{"total":88,"rows":[{"id":"2402.09283","title":"Attacks, Defenses and Evaluations for LLM Conversation Safety: A Survey","created":"2024-02-14","cited_by_count":130,"area":null},{"id":"2401.14446","title":"Black-Box Access is Insufficient for Rigorous AI Audits","created":"2024-01-25","cited_by_count":126,"area":null},{"id":"2405.16833","title":"Safe LoRA: the Silver Lining of Reducing Safety Risks when Fine-tuning Large Language Models","created":"2024-05-27","cited_by_count":125,"area":null},{"id":"2311.12786","title":"Mechanistically analyzing the effects of fine-tuning on procedurally defined tasks","created":"2023-11-21","cited_by_count":115,"area":null},{"id":"2402.18540","title":"Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates","created":"2024-02-28","cited_by_count":96,"area":null},{"id":"2406.05761","title":"The BiGGen Bench: A Principled Benchmark for Fine-grained Evaluation of Language Models with Language Models","created":"2024-06-09","cited_by_count":93,"area":null},{"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":"2406.20053","title":"Covert Malicious Finetuning: Challenges in Safeguarding LLM Adaptation","created":"2024-06-28","cited_by_count":80,"area":null},{"id":"2405.03644","title":"When LLMs Meet Cybersecurity: A Systematic Literature Review","created":"2024-05-06","cited_by_count":71,"area":null},{"id":"2407.09121","title":"Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training","created":"2024-07-12","cited_by_count":69,"area":null},{"id":"2403.07918","title":"On the Societal Impact of Open Foundation Models","created":"2024-02-27","cited_by_count":65,"area":null},{"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":"2510.26243","title":"Angular Steering: Behavior Control via Rotation in Activation Space","created":"2025-10-30","cited_by_count":53,"area":null},{"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":"2402.04247","title":"Risks of AI Scientists: Prioritizing Safeguarding Over Autonomy","created":"2024-02-06","cited_by_count":49,"area":null},{"id":"2405.19550","title":"Stress-Testing Capability Elicitation With Password-Locked Models","created":"2024-05-29","cited_by_count":46,"area":null},{"id":"2410.07137","title":"Cheating Automatic LLM Benchmarks: Null Models Achieve High Win Rates","created":"2024-10-09","cited_by_count":45,"area":null},{"id":"2401.17633","title":"Navigating the OverKill in Large Language Models","created":"2024-01-31","cited_by_count":44,"area":null},{"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":"2505.17306","title":"Refusal Direction is Universal Across Safety-Aligned Languages","created":"2025-05-22","cited_by_count":38,"area":null},{"id":"2311.00117","title":"BadLlama: cheaply removing safety fine-tuning from Llama 2-Chat 13B","created":"2023-10-31","cited_by_count":36,"area":null},{"id":"2402.16382","title":"Immunization against harmful fine-tuning attacks","created":"2024-02-26","cited_by_count":35,"area":null},{"id":"2410.10760","title":"Denial-of-Service Poisoning Attacks against Large Language Models","created":"2024-10-14","cited_by_count":35,"area":null},{"id":"2412.13879","title":"Crabs: Consuming Resource via Auto-generation for LLM-DoS Attack under Black-box Settings","created":"2024-12-18","cited_by_count":35,"area":null},{"id":"2410.10014","title":"Safety-Aware Fine-Tuning of Large Language Models","created":"2024-10-13","cited_by_count":34,"area":null},{"id":"2502.13175","title":"Towards Robust and Secure Embodied AI: A Survey on Vulnerabilities and Attacks","created":"2025-02-18","cited_by_count":34,"area":null},{"id":"2402.14968","title":"Mitigating Fine-tuning based Jailbreak Attack with Backdoor Enhanced Safety Alignment","created":"2024-02-22","cited_by_count":31,"area":null},{"id":"2405.02501","title":"PICLe: Eliciting Diverse Behaviors from Large Language Models with Persona In-Context Learning","created":"2024-05-03","cited_by_count":29,"area":null},{"id":"2402.11725","title":"How Susceptible are Large Language Models to Ideological Manipulation?","created":"2024-02-18","cited_by_count":26,"area":null},{"id":"2312.12736","title":"Learning and Forgetting Unsafe Examples in Large Language Models","created":"2023-12-20","cited_by_count":24,"area":null},{"id":"2403.13784","title":"The Model Openness Framework: Promoting Completeness and Openness for Reproducibility, Transparency, and Usability in Artificial