AI Paper Atlas

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

Rethinking Policy Diversity in Ensemble Policy Gradient in Large-Scale Reinforcement Learning

Naoki Shitanda, Motoki Omura, Tatsuya Harada, Takayuki Osa

ICLR 2026 · Reinforcement learning › Policy gradient and actor-critic · approach: Trajectory reuse in policy gradients · 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 KL constraint and adversarial reward to SAPG (CPO)SAPG · ShadowHandfinal return: 12882 -> 13762Improved
add KL constraint and adversarial reward to SAPG (CPO)SAPG · AllegroHandfinal return: 11989 -> 14421Improved
add KL constraint and adversarial reward to SAPG (CPO)SAPG · Reorientationfinal return: 38.79 -> 43.75Improved
add KL constraint and adversarial reward to SAPG (CPO)SAPG · Two-Arms Reorientationfinal return: 5.11 -> 35.30Improved
add KL constraint and adversarial reward to SAPG (CPO)SAPG · Regraspingfinal return: 37.20 -> 37.44No clear effect
add KL constraint and adversarial reward to SAPG (CPO)SAPG · Throwfinal return: 22.51 -> 21.69No clear effect
strengthen KL constraint lambda_f from 0.5 to 0.05 (ablation)CPO · ShadowHandmean IS ratio deviation: 0.403 -> 0.187Improved
strengthen KL constraint lambda_f from 0.5 to 0.05 (ablation)CPO · ShadowHandESS rate: 0.763 -> 0.941Improved
add KL constraint (CPO 0.5) to SAPG (ablation)SAPG · ShadowHandESS rate: 0.0223 -> 0.763Improved
add KL constraint (CPO 0.5) to SAPG (ablation)SAPG · AllegroKuka Reorientationmean IS ratio deviation: 0.608 -> 0.420Improved
vary KL coefficient lambda_f across a wide range (ablation)CPO · ShadowHand and AllegroKuka Reorientationtraining curvesNo clear effect
use CPO in locomotion tasks instead of SAPGSAPG · Locomotion tasksconvergence speedNo clear effect

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