This is the training dataset of MUFFIN (Multi-Faceted Instructions).

Please refer to our project website for more details: Website

JSON Format

The download data can be read as a Python list.

In this list, each elemental Python dictionary has one input text.

This input text has multiple task instructions and the corresponding outputs.

[
  {
    "input": "XXX",
    "instances": [
      { "instruction": "III", "output": "YYY" },
      { "instruction": "III", "output": "YYY" }
    ]
  }
  ,
  {
    "input": "XXX",
    "instances": [
      { "instruction": "III", "output": "YYY" }
    ]
  }
]

Data Statistics

There are a total of 1,463 input texts, where each input is equipped with multiple task instructions (~46.48 instructions per input), resulting in 68,014 training instances in total.

The detailed statistics are shown below:

🥳 Citation

Please kindly cite our paper if you use our dataset:

@inproceedings{Lou2023MUFFIN,
   title={{MUFFIN}: Curating Multi-Faceted Instructions for Improving Instruction Following},
   author={Renze Lou and Kai Zhang and Jian Xie and Yuxuan Sun and Janice Ahn and Hanzi Xu and Yu su and Wenpeng Yin},
   booktitle={The Twelfth International Conference on Learning Representations},
   year={2024},
   url={https://openreview.net/forum?id=1vrS1zwekw}
}