The EuroLegalLLM (European Legal Large Language Model) initiative aims to revolutionize the European legal landscape through the creation of advanced bilingual and multilingual language models tailored for legal use.
The team has pioneered SaulLM, the premier large language model for Law, and this endeavor represents an expansion into a multilingual context, prioritizing four additional languages: German, French, Spanish, and Italian.
From a societal perspective, EuroLegalLLM endeavors to dismantle the dominance of closed English-centric and non-specialized models. The customization of models for five European jurisdictions and high-resource languages, including English, German, French, Spanish, and Italian, will empower legal practitioners.
Scientifically, EuroLegalLLM addresses the crucial challenge of adapting existing LLMs to excel in both multilingual and legal contexts, thereby reshaping the European legal landscape.EuroLegalLLM emphasizes transparency and openness across various facets, including data, training/adaptation methodologies, evaluation criteria, and model weights.
As such, the project stands as a scientifically grounded endeavor and a sturdy foundation for legal research and practice in linguistically diverse Europe. EuroLegalLLM seeks to fill this void and pave the way for advancements in multilingual legal AI research and application.
Pierre Colombo, CentraleSupelec - Université Paris Saclay - France