{"total":84,"rows":[{"id":"2412.14093","title":"Alignment faking in large language models","created":"2024-12-18","cited_by_count":360,"area":"alignment"},{"id":"2412.04984","title":"Frontier Models are Capable of In-context Scheming","created":"2024-12-06","cited_by_count":287,"area":null},{"id":"2502.17424","title":"Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs","created":"2025-02-24","cited_by_count":172,"area":"safety"},{"id":"2511.18397","title":"Natural Emergent Misalignment from Reward Hacking in Production RL","created":"2025-11-23","cited_by_count":136,"area":"alignment"},{"id":"2506.11613","title":"Model Organisms for Emergent Misalignment","created":"2025-06-13","cited_by_count":91,"area":null},{"id":"2506.11618","title":"Convergent Linear Representations of Emergent Misalignment","created":"2025-06-13","cited_by_count":71,"area":null},{"id":"2410.03642","title":"Aligning LLMs with Individual Preferences via Interaction","created":"2024-10-04","cited_by_count":63,"area":"alignment"},{"id":"2410.21514","title":"Sabotage Evaluations for Frontier Models","created":"2024-10-28","cited_by_count":52,"area":null},{"id":"2501.17315","title":"A sketch of an AI control safety case","created":"2025-01-28","cited_by_count":39,"area":"agents"},{"id":"2505.01420","title":"Evaluating Frontier Models for Stealth and Situational Awareness","created":"2025-05-02","cited_by_count":37,"area":null},{"id":"2507.01786","title":"Probing and Steering Evaluation Awareness of Language Models","created":"2025-07-02","cited_by_count":35,"area":null},{"id":"2411.03336","title":"Towards evaluations-based safety cases for AI scheming","created":"2024-10-29","cited_by_count":33,"area":null},{"id":"2505.17120","title":"Self-Interpretability: LLMs Can Describe Complex Internal Processes that Drive Their Decisions","created":"2025-05-21","cited_by_count":25,"area":null},{"id":"2506.05068","title":"Does It Make Sense to Speak of Introspection in Large Language Models?","created":"2025-06-05","cited_by_count":25,"area":null},{"id":"2504.05259","title":"How to evaluate control measures for LLM agents? A trajectory from today to superintelligence","created":"2025-04-07","cited_by_count":20,"area":"agents"},{"id":"2601.11369","title":"Institutional AI: Governing LLM Collusion in Multi-Agent Cournot Markets via Public Governance Graphs","created":"2026-01-16","cited_by_count":20,"area":"agents"},{"id":"2508.13180","title":"Search-Time Data Contamination","created":"2025-08-12","cited_by_count":19,"area":null},{"id":"2501.03151","title":"Large language models for artificial general intelligence (AGI): A survey of foundational principles and approaches","created":"2025-01-06","cited_by_count":18,"area":"mllm"},{"id":"2502.08301","title":"Compromising Honesty and Harmlessness in Language Models via Deception Attacks","created":"2025-02-12","cited_by_count":17,"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":"2410.18819","title":"From Imitation to Introspection: Probing Self-Consciousness in Language Models","created":"2024-10-24","cited_by_count":14,"area":null},{"id":"2505.17815","title":"Evaluation Faking: Unveiling Observer Effects in Safety Evaluation of Frontier AI Systems","created":"2025-05-23","cited_by_count":14,"area":"lr"},{"id":"2509.21545","title":"Evidence for Limited Metacognition in LLMs","created":"2025-09-25","cited_by_count":12,"area":null},{"id":"2410.06195","title":"EgoSocialArena: Benchmarking the Social Intelligence of Large Language Models from a First-person Perspective","created":"2024-10-08","cited_by_count":10,"area":null},{"id":"2512.24661","title":"Do Large Language Models Know What They Are Capable Of?","created":"2025-12-31","cited_by_count":10,"area":null},{"id":"2504.03553","title":"Agentic Knowledgeable Self-awareness","created":"2025-04-04","cited_by_count":8,"area":"agents"},{"id":"2509.04373","title":"Measuring Bias or Measuring the Task: Understanding the Brittle Nature of LLM Gender Biases","created":"2025-09-04","cited_by_count":8,"area":null},{"id":"2504.20084","title":"AI Awareness","created":"2025-04-25","cited_by_count":7,"area":null},{"id":"2603.16928","title":"Noticing