{"total":88,"rows":[{"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":"2509.02350","title":"Implicit Reasoning in Large Language Models: A Comprehensive Survey","created":"2025-09-02","cited_by_count":43,"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":"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":"2603.05706","title":"Reasoning Models Struggle to Control their Chains of Thought","created":"2026-03-05","cited_by_count":25,"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":"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":"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":"2510.27484","title":"Thought Branches: Interpreting LLM Reasoning Requires Resampling","created":"2025-10-31","cited_by_count":16,"area":"lr"},{"id":"2507.19218","title":"Technological folie \\`a deux: Feedback Loops Between AI Chatbots and Mental Illness","created":"2025-07-25","cited_by_count":15,"area":null},{"id":"2510.23966","title":"A Pragmatic Way to Measure Chain-of-Thought Monitorability","created":"2025-10-28","cited_by_count":13,"area":"lr"},{"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":"2601.21571","title":"Shaping capabilities with token-level data filtering","created":"2026-01-29","cited_by_count":13,"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":"2602.18297","title":"Analyzing and Improving Chain-of-Thought Monitorability Through Information Theory","created":"2026-02-20","cited_by_count":10,"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":"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":"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":"2603.09786","title":"Quantifying the Necessity of Chain of Thought through Opaque Serial Depth","created":"2026-03-10","cited_by_count":8,"area":"lr"},{"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":"2602.22303","title":"Training Agents to Self-Report Misbehavior","created":"2026-02-25","cited_by_count":6,"area":"agents"},{"id":"2512.12777","title":"State over Tokens: Characterizing the Role of Reasoning Tokens","created":"2025-12-14","cited_by_count":5,"area":"lr"},{"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":"2603.30036","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":"2604.16242","title":"Detecting and Suppressing Reward Hacking with Gradient Fingerprints","created":"2026-04-17","cited_by_count":5,"area":"alignment"},{"id":"2605.18549","title":"Monitoring the Internal Monologue: Probe Trajectories Reveal Reasoning Dynamics","created":"2026-05-18","cited_by_count":5,"area":"lr"},{"id":"2512.01222","title":"Unsupervised decoding of encoded reasoning using language model interpretability","created":"2025-12-01","cited_by_count":4,"area":"lr"},{"id":"2602.01148","title":"Capabilities and Fundamental Limits of Latent Chain-of-Thought","created":"2026-02-01","cited_by_count":4,"area":"lr"},{"id":"2602.03978","title":"Monitorability as a Free Gift: How RLVR Spontaneously Aligns Reasoning","created":"2026-02-03","cited_by_count":4,"area":"lr"},{"id":"2510.01069","title":"Typed Chain-of-Thought: A Curry-Howard Framework for Verifying LLM Reasoning","created":"2025-10-01","cited_by_count":3,"area":"lr"},{"id":"2511.15613","title":"When to Think and When to Look: Uncertainty-Guided Lookback","created":"2025-11-19","cited_by_count":3,"area":"lr"},{"id":"2512.00742","title":"On the Regulatory Potential of User Interfaces for AI Agent Governance","created":"2025-11-30","cited_by_count":3,"area":null},{"id":"2604.03121","title":"An Independent Safety Evaluation of Kimi K2.5","created":"2026-04-03","cited_by_count":3,"area":"agents"},{"id":"2604.27865","title":"KellyBench: A Benchmark for Long-Horizon Sequential Decision Making","created":"2026-04-30","cited_by_count":3,"area":null},{"id":"2605.15257","title":"Training on Documents About Monitoring Leads to CoT Obfuscation","created":"2026-05-14","cited_by_count":3,"area":"lr"},{"id":"2606.20560","title":"How Transparent is DiffusionGemma?","created":"2026-06-18","cited_by_count":3,"area":"lr"},{"id":"2606.25013","title":"Do Thinking Tokens Help with Safety?","created":"2026-06-23","cited_by_count":3,"area":"lr"},{"id":"2607.14345","title":"Value Leakage: An LLM's Answers Are Silently Shaped by Its Own Values","created":"2026-07-15","cited_by_count":3,"area":null},{"id":"2608.23875","title":"AI Finds A Way","created":"2026-08-24","cited_by_count":3,"area":null},{"id":"2510.17057","title":"The Ends Justify the Thoughts: RL-Induced Motivated Reasoning in LLM CoTs","created":"2025-10-20","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":"2510.22362","title":"Mapping Faithful Reasoning in Language Models","created":"2025-10-25","cited_by_count":2,"area":"lr"},{"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":"2604.06427","title":"The Depth Ceiling: On the Limits of Large Language Models in Discovering Latent Planning","created":"2026-04-07","cited_by_count":2,"area":"lr"},{"id":"2605.11746","title":"When Reasoning Traces Become Performative: Step-Level Evidence that Chain-of-Thought Is an Imperfect Oversight Channel","created":"2026-05-12","cited_by_count":2,"area":"lr"},{"id":"2606.11172","title":"Predicting Future Behaviors in Reasoning Models Enables Better Steering","created":"2026-06-09","cited_by_count":2,"area":"lr"},{"id":"2607.03502","title":"Reading