When AI models reason about images, small perceptual errors compound across multiple steps and produce wrong answers. Alibaba’s HopChain framework tackles this by generating multi-stage image questions that break complex problems into linked individual steps, forcing models to verify each visual detail before drawing conclusions. The approach improves 20 out of 24 benchmarks.
Reinforcement learning hits a wall with reasoning models because every token gets the same reward. A new algorithm from Alibaba’s Qwen team fixes this by weighting each step based on how much it shapes what comes next, doubling the length of thought processes in the process.
Microsoft is investing $10 billion in Japan from 2026 to 2029, its largest ever commitment to the country. Vice Chairman and President Brad Smith presented the plans during a visit to Tokyo. The funding covers AI infrastructure, cybersecurity, and workforce training.
Together with SoftBank and Sakura Internet, Microsoft plans to offer GPU-based AI services through Azure, with all data staying in Japan. This should make it possible to develop Japanese language models. On the security side, Microsoft is deepening its partnership with Japan's cybersecurity agency and the National Police to better detect and prevent cyberattacks.
Japan is projected to be short 3.26 million AI and robotics specialists by 2040, according to the Ministry of Economy. To address that, Microsoft is teaming up with Fujitsu, Hitachi, NEC, NTT Data, and SoftBank to train one million engineers and developers by 2030. Microsoft had already invested $2.9 billion in Japan in 2024.
Chinese AI startup Zhipu AI has released GLM-5V-Turbo, a multimodal model that processes images, video, and text and is designed for use in agent workflows.
