Beyond Code Pairs: Dialogue-Based Data Generation for LLM Code Translation
Le Chen 1 Nuo Xu 2 Winson Chen 3 Bin Lei 2 Pei-Hung Lin 4 Dunzhi Zhou 2 Rajeev Thakur 1 Caiwen Ding 2 Ali Jannesari 3 Chunhua Liao 4 1 Argonne National Laboratory, Lemont, USA 2 University of Minnesota, Minneapolis, USA 3 Iowa State University, Ames, USA 4 Lawrence Livermore National Laboratory, Liv · 2025
Read Beyond Code Pairs: Dialogue-Based Data Generation for LLM Code Translation by Le Chen 1 Nuo Xu 2 Winson Chen 3 Bin Lei 2 Pei-Hung Lin 4 Dunzhi Zhou 2 Rajeev Thakur 1 Caiwen Ding 2 Ali Jannesari 3 Chunhua Liao 4 1 Argonne National Laboratory, Lemont, USA 2 University of Minnesota, Minneapolis, USA 3 Iowa State University, Ames, USA 4 Lawrence Livermore National Laboratory, Liv (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-03086.txt.
Description
Translating code between languages needs good training data. Le Chen and a large team address a shortage of such data in this 2025 work. Language models do well at code translation, but they falter in rare fields like Fortran and new tools like CUDA.
The team builds a pipeline that makes training data through dialogue. Two models work as questioner and solver, guided by compiler and runtime feedback. That loop helps keep the code correct. More good data could help firms update old code and use new hardware without hand work. Compilers in the loop help keep generated code pairs honest, rather than merely plausible.
Catalog record
| CREATOR | Le Chen 1 Nuo Xu 2 Winson Chen 3 Bin Lei 2 Pei-Hung Lin 4 Dunzhi Zhou 2 Rajeev Thakur 1 Caiwen Ding 2 Ali Jannesari 3 Chunhua Liao 4 1 Argonne National Laboratory, Lemont, USA 2 University of Minnesota, Minneapolis, USA 3 Iowa State University, Ames, USA 4 Lawrence Livermore National Laboratory, Liv |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 75 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-03086