Two powers vie for supremacy, while Chinese universities rapidly rise. Author | Zhang Chi
Editor | Ma Xiaoning
CoRL 2026 will be held in Austin, Texas, USA, from November 9–12. As a premier international academic conference at the intersection of robotics and machine learning, CoRL has always been the best place to learn about the latest research in robotics. Leiphone (WeChat official account: Leiphone) will also be on site in Austin to bring readers first-hand reports. Welcome to scan the QR code at the end of the article to join our community and exchange ideas with us.
Unlike other conferences, which accept thousands of papers, CoRL has historically accepted relatively few papers, with around 200 each year. However, this year's number of accepted papers has grown dramatically over last year's, reaching 687.
So, based on public data, we compiled statistics on the institutions, companies, and countries/regions represented among the papers accepted in recent years, to see how the structure has changed across the six editions of the conference from 2021 to 2026.
01
Data Description
The data for 2021–2025 is based on the open-source Paper Copilot project. The data comes from OpenReview, official conference channels, and the PMLR paper collection, among other sources, and information such as authors, institutions, and countries is relatively complete. Quite a few domestic universities have already used this project, and it appears to be fairly reliable.
The 2026 data comes from the Conference Explorers project, whose author is an assistant professor at DGIST in South Korea. Official conference agenda snapshots and paper information on OpenReview were referenced during data processing. However, some of the data is incomplete: of the 687 papers, only 480 have standardized institutional information. Since no other, more complete 2026 data could be found, and given the limited technical capability in web crawling, and also because the largely complete information is already sufficient to reflect the overall trend, we ultimately decided to use this project's data for the analysis.
02
Analysis Results
CoRL's paper output grew in a fairly steady, near-linear fashion from 2021 to 2025 (153 → 263 papers, +14.5% per year on average), and then jumped straight to 687 papers in 2026, indirectly showing how hot the robotics research field has been in the past two years.
For papers by country, the statistics are based on "whether a paper lists an institution from that country" (a paper is counted as long as it contains at least one institution from that country, and a single paper may be counted for multiple countries).
The United States fell from 76% in 2021 to 51% in 2026, while China rose from 5.9% to 37.5%, first exceeding 20% in 2024. China's academic community has an increasingly strong voice at top-tier conferences.
Note that the statistics only count participation, not the degree of contribution; that is, there is no distinction between first and second author.
For the two academic heavyweights, China and the United States, the paper counts and their shares (with overlaps) are as follows:
Looking more specifically at the institutions the paper authors are affiliated with, this year's highlight is that Tsinghua's paper count surpassed CMU and Stanford for the first time.
Among the top five in paper count in 2026, three are Chinese universities. Tsinghua took the lead with 58 papers (20 in 2025), followed by Shanghai Jiao Tong University with 36 papers and Peking University with 33 papers.
Meanwhile, a group of institutions that only appeared at CoRL for the first time in 2026 jumped straight into the leading ranks, such as the Shanghai Innovation Institute with 21 papers, A*STAR with 6 papers, and Hessian.AI with 6 papers.
One caveat, however: more than 200 papers in the 2026 data do not specify an affiliated institution, so the figures are not fully accurate. This should not, however, affect the overall distribution.
For the specific growth patterns, see the share-multiplier table below. Share multiplier = 2026 share / 2025 share.
Only institutions with at least 5 papers in 2025 are included (34 institutions in total), because multipliers based on samples that are too small are unstable. Institutions newly appearing in 2026 that have no 2025 baseline are also excluded.
When expressed as share multipliers, UCLA has outperformed the overall market by the greatest margin (it had 5 papers in 2025, so its base was not large either, but it has 16 in 2026), although it is now surrounded by Chinese universities. It is followed by Tsinghua University, the University of Hong Kong, Shanghai Jiao Tong University, Zhejiang University, and Peking University.
The charts above can make it seem that most paper contributions come from leading universities, but the share of papers taken by the top 10 institutions fell from 49.8% in 2021 to 26.2% in 2026, while the total number of institutions listed in paper affiliations rose from 92 to 350.
This is typical long-tail growth, indicating that research across the entire field of robotics is accelerating.
Over the past two years, research in China’s robotics field has expanded markedly, but looking at the six-year totals, U.S. institutions still lead in overall research output, and research conducted by U.S. companies also accounts for a substantial share.
Below are the top 25 institutions ranked by cumulative paper count for 2021–2026. Because coverage of institutional information is incomplete for 2026, all the figures represent only lower bounds of the actual data.
Combining the paper shares of institutions across the years makes changes in their research strength more evident.
Below is a heatmap showing the distribution of the top five institutions by paper share at each CoRL edition from 2021 to 2026. Across the six editions, only nine institutions have ranked in the top five by share.
The line chart below shows the turnover process more clearly (also based on rankings by share).
As can be seen from the chart, Stanford and CMU have remained consistently strong, while Chinese universities have caught up and moved ahead with tremendous momentum.
Finally, it should be noted that the above data are not necessarily fully accurate.
First, the 2026 data lack sufficient completeness and will need to be verified once better data sources become available.
Second, some information may come from author profile pages, where current, past, or visiting institutional affiliations may change, so it may not necessarily reflect the situation at the time the paper was published.
In addition, differences in how data sources identify an institution versus its parent institution, or different campuses of the same university, can also affect the results.
Data sources:
Paper Copilot: https://github.com/papercopilot/paperlists/tree/main/corl
Conference Explorers: https://github.com/gisbi-kim/conference-explorers
