Sign in to view source links and access this dataset
Description
51,760 entries of instruction-following conversation data, originally created by Stanford University for training language models. This version has been cleaned, fixed, and converted into the ShareGPT format. The dataset was last updated on HuggingFace by SicariusSicariiStuff in November 2025.
Use Cases
Fine-tune language models for instruction-following based on the cleaned conversation examples.
Benchmark conversational AI performance using a standardized ShareGPT format.
Train models to generate human-like responses based on structured prompts.
Study the effectiveness of data cleaning and format conversion on model training outcomes.
Strengths
Contains exactly 51,760 entries, providing a substantial volume of training examples.
Derived from the original Stanford Alpaca dataset, a well-known benchmark in the field.
Has undergone a cleaning and format conversion process to the ShareGPT standard.
Limitations
Description metadata is limited; actual data quality requires manual inspection after download.
Column-level documentation is absent; field semantics must be inferred after download.
The original source's potential biases may be inherited by this cleaned version.
Provenance
Source
Stanford University (original Alpaca dataset)
Collection Method
Cleaned, fixed, and converted from the original Alpaca format to ShareGPT format.
Freshness
Last updated 2025-11-04 05:43:42; freshness should be verified.
License is unknown; users must verify licensing terms before use.