---
dataset_info:
  features:
  - name: id
    dtype: string
  - name: question
    dtype: string
  - name: answer
    dtype: string
  - name: category
    dtype: string
  - name: difficulty
    dtype: string
  - name: prompt
    dtype: string
  splits:
  - name: train
    num_bytes: 236134
    num_examples: 299
  - name: test
    num_bytes: 160017
    num_examples: 201
  download_size: 150919
  dataset_size: 396151
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
---

# TeleMath, processed

The [netop/TeleMath](https://huggingface.co/datasets/netop/TeleMath) benchmark, 500 telecom
mathematics problems with numerical answers, split into train and test.

TeleMath ships a single split. This one is stratified by category, 299 questions for
train and 201 for test, with a fixed seed (0), so every category appears on both sides
in proportion to its size. It is the split every experiment in
[TeleMath-router](https://huggingface.co/datasets/ymoslem/TeleMath-router) uses.

| Column | Meaning |
|---|---|
| `id` | the question's id in netop/TeleMath |
| `question` | the question as published |
| `answer` | the numerical reference answer |
| `category`, `difficulty` | as published |
| `prompt` | the question followed by the answer-format instruction the models were given |

The instruction appended to every question:

> Solve the problem and give the final numerical answer, in the unit stated in the question, inside \boxed{}.