Pith. sign in

REVIEW 3 major objections 5 minor 52 references

Scientific exploration, collaboration and labor division in the large language model era

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that, after large language models became widely available in late 2022, scientists in a large biomedical sample began publishing in more intellectually distant fields, entering new fields more often, collaborating with…

desk verdict Large-scale descriptive evidence that post-2022 science became more exploratory and more modularly staffed; the AI-writing-intensity comparisons are suggestive but rest on a proxy that needs further validation. read the letter →

arxiv 2607.20923 v1 pith:5A3YWYZS submitted 2026-07-23 cs.DL cs.AIcs.CY

Xiang Zheng , Xi Hong , Jialin Liu , Chaoqun Ni This is my paper · ORCID
classification cs.DLcs.AIcs.CY
keywords largelanguagemodelsscientificcollaborationinterdisciplinarityresearchexplorationlabordivisionCRediTAI-assistedwritingscienceof
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

After late 2022, when large language models became widely available, scientists in this large sample changed how they choose research directions, whom they collaborate with, and how they divide tasks. The paper finds that researchers publish across more intellectually distant fields, enter fields new to them at higher rates, and collaborate with a more disciplinarily diverse set of coauthors, with the biggest shifts among established scientists and researchers from non-English-speaking low- and middle-income countries. Authors whose papers show stronger AI-writing signals were already more interdisciplinary and exploratory before 2022, and the gap with low-AI-writing authors widened afterward. Within teams, contributors report fewer and more distinct roles, with software and validation duties growing while conceptual and management duties shrink. The authors read this as a broad reorganization of scientific work in the LLM era, not as proof that LLMs caused these changes.

What carries the argument

The argument is carried by two instruments. The first is an author-level AI-writing fraction, estimated from a word-frequency mixture model that treats each paper as a mix of human-written and AI-generated sentences and fits the mixing fraction by maximum likelihood; it is the paper's proxy for LLM exposure. The second is a family of author-year portfolio measures built from field distances computed as cosine distances between fields' reference profiles, including a field-diversity index, a Rao-Stirling interdisciplinarity index, and a pivot-size measure that compares each paper's references to the author's prior three-year reference profile. For division of labor, the paper uses CRediT role statements to compute roles per author, pairwise Jaccard similarity of coauthor role sets, and a degree-preserving within-paper shuffle that tests whether role combinations deviate from random allocation. These measures together let the paper connect textual AI-writing signals to behavioral changes at the portfolio, network, and team levels.

What would settle it

Take the same author population and measure LLM exposure directly, for example through linked survey responses or keystroke and editor logs, and check whether the high-low gaps in interdisciplinarity and role differentiation persist; if the AI-writing fraction does not track actual LLM use or the gaps vanish when actual use is controlled, the central claim is falsified. A complementary check is to apply the AI-writing proxy to a placebo corpus written before LLMs existed and see whether the same high-low gaps appear in counterfactual pre-2022 years; they should not.

Watch

Extended reading notes

Core claim

Using publication histories for 775,323 scientists and contribution statements from 137,120 multi-author papers, the paper establishes that the post-2022 period coincides with systematic increases in three linked dimensions: portfolio breadth and exploration, interdisciplinary collaboration, and differentiation of labor. It further shows that scientists with high AI-writing rates were already more interdisciplinary and exploratory before LLM diffusion and that the gap widened after 2022, supporting a selection-plus-reinforcement interpretation rather than a simple treatment effect. On division of labor, the decline in individual role counts is driven mainly by reduced sharing of the same roles among coauthors, and role combinations move closer to a randomized within-paper baseline, indicating weaker fixed role bundling. The authors are explicit that the design is descriptive and cannot separate LLM adoption from other post-2022 changes.

Load-bearing premise

The load-bearing premise is that the paper-level AI-writing fraction, a word-frequency score, is a valid author-level proxy for actual LLM exposure; if it is biased by field, time, or writing style, the high-versus-low AI-writing comparisons and the selection-plus-reinforcement conclusion could be artifacts.

