Overview
Velvet is an Italian family of Large Language Models developed from scratch and trained on the HPC Leonardo infrastructure hosted by CINECA. This model was built utilizing public data that underwent extensive curation.
Role
- Ethics assessment and AI Alignment: Responsible for evaluating ethical considerations and ensuring the model aligns with human values and safety standards
- Technical Team Member: Contributed to the technical development and implementation
Key Features
- Trained specifically for the Italian language
- Developed using high-performance computing resources
- Emphasis on ethical AI development
- Extensive data curation process
- Built from scratch with careful attention to alignment
Project Details
Year: 2025
Infrastructure: HPC Leonardo (CINECA)
Focus: Italian language processing, AI Safety, Alignment