{"total":118,"rows":[{"id":"2501.14249","title":"Humanity's Last Exam","created":"2025-01-24","cited_by_count":491,"area":null},{"id":"2406.07358","title":"AI Sandbagging: Language Models can Strategically Underperform on Evaluations","created":"2024-06-11","cited_by_count":145,"area":null},{"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":"2405.03644","title":"When LLMs Meet Cybersecurity: A Systematic Literature Review","created":"2024-05-06","cited_by_count":71,"area":null},{"id":"2407.21792","title":"Safetywashing: Do AI Safety Benchmarks Actually Measure Safety Progress?","created":"2024-07-31","cited_by_count":61,"area":null},{"id":"2410.21514","title":"Sabotage Evaluations for Frontier Models","created":"2024-10-28","cited_by_count":52,"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":"2412.08686","title":"LatentQA: Teaching LLMs to Decode Activations Into Natural Language","created":"2024-12-11","cited_by_count":51,"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":"2406.18664","title":"Evaluating Copyright Takedown Methods for Language Models","created":"2024-06-26","cited_by_count":43,"area":null},{"id":"2504.16980","title":"Safety Pretraining: Toward the Next Generation of Safe AI","created":"2025-04-23","cited_by_count":43,"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":"2410.04332","title":"Gradient Routing: Masking Gradients to Localize Computation in Neural Networks","created":"2024-10-06","cited_by_count":34,"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":"2406.14595","title":"Adversaries Can Misuse Combinations of Safe Models","created":"2024-06-20","cited_by_count":32,"area":null},{"id":"2504.16137","title":"Virology Capabilities Test (VCT): A Multimodal Virology Q&A Benchmark","created":"2025-04-21","cited_by_count":31,"area":null},{"id":"2405.10632","title":"Towards interactive evaluations for interaction harms in human-AI systems","created":"2024-05-17","cited_by_count":29,"area":null},{"id":"2503.10728","title":"DarkBench: Benchmarking Dark Patterns in Large Language Models","created":"2025-03-13","cited_by_count":28,"area":null},{"id":"2406.19354","title":"Fundamental Problems With Model Editing: How Should Rational Belief Revision Work in LLMs?","created":"2024-06-27","cited_by_count":24,"area":null},{"id":"2412.01784","title":"Noise Injection Reveals Hidden Capabilities of Sandbagging Language Models","created":"2024-12-02","cited_by_count":24,"area":null},{"id":"2505.18893","title":"Reality Check: A New Evaluation Ecosystem Is Necessary to Understand AI's Real World Effects","created":"2025-05-24","cited_by_count":19,"area":null},{"id":"2507.12691","title":"Benchmarking Deception Probes via Black-to-White Performance Boosts","created":"2025-07-16","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":"2502.02180","title":"The Elicitation Game: Evaluating Capability Elicitation Techniques","created":"2025-02-04","cited_by_count":18,"area":null},{"id":"2412.12480","title":"Subversion Strategy Eval: Can language models statelessly strategize to subvert control protocols?","created":"2024-12-17","cited_by_count":15,"area":null},{"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":"2410.09247","title":"Benchmark Inflation: Revealing LLM Performance Gaps Using Retro-Holdouts","created":"2024-10-11","cited_by_count":14,"area":null},{"id":"2410.18819","title":"From Imitation to Introspection: Probing Self-Consciousness in Language Models","created":"2024-10-24","cited_by_count":14,"area":null},{"id":"2604.11806","title":"Detecting Safety Violations Across Many Agent Traces","created":"2026-04-13","cited_by_count":14,"area":null},{"id":"2511.22662","title":"Difficulties with Evaluating a Deception Detector for AIs","created":"2025-11-27","cited_by_count":13,"area":null},{"id":"2601.21571","title":"Shaping capabilities with token-level data filtering","created":"2026-01-29","cited_by_count":13,"area":null},{"id":"2502.15797","title":"OCCULT: Evaluating Large Language Models for Offensive Cyber Operation Capabilities","created":"2025-02-18","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":"2501.11739","title":"Episodic memory in AI agents poses risks that should be studied and mitigated","created":"2025-01-20","cited_by_count":11,"area":null},{"id":"2405.15523","title":"The Mosaic Memory of Large Language Models","created":"2024-05-24","cited_by_count":10,"area":null},{"id":"2407.13059","title":"Prioritizing High-Consequence Biological Capabilities in Evaluations of Artificial Intelligence Models","created":"2024-05-25","cited_by_count":10,"area":null},{"id":"2506.20702","title":"The Singapore Consensus on Global AI Safety