{"total":384,"rows":[{"id":"2505.05410","title":"Reasoning Models Don't Always Say What They Think","created":"2025-05-08","cited_by_count":415,"area":"lr"},{"id":"2507.11473","title":"Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety","created":"2025-07-15","cited_by_count":249,"area":"lr"},{"id":"2509.08827","title":"A Survey of Reinforcement Learning for Large Reasoning Models","created":"2025-09-10","cited_by_count":180,"area":"lr"},{"id":"2510.05179","title":"Agentic Misalignment: How LLMs Could Be Insider Threats","created":"2025-10-05","cited_by_count":158,"area":"agents"},{"id":"2511.18397","title":"Natural Emergent Misalignment from Reward Hacking in Production RL","created":"2025-11-23","cited_by_count":136,"area":"alignment"},{"id":"2506.19143","title":"Thought Anchors: Which LLM Reasoning Steps Matter?","created":"2025-06-23","cited_by_count":129,"area":"lr"},{"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":"2507.14805","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","created":"2025-07-20","cited_by_count":95,"area":null},{"id":"2506.19823","title":"Persona Features Control Emergent Misalignment","created":"2025-06-24","cited_by_count":86,"area":null},{"id":"2506.15740","title":"SHADE-Arena: Evaluating Sabotage and Monitoring in LLM Agents","created":"2025-06-17","cited_by_count":71,"area":"agents"},{"id":"2508.17511","title":"School of Reward Hacks: Hacking harmless tasks generalizes to misaligned behavior in LLMs","created":"2025-08-24","cited_by_count":66,"area":"alignment"},{"id":"2503.24370","title":"Effectively Controlling Reasoning Models through Thinking Intervention","created":"2025-03-31","cited_by_count":65,"area":"lr"},{"id":"2506.13206","title":"Thought Crime: Backdoors and Emergent Misalignment in Reasoning Models","created":"2025-06-16","cited_by_count":65,"area":"lr"},{"id":"2505.23575","title":"CoT Red-Handed: Stress Testing Chain-of-Thought Monitoring","created":"2025-05-29","cited_by_count":54,"area":"lr"},{"id":"2506.10805","title":"Detecting High-Stakes Interactions with Activation Probes","created":"2025-06-12","cited_by_count":53,"area":null},{"id":"2504.21277","title":"Reinforced MLLM: A Survey on RL-Based Reasoning in Multimodal Large Language Models","created":"2025-04-30","cited_by_count":44,"area":"lr"},{"id":"2505.13763","title":"Language Models Are Capable of Metacognitive Monitoring and Control of Their Internal Activations","created":"2025-05-19","cited_by_count":41,"area":null},{"id":"2505.13775","title":"Beyond Semantics: The Unreasonable Effectiveness of Reasonless Intermediate Tokens","created":"2025-05-19","cited_by_count":41,"area":"lr"},{"id":"2603.05488","title":"Reasoning Theater: Disentangling Model Beliefs from Chain-of-Thought","created":"2026-03-05","cited_by_count":40,"area":"lr"},{"id":"2504.09762","title":"Position: Stop Anthropomorphizing Intermediate Tokens as Reasoning/Thinking Traces!","created":"2025-04-14","cited_by_count":37,"area":"lr"},{"id":"2506.19248","title":"Inference-Time Reward Hacking in Large Language Models","created":"2025-06-24","cited_by_count":34,"area":"alignment"},{"id":"2604.13602","title":"Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges","created":"2026-04-15","cited_by_count":33,"area":"alignment"},{"id":"2506.01926","title":"Large language models can learn and generalize steganographic chain-of-thought under process supervision","created":"2025-06-02","cited_by_count":32,"area":"lr"},{"id":"2508.19461","title":"Reliable Weak-to-Strong Monitoring of LLM Agents","created":"2025-08-26","cited_by_count":31,"area":"agents"},{"id":"2505.14617","title":"The Hawthorne Effect in Reasoning Models: Evaluating and Steering Test Awareness","created":"2025-05-20","cited_by_count":30,"area":"lr"},{"id":"2505.18807","title":"Mitigating Deceptive Alignment via Self-Monitoring","created":"2025-05-24","cited_by_count":29,"area":"lr"},{"id":"2505.18889","title":"Security Concerns for Large Language Models: A Survey","created":"2025-05-24","cited_by_count":29,"area":null},{"id":"2509.03403","title":"Beyond Correctness: Harmonizing Process and Outcome Rewards through RL Training","created":"2025-09-03","cited_by_count":29,"area":"alignment"},{"id":"2506.15674","title":"Leaky Thoughts: Large Reasoning Models Are Not Private Thinkers","created":"2025-06-18","cited_by_count":28,"area":"lr"},{"id":"2511.21654","title":"EvilGenie: A Reward Hacking Benchmark","created":"2025-11-26","cited_by_count":28,"area":"alignment"},{"id":"2604.15149","title":"LLMs Gaming Verifiers: RLVR can Lead to Reward Hacking","created":"2026-04-16","cited_by_count":28,"area":"lr"},{"id":"2605.21384","title":"SpecBench: Measuring Reward Hacking in Long-Horizon Coding Agents","created":"2026-05-20","cited_by_count":28,"area":"agents"},{"id":"2510.01367","title":"Is It Thinking or Cheating? Detecting Implicit Reward Hacking by Measuring Reasoning Effort","created":"2025-10-01","cited_by_count":27,"area":"lr"},{"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":"2509.03871","title":"A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Models","created":"2025-09-04","cited_by_count":25,"area":"lr"},{"id":"2603.05706","title":"Reasoning Models Struggle to Control their Chains of Thought","created":"2026-03-05","cited_by_count":25,"area":"lr"},{"id":"2506.14261","title":"RL-Obfuscation: Can Language Models Learn to Evade Latent-Space Monitors?","created":"2025-06-17","cited_by_count":24,"area":null},{"id":"2505.13774","title":"Measuring the Faithfulness of Thinking Drafts in Large Reasoning Models","created":"2025-05-19","cited_by_count":23,"area":"lr"},{"id":"2606.07682","title":"SWE-Marathon: Can Agents Autonomously Complete Ultra-Long-Horizon Software Work?","created":"2026-06-05","cited_by_count":23,"area":null},{"id":"2505.22203","title":"From Accuracy to Robustness: A Study of Rule- and Model-based Verifiers in Mathematical Reasoning","created":"2025-05-28","cited_by_count":22,"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":"2512.13526","title":"Async Control: Stress-testing Asynchronous Control Measures for LLM Agents","created":"2025-12-15","cited_by_count":22,"area":"agents"},{"id":"2506.22777","title":"Teaching Models to Verbalize Reward Hacking in Chain-of-Thought Reasoning","created":"2025-06-28","cited_by_count":21,"area":"lr"},{"id":"2602.06176","title":"Large Language Model Reasoning Failures","created":"2026-02-05","cited_by_count":21,"area":"lr"},{"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":"2510.09312","title":"Verifying Chain-of-Thought Reasoning via Its Computational Graph","created":"2025-10-10","cited_by_count":20,"area":"lr"},{"id":"2504.16728","title":"IRIS: Interactive Research Ideation System for Accelerating Scientific Discovery","created":"2025-04-23","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":"2506.10922","title":"Robustly Improving LLM Fairness in Realistic Settings via Interpretability","created":"2025-06-12","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":"2507.12428","title":"Can We Predict Alignment Before Models Finish Thinking? Towards Monitoring Misaligned Reasoning Models","created":"2025-07-16","cited_by_count":17,"area":"lr"},{"id":"2510.27378","title":"Measuring Chain-of-Thought Monitorability Through Faithfulness and Verbosity","created":"2025-10-31","cited_by_count":17,"area":"lr"},{"id":"2602.20710","title":"Counterfactual Simulation Training for Chain-of-Thought Faithfulness","created":"2026-02-24","cited_by_count":17,"area":"lr"},{"id":"2603.07084","title":"Countdown-Code: A Testbed for Studying The Emergence and Generalization of Reward Hacking in RLVR","created":"2026-03-07","cited_by_count":17,"area":"alignment"},{"id":"2603.15714","title":"How Vulnerable Are AI Agents to Indirect Prompt Injections? Insights from a Large-Scale Public Competition","created":"2026-03-16","cited_by_count":17,"area":"agents"},{"id":"2509.09396","title":"LLMs Don't Know Their Own Decision Boundaries: The Unreliability of Self-Generated Counterfactual Explanations","created":"2025-09-11","cited_by_count":16,"area":"synth"},{"id":"2508.06249","title":"In-Training Defenses against Emergent Misalignment in Language Models","created":"2025-08-08","cited_by_count":15,"area":"alignment"},{"id":"2510.24941","title":"Can Aha Moments Be Fake? Towards Quantifying Decorative and True Thinking in Chain-of-Thought","created":"2025-10-28","cited_by_count":15,"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":"2507.11662","title":"Let's Think in Two Steps: Mitigating Agreement Bias in MLLMs with Self-Grounded Verification","created":"2025-07-15","cited_by_count":14,"area":"lr"},{"id":"2512.19027","title":"Recontextualization Mitigates Specification Gaming without Modifying the Specification","created":"2025-12-22","cited_by_count":14,"area":null},{"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":"2603.04069","title":"Monitoring Emergent Reward Hacking During Generation via Internal Activations","created":"2026-03-04","cited_by_count":14,"area":"alignment"},{"id":"2603.04474","title":"From Spark to Fire: Modeling and Mitigating Error Cascades in LLM-Based Multi-Agent Collaboration","created":"2026-03-04","cited_by_count":14,"area":"agents"},{"id":"2604.11806","title":"Detecting Safety Violations Across Many Agent Traces","created":"2026-04-13","cited_by_count":14,"area":null},{"id":"2605.12673","title":"Do Androids Dream of