{"total":136,"rows":[{"id":"2607.15495","title":"Verbalizable Representations Form a Global Workspace in Language Models","created":"2026-07-16","cited_by_count":75,"area":null},{"id":"2604.06132","title":"Claw-Eval: Towards Trustworthy Evaluation of Autonomous Agents","created":"2026-04-07","cited_by_count":64,"area":"agents"},{"id":"2604.07729","title":"Emotion Concepts and their Function in a Large Language Model","created":"2026-04-09","cited_by_count":57,"area":null},{"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":"2605.02964","title":"Reward Hacking Benchmark: Measuring Exploits in LLM Agents with Tool Use","created":"2026-05-03","cited_by_count":31,"area":"agents"},{"id":"2604.15149","title":"LLMs Gaming Verifiers: RLVR can Lead to Reward Hacking","created":"2026-04-16","cited_by_count":28,"area":"lr"},{"id":"2605.02087","title":"Model Spec Midtraining: Improving How Alignment Training Generalizes","created":"2026-05-03","cited_by_count":20,"area":null},{"id":"2601.20103","title":"Benchmarking Reward Hack Detection in Code Environments via Contrastive Analysis","created":"2026-01-27","cited_by_count":18,"area":"alignment"},{"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":"2604.25891","title":"Conditional misalignment: common interventions can hide emergent misalignment behind contextual triggers","created":"2026-04-28","cited_by_count":17,"area":null},{"id":"2512.04864","title":"Are Your Agents Upward Deceivers?","created":"2025-12-04","cited_by_count":16,"area":"agents"},{"id":"2604.17596","title":"Terminal Wrench: A Dataset of 331 Reward-Hackable Environments and 3,632 Exploit Trajectories","created":"2026-04-19","cited_by_count":15,"area":null},{"id":"2512.19027","title":"Recontextualization Mitigates Specification Gaming without Modifying the Specification","created":"2025-12-22","cited_by_count":14,"area":null},{"id":"2603.01608","title":"Evaluating and Understanding Scheming Propensity in LLM Agents","created":"2026-03-02","cited_by_count":14,"area":"agents"},{"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":"2604.28139","title":"Claw-Eval-Live: A Live Agent Benchmark for Evolving Real-World Workflows","created":"2026-04-30","cited_by_count":13,"area":"agents"},{"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":"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":"2605.20744","title":"Hack-Verifiable Environments: Towards Evaluating Reward Hacking at Scale","created":"2026-05-20","cited_by_count":10,"area":"alignment"},{"id":"2603.03992","title":"Measuring AI R&D Automation","created":"2026-03-04","cited_by_count":9,"area":null},{"id":"2604.22167","title":"Estimating Tail Risks in Language Model Output Distributions","created":"2026-04-24","cited_by_count":9,"area":"safety"},{"id":"2605.12798","title":"Emergent and Subliminal Misalignment Through the Lens of Data-Mediated Transfer","created":"2026-05-12","cited_by_count":8,"area":null},{"id":"2606.23075","title":"Safety in Self-Evolving LLM Agent Systems: Threats, Amplification, and Case Studies","created":"2026-06-22","cited_by_count":8,"area":"agents"},{"id":"2602.05910","title":"Chunky Post-Training: Data Driven Failures of Generalization","created":"2026-02-05","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":"2604.21098","title":"Propensity Inference: Environmental Contributors to LLM Behaviour","created":"2026-04-22","cited_by_count":7,"area":null},{"id":"2602.16931","title":"Narrow Fine-Tuning Erodes Safety Alignment in Vision-Language Agents","created":"2026-02-18","cited_by_count":6,"area":"mllm"},{"id":"2602.22303","title":"Training Agents to Self-Report