AI Paper Atlas

Research map / Reinforcement learning / Policy gradient and actor-critic

Reward Scale Robustness for Proximal Policy Optimization via DreamerV3 Tricks

Ryan Sullivan, Akarsh Kumar, Shengyi Huang, John P Dickerson, Joseph Suarez

NeurIPS 2023 · Reinforcement learning › Policy gradient and actor-critic · approach: Tree search and policy optimization · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add all evaluated DreamerV3 tricksPPO · Arcade Learning Environment, reward clippingmedian human-normalized scoreNo clear effect
Add all evaluated DreamerV3 tricksPPO · Arcade Learning Environment, no reward clippingmedian human-normalized scoreImproved
Add all evaluated DreamerV3 tricksPPO · DeepMind Control Suitemedian normalized returnHurt
Add symlog predictions (ablation)PPO · Arcade Learning Environment, no reward clippingmedian human-normalized scoreImproved
Add symlog predictions (ablation)PPO · Arcade Learning Environment, reward clippingmedian human-normalized scoreHurt
Add twohot encoding (ablation)PPO · Arcade Learning Environmentmedian human-normalized scoreHurt
Add twohot encoding (ablation)PPO · DeepMind Control Suitemedian normalized returnHurt
Add critic EMA regularizer (ablation)PPO · DeepMind Control Suitemedian normalized returnHurt
Add percentile scaling (ablation)PPO · DeepMind Control Suitemedian normalized returnImproved
Add symlog predictions (ablation)PPO with all tricks · Arcade Learning Environmentmedian human-normalized scoreImproved
Add percentile scaling (ablation)PPO · Atari with reward clipping versus no reward clippingmedian human-normalized scoreMixed
Add critic EMA regularizer (ablation)PPO · Atari with reward clipping versus no reward clippingmedian human-normalized scoreMixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy