Research map / Reinforcement learning / Visual and representation RL
Iteratively Learn Diverse Strategies with State Distance Information
Wei Fu, Weihua Du, Jingwei Li, Sunli Chen, Jingzhao Zhang, Yi Wu
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 |
|---|---|---|---|
| iterative learning (ITR) instead of population-based training (PBT), N_L=4 | policy diversity search · 2-D navigation | landmarks discovered: 2.0 -> 3.5 | Improved |
| iterative learning (ITR) instead of population-based training (PBT), N_L=5 | policy diversity search · 2-D navigation | landmarks discovered: 2.2 -> 4.5 | Improved |
| fix Lagrange multiplier instead of two-timescale GDA (ablation) | SIPO-WD · GRF 3v1 | distinct strategies: 3.0 -> 1.0 | Hurt |
| replace state-distance intrinsic reward with action cross-entropy (ablation) | SIPO-WD · GRF corner | distinct strategies: 3.0 -> 1.7 | Hurt |
| replace state-distance intrinsic reward with action cross-entropy (ablation) | SIPO-WD · GRF CA | distinct strategies: 3.0 -> 2.3 | Hurt |
| replace state-distance intrinsic reward with action cross-entropy (ablation) | SIPO-WD · GRF 3v1 | distinct strategies: 3.0 -> 2.7 | Hurt |
| use filtering-based optimization instead of Lagrangian constraint (ablation) | SIPO-WD · GRF 3v1 | distinct strategies: 3.0 -> 1.3 | Hurt |
| use filtering-based optimization instead of Lagrangian constraint (ablation) | SIPO-WD · GRF corner | distinct strategies: 3.0 -> 1.0 | Hurt |
| train M policies simultaneously (PBT) instead of iteratively (ablation) | SIPO-WD · GRF 3v1 | distinct strategies: 3.0 -> 2.7 | Hurt |
| SIPO-WD instead of RSPO | SIPO · Humanoid | diversity: 0.53 -> 0.71 | Improved |
| SIPO-RBF instead of RSPO | SIPO · Humanoid | diversity: 0.53 -> 0.53 | No clear effect |
| SIPO-RBF instead of RSPO | SIPO · SMAC 2m_vs_1z | kNN state entropy: 0.032 -> 0.038 | Improved |
Closest papers by meaning
- DGPO: Discovering Multiple Strategies with Diversity-Guided Policy Optimization (AAAI 2024 · similarity 0.91)
- Controlling Behavioral Diversity in Multi-Agent Reinforcement Learning (ICML 2024 · similarity 0.90)
- Proximal Policy Gradient Arborescence for Quality Diversity Reinforcement Learning (ICLR 2024 · similarity 0.89)
- Quality Diversity through Human Feedback: Towards Open-Ended Diversity-Driven Optimization (ICML 2024 · similarity 0.89)
- Enhancing Diversity In Parallel Agents: A Maximum State Entropy Exploration Story (ICML 2025 · similarity 0.88)
- DUO: Diverse, Uncertain, On-Policy Query Generation and Selection for Reinforcement Learning from Human Feedback (AAAI 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.