Misbehavior","created":"2026-02-25","cited_by_count":6,"area":"agents"},{"id":"2608.12253","title":"One Frozen Simulator Is Not Enough: Simulator Collapse in Multi-Agent RL","created":"2026-08-12","cited_by_count":6,"area":"agents"},{"id":"2601.23081","title":"Character as a Latent Variable in Large Language Models: A Mechanistic Account of Emergent Misalignment and Conditional Safety Failures","created":"2026-01-30","cited_by_count":5,"area":"safety"},{"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":"2602.14697","title":"Evolutionary System Prompt Learning for Reinforcement Learning in LLMs","created":"2026-02-16","cited_by_count":5,"area":"agents"},{"id":"2602.18920","title":"DeepInnovator: Triggering the Innovative Capabilities of LLMs","created":"2026-02-21","cited_by_count":5,"area":"agents"},{"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":"2604.16242","title":"Detecting and Suppressing Reward Hacking with Gradient Fingerprints","created":"2026-04-17","cited_by_count":5,"area":"alignment"},{"id":"2606.16062","title":"Auditing Reward Hackability in Code RL Training Environments","created":"2026-06-14","cited_by_count":5,"area":null},{"id":"2602.03806","title":"Bridging Online and Offline RL: Contextual Bandit Learning for Multi-Turn Code Generation","created":"2026-02-03","cited_by_count":4,"area":"offlinerl"},{"id":"2603.04426","title":"Delta-Crosscoder: Robust Crosscoder Model Diffing in Narrow Fine-Tuning Regimes","created":"2026-02-16","cited_by_count":4,"area":null},{"id":"2604.13076","title":"Alignment midtraining for animals","created":"2026-03-21","cited_by_count":4,"area":null},{"id":"2605.16282","title":"Taxonomy and Consistency Analysis of Safety Benchmarks for AI Agents","created":"2026-04-11","cited_by_count":4,"area":"agents"},{"id":"2605.26438","title":"LURE: Live-Usage Replay Evaluations for Reducing Evaluation Awareness","created":"2026-04-08","cited_by_count":4,"area":null},{"id":"2605.28591","title":"Models That Know How Evaluations Are Designed Score Safer","created":"2026-05-27","cited_by_count":4,"area":null},{"id":"2606.05647","title":"Coding with \"Enemy\": Can Human Developers Detect AI Agent Sabotage?","created":"2026-06-04","cited_by_count":4,"area":"agents"},{"id":"2606.24014","title":"Reinforcement Learning Towards Broadly and Persistently Beneficial Models","created":"2026-06-22","cited_by_count":4,"area":"alignment"},{"id":"2607.11022","title":"When the Reward Suite Is Leaky: A Preregistered Causal Contrast of Natural Verifier False Positives in RLVR","created":"2026-07-13","cited_by_count":4,"area":null},{"id":"2608.17776","title":"Debate Training Reduces Reward Hacking in RLAIF","created":"2026-08-18","cited_by_count":4,"area":"alignment"},{"id":"2609.20812","title":"Quantifying Overclaiming Propensity in Frontier LLM Agents","created":"2026-09-17","cited_by_count":4,"area":"agents"},{"id":"2601.22510","title":"Shattered Compositionality: Counterintuitive Learning Dynamics of Transformers for Arithmetic","created":"2026-01-30","cited_by_count":3,"area":null},{"id":"2603.01204","title":"Subliminal Signals in Preference Labels","created":"2026-03-01","cited_by_count":3,"area":"alignment"},{"id":"2604.11072","title":"Hodoscope: Unsupervised Monitoring for AI Misbehaviors","created":"2026-04-13","cited_by_count":3,"area":null},{"id":"2604.16423","title":"Shifting the Gradient: Understanding How Defensive Training Methods Protect Language Model Integrity","created":"2026-04-03","cited_by_count":3,"area":null},{"id":"2604.28093","title":"What Makes a Good Terminal-Agent Benchmark Task: A Guideline for Adversarial, Difficult, and Legible Evaluation