Skip to content

Dataset · Open-Orca

Cleaned instructions with GPT-4 answers

Open-Orca/SlimOrca

About 500k GPT-4 answers filtered by FLAN annotations for training instruction models.

What's inside

An English-language text dataset of instructions and answers, being a cleaned sub-sample of OpenOrca. It contains about 500k GPT-4 answers; an additional GPT-4 pass removed answers deemed incorrect based on human FLAN annotations. Tasks include question answering, summarization, classification, and others.

What you can do with it

  • SFT of instruction LLMs
  • LoRA fine-tuning for QA and summarization
  • Comparison with the full OpenOrca
Tasks
Text classification, Token classification, Table question answering, Question answering, Zero shot classification, Summarization
Languages
English
Size
100K–1M rows
Downloads this month
3.1K
Likes
311

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Open-Orca/SlimOrca")