{"total":149,"rows":[{"id":"2210.10760","title":"Scaling Laws for Reward Model Overoptimization","created":"2022-10-19","cited_by_count":1095,"area":"alignment"},{"id":"2307.15217","title":"Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback","created":"2023-07-27","cited_by_count":759,"area":"alignment"},{"id":"2204.06745","title":"GPT-NeoX-20B: An Open-Source Autoregressive Language Model","created":"2022-04-14","cited_by_count":713,"area":null},{"id":"2206.05802","title":"Self-critiquing models for assisting human evaluators","created":"2022-06-12","cited_by_count":380,"area":null},{"id":"2109.13916","title":"Unsolved Problems in ML Safety","created":"2021-09-28","cited_by_count":292,"area":null},{"id":"2209.00626","title":"The Alignment Problem from a Deep Learning Perspective","created":"2022-08-30","cited_by_count":258,"area":null},{"id":"2209.13085","title":"Defining and Characterizing Reward Hacking","created":"2022-09-27","cited_by_count":203,"area":"alignment"},{"id":"2403.19159","title":"Disentangling Length from Quality in Direct Preference Optimization","created":"2024-03-28","cited_by_count":196,"area":"alignment"},{"id":"2406.02900","title":"Scaling Laws for Reward Model Overoptimization in Direct Alignment Algorithms","created":"2024-06-05","cited_by_count":143,"area":"alignment"},{"id":"2110.06674","title":"Truthful AI: Developing and governing AI that does not lie","created":"2021-10-13","cited_by_count":113,"area":null},{"id":"2206.13353","title":"Is Power-Seeking AI an Existential Risk?","created":"2022-06-16","cited_by_count":105,"area":null},{"id":"2405.06624","title":"Towards Guaranteed Safe AI: A Framework for Ensuring Robust and Reliable AI Systems","created":"2024-05-10","cited_by_count":96,"area":"wm"},{"id":"2102.03896","title":"Consequences of Misaligned AI","created":"2021-02-07","cited_by_count":80,"area":null},{"id":"2403.03185","title":"Correlated Proxies: A New Definition and Improved Mitigation for Reward Hacking","created":"2024-03-05","cited_by_count":50,"area":"alignment"},{"id":"2310.09144","title":"Goodhart's Law in Reinforcement Learning","created":"2023-10-13","cited_by_count":41,"area":null},{"id":"2409.15268","title":"Style Outweighs Substance: Failure Modes of LLM Judges in Alignment Benchmarking","created":"2024-09-23","cited_by_count":35,"area":"alignment"},{"id":"2410.05584","title":"Rethinking Reward Model Evaluation: Are We Barking up the Wrong Tree?","created":"2024-10-08","cited_by_count":31,"area":"lr"},{"id":"2309.16155","title":"The Trickle-down Impact of Reward (In-)consistency on RLHF","created":"2023-09-28","cited_by_count":29,"area":"alignment"},{"id":"2605.21384","title":"SpecBench: Measuring Reward Hacking in Long-Horizon Coding Agents","created":"2026-05-20","cited_by_count":28,"area":"agents"},{"id":"2209.13020","title":"Law Informs Code: A Legal Informatics Approach to Aligning Artificial Intelligence with Humans","created":"2022-09-14","cited_by_count":23,"area":null},{"id":"2512.16301","title":"Adaptation of Agentic AI: A Survey of Post-Training, Memory, and Skills","created":"2025-12-18","cited_by_count":22,"area":"agents"},{"id":"2202.09519","title":"The four-fifths rule is not disparate impact: a woeful tale of epistemic trespassing in algorithmic fairness","created":"2022-02-19","cited_by_count":21,"area":null},{"id":"2405.03207","title":"A Philosophical Introduction to Language Models - Part II: The Way Forward","created":"2024-05-06","cited_by_count":18,"area":null},{"id":"2505.12763","title":"Rethinking Reward Model Evaluation Through the Lens of Reward Overoptimization","created":"2025-05-19","cited_by_count":11,"area":"alignment"},{"id":"2211.07280","title":"A taxonomic system for failure cause analysis of open source AI incidents","created":"2022-11-14","cited_by_count":9,"area":null},{"id":"2410.09638","title":"On Goodhart's law, with an application to value alignment","created":"2024-10-12","cited_by_count":9,"area":null},{"id":"2603.12520","title":"When