README.md

---
license: mit
annotations_creators:
- user-generated
language:
- en
task_categories:
- text-classification
dataset_info:
  features:
  - name: id
    dtype: string
    description: Unique identifier for each article
  - name: date
    dtype: string
    description: Date of the article
  - name: time_period
    dtype: string
    description: Historical period classification
  - name: weather_type
    dtype: string
    description: Type of weather event
  - name: text
    dtype: string
    description: Article text content
  - name: infrastructural_impact
    dtype: int64
    description: Binary label for infrastructural impact (0 or 1)
  - name: political_impact
    dtype: int64
    description: Binary label for political impact (0 or 1)
  - name: financial_impact
    dtype: int64
    description: Binary label for financial impact (0 or 1)
  - name: ecological_impact
    dtype: int64
    description: Binary label for ecological impact (0 or 1)
  - name: agricultural_impact
    dtype: int64
    description: Binary label for agricultural impact (0 or 1)
  - name: human_health_impact
    dtype: int64
    description: Binary label for human health impact (0 or 1)
  splits:
    train:
      name: train
      num_bytes: 3300000
      num_examples: 970
    validation:
      name: validation
      num_bytes: 669000
      num_examples: 208
    test:
      name: test
      num_bytes: 582000
      num_examples: 208
  download_size: 4551000
  dataset_size: 4551000
  configs:
  - config_name: default
    data_files:
    - split: train
      path: train.csv
    - split: validation
      path: validation.csv
    - split: test
      path: test.csv
size_categories:
- 1K<n<10K
---
# WX Impact Benchmark - Mixed Context Dataset

This dataset contains the Mixed Context Weather Impact Benchmark Dataset with multi-label impact classifications.

## Dataset Description

This is the MixedCTX_Dataset(1386) subset of the WX Impact Benchmark, containing weather-related articles with various context lengths and complexity levels.

- **Train set**: Training data for model development
- **Validation set**: Validation data for hyperparameter tuning
- **Test set**: Test data for final evaluation

## Features

- `id`: Unique identifier for each article
- `date`: Date of the article
- `time_period`: Historical period classification
- `weather_type`: Type of weather event
- `text`: Article text content
- `infrastructural_impact`: Binary label for infrastructural impact
- `political_impact`: Binary label for political impact
- `financial_impact`: Binary label for financial impact
- `ecological_impact`: Binary label for ecological impact
- `agricultural_impact`: Binary label for agricultural impact
- `human_health_impact`: Binary label for human health impact

## Usage

```python
from datasets import load_dataset

# Load the Mixed Context dataset
dataset = load_dataset("Michaelyya/wximpactbench-1386")

# Access splits
train_data = dataset["train"]
val_data = dataset["validation"]
test_data = dataset["test"]
```

## Cite
@misc{yu2025wximpactbenchdisruptiveweatherimpact,
      title={WXImpactBench: A Disruptive Weather Impact Understanding Benchmark for Evaluating Large Language Models}, 
      author={Yongan Yu and Qingchen Hu and Xianda Du and Jiayin Wang and Fengran Mo and Renee Sieber},
      year={2025},
      eprint={2505.20249},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2505.20249}, 
}