Research map / Learning theory and algorithms / Combinatorial and search optimization
Neural-Guided Deductive Search for Real-Time Program Synthesis from Examples
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 branch-score predictions to guide deductive search with branch-and-bound | PROSE on real-world string manipulation tasks · Real-world string tasks test set | accuracy: 67.12 -> 68.49 | Helped |
| Use neural branch-score predictions to guide deductive search with branch-and-bound | PROSE on real-world string manipulation tasks · Real-world string tasks test set | speed-up vs. PROSE: 1.00 -> 1.67 | Helped |
| Replace branch-and-bound with a threshold controller (ablation) | NGDS with T₁ score model · String manipulation test set | accuracy: 68.49 -> 67.12 | Hurt |
| Replace branch-and-bound with a threshold controller (ablation) | NGDS with T₁ score model · String manipulation test set | speed-up vs. PROSE: 1.22 -> 1.27 | Helped |
| Add a 0.2 threshold predecessor to branch-and-bound (ablation) | NGDS with T₁ score model · String manipulation test set | accuracy: 68.49 -> 67.12 | Hurt |
| Add a PP score model (ablation) | NGDS(T₁, BB) · String manipulation test set | accuracy: 68.49 -> 72.60 | Helped |
| Add a PP score model (ablation) | NGDS(T₁, BB) · String manipulation test set | speed-up vs. PROSE: 1.22 -> 0.89 | Hurt |
| Add a POS score model (ablation) | NGDS(T₁, BB) · String manipulation test set | accuracy: 68.49 -> 68.49 | No clear effect |
| Add a POS score model (ablation) | NGDS(T₁, BB) · String manipulation test set | speed-up vs. PROSE: 1.22 -> 1.67 | Helped |
| Use attention in the score prediction model (ablation) | NGDS branch-score predictor · Train/validation/test score-flip evaluation | score flip accuracy: 96.09/91.24/91.12% -> 99.57/90.4/96.4% | Mixed |
| Use attention in the score prediction model (ablation) | NGDS branch-score predictor · Synthesis-time evaluation | speed-up | Hurt |
| Use the T₁ branch-and-bound model with the ML-based ranker | PROSE · String manipulation test set with ML ranker | speed-up vs. PROSE: 1.00 -> 2.15 | Helped |
Closest papers by meaning
- Program Synthesis via Test-Time Transduction (NeurIPS 2025 · similarity 0.89)
- Execution-Guided Neural Program Synthesis (ICLR 2019 · similarity 0.88)
- Latent Execution for Neural Program Synthesis Beyond Domain-Specific Languages (NeurIPS 2021 · similarity 0.88)
- Neuro-Symbolic Program Synthesis (ICLR 2017 · similarity 0.88)
- Learning to Infer Program Sketches (ICML 2019 · similarity 0.88)
- HYSYNTH: Context-Free LLM Approximation for Guiding Program Synthesis (NeurIPS 2024 · similarity 0.87)
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.