Research map / Large language models / Agents, code and math
Neural program synthesis and transpilation
Guide program synthesis, translation, and robot programs with learned models and probabilistic programs.
31 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Agents, code and math, keyword cluster “search · optimization · decision”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
12 papers in 2023–24 and 8 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 3.3% to 0.6%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2017 · DeepFix: Fixing Common C Language Errors by Deep Learning (AAAI 2017 · 423 citations)
- 2018 · Disjunctive Program Synthesis: A Robust Approach to Programming by Example (AAAI 2018 · 16 citations)
- 2019 · Automatic Code Review by Learning the Revision of Source Code (AAAI 2019 · 47 citations)
- 2020 · Unsupervised Translation of Programming Languages (NeurIPS 2020 · 62 citations)
- 2021 · Computer-Aided Design as Language (NeurIPS 2021 · 40 citations)
- 2022 · Multilingual Code Snippets Training for Program Translation (AAAI 2022 · 38 citations)
- 2023 · DocPrompting: Generating Code by Retrieving the Docs (ICLR 2023 · 19 citations)
- 2024 · IaC-Eval: A Code Generation Benchmark for Cloud Infrastructure-as-Code Programs (NeurIPS 2024 · 14 citations)
- 2025 · DiSciPLE: Learning Interpretable Programs for Scientific Visual Discovery (CVPR 2025 · 0 citations)
- 2026 · GIFT: Bootstrapping Image-to-CAD Program Synthesis via Geometric Feedback (ICML 2026 · 0 citations)
Earliest papers here
- DeepFix: Fixing Common C Language Errors by Deep Learning (AAAI 2017 · 423 citations)
- Automated Data Extraction Using Predictive Program Synthesis (AAAI 2017 · 34 citations)
- Learning to superoptimize programs (ICLR 2017 · 17 citations)
What has been tried
23 papers here report 196 measured changes (a component added, removed or swapped, compared with a baseline): helped 69%, no clear effect 9%, hurt 18%, mixed 4%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add few-shot prompting | IaC-Eval · pass@1 accuracy | 19.36% -> 10.64% | Hurt | IaC-Eval: A Code Generation Benchmark for Cloud Infrastructure-as-Code Programs (NeurIPS 2024 · 14 citations) |
| Add synthetic-sketch training | Reference figures · average performance across reported metrics | MTE -1.957; CBLEU +0.327; TED -0.822; DSIM +2.433; SSIM +1.318; KID +3.296 | Mixed | DeTikZify: Synthesizing Graphics Programs for Scientific Figures and Sketches with TikZ (NeurIPS 2024 · 3 citations) |
| Add causal-discovery feedback across rounds instead of using DATA's single round | AppleGastronome · F1 | 0.72±0.04 -> 0.85±0.11 | Helped | Discovery of the Hidden World with Large Language Models (NeurIPS 2024 · 3 citations) |
| Use feature-set prediction instead of directly prompting for a predictor | Population Density · L2 log | 0.3159 -> 0.2906 | Helped | DiSciPLE: Learning Interpretable Programs for Scientific Visual Discovery (CVPR 2025 · 0 citations) |
| Add functional specification learning only | Performance optimization (HQ test) · OPT@8 | 47.5 -> 49.2 | Helped | EditLord: Learning Code Transformation Rules for Code Editing (ICML 2025 · 0 citations) |
| Use AI Programmer data rather than the other individual data sources | Assist Programming Eval · Pass@1 | Helped | CursorCore: Assist Programming through Aligning Anything (ICML 2025 · 0 citations) | |
| vary confidence threshold gamma from 0.8 to 0.95 | Project Euler · number of corrected programs | No clear effect | Guess & Sketch: Language Model Guided Transpilation (ICLR 2024 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 9 papers of this approach from 2024–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- The Stack · 2 of 9 papers
Baselines
- GPT-4o · 2 of 9 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 20 papers; from OpenAlex and the papers' affiliations).
- Carnegie Mellon University · 3 papers
- Columbia University · 2 papers
- Cornell University · 2 papers
Use it
Download BibTeX of 29 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- LLM-guided program synthesis (93 papers)
- Execution-based code verification (90 papers)
- LLM planning with symbolic tools (44 papers)
Elsewhere in the topic
- Mathematical and structural reasoning benchmarks (304 papers)
- Agent instruction-following benchmarks (176 papers)
- Inductive and abstract reasoning evaluation (158 papers)
- Long-horizon LLM agents (155 papers)
- LLMs for optimization and heuristics (147 papers)
- Software engineering agents (140 papers)
- LLMs for scientific discovery (100 papers)
- Code generation benchmarks (96 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).