https://github.com/orhonovich/unnatural-instructions

https://github.com/orhonovich/unnatural-instructions

Last synced: 24 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/orhonovich/unnatural-instructions
• Owner: orhonovich
• License: mit
• Created: 2022-12-19T14:01:06.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-02-23T06:26:18.000Z (over 1 year ago)
• Last Synced: 2024-01-14T05:59:31.773Z (6 months ago)
• Size: 27.6 MB
• Stars: 162
• Watchers: 7
• Forks: 9
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Lists

• awesome-chatgpt-dataset - Unnatural Instructions - ative and diverse instructions, collected with virtually no human labor. | MIT | (Dataset Detail)
• awesome-prompt-engineering - Unnatural Instruction
• Awesome-instruction-tuning - Unnatural Inst. - LM-Unnat. Inst. | T5-LM | 11B | (Datasets and Models / Modified from Traditional NLP)
• awesome-stars - orhonovich/unnatural-instructions - (Others)
• awesome-rlhf - unnatural-instructions
• awesome-instruction-dataset - (orhonovich/unnatural-instructions)|240K|EN|MT|MIX
• awesome-instruction-datasets - Dataset Link
• Awesome-LLM - https://github.com/orhonovich/unnatural-instructions

README

        
# Unnatural Instructions

This repository contains the Unnatural Instructions dataset. Unnatural Instructions is a dataset of instructions automatically generated by a Large Language model.

See full details in the paper: "[Unnatural Instructions: Tuning Language Models with (Almost) No Human Labor](https://arxiv.org/abs/2212.09689)"

## 🗃️ Content

The `data` folder contains two files: `core_data.jsonl`, containing the Unnatural Instructions core dataset of 68,478 instruction-input-output triplets, and `full_data.jsonl`, containing the full 240,670 Unnatural Instructions examples. The full data was constructed by expanding the core data with automatically generated instruction paraphrases.

## 📄 Format

### Core data

Each line in `core_data.jsonl` is a JSON object with two fields - `instruction`, which is a natural language instruction describing a task, and `instances`, an array of JSON objects, each contains

- `input`: An input for the task described by the `instruction`

- `instruction_with_input`: The instruction concatenated with the `input`

- `constraints`: The task's output space constraints

- `output`: The output of executing `instruction` with the given `input`

### Full data

`core_data.jsonl` has the same structure as `core_data.jsonl`, but with one additional field - `reformulations`. `reformulations` is an array of JSON objects, each corresponds to an automatically generated paraphrase for the given instruction. Each reformulation contains the fields:

- `instruction`: A paraphrase of the original instruction

- `input`: An input for the task described by the `instruction`

- `instruction_with_input`: The paraphrased instruction concatenated with the `input`

- `output`: The output of executing `instruction` with the given `input`

## 📘 Citation

If you make use of Unnatural Instructions, please cite the following paper:

```

@misc{honovich2022unnatural,

      title = {Unnatural Instructions: Tuning Language Models with (Almost) No Human Labor},

      author = {Honovich, Or and Scialom, Thomas and Levy, Omer and Schick, Timo},

      url = {https://arxiv.org/abs/2212.09689},

      publisher = {arXiv},

      year={2022}

}

```