Breaking the Game? Systematically Auditing AI Agent Benchmarks with BenchJack","created":"2026-05-12","cited_by_count":14,"area":null},{"id":"2511.11584","title":"Output Supervision Can Obfuscate the Chain of Thought","created":"2025-10-11","cited_by_count":13,"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":"2512.18160","title":"Propose, Solve, Verify: Self-Play Through Formal Verification","created":"2025-12-20","cited_by_count":13,"area":"synth"},{"id":"2601.21571","title":"Shaping capabilities with token-level data filtering","created":"2026-01-29","cited_by_count":13,"area":null},{"id":"2602.02639","title":"A Positive Case for Faithfulness: LLM Self-Explanations Help Predict Model Behavior","created":"2026-02-02","cited_by_count":13,"area":"lr"},{"id":"2505.17767","title":"Position: The Real Barrier to LLM Agent Usability is Agentic ROI","created":"2025-05-23","cited_by_count":12,"area":"agents"},{"id":"2510.27269","title":"Why Do Multilingual Reasoning Gaps Emerge in Reasoning Language Models?","created":"2025-10-31","cited_by_count":12,"area":"lr"},{"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":"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":"2506.10139","title":"Unsupervised Elicitation of Language Models","created":"2025-06-11","cited_by_count":11,"area":null},{"id":"2509.24393","title":"Towards Safe Reasoning in Large Reasoning Models via Corrective Intervention","created":"2025-09-29","cited_by_count":11,"area":"lr"},{"id":"2602.01750","title":"Adversarial Reward Auditing for Active Detection and Mitigation of Reward Hacking","created":"2026-02-02","cited_by_count":11,"area":"alignment"},{"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":"2506.20702","title":"The Singapore Consensus on Global AI Safety Research Priorities","created":"2025-06-25","cited_by_count":10,"area":null},{"id":"2507.15886","title":"Combining Cost-Constrained Runtime Monitors for AI Safety","created":"2025-07-19","cited_by_count":10,"area":null},{"id":"2601.05752","title":"AutoMonitor-Bench: Evaluating the Reliability of LLM-Based Misbehavior Monitor","created":"2026-01-09","cited_by_count":10,"area":null},{"id":"2602.18297","title":"Analyzing and Improving Chain-of-Thought Monitorability Through Information Theory","created":"2026-02-20","cited_by_count":10,"area":"lr"},{"id":"2603.00829","title":"Constitutional Black-Box Monitoring for Scheming in LLM Agents","created":"2026-02-28","cited_by_count":10,"area":"agents"},{"id":"2605.20744","title":"Hack-Verifiable Environments: Towards Evaluating Reward Hacking at Scale","created":"2026-05-20","cited_by_count":10,"area":"alignment"},{"id":"2504.05294","title":"Truthful or Fabricated? Using Causal Attribution to Mitigate Reward Hacking in Explanations","created":"2025-04-07","cited_by_count":9,"area":"lr"},{"id":"2505.13787","title":"Preference Learning with Lie Detectors can Induce Honesty or Evasion","created":"2025-05-20","cited_by_count":9,"area":"alignment"},{"id":"2505.17244","title":"ReasoningShield: Safety Detection over Reasoning Traces of Large Reasoning Models","created":"2025-05-22","cited_by_count":9,"area":"lr"},{"id":"2506.23678","title":"Interactive Reasoning: Visualizing and Controlling Chain-of-Thought Reasoning in Large Language Models","created":"2025-06-30","cited_by_count":9,"area":"lr"},{"id":"2508.15842","title":"Lexical Hints of Accuracy in LLM Reasoning Chains","created":"2025-08-19","cited_by_count":9,"area":"lr"},{"id":"2509.07339","title":"Performative Thinking? The Brittle Correlation Between CoT Length and Problem Complexity","created":"2025-09-09","cited_by_count":9,"area":"lr"},{"id":"2510.27338","title":"Reasoning Models Sometimes Output Illegible Chains of Thought","created":"2025-10-31","cited_by_count":9,"area":"lr"},{"id":"2511.08525","title":"Investigating CoT Monitorability in Large Reasoning Models","created":"2025-11-11","cited_by_count":9,"area":"lr"},{"id":"2511.21399","title":"Steering Awareness: Detecting Activation Steering from Within","created":"2025-11-26","cited_by_count":9,"area":"mechinterp"},{"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":"2601.00514","title":"The Illusion of Insight in Reasoning Models","created":"2026-01-02","cited_by_count":9,"area":"lr"},{"id":"2604.04902","title":"Are Latent Reasoning Models Easily Interpretable?","created":"2026-04-06","cited_by_count":9,"area":"lr"},{"id":"2608.09867","title":"Stealing Reasoning Traces from Proprietary LLM APIs","created":"2026-08-10","cited_by_count":9,"area":"lr"},{"id":"2510.01925","title":"Enhancing Large Language Model Reasoning with Reward Models: An Analytical Survey","created":"2025-10-02","cited_by_count":8,"area":"lr"}],"next_cursor":"100"}