Research map / Reinforcement learning / Visual and representation RL
Learning Compositional Neural Programs with Recursive Tree Search and Planning
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 |
|---|---|---|---|
| add MCTS planning at deployment (ablation) | AlphaNPI iterative BUBBLESORT · list length 100 | accuracy: 10% -> 40% | Helped |
| add MCTS planning at deployment (ablation) | AlphaNPI iterative BUBBLESORT · list length 60 | accuracy: 40% -> 95% | Helped |
| add MCTS planning at deployment (ablation) | AlphaNPI recursive BUBBLESORT · list length 100 | accuracy: 10% -> 100% | Helped |
| add program hierarchy to sorting (ablation) | AlphaNPI iterative sorting · iterative sorting, list length 3-6 | accuracy: no hierarchy: 94% -> 42% -> 10% -> 1% (length 3 to 6, with planning) | Helped |
| use recursive instead of iterative programs | AlphaNPI BUBBLESORT · list length 100 with planning | accuracy: 40% -> 100% | Helped |
| add MCTS planning at deployment (ablation) | AlphaNPI Tower of Hanoi · 2 to 12 disks | accuracy: 100% -> 100% | No clear effect |
| add MCTS planning at deployment for no-hierarchy sorting (ablation) | AlphaNPI atomic-only sorting · list length 3 | accuracy: 78% -> 94% | Helped |
Closest papers by meaning
- Improving the Universality and Learnability of Neural Programmer-Interpreters with Combinator Abstraction (ICLR 2018 · similarity 0.87)
- Gradient-Based Program Synthesis with Neurally Interpreted Languages (ICLR 2026 · similarity 0.86)
- Neural Program Lattices (ICLR 2017 · similarity 0.86)
- Making Neural Programming Architectures Generalize via Recursion (ICLR 2017 · similarity 0.86)
- Synthesizing Programmatic Reinforcement Learning Policies with Large Language Model Guided Search (ICLR 2025 · similarity 0.85)
- Efficient Reinforcement Learning by Discovering Neural Pathways (NeurIPS 2024 · similarity 0.84)
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.