README.md

metadata

license: apache-2.0
size_categories:
  - n<1K
language:
  - ru
pretty_name: Russian Math Problems
dataset_info:
  - config_name: RussianMath
    features:
      - name: task
        dtype: string
      - name: solution
        dtype: string
      - name: short answer
        dtype: string
      - name: class
        dtype: string
      - name: grade
        dtype: int64
    splits:
      - name: train
        num_examples: 199
        num_bytes: 756742
configs:
  - config_name: RussianMath
    data_files:
      - split: train
        path: data/0000.parquet
extra_gated_prompt: >-
  You agree to not use the dataset to conduct experiments that cause harm to
  human subjects, or train on it
extra_gated_fields:
  Company: text
  Name: text
  Surname: text
  Country: country
  Email: text
  I want to use this dataset for:
    type: select
    options:
      - Research
      - Education
      - Industry
      - label: Other
        value: other
  I agree to use this dataset for non-commercial use ONLY: checkbox

Russian math

Датасет собранный из матетматических олимпиад разного уровня, class означает происхождение задачи по уровню олимпиады.

Dataset collected from matetmatic olympiads of different levels, class means the origin of the problem by the level of the olympiad.

@misc{nikolich2024vikhr,
      title={Vikhr: The Family of Open-Source Instruction-Tuned Large Language Models for Russian},
      author={Aleksandr Nikolich and Konstantin Korolev and Artem Shelmanov and Igor Kiselev},
      year={2024},
      eprint={2405.13929},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2405.13929},
}