Peking University Proposes RepoZero, the First Verifiable Repository-Level Generation Benchmark to Evaluate LLM's Ability to Generate Code Repositories from Scratch
夕小瑶科技说73 字 (约 1 分钟)
70
Peking University has developed RepoZero, the first verifiable repository-level generation benchmark to assess LLMs' capability of generating complete code repositories from scratch, revealing significant limitations in current models for complex code structure generation.
入选理由:RepoZero通过功能完整性、代码质量、文档一致性等12项指标,量化评估LLM生成代码仓库的可行性
FeaturedArticle#Code Generation#LLM Evaluation#Benchmark Testing#Peking University中文