Editorial extensions

If this is right

  • If the portfolio-breadth finding holds, research evaluations that treat field concentration as a stability metric will need to view post-2022 breadth as a real behavioral shift rather than an indexing artifact.
  • If high-AI-writing authors were already more exploratory, providing LLM access alone will not make all scientists equally interdisciplinary; the gap between early adopters and others may widen.
  • If the weaker link between collaborator diversity and paper interdisciplinarity holds for high-AI-writing authors, individual-level AI tool use may be partly substituting for the knowledge-integration role that cross-field collaborators previously played.
  • If role differentiation continues, contribution-reporting norms and evaluation rubrics that reward broad, shared role profiles will need to accommodate narrower, more modular task assignments.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • My inference: the selection-plus-reinforcement pattern suggests a widening rather than a converging trajectory, with already-exploratory scientists gaining the most from LLM tools and less-exploratory scientists potentially falling further behind even as tool access becomes universal.
  • My inference: the CRediT role shifts may partly reflect changes in how journals ask authors to report contributions rather than in how work is actually performed; comparing the same journals' reporting instructions over time would isolate the reporting-norm component.
  • My inference: if AI tools are substituting for some collaborative functions, one testable extension is that team sizes in high-AI-writing venues will stagnate or decline relative to pre-2022 trends after controlling for field effects, a pattern this paper does not directly test.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This paper uses a linked PubMed Central/OpenAlex sample of 775,323 authors and 137,120 multi-author CRediT papers to describe changes in scientific exploration, collaboration, and the division of labor after the 2022 public diffusion of large language models. The authors report that scientists' field portfolios became broader and more exploratory after 2022, that authors with stronger AI-writing signals were already more interdisciplinary and exploratory before that date and the gap widened afterward, that collaboration networks became more interdisciplinary, and that reported CRediT role sets became narrower, less overlapping, and less rigidly bundled. The paper is explicitly descriptive and repeatedly cautions that the AI-writing measure is a textual proxy rather than direct evidence of LLM use.

Significance. If the descriptive findings hold, this is an important and timely contribution to the science-of-science literature, documenting a broad reorganization of scientific work in the LLM era. The study's strengths include its very large linked dataset, the use of multiple complementary measures (field counts, Shannon entropy, HHI, Rao-Stirling index, pivot size, reference-field diversity), author-level fixed-effects event studies, CEM matching, and a candid limitations section. The temporal trends are supported by several robustness checks. However, the AI-intensity comparisons and the selection-plus-reinforcement interpretation are more fragile because the AI-writing proxy's measurement invariance is not established; the quantitative counterfactual claims also lack uncertainty quantification. These issues are fixable with additional analyses, so the paper is promising but not yet ready in its current form.

