What is counted
- Venues: AAAI, IJCAI, ICLR, NeurIPS, ICML, CVPR, ICCV, ECCV, ACL and EMNLP, from 2020 on, grouped as CSRankings groups them: AI (AAAI, IJCAI), ML (ICLR, NeurIPS, ICML), CV (CVPR, ICCV, ECCV) and NLP (ACL, EMNLP). Years are conference years, so an ICLR 2026 paper was submitted in autumn 2025. ICCV is held in odd years and ECCV in even years. IJCAI-PRICAI 2020, held in January 2021, counts as 2020. EMNLP 2026 counts once its papers are in the ACL Anthology.
- People: Vietnamese lecturers, assistant professors, associate professors and professors whose main post is now at a university in Vietnam. Students, researchers at companies or institutes, and faculty from other countries are not tracked. Lecturers are included because the titles of associate professor and professor in Vietnam are conferred by a national council and are harder to obtain than in most other countries, so many lecturers do the work an assistant or associate professor does elsewhere. The list is kept by hand, and nobody appears before their entry has been approved and they have at least one counted submission at the ten venues, accepted or not.
- Only while in Vietnam: a paper counts from the year the professor joined a Vietnamese institution. Earlier years are greyed out on the professor’s page.
- At their institution: an accepted paper counts for a professor only if the paper lists them at their institution, as printed on the paper itself (checked against the PDF where the official list takes affiliations from author profiles). A paper written only at a company or research institute, such as Qualcomm AI Research (formerly VinAI), does not count. Unofficial papers, which print no affiliation yet, count.
- Main track only: workshops, position-paper tracks (ICML, NeurIPS), the NeurIPS Datasets & Benchmarks track, journal tracks, ICLR blog posts and Tiny Papers are left out. A professor’s page lists such papers under “Not main track”.
- Findings counts as rejected: a paper placed in Findings (ACL, EMNLP, CVPR or any other conference) was not accepted to the main track, so it counts as rejected there, and it is also listed and counted with the workshop papers.
- Rejected adds up papers rejected by reviewers, withdrawn, desk-rejected and placed in Findings. Submitted is accepted plus rejected.
Co-authors and topics
These figures are taken over all of a professor’s counted papers, accepted or not. Rejected papers come from OpenReview, which gives author names but no affiliations, so they enter the first figure only.
- Co-authors per paper: the average number of other authors on a paper.
- Min and max: the fewest co-authors on a single accepted paper, and the most co-authors on a single paper, accepted or not. A minimum of 0 means a paper written alone.
- Foreign co-authors per paper: the average number of other authors whose affiliation on the paper names no Vietnamese institution. An author who lists both a Vietnamese and a foreign institution is not foreign.
- Domestic per paper: the average number of other authors whose affiliation on the paper is an institution in Vietnam, the professor’s own included.
- Same institution per paper: the average number of other authors whose affiliation on the paper is the professor’s own institution.
- Other institutions per paper: the average number of other authors whose affiliation on the paper is another institution, in Vietnam or abroad.
- Topics: the number of different topic labels the conference gave the papers, each an area and a topic within it, for example “Deep Learning / Large Language Models” or “Reinforcement Learning / Planning”. They are the rows of the topic table on the professor’s page.
The two affiliation figures use only papers with an affiliation for each author: from the official list, or, where the list prints none (ACL, EMNLP, AAAI, IJCAI, CVPR before 2023 and in 2026), read from the title page of the paper’s PDF and checked by hand. Papers known only from personal pages are left out of them. Topic labels exist only for ICLR and ICML from 2024 and for NeurIPS from 2023 to 2025. The label sets differ between conferences and years, and the authors choose the label, so two papers on the same subject can carry different labels. Each professor’s page says how many papers a figure is based on.
The chart of submitted and accepted papers
Each professor’s page has a chart of papers per year. Two parts of it are counted: the accepted papers at all ten venues, and the rejections on record (every ICLR rejection, public NeurIPS rejections and papers placed in Findings). The third part is an estimate of the papers that the other ten venues did not accept, which nobody outside can see.
The estimate is made professor by professor, from their own ICLR record:
With 9 ICLR submissions and 1 accepted, that is 10 / 2 = 5, so each paper accepted elsewhere is taken to stand for 5 submissions. Adding 1 on both sides is the usual way to estimate attempts per success from a short record: the result stays finite when nothing was accepted, and one paper more or less does not swing it. A professor with no ICLR submissions gets 1, so the chart adds nothing for them.
This is a projection, not a measurement. It assumes the other venues treated the professor’s papers as ICLR did, and it counts attempts: a paper rejected at one conference and accepted at the next appears twice.
