Research map / Learning theory and algorithms / Combinatorial and search optimization
DeepCoder: Learning to Write Programs
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 |
|---|---|---|---|
| Use neural network function predictions to order DFS choices | DFS · Generated programs, T=3; solve 20% | timeout: 41ms -> 2.7ms | Helped |
| Use neural network function predictions to order DFS choices | DFS · Generated programs, T=3; solve 40% | timeout: 126ms -> 33ms | Helped |
| Use neural network function predictions to order DFS choices | DFS · Generated programs, T=3; solve 60% | timeout: 314ms -> 110ms | Helped |
| Use neural network function predictions in sort-and-add enumeration | Enumeration · Generated programs, T=3; solve 20% | timeout: 80ms -> 1.3ms | Helped |
| Use neural network function predictions in sort-and-add enumeration | Enumeration · Generated programs, T=3; solve 40% | timeout: 335ms -> 6.1ms | Helped |
| Use neural network function predictions in sort-and-add enumeration | Enumeration · Generated programs, T=3; solve 60% | timeout: 861ms -> 27ms | Helped |
| Use neural network function predictions in sort-and-add search | λ² · Generated programs, T=3; solve 20% | timeout: 18.9s -> 0.23s | Helped |
| Use neural network function predictions in sort-and-add search | λ² · Generated programs, T=3; solve 40% | timeout: 49.6s -> 0.52s | Helped |
| Use neural network function predictions in sort-and-add search | λ² · Generated programs, T=3; solve 60% | timeout: 84.2s -> 13.5s | Helped |
| Use neural network function predictions in sort-and-add search | Sketch · Generated programs, T=3; solve 20% | timeout: >10^3s -> 2.13s | Helped |
| Use neural network function predictions in sort-and-add search | Sketch · Generated programs, T=3; solve 40% | timeout: >10^3s -> 455s | Helped |
| Replace the feed-forward input-output encoder with a GRU encoder (ablation) | DeepCoder · Function prediction from input-output examples | performance | No clear effect |
Closest papers by meaning
- Learning to Infer Program Sketches (ICML 2019 · similarity 0.87)
- Towards Synthesizing Complex Programs From Input-Output Examples (ICLR 2018 · similarity 0.87)
- Leveraging Language to Learn Program Abstractions and Search Heuristics (ICML 2021 · similarity 0.86)
- Gradient-Based Program Synthesis with Neurally Interpreted Languages (ICLR 2026 · similarity 0.86)
- Searching Latent Program Spaces (NeurIPS 2025 · similarity 0.85)
- CPRet: A Dataset, Benchmark, and Model for Retrieval in Competitive Programming (NeurIPS 2025 · similarity 0.85)
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.