umaimakhan01/domain-code-bench

DomainCodeBench是一个针对代码生成模型的领域特定基准测试数据集，专注于评估模型在医疗保健系统、金融算法、分子模拟和法律文档处理四个专业领域的性能。该数据集通过功能正确性、合规性评分、领域API覆盖率、代码质量和参考解决方案相似性等多个维度，全面评估模型在特定领域的代码生成能力。数据集包含20个任务，每个任务都设计有特定的难度和关键挑战，旨在测试模型在复杂领域特定要求下的表现。

DomainCodeBench is a domain-specific code generation benchmark that evaluates models across four specialized domains: Healthcare Systems, Financial Algorithms, Molecular Simulation, and Legal Document Processing. Unlike general-purpose benchmarks, it assesses domain-specific quality through multiple dimensions including functional correctness, compliance score, domain API coverage, code quality, and reference similarity. The benchmark consists of 20 tasks, each designed with specific difficulty levels and key challenges to test models performance under complex domain-specific requirements.