US China AI Collaboration NeurIPS 2025: Data Shows Deep Academic Ties

They're shaking hands while keeping their fists clenched. While Washington and Beijing trade blows over AI supremacy, the research labs of both nations remain surprisingly enmeshed. A closer look at recent academic output suggests that the scientific community isn't as divided as the political headlines imply.

Mapping US China AI Collaboration NeurIPS Trends

According to an analysis by WIRED, the NeurIPS 2025 conference—held in December 2025—showcased a significant level of cross-border partnership. Out of 5,290 total papers, 141 (roughly 3%) involved co-authors from both US and Chinese institutions. This rate has remained steady, with 134 collaborative papers out of 4,497 recorded in 2024.

Algorithms Crossing the Pacific

The sharing of models and architectures is even more pervasive. The Transformer architecture, originally a Google breakthrough, appeared in 292 papers from Chinese institutions. Meta'sLlama models were featured in 106 of those studies. Conversely, Alibaba'sQwen model was a key element in 63 papers involving American researchers.

Jeffrey Ding, an assistant professor at George Washington University, noted that the two ecosystems are "inextricably enmeshed." Regardless of political pressure, researchers continue to leverage the best tools available, regardless of their origin.