major comments (3)
  1. [Supplementary Note 1 and Fig. 2(c)-(i)] The AI-writing fraction is estimated from a word-frequency mixture model trained on bioRxiv text and applied to PMC papers, but no evidence is provided for measurement invariance across the 26 fields or over time. Because high- and low-AI-writing groups are defined from the same 2023-2025 window in which the outcome measures are computed, any field- or time-specific stylistic feature of interdisciplinary or exploratory papers that resembles LLM text could masquerade as 'AI-writing signal'. The validation in Supplementary Fig. S13 uses only abstracts and commercial detectors, not full-text, per-field, or per-year calibration. This is load-bearing because the 'selection-plus-reinforcement' interpretation in the Discussion and the widening-gap claims in Fig. 2 and Fig. 3 depend on the proxy. I request a holdout-style validation: report field-specific and year-specific false-positive rates against labeled human/AI text, or demonstrate that the association between the proxy and Rao-Stirling index, pivot size, and collaborator diversity survives within-field and within-subfield calibration.
  2. [Figs. 1-3, Results sections on research portfolios and collaboration] The quantitative claims of the form '15.6% higher than expected', '13.0% higher', and '23.2% higher' are based on linear extrapolation of 2011-2022 trends, but no uncertainty is attached to the dashed counterfactual extensions. With only two to three post-2022 observation points and noticeable pre-trend noise (e.g., the pivot measure declining before rebounding), these percentage excesses may be sensitive to the choice of pre-period window and functional form. I ask for forecast intervals or bootstrap confidence bands on the excess values, and for sensitivity analyses using alternative counterfactuals (e.g., a shorter pre-COVID window, a damped trend, or a placebo break year).
  3. [Fig. 3h and Supplementary Note 2] The claim that the association between paper-based and collaborator-based interdisciplinarity is 'consistently weaker' for high-AI-writing authors and weakens further after 2022 is central to the individual-level knowledge integration interpretation, but the manuscript does not report the estimating equation, coefficient estimates, or confidence intervals for this interaction. The event-study models in Supplementary Note 2 are described for the high-low comparisons in Supplementary Fig. S9, not for the Fig. 3h association. Please specify the model, present the interaction coefficients with standard errors, and ideally show an event-study plot with the high-low difference in the paper-collaborator Rao-Stirling slope over time.
minor comments (5)
  1. [Methods and Fig. 1 caption] Career-stage bins are inconsistent: the Methods section defines early career as %u22645 years, mid-career as 6-15 years, and senior as %u226516 years, while Fig. 1's caption refers to Junior (<6), Early career (6-15), and Advanced (>15 yr). Please align the definitions.
  2. [Throughout] The paper alternates between 'after 2022', 'post-2022', and dashed vertical lines at 2023. Please standardize the epoch boundary description (e.g., 'the 2023-2025 period') to avoid ambiguity about whether 2022 is included in the pre- or post-period.
  3. [Methods, Eq. (1) and Research Pivot Measures] The pivot measure is described as 'cosine distance ... using Eq. 1', but Eq. (1) defines field distance from reference vectors of fields. Please state explicitly that the pivot measure applies the same cosine-distance formula to focal and prior reference vectors, and clarify the vector construction in both cases.
  4. [Supplementary Note 2, Eq. (S8)] In Eq. (S8), the term n_it appears without a coefficient, making it look like a regressor without a parameter. Please write the model with an explicit coefficient for publication counts or state that it is included as a control.
  5. [Results, Country heterogeneity paragraph] The sentence 'This is most likely driven by Chinese authors' would be stronger with a formal statistical test of country-by-year interactions rather than a visual inspection of Supplementary Fig. S7.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: AI-writing proxy is externally anchored and bibliometric outcomes are independently measured.

full rationale

The paper's central claims are empirical associations and temporal descriptions, not derivations from the AI-writing measure. The AI-writing fraction is a text-based mixture estimate imported from external prior work (Liang et al.; Kobak et al.), with word distributions estimated on bioRxiv, and the high/low thresholds (0.15/0.05) are anchored to the 2021-2022 paper-level baseline rather than fitted to the portfolio, collaboration, or role outcomes. Field diversity, pivot, collaboration, and CRediT role measures are computed from OpenAlex metadata and contribution statements that are not defined in terms of the AI-writing signal. The CRediT decomposition is an explicit accounting identity used to separate components, not a prediction derived from the identity, and the role-pair shuffle is a within-paper null model. Self-citations (refs 11 and 36; Supplementary refs 6-7) provide prior validation and contextual estimates for the proxy, but the proxy itself is external and the validation is externally falsifiable against commercial detectors; these citations are supporting, not load-bearing. The paper also explicitly disclaims causal interpretation and notes the proxy's limitations. The same post-2022 window supplying both group definition and some outcome measurements is a confounding/validity concern, not a definitional equivalence, so it does not constitute circularity under the stated criteria.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claims rely on measurement and counterfactual assumptions rather than invented theoretical entities. The free parameters are threshold choices and discretization granularity; none is fitted to the outcome variables, which keeps circularity low, but the thresholds and the linear-trend counterfactual should be tested for sensitivity.

