Research map / Large language models / Agents, code and math
Unsupervised Translation of Programming Languages
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 |
|---|---|---|---|
| increase beam size from 1 to 25 | TransCoder · C++ → Java test set | computational accuracy: 60.91 -> 74.84 | Helped |
| increase beam size from 1 to 25 | TransCoder · C++ → Python test set | computational accuracy: 44.49 -> 67.17 | Helped |
| increase beam size from 1 to 25 | TransCoder · Java → C++ test set | computational accuracy: 80.9 -> 91.63 | Helped |
| increase beam size from 1 to 25 | TransCoder · Java → Python test set | computational accuracy: 34.99 -> 68.68 | Helped |
| increase beam size from 1 to 25 | TransCoder · Python → C++ test set | computational accuracy: 32.19 -> 57.29 | Helped |
| increase beam size from 1 to 25 | TransCoder · Python → Java test set | computational accuracy: 24.74 -> 56.13 | Helped |
| return only the highest-log-probability hypothesis instead of checking all beam hypotheses | TransCoder Beam 10 · all six language directions | computational accuracy: 73.39 -> 65.07; 61.98 -> 46.86; 89.27 -> 79.82; 64.36 -> 49.03; 49.57 -> 32.40; 51.14 -> 36.59 | Hurt |
| remove comments from training code (ablation) | TransCoder · C++ → Java test set, Beam 1 | computational accuracy: 62.16 -> 60.91 | No clear effect |
| remove comments from training code (ablation) | TransCoder · C++ → Python test set, Beam 1 | computational accuracy: 40.82 -> 44.49 | Hurt |
| remove comments from training code (ablation) | TransCoder · Java → C++ test set, Beam 1 | computational accuracy: 76.82 -> 80.9 | Hurt |
| remove comments from training code (ablation) | TransCoder · Java → Python test set, Beam 1 | computational accuracy: 46.436 -> 34.99 | Helped |
| remove comments from training code (ablation) | TransCoder · Python → C++ test set, Beam 1 | computational accuracy: 34.12 -> 32.19 | No clear effect |
Closest papers by meaning
- CodeRosetta: Pushing the Boundaries of Unsupervised Code Translation for Parallel Programming (NeurIPS 2024 · similarity 0.88)
- Automatically Testing Functional Properties of Code Translation Models (AAAI 2024 · similarity 0.87)
- Guess & Sketch: Language Model Guided Transpilation (ICLR 2024 · similarity 0.87)
- QiMeng-NeuComBack: Self-Evolving Translation from IR to Assembly Code (NeurIPS 2025 · similarity 0.85)
- QiMeng-MuPa: Mutual-Supervised Learning for Sequential-to-Parallel Code Translation (NeurIPS 2025 · similarity 0.85)
- Function-to-Style Guidance of LLMs for Code Translation (ICML 2025 · 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.