Unverified85% confidenceFactTime unknown
Codex模型的评估通常使用HumanEval和MBPP等编码基准测试
1
Sources
85%
Confidence
Long-term
Relevance
6/2/2026
First Seen
Sources
用OpenAI API构建智能编程助手实战指南
bilibiliKurashizu
Related Entities
Related Claims
Unverified代码大语言模型采用HumanEval、MBPP等可执行基准评估，即直接运行生成代码并检验输出是否正确79% similarUnverified传统代码评测基准（如HumanEval、MBPP）主要测试模型在孤立函数级别的生成准确率77% similarUnverifiedHumanEval、MBPP、SWE-bench是AI编程评测领域的行业标准基准测试76% similarUnverifiedHumanEval、MBPP、SWE-bench是专门针对代码生成的标准评测集74% similarUnverifiedHumanEval 和 MBPP 主要衡量模型代码生成能力，而代码审查是本质不同的任务，业界目前缺乏标准化的代码审查基准测试74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/37190API
curl https://kongchang.com/api/v1/knowledge/claims/37190MCP
get_claim(id=37190)