Research map / Learning theory and algorithms / Combinatorial and search optimization
Glass-Box Program Synthesis: A Machine Learning Approach
Konstantina Christakopoulou, Adam Kalai
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use domain-specific GlassPS learning instead of brute-force search | GlassPS · Challenge 1: GCD | solution time: >30:00 -> 0:10 | Improved |
| Use all-domain GlassPS learning instead of brute-force search | GlassPS · Challenge 3: most frequent character | solution time: >2:15 -> 0:01 | Improved |
| Use domain-specific GlassPS learning instead of brute-force search | GlassPS · Challenge 4: alphabetically first character | solution time: 0:05 -> 0:01 | Improved |
| Use domain-specific GlassPS learning instead of brute-force search | GlassPS · Challenge 5: solve log(y) - (x * x) / 2 = 0 | solution time: 2:15 -> 0:01 | Improved |
| Use domain-specific GlassPS learning instead of brute-force search | GlassPS · Challenge 6: solve y + pow(x, 2 * 2) / 2 = 0 | solution time: >2:15 -> 0:01 | Improved |
| Use domain-specific GlassPS learning instead of brute-force search | GlassPS · Challenge 7: sum of squares | solution time: >2:15 -> 0:05 | Improved |
| Use all-domain GlassPS learning instead of brute-force search | GlassPS · Challenge 8: sum of cubes | solution time: >2:15 -> 0:01 | Improved |
| Use all-domain GlassPS learning instead of brute-force search | GlassPS · Challenge 10: sum of reciprocal product | solution time: >2:15 -> 0:01 | Improved |
| Use domain-specific rather than all-domain learning | GlassPS · Challenge 1: GCD | solution time: >2:15 -> 0:10 | Improved |
| Use domain-specific rather than all-domain learning | GlassPS · Challenge 6: solve y + pow(x, 2 * 2) / 2 = 0 | solution time: 0:03 -> 0:01 | Improved |
| Use domain-specific rather than all-domain learning | GlassPS · Challenge 7: sum of squares | solution time: >2:15 -> 0:05 | Improved |
| Use domain-specific rather than all-domain learning | GlassPS · Challenge 2: greatest non-trivial factor | solution time: >2:15 -> >2:15 | No clear effect |
Closest papers by meaning
- CrossBeam: Learning to Search in Bottom-Up Program Synthesis (ICLR 2022 · similarity 0.87)
- Differentiable Synthesis of Program Architectures (NeurIPS 2021 · similarity 0.87)
- Program Synthesis with Best-First Bottom-Up Search (Abstract Reprint) (AAAI 2024 · similarity 0.85)
- Eco Search: A No-delay Best-First Search Algorithm for Program Synthesis (AAAI 2025 · similarity 0.85)
- Bayesian Program Learning by Decompiling Amortized Knowledge (ICML 2024 · similarity 0.85)
- Learning to Reason via Program Generation, Emulation, and Search (NeurIPS 2024 · 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.