free parameters (4)
  • AI-writing group thresholds = 0.05 / 0.15
    Low-AI-writing authors are <=0.05 and high are >=0.15 author-level post-2023 AI-writing rate; thresholds anchored to 2021-2022 paper-level means (about 0.05) and 3x baseline, chosen by the authors on the same data.
  • Specialist-generalist HHI cutoff = 0.5
    Authors with trailing three-year field HHI at least 0.5 are classified as specialists; the cutoff is conventional but hand-chosen.
  • Field discretization = 26 fields, 254 subfields
    Papers are assigned to one primary OpenAlex field; all distance and diversity measures depend on this granularity, which is only partially tested in robustness checks.
  • Career-stage bins = 5, 6-15, 16+ years
    Used for matching and descriptive splits; standard but arbitrary boundaries.
assumptions (5)
  • domain assumption The Liang et al. mixture model word probabilities from bioRxiv human and AI sentences transfer to PMC full-text biomedical writing.
    Supplementary Note 1 uses these probabilities directly to estimate the AI-writing fraction; field-specific or time-varying vocabulary shifts could bias the estimate.
  • domain assumption OpenAlex topic-to-field assignments and PMC-to-OpenAlex author disambiguation are accurate enough that yearly field portfolios reflect real research activities.
    All field diversity, exploration, and collaboration measures derive from this linkage, as described in the Data section.
  • domain assumption CRediT role statements reflect actual task division and are reported consistently over time within the studied journals.
    The labor-division analysis interprets role strings as behavioral facts; the authors themselves note that roles are reported categories, not direct observations, in the Limitations section.
  • ad hoc to paper Pre-2023 linear trends are the correct counterfactual for post-2022 expectations.
    Figures 1-3 use linear fits from 2011-2022 to define expected post-2022 levels; this assumes no other nonlinear drivers would have changed the trajectories.
  • domain assumption The shuffled within-paper role baseline is a valid null model for role bundling.
    Role-pair deviations are judged against a degree-preserving random shuffle, as described in the Methods section, which assumes conditional random allocation is the relevant null.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Scientific exploration, collaboration and labor division in the large language model era." pith.science (2026). https://pith.science/paper/5A3YWYZS

@misc{pith2026260720923,
  author       = {Pith},
  title        = {Pith review of: Scientific exploration, collaboration and labor division in the large language model era},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5A3YWYZS}},
  note         = {Machine review of arXiv:2607.20923}
}
read the original abstract

Large language models (LLMs) have rapidly and significantly entered scientific workflows, but it remains unclear how their diffusion is associated with changes in scientists' strategies in research directions and team building. We link PubMed Central full text with OpenAlex publication and collaboration histories for 775,323 scientists and analyze CRediT contribution statements from 137,120 multi-author papers. After 2022, scientists increasingly published across more intellectually distant fields and entered fields in which they had not previously worked. These increases in interdisciplinarity and exploration were especially pronounced among established scientists and scientists from non-English-speaking low- and middle-income countries. Authors with stronger AI-writing signals were already more interdisciplinary and exploratory before the widespread adoption of LLMs, and the gap widened further after 2022 compared with authors with weaker AI-writing signals. Scientists' collaboration networks also became more interdisciplinary after 2022. Yet, among authors with stronger AI-writing signals, research interdisciplinarity was less closely tied to the disciplinary diversity of their collaborators. The division of labor within research teams also became more differentiated. Contributors on papers published after 2022 reported narrower role sets on average, coauthors shared fewer roles in common, and their role profiles became less rigid and more fluid. Software and validation roles increased, while conceptual and management roles decreased. These patterns suggest that team members are taking on more distinct responsibilities and may rely less on one another to perform research tasks. Overall, this study indicates that the LLM era coincides with a broader reorganization of scientific exploration, collaboration, and the division of labor.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 28 canonical work pages

  1. [1]

    The Rise of Large Language Models.Nature Computational Science5(9), 689–690 (2025), doi:10.1038/s43588-025-00890-x

    work page doi:10.1038/s43588-025-00890-x 2025

  2. [2]

