README.md

metadata

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}
}