AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Verifiable Reinforcement Learning via Policy Extraction

Osbert Bastani, Yewen Pu, Armando Solar-Lezama

NeurIPS 2018 · Reinforcement learning › Visual and representation RL · approach: Scalable RL systems and environments · 159 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use Q-weighted policy extraction instead of DAgger with 0-1 lossDecision-tree policy extraction on Atari Pong · Atari Pong, reward 21nodes: 7967 -> 769Improved
Distill the DNN oracle into a decision treeAtari Pong · Atari Pongreward: 21.0 -> 21.0No clear effect
Distill the DNN oracle into a decision treeToy Pong · Toy Pongreward: 250 -> 250No clear effect
Distill the oracle into a 3-node decision treeCart-pole controller · Cart-polereward: 200.0 -> 200.0No clear effect
Use fitted Q iteration instead of VIPERCart-pole · Cart-polereward: 104.3Hurt
Use a PPO-oracle regression tree instead of the neural-network oracleHalf-cheetah controller · Half-cheetahcumulative reward: 4189 -> 4014Hurt
Use an iLQR oracle for extraction instead of the reward-trained oracle (ablation)Cart-pole stability controller · Cart-polestabilityImproved
Add a top-level node to choose a safer action in the identified region (ablation)Toy Pong decision tree · Toy Pongverified correctnessImproved
Extend the paddle length to L=9/2 (ablation)Toy Pong controller · Toy Pongverified correctnessImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy