G-TransEval

G-TransEval数据集由上海交通大学软件学院创建，旨在通过细粒度的评估来提高代码翻译模型的性能。该数据集包含400个样本，涵盖了四种不同复杂度和知识依赖的翻译类型：令牌级、语法级、库级和算法级。数据集通过多样化的源代码收集和语言扩展，确保了评估的全面性和准确性。此外，每个样本都配备了单元测试，以验证翻译的正确性。G-TransEval数据集的应用领域主要集中在提高代码翻译系统的准确性和可靠性，特别是在处理复杂和知识依赖性强的翻译任务时。

The G-TransEval dataset was developed by the School of Software Engineering, Shanghai Jiao Tong University, aiming to improve the performance of code translation models through fine-grained evaluation. This dataset contains 400 samples covering four translation types with distinct complexity and knowledge dependency: token-level, syntax-level, library-level, and algorithm-level. The comprehensiveness and accuracy of the evaluation are ensured via diversified source code collection and language expansion. Additionally, each sample is equipped with unit tests to verify the correctness of the translated code. The primary application scenarios of the G-TransEval dataset focus on enhancing the accuracy and reliability of code translation systems, especially when handling complex and knowledge-intensive translation tasks.