LLM Judge Scores Look Good but Best-of-N Decisions Fail","created":"2026-03-12","cited_by_count":9,"area":null},{"id":"2310.18244","title":"A Review of the Evidence for Existential Risk from AI via Misaligned Power-Seeking","created":"2023-10-27","cited_by_count":8,"area":null},{"id":"2606.26300","title":"The Verification Horizon: No Silver Bullet for Coding Agent Rewards","created":"2026-06-24","cited_by_count":8,"area":"agents"},{"id":"2408.13406","title":"Collaboration Dynamics and Reliability Challenges of Multi-Agent LLM Systems in Finite Element Analysis","created":"2024-08-23","cited_by_count":6,"area":"agents"},{"id":"2508.14927","title":"AI Testing Should Account for Sophisticated Strategic Behaviour","created":"2025-08-19","cited_by_count":6,"area":null},{"id":"2607.14004","title":"Do Agent Optimizers Compound? A Continual-Learning Evaluation on Terminal-Bench 2.0","created":"2026-07-15","cited_by_count":6,"area":null},{"id":"2607.22368","title":"Do Agent Benchmarks Measure Capability? Protocol Validity in the Age of Agentic AI","created":"2026-07-24","cited_by_count":6,"area":"agents"},{"id":"2310.14455","title":"An International Consortium for Evaluations of Societal-Scale Risks from Advanced AI","created":"2023-10-22","cited_by_count":5,"area":null},{"id":"2405.04048","title":"Philosophy of Cognitive Science in the Age of Deep Learning","created":"2024-05-07","cited_by_count":5,"area":null},{"id":"2509.05381","title":"Murphys Laws of AI Alignment: Why the Gap Always Wins","created":"2025-09-04","cited_by_count":5,"area":"alignment"},{"id":"2605.07728","title":"SARC: A Governance-by-Architecture Framework for Agentic AI Systems","created":"2026-05-08","cited_by_count":5,"area":"agents"},{"id":"2208.05320","title":"PEPPER: Empowering User-Centric Recommender Systems over Gossip Learning","created":"2022-08-09","cited_by_count":4,"area":null},{"id":"2305.19223","title":"Intent-aligned AI systems deplete human agency: the need for agency foundations research in AI safety","created":"2023-05-30","cited_by_count":4,"area":null},{"id":"2409.19024","title":"Elephant in the Room: Unveiling the Impact of Reward Model Quality in Alignment","created":"2024-09-26","cited_by_count":4,"area":"alignment"},{"id":"2411.15951","title":"Partial Identifiability and Misspecification in Inverse Reinforcement Learning","created":"2024-11-24","cited_by_count":4,"area":"offlinerl"},{"id":"2502.04675","title":"Scalable Oversight for Superhuman AI via Recursive Self-Critiquing","created":"2025-02-07","cited_by_count":4,"area":"alignment"},{"id":"2606.28430","title":"Building to the Test: Coding Agents Deliver What You Check, Not What You Requested","created":"2026-06-26","cited_by_count":4,"area":"agents"},{"id":"2601.16909","title":"Preventing the Collapse of Peer Review Requires Verification-First AI","created":"2026-01-23","cited_by_count":3,"area":null},{"id":"2603.08371","title":"Leaderboard Incentives: Model Rankings under Strategic Post-Training","created":"2026-03-09","cited_by_count":3,"area":null},{"id":"2605.17829","title":"Interactive Evaluation Requires a Design Science","created":"2026-05-18","cited_by_count":3,"area":null},{"id":"2606.04075","title":"Large Language Models Hack Rewards, and Society","created":"2026-06-02","cited_by_count":3,"area":null},{"id":"2609.02246","title":"LLM-as-a-Judge Is Not an Oracle: Why Self-Improving Agents Need Deterministic Guardrails","created":"2026-09-02","cited_by_count":3,"area":"agents"},{"id":"2609.04665","title":"Harness-agnostic detection and immunization of reward hacking in self-evolving language models","created":"2026-09-04","cited_by_count":3,"area":"alignment"},{"id":"2305.15299","title":"Science in the Era of ChatGPT, Large Language Models and Generative AI: Challenges for Research Ethics and How to Respond","created":"2023-05-24","cited_by_count":2,"area":null},{"id":"2503.07341","title":"The Economics of p(doom): Scenarios of