Tsinghua Tops ICLR 2026; Chinese Universities Claim Top Five, US Holds Strong Oral Share
The 2026 ICLR conference saw Chinese institutions contribute 5,000 papers—dominating the top five slots—with Tsinghua leading at 331, while U.S. labs secured 222 of the 4% oral papers, highlighting a robust American presence despite China's overall volume advantage.
ICLR 2026 Paper Statistics
Chinese institutions contributed 5,000 papers (4,842 posters, 158 oral), accounting for roughly 40% of all poster papers and 30% of oral papers worldwide.
U.S. institutions contributed 3,785 papers; oral papers made up about 40% of their output and posters about 30%.
European institutions contributed 1,594 papers, and institutions from other regions contributed 2,192 papers.
Top Institutions by Total Papers
Tsinghua University leads with 331 papers (312 posters, 19 oral). It is followed by Shanghai Jiao Tong University, Peking University, Zhejiang University, and the Chinese Academy of Sciences – all Chinese institutions.
Oral Paper Distribution
Oral papers represent 4% of all accepted papers. U.S. institutions secured 222 oral papers (40.5% of the oral pool). Chinese institutions contributed 158 oral papers (28.8%). Europe and other regions contributed 85 and 84 oral papers respectively.
Among U.S. labs, Microsoft obtained 15 oral papers, Meta 14, MIT 13, Carnegie Mellon University 12, UC Berkeley 11, and Stanford 10.
Methodology
The ICLR 2026 figures use a stricter direct paper‑authorship attribution method, unlike the 2025 NeurIPS analysis that relied on author registration data from the OpenReview platform.
Chinese Institutional Composition
Universities and research institutes produced 3,754 papers (≈75.1% of Chinese output); companies produced 1,246 papers (≈24.9%).
Alibaba led Chinese companies with 137 papers (131 posters, 6 oral), Shanghai AI Lab followed with 130 papers, and ByteDance published 124 papers, including 13 oral papers.
Resource Concentration Trends
High‑impact AI research increasingly depends on massive compute, large datasets, dedicated engineering teams, and long‑term experimentation, concentrating resources in top universities, major corporate labs, and national platforms.
Global Landscape Beyond China and the U.S.
Active institutions also appear in Singapore, South Korea, Canada, Australia, Japan, and the United Arab Emirates. Europe’s core contributions come mainly from the United Kingdom and Switzerland; Singapore and South Korea’s output is comparable to the entire EU.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
SuanNi
A community for AI developers that aggregates large-model development services, models, and compute power.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
