{"total":145,"rows":[{"id":"2504.01848","title":"PaperBench: Evaluating AI's Ability to Replicate AI Research","created":"2025-04-02","cited_by_count":321,"area":null},{"id":"2412.04984","title":"Frontier Models are Capable of In-context Scheming","created":"2024-12-06","cited_by_count":287,"area":null},{"id":"2510.05179","title":"Agentic Misalignment: How LLMs Could Be Insider Threats","created":"2025-10-05","cited_by_count":158,"area":"agents"},{"id":"2410.13787","title":"Looking Inward: Language Models Can Learn About Themselves by Introspection","created":"2024-10-17","cited_by_count":113,"area":null},{"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":"2502.06559","title":"Can We Trust AI Benchmarks? An Interdisciplinary Review of Current Issues in AI Evaluation","created":"2025-02-10","cited_by_count":60,"area":null},{"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":"2502.05209","title":"Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilities","created":"2025-02-03","cited_by_count":39,"area":null},{"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":"2505.18807","title":"Mitigating Deceptive Alignment via Self-Monitoring","created":"2025-05-24","cited_by_count":29,"area":"lr"},{"id":"2502.11355","title":"Nuclear Deployed: Analyzing Catastrophic Risks in Decision-making of Autonomous LLM Agents","created":"2025-02-17","cited_by_count":27,"area":"lr"},{"id":"2412.01784","title":"Noise Injection Reveals Hidden Capabilities of Sandbagging Language Models","created":"2024-12-02","cited_by_count":24,"area":null},{"id":"2506.04018","title":"AgentMisalignment: Measuring the Propensity for Misaligned Behaviour in LLM-Based Agents","created":"2025-06-04","cited_by_count":22,"area":"agents"},{"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":"2507.12691","title":"Benchmarking Deception Probes via Black-to-White Performance Boosts","created":"2025-07-16","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":"2508.00943","title":"LLMs Can Covertly Sandbag on Capability Evaluations Against Chain-of-Thought Monitoring","created":"2025-07-31","cited_by_count":18,"area":"lr"},{"id":"2602.11865","title":"Intelligent AI Delegation","created":"2026-02-12","cited_by_count":18,"area":null},{"id":"2510.27378","title":"Measuring Chain-of-Thought Monitorability Through Faithfulness and Verbosity","created":"2025-10-31","cited_by_count":17,"area":"lr"},{"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":"2511.22662","title":"Difficulties with Evaluating a Deception Detector for AIs","created":"2025-11-27","cited_by_count":13,"area":null},{"id":"2602.15515","title":"The Obfuscation Atlas: Mapping Where Honesty Emerges in RLVR with Deception Probes","created":"2026-02-17","cited_by_count":12,"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":"2601.06663","title":"SafePro: Evaluating the Safety of Professional-Level AI Agents","created":"2026-01-10","cited_by_count":10,"area":"agents"},{"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":"2512.11949","title":"Neural Chameleons: Language Models Can Learn to Hide Their Thoughts from Unseen Activation Monitors","created":"2025-12-12","cited_by_count":9,"area":"lr"},{"id":"2502.06470","title":"A Survey of Theory of Mind in Large Language Models: Evaluations, Representations, and Safety Risks","created":"2025-02-10","cited_by_count":8,"area":null},{"id":"2503.04744","title":"Safety Cases: A Scalable Approach to Frontier AI Safety","created":"2025-02-05","cited_by_count":8,"area":null},{"id":"2506.23706","title":"Attestable Audits: Verifiable AI Safety Benchmarks Using Trusted Execution Environments","created":"2025-06-30","cited_by_count":8,"area":null},{"id":"2510.08211","title":"LLMs Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions","created":"2025-10-09","cited_by_count":8,"area":null},{"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":"2504.20084","title":"AI Awareness","created":"2025-04-25","cited_by_count":7,"area":null},{"id":"2602.18971","title":"When Do LLM Preferences Predict Downstream Behavior?","created":"2026-02-21","cited_by_count":7,"area":"alignment"},{"id":"2603.10011","title":"Gemma Needs Help: Investigating and Mitigating Emotional Instability in LLMs","created":"2026-02-17","cited_by_count":7,"area":null},{"id":"2604.01476","title":"From Rebound to Remedy: Understanding and Mitigating Reward Hacking via Representation Engineering","created":"2026-04-01","cited_by_count":7,"area":"alignment"},{"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":"2605.13329","title":"Tracing Persona Vectors Through LLM