China AI: Researchers Publish Novel Image and Video Tool-Use Model

In a recent development within China's AI research landscape, a paper titled 'AdaTooler-V: Adaptive Tool-Use for Images and Videos' was published on arXiv on December 18, 2025. The research, co-authored by individuals with Chinese names, introduces a novel approach to multimodal large language models (MLLMs) that enhances their ability to interact with vision tools. This work aims to improve MLLMs' reasoning and performance by integrating interleaved chain-of-thought processes with visual tool utilization.

While major Chinese tech companies like Tencent, Alibaba, Baidu, and Huawei have not announced significant AI breakthroughs in the past five days, this academic publication signifies ongoing progress in fundamental AI research within China. The development of advanced MLLMs capable of complex tool interaction is crucial for China's broader AI ambitions. Such research contributes to the nation's efforts to compete in the global AI race, pushing the boundaries of what AI systems can achieve and potentially narrowing the technological gap with Western counterparts. The paper's focus on adaptive tool-use suggests a move towards more sophisticated and versatile AI agents.

References

1. https://arxiv.org/abs/2512.16918v1