Datasets:

Modalities:
Text
Formats:
parquet
ArXiv:
Libraries:
Datasets
pandas
License:
Calc-math_qa / README.md
prompteus's picture
Update README.md
102358b
metadata
license: apache-2.0
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
  - config_name: original-splits
    data_files:
      - split: train
        path: original-splits/train-*
      - split: validation
        path: original-splits/validation-*
      - split: test
        path: original-splits/test-*
dataset_info:
  - config_name: default
    features:
      - name: id
        dtype: string
      - name: question
        dtype: string
      - name: chain
        dtype: string
      - name: result
        dtype: string
      - name: result_float
        dtype: float64
      - name: question_without_options
        dtype: string
      - name: options
        struct:
          - name: A
            dtype: string
          - name: B
            dtype: string
          - name: C
            dtype: string
          - name: D
            dtype: string
          - name: E
            dtype: string
      - name: annotated_formula
        dtype: string
      - name: linear_formula
        dtype: string
      - name: rationale
        dtype: string
      - name: category
        dtype: string
    splits:
      - name: train
        num_bytes: 25058735
        num_examples: 20868
    download_size: 11157481
    dataset_size: 25058735
  - config_name: original-splits
    features:
      - name: id
        dtype: string
      - name: question
        dtype: string
      - name: chain
        dtype: string
      - name: result
        dtype: string
      - name: result_float
        dtype: float64
      - name: question_without_options
        dtype: string
      - name: options
        struct:
          - name: A
            dtype: string
          - name: B
            dtype: string
          - name: C
            dtype: string
          - name: D
            dtype: string
          - name: E
            dtype: string
      - name: annotated_formula
        dtype: string
      - name: linear_formula
        dtype: string
      - name: rationale
        dtype: string
      - name: category
        dtype: string
    splits:
      - name: train
        num_bytes: 25058735
        num_examples: 20868
      - name: validation
        num_bytes: 3722848
        num_examples: 3102
      - name: test
        num_bytes: 2423833
        num_examples: 2029
    download_size: 13928430
    dataset_size: 31205416

Dataset Card for Calc-math_qa

Summary

This dataset is an instance of math_qa dataset, converted to a simple HTML-like language that can be easily parsed (e.g. by BeautifulSoup). The data contains 3 types of tags:

  • gadget: A tag whose content is intended to be evaluated by calling an external tool (sympy-based calculator in this case)
  • output: An output of the external tool
  • result: The final answer of the mathematical problem (correct option)

Supported Tasks

The dataset is intended for training Chain-of-Thought reasoning models able to use external tools to enhance the factuality of their responses. This dataset presents in-context scenarios where models can outsource the computations in the reasoning chain to a calculator.

Construction Process

We took the original math_qa dataset, parsed the nested formulas, linearized them into a sequence (chain) of operations, and replaced all advanced function calls (such as circle_area) with explicit elementary operations. We evaluate all the steps in each example and filter out examples if their evaluation does not match the answer selected as correct in the data with a 5% tolerance, with about 26k examples remaining. The sequence of steps is then saved in HTML-like language in the chain column.

We also perform in-dataset and cross-dataset data-leak detection within Calc-X collection. Specifically for MathQA, we found that majority of validation and test examples are near-duplicates of some example in the train set, and that all validation and test examples likely originate from the Aqua-RAT train split. We do not recommend to original validation and test splits of the MathQA dataset.

You can read more information about this process in our Calc-X paper.

Data splits

In our default configuration, test and validation splits are removed and we recommend using MathQA for training only. You can load it using:

datasets.load_dataset("MU-NLPC/calc-math_qa")

If you want to use the original dataset splits, you can load it using:

datasets.load_dataset("MU-NLPC/calc-math_qa", "original-splits")

Atributes

  • id - id of the example
  • question - the description of a mathematical problem in natural language, and includes the options to be selected from
  • chain - solution in the form of step-by-step calculations encoded in simple html-like language. computed from annotated_formula column
  • result - the correct option
  • result_float - the result converted to a float
  • question_without_options - same as question, but does not contain the options
  • options - dictionary of options to choose from, one is correct, keys are "A".."E"
  • annotated_formula - human-annotated nested expression that (approximately) evaluates to the selected correct answer
  • linear_formula - same as annotated_formula, but linearized by original math_qa authors
  • rationale - human-annotated free-text reasoning that leads to the correct answer
  • category - category of the math problem

Attributes id, question, chain, and result are present in all datasets in Calc-X collection.

Sources

Related work

This dataset was created as a part of a larger effort in training models capable of using a calculator during inference, which we call Calcformers. We have released a collection of datasets on solving math problems with calculator interactions on HuggingFace called Calc-X collection. You can find the models we trained in the Calcformers collection. You can read more in our paper Calc-X and Calcformers.

Licence

Apache 2.0, consistently with the original dataset.

Cite

If you use this version of dataset in research, please cite the original MathQA paper, and Calc-X paper as follows:

@inproceedings{kadlcik-etal-2023-soft,
    title = "Calc-X and Calcformers: Empowering Arithmetical Chain-of-Thought through Interaction with Symbolic Systems",
    author = "Marek Kadlčík and Michal Štefánik and Ondřej Sotolář and Vlastimil Martinek",
    booktitle = "Proceedings of the The 2023 Conference on Empirical Methods in Natural Language Processing: Main track",
    month = dec,
    year = "2023",
    address = "Singapore, Singapore",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/2305.15017",
}