Research map / Reinforcement learning / MDPs and planning
ODE-based Recurrent Model-free Reinforcement Learning for POMDPs
Xuanle Zhao, Duzhen Zhang, Liyuan Han, Tielin Zhang, Bo XU
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 |
|---|---|---|---|
| separate context encoders for policy and value instead of shared (ablation) | PODMP · Walker-P | final return: 1336 -> 1487 | Improved |
| separate context encoders for policy and value instead of shared (ablation) | PODMP · Walker-V | final return: 584 -> 612 | Improved |
| context encoder input o,a,r instead of o only (ablation) | PODMP · Walker-P | final return: 1276 -> 1487 | Improved |
| add previous action to observation input (oa) vs o (ablation) | PODMP · Walker-P | final return: 1276 -> 1065 | Hurt |
| add reward to observation and action input (oar vs oa) (ablation) | PODMP · Walker-P | final return: 1065 -> 1487 | Improved |
| RK4 solver instead of Euler (ablation) | PODMP · Ant-P | final return: 1243 -> 1183 | No clear effect |
| RK4 solver instead of Euler (ablation) | PODMP · Walker-V | final return: 612 -> 683 | No clear effect |
| Heun solver instead of Euler (ablation) | PODMP · Ant-P | final return: 1243 -> 1123 | No clear effect |
| RK4 solver instead of Euler (ablation) | PODMP · AntBLT-P training time | hours: 12 hours -> 20 hours | Hurt |
| Heun solver instead of Euler (ablation) | PODMP · AntBLT-P training time | hours: 12 hours -> 17 hours | Hurt |
| GRU-ODE encoder (ours) vs RMF baseline | PODMP · Hopper-P | final return: 2133 -> 2455 | Improved |
| GRU-ODE encoder (ours) vs RMF baseline | PODMP · Ant-V | final return: 1021 -> 998 | Hurt |
Closest papers by meaning
- Rethinking Transformers in Solving POMDPs (ICML 2024 · similarity 0.90)
- Real-Time Recurrent Reinforcement Learning (AAAI 2025 · similarity 0.89)
- Efficient Recurrent Off-Policy RL Requires a Context-Encoder-Specific Learning Rate (NeurIPS 2024 · similarity 0.88)
- Model-based Reinforcement Learning for Confounded POMDPs (ICML 2024 · similarity 0.88)
- Periodic agent-state based Q-learning for POMDPs (NeurIPS 2024 · similarity 0.88)
- Scalable Policy-Based RL Algorithms for POMDPs (NeurIPS 2025 · 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.
