A research collaboration between Shanghai Jiao Tong University, Shanghai Qi Zhi Institution, and Huawei Technologies has introduced “BAFT”, a cutting-edge auto-save system for AI training that minimizes downtime and optimizes efficiency. Designed to leverage idle moments in training workflows, BAFT significantly enhances fault tolerance while reducing computational overhead, setting a new industry benchmark for reliable AI model development.

Researchers from the National University of Defense Technology (China) have developed an innovative processor that enables seamless execution of legacy software on next-generation hardware without costly redevelopment. This breakthrough technology integrates hardware-based translation, allowing software originally designed for ARM-based chips to run efficiently on the open-source RISC-V architecture while maintaining up to 65% of native performance. By eliminating software compatibility barriers, this solution offers a cost-effective and power-efficient path for industries relying on embedded systems, including smart devices, medical sensors, and industrial automation.