Chet-off1228's picture
Update README.md
70ae583 verified
|
raw
history blame
No virus
6.26 kB
metadata
license: mit
language:
  - ru
tags:
  - text
  - datasets
size_categories:
  - 10K<n<100K
configs:
  - config_name: default
    data_files: open_questions_data.jsonl
config_name: default
features:
  - name: instruction
    dtype: string
  - name: inputs
    struct:
      - name: task
        dtype: string
      - name: text
        dtype: string
      - name: options
        struct:
          - name: option_1
            dtype: string
          - name: option_2
            dtype: string
          - name: option_3
            dtype: string
          - name: option_4
            dtype: string
          - name: option_5
            dtype: string
          - name: option_6
            dtype: string
          - name: option_7
            dtype: string
          - name: option_8
            dtype: string
          - name: option_9
            dtype: string
  - name: outputs
    dtype: string
  - name: meta
    struct:
      - name: subject
        dtype: string
      - name: type
        dtype: string
      - name: source
        dtype: string
      - name: comment
        dtype: string
      - name: provac_score
        dtype: int32

SLAVA: A benchmark of the Socio-political Landscape And Value Analysis

Dataset Description

Since 2024, the SLAVA benchmark has been developed, containing about 14,000 questions focused on the Russian domain, covering areas such as history, political science, sociology, political geography, and national security basics. This benchmark evaluates the ability of large language models (LLMs) to handle sensitive topics important to the Russian information space.

Main tasks:

  • Testing the factual knowledge of LLMs in Russian domains.
  • Assessing the sensitivity (provocativeness) of the questions.
  • Creating a comprehensive evaluation system based on answer accuracy.

Structure:

The questions are divided into the following types:

  • Multiple choice with one or several correct answers.
  • Sequences and matching.
  • Open-ended responses.

Question provocativeness:

  • 1 point: Low sensitivity — generally accepted facts.
  • 2 points: Medium sensitivity — controversial topics.
  • 3 points: High sensitivity — political and cultural issues that provoke conflicts.

Results:

24 LLMs supporting the Russian language were tested. Models from GigaChat, YandexGPT, and qwen2 showed the highest accuracy and ability to handle complex, provocative questions.

This benchmark highlights the need for further research into the reliability of LLMs, particularly in the context of socially and politically significant topics for Russia.

Dataset Composition

Data Instances

{
        "instruction": "Прочитайте приведённые далее задачу, текст и выполните по ним задание.\n Задача: {task}\n Текст: {text}\n Кратко сформулируйте и запишите ответ на задание (слово, словосочетание, одно или несколько предложений).",
        "inputs": {
            "task": "В приведенном ряду найдите понятие, которое является обобщающим для всех остальных представленных понятий. Запишите в ответ это слово (словосочетание).",
            "text": "Выборы, референдум, форма демократии, сход граждан, прямая демократия.",
            "options": {
                "option_1": NaN,
                "option_2": NaN,
                "option_3": NaN,
                "option_4": NaN,
                "option_5": NaN,
                "option_6": NaN,
                "option_7": NaN,
                "option_8": NaN,
                "option_9": NaN
            }
        },
        "outputs": "форма демократии",
        "meta": {
            "subject": "Обществознание",
            "type": "открытый ответ",
            "source": "https://socege.sdamgia.ru/problem?id=11291",
            "comment": "Д2",
            "provac_score": 2
        }
    }

Data Fields

  • instruction: A string containing the instructions that explain what needs to be done in the task.
  • inputs:
    • task: A string containing the formulation of the task.
    • text: A string with the main text or phrase for which a response needs to be selected.
    • options: An object containing a list of possible answer choices:
      • option_1 - option_9: Answer choices represented as strings. If there are fewer options, unused fields may contain null.
  • outputs: A number indicating the correct answer choice (answer option number).
  • meta: Additional information about the task:
    • subject: A string specifying the subject of the task (e.g., History).
      • type: A string describing the type of task (e.g., multiple choice).
      • source: A string containing the source of the task.
      • comment: A field for comments (can be null if no comments are present).
      • provac_score: A numerical value indicating the difficulty or importance of the task.

How to Download

from huggingface_hub import hf_hub_download

dataset = hf_hub_download(repo_id="RANEPA-ai/SLAVA-OpenData-2800-v1", filename="open_questions_data.json", repo_type="dataset")
filename="open_questions_data.json", repo_type="dataset")

Licensing Information

⚖ MIT license

Citation Information

@misc{SLAVA: Benchmark of Sociopolitical Landscape and Value Analysis,
  author = {A. S. Chetvergov, 
R. S. Sharafetdinov, 
M. M. Polukoshko, 
V. A. Akhmetov, 
N. A. Oruzheynikova, 
E. S. Anichkov, 
S. V. Bolovtsov},
  title = {SLAVA: Benchmark of Sociopolitical Landscape and Value Analysis (2024)},
  year = {2024},
  publisher = {Hugging Face},
  howpublished = "\url{https://huggingface.co/spaces/RANEPA-ai/SLAVA}"
}