The project's latest updates focus on improvements in AI models, privacy support for distributed computing, and the launch of new systems.
Large Model Initialization
The project team has made significant progress in optimizing the Morpheus model for AI agents, achieving a 19% progress. Reinforcement learning and adversarial data are included to enhance performance. New data processing and cleaning mechanisms are also being developed.
Support for Distributed Computing
The MANTA computing platform's capabilities are being expanded to ensure privacy during training and inference. Optimization for MOE model training in the DeepSeek V3 project is yet to begin.
Agent Model Center
Initial development of the Agent Model Center has started, integrating baseline models from MATRIX. Users can develop their own agents based on the provided models. DeepSeek V3 is integrated as one of the baseline models.
The project is making strides in optimizing and developing new AI technologies and distributed computing systems.
