Open datasets

Open Data Directory

Datasets for training, evaluation, and retrieval-augmented generation, with health and fallback visibility.

Loading data health...

AI Data Stack: training → embeddings → LoRA → RAG

Use these shortcuts to jump into the most common open datasets and resources for building local models and grounded assistants.

Large corpora and knowledge bases used to train foundation models. Always verify license and dataset terms for your use case.

Common Crawl FineWeb Dolma Wikimedia Dumps

High-signal instruction datasets and practical fine-tuning references (LoRA / QLoRA / PEFT).

OpenAssistant FLAN LoRA QLoRA

Benchmarks and datasets used to measure retrieval quality (and to train/evaluate embedding models).

BEIR MTEB MS MARCO Embedding Benchmarks

Open corpora to ingest for RAG and references for building + evaluating grounded QA workflows.

arXiv Stack Exchange RAG Learning Hub