Research map / Reinforcement learning / Policy gradient and actor-critic
Imitation-Projected Programmatic Reinforcement Learning
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 |
|---|---|---|---|
| PROPEL (iterative lift-and-project) instead of one-shot distillation NDPS (high-level language) | programmatic policy · TORCS G-Track | median lap time (s): 108.25 -> 93.67 | Helped |
| PROPEL-Tree vs VIPER | tree policy · TORCS E-Road | median lap time (s): 87.53 -> 79.39 | Helped |
| PROPEL-Prog vs NDPS on hard track | programmatic policy · TORCS Ruudskogen | median lap time (s): CR -> 124.58 | Helped |
| PROPEL-Prog vs DDPG | policy · TORCS G-Track | crash ratio: 0.24 -> 0.04 | Helped |
| PROPEL-Prog vs DDPG on lap time | policy · TORCS G-Track | median lap time (s): 78.82 -> 93.67 | Hurt |
| PROPEL-Prog vs DDPG | policy · TORCS Alpine-2 | median lap time (s): CR -> 256.59 | Helped |
| PROPEL-Prog vs prior PID controller | policy · TORCS G-Track | median lap time (s): 312.92 -> 93.67 | Helped |
| train on one track and test on others, PROPEL-Prog vs DDPG | policy · TORCS generalization | median lap time (s): CR -> 124 (G-Track to E-Road) | Helped |
| initialize PROPEL with DDPG neural policy instead of simple programmatic prior (ablation) | PROPEL · TORCS | completion | Hurt |
Closest papers by meaning
- Programmatic Reinforcement Learning without Oracles (ICLR 2022 · similarity 0.90)
- Synthesizing Programmatic Reinforcement Learning Policies with Large Language Model Guided Search (ICLR 2025 · similarity 0.88)
- Reclaiming the Source of Programmatic Policies: Programmatic versus Latent Spaces (ICLR 2024 · similarity 0.86)
- Foundation Policies with Hilbert Representations (ICML 2024 · similarity 0.86)
- $\mathcal{B}$-Coder: Value-Based Deep Reinforcement Learning for Program Synthesis (ICLR 2024 · similarity 0.86)
- Simplifying Constraint Inference with Inverse Reinforcement Learning (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.