    A. R. Doshi, O. P. Hauser, Generative AI enhances individual creativity but reduces the collective diversity of novel content.Science Advances10(28), eadn5290 (2024), doi:10. 1126/sciadv.adn5290

    work page 2024

  3. [3]

    B. C. Lee, J. J. Chung, An empirical investigation of the impact of ChatGPT on creativity. Nature Human Behaviour8(10), 1906–1914 (2024), doi:10.1038/s41562-024-01953-1

    doi:10.1038/s41562-024-01953-1 2024

  4. [4]

    C. Si, D. Yang, T. Hashimoto, Can LLMs Generate Novel Research Ideas? A Large- Scale Human Study with 100+ NLP Researchers, inInternational Conference on Learn- ing Representations, Y. Yue, A. Garg, N. Peng, F. Sha, R. Yu, Eds., vol. 2025 (2025), pp. 94003–94092,https://proceedings.iclr.cc/paper_files/paper/2025/file/ ea94957d81b1c1caf87ef5319fa6b467-Pape...

    work page 2025

  5. [5]

    Yang,et al., Can Large Language Models Help Young Researchers Develop New Clinical Research Ideas?The Lancet Digital Health8(4) (2026), doi:10.1016/j.landig.2026.100983

    Y. Yang,et al., Can Large Language Models Help Young Researchers Develop New Clinical Research Ideas?The Lancet Digital Health8(4) (2026), doi:10.1016/j.landig.2026.100983

    work page arXiv 2026

  6. [6]

    Stokel-Walker, R

    C. Stokel-Walker, R. Van Noorden, What ChatGPT and generative AI mean for science.Nature 614(7947), 214–216 (2023), doi:10.1038/d41586-023-00340-6

    work page doi:10.1038/d41586-023-00340-6 2023

  7. [8]

    Chugunova,et al., Who uses AI in research, and for what? Large-scale survey evidence from Germany.Research Policy55(2), 105381 (2026), doi:10.1016/j.respol.2025.105381

    M. Chugunova,et al., Who uses AI in research, and for what? Large-scale survey evidence from Germany.Research Policy55(2), 105381 (2026), doi:10.1016/j.respol.2025.105381

    work page arXiv 2026

  8. [13]

    Van Noorden, J

    R. Van Noorden, J. M. Perkel, AI and science: what 1,600 researchers think.Nature621(7980), 672–675 (2023), doi:10.1038/d41586-023-02980-0

    doi:10.1038/d41586-023-02980-0 2023

Show all 52 references
  1. [14]

    J. Gao, D. Wang, Quantifying the Use and Potential Benefits of Artificial Intelligence in Scientific Research.Nature Human Behaviour8(12), 2281–2292 (2024), doi:10.1038/ s41562-024-02020-5

    2024

  2. [15]

    Kusumegi,et al., Scientific production in the era of large language models.Science 390(6779), 1240–1243 (2025), doi:10.1126/science.adw3000

    K. Kusumegi,et al., Scientific production in the era of large language models.Science 390(6779), 1240–1243 (2025), doi:10.1126/science.adw3000

    2025 doi

  3. [16]

    S. Noy, W. Zhang, Experimental evidence on the productivity effects of generative artificial intelligence.Science381(6654), 187–192 (2023), doi:10.1126/science.adh2586

    2023 doi

  4. [17]

    J. G. March, Exploration and Exploitation in Organizational Learning.Organization Science 2(1), 71–87 (1991)

    1991

  5. [18]

    Hill,et al., The pivot penalty in research.Nature642(8069), 999–1006 (2025), doi: 10.1038/s41586-025-09048-1

    R. Hill,et al., The pivot penalty in research.Nature642(8069), 999–1006 (2025), doi: 10.1038/s41586-025-09048-1

    2025 doi

  6. [19]

    Tripodi,et al., Tenure and research trajectories.Proceedings of the National Academy of Sciences122(30), e2500322122 (2025), doi:10.1073/pnas.2500322122

