{"total":69,"rows":[{"id":"2408.15221","title":"LLM Defenses Are Not Robust to Multi-Turn Human Jailbreaks Yet","created":"2024-08-27","cited_by_count":151,"area":"safety"},{"id":"2410.10700","title":"LLMs know their vulnerabilities: Uncover Safety Gaps through Natural Distribution Shifts","created":"2024-10-14","cited_by_count":86,"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":"2502.11355","title":"Nuclear Deployed: Analyzing Catastrophic Risks in Decision-making of Autonomous LLM Agents","created":"2025-02-17","cited_by_count":27,"area":"lr"},{"id":"2602.02132","title":"There Is More to Refusal in Large Language Models than a Single Direction","created":"2026-02-02","cited_by_count":26,"area":"mechinterp"},{"id":"2506.14922","title":"FORTRESS: Frontier Risk Evaluation for National Security and Public Safety","created":"2025-06-17","cited_by_count":24,"area":"safety"},{"id":"2410.10343","title":"Locking Down the Finetuned LLMs Safety","created":"2024-10-14","cited_by_count":22,"area":null},{"id":"2506.09996","title":"From Judgment to Interference: Early Stopping LLM Harmful Outputs via Streaming Content Monitoring","created":"2025-06-11","cited_by_count":22,"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":"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":"2503.00187","title":"Steering Dialogue Dynamics for Robustness against Multi-turn Jailbreaking Attacks","created":"2025-02-28","cited_by_count":20,"area":"safety"},{"id":"2503.18991","title":"Inverse Reinforcement Learning with Dynamic Reward Scaling for LLM Alignment","created":"2025-03-23","cited_by_count":20,"area":"alignment"},{"id":"2505.16186","title":"SafeKey: Amplifying Aha-Moment Insights for Safety Reasoning","created":"2025-05-22","cited_by_count":20,"area":"lr"},{"id":"2410.01294","title":"Endless Jailbreaks with Bijection Learning","created":"2024-10-02","cited_by_count":19,"area":"safety"},{"id":"2602.11865","title":"Intelligent AI Delegation","created":"2026-02-12","cited_by_count":18,"area":null},{"id":"2410.02684","title":"HiddenGuard: Fine-Grained Safe Generation with Specialized Representation Router","created":"2024-10-03","cited_by_count":16,"area":"safety"},{"id":"2502.04040","title":"Safety Reasoning with Guidelines","created":"2025-02-06","cited_by_count":15,"area":"lr"},{"id":"2505.20259","title":"Lifelong Safety Alignment for Language Models","created":"2025-05-26","cited_by_count":15,"area":"alignment"},{"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":"2407.16637","title":"Course-Correction: Safety Alignment Using Synthetic Preferences","created":"2024-07-23","cited_by_count":13,"area":"alignment"},{"id":"2511.04962","title":"Too Good to be Bad: On the Failure of LLMs to Role-Play Villains","created":"2025-11-07","cited_by_count":11,"area":null},{"id":"2601.21900","title":"TraceRouter: Robust Safety for Large Foundation Models via Path-Level Intervention","created":"2026-01-29","cited_by_count":11,"area":"mechinterp"},{"id":"2502.09638","title":"Jailbreaking to Jailbreak","created":"2025-02-09","cited_by_count":9,"area":"safety"},{"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":"2506.01307","title":"Align is not Enough: Multimodal Universal Jailbreak Attack against Multimodal Large Language Models","created":"2025-06-02","cited_by_count":9,"area":"mllm"},{"id":"2410.12999","title":"POROver: Improving Safety and Reducing Overrefusal in Large Language Models with Overgeneration and Preference Optimization","created":"2024-10-16","cited_by_count":8,"area":"alignment"},{"id":"2411.12665","title":"Regression for the Mean: Auto-Evaluation and Inference with Few Labels through Post-hoc Regression","created":"2024-11-19","cited_by_count":7,"area":null},{"id":"2505.23856","title":"OMNIGUARD: An Efficient Approach for AI Safety Moderation Across Languages and Modalities","created":"2025-05-29","cited_by_count":7,"area":"safety"},{"id":"2601.03868","title":"What Matters For Safety Alignment?","created":"2026-01-07","cited_by_count":7,"area":"alignment"},{"id":"2501.13677","title":"HumorReject: Decoupling LLM Safety from Refusal Prefix via A Little Humor","created":"2025-01-23","cited_by_count":6,"area":"safety"},{"id":"2510.04320","title":"Read the Scene, Not the Script: Outcome-Aware Safety for LLMs","created":"2025-10-05","cited_by_count":6,"area":"alignment"},{"id":"2502.10450","title":"Trustworthy AI: Safety, Bias, and Privacy -- A Survey","created":"2025-02-11","cited_by_count":5,"area":"safety"},{"id":"2505.13862","title":"PandaGuard: Systematic Evaluation of LLM Safety against Jailbreaking Attacks","created":"2025-05-20","cited_by_count":4,"area":"alignment"},{"id":"2604.04842","title":"Do No Harm: Exposing Hidden Vulnerabilities of LLMs via Persona-based Client Simulation Attack in Psychological Counseling","created":"2026-04-06","cited_by_count":4,"area":"safety"},{"id":"2410.11437","title":"Difficult Task Yes but Simple Task No: Unveiling the Laziness in Multimodal LLMs","created":"2024-10-15","cited_by_count":3,"area":"mllm"},{"id":"2502.15873","title":"Practical Principles for AI Cost and Compute Accounting","created":"2025-02-21","cited_by_count":3,"area":null},{"id":"2602.08813","title":"Robust Policy