Datasets:

TheFinAI
/

en-ccf

pretty_name: CALM · Credit Card Fraud (fraud detection)
license: other
license_name: dbcl-1.0
language:
  - en
task_categories:
  - text-classification
size_categories:
  - 10K<n<100K
source_datasets:
  - extended
tags:
  - finance
  - benchmark
  - thefinai
  - calm
  - flare
extra_gated_heading: Request access to CALM · Credit Card Fraud (fraud detection)
extra_gated_description: >-
  This CALM task is released by The Fin AI for research. Access is granted
  automatically after you complete this short form.
extra_gated_button_content: Agree and access
extra_gated_prompt: >-
  By accessing this dataset you agree to the license of the original source
  ((DbCL) v1.0) and to cite the CALM paper and the original dataset in any
  resulting publication.
extra_gated_fields:
  Full name: text
  Affiliation: text
  Country: country
  Intended use:
    type: select
    options:
      - Research
      - Education
      - Commercial evaluation
      - Other
  I agree to the terms above and will cite the papers: checkbox

CALM · Credit Card Fraud (fraud detection)

📄 Paper · 💻 Code · 🌐 The Fin AI

Part of CALM — Empowering Many, Biasing a Few: Generalist Credit Scoring through Large Language Models (arXiv:2310.00566).


Task	fraud detection
Original dataset	Credit Card Fraud (anonymized, PCA features)
Evaluation metric	Accuracy, MCC, F1, Miss
Source license	(DbCL) v1.0
Language	en

Quick Start

from datasets import load_dataset

ds = load_dataset("TheFinAI/en-ccf", split="test")
print(ds[0])

Example prompt

Detect the credit card fraud using the following financial table attributes. Respond with only 'yes' or 'no', and do not provide any additional information. Therein, the data contains 28 numerical input variables V1, V2, ..., and V28 which are the result of a PCA transformation and 1 input variable Amount which has not been transformed with PCA. The feature 'Amount' is the transaction Amount, this…

Dataset Structure

Split	Rows
`train`	7,974
`validation`	1,139
`test`	2,279

Field	Description
`id`	Example id
`query`	Full instruction prompt given to the model
`answer`	Gold answer / label text
`choices`	Label space
`gold`	Index of the gold label in `choices`
`text`	Raw input text (without instruction)

License

The original data is released under (DbCL) v1.0 (CALM paper, Table 1).

Citation

Please cite CALM and the original dataset (Credit Card Fraud (anonymized, PCA features)):

@misc{feng2024empoweringmanybiasingfew,
      title={Empowering Many, Biasing a Few: Generalist Credit Scoring through Large Language Models},
      author={Duanyu Feng and Yongfu Dai and Jimin Huang and Yifang Zhang and Qianqian Xie and Weiguang Han and Zhengyu Chen and Alejandro Lopez-Lira and Hao Wang},
      year={2024},
      eprint={2310.00566},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2310.00566},
}