---
language:
- en
license: apache-2.0
task_categories:
- text-classification
task_ids:
- natural-language-inference
dataset_info:
  config_name: atomic
  features:
  - name: Premise
    dtype: string
  - name: Hypothesis
    dtype: string
  - name: Update
    dtype: string
  - name: UpdateType
    dtype: string
  splits:
  - name: train
    num_bytes: 4463463
    num_examples: 35002
  - name: validation
    num_bytes: 487230
    num_examples: 3840
  - name: test
    num_bytes: 526381
    num_examples: 4138
  download_size: 1871664
  dataset_size: 5477074
configs:
- config_name: atomic
  data_files:
  - split: train
    path: atomic/train-*
  - split: validation
    path: atomic/validation-*
  - split: test
    path: atomic/test-*
---
https://github.com/rudinger/defeasible-nli
```
@inproceedings{rudinger2020thinking,
  title={Thinking like a skeptic: 
  feasible inference in natural language},
  author={Rudinger, Rachel and Shwartz, Vered and Hwang, Jena D and Bhagavatula, Chandra and Forbes, Maxwell and Le Bras, Ronan and Smith, Noah A and Choi, Yejin},
  booktitle={Findings of the Association for Computational Linguistics: EMNLP 2020},
  pages={4661--4675},
  year={2020}
}
```