The dataset viewer is not available for this subset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/harbor/harbor.py", line 171, in _split_generators
raise DataFilesNotFoundError("No task.toml or instruction.md files found")
datasets.exceptions.DataFilesNotFoundError: No task.toml or instruction.md files found
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Russian University Mathematics — Handwritten and LaTeX Solutions
A collection of real handwritten mathematical solutions paired with their corresponding expert-written digital solutions.
This dataset is a 50-problem evaluation sample from a larger collection of 1,740 Russian-language mathematics problems, created for research and development in mathematical OCR, document understanding, multimodal AI and mathematical reasoning.
Dataset overview
| Property | Value |
|---|---|
| Problems | 50 |
| Mathematical branches | 19 |
| Different topics | 49 |
| Original handwritten pages | 117 |
| Problems containing figures | 29 |
| Mathematical formulas | Approximately 1,600 |
| Language | Russian |
| Handwriting author | One human expert |
| Creation period | 2010–2020 |
What is included?
Each problem folder contains the available source materials:
- Original handwritten solution pages (JPEG/PNG)
- Typed solution in Microsoft Word (DOCX)
- Typed solution in PDF
- Markdown solution with LaTeX mathematical formulas
- Standalone LaTeX document
- Mathematical figures and diagrams
- Problem statement and plain-text summary
The handwritten solutions and typed solutions correspond to the same problem. They are aligned at the problem level, not necessarily line by line.
Mathematical coverage
The sample covers 19 mathematical branches:
- Algebra and Linear Algebra
- Mathematical Analysis
- Analytic Geometry
- Calculus of Variations
- Geometry
- Discrete Mathematics
- Differential Geometry
- Differential Equations
- Integrals and Field Theory
- Complex Analysis
- Mathematical Physics and PDEs
- Mathematical Olympiad Problems
- Numerical Methods
- Probability Theory
- Topology
- Russian State Examination (ЕГЭ)
- Mathematical Economics
- Functional Analysis
- Mathematical Statistics
Intended applications
Handwritten mathematical OCR
Training and evaluating models that recognize handwritten mathematical documents containing Cyrillic text, formulas and mathematical notation.
Image-to-LaTeX
Developing and evaluating systems that convert complete handwritten mathematical solutions into structured LaTeX.
LLM mathematical reasoning
Research into supervised fine-tuning and evaluation of models on multi-step mathematical problems.
Multimodal document understanding
Connecting handwritten pages, mathematical figures, problem statements and corresponding typed solutions.
Data creation and provenance
All solutions were originally handwritten and subsequently typed by the same mathematics expert, Tigran Grigoryan, between 2010 and 2020.
The typed versions were prepared in Microsoft Word with native mathematical equations. Markdown and LaTeX versions were subsequently generated from the Word documents and checked.
The problem statements originate from standard educational sources, including textbooks, problem collections, coursework and examination materials. No ownership of the original problem statement texts is claimed.
The handwriting belongs to the dataset author. No third-party customer documents or third-party personal data are included.
Dataset structure
Each mathematical problem has its own directory organized by mathematical branch and topic.
Files include DOCX, PDF, Markdown, TEX, JPEG and PNG formats.
The sample includes manifest.csv and manifest.jsonl describing its contents.
Important limitations
- The dataset is Russian-language only.
- The handwritten and typed versions are paired by problem, not line by line.
- Handwriting page resolution and writing styles may vary.
- The sample is intended for evaluation and research, not redistribution.
- Some mathematical problem statements originate from third-party educational sources.
Licensing and access
This repository provides a free 50-problem evaluation sample only.
Access is subject to the terms in SAMPLE_LICENSE.md.
The sample is not licensed for unrestricted redistribution or commercial training.
Full commercial dataset available
The complete collection contains:
- 1,740 mathematical problems
- 3,450 handwritten solution pages
- 50,603 mathematical formulas
- 927 mathematical figures
- 19 branches and 127 topics
- Approximately 2.2 GB of data
Available formats include DOCX, PDF, Markdown, LaTeX, JPEG and PNG.
Full dataset price: USD 45,000
Selected mathematical branches are also available.
A non-exclusive commercial AI training and evaluation licence is offered for the full dataset. Commercial model deployment and publication of trained model weights are permitted under the commercial licence. Redistribution of raw data is not permitted.
Contact and commercial inquiries
Tigran Grigoryan
Individual Entrepreneur — Armenia
Email: tigran.grigoryan83@gmail.com
For commercial licensing, larger dataset access, custom subsets or further information, please contact me directly.
Serious inquiries from AI companies, OCR developers, research laboratories and universities are welcome.
- Downloads last month
- 170