Datasets:

Modalities:
Text
Formats:
parquet
ArXiv:
Libraries:
Datasets
pandas
License:
prompteus commited on
Commit
bffb5d9
1 Parent(s): 7656e0c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +21 -12
README.md CHANGED
@@ -150,19 +150,19 @@ datasets.load_dataset("MU-NLPC/calc-math_qa", "original-splits")
150
 
151
  ## Atributes
152
 
153
- - `id` - id of the example
154
- - `question` - the description of a mathematical problem in natural language, and includes the options to be selected from
155
- - `chain` - solution in the form of step-by-step calculations encoded in simple html-like language. computed from `annotated_formula` column
156
- - `result` - the correct option
157
- - `result_float` - the result converted to a float
158
- - `question_without_options` - same as `question`, but does not contain the options
159
- - `options` - dictionary of options to choose from, one is correct, keys are "A".."E"
160
- - `annotated_formula` - human-annotated nested expression that (approximately) evaluates to the selected correct answer
161
- - `linear_formula` - same as `annotated_formula`, but linearized by original math_qa authors
162
- - `rationale` - human-annotated free-text reasoning that leads to the correct answer
163
- - `category` - category of the math problem
164
 
165
- Attributes `id`, `question`, `chain`, and `result` are present in all datasets in [Calc-X collection](https://huggingface.co/collections/MU-NLPC/calc-x-652fee9a6b838fd820055483).
166
 
167
 
168
  ## Sources
@@ -171,6 +171,15 @@ Attributes `id`, `question`, `chain`, and `result` are present in all datasets i
171
  - [official website](https://math-qa.github.io/)
172
 
173
 
 
 
 
 
 
 
 
 
 
174
  ## Licence
175
 
176
  Apache 2.0, consistently with the original dataset.
 
150
 
151
  ## Atributes
152
 
153
+ - **id** - id of the example
154
+ - **question** - the description of a mathematical problem in natural language, and includes the options to be selected from
155
+ - **chain** - solution in the form of step-by-step calculations encoded in simple html-like language. computed from `annotated_formula` column
156
+ - **result** - the correct option
157
+ - **result_float** - the result converted to a float
158
+ - **question_without_options** - same as `question`, but does not contain the options
159
+ - **options** - dictionary of options to choose from, one is correct, keys are "A".."E"
160
+ - **annotated_formula** - human-annotated nested expression that (approximately) evaluates to the selected correct answer
161
+ - **linear_formula** - same as `annotated_formula`, but linearized by original math_qa authors
162
+ - **rationale** - human-annotated free-text reasoning that leads to the correct answer
163
+ - **category** - category of the math problem
164
 
165
+ Attributes **id**, **question**, **chain**, and **result** are present in all datasets in [Calc-X collection](https://huggingface.co/collections/MU-NLPC/calc-x-652fee9a6b838fd820055483).
166
 
167
 
168
  ## Sources
 
171
  - [official website](https://math-qa.github.io/)
172
 
173
 
174
+ ## Related work
175
+
176
+ This dataset was created as a part of a larger effort in training models capable of using a calculator during inference.
177
+ You can read more in our paper [Calc-X and Calcformers](https://arxiv.org/abs/2305.15017).
178
+ We have released a collection of datasets on solving math problems with calculator interactions on HuggingFace called [Calc-X collection](https://huggingface.co/collections/MU-NLPC/calc-x-652fee9a6b838fd820055483).
179
+ We have also trained and published several T5-based models that use calculator during inference to perform arithmetics correctly, which we call [Calcformers](https://huggingface.co/collections/MU-NLPC/calcformers-65367392badc497807b3caf5).
180
+
181
+
182
+
183
  ## Licence
184
 
185
  Apache 2.0, consistently with the original dataset.