---
dataset_info:
  features:
  - name: idx
    dtype: int64
  - name: context
    dtype: string
  - name: marker
    dtype: string
  - name: option_0
    dtype: string
  - name: option_1
    dtype: string
  - name: option_2
    dtype: string
  - name: option_3
    dtype: string
  - name: label
    dtype: int64
  splits:
  - name: train
    num_bytes: 4950296
    num_examples: 9299
  - name: test
    num_bytes: 2038685
    num_examples: 3757
  download_size: 4621420
  dataset_size: 6988981
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
license: apache-2.0
source_datasets:
- prajjwal1/discosense
---

# discosense

DiscoSense: four-choice commonsense reasoning with discourse connectives.

Original distribution: https://github.com/prajjwal1/discosense.
Source paper: https://arxiv.org/abs/2210.12478.
All native fields, option order, integer gold indices and train/test partitions
are preserved verbatim. No validation partition or new distractors are created.
Use both `context` and `marker` as input; `label` indexes `option_0`–`option_3`.
The pinned source Hub README declares Apache-2.0; see provenance for evidence.
Rebuild with `python scripts/upload_repackaged.py discosense`.

Original data: [prajjwal1/discosense](https://huggingface.co/datasets/prajjwal1/discosense), [original distribution](https://github.com/prajjwal1/discosense). Repackaged as parquet for [tasksource](https://github.com/sileod/tasksource) by [scripts/repackage_dataset/text.py](https://github.com/sileod/tasksource/blob/main/scripts/repackage_dataset/text.py#L39).

[Conversion provenance](provenance.json): source revisions, hashes, configuration and exclusions.