LLM Datasets: Find data for LLM post-training

Understand the project

Practical overview

Use it when you need a starting directory for finding post-training datasets and want a quality checklist to guide candidate assessment.

LLM developers selecting data for supervised fine-tuning or other post-training work.
Researchers and practitioners comparing datasets across general instruction following and subject areas such as mathematics and science.

Browse curated dataset listings by task or subject area.
Follow listings to source datasets and related resources.
Assess candidate data using accuracy, diversity and complexity as quality dimensions.
Consider review, filtering and model-based scoring approaches when evaluating dataset quality.

Before you choose

The repository is a directory of listings; dataset access is through linked resources, not established as repository-hosted data.
Dataset terms may differ: the README notes permissive licenses generally unless an entry specifies otherwise, so check each linked dataset’s terms.
The quality guidance emphasizes accuracy, diversity and complexity; the appropriate balance depends on the model task and intended coverage.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Compare candidate datasets against accuracy, diversity, and complexity criteria before choosing one for post-training.

Browse the list’s dataset categories and notes to identify candidates relevant to the intended use.
Assess whether samples are factually correct and relevant; use solvers for math or unit tests for code where applicable.
Consider whether the data covers varied use cases and includes suitable multi-turn, multilingual, well-written, or reasoning-oriented examples.
Combine manual review with rule-based filtering and scoring by judge LLMs or reward models to assess dataset quality.

A candidate dataset or set of candidates assessed against the listed quality characteristics and review techniques.

The list is a directory of datasets and tools; it does not establish that it hosts the datasets or provides a programmatic API.
Dataset licensing varies; check the individual dataset’s terms rather than assuming a permissive license.
The repository does not document a standardized evaluation procedure or threshold for accepting a dataset.

Evidence and transparency

Source record details

Source availability: Active
Sources checked: 15 October 2026
Original website title: mlabonne/llm-datasets · Official homepage title · retained in its source language