GPT-4 generated datasets

Leon-Leee 's Collections

Useful Pretrain-Datasets

Code, Math, and Reasoning related Instruct datasets

High Quality Instruct Dataset

Code Benchmarks

updated Apr 16

Collection of some GPT-4 generated datasets. It may be useful for those looking for the best-quality datasets to train competitive LLMs.

Upvote

openbmb/UltraInteract_sft

Viewer • Updated Apr 5 • 289k • 19.8k • 119

Note Should bring benefits for coding and reason since it's crafted by a slightly different prompt method and generated by GPT-4.
ise-uiuc/Magicoder-Evol-Instruct-110K

Viewer • Updated Dec 28, 2023 • 111k • 1.25k • 154

Note They did some cleaning; the original dataset is from https://huggingface.co/datasets/theblackcat102/evol-codealpaca-v1.
microsoft/orca-math-word-problems-200k

Viewer • Updated Mar 4 • 200k • 1.05k • 424

Note They used this dataset to build a strong math model.
Leon-Leee/Wizardlm_Evol_Instruct_v2_196K_backuped

Viewer • Updated Apr 16 • 143k • 45 • 1

Note A high-quality dataset by evol-instruct and GPT4.
CollectiveCognition/chats-data-2023-09-22

Viewer • Updated Sep 23, 2023 • 156 • 72 • 18

Note Not sure if all are by gpt-4.
jondurbin/airoboros-3.2

Viewer • Updated Jan 2 • 58.7k • 107 • 44

Note Not sure if all are by gpt-4.
camel-ai/biology

Viewer • Updated May 23, 2023 • 20k • 328 • 37

Note might be out-dated.
camel-ai/physics

Viewer • Updated May 23, 2023 • 20k • 366 • 68

Note might be out-dated.
Open-Orca/1million-gpt-4

Viewer • Updated Feb 7 • 995k • 4.06k • 41

Note No description is really bad. And they have been quiet for a while.
teknium/OpenHermes-2.5

Viewer • Updated Apr 15 • 1M • 2.37k • 692

Note A huge collection and caution: some of the dataset are not by gpt4.
openchat/openchat_sharegpt4_dataset

Updated Jul 1, 2023 • 185 • 167

Note Their models are really strong.
teknium/GPTeacher-General-Instruct

Viewer • Updated Apr 29, 2023 • 89.3k • 86 • 43

Note Not sure if out-dated.
Tony-Yuan/TheElements

Viewer • Updated Nov 27, 2023 • 1.56k • 44

Note Not sure if it is really good.
FreedomIntelligence/Evol-Instruct-Chinese-GPT4

Viewer • Updated Dec 6, 2023 • 70k • 61 • 45
shibing624/sharegpt_gpt4

Viewer • Updated Feb 23 • 103k • 460 • 114
MAsad789565/Coding_GPT4_Data

Viewer • Updated Mar 18 • 2.21k • 65 • 4
bz-arc13/evol_instruct_zh_gpt4

Viewer • Updated Mar 18 • 68.9k • 34
LNTANOooo/evol_instruct_zh_gpt4_v3

Viewer • Updated Mar 15 • 68.9k • 37 • 1

Upvote