Research Priorities","created":"2025-06-25","cited_by_count":10,"area":null},{"id":"2505.06108","title":"LLMs Outperform Experts on Challenging Biology Benchmarks","created":"2025-05-09","cited_by_count":9,"area":null},{"id":"2511.17408","title":"The Impact of Off-Policy Training Data on Probe Generalisation","created":"2025-11-21","cited_by_count":9,"area":null},{"id":"2506.06414","title":"Benchmarking Misuse Mitigation Against Covert Adversaries","created":"2025-06-06","cited_by_count":8,"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":"2410.06704","title":"PII-Scope: A Comprehensive Study on Training Data PII Extraction Attacks in LLMs","created":"2024-10-09","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":"2505.22860","title":"Permissioned LLMs: Enforcing Access Control in Large Language Models","created":"2025-05-28","cited_by_count":7,"area":null},{"id":"2406.18326","title":"PaCoST: Paired Confidence Significance Testing for Benchmark Contamination Detection in Large Language Models","created":"2024-06-26","cited_by_count":6,"area":null},{"id":"2602.16703","title":"Measuring Mid-2025 LLM-Assistance on Novice Performance in Biology","created":"2026-02-18","cited_by_count":6,"area":null},{"id":"2503.15815","title":"Attention Pruning: Automated Fairness Repair of Language Models via Surrogate Simulated Annealing","created":"2025-03-20","cited_by_count":5,"area":null},{"id":"2505.16789","title":"Accidental Vulnerability: Factors in Fine-Tuning that Shift Model Safeguards","created":"2025-05-22","cited_by_count":5,"area":null},{"id":"2408.04983","title":"Get Confused Cautiously: Textual Sequence Memorization Erasure with Selective Entropy Maximization","created":"2024-08-09","cited_by_count":4,"area":null},{"id":"2409.02228","title":"Unforgettable Generalization in Language Models","created":"2024-09-03","cited_by_count":4,"area":null},{"id":"2409.16125","title":"Analyzing Probabilistic Methods for Evaluating Agent Capabilities","created":"2024-09-24","cited_by_count":4,"area":null},{"id":"2505.15811","title":"On the creation of narrow AI: hierarchy and nonlocality of neural network skills","created":"2025-05-21","cited_by_count":4,"area":null},{"id":"2505.18003","title":"An Example Safety Case for Safeguards Against Misuse","created":"2025-05-23","cited_by_count":4,"area":null},{"id":"2509.16332","title":"Psychometric Personality Shaping Modulates Capabilities and Safety in Language Models","created":"2025-09-19","cited_by_count":4,"area":null},{"id":"2503.17388","title":"AI Companies Should Report Pre- and Post-Mitigation Safety Evaluations","created":"2025-03-17","cited_by_count":3,"area":null},{"id":"2509.16660","title":"Redefining Experts: Interpretable Decomposition of Language Models for Toxicity Mitigation","created":"2025-09-20","cited_by_count":3,"area":null},{"id":"2602.14135","title":"ForesightSafety Bench: A Frontier Risk Evaluation and Governance Framework towards Safe AI","created":"2026-02-15","cited_by_count":3,"area":null},{"id":"2604.20995","title":"Value-Conflict Diagnostics Reveal Widespread Alignment Faking in Language Models","created":"2026-04-22","cited_by_count":3,"area":null},{"id":"2606.11660","title":"Bergson: An Open Source Library for Data Attribution","created":"2026-06-10","cited_by_count":3,"area":null},{"id":"2607.08077","title":"Modular Pretraining Enables Access Control","created":"2026-07-09","cited_by_count":3,"area":null},{"id":"2607.27539","title":"Can an AI Assistant Really Forget? Auditable Deletion from Addressable Memory","created":"2026-07-30","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":"2412.00789","title":"A Cognac Shot To Forget Bad Memories: Corrective Unlearning for Graph Neural Networks","created":"2024-12-01","cited_by_count":2,"area":null},{"id":"2501.06137","title":"Supervision policies can shape long-term risk management in general-purpose AI models","created":"2025-01-10","cited_by_count":2,"area":null},{"id":"2502.05242","title":"Beyond External Monitors: Enhancing Transparency of Large Language Models for Easier Monitoring","created":"2025-02-07","cited_by_count":2,"area":null},{"id":"2504.14039","title":"MEQA: A Meta-Evaluation Framework for Question & Answer LLM Benchmarks","created":"2025-04-18","cited_by_count":2,"area":null},{"id":"2602.00513","title":"Minerva: Reinforcement Learning with Verifiable Rewards for Cyber Threat Intelligence LLMs","created":"2026-01-31","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":"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":"2606.01189","title":"The Case for Model Science: Verify, Explore, Steer, Refine","created":"2026-05-31","cited_by_count":2,"area":null},{"id":"2608.02684","title":"A