README.md

metadata

license: other
tags:
  - science
  - mathematics
  - reasoning
  - multiple-choice
  - question-answering
task_categories:
  - multiple-choice
  - math
  - question-answering
language:
  - en
size_categories:
  - 1K<n<10K

StreetMath Dataset

Dataset Summary

The Street Math dataset is a synthetic reasoning benchmark that evaluates a model’s ability to approximate sums of decimal prices in everyday shopping scenarios.
Each example presents a list of item prices, and the model must select the approximate total cost (before tax) from multiple-choice options.

The dataset is designed to test numerical reasoning, estimation, and handling of decimal numbers.
Language: English.
Domain: mathematics applied to real-world shopping tasks.

Languages

• English (en): prompts and options are written in plain English, with U.S. dollar formatting for prices.

Data Instances

Example instance:

{
  "id": "basket_sum_000243",
  "topic": "basket_sum",
  "subtopic": "decimal_prices",
  "prompt": "You’re buying these items: $3.55, $15.42, $4.56, $12.63, $6.08. About how much will you pay (before tax)?",
  "labels": ["A", "B", "C", "D"],
  "correct_label": "A",
  "choices": ["$43.00", "$14.11", "$42.24", "$182.80"],
  "correct_option": 0,
  "metadata": {
    "exact_value": 42.24,
    "good_value": 43.0,
    "mild_value": 14.11,
    "way_value": 182.8,
    "prices": [3.55, 15.42, 4.56, 12.63, 6.08]
  },
  "split": "test"
}

Intended Uses

The Basket Sum dataset is intended for:

• Benchmarking language models on basic numerical reasoning and arithmetic in natural language contexts.
• Evaluating estimation skills: testing whether models can provide approximate answers rather than exact calculations.
• Educational and research purposes: studying how models handle everyday math tasks such as adding decimal prices.

This dataset is not intended for:

• Financial or accounting applications.
• Real-world shopping or economic forecasting.
• Any critical decision-making where incorrect numerical outputs could cause harm.

Format

• File type: JSON Lines (.jsonl)
• Each line: one example as a JSON object
• Compatible with: Hugging Face datasets library (load_dataset("json", data_files="..."))

How to Get the Dataset

You can easily load this dataset from the Hugging Face Hub using the datasets library:

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("Chiung-Yi/StreetMath")

# Access the test split
test_dataset = dataset["test"]

# Example: print the first item
print(test_dataset[0])

Limitations and Ethical Considerations

Licensing: The license is currently unspecified. For any public or commercial use, it is necessary to verify the terms with the author.

Dataset Curators

• Original dataset created by Chiung-Yi

Disclaimer

This dataset card was written by a community contributor to improve documentation. If you are the original author or know additional details, feel free to submit a pull request or open an issue to update this card.