AI Paper Atlas

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

Harmonized Dual Policy Improvement for Modelic Reinforcement Learning

Guojian Zhan, Likun Wang, Feihong Zhang, Yang Guan, Shengbo Eben Li

ICML 2026 · Reinforcement learning › Policy gradient and actor-critic · approach: Model-based RL with guarantees

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 harmonized dual policy gradient (HDPI) to BOOMBOOM · DMC Suite (7 tasks)mean TAR: 803.1 -> 852.6Improved
add harmonized dual policy gradient (HDPI) to BOOMBOOM · Humanoid-Bench (7 tasks)mean TAR: 639.0 -> 746.1Improved
add harmonized dual policy gradient (HDPI) to BOOMBOOM · H1-RunTAR: 330.3 -> 752.1Improved
add harmonized dual policy gradient (HDPI) to BOOMBOOM · H1-WalkTAR: 902.1 -> 929.4No clear effect
add harmonization step to policy updateBOOM · wall-clock for 10k iterationsseconds: 312 -> 339Hurt
vary harmonization coefficient rho by 0.1x, 1x, 10x (ablation)BOOM-H · Dog-runreturn and policy lossNo clear effect
vary alignment coefficient lambda by 0.1x, 1x, 10x (ablation)BOOM-H · Dog-runasymptotic returnImproved
replace naive weighted combination with harmonized update (ablation)BOOM · Dog-run and H1hand-runpolicy lossImproved

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