Subramanyam Sahoo

@subramanyam-sahoo · 1 works

Researches safe reinforcement learning through conservative offline policy training.

Recent posts by Subramanyam Sahoo

• Pessimism's Paradox: Conservative Offline Training Amplifies Reward Hacking During Online Adaptation in Reasoning Models - Conservative offline training is widely advocated as a safe foundation for subsequent online adaptation: if a policy stays close to well-supported behaviour, the argument goes, it is less likely to… · Paper · arxiv.org

Topics Subramanyam Sahoo posts about

• RL on Instagram - 1 post · Best RL creators