AI Paper Atlas

Research map / Reinforcement learning / Offline RL

Critic Regularized Regression

Ziyu Wang, Alexander Novikov, Konrad Zolna, Josh S Merel, Jost Tobias Springenberg, Scott E Reed, Bobak Shahriari, Noah Siegel, Caglar Gulcehre, Nicolas Heess, Nando de Freitas

NeurIPS 2020 · Reinforcement learning › Offline RL · approach: Simple practical offline RL · 93 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
CRR exp vs ABMoffline RL · Humanoid Runepisode return: 302 -> 586Improved
CRR exp vs ABMoffline RL · Humanoid Corridorepisode return: 64 -> 918Improved
CRR vs ABMoffline RL · Rodent Mazesepisode return: 478 -> 459Hurt
CRR binary vs D4PGoffline RL · Manipulator Insert Ballepisode return: 154 -> 654Improved
advantage-filtered regression (CRR exp) vs plain BC (ablation)same network · Humanoid Corridorepisode return: 220 -> 918Improved
binary filter instead of exp weighting (ablation)CRR · Cartpole Swingupepisode return: 664 -> 860Improved
binary filter instead of exp weighting (ablation)CRR · Humanoid Runepisode return: 586 -> 412Hurt
binary max (A_max) instead of binary (A_mean) (ablation)CRR binary · Humanoid Runepisode return: 412 -> 226Hurt
binary max (A_max) instead of binary (A_mean) (ablation)CRR binary · Finger Turn Hardepisode return: 755 -> 833Improved
critic-weighted policy (CWP) action selection (ablation)CRR · DCS/locomotion selected envsepisode returnImproved
turn off Gaussian noise at evaluation (ablation)CRR · low/moderate action-dim envsepisode returnMixed

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