Recently, the list of winners for the NeurIPS 2025 Best Paper Awards was officially disclosed, with four outstanding papers taking home the accolades. A significant number of these winning papers were spearheaded by Chinese scholars. Their research spanned a diverse range of topics, including language model diversity, attention mechanisms within large-scale models, deep reinforcement learning, and diffusion model theory.
Remarkably, the Alibaba Tongyi Qianwen team stood out by clinching the Best Paper Award for their groundbreaking research on attention mechanisms in large models. This achievement marks them as the sole Chinese team to be bestowed with this prestigious honor. The fruits of their labor have already been integrated into the Qwen3-Next model, resulting in a substantial boost to its performance and robustness.