the Watcher: LLM Agents Can Infer CoT Monitoring from Blocking Feedback","created":"2026-03-14","cited_by_count":7,"area":"lr"},{"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":"2606.07612","title":"Position: Anthropomorphic Misalignment Research Needs Stronger Evidence","created":"2026-05-29","cited_by_count":7,"area":null},{"id":"2510.00857","title":"ManagerBench: Evaluating the Safety-Pragmatism Trade-off in Autonomous LLMs","created":"2025-10-01","cited_by_count":6,"area":null},{"id":"2506.22957","title":"Agent-to-Agent Theory of Mind: Testing Interlocutor Awareness among Large Language Models","created":"2025-06-28","cited_by_count":5,"area":"agents"},{"id":"2507.12872","title":"Manipulation Attacks by Misaligned AI: Risk Analysis and Safety Case Framework","created":"2025-07-17","cited_by_count":5,"area":null},{"id":"2604.13051","title":"The Consciousness Cluster: Emergent preferences of Models that Claim to be Conscious","created":"2026-03-17","cited_by_count":5,"area":"alignment"},{"id":"2512.12411","title":"Detecting the Disturbance: A Nuanced View of Introspective Abilities in LLMs","created":"2025-12-13","cited_by_count":4,"area":null},{"id":"2605.28591","title":"Models That Know How Evaluations Are Designed Score Safer","created":"2026-05-27","cited_by_count":4,"area":null},{"id":"2606.32038","title":"Introspective Coupling: Self-Explanation Training Tracks Behavioral Change Despite Fixed Supervision","created":"2026-06-30","cited_by_count":4,"area":null},{"id":"2505.24539","title":"Localizing Persona Representations in LLMs","created":"2025-05-30","cited_by_count":3,"area":null},{"id":"2606.23583","title":"Evaluation Awareness Is Not One Capability: Evidence from Open Language Models","created":"2026-06-22","cited_by_count":3,"area":null},{"id":"2607.18966","title":"Measuring Reward-Seeking via Contrastive Belief Updates","created":"2026-07-21","cited_by_count":3,"area":"alignment"},{"id":"2412.03904","title":"MISR: Measuring Instrumental Self-Reasoning in Frontier Models","created":"2024-12-05","cited_by_count":2,"area":"lr"},{"id":"2508.19546","title":"Language Models Identify Ambiguities and Exploit Loopholes","created":"2025-08-27","cited_by_count":2,"area":null},{"id":"2510.05169","title":"From Poisoned to Aware: Fostering Backdoor Self-Awareness in LLMs","created":"2025-10-05","cited_by_count":2,"area":"safety"},{"id":"2604.28182","title":"Exploration Hacking: Can LLMs Learn to Resist RL Training?","created":"2026-04-30","cited_by_count":2,"area":"lr"},{"id":"2607.04277","title":"Self-Reference in Large Language Models: The Introspection Threshold for Recursive Self-Improvement","created":"2026-07-05","cited_by_count":2,"area":"synth"},{"id":"2607.20436","title":"Routing Subspaces: Auditing Evaluation-to-Deployment Mismatch in Fine-Tuned Language Models","created":"2026-05-11","cited_by_count":2,"area":null},{"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":"2603.09200","title":"The Reasoning Trap -- Logical Reasoning as a Mechanistic Pathway to Situational Awareness","created":"2026-03-10","cited_by_count":1,"area":"lr"},{"id":"2605.29729","title":"Realistic honeypot evaluations for scheming propensity","created":"2026-05-28","cited_by_count":1,"area":null},{"id":"2606.08243","title":"Building Comparative Motivation Profiles with Instrumental Interventions","created":"2026-06-06","cited_by_count":1,"area":"mechinterp"},{"id":"2606.12429","title":"Muse Spark Safety & Preparedness Report","created":"2026-05-14","cited_by_count":1,"area":null},{"id":"2606.23700","title":"Self-Recognition Finetuning can Prevent and Reverse Emergent Misalignment","created":"2026-06-04","cited_by_count":1,"area":"safety"},{"id":"2606.29196","title":"Representational Depth of Evaluation Awareness Shifts With Scale in Open-Weight Language Models","created":"2026-06-28","cited_by_count":1,"area":null},{"id":"2607.14111","title":"Introspection Fine-Tuning (IFT): Training Small LLMs to Introspect","created":"2026-05-08","cited_by_count":1,"area":null},{"id":"2608.26159","title":"Self-Generated Text Recognition: