RaymondLi/perturbed_humaneval

---
license: apache-2.0
---

# Dataset Card for Dataset Name

## Dataset Description

- **Repository:** https://github.com/amazon-science/recode/tree/main
- **Paper:** https://arxiv.org/abs/2212.10264

### Dataset Summary

The Recode benchmark proposes to apply code and natural language transformations to code-generation benchmarks to evaluate the robustness of code-generation models.
This dataset contains the perturbed version of HumanEval that they released.
It was automatically generated from the [HumanEval](https://huggingface.co/datasets/openai_humaneval) dataset.

### Subsets

There are four transformation categories that form the subsets of this dataset: `func_name`, `nlaugmenter`, `natgen` and `format`.

### Languages

The programming problems are written in Python and contains docstrings and comments in English.

## Dataset Structure

### Data Instances

[More Information Needed]

### Data Fields

- `task_id`: ID of the original HumanEval example
- `prompt`: the perturbed prompt
- `entry_point`: entry point for test
- `canonical_solution`: solution for the problem in the `prompt`
- `test`: contains function to test generated code for correctness
- `seed`: seed of the perturbed prompt
- `perturbation_name`: name of the perturbation
- `partial`: partial solution to the problem. This field is only present for transformation categories that affect a partial solution: `natgen` and `format`.

### Data Splits

The dataset only has a test split.

## Dataset Creation

### Curation Rationale

[More Information Needed]

### Source Data

#### Initial Data Collection and Normalization

[More Information Needed]

#### Who are the source language producers?

[More Information Needed]

### Annotations

#### Annotation process

[More Information Needed]

#### Who are the annotators?

[More Information Needed]

### Personal and Sensitive Information

[More Information Needed]

## Considerations for Using the Data

### Social Impact of Dataset

[More Information Needed]

### Discussion of Biases

[More Information Needed]

### Other Known Limitations

[More Information Needed]

## Additional Information

### Dataset Curators

[More Information Needed]

### Licensing Information

[More Information Needed]

### Citation Information

```
@article{wang2022recode,
title={ReCode: Robustness Evaluation of Code Generation Models},
author={Wang, Shiqi and Li, Zheng and Qian, Haifeng and Yang, Chenghao and Wang, Zijian and Shang, Mingyue and Kumar, Varun and Tan, Samson and Ray, Baishakhi and Bhatia, Parminder and others},
journal={arXiv preprint arXiv:2212.10264},
year={2022}
}
```

### Contributions

[More Information Needed]

原始信息汇总

数据集概述

数据集描述

• 名称: Recode Benchmark
• 目的: 评估代码生成模型的鲁棒性
• 来源: 自动生成的扰动版本HumanEval数据集
• 语言: Python编程语言，包含英文的docstrings和comments

数据集结构

数据字段

• task_id: 原始HumanEval示例的ID
• prompt: 扰动后的提示
• entry_point: 测试入口点
• canonical_solution: 问题的解决方案
• test: 用于测试生成代码正确性的函数
• seed: 扰动提示的种子
• perturbation_name: 扰动名称
• partial: 问题的部分解决方案（仅在natgen和format类别中存在）

数据分割

• 分割: 仅包含测试集

数据集创建

数据集子集

• 子集: 四个变换类别：func_name, nlaugmenter, natgen, format

数据来源

• 原始数据: HumanEval数据集
• 扰动生成: 自动生成

许可证

• 许可证: Apache-2.0