Pretraining","created":"2026-05-13","cited_by_count":7,"area":null},{"id":"2510.24626","title":"Relative Scaling Laws for LLMs","created":"2025-10-28","cited_by_count":6,"area":null},{"id":"2602.04930","title":"Attack Selection Reduces Safety in Concentrated AI Control Settings against Trusted Monitoring","created":"2026-02-04","cited_by_count":6,"area":"safety"},{"id":"2602.11079","title":"Probe-Based Data Attribution: Discovering and Mitigating Undesirable Behaviors in LLM Post-Training","created":"2026-02-11","cited_by_count":6,"area":null},{"id":"2602.22303","title":"Training Agents to Self-Report Misbehavior","created":"2026-02-25","cited_by_count":6,"area":"agents"},{"id":"2607.01033","title":"The Model Organism Lottery: Model Organism Interpretability Strongly Depends on Training Methodology","created":"2026-07-01","cited_by_count":6,"area":"mechinterp"},{"id":"2412.06845","title":"7B Fully Open Source Moxin-LLM/VLM -- From Pretraining to GRPO-based Reinforcement Learning Enhancement","created":"2024-12-08","cited_by_count":5,"area":"lr"},{"id":"2504.13707","title":"OpenDeception: Learning Deception and Trust in Human-AI Interaction via Multi-Agent Simulation","created":"2025-04-18","cited_by_count":5,"area":"agents"},{"id":"2505.20435","title":"The Shape of Adversarial Influence: Characterizing LLM Latent Spaces with Persistent Homology","created":"2025-05-26","cited_by_count":5,"area":"safety"},{"id":"2506.10236","title":"Prompt Attacks Reveal Superficial Knowledge Removal in Unlearning Methods","created":"2025-06-11","cited_by_count":5,"area":"safety"},{"id":"2508.06361","title":"Beyond Prompt-Induced Lies: Investigating LLM Deception on Benign Prompts","created":"2025-08-08","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":"2609.01836","title":"Agent Memory Is a Surface for Endogenous Authorization Laundering","created":"2026-09-01","cited_by_count":5,"area":"agents"},{"id":"2411.12820","title":"Declare and Justify: Explicit assumptions in AI evaluations are necessary for effective regulation","created":"2024-11-19","cited_by_count":4,"area":null},{"id":"2412.08653","title":"What AI evaluations for preventing catastrophic risks can and cannot do","created":"2024-11-26","cited_by_count":4,"area":null},{"id":"2604.22082","title":"Removing Sandbagging in LLMs by Training with Weak Supervision","created":"2026-04-23","cited_by_count":4,"area":null},{"id":"2504.07831","title":"Deceptive Automated Interpretability: Language Models Coordinating to Fool Oversight Systems","created":"2025-04-10","cited_by_count":3,"area":"mechinterp"},{"id":"2510.07761","title":"Test-Time Reasoners Are Strategic Multiple-Choice Test-Takers","created":"2025-10-09","cited_by_count":3,"area":"lr"},{"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":"2603.26676","title":"Evaluating Human-AI Safety: A Framework for Measuring Harmful Capability Uplift","created":"2026-03-06","cited_by_count":3,"area":null},{"id":"2604.09104","title":"Scheming in the wild: detecting real-world AI scheming incidents with open-source intelligence","created":"2026-04-10","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":"2606.29604","title":"Mechanistically Eliciting Latent Behaviors in Language Models","created":"2026-06-28","cited_by_count":3,"area":null},{"id":"2506.11604","title":"VLM@school -- Evaluation of AI image understanding on German middle school knowledge","created":"2025-06-13","cited_by_count":2,"area":"lr"},{"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":"2602.05656","title":"Alignment Verifiability in Large Language Models: Normative Indistinguishability under Behavioral Evaluation","created":"2026-02-05","cited_by_count":2,"area":null},{"id":"2604.02174","title":"Quantifying Self-Preservation Bias in Large Language Models","created":"2026-04-02","cited_by_count":2,"area":"alignment"},{"id":"2604.28182","title":"Exploration Hacking: Can LLMs Learn to Resist RL Training?","created":"2026-04-30","cited_by_count":2,"area":"lr"},{"id":"2607.08066","title":"Persuasion Attacks Can Decrease Effectiveness of CoT Monitoring","created":"2026-07-09","cited_by_count":2,"area":"lr"},{"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":"2609.19101","title":"Monitoring and Discovering Reward Hacking with Internal Representations during LLM Evaluations","created":"2026-09-16","cited_by_count":2,"area":"alignment"},{"id":"2505.07846","title":"Winning at All Cost: A Small Environment for Eliciting Specification Gaming Behaviors in Large Language Models","created":"2025-05-07","cited_by_count":1,"area":null},{"id":"2509.22735","title":"Regulating the Agency of LLM-based