To ad- vance research on code translation and meet diverserequirementsofreal-worldapplications, we construct CodeTransOcean , a large-scale comprehensive benchmark that supports the largest …
To that end, we present a large-scale empirical study to investigate the ability of LLMs, including general LLMs and code LLMs, for code translation across pairs of different languages, including C, C++, Go, …
Reason 1: Testing cannot capture full equivalence. Even simple translation tasks, such as Figure 1 (from Cheung et al. (2013) on verified code translation of web applications) illus-trates the flaws in a testing …
Our work fits into this gap by evaluating the effect of AI support on such outcomes: the quality and completeness of a source code translation. We begin by summarizing recent developments in the …
We develop the first property-guided search procedure for code translation models to generate alternative trans-lations that potentially satisfy a given set of properties.
We describe the steps involved in glue code generation for the Java-to-Python code translation for an entire project with several interdependent Java classes. We identify and describe the inherent …
Our human-language code translator was inspired by a desire to make programming more accessible [6]. An accurate and useful translator would enable faster localization of instruction materials and it …