Dataset · alpaca-cleaned

Alpaca-Cleaned

Builder: gururise / yahma (community)
Availability: available · checked 2026-09-24 · source

Raw record: /data/datasets/alpaca-cleaned.json

Fields

id

alpaca-cleaned

identifiers

huggingface

yahma/alpaca-cleaned

builder

gururise / yahma (community)

release_date

2023-03-24partial · source

note:  HF dataset repo creation date, not the builder's own release announcement.

availability

available · checked 2026-09-24 · source

content

Card: 'a cleaned version of the original Alpaca Dataset released by Stanford', fixing hallucinations, empty outputs and similar issues. Derivation from `alpaca-52k` is recorded in prose (no dataset-to-dataset edge). The card refers to the original outputs as GPT3's.recorded · source

primary_sources

https://huggingface.co/datasets/yahma/alpaca-cleaned

https://github.com/gururise/AlpacaDataCleaned

record_history

date:2026-09-24 · change:created from primary sources (tranche 2 dataset records) · by:wilson-pruitt + claude ·

Parents

Influence without weights

• distilled_from_outputs → text-davinci-003 declared source Inherited through Alpaca: Stanford CRFM generated the original 52K with text-davinci-003; the cleaner's card describes this set as a cleaned version of it. Cleaning changed some outputs; the fraction is not_recorded.

Children

Training data

• ← trained_on alpaca-lora-7b declared source Card metadata: datasets: [yahma/alpaca-cleaned]; README training command --data_path 'yahma/alpaca-cleaned'. Uploader tloen is the developer. Card prose instead says 'Stanford Alpaca dataset'; see training_data note. Reviewer: if you prefer the prose reading, swap to alpaca-52k.

Read in

No station on the reading path has touched this record yet.