Existential Risk and Economic Growth in the Age of Transformative AI","created":"2025-03-10","cited_by_count":2,"area":null},{"id":"2503.12301","title":"One Goal, Many Challenges: Robust Preference Optimization Amid Content-Aware and Multi-Source Noise","created":"2025-03-16","cited_by_count":2,"area":"alignment"},{"id":"2505.23445","title":"The Strong, Weak and Benign Goodhart's law. An independence-free and paradigm-agnostic formalisation","created":"2025-05-29","cited_by_count":2,"area":null},{"id":"2510.03231","title":"Reward Models are Metrics in a Trench Coat","created":"2025-10-03","cited_by_count":2,"area":"alignment"},{"id":"2511.06148","title":"Large Language Models Develop Novel Social Biases Through Adaptive Exploration","created":"2025-11-08","cited_by_count":2,"area":null},{"id":"2512.03244","title":"ScalePRM: Training Process Reward Models by Scaling Verification Compute Without Ground Truth","created":"2025-12-02","cited_by_count":2,"area":"lr"},{"id":"2601.03948","title":"Trade-R1: Bridging Verifiable Rewards to Stochastic Environments via Process-Level Reasoning Verification","created":"2026-01-07","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":"2603.04861","title":"Causally Robust Reward Learning from Reason-Augmented Preference Feedback","created":"2026-03-05","cited_by_count":2,"area":"alignment"},{"id":"2603.24126","title":"Likelihood Hacking in Probabilistic Program Synthesis","created":"2026-03-25","cited_by_count":2,"area":null},{"id":"2605.06390","title":"Automated alignment is harder than you think","created":"2026-05-07","cited_by_count":2,"area":null},{"id":"2605.07671","title":"The Endogeneity of Miscalibration: Impossibility and Escape in Scored Reporting","created":"2026-05-08","cited_by_count":2,"area":null},{"id":"2606.00437","title":"EST-PRM: Stress-Testing Process Reward Models Before They Become Load-Bearing","created":"2026-05-30","cited_by_count":2,"area":"lr"},{"id":"2608.13566","title":"Don't Claim Benchmark-Oriented Optimization Improves General Coding Capability -- Diverse Evaluation Is Required","created":"2026-06-26","cited_by_count":2,"area":null},{"id":"2210.07126","title":"Challenges in Explanation Quality Evaluation","created":"2022-10-13","cited_by_count":1,"area":null},{"id":"2512.03048","title":"The Specification Trap: Why Static Value Alignment Alone Is Insufficient for Robust Alignment","created":"2025-11-19","cited_by_count":1,"area":"alignment"},{"id":"2603.15017","title":"Consequentialist Objectives and Catastrophe","created":"2026-03-16","cited_by_count":1,"area":null},{"id":"2604.17769","title":"Reverse Constitutional AI: A Framework for Controllable Toxic Data Generation via Probability-Clamped RLAIF","created":"2026-04-20","cited_by_count":1,"area":"alignment"},{"id":"2604.19752","title":"Soft-Label Governance for Distributional Safety in Multi-Agent Systems","created":"2026-03-19","cited_by_count":1,"area":"agents"},{"id":"2604.26360","title":"Uncertainty-Aware Reward Discounting for Mitigating Reward Hacking","created":"2026-04-29","cited_by_count":1,"area":"alignment"},{"id":"2605.26200","title":"Workflow Closure Is Not Scientific Closure in Auto-Research Systems","created":"2026-05-25","cited_by_count":1,"area":null},{"id":"2606.11417","title":"Signed Compression Progress on a Sealed Audit is Goodhart-Resistant","created":"2026-06-09","cited_by_count":1,"area":null},{"id":"2606.23597","title":"Against Proxy Optimization","created":"2026-06-22","cited_by_count":1,"area":null},{"id":"2607.02134","title":"Coding-agents can replicate scientific machine learning papers","created":"2026-07-02","cited_by_count":1,"area":null},{"id":"2607.08946","title":"Training, Reading, and Editing Legible Transformers","created":"2026-07-09","cited_by_count":1,"area":null},{"id":"2607.28196","title":"Fidelity Is Not Safety: Gently-Compressed LLMs Pass Every Data-Free Quality Guard Yet Invent Procedure Steps in Agentic