Intelligence","created":"2024-03-20","cited_by_count":23,"area":null},{"id":"2410.10343","title":"Locking Down the Finetuned LLMs Safety","created":"2024-10-14","cited_by_count":22,"area":null},{"id":"2502.20968","title":"Beware of Your Po! Measuring and Mitigating AI Safety Risks in Role-Play Fine-Tuning of LLMs","created":"2025-02-28","cited_by_count":22,"area":null},{"id":"2407.12344","title":"The Better Angels of Machine Personality: How Personality Relates to LLM Safety","created":"2024-07-17","cited_by_count":21,"area":null},{"id":"2311.02105","title":"Making Harmful Behaviors Unlearnable for Large Language Models","created":"2023-11-02","cited_by_count":20,"area":null},{"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":"2506.00085","title":"COSMIC: Generalized Refusal Direction Identification in LLM Activations","created":"2025-05-30","cited_by_count":19,"area":null},{"id":"2508.03153","title":"Estimating Worst-Case Frontier Risks of Open-Weight LLMs","created":"2025-08-05","cited_by_count":19,"area":null},{"id":"2505.12038","title":"Safe Delta: Consistently Preserving Safety when Fine-Tuning LLMs on Diverse Datasets","created":"2025-05-17","cited_by_count":18,"area":null},{"id":"2405.19524","title":"AI Risk Management Should Incorporate Both Safety and Security","created":"2024-05-29","cited_by_count":17,"area":null},{"id":"2410.15821","title":"The effect of fine-tuning on language model toxicity","created":"2024-10-21","cited_by_count":16,"area":null},{"id":"2311.09585","title":"LifeTox: Unveiling Implicit Toxicity in Life Advice","created":"2023-11-16","cited_by_count":14,"area":null},{"id":"2505.17646","title":"Unveiling the Basin-Like Loss Landscape in Large Language Models","created":"2025-05-23","cited_by_count":13,"area":null},{"id":"2409.09041","title":"Acceptable Use Policies for Foundation Models","created":"2024-08-29","cited_by_count":12,"area":null},{"id":"2509.13281","title":"RepIt: Steering Language Models with Concept-Specific Refusal Vectors","created":"2025-09-16","cited_by_count":12,"area":null},{"id":"2412.19512","title":"Safeguard Fine-Tuned LLMs Through Pre- and Post-Tuning Model Merging","created":"2024-12-27","cited_by_count":11,"area":null},{"id":"2508.17158","title":"Towards Safeguarding LLM Fine-tuning APIs against Cipher Attacks","created":"2025-08-23","cited_by_count":11,"area":null},{"id":"2405.10295","title":"Societal Adaptation to Advanced AI","created":"2024-05-16","cited_by_count":10,"area":null},{"id":"2508.20697","title":"Token Buncher: Shielding LLMs from Harmful Reinforcement Learning Fine-Tuning","created":"2025-08-28","cited_by_count":8,"area":null},{"id":"2604.03976","title":"Quantifying Trust: Financial Risk Management for Trustworthy AI Agents","created":"2026-04-05","cited_by_count":8,"area":null},{"id":"2410.09838","title":"Uncovering, Explaining, and Mitigating the Superficial Safety of Backdoor Defense","created":"2024-10-13","cited_by_count":7,"area":null},{"id":"2501.19040","title":"Towards the Worst-case Robustness of Large Language Models","created":"2025-01-31","cited_by_count":7,"area":null},{"id":"2502.13593","title":"Toward Robust Non-Transferable Learning: A Survey and Benchmark","created":"2025-02-19","cited_by_count":7,"area":null},{"id":"2510.01529","title":"Bypassing Prompt Guards in Production with Controlled-Release Prompting","created":"2025-10-02","cited_by_count":7,"area":null},{"id":"2404.10552","title":"Unveiling the Misuse Potential of Base Large Language Models via In-Context Learning","created":"2024-04-16","cited_by_count":6,"area":null},{"id":"2408.14792","title":"Measuring Human Contribution in AI-Assisted Content Generation","created":"2024-08-27","cited_by_count":6,"area":null},{"id":"2409.03274","title":"Recent Advances in Attack and Defense Approaches of Large Language Models","created":"2024-09-05","cited_by_count":6,"area":null},{"id":"2510.24222","title":"HACK: Hallucinations Along Certainty and Knowledge Axes","created":"2025-10-28","cited_by_count":6,"area":null},{"id":"2410.16672","title":"The Tug of War Within: Mitigating the Fairness-Privacy Conflicts in Large Language Models","created":"2024-10-22","cited_by_count":5,"area":null},{"id":"2410.22944","title":"Focus On This, Not That! Steering LLMs with Adaptive Feature Specification","created":"2024-10-30","cited_by_count":5,"area":null},{"id":"2509.07287","title":"Paladin: Defending LLM-enabled Phishing Emails