Researchers at Google Cloud and UCLA have proposed a new reinforcement learning framework that significantly improves the ability of language models to learn very challenging multi-step reasoning tasks. Supervised Reinforcement Learning (SRL) reformulates problem-solving as a sequence of logical “actions,” providing rich learning signals during the training process. This approach enables smaller models to learn complex problems that were previously out of reach for other common training techniques.

Researchers at Google Cloud and UCLA have proposed a new reinforcement learning framework that significantly improves the ability of language models to learn very challenging multi-step reasoning tasks. Supervised Reinforcement Learning (SRL) reformulates problem-solving as a sequence of logical “actions,” providing rich learning signals during the training process. This approach enables smaller models to learn complex problems that were previously out of reach for other common training techniques.

Face aux tensions entre les États-Unis et l'Union européenne et à la crainte d'une dépendance technologique américaine, Nextcloud a vu ses demandes de clients tripler au premier semestre. Son PDG nous explique comment cette plateforme open source, désormais adoptée par des dizaines de millions d'utilisateurs en Europe, symbolise aujourd'hui la souveraineté numérique face aux géants du cloud américain.