How many arXiv preprints are published in journals? Data by field (2022–2025)
Key numbers
- 40% of the arXiv papers submitted in 2022 now show a journal reference or a DOI on arXiv (75,197 of 185,692).
- Highest: Astrophysics (astro-ph), 87%.
- Lowest: Economics (econ), 18%.
- For papers submitted in 2025, the share is still 21%: publication takes time.
- Counting the papers whose arXiv comments say they were accepted (mostly at conferences), computer science goes from 20% to 40%.
Counted by SummarizePaper from the metadata of every arXiv paper submitted from 2022 to 2025 (arXiv OAI-PMH, harvested on 21 November 2026). A journal reference is added by the authors, or by arXiv from some publishers, so these shares are a lower bound.
What share of arXiv preprints are published in journals?
At least 40% of the papers submitted to arXiv in 2022 point to a published version: their arXiv record carries a journal reference or a DOI. The share depends on how long papers have had to be published:
| Submitted in | Papers | With a journal reference or DOI | Share |
|---|---|---|---|
| 2022 | 185,692 | 75,197 | 40% |
| 2023 | 208,493 | 64,196 | 31% |
| 2024 | 244,031 | 62,923 | 26% |
| 2025 | 284,486 | 60,162 | 21% |
Which fields publish their preprints in journals most?
| Field | Submitted 2022 | Submitted 2023 | Submitted 2024 | Submitted 2025 |
|---|---|---|---|---|
| Astrophysics (astro-ph) | 87% | 58% | 45% | 46% |
| General relativity and quantum cosmology (gr-qc) | 81% | 64% | 58% | 52% |
| Nuclear physics (nucl-ex, nucl-th) | 77% | 55% | 48% | 44% |
| Condensed matter (cond-mat) | 74% | 61% | 55% | 47% |
| High-energy physics, theory (hep-th) | 74% | 46% | 37% | 33% |
| High-energy physics, phenomenology (hep-ph) | 72% | 49% | 40% | 38% |
| Quantum physics (quant-ph) | 69% | 59% | 52% | 39% |
| Nonlinear sciences (nlin) | 66% | 44% | 35% | 29% |
| High-energy physics, experiment (hep-ex) | 60% | 57% | 58% | 51% |
| Other physics (physics.*) | 57% | 41% | 35% | 31% |
| Mathematical physics (math-ph) | 55% | 43% | 34% | 23% |
| Quantitative biology (q-bio) | 29% | 23% | 21% | 16% |
| Electrical engineering and systems science (eess) | 25% | 21% | 20% | 16% |
| Mathematics (math) | 24% | 20% | 17% | 10% |
| Statistics (stat) | 20% | 18% | 16% | 11% |
| Computer science (cs) | 20% | 17% | 16% | 13% |
| Quantitative finance (q-fin) | 20% | 18% | 15% | 12% |
| Economics (econ) | 18% | 14% | 13% | 10% |
Physics and astronomy papers are usually submitted to arXiv when they go to a journal, and authors update the record once published. In computer science most papers go to conferences, whose proceedings are often not entered as a journal reference, so the low share there does not mean the papers stay unpublished.
What about conference papers?
Many authors write in the arXiv comments where a paper was accepted ("Accepted at NeurIPS 2022", "To appear in ..."), without adding a journal reference. Counting those papers too, for papers submitted in 2022:
| Field | Journal reference or DOI | Including accepted per the comments |
|---|---|---|
| Computer science (cs) | 20% | 40% |
| Electrical engineering and systems science (eess) | 25% | 39% |
| Statistics (stat) | 20% | 26% |
| Mathematics (math) | 24% | 30% |
| Astrophysics (astro-ph) | 87% | 91% |
| High-energy physics, experiment (hep-ex) | 60% | 63% |
| Quantitative biology (q-bio) | 29% | 33% |
| High-energy physics, phenomenology (hep-ph) | 72% | 76% |
| Quantitative finance (q-fin) | 20% | 23% |
| General relativity and quantum cosmology (gr-qc) | 81% | 83% |
| Nuclear physics (nucl-ex, nucl-th) | 77% | 80% |
| High-energy physics, theory (hep-th) | 74% | 76% |
| Nonlinear sciences (nlin) | 66% | 68% |
| Economics (econ) | 18% | 20% |
| Mathematical physics (math-ph) | 55% | 57% |
| Other physics (physics.*) | 57% | 59% |
| Quantum physics (quant-ph) | 69% | 70% |
| Condensed matter (cond-mat) | 74% | 75% |
The comments are free text written by the authors; "accepted" is read from words such as accepted, to appear or camera-ready, and comments that say "submitted to" or "under review" are not counted.
