Research map / Reinforcement learning / Visual and representation RL
Verifiable Reinforcement Learning via Policy Extraction
Osbert Bastani, Yewen Pu, Armando Solar-Lezama
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 |
|---|---|---|---|
| Use Q-weighted policy extraction instead of DAgger with 0-1 loss | Decision-tree policy extraction on Atari Pong · Atari Pong, reward 21 | nodes: 7967 -> 769 | Improved |
| Distill the DNN oracle into a decision tree | Atari Pong · Atari Pong | reward: 21.0 -> 21.0 | No clear effect |
| Distill the DNN oracle into a decision tree | Toy Pong · Toy Pong | reward: 250 -> 250 | No clear effect |
| Distill the oracle into a 3-node decision tree | Cart-pole controller · Cart-pole | reward: 200.0 -> 200.0 | No clear effect |
| Use fitted Q iteration instead of VIPER | Cart-pole · Cart-pole | reward: 104.3 | Hurt |
| Use a PPO-oracle regression tree instead of the neural-network oracle | Half-cheetah controller · Half-cheetah | cumulative reward: 4189 -> 4014 | Hurt |
| Use an iLQR oracle for extraction instead of the reward-trained oracle (ablation) | Cart-pole stability controller · Cart-pole | stability | Improved |
| Add a top-level node to choose a safer action in the identified region (ablation) | Toy Pong decision tree · Toy Pong | verified correctness | Improved |
| Extend the paddle length to L=9/2 (ablation) | Toy Pong controller · Toy Pong | verified correctness | Improved |
Closest papers by meaning
- SkillTree: Explainable Skill-Based Deep Reinforcement Learning for Long-Horizon Control Tasks (AAAI 2025 · similarity 0.87)
- Mitigating Information Loss in Tree-Based Reinforcement Learning via Direct Optimization (ICLR 2025 · similarity 0.87)
- In Search of Trees: Decision-Tree Policy Synthesis for Black-Box Systems via Search (AAAI 2025 · similarity 0.86)
- Blending Imitation and Reinforcement Learning for Robust Policy Improvement (ICLR 2024 · similarity 0.86)
- RGMDT: Return-Gap-Minimizing Decision Tree Extraction in Non-Euclidean Metric Space (NeurIPS 2024 · similarity 0.86)
- Confounding Robust Deep Reinforcement Learning: A Causal Approach (NeurIPS 2025 · 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.
