OcciGlot - New Open Source Language Models for Europe released 🇪🇺

Researchers from DFKI and hessian.AI have launched the OcciGlot initiative to develop generative open source language models for European languages.

👉🏼 dfki.de/en/web/news/oc…

📢Máme málo neanglických LLM benchmarků! Pojďte se zapojit do Hugging Face 🤗komunitního projektu a pojďmě společně přeložit 500 anglických promptů do čestiny pro následné automatické evaluace.

Více na huggingface.co/spaces/DIBT-Cz….

Common corpus is an international initiative coordinated by @pleias_fr with the support of the state start-up LANGU:IA, supported by French Ministry of Culture and Direction interministérielle du numérique Also involving the open science LLM (Occiglot, Eleuther AI) and cultural heritage communities (@storytracer)

500B public domain dataset released by pleias today. And of course it’s multilingual.

We’re very excited about our ongoing collaboration. More cool things to come 🚀

Interesting comparison of 'German' LLMs on Reddit - finds that DiscoResearch DiscoLM German 7b outputs best 'native-sounding' quality despite being behind in benchmarks - followed by #Occiglot DE EN Instruct which also uses our data 🙂🙌 1/2

We just made a large-scale evaluation sweep of tokenizer performance across European languages.
We’re sharing it publicly as part of our commitment to transparent research, and hoping it might be helpful for others.

occiglot.github.io/occiglot/posts…

Meet Occiglot: A Large-Scale Research Collective for Open-Source Development of Large Language Models by and for Europe

marktechpost.com/2024/03/08/mee…

#ArtificialIntelligence #DataScience #LLMs OcciGlot

Checkout #Occiglot a large-scale research collective for open-source development of LLMs: huggingface.co/occiglot

A great open model initative by many of our friends hessian.AI , DFKI, TU Darmstadt and we are happy to contribute a (small) dataset for instruction tuning 🙂.