Where are arXiv preprints published?
The most frequent journals and conferences of papers submitted in 2022, read from their journal reference, DOI or arXiv comments, by field:
| Field | Most frequent journals and conferences (papers) |
|---|---|
| Computer science (cs) | NeurIPS (1,060), CVPR (1,006), AAAI (702), ECCV (633), EMNLP (586) |
| Mathematics (math) | J. Comput. Phys. (132), Algebr. Geom. Topol. (97), Advances in Mathematics (73), SIGMA (68), Alg. Number Th. (56) |
| Condensed matter (cond-mat) | Phys. Rev. B (3,099), Phys. Rev. Lett. (700), Phys. Rev. E (592), Nature Commun. (333), Phys. Rev. Research (321) |
| Astrophysics (astro-ph) | MNRAS (3,513), Astrophys. J. (2,477), Astron. Astrophys. (1,960), Astrophys. J. Lett. (581), Phys. Rev. D (521) |
| Other physics (physics.*) | Phys. Rev. A (317), JINST (238), J. Fluid Mech. (233), Opt. Express (206), Phys. Rev. E (205) |
| Electrical engineering and systems science (eess) | Interspeech (199), ICASSP (184), MICCAI (112), IEEE Trans. Signal Process. (55), IEEE Access (38) |
| Quantum physics (quant-ph) | Phys. Rev. A (986), Phys. Rev. Lett. (366), Quantum (355), Phys. Rev. Research (231), Phys. Rev. B (171) |
| High-energy physics, phenomenology (hep-ph) | Phys. Rev. D (1,232), JHEP (693), Eur. Phys. J. C (311), Phys. Lett. B (231), JCAP (134) |
| Statistics (stat) | NeurIPS (92), ICML (62), AISTATS (32), Journal of Machine Learning Research (26), ICLR (21) |
| High-energy physics, theory (hep-th) | JHEP (1,207), Phys. Rev. D (676), Eur. Phys. J. C (148), Phys. Lett. B (132), SciPost Phys. (108) |
| General relativity and quantum cosmology (gr-qc) | Phys. Rev. D (898), Eur. Phys. J. C (248), Class. Quantum Grav. (199), JCAP (142), Universe (105) |
| Quantitative biology (q-bio) | Phys. Rev. E (35), NeurIPS (21), Sci. Rep. (16), Chaos (12), PLOS ONE (12) |
| Nuclear physics (nucl-ex, nucl-th) | Phys. Rev. C (492), Phys. Lett. B (104), Phys. Rev. D (91), Phys. Rev. Lett. (65), Eur. Phys. J. A (55) |
| Economics (econ) | Journal of Econometrics (5), Econom. Theory (5) |
| Mathematical physics (math-ph) | J. Phys. A (38), J. Math. Phys. (29), J. Phys. A: Math. Theor. (27), Commun. Math. Phys. (15), Ann. Henri Poincaré (12) |
| High-energy physics, experiment (hep-ex) | Phys. Rev. D (131), JHEP (99), Phys. Rev. Lett. (83), Eur. Phys. J. C (63), Phys. Lett. B (36) |
| Quantitative finance (q-fin) | Entropy (9) |
| Nonlinear sciences (nlin) | Phys. Rev. E (68), Physica D (29), Chaos (24), J. Phys. A (19), Phys. Rev. Lett. (11) |
How to know if a preprint was published
- On arXiv, look for the "Journal reference" and "DOI" lines under the abstract; they are often missing even when the paper is published.
- On SummarizePaper, a paper page shows its journal version when one is found, and Premium readers can be alerted when a followed preprint is published.
- For a whole reading list, Premium watches a collection and emails you when one of its papers is published, retracted or corrected.
More research publishing data
- arXiv statistics: submissions per year and field
- The fastest-growing research topics
- How many authors per paper, by field
Methodology and limits
- Papers: every arXiv paper whose identifier starts with the year (2201–2212 for 2022, and so on), from the arXiv metadata of arXiv's OAI-PMH interface, harvested on 21 November 2026.
- Published: the arXiv record has a journal reference or a DOI. Both are entered by the authors or added by arXiv for some publishers; many published papers have neither, so every share here is a lower bound, lowest in fields that publish in conference proceedings.
- Field: the archive of the primary category; physics.* is "physics (other)", nucl-ex and nucl-th are nuclear physics. Fields with fewer than 300 papers in 2022 are left out.
- Journals: names read from the journal reference, normalized (abbreviations and spacing), or else from the DOI prefix of the main publishers, or else from a conference named in comments that say the paper was accepted; a paper whose journal cannot be recognized is not counted in that table.