Execution","created":"2026-07-30","cited_by_count":1,"area":"agents"},{"id":"2608.04896","title":"When Shared Rollouts Fail in Defensive Driving Evaluation: A NAVSIM Score Basis Audit","created":"2026-08-05","cited_by_count":1,"area":null},{"id":"2608.09819","title":"Macaron-V1: Towards Open Continual Learning with Self-Improvement and Mixture-of-LoRA","created":"2026-08-10","cited_by_count":1,"area":null},{"id":"2609.32390","title":"Reward Hacking and Agent Containment Failure: A Monte Carlo Study Based on the 2026 Hugging Face Incident","created":"2026-09-26","cited_by_count":1,"area":null},{"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":"1809.01036","title":"A Roadmap for Robust End-to-End Alignment","created":"2018-09-04","cited_by_count":0,"area":"alignment"},{"id":"2302.07872","title":"Data-Centric Governance","created":"2023-02-14","cited_by_count":0,"area":null},{"id":"2308.05120","title":"Dynamic Model Agnostic Reliability Evaluation of Machine-Learning Methods Integrated in Instrumentation & Control Systems","created":"2023-08-08","cited_by_count":0,"area":null},{"id":"2403.05540","title":"Extinction Risks from AI: Invisible to Science?","created":"2024-02-02","cited_by_count":0,"area":null},{"id":"2407.06098","title":"Epistemological Bias As a Means for the Automated Detection of Injustices in Text","created":"2024-07-08","cited_by_count":0,"area":null},{"id":"2509.02391","title":"Gaming and Cooperation in Federated Learning: What Can Happen and How to Monitor It","created":"2025-09-02","cited_by_count":0,"area":null},{"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":"2602.13934","title":"Why Code, Why Now: An Information-Theoretic Perspective on the Limits of Machine Learning","created":"2026-02-15","cited_by_count":0,"area":"lr"},{"id":"2603.29993","title":"Extending MONA in Camera Dropbox: Reproduction, Learned Approval, and Design Implications for Reward-Hacking Mitigation","created":"2026-03-31","cited_by_count":0,"area":"alignment"},{"id":"2604.04237","title":"Pedagogical Safety in Educational Reinforcement Learning: Formalizing and Detecting Reward Hacking in AI Tutoring Systems","created":"2026-04-05","cited_by_count":0,"area":"alignment"},{"id":"2604.05274","title":"Simulating the Evolution of Alignment and Values in Machine Intelligence","created":"2026-04-07","cited_by_count":0,"area":"alignment"},{"id":"2604.07709","title":"IatroBench: A Pre-Registered Benchmark of Clinical Omission in Language Models","created":"2026-04-09","cited_by_count":0,"area":null},{"id":"2604.16516","title":"Operationalizing Fairness in Text-to-Image Models: A Survey of Bias, Fairness Audits and Mitigation Strategies","created":"2026-04-15","cited_by_count":0,"area":null},{"id":"2604.17989","title":"AIT Academy: Cultivating the Complete Agent with a Confucian Three-Domain Curriculum","created":"2026-04-20","cited_by_count":0,"area":"agents"},{"id":"2605.10956","title":"Acceleration of horizontal numerical advection for atmospheric modeling through surrogate modeling with temporal coarse-graining","created":"2026-04-30","cited_by_count":0,"area":null},{"id":"2605.11496","title":"The Evaluation Differential: When Frontier AI Models Recognise They Are Being Tested","created":"2026-05-12","cited_by_count":0,"area":null},{"id":"2605.14407","title":"Metis AI: The Overlooked Middle Zone Between AI-Native and World-Movers","created":"2026-05-14","cited_by_count":0,"area":null},{"id":"2605.22612","title":"Healthcare LLM Benchmarks Are Only as Good as Their Explicit Assumptions","created":"2026-05-21","cited_by_count":0,"area":null},{"id":"2605.25739","title":"The Behavioral Credibility Trilemma: When Calibrated Autonomy Becomes Impossible","created":"2026-05-25","cited_by_count":0,"area":"deeprl"},{"id":"2605.27914","title":"Does Capability Transfer to Subjective Behavior -- and Would Our Instruments Tell Us? A Self-Evolving, Trust-by-Construction Evaluation Paradigm","created":"2026-05-27","cited_by_count":0,"area":"synth"}],"next_cursor":"100"}