Quality Heuristics, Cross-Task Transfer, and Downstream Bias in LLM Evaluation","created":"2026-07-07","cited_by_count":1,"area":"synth"},{"id":"2609.01491","title":"GlossoGen: Emergent Language in Complex Multi-Agent LLM Interactions","created":"2026-09-01","cited_by_count":1,"area":"agents"},{"id":"2506.03614","title":"VLMs Can Aggregate Scattered Training Patches","created":"2025-06-04","cited_by_count":0,"area":"mllm"},{"id":"2508.12277","title":"The Self-Execution Benchmark: Measuring LLMs' Attempts to Overcome Their Lack of Self-Execution","created":"2025-08-17","cited_by_count":0,"area":"lr"},{"id":"2510.02840","title":"Take Goodhart Seriously: Principled Limit on General-Purpose AI Optimization","created":"2025-10-03","cited_by_count":0,"area":null},{"id":"2603.04407","title":"Semantic Containment as a Fundamental Property of Emergent Misalignment","created":"2026-02-02","cited_by_count":0,"area":null},{"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":"2605.20382","title":"Do as I Say, Not as I Do: Instruction-Induction Conflict in LLMs","created":"2026-05-19","cited_by_count":0,"area":null},{"id":"2606.23343","title":"Group Selection Promotes Prosocial Prompts in Populations of LLM Agents","created":"2026-06-22","cited_by_count":0,"area":"agents"},{"id":"2606.23671","title":"Can LLMs Reliably Self-Report Adversarial Prefills, and How?","created":"2026-06-22","cited_by_count":0,"area":null},{"id":"2606.31644","title":"Moral Safety in LLMs: Exposing Performative Compliance with Puzzled Cues","created":"2026-06-30","cited_by_count":0,"area":null},{"id":"2607.14250","title":"The Severance Problem: LLMs are Unaware of the Person Beyond the Prompt","created":"2026-07-15","cited_by_count":0,"area":null},{"id":"2608.13329","title":"A Probe Direction Is a Property of Its Prompt","created":"2026-08-13","cited_by_count":0,"area":null},{"id":"2608.27348","title":"INTENT-AS-A-TOOL Makes it Easy to Track Agentic Misalignment","created":"2026-08-27","cited_by_count":0,"area":"agents"},{"id":"2609.01611","title":"EvalDetectBench: A Benchmark for Measuring Evaluation Awareness in Frontier Language Models","created":"2026-06-08","cited_by_count":0,"area":null},{"id":"2609.05009","title":"Language models judge war differently when tested for alignment","created":"2026-09-04","cited_by_count":0,"area":"alignment"},{"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":"2609.09899","title":"Strangers to Themselves: What Language Models Say About Themselves Is Generic","created":"2026-09-09","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":"2609.22119","title":"Evaluation Awareness Shifts from Format to Context with Model Scale","created":"2026-08-20","cited_by_count":0,"area":null},{"id":"2609.25021","title":"\"As a Language Model...\": Chat Template Switches LLM Self-Referential Voice and Activation Steering Reproduces It","created":"2026-08-09","cited_by_count":0,"area":"mechinterp"},{"id":"2609.30048","title":"Style, Not Self: Surface Cues Explain Zero-Shot Code Attribution by Large Language Models","created":"2026-09-24","cited_by_count":0,"area":null},{"id":"2609.35261","title":"Imprint Reader: From Weight-Update Readout to Behavioral Intervention","created":"2026-09-28","cited_by_count":0,"area":null},{"id":"2609.38948","title":"DrivingBench: Can Vision-Language Models Drive a Toyota Corolla?","created":"2026-09-30","cited_by_count":0,"area":"mllm"},{"id":"2610.08540","title":"Toward Alignment Scaling Laws: A Framework and First Preregistered Measurements","created":"2026-10-06","cited_by_count":0,"area":"safety"},{"id":"2610.09944","title":"AgentTime: Can Agents Estimate and Control Their Own Runtime?","created":"2026-10-07","cited_by_count":0,"area":"agents"},{"id":"2610.15153","title":"Can agents whose sole duty is to keep humans alive be hijacked to kill? Delayed multi-turn derived-action trap attacks in simulated high-stakes environments","created":"2026-10-10","cited_by_count":0,"area":"safety"},{"id":"2610.17495","title":"A Sense of Proprioception in Language Models","created":"2026-10-12","cited_by_count":0,"area":null}],"next_cursor":null}