Research map / Reinforcement learning / Visual and representation RL
Sub-policy Adaptation for Hierarchical Reinforcement Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| randomize the skill time-commitment instead of fixing it (ablation) | HiPPO · Snake Gather and Ant Gather, dynamics perturbations | drop in average return | Mixed |
| adapt pretrained sub-policies together with the manager instead of keeping them fixed (ablation) | HiPPO · Block task and Gather | final performance | Helped |
| use HiPPO instead of flat PPO | learning from scratch · Block and Gather environments | performance | Helped |
| use HiPPO instead of PPO with action repeat (ablation) | learning from scratch · Block and Gather environments | final performance | Helped |
| set manager period p=1 (ablation) | HiPPO · learning from scratch | performance | Hurt |
| use a skill-conditioned baseline (ablation) | HiPPO · learning from scratch | performance | Helped |
| remove PPO's clipped trust-region-like update and use HierVPG (ablation) | hierarchical policy optimization · learning from scratch | training stability | Hurt |
| randomize the skill time-commitment (ablation) | HiPPO · zero-shot adaptation to similar tasks | adaptation performance | Helped |
| fine-tune pretrained skills for the downstream task (ablation) | HiPPO · Gather with a cautious-behavior penalty | final performance | Helped |
Closest papers by meaning
- Hierarchical Reinforcement Learning with Advantage-Based Auxiliary Rewards (NeurIPS 2019 · similarity 0.91)
- Hierarchical Reinforcement Learning via Advantage-Weighted Information Maximization (ICLR 2019 · similarity 0.90)
- Latent Space Policies for Hierarchical Reinforcement Learning (ICML 2018 · similarity 0.90)
- Learning Multi-Level Hierarchies with Hindsight (ICLR 2019 · similarity 0.90)
- Deep Reinforcement Learning from Hierarchical Preference Design (ICML 2025 · similarity 0.88)
- Hierarchical Reinforcement Learning with Uncertainty-Guided Diffusional Subgoals (ICML 2025 · 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.