Copertina

Do Synthetic Trajectories Reflect Real Reward Hacking? A Systematic Study on Monitoring In-the-Wild Hacking in Code Generation

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Li, Lichen, Zhou, Hengguang, Liang, Yijun, Zhou, Tianyi, Hsieh, Cho-Jui
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Machine Learning
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Understanding Reward Hacking in Text-to-Image Reinforcement Learning
di: Hong, Yunqi, et al.
Pubblicazione: (2026)

ODIN: Disentangled Reward Mitigates Hacking in RLHF
di: Chen, Lichang, et al.
Pubblicazione: (2024)

Hack-Verifiable Environments: Towards Evaluating Reward Hacking at Scale
di: Roth, Amit, et al.
Pubblicazione: (2026)

Defining and Characterizing Reward Hacking
di: Skalse, Joar, et al.
Pubblicazione: (2022)

R1-Zero's "Aha Moment" in Visual Reasoning on a 2B Non-SFT Model
di: Zhou, Hengguang, et al.
Pubblicazione: (2025)

Countdown-Code: A Testbed for Studying The Emergence and Generalization of Reward Hacking in RLVR
di: Khalifa, Muhammad, et al.
Pubblicazione: (2026)

SoliReward: Mitigating Susceptibility to Reward Hacking and Annotation Noise in Video Generation Reward Models
di: Lian, Jiesong, et al.
Pubblicazione: (2025)

MOSSBench: Is Your Multimodal Language Model Oversensitive to Safe Queries?
di: Li, Xirui, et al.
Pubblicazione: (2024)

Reward Shaping to Mitigate Reward Hacking in RLHF
di: Fu, Jiayi, et al.
Pubblicazione: (2025)

Information-Theoretic Reward Modeling for Stable RLHF: Detecting and Mitigating Reward Hacking
di: Miao, Yuchun, et al.
Pubblicazione: (2025)

Reward Hacking Mitigation using Verifiable Composite Rewards
di: Tarek, Mirza Farhan Bin, et al.
Pubblicazione: (2025)

Repairing Reward Functions with Feedback to Mitigate Reward Hacking
di: Hatgis-Kessell, Stephane, et al.
Pubblicazione: (2025)

Detecting and Suppressing Reward Hacking with Gradient Fingerprints
di: Wang, Songtao, et al.
Pubblicazione: (2026)

Robust Optimization for Mitigating Reward Hacking with Correlated Proxies
di: Liu, Zixuan, et al.
Pubblicazione: (2026)

Inference-Time Reward Hacking in Large Language Models
di: Khalaf, Hadi, et al.
Pubblicazione: (2025)

Benchmarking Reward Hack Detection in Code Environments via Contrastive Analysis
di: Deshpande, Darshan, et al.
Pubblicazione: (2026)

Adversarial Reward Auditing for Active Detection and Mitigation of Reward Hacking
di: Beigi, Mohammad, et al.
Pubblicazione: (2026)

EvilGenie: A Reward Hacking Benchmark
di: Gabor, Jonathan, et al.
Pubblicazione: (2025)

Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges
di: Wang, Xiaohua, et al.
Pubblicazione: (2026)

The Energy Loss Phenomenon in RLHF: A New Perspective on Mitigating Reward Hacking
di: Miao, Yuchun, et al.
Pubblicazione: (2025)

InfoRM: Mitigating Reward Hacking in RLHF via Information-Theoretic Reward Modeling
di: Miao, Yuchun, et al.
Pubblicazione: (2024)

Helping or Herding? Reward Model Ensembles Mitigate but do not Eliminate Reward Hacking
di: Eisenstein, Jacob, et al.
Pubblicazione: (2023)

Beyond Reward Hacking: Causal Rewards for Large Language Model Alignment
di: Wang, Chaoqi, et al.
Pubblicazione: (2025)

GARDO: Reinforcing Diffusion Models without Reward Hacking
di: He, Haoran, et al.
Pubblicazione: (2025)

Rethinking RL Evaluation: Can Benchmarks Truly Reveal Failures of RL Methods?
di: Chen, Zihan, et al.
Pubblicazione: (2025)

Uncertainty-Aware Reward Discounting for Mitigating Reward Hacking
di: Singha, Disha
Pubblicazione: (2026)

Directional Alignment Mitigates Reward Hacking in Reinforcement Learning for Language Models
di: Deng, Wenlong, et al.
Pubblicazione: (2026)

Causal Reward Adjustment: Mitigating Reward Hacking in External Reasoning via Backdoor Correction
di: Song, Ruike, et al.
Pubblicazione: (2025)

From Curiosity to Caution: Mitigating Reward Hacking for Best-of-N with Pessimism
di: Yu, Zhuohao, et al.
Pubblicazione: (2026)

LLMs Gaming Verifiers: RLVR can Lead to Reward Hacking
di: Helff, Lukas, et al.
Pubblicazione: (2026)

Hacking Predictors Means Hacking Cars: Using Sensitivity Analysis to Identify Trajectory Prediction Vulnerabilities for Autonomous Driving Security
di: Gibson, Marsalis, et al.
Pubblicazione: (2024)

UMM-RM: An Upcycle-and-Merge MoE Reward Model for Mitigating Reward Hacking
di: Fu, Lingling, et al.
Pubblicazione: (2025)

Mitigating Reward Hacking in RLHF via Advantage Sign Robustness
di: Ono, Shinnosuke, et al.
Pubblicazione: (2026)

Feedback Loops With Language Models Drive In-Context Reward Hacking
di: Pan, Alexander, et al.
Pubblicazione: (2024)

Reward Hacking Benchmark: Measuring Exploits in LLM Agents with Tool Use
di: Thaman, Kunvar
Pubblicazione: (2026)

When Reward Hacking Rebounds: Understanding and Mitigating It with Representation-Level Signals
di: Wu, Rui, et al.
Pubblicazione: (2026)

Correlated Proxies: A New Definition and Improved Mitigation for Reward Hacking
di: Laidlaw, Cassidy, et al.
Pubblicazione: (2024)

Generative Adversarial Post-Training Mitigates Reward Hacking in Live Human-AI Music Interaction
di: Wu, Yusong, et al.
Pubblicazione: (2025)

Hacking Task Confounder in Meta-Learning
di: Wang, Jingyao, et al.
Pubblicazione: (2023)

MIRA: Towards Mitigating Reward Hacking in Inference-Time Alignment of T2I Diffusion Models
di: Zhai, Kevin, et al.
Pubblicazione: (2025)