Scholé: Using Federated Reinforcement Learning with Human Feedback for Personalized Learning at Scale

• Unité de recherche: INNOSUISSE
• Numéro de projet: 125.499 IP-ICT
• Titre du projet anglais: Scholé: Using Federated Reinforcement Learning with Human Feedback for Personalized Learning at Scale

Mots-clé (Anglais)

ICT, Artificial/local/extended/networked intelligence

Résumé des résultats (Abstract) (Anglais)

Scholé is a personalized learning framework that learns as it teaches. Our AI for education research features a combination of federated learning, reinforcement learning with human feedback, and memory updates in large language models that has never been attempted before in a real-world system.