification of LLMs Using Sparse Autoencoders","created":"2025-05-20","cited_by_count":9,"area":"mechinterp"},{"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":"2508.05775","title":"Guardians and Offenders: A Survey on Harmful Content Generation and Safety Mitigation of LLM","created":"2025-08-07","cited_by_count":9,"area":"alignment"},{"id":"2510.14301","title":"A Guardrail for Safety Preservation: When Safety-Sensitive Subspace Meets Harmful-Resistant Null-Space","created":"2025-10-16","cited_by_count":9,"area":"alignment"},{"id":"2410.04524","title":"Toward Secure Tuning: Mitigating Security Risks from Instruction Fine-Tuning","created":"2024-10-06","cited_by_count":8,"area":null},{"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":"2602.00038","title":"LSSF: Safety Alignment for Large Language Models through Low-Rank Safety Subspace Fusion","created":"2026-01-19","cited_by_count":8,"area":"alignment"},{"id":"2502.00840","title":"Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense","created":"2025-02-02","cited_by_count":7,"area":null},{"id":"2502.16770","title":"LED-Merging: Mitigating Safety-Utility Conflicts in Model Merging with Location-Election-Disjoint","created":"2025-02-24","cited_by_count":7,"area":null},{"id":"2503.19041","title":"LookAhead Tuning: Safer Language Models via Partial Answer Previews","created":"2025-03-24","cited_by_count":7,"area":"alignment"},{"id":"2504.04215","title":"Towards Understanding and Improving Refusal in Compressed Models via Mechanistic Interpretability","created":"2025-04-05","cited_by_count":7,"area":"mechinterp"}],"next_cursor":"100"}