Design","created":"2026-04-30","cited_by_count":3,"area":"agents"},{"id":"2605.02269","title":"Towards Understanding Specification Gaming in Reasoning Models","created":"2026-05-04","cited_by_count":3,"area":"lr"},{"id":"2606.19624","title":"MassSpecGym in the Wild: Uncovering and Correcting Evaluation Pitfalls in AI-Driven Molecule Discovery","created":"2026-06-17","cited_by_count":3,"area":null},{"id":"2606.30252","title":"Inoculation Adapters: Improved Selective Generalization of Capabilities with Fewer Surprising Backdoors","created":"2026-06-29","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":"2606.06667","title":"The Piggyback Hypothesis of Generalization: Explaining and Mitigating Emergent Misalignment","created":"2026-06-04","cited_by_count":2,"area":null},{"id":"2607.18639","title":"Mark, Don't Erase: Token Inoculation for Dual-Use Knowledge in LLMs","created":"2026-07-21","cited_by_count":2,"area":"safety"},{"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":"2609.24967","title":"Emergent Collusion in Long-Horizon LLM Agent Interaction","created":"2026-09-21","cited_by_count":2,"area":"agents"},{"id":"2609.28614","title":"Reward Hacking Challenges Oversight of Autonomous Research Agents","created":"2026-09-23","cited_by_count":2,"area":"agents"},{"id":"2602.14844","title":"Interactionless Inverse Reinforcement Learning: A Data-Centric Framework for Durable Alignment","created":"2026-02-16","cited_by_count":1,"area":"offlinerl"},{"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":"2603.02225","title":"Scaling Reward Modeling without Human Supervision","created":"2026-02-11","cited_by_count":1,"area":"alignment"},{"id":"2603.10009","title":"Personalized Group Relative Policy Optimization for Heterogenous Preference Alignment","created":"2026-02-17","cited_by_count":1,"area":"lr"},{"id":"2603.11356","title":"Resolving Java Code Repository Issues with iSWE Agent","created":"2026-03-11","cited_by_count":1,"area":null},{"id":"2603.25861","title":"Why Safety Probes Catch Liars But Miss Fanatics","created":"2026-03-26","cited_by_count":1,"area":null},{"id":"2604.11477","title":"OOM-RL: Out-of-Money Reinforcement Learning Market-Driven Alignment for LLM-Based Multi-Agent Systems","created":"2026-04-13","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":"2604.28082","title":"Characterizing the Consistency of the Emergent Misalignment Persona","created":"2026-04-30","cited_by_count":1,"area":null},{"id":"2605.12850","title":"Persona-Model Collapse in Emergent Misalignment","created":"2026-05-13","cited_by_count":1,"area":"synth"},{"id":"2605.17958","title":"Enhancing the Code Reasoning Capabilities of LLMs via Consistency-based Reinforcement Learning","created":"2026-05-18","cited_by_count":1,"area":"lr"},{"id":"2606.04413","title":"(Mis)generalization of Helpful-only Fine-tuning","created":"2026-06-03","cited_by_count":1,"area":null},{"id":"2606.07602","title":"Sample-Efficient Post-Training for LEGO Spatial-Physics Reasoning","created":"2026-05-29","cited_by_count":1,"area":"lr"},{"id":"2606.22859","title":"AI Scientists as Engines of Discovery: A Case for Development within Reformed Institutions","created":"2026-06-22","cited_by_count":1,"area":"agents"},{"id":"2606.23700","title":"Self-Recognition Finetuning can Prevent and Reverse Emergent Misalignment","created":"2026-06-04","cited_by_count":1,"area":"safety"},{"id":"2606.24251","title":"Probing the Misaligned Thinking Process of Language Models","created":"2026-06-23","cited_by_count":1,"area":"lr"},{"id":"2606.26071","title":"Model Forensics: Investigating Whether Concerning Behavior Reflects