Blind Spot in Alignment: Quantifying Biosecurity Risks in Large Language Models","created":"2026-08-03","cited_by_count":2,"area":null},{"id":"2408.01449","title":"AI Act for the Working Programmer","created":"2024-07-23","cited_by_count":1,"area":null},{"id":"2503.15182","title":"Foundation models may exhibit staged progression in novel CBRN threat disclosure","created":"2025-03-19","cited_by_count":1,"area":null},{"id":"2503.23424","title":"What Makes an Evaluation Useful? Common Pitfalls and Best Practices","created":"2025-03-30","cited_by_count":1,"area":null},{"id":"2504.05059","title":"MIAT: Maneuver-Intention-Aware Transformer for Spatio-Temporal Trajectory Prediction","created":"2025-04-07","cited_by_count":1,"area":null},{"id":"2510.27629","title":"Best Practices for Biorisk Evaluations on Open-Weight Bio-Foundation Models","created":"2025-10-31","cited_by_count":1,"area":null},{"id":"2511.18298","title":"Cross-Disciplinary Knowledge Retrieval and Synthesis: A Compound AI Architecture for Scientific Discovery","created":"2025-11-23","cited_by_count":1,"area":null},{"id":"2601.23157","title":"No More, No Less: Least-Privilege Language Models","created":"2026-01-30","cited_by_count":1,"area":null},{"id":"2603.18019","title":"BenchBrowser: Retrieving Evidence for Evaluating Benchmark Validity","created":"2026-02-25","cited_by_count":1,"area":null},{"id":"2604.20389","title":"CyberCertBench: Evaluating LLMs in Cybersecurity Certification Knowledge","created":"2026-04-22","cited_by_count":1,"area":null},{"id":"2606.12429","title":"Muse Spark Safety & Preparedness Report","created":"2026-05-14","cited_by_count":1,"area":null},{"id":"2606.18936","title":"SciRisk-Bench: A Risk-Dimension-Aware Benchmark for AI4Science Safety","created":"2026-06-17","cited_by_count":1,"area":null},{"id":"2606.28643","title":"Obliviate: Erasing Concepts from Autoregressive Image Generation Models","created":"2026-06-26","cited_by_count":1,"area":null},{"id":"2607.13039","title":"Safeguard-Conditioned Uplift: Measuring Utility-Risk Frontiers for Dual-Use Biology Assistants","created":"2026-06-12","cited_by_count":1,"area":null},{"id":"2607.18665","title":"SciHazard: A Benchmark for Measuring Scientific Safety Risks with Decomposed Harm Scoring","created":"2026-07-21","cited_by_count":1,"area":null},{"id":"2607.27951","title":"Safeguards Based on Copyable Context Cannot Provide Reliable Safety for LLMs","created":"2026-07-30","cited_by_count":1,"area":null},{"id":"2608.29458","title":"Reference-Grafting Matches Fine-Tuning at Eliciting Sandbagged Capabilities","created":"2026-08-29","cited_by_count":1,"area":null},{"id":"2412.10257","title":"Targeted Angular Reversal of Weights (TARS) for Knowledge Removal in Large Language Models","created":"2024-12-13","cited_by_count":0,"area":null},{"id":"2505.05064","title":"WaterDrum: Watermarking for Data-centric Unlearning Metric","created":"2025-05-08","cited_by_count":0,"area":null},{"id":"2509.19485","title":"Identifying and Addressing User-level Security Concerns in Smart Homes Using \"Smaller\" LLMs","created":"2025-09-23","cited_by_count":0,"area":null},{"id":"2510.13163","title":"A Matter of Representation: Towards Graph-Based Abstract Code Generation","created":"2025-10-15","cited_by_count":0,"area":null},{"id":"2510.21184","title":"Reducing the Probability of Undesirable Outputs in Language Models Using Probabilistic Inference","created":"2025-10-24","cited_by_count":0,"area":null},{"id":"2512.12469","title":"Sparse Concept Anchoring for Interpretable and Controllable Neural Representations","created":"2025-12-13","cited_by_count":0,"area":null},{"id":"2603.14672","title":"Seamless Deception: Larger Language Models Are Better Knowledge Concealers","created":"2026-03-15","cited_by_count":0,"area":null},{"id":"2603.17884","title":"DebugLM: Learning Traceable Training Data Provenance for LLMs","created":"2026-03-18","cited_by_count":0,"area":null},{"id":"2603.22869","title":"Chain-of-Authorization: Embedding authorization into large language models","created":"2026-03-24","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":"2604.05872","title":"Swiss-Bench 003: Evaluating LLM Reliability and Adversarial Security for Swiss Regulatory Contexts","created":"2026-04-07","cited_by_count":0,"area":null},{"id":"2604.24966","title":"Risk Reporting for Developers' Internal AI Model Use","created":"2026-04-27","cited_by_count":0,"area":null},{"id":"2605.10575","title":"Acceptance Cards:A Four-Diagnostic Standard for Safe Fine-Tuning Defense Claims","created":"2026-05-11","cited_by_count":0,"area":null}],"next_cursor":"100"}