acceptodds

NeurIPS 2026 reviews are up: reviews, scores, rebuttals and meta-reviews for 7,877 papers. See the highest rated

acceptodds: the papers of NeurIPS 2026

44 results for “"reward hacking" OR "specification gaming"” in NeurIPS 2026. clearsort:relevanceratingnewestmost read
[abs]Model Incrimination: Investigating Whether Concerning Behavior Reflects MisalignmentGerson Kroiz, Aditya Singh, Senthooran Rajamanoharan et al.5.00
[PDF]Verifier-Backed Hard Problem Generation for Mathematical ReasoningYuhang Lai, Jiazhan Feng, Yee Whye Teh et al.4.00
[abs]Locality-Controlled OOD Guidance for Robust Regulatory DNA Sequence DesignDabin Jeong, Marwane Bourdim, Mohammad Lotfollahi4.50
[PDF]Visual-ERM: Reward Modeling for Visual EquivalenceZiyu Liu, Shengyuan Ding, Xinyu Fang et al.4.33
[PDF]Step-wise Rubric Rewards for LLM ReasoningWeichu Xie, Haozhe Zhao, Wenpu Liu et al.4.25
[abs]MemeEconomy : Do LLM Agents Trade Ethics for Survival?Syed Nazmus Sakib, Nafiul Haque, Ahnaf Manan et al.4.33
[PDF]Reinforcement World Model Learning for LLM-based AgentsXiao Yu, Baolin Peng, Ruize Xu et al.4.00
[PDF]Training on Documents About Monitoring Leads to CoT ObfuscationReilly Haskins, Bilal Chughtai, Joshua Engels4.33
[PDF]G-Zero: Self-Play for Open-Ended Generation from Zero DataChengsong Huang, Haolin Liu, Tong Zheng et al.4.00
[abs]Mitigating Reward Hacking via Task RepresentationsLillian Sun, Joe Benton, Eric Easley5.33
[PDF]Theoretical Limits of Language Model AlignmentLucas Monteiro Paes, Natalie Mackraz, Barry-John Theobald et al.4.67
[PDF]Reward Hacking in Rubric-Based Reinforcement LearningAnas Mahmoud, MohammadHossein Rezaei, Zihao Wang et al.4.67
[PDF]Two is better than one: A Collapse-free Multi-Reward RLIF Training FrameworkShourov Joarder, Diganta Sikdar, Ahsan H Akash et al.3.67
[PDF]Imperfect World Models are ExploitableLogan M Bhamidipaty, Esmeralda S Whitammer, David Abel et al.4.67
[PDF]Hide to Guide: Learning via Semantic MaskingRuitao Liu, Qinghao Hu, Alex Hu et al.4.67

Privacy