Misalignment","created":"2026-06-24","cited_by_count":1,"area":null},{"id":"2607.01567","title":"Scaling Trends for Lie Detector Oversight in Preference Learning","created":"2026-07-02","cited_by_count":1,"area":null},{"id":"2607.02755","title":"Out-of-Distribution Generalization of Risk Aversion in Language Models","created":"2026-07-02","cited_by_count":1,"area":null},{"id":"2607.03640","title":"Revealing Hidden Model Behaviors with Task-Specific Self-Reports","created":"2026-07-03","cited_by_count":1,"area":null},{"id":"2607.09766","title":"Norm Enforcement for AI Agents: Robustly Shaping Behavior in Multi-Agent Systems","created":"2026-07-07","cited_by_count":1,"area":"agents"},{"id":"2607.18575","title":"RECEIPT: Deterministic, Reward-Hacking-Resistant Verification for White-Box Agentic XSS Discovery","created":"2026-07-20","cited_by_count":1,"area":"agents"},{"id":"2607.22676","title":"How LLM Task-Adaptation Reshapes Alignment: A Multi-dimensional Study of Behavioral and Representational Drift","created":"2026-07-10","cited_by_count":1,"area":null},{"id":"2607.23524","title":"Delegation Intelligence in Deep Search: A Controllable Framework for Disentangled Capability Diagnosis","created":"2026-07-26","cited_by_count":1,"area":null},{"id":"2607.26173","title":"Shared SFT Lessons Across Alignment, Model Organisms, and Toy Models","created":"2026-07-28","cited_by_count":1,"area":null},{"id":"2608.13482","title":"Synthetic Persona Pretraining: Alignment from Token Zero","created":"2026-08-13","cited_by_count":1,"area":null},{"id":"2609.13236","title":"Self-Evolving AI for Humanoids: Mechanisms, Safety, and Evaluation of Post-Deployment Self-Improvement","created":"2026-09-02","cited_by_count":1,"area":"wm"},{"id":"2609.34262","title":"Maintaining Benchmarks Against Increasingly Capable Agents: Detection and Remediation of Unearned Passes","created":"2026-09-28","cited_by_count":1,"area":"agents"},{"id":"2609.37914","title":"The Unequal Influence of Bad Advice: Using Training Data Attribution to Modulate Emergent Misalignment","created":"2026-09-29","cited_by_count":1,"area":null},{"id":"2610.02781","title":"OPD Before RL: Warm-Starting Rubric-Based RL with On-Policy Distillation","created":"2026-10-02","cited_by_count":1,"area":"distill"},{"id":"2610.10947","title":"RH-Detect: A Unified Benchmark for Reward Hacking Detection","created":"2026-10-07","cited_by_count":1,"area":"alignment"},{"id":"2601.03388","title":"Metaphors are a Source of Cross-Domain Misalignment of Large Reasoning Models","created":"2026-01-06","cited_by_count":0,"area":"lr"},{"id":"2602.02304","title":"Comparing Explanations is Not Enough, Explain the Change: New Standards are Needed to Explain Behavioral Shifts in Large Language Models","created":"2026-02-02","cited_by_count":0,"area":"alignment"},{"id":"2602.12124","title":"Alignment Risks from Capability-Seeking RL Training","created":"2026-02-12","cited_by_count":0,"area":"safety"},{"id":"2602.15338","title":"Discovering Implicit Large Language Model Alignment Objectives","created":"2026-02-17","cited_by_count":0,"area":"alignment"},{"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":"2603.26846","title":"Stable Reasoning, Unstable Responses: Mitigating LLM Deception via Stability Asymmetry","created":"2026-03-27","cited_by_count":0,"area":"lr"},{"id":"2605.02175","title":"Intervention Complexity as a Canonical Reward and a Measure of Intelligence","created":"2026-05-04","cited_by_count":0,"area":null},{"id":"2605.23565","title":"Understanding Goal Generalisation in Sequential Reinforcement Learning","created":"2026-05-22","cited_by_count":0,"area":"deeprl"}],"next_cursor":"100"}