    G. Tripodi,et al., Tenure and research trajectories.Proceedings of the National Academy of Sciences122(30), e2500322122 (2025), doi:10.1073/pnas.2500322122

    2025 doi

  7. [20]

    Death of the Renaissance Man

    B. F. Jones, The Burden of Knowledge and the “Death of the Renaissance Man”: Is Innovation Getting Harder?The Review of Economic Studies76(1), 283–317 (2009), doi:10.1111/j. 1467-937X.2008.00531.x

    2009 arXiv

  8. [21]

    Leahey, C

    E. Leahey, C. M. Beckman, T. L. Stanko, Prominent but Less Productive: The Impact of Interdisciplinarity on Scientists’ Research*.Administrative Science Quarterly62(1), 105– 139 (2017), doi:10.1177/0001839216665364. 19

    2017 doi

  9. [22]

    L. Li,et al., Deciphering Scientific Collaboration in Biomedical LLM Research: Dynamics, Institutional Participation, and Resource Disparities.arXiv preprint arXiv:2511.00818(2025), doi:10.48550/arXiv.2511.00818,http://arxiv.org/abs/2511.00818

    2025 doi

  10. [23]

    H. Xu, H. Liu, Y. Leng, Y. Ding, Adapting to LLMs: How Insiders and Outsiders Reshape Scientific Knowledge Production.arXiv preprint arXiv:2505.12666(2025), doi:10.48550/ arXiv.2505.12666,http://arxiv.org/abs/2505.12666

    2025 doi arXiv

  11. [24]

    K. L. Hall,et al., The science of team science: A review of the empirical evidence and research gaps on collaboration in science.American Psychologist73(4), 532–548 (2018), doi:10.1037/amp0000319

    2018 doi

  12. [25]

    Larivi`ere, S

    V. Larivi`ere, S. Haustein, K. B¨orner, Long-distance interdisciplinarity leads to higher scientific impact.PloS One10(3), e0122565 (2015), doi:10.1371/journal.pone.0122565

    2015 doi

  13. [26]

    Salatino, F

    A. Salatino, F. Osborne, D. R. Recupero, S. Angioni, E. Motta, Does Diversity of Expertise Drive Citation Impact? Evidence from Computer Science.Scientometrics131(2), 1119–1146 (2026), doi:10.1007/s11192-026-05560-x

    2026 doi

  14. [27]

    L. Li, L. Dinh, S. Hu, L. Hemphill, Academic Collaboration on Large Language Model Studies Increases Overall but Varies across Disciplines doi:10.48550/arXiv.2408.04163,http:// arxiv.org/abs/2408.04163

    doi arXiv

  15. [28]

    C. Lu,et al., Co-contributorship network and division of labor in individual scientific col- laborations.Journal of the Association for Information Science and Technology71(10), 1162–1178 (2020), doi:10.1002/asi.24321

    2020 doi

  16. [29]

    D. A. Boiko, R. MacKnight, B. Kline, G. Gomes, Autonomous chemical research with large language models.Nature624(7992), 570–578 (2023), doi:10.1038/s41586-023-06792-0

    2023 doi

  17. [30]

    Wang, AI agents are ‘aeroplanes for the mind’: five ways to ensure that scientists are responsible pilots.Nature651(8104), 32–34 (2026), doi:10.1038/d41586-026-00665-y

    D. Wang, AI agents are ‘aeroplanes for the mind’: five ways to ensure that scientists are responsible pilots.Nature651(8104), 32–34 (2026), doi:10.1038/d41586-026-00665-y

    2026 doi

  18. [31]

    J. W. Burton,et al., How Large Language Models Can Reshape Collective Intelligence.Nature Human Behaviour8(9), 1643–1655 (2024), doi:10.1038/s41562-024-01959-9

    2024 doi

  19. [32]

    A. W. Woolley, Generative AI and Collaboration: Opportunities for Cultivating Collec- tive Intelligence.Journal of Organization Design15(1), 45–50 (2026), doi:10.1007/ s41469-025-00199-z