Optimization to Prevent Catastrophic Forgetting","created":"2026-02-09","cited_by_count":3,"area":"alignment"},{"id":"2602.13547","title":"AISA: Awakening Intrinsic Safety Awareness in Large Language Models against Jailbreak Attacks","created":"2026-02-14","cited_by_count":3,"area":"safety"},{"id":"2605.14454","title":"LiSA: Lifelong Safety Adaptation via Conservative Policy Induction","created":"2026-05-14","cited_by_count":3,"area":"safety"},{"id":"2502.11681","title":"RIDE: Enhancing Large Language Model Alignment through Restyled In-Context Learning Demonstration Exemplars","created":"2025-02-17","cited_by_count":2,"area":"alignment"},{"id":"2505.14226","title":"Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs","created":"2025-05-20","cited_by_count":2,"area":null},{"id":"2506.05376","title":"A Red Teaming Roadmap Towards System-Level Safety","created":"2025-05-30","cited_by_count":2,"area":"safety"},{"id":"2508.16325","title":"ConceptGuard: Neuro-Symbolic Safety Guardrails via Sparse Interpretable Jailbreak Concepts","created":"2025-08-22","cited_by_count":2,"area":"safety"},{"id":"2510.03969","title":"How Catastrophic is Your LLM? Certifying Risk in Conversation","created":"2025-10-04","cited_by_count":2,"area":null},{"id":"2602.05164","title":"Position: Capability Control Should be a Separate Goal From Alignment","created":"2026-02-05","cited_by_count":2,"area":null},{"id":"2602.16835","title":"NeST: Neuron Selective Tuning for LLM Safety","created":"2026-02-18","cited_by_count":2,"area":"alignment"},{"id":"2503.11742","title":"Safe Vision-Language Models via Unsafe Weights Manipulation","created":"2025-03-14","cited_by_count":1,"area":"mllm"},{"id":"2508.08855","title":"BiasGym: A Simple and Generalizable Framework for Analyzing and Removing Biases through Injection","created":"2025-08-12","cited_by_count":1,"area":null},{"id":"2511.19009","title":"Understanding and Mitigating Over-refusal for Large Language Models via Representation Intervention","created":"2025-11-24","cited_by_count":1,"area":"safety"},{"id":"2512.21354","title":"Reflection-Driven Control for Trustworthy Code Agents","created":"2025-12-22","cited_by_count":1,"area":"agents"},{"id":"2603.22829","title":"Improving Safety Alignment via Balanced Direct Preference Optimization","created":"2026-03-24","cited_by_count":1,"area":"alignment"},{"id":"2604.04992","title":"FreakOut-LLM: The Effect of Emotional Stimuli on Safety Alignment","created":"2026-04-05","cited_by_count":1,"area":"alignment"},{"id":"2604.17843","title":"Learning from AVA: Early Lessons from a Curated and Trustworthy Generative AI for Policy and Development Research","created":"2026-04-20","cited_by_count":1,"area":null},{"id":"2605.24154","title":"Palette: A Modular, Controllable, and Efficient Framework for On-demand Authorized Safety Alignment Relaxation in LLMs","created":"2026-05-22","cited_by_count":1,"area":"alignment"},{"id":"2608.03201","title":"When Refusal Looks Safe: The Refusal-Cue Shortcut in Safety Guard Models","created":"2026-08-04","cited_by_count":1,"area":"safety"},{"id":"2608.14577","title":"HarmProfile: Characterizing Harmful Distributions in Frontier LLMs","created":"2026-06-11","cited_by_count":1,"area":"safety"},{"id":"2608.30856","title":"You Shouldn't Have Asked: A Pragmatics-Inspired Taxonomy for Evaluating LLM Refusals","created":"2026-08-31","cited_by_count":1,"area":null},{"id":"2505.18672","title":"Concept Concentration for Faithful Representation Intervention","created":"2025-05-24","cited_by_count":0,"area":"safety"},{"id":"2601.09721","title":"Safety boundary maintenance in consumer AI systems responding to pediatric health queries: a cross-platform benchmark evaluation under naturalistic and adversarially pressured conditions","created":"2025-12-26","cited_by_count":0,"area":null},{"id":"2605.01749","title":"Only Say What You Know: Calibration-Aware Generation for Long-Form Factuality","created":"2026-05-03","cited_by_count":0,"area":"lr"},{"id":"2605.07883","title":"Beyond \"I cannot fulfill this request\": Alleviating Rigid Rejection in LLMs via Label Enhancement","created":"2026-05-08","cited_by_count":0,"area":"alignment"},{"id":"2605.20286","title":"Adaptive Probe-based Steering for Robust LLM Jailbreaking","created":"2026-05-19","cited_by_count":0,"area":"safety"},{"id":"2606.00686","title":"Dialectics of Alignment: Harnessing Unsafe Knowledge for Dynamic Safety Routing","created":"2026-05-30","cited_by_count":0,"area":"moe"},{"id":"2607.16242","title":"TRACE: Trajectory-Based Safety Patch Learning for LLM Post-Training Realignment","created":"2026-06-26","cited_by_count":0,"area":"alignment"},{"id":"2608.13304","title":"Refusing Intent, Not Form: Wrapper-Based Intent-Group Supervision for LLM Safety","created":"2026-08-13","cited_by_count":0,"area":"safety"},{"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":"2610.03938","title":"MLLMs Fail to Refuse when Using Tools Agentically","created":"2026-10-02","cited_by_count":0,"area":"lr"},{"id":"2610.11005","title":"How Narrative Wrapping Affects LLM Refusal: A Cross-Language Benchmark and Defense","created":"2026-10-07","cited_by_count":0,"area":"alignment"}],"next_cursor":null}