Datasets:

Modalities:
Text
Formats:
json
Languages:
Chinese
Libraries:
Datasets
Dask
License:
CPsyExam / README.md
ChillWay's picture
Update README.md
90ec25f verified
metadata
language: zh
license: cc-by-4.0
datasets:
  - CPsyExam

CPsyExam Dataset

CPsyExam is a comprehensive dataset designed for evaluating Chinese psychological examination capabilities in large language models.

Dataset Description

The CPsyExam dataset is structured to assess psychological examination competencies in a Chinese context. It provides a standardized way to evaluate how well language models can handle psychological assessment tasks. The dataset contains multiple-choice questions from various psychological domains, including clinical and counseling psychology.

News 🔥

  • December 9, 2024: Official website launched at https://cpsyexam.xinhai.co
  • November 30, 2024: Our paper "CPsyExam: Chinese Benchmark for Evaluating Psychology using Examinations" has been accepted by COLING 2025! 🎉

Data Format

The dataset is organized in JSON format with the following splits:

  • Training set
  • Validation set
  • Test set

Each example in the dataset follows this structure:

{
    "subject_name": "心理学科目名称",
    "question_type": "multi/single",
    "kind": "knowledge",
    "question": "题目内容",
    "options": {
        "A": "选项A内容",
        "B": "选项B内容",
        "C": "选项C内容",
        "D": "选项D内容",
        "E": "选项E内容"
    },
    "explanation": "题目解析说明",
    "id": "唯一标识ID"
}

Field Descriptions:

  • subject_name: The specific subject area in psychology
  • question_type: Type of question (e.g., "multi" for multiple choice)
  • kind: Category of the question (e.g., "knowledge")
  • question: The actual question text
  • options: A dictionary containing the multiple choice options (A through E)
  • explanation: Detailed explanation of the correct answer
  • id: Unique identifier for the question

Usage

Loading the Dataset

from datasets import load_dataset

dataset = load_dataset("CAS-SIAT-XinHai/CPsyExam")

Data Preparation for Fine-tuning

For those interested in using the dataset for fine-tuning, you can prepare the data according to the SFT section.

Leaderboard

Track model performance and submit your results at our leaderboard.

Citation

If you use this dataset in your research, please cite:

@misc{CPsyExam benchmark,
  title={CPsyExam: Chinese Benchmark for Evaluating Psychology using Examinations},
  author={Jiahao Zhao, Jingwei Zhu, Minghuan Tan, Min Yang, Di Yang, 
          Chenhao Zhang, Guancheng Ye, Chengming Li, Xiping Hu},
  year = {2024},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/CAS-SIAT-XinHai/CPsyExam}}
}

License

This dataset is licensed under CC-BY-4.0.

Contact

For more information, visit our official website.