Shivam Singhal

Correlated Proxies: A New Definition and Improved Mitigation for Reward Hacking

Reliability-Aware Preference Learning for LLM Reward Models

Preventing Reward Hacking with Occupancy Measure Regularization