    2026

  20. [33]

    L. Ding, C. Lawson, P. Shapira, Rise of Generative Artificial Intelligence in Science.Scien- tometrics130(9), 5093–5114 (2025), doi:10.1007/s11192-025-05413-z

    2025 doi

  21. [34]

    Q. Hao, F. Xu, Y. Li, J. Evans, Artificial Intelligence Tools Expand Scientists’ Im- pact but Contract Science’s Focus.Nature649(8099), 1237–1243 (2026), doi:10.1038/ s41586-025-09922-y

    2026

  22. [35]

    Ben-Zion, E

    Y. Ben-Zion, E. Cohen, N. Davidovitch, Changes in Manuscript Length, Research Team Size, and International Collaboration in the Post-2022 Period: Evidence from PLOS ONE (2026), doi:10.48550/arXiv.2603.01718. 20

    2026 doi

  23. [36]

    E. Yan, C. Ni, AI-assisted Writing and the Reorganization of Scientific Knowledge (2026), doi:10.48550/arXiv.2604.14126

    2026 doi arXiv

  24. [38]

    Z. Liao,et al., LLMs as Research Tools: A Large Scale Survey of Researchers’ Usage and Perceptions.arXiv preprint arXiv:2411.05025(2024), doi:10.48550/arXiv.2411.05025, https://ui.adsabs.harvard.edu/abs/2024arXiv241105025L

    2024 doi arXiv

  25. [39]

    Y. Lin, Z. Risha, E. Leahey, L. Wu, The Longevity of Innovation.arXiv preprint arXiv:2606.29777(2026), doi:10.48550/arXiv.2606.29777,http://arxiv.org/abs/ 2606.29777

    2026 doi arXiv

  26. [40]

    Zhang, Z

    L. Zhang, Z. Wang, Z. Gou, E. Yan, G. Sivertsen, How Scientific Labor Is Divided in Interna- tional Collaboration: Comparative Evidence from China and the United States.Scientometrics (2026), doi:10.1007/s11192-026-05669-z

    2026 doi

  27. [41]

    L. Guo, Y. Wang, M. Li, Exploration, Exploitation and Funding Success: Evidence from Junior Scientists Supported by the Chinese Young Scientists Fund.Journal of Informet- rics18(2), 101492 (2024), doi:10.1016/j.joi.2024.101492,https://www.sciencedirect. com/science/article/pii...

    2024

  28. [42]

    rep., Elsevier (2025)

    Elsevier,Researcher of the Future: A Confidence in Research report, Tech. rep., Elsevier (2025)

    2025

  29. [43]

    Branda, M

    F. Branda, M. Ciccozzi, F. Scarpa, Artificial intelligence in scientific research: Challenges, opportunities and the imperative of a human-centric synergy.Journal of Informetrics19(4), 101727 (2025), doi:10.1016/j.joi.2025.101727

    2025

  30. [44]

    T. Zheng,et al., From Automation to Autonomy: A Survey on Large Language Models in Scientific Discovery, inProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing(2025), pp. 17733–17750, doi:10.18653/v1/2025.emnlp-main. 895,https://aclanthology.o...

    2025 doi

  31. [45]

    F. Dell’ Acqua,et al.,The Cybernetic Teammate: A Field Experiment on Generative AI Reshap- ing Teamwork and Expertise, Working Paper 33641, National Bureau of Economic Research (2025), doi:10.3386/w33641,https://www.nber.org/papers/w33641

    2025 doi

  32. [46]

    M. Chelli,et al., Hallucination Rates and Reference Accuracy of ChatGPT and Bard for Systematic Reviews: Comparative Analysis.Journal of Medical Internet Research26(1), e53164 (2024), doi:10.2196/53164

    2024 doi

  33. [47]

    Li,et al., BibAgent: An Agentic Framework for Traceable Miscitation Detection in Scientific Literature (2026), doi:10.48550/arXiv.2601.16993

    P. Li,et al., BibAgent: An Agentic Framework for Traceable Miscitation Detection in Scientific Literature (2026), doi:10.48550/arXiv.2601.16993

    2026 doi

  34. [48]

    Zhao,et al., LLM hallucinations in the wild: Large-scale evidence from non-existent citations (2026),https://arxiv.org/abs/2605.07723

    Z. Zhao,et al., LLM hallucinations in the wild: Large-scale evidence from non-existent citations (2026),https://arxiv.org/abs/2605.07723. 21 Supplementary Materials for Scientific exploration, collaboration and labor division in the large language model era Xiang Zheng1, Xi Ho...

