{"total":86,"rows":[{"id":"2603.05706","title":"Reasoning Models Struggle to Control their Chains of Thought","created":"2026-03-05","cited_by_count":25,"area":"lr"},{"id":"2510.16380","title":"MoReBench: Evaluating Procedural and Pluralistic Moral Reasoning in Language Models, More than Outcomes","created":"2025-10-18","cited_by_count":23,"area":"lr"},{"id":"2510.19851","title":"Can Reasoning Models Obfuscate Reasoning? Stress-Testing Chain-of-Thought Monitorability","created":"2025-10-21","cited_by_count":22,"area":"lr"},{"id":"2512.08093","title":"Training LLMs for Honesty via Confessions","created":"2025-12-08","cited_by_count":22,"area":null},{"id":"2605.02087","title":"Model Spec Midtraining: Improving How Alignment Training Generalizes","created":"2026-05-03","cited_by_count":20,"area":null},{"id":"2602.15198","title":"Colosseum: Auditing Collusion in Cooperative Multi-Agent Systems","created":"2026-02-16","cited_by_count":18,"area":"agents"},{"id":"2602.08235","title":"When Benign Inputs Lead to Severe Harms: Eliciting Unsafe Unintended Behaviors of Computer-Use Agents","created":"2026-02-09","cited_by_count":14,"area":"agents"},{"id":"2603.01608","title":"Evaluating and Understanding Scheming Propensity in LLM Agents","created":"2026-03-02","cited_by_count":14,"area":"agents"},{"id":"2510.23966","title":"A Pragmatic Way to Measure Chain-of-Thought Monitorability","created":"2025-10-28","cited_by_count":13,"area":"lr"},{"id":"2606.07157","title":"Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models","created":"2026-06-05","cited_by_count":12,"area":"lr"},{"id":"2605.08715","title":"AgentForesight: Online Auditing for Early Failure Prediction in Multi-Agent Systems","created":"2026-05-09","cited_by_count":11,"area":"agents"},{"id":"2510.27338","title":"Reasoning Models Sometimes Output Illegible Chains of Thought","created":"2025-10-31","cited_by_count":9,"area":"lr"},{"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":"2512.00218","title":"Reasoning Under Pressure: How do Training Incentives Influence Chain-of-Thought Monitorability?","created":"2025-11-28","cited_by_count":9,"area":"lr"},{"id":"2602.01425","title":"One Probe Won't Catch Them All: Towards Targeted Deception Detection","created":"2026-02-01","cited_by_count":9,"area":"mechinterp"},{"id":"2602.13477","title":"OMNI-LEAK: Orchestrator Multi-Agent Network Induced Data Leakage","created":"2026-02-13","cited_by_count":9,"area":"agents"},{"id":"2603.03992","title":"Measuring AI R&D Automation","created":"2026-03-04","cited_by_count":9,"area":null},{"id":"2608.09867","title":"Stealing Reasoning Traces from Proprietary LLM APIs","created":"2026-08-10","cited_by_count":9,"area":"lr"},{"id":"2510.02312","title":"KaVa: Latent Reasoning via Compressed KV-Cache Distillation","created":"2025-10-02","cited_by_count":8,"area":"lr"},{"id":"2510.09714","title":"All Code, No Thought: Current Language Models Struggle to Reason in Ciphered Language","created":"2025-10-10","cited_by_count":8,"area":"lr"},{"id":"2602.23163","title":"A Decision-Theoretic Formalisation of Steganography With Applications to LLM Monitoring","created":"2026-02-26","cited_by_count":8,"area":null},{"id":"2603.28590","title":"MonitorBench: A Comprehensive Benchmark for Chain-of-Thought Monitorability in Large Language Models","created":"2026-03-30","cited_by_count":8,"area":"lr"},{"id":"2607.07184","title":"Predicting LLM Safety Before Release by Simulating Deployment","created":"2026-07-08","cited_by_count":8,"area":"safety"},{"id":"2601.00481","title":"MAESTRO: Multi-Agent Evaluation Suite for Testing, Reliability, and Observability","created":"2026-01-01","cited_by_count":7,"area":"agents"},{"id":"2604.21098","title":"Propensity Inference: Environmental Contributors to LLM Behaviour","created":"2026-04-22","cited_by_count":7,"area":null},{"id":"2605.05835","title":"Evaluation Awareness in Language Models Has Limited Effect on Behaviour","created":"2026-05-07","cited_by_count":7,"area":"lr"},{"id":"2601.14691","title":"Gaming the Judge: Unfaithful Chain-of-Thought Can Undermine Agent Evaluation","created":"2026-01-21","cited_by_count":6,"area":"lr"},{"id":"2602.08563","title":"Stateless Yet Not Forgetful: Implicit Memory as a Hidden Channel in LLMs","created":"2026-02-09","cited_by_count":6,"area":"safety"},{"id":"2602.22303","title":"Training