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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add KL constraint and adversarial reward to SAPG (CPO) | SAPG · ShadowHand | final return: 12882 -> 13762 | Improved |
| add KL constraint and adversarial reward to SAPG (CPO) | SAPG · AllegroHand | final return: 11989 -> 14421 | Improved |
| add KL constraint and adversarial reward to SAPG (CPO) | SAPG · Reorientation | final return: 38.79 -> 43.75 | Improved |
| add KL constraint and adversarial reward to SAPG (CPO) | SAPG · Two-Arms Reorientation | final return: 5.11 -> 35.30 | Improved |
| add KL constraint and adversarial reward to SAPG (CPO) | SAPG · Regrasping | final return: 37.20 -> 37.44 | No clear effect |
| add KL constraint and adversarial reward to SAPG (CPO) | SAPG · Throw | final return: 22.51 -> 21.69 | No clear effect |
| strengthen KL constraint lambda_f from 0.5 to 0.05 (ablation) | CPO · ShadowHand | mean IS ratio deviation: 0.403 -> 0.187 | Improved |
| strengthen KL constraint lambda_f from 0.5 to 0.05 (ablation) | CPO · ShadowHand | ESS rate: 0.763 -> 0.941 | Improved |
| add KL constraint (CPO 0.5) to SAPG (ablation) | SAPG · ShadowHand | ESS rate: 0.0223 -> 0.763 | Improved |
| add KL constraint (CPO 0.5) to SAPG (ablation) | SAPG · AllegroKuka Reorientation | mean IS ratio deviation: 0.608 -> 0.420 | Improved |
| vary KL coefficient lambda_f across a wide range (ablation) | CPO · ShadowHand and AllegroKuka Reorientation | training curves | No clear effect |
| use CPO in locomotion tasks instead of SAPG | SAPG · Locomotion tasks | convergence speed | No clear effect |
Closest papers by meaning
- Iteratively Learn Diverse Strategies with State Distance Information (NeurIPS 2023 · similarity 0.91)
- DGPO: Discovering Multiple Strategies with Diversity-Guided Policy Optimization (AAAI 2024 · similarity 0.90)
- The Curse of Diversity in Ensemble-Based Exploration (ICLR 2024 · similarity 0.89)
- Enhancing Diversity In Parallel Agents: A Maximum State Entropy Exploration Story (ICML 2025 · similarity 0.89)
- Controlling Behavioral Diversity in Multi-Agent Reinforcement Learning (ICML 2024 · similarity 0.88)
- Sample Efficient Myopic Exploration Through Multitask Reinforcement Learning with Diverse Tasks (ICLR 2024 · similarity 0.88)
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.
