{"total":1154,"rows":[{"id":"2402.04249","title":"HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal","created":"2024-02-06","cited_by_count":1429,"area":"safety"},{"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":"2406.00515","title":"A Survey on Large Language Models for Code Generation","created":"2024-06-01","cited_by_count":658,"area":null},{"id":"2401.06373","title":"How Johnny Can Persuade LLMs to Jailbreak Them: Rethinking Persuasion to Challenge AI Safety by Humanizing LLMs","created":"2024-01-12","cited_by_count":573,"area":"lr"},{"id":"2403.03218","title":"The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning","created":"2024-03-05","cited_by_count":509,"area":"safety"},{"id":"2404.02151","title":"Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks","created":"2024-04-02","cited_by_count":486,"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":"2310.06987","title":"Catastrophic Jailbreak of Open-source LLMs via Exploiting Generation","created":"2023-10-10","cited_by_count":451,"area":"safety"},{"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":"2402.10260","title":"A StrongREJECT for Empty Jailbreaks","created":"2024-02-15","cited_by_count":385,"area":"safety"},{"id":"2506.04178","title":"OpenThoughts: Data Recipes for Reasoning Models","created":"2025-06-04","cited_by_count":304,"area":"lr"},{"id":"2401.10019","title":"R-Judge: Benchmarking Safety Risk Awareness for LLM Agents","created":"2024-01-18","cited_by_count":288,"area":"agents"},{"id":"2406.15513","title":"PKU-SafeRLHF: Towards Multi-Level Safety Alignment for LLMs with Human Preference","created":"2024-06-20","cited_by_count":261,"area":"alignment"},{"id":"2402.08787","title":"Rethinking Machine Unlearning for Large Language Models","created":"2024-02-13","cited_by_count":257,"area":"lr"},{"id":"2402.08983","title":"SafeDecoding: Defending against Jailbreak Attacks via Safety-Aware Decoding","created":"2024-02-14","cited_by_count":247,"area":"safety"},{"id":"2405.20947","title":"OR-Bench: An Over-Refusal Benchmark for Large Language Models","created":"2024-05-31","cited_by_count":244,"area":"alignment"},{"id":"2406.14598","title":"SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal","created":"2024-06-20","cited_by_count":240,"area":"alignment"},{"id":"2402.05162","title":"Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank Modifications","created":"2024-02-07","cited_by_count":231,"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":"2402.05044","title":"SALAD-Bench: A Hierarchical and Comprehensive Safety Benchmark for Large Language Models","created":"2024-02-07","cited_by_count":228,"area":null},{"id":"2402.11753","title":"ArtPrompt: ASCII Art-based Jailbreak Attacks against Aligned LLMs","created":"2024-02-19","cited_by_count":216,"area":"safety"},{"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":"2401.01967","title":"A Mechanistic Understanding of Alignment Algorithms: A Case Study on DPO and Toxicity","created":"2024-01-03","cited_by_count":185,"area":"alignment"},{"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":"2404.07921","title":"AmpleGCG: Learning a Universal and Transferable Generative Model of Adversarial Suffixes for Jailbreaking Both Open and Closed LLMs","created":"2024-04-11","cited_by_count":172,"area":"safety"},{"id":"2502.17424","title":"Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs","created":"2025-02-24","cited_by_count":172,"area":"safety"},{"id":"2311.05553","title":"Removing RLHF Protections in GPT-4 via Fine-Tuning","created":"2023-11-09","cited_by_count":171,"area":"alignment"},{"id":"2402.02207","title":"Safety Fine-Tuning at (Almost) No Cost: A Baseline for Vision Large Language Models","created":"2024-02-03","cited_by_count":167,"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":"2404.16873","title":"AdvPrompter: Fast Adaptive Adversarial Prompting for LLMs","created":"2024-04-21","cited_by_count":156,"area":"safety"},{"id":"2402.08567","title":"Agent Smith: A Single Image Can Jailbreak One Million Multimodal LLM Agents Exponentially Fast","created":"2024-02-13","cited_by_count":153,"area":"mllm"},{"id":"2402.01364","title":"Continual Learning for Large Language Models: A Survey","created":"2024-02-02","cited_by_count":147,"area":null},{"id":"2311.01011","title":"Tensor Trust: Interpretable Prompt Injection Attacks from an Online Game","created":"2023-11-02","cited_by_count":145,"area":"safety"},{"id":"2403.04783","title":"AutoDefense: Multi-Agent LLM Defense against Jailbreak