Agents","created":"2025-09-25","cited_by_count":1,"area":"agents"},{"id":"2510.01266","title":"OpenAI's GPT-OSS-20B Model and Safety Alignment Issues in a Low-Resource Language","created":"2025-09-26","cited_by_count":1,"area":"alignment"},{"id":"2510.21524","title":"EU-Agent-Bench: Measuring Illegal Behavior of LLM Agents Under EU Law","created":"2025-10-24","cited_by_count":1,"area":"agents"},{"id":"2512.00349","title":"Debate with Images: Detecting Deceptive Behaviors in Multimodal Large Language Models","created":"2025-11-29","cited_by_count":1,"area":"mllm"},{"id":"2602.08964","title":"A Behavioural and Representational Evaluation of Goal-Directedness in Language Model Agents","created":"2026-02-09","cited_by_count":1,"area":"agents"},{"id":"2604.17663","title":"ATLAS: Constitution-Conditioned Latent Geometry and Redistribution Across Language Models and Neural Perturbation Data","created":"2026-04-19","cited_by_count":1,"area":null},{"id":"2604.18970","title":"Mechanistic Anomaly Detection via Functional Attribution","created":"2026-04-21","cited_by_count":1,"area":"safety"},{"id":"2604.25249","title":"Below-Chance Blindness: Prompted Underperformance in Small LLMs Produces Positional Bias Rather than Answer Avoidance","created":"2026-04-28","cited_by_count":1,"area":null},{"id":"2606.04413","title":"(Mis)generalization of Helpful-only Fine-tuning","created":"2026-06-03","cited_by_count":1,"area":null},{"id":"2606.06286","title":"LLMs Can Leak Training Data But Do They Want To? A Propensity-Aware Evaluation of Memorization in LLMs","created":"2026-06-04","cited_by_count":1,"area":null},{"id":"2606.09711","title":"Proxy Reward Internalization and Mechanistic Exploitation: A Learned Precursor to Reward Hacking and Its Generalization","created":"2026-06-08","cited_by_count":1,"area":"alignment"},{"id":"2606.12016","title":"Generalization Hacking: Models Can Game Reinforcement Learning by Preventing Behavioral Generalization","created":"2026-06-10","cited_by_count":1,"area":"alignment"},{"id":"2606.24251","title":"Probing the Misaligned Thinking Process of Language Models","created":"2026-06-23","cited_by_count":1,"area":"lr"},{"id":"2607.14399","title":"Instrument Effects in Language-Model Honesty Evaluation: An Auditable Single-System Demonstration","created":"2026-07-15","cited_by_count":1,"area":null},{"id":"2607.25379","title":"Cyber-Capable AI Agents: Vulnerabilities, Evaluation Containment, and Defensive Response","created":"2026-07-28","cited_by_count":1,"area":"agents"},{"id":"2608.05086","title":"Item Response Theory for AI Safety","created":"2026-08-05","cited_by_count":1,"area":null},{"id":"2608.14577","title":"HarmProfile: Characterizing Harmful Distributions in Frontier LLMs","created":"2026-06-11","cited_by_count":1,"area":"safety"},{"id":"2608.29458","title":"Reference-Grafting Matches Fine-Tuning at Eliciting Sandbagged Capabilities","created":"2026-08-29","cited_by_count":1,"area":null},{"id":"2608.29461","title":"A Causal Model for Locating and Unlocking Sandbagging in Model Organisms","created":"2026-08-29","cited_by_count":1,"area":null},{"id":"2609.36490","title":"LLMs Learn to Evade Latent Monitors from Prior Feedback Alone","created":"2026-09-29","cited_by_count":1,"area":"agents"},{"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":"2506.15735","title":"ContextBench: Modifying Contexts for Targeted Latent Activation","created":"2025-06-15","cited_by_count":0,"area":null},{"id":"2506.22977","title":"On the Generalizability of \"Competition of Mechanisms: Tracing How Language Models Handle Facts and Counterfactuals\"","created":"2025-06-28","cited_by_count":0,"area":null},{"id":"2510.01259","title":"In AI Sweet Harmony: Sociopragmatic Guardrail Bypasses and Evaluation-Awareness in OpenAI gpt-oss-20b","created":"2025-09-25","cited_by_count":0,"area":"safety"},{"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":"2602.08449","title":"When Evaluation Becomes a Side Channel: Regime Leakage and Structural Mitigations for Alignment Assessment","created":"2026-02-09","cited_by_count":0,"area":null},{"id":"2602.14529","title":"Disentangling Deception and Hallucination Failures in LLMs","created":"2026-02-16","cited_by_count":0,"area":null},{"id":"2602.17127","title":"The Emergence of Lab-Driven Alignment Signatures: A Psychometric Framework for Auditing Latent Bias and Compounding Risk in Generative AI","created":"2026-02-19","cited_by_count":0,"area":"lr"},{"id":"2603.13791","title":"DeceptGuard :A Constitutional Oversight Framework For Detecting Deception in LLM Agents","created":"2026-03-14","cited_by_count":0,"area":"agents"}],"next_cursor":"100"}