    2026 arXiv

  35. [51]

    Liang,et al., Quantifying large language model usage in scientific papers.Nature Human Behaviour9(12), 2599–2609 (2025), doi:10.1038/s41562-025-02273-8

    W. Liang,et al., Quantifying large language model usage in scientific papers.Nature Human Behaviour9(12), 2599–2609 (2025), doi:10.1038/s41562-025-02273-8

    2025 doi

  36. [52]

    Kobak, R

    D. Kobak, R. Gonz ´alez-M´arquez, E.- ´A. Horv´at, J. Lause, Delving into LLM-assisted writing in biomedical publications through excess vocabulary.Science Advances11(27), eadt3813 (2025), doi:10.1126/sciadv.adt3813

    2025 doi

  37. [53]

    Y. He, Y. Bu, Academic journals’ AI policies fail to curb the surge in AI-assisted academic writing.Proceedings of the National Academy of Sciences123(9), e2526734123 (2026), doi:10.1073/pnas.2526734123

    2026 doi

  38. [54]

    Siler, The diffusion of large language models in published academic articles.Proceed- ings of the National Academy of Sciences123(22), e2605754123 (2026), doi:10.1073/pnas

    K. Siler, The diffusion of large language models in published academic articles.Proceed- ings of the National Academy of Sciences123(22), e2605754123 (2026), doi:10.1073/pnas. 2605754123

    2026 doi

  39. [55]

    W. Liang,et al., Monitoring AI-modified content at scale: a case study on the impact of ChatGPT on AI conference peer reviews.Proceedings of the 41st International Conference on Machine Learning235, 29575–29620 (2024)

    2024

  40. [56]

    J. Liu, Y. Bu, Towards the relationship between AIGC in manuscript writing and author profiles: evidence from preprints in LLMs.arXiv preprint arXiv:2404.15799(2024), doi:10.48550/ arXiv.2404.15799,https://ui.adsabs.harvard.edu/abs/2024arXiv240415799L

    2024 doi arXiv

  41. [57]

    J. Liu, Y. He, Z. Zheng, Y. Bu, C. Ni, AI-Assisted Writing Is Growing Fastest Among Non- English-Speaking and Less Established Scientists.arXiv preprint arXiv:2511.15872(2025), doi:10.48550/arXiv.2511.15872,http://arxiv.org/abs/2511.15872

    2025 doi

  42. [58]

    Zeng,et al., Increasing Trend of Scientists to Switch between Topics.Nature Communica- tions10(1), 3439 (2019), doi:10.1038/s41467-019-11401-8

    A. Zeng,et al., Increasing Trend of Scientists to Switch between Topics.Nature Communica- tions10(1), 3439 (2019), doi:10.1038/s41467-019-11401-8. S25

    2019 doi

  43. [2023]

    The outcome measures are from Figs

    Error bands denote 95% confidence intervals. The outcome measures are from Figs. 1-2. Positive post-2022 coefficients indicate that high-AI-writing authors increased more than matched low-AI-writing authors, relative to the corresponding high-low difference in 2022. S15 Share ...

    2022

  44. [2025]

    Point colors identify the two fields in each pair, and larger points represent larger absolute changes in author share

    Each point is an unordered pair of paper primary fields, with the x-axis showing the share of authors publishing in that field pair during 2023-2025 and the y-axis showing the corresponding share during 2020-2022; points below the 45-degree line indicate field pairs that becam...

    2023

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.