{"total":486,"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":"2410.09024","title":"AgentHarm: A Benchmark for Measuring Harmfulness of LLM Agents","created":"2024-10-11","cited_by_count":437,"area":"agents"},{"id":"2406.04313","title":"Improving Alignment and Robustness with Circuit Breakers","created":"2024-06-06","cited_by_count":322,"area":"safety"},{"id":"2412.16339","title":"Deliberative Alignment: Reasoning Enables Safer Language Models","created":"2024-12-20","cited_by_count":308,"area":"lr"},{"id":"2406.14598","title":"SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal","created":"2024-06-20","cited_by_count":240,"area":"alignment"},{"id":"2407.04295","title":"Jailbreak Attacks and Defenses Against Large Language Models: A Survey","created":"2024-07-05","cited_by_count":230,"area":"safety"},{"id":"2506.23918","title":"Thinking with Images for Multimodal Reasoning: Foundations, Methods, and Future Frontiers","created":"2025-06-30","cited_by_count":199,"area":"lr"},{"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":"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":"2510.09023","title":"The Attacker Moves Second: Stronger Adaptive Attacks Bypass Defenses Against Llm Jailbreaks and Prompt Injections","created":"2025-10-10","cited_by_count":129,"area":"safety"},{"id":"2503.00061","title":"Adaptive Attacks Break Defenses Against Indirect Prompt Injection Attacks on LLM Agents","created":"2025-02-27","cited_by_count":125,"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":"2405.15589","title":"Efficient Adversarial Training in LLMs with Continuous Attacks","created":"2024-05-24","cited_by_count":113,"area":"safety"},{"id":"2506.14866","title":"OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents","created":"2025-06-17","cited_by_count":111,"area":"agents"},{"id":"2405.21018","title":"Improved Techniques for Optimization-Based Jailbreaking on Large Language Models","created":"2024-05-31","cited_by_count":110,"area":"safety"},{"id":"2406.05870","title":"Machine Against the RAG: Jamming Retrieval-Augmented Generation with Blocker Documents","created":"2024-06-09","cited_by_count":102,"area":"rag"},{"id":"2502.02542","title":"OverThink: Slowdown Attacks on Reasoning LLMs","created":"2025-02-04","cited_by_count":96,"area":"lr"},{"id":"2407.11969","title":"Does Refusal Training in LLMs Generalize to the Past Tense?","created":"2024-07-16","cited_by_count":79,"area":"safety"},{"id":"2412.03556","title":"Best-of-N Jailbreaking","created":"2024-12-04","cited_by_count":76,"area":"safety"},{"id":"2406.00799","title":"Get my drift? Catching LLM Task Drift with Activation Deltas","created":"2024-06-02","cited_by_count":72,"area":"rag"},{"id":"2406.01288","title":"Improved Few-Shot Jailbreaking Can Circumvent Aligned Language Models and Their Defenses","created":"2024-06-03","cited_by_count":70,"area":"safety"},{"id":"2410.09102","title":"Instructional Segment Embedding: Improving LLM Safety with Instruction Hierarchy","created":"2024-10-09","cited_by_count":67,"area":"safety"},{"id":"2410.13886","title":"Refusal-Trained LLMs Are Easily Jailbroken As Browser Agents","created":"2024-10-11","cited_by_count":65,"area":"agents"},{"id":"2409.20089","title":"Robust LLM safeguarding via refusal feature adversarial training","created":"2024-09-30","cited_by_count":64,"area":null},{"id":"2406.05498","title":"SelfDefend: LLMs Can Defend Themselves against Jailbreaking in a Practical Manner","created":"2024-06-08","cited_by_count":63,"area":"safety"},{"id":"2404.12241","title":"Introducing v0.5 of the AI Safety Benchmark from MLCommons","created":"2024-04-18","cited_by_count":62,"area":null},{"id":"2501.18841","title":"Trading Inference-Time Compute for Adversarial Robustness","created":"2025-01-31","cited_by_count":61,"area":"lr"},{"id":"2410.06625","title":"ETA: Evaluating Then Aligning Safety of Vision Language Models at Inference Time","created":"2024-10-09","cited_by_count":60,"area":"mllm"},{"id":"2404.16369","title":"Don't Say No: Jailbreaking LLM by Suppressing Refusal","created":"2024-04-25","cited_by_count":56,"area":"alignment"},{"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":"2503.12188","title":"Multi-Agent Systems Execute Arbitrary Malicious Code","created":"2025-03-15","cited_by_count":53,"area":"agents"},{"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":"2508.09224","title":"From Hard Refusals to Safe-Completions: Toward Output-Centric Safety Training","created":"2025-08-12","cited_by_count":52,"area":null},{"id":"2406.09324","title":"Bag of Tricks: Benchmarking of Jailbreak Attacks on LLMs","created":"2024-06-13","cited_by_count":51,"area":"safety"},{"id":"2410.17520","title":"MobileSafetyBench: