Research map / Reinforcement learning / Visual and representation RL
Diversity-Driven Extensible 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 |
|---|---|---|---|
| Use DEHRL's predictor-based intrinsic reward that favors states most different from alternative predicted outcomes | OverCooked subpolicy discovery · OverCooked, no extrinsic reward, 10 million steps, 5 subpolicies | number of useful subpolicies discovered: 1 vs. 5 | Helped |
| Increase the number of discovered subpolicies from 5 to 20 (ablation) | DEHRL · OverCooked, no extrinsic reward, 20 million steps | inclusion of five useful subpolicies: 20 | No clear effect |
| Increase the number of discovered subpolicies from 5 to 20 (ablation) | SNN · OverCooked, no extrinsic reward, 20 million steps | number of useful subpolicies discovered: 1 | No clear effect |
| Replace the sum of distances to alternative predicted states with the minimum distance (ablation) | DEHRL intrinsic reward estimator · DEHRL | intrinsic reward estimation quality | Helped |
| Add DEHRL's multi-level hierarchy and diversity-driven intrinsic rewards | OverCooked sparse-reward task · OverCooked, reward-level 2, all goal types | final performance score: 0.95 vs. 0.00; 0.93 vs. 0.00; 0.81 vs. 0.00 | Helped |
| Use DEHRL rather than MLSH for higher-level meta-subpolicies | Meta HRL · OverCooked, reward-level 2 | meta-learning ability | Helped |
| Use DEHRL rather than MLSH | Meta HRL · OverCooked, reward-level 1 | meta-learning performance | No clear effect |
| Increase the number of hierarchy levels (ablation) | DEHRL · MineCraft, intrinsic-reward-only play | world complexity (Valid Operations) | Helped |
| Use DEHRL rather than a random policy | MineCraft world building · MineCraft, intrinsic-reward-only play | world complexity (Valid Operations) | Helped |
Closest papers by meaning
- Data-Efficient Hierarchical Reinforcement Learning (NeurIPS 2018 · similarity 0.90)
- Hierarchical Reinforcement Learning with Targeted Causal Interventions (ICML 2025 · similarity 0.88)
- Probabilistic Subgoal Representations for Hierarchical Reinforcement Learning (ICML 2024 · similarity 0.87)
- Hierarchical Reinforcement Learning with Uncertainty-Guided Diffusional Subgoals (ICML 2025 · similarity 0.87)
- DUO: Diverse, Uncertain, On-Policy Query Generation and Selection for Reinforcement Learning from Human Feedback (AAAI 2025 · similarity 0.87)
- Controlling Behavioral Diversity in Multi-Agent Reinforcement Learning (ICML 2024 · similarity 0.87)
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.