ICLR 2026 by the Numbers: Tsinghua Leads Overall, US Dominates Oral Papers

Overall paper count

ICLR 2026 recorded 5,000 papers from Chinese institutions, the largest total among all regions. The top five institutions worldwide are all Chinese: Tsinghua University (331 papers – 312 posters, 19 oral), Shanghai Jiao Tong University (240), Peking University (231), Zhejiang University (220) and the Chinese Academy of Sciences (206).

Oral paper distribution

When limited to oral papers, the balance shifts. U.S. institutions contributed 222 oral papers (40.5 % of all oral papers), China 158 (28.8 %), Europe 85 and other regions 84. Tsinghua remains the leading Chinese institution with 19 oral papers, ranking first globally in the oral category.

U.S. institutions in the oral leaderboard

Microsoft leads U.S. oral contributions with 15 papers, followed by Meta (14), MIT (13), Carnegie Mellon University (12), UC Berkeley (11) and Stanford (10). The oral papers from the United States are concentrated in a small set of industrial labs and top universities.

Chinese institutions in the oral leaderboard

Beyond Tsinghua, ByteDance secured 13 oral papers, Shanghai Jiao Tong University and Peking University each 11, Fudan University 9, the Chinese Academy of Sciences 8 and Hong Kong University of Science and Technology 7.

Enterprise contributions

Among Chinese enterprises, Alibaba leads with 137 papers (131 posters, 6 oral). Shanghai AI Lab contributed 130 papers, ByteDance 124, Huawei 114, Tencent 96, Ant Group 52, Kuaishou 40 and Meituan 29. Enterprises account for 1,246 papers (≈24.9 % of Chinese output), while universities and research institutes provide 3,754 papers (≈75.1 %).

Comparison with previous conferences

The shift toward Chinese dominance in total paper volume mirrors the pattern observed at NeurIPS 2025, where Chinese universities also moved to the top of the overall count. However, as at NeurIPS 2025, the oral tier remains led by U.S. institutions.

Interpretation

The data indicate that while Chinese universities now dominate the quantity of ICLR 2026 submissions, the high‑impact oral tier is still concentrated in U.S. labs and universities, reflecting the current distribution of compute, data and engineering resources. The next competitive frontier for Chinese research is increasing the depth and quality of oral contributions.

Code example

本文
约2000字
，建议阅读
5
分钟
清华 331 篇全球第一，Oral 数量同样领跑，中国高校集体冲上 ICLR 前排。