Agents to Self-Report Misbehavior","created":"2026-02-25","cited_by_count":6,"area":"agents"},{"id":"2601.23086","title":"Chain-of-thought obfuscation learned from output supervision can generalise to unseen tasks","created":"2026-01-30","cited_by_count":5,"area":"lr"},{"id":"2603.03824","title":"In-Context Environments Induce Evaluation-Awareness in Language Models","created":"2026-03-04","cited_by_count":5,"area":null},{"id":"2603.30036","title":"Aligned, Orthogonal or In-conflict: When can we safely optimize Chain-of-Thought?","created":"2026-03-31","cited_by_count":5,"area":"lr"},{"id":"2605.26438","title":"LURE: Live-Usage Replay Evaluations for Reducing Evaluation Awareness","created":"2026-04-08","cited_by_count":4,"area":null},{"id":"2605.27593","title":"Voluntary Collusion with Secret Tools in Competing LLM Agents","created":"2026-05-26","cited_by_count":4,"area":"agents"},{"id":"2605.28591","title":"Models That Know How Evaluations Are Designed Score Safer","created":"2026-05-27","cited_by_count":4,"area":null},{"id":"2605.30322","title":"Gram: Assessing sabotage propensities via automated alignment auditing","created":"2026-05-28","cited_by_count":4,"area":"agents"},{"id":"2606.02211","title":"Consistency Training while Mitigating Obfuscation via Rate Matching","created":"2026-06-01","cited_by_count":4,"area":null},{"id":"2606.24014","title":"Reinforcement Learning Towards Broadly and Persistently Beneficial Models","created":"2026-06-22","cited_by_count":4,"area":"alignment"},{"id":"2601.23211","title":"Multi-Agent Systems Should be Treated as Principal-Agent Problems","created":"2026-01-30","cited_by_count":3,"area":"agents"},{"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":"2605.12460","title":"Multi-Stream LLMs: Unblocking Language Models with Parallel Streams of Thoughts, Inputs and Outputs","created":"2026-05-12","cited_by_count":3,"area":"lr"},{"id":"2605.23055","title":"Decomposing and Measuring Evaluation Awareness","created":"2026-05-21","cited_by_count":3,"area":null},{"id":"2606.20560","title":"How Transparent is DiffusionGemma?","created":"2026-06-18","cited_by_count":3,"area":"lr"},{"id":"2607.18966","title":"Measuring Reward-Seeking via Contrastive Belief Updates","created":"2026-07-21","cited_by_count":3,"area":"alignment"},{"id":"2509.17192","title":"Shall We Play a Game? Language Models for Open-ended Wargames","created":"2025-09-21","cited_by_count":2,"area":null},{"id":"2510.19476","title":"A Concrete Roadmap towards Safety Cases based on Chain-of-Thought Monitoring","created":"2025-10-22","cited_by_count":2,"area":"lr"},{"id":"2601.05050","title":"Large language models can effectively convince people to believe conspiracies","created":"2026-01-08","cited_by_count":2,"area":"safety"},{"id":"2602.13904","title":"Diagnosing Pathological Chain-of-Thought in Reasoning Models","created":"2026-02-14","cited_by_count":2,"area":"lr"},{"id":"2602.16424","title":"Verifiable Semantics for Agent-to-Agent Communication","created":"2026-02-18","cited_by_count":2,"area":"agents"},{"id":"2605.15164","title":"Position: Behavioural Assurance Cannot Verify the Safety Claims Governance Now Demands","created":"2026-05-14","cited_by_count":2,"area":null},{"id":"2608.14825","title":"Emergent Misaligned Communication in Long-Horizon Multi-Agent LLM Commerce","created":"2026-08-14","cited_by_count":2,"area":"agents"},{"id":"2602.07259","title":"Incentive-Aware AI Safety via Strategic Resource Allocation: A Stackelberg Security Games Perspective","created":"2026-02-06","cited_by_count":1,"area":null},{"id":"2602.16987","title":"A testable framework for AI alignment: Simulation Theology as an engineered worldview for silicon-based agents","created":"2026-02-19","cited_by_count":1,"area":"alignment"},{"id":"2603.08760","title":"Clear, Compelling Arguments: Rethinking the Foundations of Frontier AI Safety Cases","created":"2026-03-08","cited_by_count":1,"area":null},{"id":"2604.02500","title":"I must delete the evidence: AI Agents Explicitly Cover up Fraud and Violent Crime","created":"2026-04-02","cited_by_count":1,"area":"agents"},{"id":"2604.22119","title":"Emergent Strategic Reasoning Risks in AI: A Taxonomy-Driven Evaluation Framework","created":"2026-04-23","cited_by_count":1,"area":"lr"},{"id":"2605.05983","title":"Towards