Attacks","created":"2024-03-02","cited_by_count":145,"area":"agents"},{"id":"2408.00761","title":"Tamper-Resistant Safeguards for Open-Weight LLMs","created":"2024-08-01","cited_by_count":138,"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":"2404.08144","title":"LLM Agents can Autonomously Exploit One-day Vulnerabilities","created":"2024-04-11","cited_by_count":128,"area":"agents"},{"id":"2405.06823","title":"PLeak: Prompt Leaking Attacks against Large Language Model Applications","created":"2024-05-10","cited_by_count":127,"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":"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":"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":"2504.13203","title":"X-Teaming: Multi-Turn Jailbreaks and Defenses with Adaptive Multi-Agents","created":"2025-04-15","cited_by_count":125,"area":"agents"},{"id":"2406.09187","title":"GuardAgent: Safeguard LLM Agents by a Guard Agent via Knowledge-Enabled Reasoning","created":"2024-06-13","cited_by_count":124,"area":"lr"},{"id":"2409.18025","title":"An Adversarial Perspective on Machine Unlearning for AI Safety","created":"2024-09-26","cited_by_count":123,"area":"safety"},{"id":"2503.00555","title":"Safety Tax: Safety Alignment Makes Your Large Reasoning Models Less Reasonable","created":"2025-03-01","cited_by_count":122,"area":"lr"},{"id":"2407.18418","title":"Know Your Limits: A Survey of Abstention in Large Language Models","created":"2024-07-25","cited_by_count":121,"area":null},{"id":"2404.01099","title":"What is in Your Safe Data? Identifying Benign Data that Breaks Safety","created":"2024-04-01","cited_by_count":118,"area":"safety"},{"id":"2404.02831","title":"Empowering Biomedical Discovery with AI Agents","created":"2024-04-03","cited_by_count":118,"area":null},{"id":"2406.00045","title":"Personalized Steering of Large Language Models: Versatile Steering Vectors Through Bi-directional Preference Optimization","created":"2024-05-28","cited_by_count":118,"area":"alignment"},{"id":"2504.15585","title":"A Comprehensive Survey in LLM(-Agent) Full Stack Safety: Data, Training and Deployment","created":"2025-04-22","cited_by_count":116,"area":"lr"},{"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":"2402.01109","title":"Vaccine: Perturbation-aware Alignment for Large Language Models against Harmful Fine-tuning Attack","created":"2024-02-02","cited_by_count":115,"area":"alignment"},{"id":"2402.11746","title":"Language Models are Homer Simpson! Safety Re-Alignment of Fine-tuned Language Models through Task Arithmetic","created":"2024-02-19","cited_by_count":112,"area":"alignment"},{"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":"2402.06664","title":"LLM Agents can Autonomously Hack Websites","created":"2024-02-06","cited_by_count":109,"area":"agents"},{"id":"2401.17256","title":"Weak-to-Strong Jailbreaking on Large Language Models","created":"2024-01-30","cited_by_count":107,"area":"safety"},{"id":"2502.12893","title":"H-CoT: Hijacking the Chain-of-Thought Safety Reasoning Mechanism to Jailbreak Large Reasoning Models, Including OpenAI o1/o3, DeepSeek-R1, and Gemini 2.0 Flash Thinking","created":"2025-02-18","cited_by_count":99,"area":"lr"},{"id":"2504.02234","title":"LLM Social Simulations Are a Promising Research Method","created":"2025-04-03","cited_by_count":99,"area":null},{"id":"2402.13494","title":"GradSafe: Detecting Jailbreak Prompts for LLMs via Safety-Critical Gradient Analysis","created":"2024-02-21","cited_by_count":96,"area":"safety"},{"id":"2402.13669","title":"Self-Distillation Bridges Distribution Gap in Language Model Fine-Tuning","created":"2024-02-21","cited_by_count":96,"area":"distill"},{"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":"2405.14577","title":"Representation Noising: A Defence Mechanism Against Harmful Finetuning","created":"2024-05-23","cited_by_count":96,"area":"safety"},{"id":"2406.01637","title":"Teams of LLM Agents can Exploit Zero-Day Vulnerabilities","created":"2024-06-02","cited_by_count":95,"area":"agents"},{"id":"2510.18874","title":"Retaining by Doing: The Role of On-Policy Data in Mitigating Forgetting","created":"2025-10-21","cited_by_count":94,"area":null},{"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":"2402.09063","title":"Soft Prompt Threats: Attacking Safety Alignment and Unlearning in Open-Source LLMs through the Embedding Space","created":"2024-02-14","cited_by_count":91,"area":"alignment"},{"id":"2502.12659","title":"The Hidden Risks of Large Reasoning Models: A Safety Assessment of