Mozilla and EleutherAI publish research on open datasets for LLM training
In June 2024, Mozilla and EleutherAI hosted the 'Dataset Convening' in Amsterdam, which brought together 30 experts from open source AI startups, nonprofit AI labs, and civil society organizations. The focus was on developing openly licensed and accessible datasets for training LLMs. The resulting research paper, “Towards Best Practices for Open Datasets for LLM Training' (released in January 2025), outlines possible tiers of openness, normative principles, and technical best practices for sourcing, processing, governing, and releasing open datasets for LLM training, as well as opportunities for policy and technical investments to help the emerging community overcome its challenges.
Read more about the the convening here.
Check out the 'Towards Best Practices for Open Datasets for LLM Training' paper here.
Related Blueprints
Document
Markdown
Convert documents to markdown format
This blueprint guides you to convert unstructured documents to Markdown format using the Docling command-line interface.
EleutherAI
Audio File
Transcript
Transcribe Audio Files with Open-Source Whisper Models
This Blueprint shows you how to transcribe audio files using open-source Whisper models via Speaches, a self-hosted server that mimics the OpenAI Whisper API.
EleutherAI
Load more Blueprints
