Code translation is the process of converting source code from one programming language to another while keeping its behavior intact. The new arXiv paper, "Hallucinations of Large Language Models in Cross-Language Code Translation: Phenomena and Mitigation," addresses a persistent issue: even as LLMs have improved at this task, they still produce hallucinated outputs.
The abstract frames the problem but does not detail the specific phenomena or mitigation techniques in the available text. The paper's title suggests it covers both a characterization of hallucinations and methods to reduce them. Because this is a single source, there is no independent comparison to note.