Research map / Reinforcement learning / MDPs and planning
Iterative Bounding MDPs: Learning Interpretable Policies via Non-Interpretable Methods
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 |
|---|---|---|---|
| Learn the tree policy directly with CUSTARD (MFEC) rather than distilling the expert with VIPER (BI) | PrereqWorld (m=7) · PrereqWorld (m=7) | depth: 6.00 -> 3.92 | Helped |
| Learn the tree policy directly with CUSTARD (DQN) rather than distilling the expert with VIPER (DQN) | PotholeWorld · PotholeWorld | reward: 46.30 -> 46.92 | Helped |
| Learn the tree policy directly with CUSTARD (PPO) rather than distilling the expert with VIPER (BI) | PotholeWorld · PotholeWorld | depth: 9.18 -> 1.04 | Helped |
| Learn the tree policy directly with CUSTARD (MFEC) rather than distilling the expert with VIPER (BI) | PotholeWorld · PotholeWorld | reward: 46.31 -> 49.18 | Helped |
| Learn a depth-limited tree directly instead of fitting a tree to an unconstrained expert | PotholeWorld · PotholeWorld, fixed tree depth | reward | Helped |
| Use CUSTARD rather than VIPER as PrereqWorld environment size increases | PrereqWorld · PrereqWorld, varying m | tree size | Helped |
| Use looser feature bounds for CartPole (ablation) | CUSTARD (PPO) · CartPole | tree depth | Hurt |
| Set PotholeWorld's discrete split-value count p to 10 instead of 3 (ablation) | CUSTARD (PPO) · PotholeWorld | policy learning | Hurt |
| Use a lower information-gathering penalty ζ=-0.01 instead of ζ=0 (ablation) | CUSTARD (PPO) · PotholeWorld | frequency of finding the optimal policy | Hurt |
| Set γw and γb to 0.99 instead of 1 (ablation) | CUSTARD (PPO) · PrereqWorld | training speed | Hurt |
Closest papers by meaning
- SPOT: Scalable Policy Optimization with Trees for Markov Decision Processes (NeurIPS 2025 · similarity 0.90)
- Mitigating Information Loss in Tree-Based Reinforcement Learning via Direct Optimization (ICLR 2025 · similarity 0.87)
- Scalable Policy-Based RL Algorithms for POMDPs (NeurIPS 2025 · similarity 0.87)
- RL in Latent MDPs is Tractable: Online Guarantees via Off-Policy Evaluation (NeurIPS 2024 · similarity 0.87)
- In Search of Trees: Decision-Tree Policy Synthesis for Black-Box Systems via Search (AAAI 2025 · similarity 0.87)
- RGMDT: Return-Gap-Minimizing Decision Tree Extraction in Non-Euclidean Metric Space (NeurIPS 2024 · similarity 0.86)
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.