PLAISEJun 3

Beyond Code Pairs: Dialogue-Based Data Generation for LLM Code Translation

arXiv:2512.0308695.35 citationsh-index: 9
AI Analysis

For practitioners needing code translation in low-resource programming languages or frameworks, this work provides a method to generate high-quality training data, significantly boosting functional correctness.

The paper tackles the problem of LLM code translation in low-resource domains (Fortran, CUDA) by introducing a dialogue-based data generation pipeline. Fine-tuning on this data improved unit test success rates by over 56% on the C++-to-CUDA task, enabling a 7B model to outperform larger proprietary systems.

Large language models (LLMs) have shown remarkable capabilities in code translation, yet their performance deteriorates in low-resource programming domains such as Fortran and emerging frameworks like CUDA, where high-quality parallel data are scarce. We present an automated dataset generation pipeline featuring a dual-LLM Questioner-Solver design that incorporates external knowledge from compilers and runtime feedback. Beyond traditional source-target code pair datasets, our approach additionally generates (1) verified translations with unit tests for assessing functional consistency and (2) multi-turn dialogues that capture the reasoning process behind translation refinement. Applied to Fortran-to-C++ and C++-to-CUDA, the pipeline yields 3.64k and 3.93k dialogues, respectively. Fine-tuning on this data yields dramatic improvements in functional correctness, boosting unit test success rates by over 56% on the challenging C++-to-CUDA task. We show that the generated data enables a 7B open-weight model to significantly outperform larger proprietary systems on key metrics like compilation success.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes