Model Releases
Bootstrapping Niche Multilingual Code Translation via Reinforcement Learning with Execution-Based Verifiable Supervision
arXiv:2608.13854v1 Announce Type: new Abstract: Code translation must preserve executable behavior across many programming languages, yet neural code translation has largely focused on a few popular l
arXiv:2608.13854v1 Announce Type: new Abstract: Code translation must preserve executable behavior across many programming languages, yet neural code translation has largely focused on a few popular languages such as C++, Java, and Python. This leaves a niche, many-to-many setting where parallel supervision is sparse, producing plausible but non-executable translations. We address this setting with preference-based reinforcement learning driven by execution-based supervision. Our pipeline firstly expands verifiable seed Python programs into a multilingual pool of execution-validated codes. Using the pool, a base LLM generates translation candidates across language pairs, which we label by their execution outcomes. The resulting preferences are used to train a reward model that scores cross-language translation quality. Finally, we optimize our base LLMs with GRPO over 600 directed language pairs (25 x 24) using the reward model as a signal. To evaluate the niche translation capability, we introduce HumanEval-X++, an execution-based benchmark that extends HumanEval-X to a broad many-to-many language space. We evaluate our approach using Qwen-3.5 4B and 9B models. On HumanEval-X++ and existing benchmarks, it yields consistent gains over the untrained baselines. In particular, the 4B model achieves an average improvement of 13% across all languages on HumanEval-X++, with a gain of 21% on mid-tier languages. Our study establishes a reliable approach of data generation, training, and benchmarking, paving the way toward further bootstrapping the quality of many-to-many translation for programming languages.
Related
- CodePivot: Bootstrapping Multilingual Transpilation in LLMs via Reinforcement Learning without Parallel Corpora
- StepCodeReasoner: Aligning Code Reasoning with Stepwise Execution Traces via Reinforcement Learning
- AutoPyVerifier: Learning Compact Executable Verifiers for Large Language Model Outputs
- Improving Small Language Models for Code Generation with Reinforcement Learning from Verification Feedback
Source: arXiv cs.CL | 2026-08-17