Datasets:
ArXiv:
License:
Update README.md
Browse files
README.md
CHANGED
@@ -8,6 +8,8 @@ dataset_info:
|
|
8 |
dtype: string
|
9 |
- name: tables
|
10 |
dtype: sequence
|
|
|
|
|
11 |
- name: topic
|
12 |
dtype: string
|
13 |
- name: python_solution
|
@@ -15,19 +17,16 @@ dataset_info:
|
|
15 |
- name: ground_truth
|
16 |
dtype: string
|
17 |
---
|
18 |
-
##
|
19 |
|
20 |
**KnowledgeMath** is a knowledge-intensive dataset focused on mathematical reasoning within the domain of finance. It requires the model to comprehend specialized financial terminology and to interpret tabular data presented in the questions.
|
21 |
**KnowledgeMath** includes **1200 QA examples** across 7 key areas in finance. These examples were collected from financial experts and feature detailed solution annotations in Python format.
|
22 |
|
23 |
-
## Dataset Information
|
24 |
-
|
25 |
- Paper: https://arxiv.org/abs/2311.09797
|
26 |
- Code: https://github.com/yale-nlp/KnowledgeMath
|
27 |
- Leaderboard: will be released soon!
|
28 |
|
29 |
-
|
30 |
-
|
31 |
All the data examples were divided into two subsets: *validation* and *test*.
|
32 |
|
33 |
- **validation**: 200 examples used for model development, validation, or for those with limited computing resources.
|
@@ -39,11 +38,7 @@ You can download this dataset by the following command:
|
|
39 |
from datasets import load_dataset
|
40 |
|
41 |
dataset = load_dataset("yale-nlp/KnowledgeMath")
|
42 |
-
```
|
43 |
|
44 |
-
Here are some examples of how to access the downloaded dataset:
|
45 |
-
|
46 |
-
```python
|
47 |
# print the first example on the validation set
|
48 |
print(dataset["validation"][0])
|
49 |
|
@@ -51,8 +46,6 @@ print(dataset["validation"][0])
|
|
51 |
print(dataset["test"][0])
|
52 |
```
|
53 |
|
54 |
-
### Data Format
|
55 |
-
|
56 |
The dataset is provided in json format and contains the following attributes:
|
57 |
|
58 |
```json
|
@@ -62,11 +55,12 @@ The dataset is provided in json format and contains the following attributes:
|
|
62 |
"tables": [list] List of Markdown-format tables associated with the question,
|
63 |
"python_solution": [string] Python-format and executable solution by financial experts. The code is written in a clear and executable format, with well-named variables and a detailed explanation,
|
64 |
"ground_truth": [integer] Executed result of `python solution`, rounded to three decimal places,
|
65 |
-
"topic": [string] The related financial area of the question
|
|
|
66 |
}
|
67 |
```
|
68 |
|
69 |
-
|
70 |
|
71 |
To automatically evaluate a model on **KnowledgeMath**, please refer to our GitHub repository [here](https://github.com/yale-nlp/KnowledgeMath).
|
72 |
|
|
|
8 |
dtype: string
|
9 |
- name: tables
|
10 |
dtype: sequence
|
11 |
+
feature:
|
12 |
+
dtype: string
|
13 |
- name: topic
|
14 |
dtype: string
|
15 |
- name: python_solution
|
|
|
17 |
- name: ground_truth
|
18 |
dtype: string
|
19 |
---
|
20 |
+
## KnowledgeMath Benchmark Description
|
21 |
|
22 |
**KnowledgeMath** is a knowledge-intensive dataset focused on mathematical reasoning within the domain of finance. It requires the model to comprehend specialized financial terminology and to interpret tabular data presented in the questions.
|
23 |
**KnowledgeMath** includes **1200 QA examples** across 7 key areas in finance. These examples were collected from financial experts and feature detailed solution annotations in Python format.
|
24 |
|
|
|
|
|
25 |
- Paper: https://arxiv.org/abs/2311.09797
|
26 |
- Code: https://github.com/yale-nlp/KnowledgeMath
|
27 |
- Leaderboard: will be released soon!
|
28 |
|
29 |
+
## KnowledgeMath Dataset Information
|
|
|
30 |
All the data examples were divided into two subsets: *validation* and *test*.
|
31 |
|
32 |
- **validation**: 200 examples used for model development, validation, or for those with limited computing resources.
|
|
|
38 |
from datasets import load_dataset
|
39 |
|
40 |
dataset = load_dataset("yale-nlp/KnowledgeMath")
|
|
|
41 |
|
|
|
|
|
|
|
42 |
# print the first example on the validation set
|
43 |
print(dataset["validation"][0])
|
44 |
|
|
|
46 |
print(dataset["test"][0])
|
47 |
```
|
48 |
|
|
|
|
|
49 |
The dataset is provided in json format and contains the following attributes:
|
50 |
|
51 |
```json
|
|
|
55 |
"tables": [list] List of Markdown-format tables associated with the question,
|
56 |
"python_solution": [string] Python-format and executable solution by financial experts. The code is written in a clear and executable format, with well-named variables and a detailed explanation,
|
57 |
"ground_truth": [integer] Executed result of `python solution`, rounded to three decimal places,
|
58 |
+
"topic": [string] The related financial area of the question,
|
59 |
+
"knowledge_terms": [list] List of knowledge terms in our constructed knowledge bank that is necessary to answer the given question. We will release this feature upon paper publication
|
60 |
}
|
61 |
```
|
62 |
|
63 |
+
## Automated Evaluation
|
64 |
|
65 |
To automatically evaluate a model on **KnowledgeMath**, please refer to our GitHub repository [here](https://github.com/yale-nlp/KnowledgeMath).
|
66 |
|