# yizhongw/self-instruct

Aligning pretrained language models with instruction data generated by themselves.

Repository: https://github.com/yizhongw/self-instruct
Canonical: https://ross.abutalabs.com/products/self-instruct
Language: Python
License: Apache-2.0
License Family: permissive
Topics: general-purpose-model, language-model, instruction-tuning
Last push: 2023-03-27T18:18:51+00:00

## Health v2 (maintenance only)
Score: 31/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 96
- inputs: {"age_days": 1352, "days_push": 1255, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4609, forks 520 (observed 2026-08-28T04:08:54.977090+00:00)

## What it is
Self-Instruct is a framework and research codebase for aligning pretrained language models with instructions using data generated by the models themselves. It implements an iterative bootstrapping pipeline that generates, filters, and curates instruction data, and releases a 52K-instruction dataset for instruction-tuning.

## Use cases
- generate synthetic instruction data for fine-tuning language models
- improve a model's instruction-following without manual annotation
- bootstrap a prompt dataset from a small seed set of tasks
- instruction-tune GPT-3 on model-generated data
- research bootstrapping methods for LLM alignment

## When to choose
- you need large-scale instruction-tuning data without human annotation
- you are researching self-generated or synthetic training data for LLMs
- you want to replicate the Self-Instruct paper pipeline

## When to avoid
- you need production-grade, actively maintained tooling
- you want to fine-tune open models with modern tooling rather than GPT-3 finetuning scripts
- you cannot tolerate noisy or biased synthetic data

## Facets
- artifact type: library
- maturity: maintenance
- function: llm-training, data-generation, prompt-engineering, machine-learning
- domain: large-language-models, artificial-intelligence, machine-learning
- platform: python, cli
- tags: instruction-tuning, synthetic-data, llm-alignment, research-code, gpt3, natural-language-processing, linux

## Member repositories
- yizhongw/self-instruct (main) score 31

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:54.977090+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:19:43.969167+00:00, confidence not recorded.
  - readme: https://github.com/yizhongw/self-instruct (fetched 2026-08-28T04:08:54.977090+00:00, sha beafa25e7be8)
- Data as of 2026-08-30T08:39:29.467469+00:00.