Tag
The article introduces the dots3-note Preview model, an open-weight multimodal AI model optimized for long-horizon tasks with TEMPO, a reinforcement learning technique that enables self-critique and adaptation.