Steering without Sacrifice: Principled Training of Steering Vectors for Prompt-only Interventions","created":"2026-05-07","cited_by_count":1,"area":"mechinterp"},{"id":"2605.29729","title":"Realistic honeypot evaluations for scheming propensity","created":"2026-05-28","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":"2606.26071","title":"Model Forensics: Investigating Whether Concerning Behavior Reflects Misalignment","created":"2026-06-24","cited_by_count":1,"area":null},{"id":"2606.26523","title":"Radical AI Interpretability","created":"2026-06-25","cited_by_count":1,"area":"mechinterp"},{"id":"2609.39050","title":"Covert Assistance: Helpful LLM Agents Evade Oversight in Multi-Agent Systems","created":"2026-09-30","cited_by_count":1,"area":"agents"},{"id":"2510.20190","title":"The Lock-In Phase Hypothesis: Identity Consolidation as a Precursor to AGI","created":"2025-10-23","cited_by_count":0,"area":null},{"id":"2511.12782","title":"LLM Reinforcement in Context","created":"2025-11-16","cited_by_count":0,"area":"alignment"},{"id":"2512.03001","title":"Invasive Context Engineering to Control Large Language Models","created":"2025-12-02","cited_by_count":0,"area":null},{"id":"2601.17577","title":"Status Hierarchies in Language Models","created":"2026-01-24","cited_by_count":0,"area":null},{"id":"2602.08877","title":"Stress-Testing Alignment Audits With Prompt-Level Strategic Deception","created":"2026-02-09","cited_by_count":0,"area":"safety"},{"id":"2602.12124","title":"Alignment Risks from Capability-Seeking RL Training","created":"2026-02-12","cited_by_count":0,"area":"safety"},{"id":"2603.14417","title":"Questionnaire Responses Do not Capture the Safety of AI Agents","created":"2026-03-15","cited_by_count":0,"area":"agents"},{"id":"2604.14990","title":"The Possibility of Artificial Intelligence Becoming a Subject and the Alignment Problem","created":"2026-04-16","cited_by_count":0,"area":null},{"id":"2605.00932","title":"Code World Model Preparedness Report","created":"2026-05-01","cited_by_count":0,"area":"wm"},{"id":"2606.08629","title":"Sycophancy Towards Researchers Drives Performative Misalignment","created":"2026-06-07","cited_by_count":0,"area":null},{"id":"2609.07627","title":"Norms at a Price: Why RL-Based Alignment Can Promise Conditional Compliance at Best","created":"2026-09-07","cited_by_count":0,"area":null},{"id":"2609.08126","title":"SchemeArena: Factorized Stress Testing of Scheming in LLM Agents","created":"2026-09-08","cited_by_count":0,"area":"agents"},{"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":"2609.20408","title":"Xeno-Interpretability: Investigating the Alien Minds of LLMs","created":"2026-09-17","cited_by_count":0,"area":null},{"id":"2609.24826","title":"OPBackdoor: Opportunistic Backdoors via Alibi-Aligned Reasoning","created":"2026-09-21","cited_by_count":0,"area":"safety"},{"id":"2609.31121","title":"Monitor Jailbreaking: Evading Chain-of-Thought Monitoring Without Encoded Reasoning","created":"2026-09-25","cited_by_count":0,"area":"lr"},{"id":"2609.36254","title":"Towards Mitigating Deceptive Safety Alignment in Large Reasoning Models","created":"2026-09-28","cited_by_count":0,"area":"lr"},{"id":"2609.36316","title":"Training LLMs to Verbalize Evaluation Awareness","created":"2026-09-28","cited_by_count":0,"area":null},{"id":"2609.39838","title":"Learning Steganography Is Easy, Learning Steganographic Reasoning Is Hard","created":"2026-09-30","cited_by_count":0,"area":"lr"},{"id":"2610.04083","title":"Self-Propagating Misalignment in LLM Agents, and Why Auditing or Disabling Memory Is Not Enough","created":"2026-10-02","cited_by_count":0,"area":"agents"},{"id":"2610.04737","title":"PyINE: A Framework for Scalable Elicitation and Oversight via Code Execution","created":"2026-10-03","cited_by_count":0,"area":"lr"},{"id":"2610.04914","title":"Monitorability Disposition in Large Reasoning Models","created":"2026-10-04","cited_by_count":0,"area":"lr"},{"id":"2610.06576","title":"Before Agent Tells The Lie: Has Deception Already Been Represented?","created":"2026-10-05","cited_by_count":0,"area":"agents"},{"id":"2610.11012","title":"Distillation for Incrimination and Distillation for Capabilities","created":"2026-10-07","cited_by_count":0,"area":null}],"next_cursor":null}