Research map / Reinforcement learning / Policy gradient and actor-critic
Action Manifold Smoothing: A Lipschitz Pathway Perspective on High-Dimensional Reinforcement Learning
Zhihao Lin
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 Neighborhood TD and LayerNorm Q-network (AMS) | SAC · Dog-Run | episodic return: 45 ± 95 -> 427 ± 25 | Improved |
| add Neighborhood TD and LayerNorm Q-network (AMS) | TD3 · Dog-Run | episodic return: 92 ± 121 -> 387 ± 32 | Improved |
| add AMS | SAC · Quadruped-Run | episodic return: 301 ± 8 -> 781 ± 61 | Improved |
| add AMS | TD3 · Quadruped-Run | episodic return: 576 ± 212 -> 765 ± 32 | Improved |
| add Neighborhood TD to LayerNorm Q-network (ablation) | AMS-TD3 · Quadruped-Run | episodic return: 730 ± 119 -> 765 ± 32 | Improved |
| add LayerNorm in Q-network to Neighborhood TD (ablation) | AMS-TD3 · Quadruped-Run | episodic return: 732 ± 59 -> 765 ± 32 | Improved |
| add Neighborhood TD to LayerNorm Q-network (ablation) | AMS-SAC · Quadruped-Run | episodic return: 714 ± 67 -> 781 ± 61 | Improved |
| add LayerNorm in Q-network to Neighborhood TD (ablation) | AMS-SAC · Quadruped-Run | episodic return: 726 ± 41 -> 781 ± 61 | Improved |
| add AMS | TD3 · H1-Walk and H1-Run | episodic return: 787 ± 57 vs 775 ± 121 (H1-Walk); 658 ± 190 vs 614 ± 278 (H1-Run) | No clear effect |
| add AMS | TD3 · Cheetah-Run | episodic return: 753 ± 27 vs 712 ± 27 | No clear effect |
| raise neighborhood radius from 0.2 to 0.4 (ablation) | AMS-TD3 · Quadruped-Run | episodic return: 766 ± 26 -> 716 ± 64 | No clear effect |
| reduce neighbors from K=8 to K=6 (ablation) | AMS-TD3 · Quadruped-Run | episodic return: 766 ± 26 -> 737 ± 20 | No clear effect |
Closest papers by meaning
- Geometry of Neural Reinforcement Learning in Continuous State and Action Spaces (ICLR 2025 · similarity 0.86)
- LipsNet++: Unifying Filter and Controller into a Policy Network (ICML 2025 · similarity 0.86)
- Local Manifold Approximation and Projection for Manifold-Aware Diffusion Planning (ICML 2025 · similarity 0.85)
- The Laplacian Keyboard: Beyond the Linear Span (NeurIPS 2026 · similarity 0.84)
- Reasoning on the Manifold: Bidirectional Consistency for Self-Verification in Diffusion Language Models (ICML 2026 · similarity 0.84)
- Hitting Time Isomorphism for Multi-Stage Planning with Foundation Policies (NeurIPS 2026 · similarity 0.84)
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.
