Datasets:

TheFinAI
/

en-finqa-reasoning

pretty_name: FinQA — reasoning-eval format
license: mit
language:
  - en
task_categories:
  - question-answering
tags:
  - finance
  - thefinai
  - fino1
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: val
        path: data/val-*
      - split: test
        path: data/test-*
dataset_info:
  features:
    - name: id
      dtype: string
    - name: Open-ended Verifiable Question
      dtype: string
    - name: Ground-True Answer
      dtype: string
  splits:
    - name: train
      num_bytes: 26435291
      num_examples: 6251
    - name: val
      num_bytes: 4731697
      num_examples: 1147
    - name: test
      num_bytes: 4731697
      num_examples: 1147
  download_size: 16335791
  dataset_size: 35898685
extra_gated_heading: Request access to FinQA — reasoning-eval format
extra_gated_description: >-
  FinQA — reasoning-eval format 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 its license and to cite the Fino1 paper
  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 paper: checkbox

FinQA — reasoning-eval format

📄 Paper · 💻 Code · 🌐 The Fin AI

Used in Fino1 — Fino1: On the Transferability of Reasoning-Enhanced LLMs and Reinforcement Learning to Finance (arXiv:2502.08127). Formerly TheFinAI/FINQA_test (the old name redirects here).

FinQA (Chen et al., 2021) in the open-ended verifiable-question format used to evaluate reasoning models in Fino1. The FinBen prompt format of FinQA is en-finqa. FinQA is MIT-licensed.


Task	numerical QA over financial reports
Language	en
License	mit

Quick Start

from datasets import load_dataset

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

Split	Rows
`train`	6,251
`val`	1,147
`test`	1,147

Columns: id, Open-ended Verifiable Question, Ground-True Answer

Citation

@misc{qian2025fino1transferabilityreasoningenhancedllms,
      title={Fino1: On the Transferability of Reasoning-Enhanced LLMs and Reinforcement Learning to Finance}, 
      author={Lingfei Qian and Weipeng Zhou and Yan Wang and Xueqing Peng and Han Yi and Yilun Zhao and Jimin Huang and Qianqian Xie and Jian-yun Nie},
      year={2025},
      eprint={2502.08127},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2502.08127}, 
}