Researchers Assess Contributions of BRICS Countries to Leading ML/AI Conferences

© iStock

The HSE Scientometrics Centre analysed more than 104,000 papers published between 2020 and 2025 and presented at ten top-level A* conferences ranked by ICORE 2026. The researchers examined the contributions of Brazil, Russia, India, China and South Africa. Together, these countries accounted for around 40,800 publications, or 39.2% of the total. However, the distribution is highly uneven, and a shared BRICS research space has yet to emerge.

Russia ranks third among the BRICS countries for the number of such publications. It also stands out for the high involvement of technology companies and the diversity of participating actors, including companies, universities and non-profit organisations. Yandex leads among companies by the number of publications, while HSE University and Skoltech are the leading university contributors, and AIRI is the leader among non-profit organisations.

Academic and Corporate Sectors: Different Models

The researchers marked up more than 3,100 organisational profiles, classifying them as academic, corporate, government-sector or mixed laboratories. The academic sector leads in all BRICS countries, accounting for more than 85% of publications. However, the role of companies varies considerably: while in Brazil, India and South Africa, the corporate sector accounts for around one-fifth of publications, in China and Russia it accounts for more than 40%.

In the overall ranking of BRICS organisations, the top positions are held by Chinese universities—Tsinghua University, Peking University and the Institute of Automation of the Chinese Academy of Sciences. Among companies, only Tencent and Alibaba make the top ten. The Indian ranking is dominated by academic institutions, while Brazil and South Africa are represented by individual organisations participating in international collaborations.

The authors describe the Russian case as a model of close cooperation between universities and technology companies. The expanded ranking of the top 100 BRICS organisations includes HSE University, Skoltech, MIPT, and AIRI, while Yandex holds the highest position among Russian representatives of the corporate sector. It is the only Russian company ranked among the top 20 BRICS corporate organisations by number of publications and, among companies of comparable size (with no more than 300 authors), ranks among the top five.

A distinctive feature of the Russian research landscape is the prominent role of a third category of organisations that belong neither to universities nor to companies. The leading organisation in this segment is the Artificial Intelligence Research Institute (AIRI), which nearly quadrupled its number of publications between 2020–2022 and 2023–2025.

A comparison of the 2020–2022 and 2023–2025 periods shows a noticeable restructuring of the institutional landscape. While research centres of foreign corporations led in terms of citations at the beginning of the period, Russian organisations have taken the top positions in recent years, with Yandex papers ranking first by weighted normalised citations. The company has maintained this high level while also producing one of the largest volumes of publications in the country. New players have also entered the rankings, including FITZ RAS, T-Technologies and MTS.

Productivity Depends on Bridge People

An analysis of 200 of the most productive research teams has shown that almost all of them bring together the academic and corporate sectors. This combination provides both access to computing resources and greater team stability. In such teams, the number of organisations and countries represented often exceeds the number of authors, as leading researchers typically have multiple affiliations.

This is particularly characteristic of Russia, where the key drivers are not individual organisations but researchers who connect universities, institutes of the Russian Academy of Sciences, and corporate laboratories. This model is common among all major technology players. Yandex has 73 joint publications with HSE University, 42 with MIPT, and 31 with Skoltech. Other technology companies are following a similar model: Sber has 42, 22, and 24 joint publications with the same organisations, respectively, while MTS and T-Technologies are also building ties with academic institutions.

China as a Separate Pole

China accounted for 93.1% of publications and 96% of citations in the entire BRICS dataset, representing 36.5% of the world's publications and ranking second globally. Chinese publications account for almost 54% of citations within China itself, which is largely unsurprising given the size of the country's research community. Other BRICS countries do not show the same degree of internal concentration: Russia accounts for around 6% of citations in domestic references, India for about 8%, and Brazil for approximately 3.5%. Moreover, Chinese authors are becoming the main source of citations for publications from Brazil, India, and Russia. According to the authors, this finding provides a practical guideline for promoting research: the audience that determines the scientific impact of BRICS research is increasingly the Chinese academic community.

BRICS Shared Research Space Still in the Formative Stage

The paper's main conclusion is that the hypothesis of a shared BRICS research space in computer science has not yet been confirmed. The United States remains the main international partner for each of the five countries. For China, Hong Kong, Singapore and Australia are also important partners, while Russia maintains significant ties with Germany, France, the United Kingdom and Saudi Arabia. Intra-BRICS collaboration remains limited: only the India–China and Russia–China pairs have produced a significant number of joint projects.

According to the authors, the current picture is less a shared market for scientific knowledge than a collection of strong research teams connected primarily to external leaders rather than to one another. At the same time, this points to an untapped opportunity: stable patterns of interaction within BRICS have yet to emerge, leaving considerable potential for expanding cooperation.

Alexey Naumov

'Machine learning is developing so rapidly that important advances are increasingly emerging at the intersection of basic science and technological development,' says Alexey Naumov, Director of the AI and Digital Science Institute at HSE FCS. 'It is therefore especially important for universities and companies not to operate in parallel but to form joint research teams, share expertise and infrastructure, and invest in long-term R&D.'

He also pointed out that for Russia, the landscape is no longer dominated by a single centre of expertise. Instead, it has become a highly competitive environment in which several strong players are developing their own research areas while also participating in academic collaborations. 'On the other hand, the study shows that, despite the high productivity of BRICS countries at leading ML conferences, stable scientific collaboration within the association has yet to emerge,' says Alexey Naumov. 'The potential for collaboration is enormous but realising it requires more than publications: it also needs personal contacts, discussions, and face-to-face meetings. In-person scientific events can provide valuable opportunities for such interaction. At HSE University, we hold the Fall into ML conference every year, bringing together authors from leading A* conferences. We see how direct communication helps researchers find common ground, build new collaborations, and move joint projects forward.'

The research findings can be used to identify cutting-edge areas and promising research teams, foster partnerships between academic organisations and companies, and inform decisions on supporting specific research areas.

The full results, including country-specific rankings of organisations, are available on the project page. The indicators are updated as publication metadata in international databases is updated.

The HSE Scientometrics Centre's research is based on bibliometric and network analysis. It covers the period from 2020 to 2025 and draws on the proceedings of ten top-level A* conferences rated by ICORE 2026. When linking publications to organisations, the researchers used only the official organisation names assigned to profiles in scientometric systems.

In computer science, research findings are more often published at conferences than in journals, as the field is evolving too rapidly for the traditional journal publication cycle. The selected conferences cover key areas including computer vision, natural language processing, machine learning, and artificial intelligence. About 83% of the papers are part of the main programme, the most rigorous and selective segment. To improve data accuracy, the researchers manually reviewed approximately 3,500 affiliations. This process also enabled them to identify more than 2,000 publications from BRICS countries that had not initially been attributed to those countries in scientometric databases.