HumanEval

HumanEval: hand-written Python programming problems (Codex paper)

164 道手工编写的 Python 编程题,每题为函数签名加文档字符串,通过运行隐藏单元测试评分,以 pass@k(通常为 pass@1)报告功能正确性。它引入了基于执行的代码评估,是 2021 至 2024 年的标准代码基准;前沿模型如今几乎能解出所有题目,且该题集已广泛存在于训练数据中。

99.3%
o4-mini-high
厂商自报
2021-07-07 → 2026-04-22: 28.8% → 76.8%
发布
2021-07
维护者
OpenAI
状态
saturated
污染风险
high
指标
pass@1 (percent, ↑)
题量
164
领域
code
人类
无实测基线

备注. Many papers report EvalPlus (HumanEval+) numbers with extra tests, which are lower and not directly comparable. OpenAI's simple-evals stopped reporting new models in July 2025.

完整账本

系统开发者分数日期来源条件
o4-mini-highOpenAI99.3%厂商自报pass_k: 1 shots: 0
simple-evals README; effectively the ceiling (163 of 164).
gpt-4-turbo-2024-04-09OpenAI88.2%厂商自报pass_k: 1 shots: 0
simple-evals README.
DeepSeek-V4-Pro-BaseDeepSeek76.8%厂商自报pass_k: 1 shots: 0
DeepSeek-V4 model card, base-model table (HF repo created 2026-04-22). Base (pre-trained, not instruction-tuned) model, so lower than chat models above.
Codex 12BOpenAI28.8%论文pass_k: 1
Codex paper abstract; Codex-S 37.7, GPT-J 11.4, GPT-3 0. 70.2% with 100 samples per problem.