# google-research/FLAN

Repository: https://github.com/google-research/FLAN
Canonical: https://ross.abutalabs.com/products/google-research-flan
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-07-02T11:25:30+00:00

## Health v2 (maintenance only)
Score: 73/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 90, release rhythm 35, longevity 100
- inputs: {"age_days": 1838, "days_push": 62, "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 1567, forks 160 (observed 2026-08-28T04:05:04.960571+00:00)

## What it is
Google Research's repository for generating the FLAN instruction tuning dataset collections, including the original Flan 2021 and the expanded Flan Collection used to train Flan-T5 and Flan-PaLM. It provides Python/SeqIO code to produce instruction tuning mixtures from many NLP tasks.

## Use cases
- generate instruction tuning data for fine-tuning LLMs
- reproduce the Flan 2021 SeqIO mixtures
- build the Flan Collection for training models like Flan-T5
- research zero-shot generalization of finetuned language models
- mix diverse NLP task datasets for prompt-based training

## When to choose
- you need the canonical FLAN instruction tuning datasets for LLM training
- you use SeqIO/T5X and want reproducible Flan mixtures
- you are researching instruction tuning or zero-shot prompting

## When to avoid
- you want a ready-to-download dataset rather than generation code
- you need a general-purpose ML framework unrelated to instruction tuning
- you want inference or serving tools for Flan models

## Facets
- artifact type: dataset
- maturity: stable
- function: machine-learning, llm-training, data-generation, etl
- domain: large-language-models, machine-learning, artificial-intelligence
- platform: python
- tags: instruction-tuning, flan, seqio, dataset-generation, research, google-research, zero-shot-learning, natural-language-processing

## Member repositories
- google-research/FLAN (main) score 73

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:04.960571+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 2026-08-30T03:59:00.395141+00:00, confidence not recorded.
  - readme: https://github.com/google-research/FLAN (fetched 2026-08-28T04:05:04.960571+00:00, sha 38a12f7b3378)
- Data as of 2026-08-30T08:39:29.467469+00:00.