with a New Trigger-Tag Paradigm","created":"2025-09-08","cited_by_count":5,"area":null},{"id":"2606.15980","title":"Do Activation Monitors Survive Model Updates? Benchmarking, Predicting, and Repairing Activation-Monitor Staleness","created":"2026-06-14","cited_by_count":5,"area":null},{"id":"2511.00382","title":"Efficiency vs. Alignment: Investigating Safety and Fairness Risks in Parameter-Efficient Fine-Tuning of LLMs","created":"2025-11-01","cited_by_count":4,"area":null},{"id":"2601.22313","title":"Hair-Trigger Alignment: Black-Box Evaluation Cannot Guarantee Post-Update Alignment","created":"2026-01-29","cited_by_count":4,"area":null},{"id":"2605.12705","title":"Early Data Exposure Improves Robustness to Subsequent Fine-Tuning","created":"2026-05-12","cited_by_count":4,"area":null},{"id":"2512.01899","title":"Provably Safe Model Updates","created":"2025-12-01","cited_by_count":3,"area":null},{"id":"2602.16729","title":"Intent Laundering: AI Safety Datasets Are Not What They Seem","created":"2026-02-17","cited_by_count":3,"area":null},{"id":"2603.01185","title":"Token-level Data Selection for Safe LLM Fine-tuning","created":"2026-03-01","cited_by_count":3,"area":null},{"id":"2410.23472","title":"Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems","created":"2024-10-30","cited_by_count":2,"area":null},{"id":"2411.07691","title":"New Emerged Security and Privacy of Pre-trained Model: a Survey and Outlook","created":"2024-11-12","cited_by_count":2,"area":null},{"id":"2505.03052","title":"Teaching Models to Understand (but not Generate) High-risk Data","created":"2025-05-05","cited_by_count":2,"area":null},{"id":"2506.20251","title":"Q-resafe: Assessing Safety Risks and Quantization-aware Safety Patching for Quantized Large Language Models","created":"2025-06-25","cited_by_count":2,"area":null},{"id":"2605.06901","title":"Reflections and New Directions for Human-Centered Large Language Models","created":"2026-05-07","cited_by_count":2,"area":null},{"id":"2605.14605","title":"A Few Steps Further: Why Defenses Against Malicious Finetuning Erode Under Continued Training","created":"2026-05-14","cited_by_count":2,"area":null},{"id":"2606.22676","title":"Skin-Deep: A Geometric Diagnostic for Alignment Fragility in Large Language Model Representations","created":"2026-06-21","cited_by_count":2,"area":null},{"id":"2411.14368","title":"RV4Chatbot: Are Chatbots Allowed to Dream of Electric Sheep?","created":"2024-11-21","cited_by_count":1,"area":null},{"id":"2606.00160","title":"DataShield: Safety-degrading Data Filtering for LLM Benign Instruction Fine-Tuning","created":"2026-05-29","cited_by_count":1,"area":null},{"id":"2607.22676","title":"How LLM Task-Adaptation Reshapes Alignment: A Multi-dimensional Study of Behavioral and Representational Drift","created":"2026-07-10","cited_by_count":1,"area":null},{"id":"2602.17546","title":"Learning to Stay Safe: Adaptive Regularization Against Safety Degradation during Fine-Tuning","created":"2026-02-19","cited_by_count":0,"area":null},{"id":"2603.04407","title":"Semantic Containment as a Fundamental Property of Emergent Misalignment","created":"2026-02-02","cited_by_count":0,"area":null},{"id":"2604.00324","title":"The Persistent Vulnerability of Aligned AI Systems","created":"2026-03-31","cited_by_count":0,"area":null},{"id":"2605.30640","title":"CSULoRA: Closest Safe Update Low-Rank Adaptation","created":"2026-05-28","cited_by_count":0,"area":null},{"id":"2606.07631","title":"Trait-space Monitoring for Emergent Misalignment During Supervised Finetuning","created":"2026-05-31","cited_by_count":0,"area":null},{"id":"2606.31591","title":"Evil Spectra: How Optimisers can Amplify or Suppress Emergent Misalignment","created":"2026-06-30","cited_by_count":0,"area":null},{"id":"2609.08418","title":"Feyospace-v1: How the Cyber Mercury Seven Trained Frontier Cyber Models","created":"2026-09-08","cited_by_count":0,"area":null},{"id":"2609.22153","title":"SafeTune: A Unified Faithful Library for Auditing and Repairing Safety Drift in Fine-Tuned LLMs","created":"2026-08-26","cited_by_count":0,"area":null},{"id":"2609.29960","title":"Beyond Average Safety: Chance-Constrained LLM Fine-tuning","created":"2026-09-24","cited_by_count":0,"area":null},{"id":"2609.35291","title":"Narrow Multimodal Fine-Tuning Can Induce Emergent Misalignment","created":"2026-09-28","cited_by_count":0,"area":null}],"next_cursor":null}