AI Paper Atlas

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

Taming Aleatoric Impulse in Off-Policy Reinforcement Learning

Zhouyang Yu, Guojian Zhan, Yang Guan, Jingliang Duan, Letian Tao, Shengbo Eben Li

ICML 2026 · Reinforcement learning › Policy gradient and actor-critic · approach: Off-policy actor-critic with expressive policies

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 aleatoric pessimism (beta_init > 0) (ablation)DSAC-AID · Humanoid-v3TAR: removing it degrades performance by 15.2%Improved
add aleatoric pessimism (beta_init > 0) (ablation)DSAC-AID · DMC Humanoid-runTAR: removing it degrades performance by 9.7%Improved
add pessimism annealing (l_beta > 0) (ablation)DSAC-AID · Humanoid-v3TAR: drastic performance collapse without itImproved
add pessimism annealing (l_beta > 0) (ablation)DSAC-AID · DMC Humanoid-runTAR: removing it drops 10.8%Improved
add optimistic exploration (lambda_UB > 0) (ablation)DSAC-AID · Humanoid-v3TAR: removing it gives a deficit of 11.4%Improved
add optimistic exploration (lambda_UB > 0) (ablation)DSAC-AID · DMC Humanoid-runTAR: removing it gives a deficit of 14.8%Improved
vary beta_init and l_beta over several settings (ablation)DSAC-AID · DMC Humanoid-runTARNo clear effect
replace baseline with DSAC-AIDBRO (strongest baseline) · AVG. Gym-MuJoConormalized TAR: 0.793 -> 0.918Improved
replace baseline with DSAC-AIDDoubleGum · AVG. DMC Suitenormalized TAR: 0.858 -> 0.926Improved

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