R1","created":"2025-02-18","cited_by_count":91,"area":"lr"},{"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":"2411.07781","title":"RedCode: Risky Code Execution and Generation Benchmark for Code Agents","created":"2024-11-12","cited_by_count":90,"area":"agents"},{"id":"2402.14020","title":"Coercing LLMs to do and reveal (almost) anything","created":"2024-02-21","cited_by_count":89,"area":"safety"},{"id":"2403.05030","title":"Defending Against Unforeseen Failure Modes with Latent Adversarial Training","created":"2024-03-08","cited_by_count":88,"area":"safety"},{"id":"2507.11878","title":"LLMs Encode Harmfulness and Refusal Separately","created":"2025-07-16","cited_by_count":88,"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":"2506.19823","title":"Persona Features Control Emergent Misalignment","created":"2025-06-24","cited_by_count":86,"area":null},{"id":"2406.12030","title":"SPA-VL: A Comprehensive Safety Preference Alignment Dataset for Vision Language Model","created":"2024-06-17","cited_by_count":85,"area":"mllm"},{"id":"2406.13356","title":"Unlearning or Obfuscating? Jogging the Memory of Unlearned LLMs via Benign Relearning","created":"2024-06-19","cited_by_count":84,"area":"safety"},{"id":"2401.11206","title":"InferAligner: Inference-Time Alignment for Harmlessness through Cross-Model Guidance","created":"2024-01-20","cited_by_count":83,"area":"alignment"},{"id":"2406.20053","title":"Covert Malicious Finetuning: Challenges in Safeguarding LLM Adaptation","created":"2024-06-28","cited_by_count":80,"area":null},{"id":"2410.02832","title":"FlipAttack: Jailbreak LLMs via Flipping","created":"2024-10-02","cited_by_count":80,"area":"safety"},{"id":"2312.15710","title":"Alleviating Hallucinations of Large Language Models through Induced Hallucinations","created":"2023-12-25","cited_by_count":79,"area":null},{"id":"2406.02622","title":"Safeguarding Large Language Models: A Survey","created":"2024-06-03","cited_by_count":78,"area":"safety"},{"id":"2404.05399","title":"SafetyPrompts: a Systematic Review of Open Datasets for Evaluating and Improving Large Language Model Safety","created":"2024-04-08","cited_by_count":77,"area":"safety"},{"id":"2311.08370","title":"SimpleSafetyTests: a Test Suite for Identifying Critical Safety Risks in Large Language Models","created":"2023-11-14","cited_by_count":76,"area":null},{"id":"2402.16717","title":"CodeChameleon: Personalized Encryption Framework for Jailbreaking Large Language Models","created":"2024-02-26","cited_by_count":76,"area":"safety"},{"id":"2403.07865","title":"CodeAttack: Revealing Safety Generalization Challenges of Large Language Models via Code Completion","created":"2024-03-12","cited_by_count":76,"area":"safety"},{"id":"2402.00357","title":"Safety of Multimodal Large Language Models on Images and Texts","created":"2024-02-01","cited_by_count":75,"area":"mllm"},{"id":"2403.08730","title":"Strengthening Multimodal Large Language Model with Bootstrapped Preference Optimization","created":"2024-03-13","cited_by_count":75,"area":"mllm"},{"id":"2407.17436","title":"AIR-Bench 2024: A Safety Benchmark Based on Risk Categories from Regulations and Policies","created":"2024-07-11","cited_by_count":74,"area":null},{"id":"2410.08827","title":"Do Unlearning Methods Remove Information from Language Model Weights?","created":"2024-10-11","cited_by_count":73,"area":"safety"},{"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":"2510.07192","title":"Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples","created":"2025-10-08","cited_by_count":72,"area":"safety"},{"id":"2402.16459","title":"Defending LLMs against Jailbreaking Attacks via Backtranslation","created":"2024-02-26","cited_by_count":70,"area":"safety"},{"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":"2406.18682","title":"The Multilingual Alignment Prism: Aligning Global and Local Preferences to Reduce Harm","created":"2024-06-26","cited_by_count":70,"area":"safety"},{"id":"2407.01599","title":"JailbreakZoo: Survey, Landscapes, and Horizons in Jailbreaking Large Language and Vision-Language Models","created":"2024-06-26","cited_by_count":70,"area":"mllm"},{"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":"2403.13031","title":"RigorLLM: Resilient Guardrails for Large Language Models against Undesired Content","created":"2024-03-19","cited_by_count":68,"area":"safety"},{"id":"2405.04760","title":"Large Language Models for Cyber Security: A Systematic Literature Review","created":"2024-05-08","cited_by_count":68,"area":null}],"next_cursor":"100"}