Between the Dots: Decoding Hidden Computation across Filler Tokens","created":"2026-07-03","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":"2608.04735","title":"Chain-of-Thought Monitoring Can Be Unreliable in Implicit-Influence Settings","created":"2026-08-05","cited_by_count":2,"area":"lr"},{"id":"2512.08645","title":"Chain-of-Image Generation: Toward Monitorable and Controllable Image Generation","created":"2025-12-09","cited_by_count":1,"area":null},{"id":"2602.15332","title":"Directional Reasoning Trajectory Change (DRTC): Identifying Critical Trace Segments in Reasoning Models","created":"2026-02-17","cited_by_count":1,"area":"lr"},{"id":"2603.12397","title":"Not Just the Destination, But the Journey: Reasoning Traces Causally Shape Generalization Behaviors","created":"2026-03-12","cited_by_count":1,"area":"lr"},{"id":"2603.12717","title":"Altered Thoughts, Altered Actions: Probing Chain-of-Thought Vulnerabilities in VLA Robotic Manipulation","created":"2026-03-13","cited_by_count":1,"area":"lr"},{"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":"2605.27901","title":"The Fragility of Chain-of-Thought Monitoring Across Typologically Diverse Languages","created":"2026-05-27","cited_by_count":1,"area":"lr"},{"id":"2605.28742","title":"CORE: Contrastive Reflection Enables Rapid Improvements in Reasoning","created":"2026-05-27","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":"2607.09786","title":"Length Penalties Make Chain-of-Thought Less Monitorable","created":"2026-07-08","cited_by_count":1,"area":"lr"},{"id":"2608.21664","title":"Measuring Activation Control in Large Language Models","created":"2026-08-21","cited_by_count":1,"area":null},{"id":"2608.29464","title":"Chain-of-Thought Faithfulness of Reasoning Models Varies with Where and How Preference Cues Are Delivered","created":"2026-08-29","cited_by_count":1,"area":"lr"},{"id":"2609.09030","title":"Answer-Distribution Trajectories: A Stochastic-Dynamics View of LLM Reasoning","created":"2026-09-08","cited_by_count":1,"area":"lr"},{"id":"2602.08877","title":"Stress-Testing Alignment Audits With Prompt-Level Strategic Deception","created":"2026-02-09","cited_by_count":0,"area":"safety"},{"id":"2602.14095","title":"NEST: Nascent Encoded Steganographic Thoughts","created":"2026-02-15","cited_by_count":0,"area":"lr"},{"id":"2602.17676","title":"Epistemic Traps: Rational Misalignment Driven by Model Misspecification","created":"2026-01-27","cited_by_count":0,"area":null},{"id":"2603.21376","title":"A transformer architecture alteration to incentivise externalised reasoning","created":"2026-03-22","cited_by_count":0,"area":"lr"},{"id":"2604.15726","title":"LLM Reasoning Is Latent, Not the Chain of Thought","created":"2026-04-17","cited_by_count":0,"area":"lr"},{"id":"2605.12746","title":"CoT-Guard: Small Models for Strong Monitoring","created":"2026-05-12","cited_by_count":0,"area":"lr"},{"id":"2608.00583","title":"A False Average: Pooled CoT-Monitor Accuracy Conceals a Reasoning-Dependent Fragility","created":"2026-08-01","cited_by_count":0,"area":"lr"},{"id":"2608.16213","title":"Process-Constituted Intelligence: A Shared Criterion for Humans and Machines","created":"2026-08-17","cited_by_count":0,"area":null},{"id":"2609.02754","title":"Untangling the Mechanisms of Misleading Context in Medical Question Answering","created":"2026-09-02","cited_by_count":0,"area":"lr"},{"id":"2609.04194","title":"Legibility is Not Interpretability: Comparing Judged and Actual Importance in Chain-Of-Thought Reasoning","created":"2026-09-03","cited_by_count":0,"area":"lr"},{"id":"2609.15989","title":"Corrupt Plans, Clean Traces: Evading Chain-of-Thought Monitoring with Plan Injection","created":"2026-09-14","cited_by_count":0,"area":"lr"},{"id":"2609.24243","title":"Taming CoT Obfuscation in VLMs: From Mechanistic Evidence to Activation Enforcement","created":"2026-09-21","cited_by_count":0,"area":"lr"},{"id":"2609.27542","title":"Control-Token Injection Suppresses Chain-of-Thought and Defeats Reasoning-Based Oversight in Tool-Using Agents","created":"2026-09-23","cited_by_count":0,"area":"agents"},{"id":"2609.30217","title":"Instrumental Monitor Evasion Emerges Under Ordinary Task Pressure","created":"2026-09-24","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":"2609.33220","title":"When Do Models Admit They Are Wrong? Failure Disclosure Is Unstable Under Reinforcement Learning","created":"2026-09-27","cited_by_count":0,"area":"lr"},{"id":"2609.35074","title":"When Confidence Rises Too Early: Detecting Shortcut Reasoning via Premature Answer Commitment","created":"2026-09-28","cited_by_count":0,"area":"lr"},{"id":"2609.37312","title":"Hidden Reasoning Must Leak, but Need Not Be Readable: Fundamental Opportunities and Limits for Chain-of-Thought Monitoring","created":"2026-09-29","cited_by_count":0,"area":"lr"},{"id":"2609.38107","title":"Correct Answers, Invalid Traces: What Verifiable Grade-School Math Reveals About Chain-of-Thought Traces","created":"2026-09-29","cited_by_count":0,"area":"lr"},{"id":"2609.39838","title":"Learning Steganography Is Easy, Learning Steganographic Reasoning Is Hard","created":"2026-09-30","cited_by_count":0,"area":"lr"},{"id":"2610.04737","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"}],"next_cursor":null}