18907305772 / Explore-Instruct

EMNLP'2023: Explore-Instruct: Enhancing Domain-Specific Instruction Coverage through Active Exploration
https://arxiv.org/abs/2310.09168
Apache License 2.0
4 stars 3 forks source link

Brainstorm dataset generation #2

Open wangchunlin opened 1 month ago

wangchunlin commented 1 month ago

Hi, Thanks for your contribution to the sysnthetic data generation. I have investigated your released code. I have a question about the raw task definition for different dataset. Whether the template of raw task is the same for different dataset such as 'brainstorm' and 'rewrite'? Look forward to your response. Thanks.

Best regards Chunlin

18907305772 commented 1 month ago

Hello, the raw task in different datasets are different. However, the template is the same as below:

{
    "raw_task_name": "...",
    "task_name": "...",
    "config_filename": "...",
    "reason": "...",
    "children": [...]
}