Evaluating Safety of Autonomous Agents in Mobile Device Control","created":"2024-10-23","cited_by_count":51,"area":"agents"},{"id":"2407.20516","title":"Machine Unlearning in Generative AI: A Survey","created":"2024-07-30","cited_by_count":49,"area":"safety"},{"id":"2502.08586","title":"Commercial LLM Agents Are Already Vulnerable to Simple Yet Dangerous Attacks","created":"2025-02-12","cited_by_count":49,"area":"agents"},{"id":"2506.07022","title":"AlphaSteer: Learning Refusal Steering with Principled Null-Space Constraint","created":"2025-06-08","cited_by_count":49,"area":"safety"},{"id":"2410.12855","title":"JAILJUDGE: A Comprehensive Jailbreak Judge Benchmark with Multi-Agent Enhanced Explanation Evaluation Framework","created":"2024-10-11","cited_by_count":47,"area":"agents"},{"id":"2406.15518","title":"Steering Without Side Effects: Improving Post-Deployment Control of Language Models","created":"2024-06-21","cited_by_count":46,"area":"safety"},{"id":"2407.14937","title":"Operationalizing a Threat Model for Red-Teaming Large Language Models (LLMs)","created":"2024-07-20","cited_by_count":46,"area":"safety"},{"id":"2410.01606","title":"Automated Red Teaming with GOAT: the Generative Offensive Agent Tester","created":"2024-10-02","cited_by_count":45,"area":"agents"},{"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":"2405.19358","title":"Robustifying Safety-Aligned Large Language Models through Clean Data Curation","created":"2024-05-24","cited_by_count":44,"area":"alignment"},{"id":"2411.18688","title":"Immune: Improving Safety Against Jailbreaks in Multi-modal LLMs via Inference-Time Alignment","created":"2024-11-27","cited_by_count":44,"area":"lr"},{"id":"2503.20823","title":"Playing the Fool: Jailbreaking LLMs and Multimodal LLMs with Out-of-Distribution Strategy","created":"2025-03-26","cited_by_count":44,"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":"2502.05209","title":"Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilities","created":"2025-02-03","cited_by_count":39,"area":null},{"id":"2503.05021","title":"Safety is Not Only About Refusal: Reasoning-Enhanced Fine-tuning for Interpretable LLM Safety","created":"2025-03-06","cited_by_count":39,"area":"lr"},{"id":"2405.20099","title":"Defensive Prompt Patch: A Robust and Interpretable Defense of LLMs against Jailbreak Attacks","created":"2024-05-30","cited_by_count":37,"area":"safety"},{"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":"2410.15236","title":"Jailbreaking and Mitigation of Vulnerabilities in Large Language Models","created":"2024-10-20","cited_by_count":34,"area":"safety"},{"id":"2507.11097","title":"The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs","created":"2025-07-15","cited_by_count":34,"area":"inference"},{"id":"2512.23738","title":"Enforcing Temporal Constraints for LLM Agents","created":"2025-12-25","cited_by_count":34,"area":"agents"},{"id":"2406.17692","title":"From Distributional to Overton Pluralism: Investigating Large Language Model Alignment","created":"2024-06-25","cited_by_count":33,"area":"alignment"},{"id":"2407.18369","title":"AI Safety in Generative AI Large Language Models: A Survey","created":"2024-07-06","cited_by_count":33,"area":null},{"id":"2410.09040","title":"AttnGCG: Enhancing Jailbreaking Attacks on LLMs with Attention Manipulation","created":"2024-10-11","cited_by_count":33,"area":"safety"},{"id":"2409.14586","title":"Backtracking Improves Generation Safety","created":"2024-09-22","cited_by_count":32,"area":"alignment"},{"id":"2406.09321","title":"JailbreakEval: An Integrated Toolkit for Evaluating Jailbreak Attempts Against Large Language Models","created":"2024-06-13","cited_by_count":31,"area":"safety"},{"id":"2406.12935","title":"ChatBug: A Common Vulnerability of Aligned LLMs Induced by Chat Templates","created":"2024-06-17","cited_by_count":30,"area":"alignment"},{"id":"2504.01550","title":"Representation Bending for Large Language Model Safety","created":"2025-04-02","cited_by_count":30,"area":"safety"},{"id":"2405.14023","title":"WordGame: Efficient & Effective LLM Jailbreak via Simultaneous Obfuscation in Query and Response","created":"2024-05-22","cited_by_count":29,"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":"2406.11614","title":"Intrinsic Test of Unlearning Using Parametric Knowledge Traces","created":"2024-06-17","cited_by_count":29,"area":"safety"},{"id":"2407.02855","title":"From Theft to Bomb-Making: The Ripple Effect of Unlearning in Defending Against Jailbreak Attacks","created":"2024-07-03","cited_by_count":29,"area":"safety"},{"id":"2502.01633","title":"Adversarial Reasoning at Jailbreaking Time","created":"2025-02-03","cited_by_count":29,"area":"lr"},{"id":"2504.10694","title":"The Jailbreak Tax: How Useful are Your Jailbreak Outputs?","created":"2025-04-14","cited_by_count":28,"area":"safety"},{"id":"2603.11088","title":"The Attack and Defense Landscape of Agentic AI: A Comprehensive Survey","created":"2026-03-11","cited_by_count":28,"area":"agents"},{"id":"2502.12964","title":"Trust Me, I'm Wrong: LLMs Hallucinate with Certainty Despite Knowing the Answer","created":"2025-02-18","cited_by_count":27,"area":null},{"id":"2406.06622","title":"Adversarial Tuning: Defending Against Jailbreak Attacks for LLMs","created":"2024-06-07","cited_by_count":26,"area":"safety"},{"id":"2506.11094","title":"The Scales of Justitia: A Comprehensive Survey on Safety Evaluation of LLMs","created":"2025-06-06","cited_by_count":26,"area":null},{"id":"2405.13077","title":"GPT-4 Jailbreaks Itself with Near-Perfect Success Using Self-Explanation","created":"2024-05-21","cited_by_count":25,"area":"safety"},{"id":"2412.10321","title":"AdvPrefix: An Objective for Nuanced LLM Jailbreaks","created":"2024-12-13","cited_by_count":25,"area":"safety"},{"id":"2412.18693","title":"Diverse and Effective Red Teaming with Auto-generated Rewards and Multi-step Reinforcement Learning","created":"2024-12-24","cited_by_count":25,"area":"safety"},{"id":"2504.18041","title":"RAG LLMs are Not Safer: A Safety Analysis of Retrieval-Augmented Generation for Large Language Models","created":"2025-04-25","cited_by_count":24,"area":"rag"},{"id":"2505.14667","title":"SAFEPATH: Preventing Harmful Reasoning in Chain-of-Thought via Early Alignment","created":"2025-05-20","cited_by_count":24,"area":"lr"},{"id":"2507.20526","title":"Security Challenges in AI Agent Deployment: Insights from a Large Scale Public Competition","created":"2025-07-28","cited_by_count":24,"area":"lr"},{"id":"2410.02298","title":"Jailbreak Antidote: Runtime Safety-Utility Balance via Sparse Representation Adjustment in Large Language Models","created":"2024-10-03","cited_by_count":23,"area":"safety"},{"id":"2411.01077","title":"Emoji Attack: Enhancing Jailbreak Attacks Against Judge LLM Detection","created":"2024-11-01","cited_by_count":23,"area":"safety"},{"id":"2502.01236","title":"Eliciting Language Model Behaviors with Investigator Agents","created":"2025-02-03","cited_by_count":23,"area":"agents"},{"id":"2412.04653","title":"Hidden in the Noise: Two-Stage Robust Watermarking for Images","created":"2024-12-05","cited_by_count":21,"area":null},{"id":"2503.01908","title":"UDora: A Unified Red Teaming Framework against LLM Agents by Dynamically Hijacking Their Own Reasoning","created":"2025-02-28","cited_by_count":21,"area":"lr"},{"id":"2506.10597","title":"SoK: Evaluating Jailbreak Guardrails for Large Language Models","created":"2025-06-12","cited_by_count":21,"area":"alignment"},{"id":"2406.07954","title":"Dataset and Lessons Learned from the 2024 SaTML LLM Capture-the-Flag Competition","created":"2024-06-12","cited_by_count":20,"area":null},{"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":"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":"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":"2411.06426","title":"SequentialBreak: Large Language Models Can be Fooled by Embedding Jailbreak Prompts into Sequential Prompt Chains","created":"2024-11-10","cited_by_count":19,"area":"safety"},{"id":"2506.00085","title":"COSMIC: Generalized Refusal Direction Identification in LLM Activations","created":"2025-05-30","cited_by_count":19,"area":null},{"id":"2506.07596","title":"TwinBreak: Jailbreaking LLM Security Alignments based on Twin Prompts","created":"2025-06-09","cited_by_count":19,"area":"safety"},{"id":"2405.13068","title":"Uncovering Logit Suppression Vulnerabilities in LLM Safety Alignment","created":"2024-05-20","cited_by_count":18,"area":"alignment"},{"id":"2412.06370","title":"Exploring Memorization and Copyright Violation in Frontier LLMs: A Study of the New York Times v. OpenAI 2023 Lawsuit","created":"2024-12-09","cited_by_count":18,"area":null},{"id":"2502.02180","title":"The Elicitation Game: Evaluating Capability Elicitation Techniques","created":"2025-02-04","cited_by_count":18,"area":null},{"id":"2503.02574","title":"LLM-Safety Evaluations Lack Robustness","created":"2025-03-04","cited_by_count":18,"area":"alignment"},{"id":"2505.14103","title":"AudioJailbreak: Jailbreak Attacks against End-to-End Large Audio-Language Models","created":"2025-05-20","cited_by_count":18,"area":"safety"},{"id":"2507.19060","title":"PurpCode: Reasoning for Safer Code Generation","created":"2025-07-25","cited_by_count":18,"area":"lr"},{"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"}],"next_cursor":"100"}