File size: 1,581 Bytes
0d17630
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
---
license: mit
task_categories:
- question-answering
- text-generation
language:
- en
tags:
- math
- reasoning
- synthetic
size_categories:
- 1K<n<10K
---

# GSM-DC Test Dataset

This dataset contains the test set for GSM-DC (Grade School Math with Distractor Chains), a synthetic math reasoning dataset with controlled complexity.

## Dataset Details

- **Total Problems**: 6300
- **Operation Counts (OP)**: 16-22 (out-of-distribution test set)
- **Problem Types**: Graph-based mathematical reasoning problems
- **Noise Levels**: Light, Medium, Hard (distractor difficulty)

## Dataset Structure

Each problem in `all_problems.json` contains:
- `problem_text`: The problem statement with all variables and relationships
- `solution`: Step-by-step ground truth solution
- `final_answer`: The numerical answer
- `n_op`: Number of operations (16-22)
- `noise_level`: Distractor difficulty (light/medium/hard)
- `graph_structure`: Internal graph representation
- `template_id`: Problem template identifier

## Usage

```python
import json

# Load the dataset
with open('all_problems.json', 'r') as f:
    problems = json.load(f)

# Access a problem
problem = problems[0]
print(problem['problem_text'])
print(problem['solution'])
print(problem['final_answer'])
```

## Citation

If you use this dataset, please cite:

```bibtex
@inproceedings{gsm-dc-2025,
    title={GSM-DC: Grade School Math with Distractor Chains},
    author={[Your Name]},
    booktitle={Proceedings of EMNLP 2025},
    year={2025}
}
```

## Paper

Published at EMNLP 2025. [Paper Link]

## License

MIT License