Behind the Paper Explosion: Author Growth and Widening Gaps in Computer Science

Jiayi Li, Bowen Niu, Bingsheng He, Zhaomin WuarXiv 2026 · arXiv

The volume of computer science research is growing at a staggering rate, leaving researchers and peer reviewers struggling to keep up. A new study, "Behind the Paper Explosion: Author Growth and Widening Gaps in Computer Science," dives into the data from 2010 to 2025 to figure out what’s really driving this surge. Is it a "productivity boom" fueled by AI, or is something else going on?

TL;DR

The "paper explosion" in computer science is almost entirely driven by a massive influx of new authors rather than increased individual productivity. While the average output per author appears stable, this masks a dramatic rise in inequality: a small group of highly prolific authors is producing more than ever, while a growing "long tail" of low-output newcomers keeps the statistical mean in check.

What Problem Does This Paper Address?

The sheer volume of CS papers has reached a breaking point, straining the peer review system and making research discovery nearly impossible. With the recent rise of Generative AI, many assumed that individual researchers were simply becoming more productive, churning out papers faster than before.

However, aggregate statistics can be misleading. An "average" of two papers per author could mean everyone is writing two papers, or it could mean one person is writing ten while four others write none. Previous studies tracked either total paper counts or individual careers, but they rarely linked the two. This paper addresses that gap, asking: Is the growth coming from more people, or from the same people writing more?

Methods

The authors employ a clever "growth decomposition" based on a simple accounting identity:

Total Papers = Number of Authors × Average Papers per Author

By applying this to longitudinal data, they can precisely attribute growth to either population expansion or changes in individual output. To ensure their findings weren't just an artifact of growing team sizes (where one paper has 20 authors), they used both fractional counting (splitting credit equally) and first-author counting.

Key technical steps included:

  • Turnover Analysis: Tracking authors over time to see who is a "new entrant," who is "returning" after a break, and who has become "inactive."
  • Identity Resolution: Using a custom trajectory model and external records (OpenAlex, ORCID) to link author identities across different platforms (arXiv, DBLP, ICLR).
  • Historical Grouping: Comparing authors who had similar output levels in 2018–2020 to see how their paths diverged by 2025.

Datasets

The study draws from a massive pool of scholarly data:

  • arXiv: The primary source for CS preprints (2010–2025).
  • DBLP: For formal journal and conference publications.
  • ICLR: Submission records (including rejected papers) to capture the full demand for peer review.
  • OpenAlex & ORCID: Used for author disambiguation and verifying prior publication histories.
  • Comparison Fields: Mathematics and Physics were used as baselines to highlight how unique the CS growth pattern is.

Key Insights and Ideas

  • It’s a Population Boom, Not a Productivity Boom: Nearly 100% of the long-term growth in CS papers is due to more people entering the field. Even after the release of ChatGPT, author growth remains the dominant factor.
  • The "Stable Mean" is a Mirage: On paper, the average output per author hasn't changed much. But this is because a massive influx of low-output newcomers (often appearing on large teams) mathematically offsets a significant surge in output from established, prolific authors.
  • Widening Inequality: Trajectories are diverging wildly. The "rich get richer" effect is pronounced: the top 1% of authors saw their output grow by over 50%, while the median author’s output remained flat.
  • AI Subfields are the Engine: Growth isn't uniform across CS. AI-related categories (Machine Learning, Computer Vision) account for over 70% of the growth in arXiv CS papers.
Figure 3: CS subfield contributions to paper-count growth.

Key Results

  • Author Growth Dominance: In arXiv CS and DBLP, author growth accounts for virtually all paper growth. In some cases, it even exceeds 100%, meaning the population grew so fast it offset a slight decline in average fractional output.
Figure 2: Cumulative contributions of author growth and output per author to CS paper growth.
  • High Turnover: The field is incredibly transient. More than half of the authors active in 2021 had stopped publishing in arXiv CS by 2025. By 2025, new entrants made up about two-thirds of all active authors.
  • Exploding Variance: The variance of first-author output grew 10 times faster than its mean between 2021 and 2025. The Gini coefficient (a measure of inequality) for output rose by 14.4% in just four years.
  • Team Size Surge: The average number of authors per paper in arXiv CS jumped from 2.64 (2010-2012) to 4.75 (2023-2025), an 80% increase.

Limitations and Future Work

  • The "AI Cause" Question: While the timing of the output surge aligns with the rise of LLMs, the paper doesn't definitively prove that AI is the cause of the widening gaps. It’s possible that only those with access to massive compute or specific skills can leverage AI to boost productivity.
  • Reviewer Quality: While the growing author pool provides more potential reviewers, the study doesn't assess the quality or experience of these new reviewers.
  • Data Lags: DBLP data for the most recent years (2024-2025) is still being indexed, so some formal publication trends may be slightly understated.

The authors suggest that future work should link individual researchers' AI access and training to their output trajectories to see if "AI equity" is becoming a new barrier in the field.

Summary updated

Abstract

The rapid growth of computer science (CS) papers increases the work needed to review research and find relevant studies. The growing use of AI tools in scientific writing makes it increasingly important to identify the authors behind this growth. Previous studies have tracked publication growth and followed researchers' careers, documenting the increasing number of authors and the many researchers who publish only briefly. However, annual averages of papers per author reflect different sets of researchers and may hide changes among the same authors. This study addresses this gap by linking individual changes to overall CS paper growth using arXiv, DBLP, and conference submission records for 2010-2025. With each paper divided equally among its authors, the total paper count equals the number of authors multiplied by their average papers per author. This relationship provides a decomposition of the two contributions of paper growth. This decomposition attributes nearly all growth in CS paper volume to an increase in the number of publishing authors. The analysis of author numbers further distinguishes expansion through new and returning arXiv authors from the replacement of authors no longer publishing there. Changes in who publishes also affect the average, motivating comparisons of the same authors over time within groups with similar past paper counts. This comparison identifies substantial replacement within the growing author group. New authors average fewer papers under this counting method, keeping the overall average nearly unchanged despite increases among authors who published in both years. Moreover, statistics show that these increases are uneven, with widening gaps even among authors who previously published similar numbers of papers. These findings provide insights into peer review, community participation, researcher evaluation, and equity within the CS community.
BibTeX
@misc{li2026behind,
  author = {Jiayi Li and Bowen Niu and Bingsheng He and Zhaomin Wu},
  title = {{Behind the Paper Explosion: Author Growth and Widening Gaps in Computer Science}},
  year = {2026},
  eprint = {2610.39831},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.39831},
}
Figure