The score
The score on the front page is the adjusted count of CSRankings: each accepted paper is worth 1 divided by its number of authors to every listed professor who is an author of it, and a professor’s score is the sum over their papers. The score of an institution follows CSRankings’ ranking of institutions: the adjusted counts of its professors are added up separately in each of the four areas (AI, ML, CV, NLP), so a paper shared by two professors of the same institution counts for each of them, and the score is the geometric mean over the areas of each count plus 1, the nth root of the product. Adding 1 keeps an area with no papers from setting the whole score to zero, and it rewards institutions with papers in several areas. Official and unofficial papers both count; the author list of an unofficial paper comes from the personal page it was found on and may be incomplete, which makes its share a little high. CSRankings also drops publications of fewer than 6 pages, which is not checked here, and counts only the last 10 years, which here is all of 2020 to 2026. The tables are ordered by the score, then by accepted papers.
Stricter counts on institution pages
Institution and professor pages have two more charts. One marks an accepted paper when more than half of its authors list the institution. The other marks it when the first author does. Affiliations come from the official list or, where it prints none, from the PDF of the paper; papers known only from personal pages have none yet and are shown in grey.
What each venue makes public
| Venue | Accepted papers | Rejected and withdrawn papers |
|---|---|---|
| ICLR | public | public for every submission |
| NeurIPS | public | public only when the authors opted in |
| ICML | public | not public |
| CVPR, ICCV, ECCV | public | not public |
| ACL, EMNLP | public | not public |
| AAAI, IJCAI | public | not public |
Submitted and rejected numbers therefore exist for ICLR only. A professor who sends most papers to ICLR will show more rejections here than one who sends them to ICML or CVPR, whatever their real acceptance rates are. The NeurIPS rejections that are public appear on the professor’s page and on the NeurIPS page, marked as partial, and are included in the rejected column on the front page.
Even ICLR is not complete: a withdrawn paper can be removed or left anonymous, and then it cannot be found.
Sources
- Accepted papers: the conference virtual sites (iclr.cc, neurips.cc, icml.cc, cvpr.thecvf.com, iccv.thecvf.com, eccv.ecva.net), the CVF Open Access lists for CVPR and ICCV, ecva.net for ECCV, the ACL Anthology volumes of long papers for ACL and the main volume for EMNLP (short papers are not counted; the EMNLP volumes mix long and short papers, so each counted paper there is checked by its page count), the technical-track issues of the AAAI proceedings (ojs.aaai.org) and the Main Track of the IJCAI proceedings (ijcai.org).
- ICLR submissions and outcomes, NeurIPS public rejections: OpenReview, queried with each professor’s own profile ID.
- Acceptance rates reported by the venues (Venues page): submitted and accepted main-track papers per year from each conference’s fact sheets, opening slides or proceedings front matter, and a few secondary sources where no official figure was found; each year and its source are in venue_rates.csv. Conferences count submissions differently (all, valid, or after desk rejections), so the rates are only roughly comparable.
Main-track accepted papers read per venue and year:
| Venue | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 | 2026 | Source of the accepted list |
|---|---|---|---|---|---|---|---|---|
| ICLR | 687 | 860 | 1095 | 1574 | 2260 | 3708 | 5353 | iclr.cc |
| NeurIPS | 1899 | 2334 | 2675 | 3218 | 4035 | 5287 | 5284‡ | neurips.cc |
| ICML | 1084 | 1183 | 1233 | 1828 | 2535 | 3258 | 6356 | icml.cc |
| CVPR | 1466† | 1660† | 2074† | 2359 | 2716 | 2871 | 4042† | openaccess.thecvf.com, cvpr.thecvf.com |
| ICCV | – | 1612† | – | 2156† | – | 2701 | – | openaccess.thecvf.com, iccv.thecvf.com |
| ECCV | 1358† | – | 1645† | – | 2387 | – | 2834 | ecva.net, eccv.ecva.net |
| ACL | 778† | 710† | 700† | 1075† | 940† | 1699† | 2296† | aclanthology.org |
| EMNLP | 751† | 847† | 828† | 1047† | 1268† | 1809† | – | aclanthology.org |
| AAAI | 1585† | 1654† | 1319† | 1578† | 2315† | 3028† | 4149† | ojs.aaai.org |
| IJCAI | 592† | 586† | 678† | 639† | 790† | 1014† | 693† | ijcai.org |
† The list gives author names without affiliations; for accepted papers of the professors here, the affiliations are read from the PDF and checked by hand. ‡ The list was published before the conference and is still being filled. In September 2026 the NeurIPS 2026 list lacked papers that their authors had already announced, which is why papers from personal pages are added for it.
Papers from personal pages
The professors’ own pages are read as well, and what they list is kept in self_reported.csv with the address and the date. They are used in three ways.
- A paper on a professor’s own page that is also in an official list counts as theirs, even when the name on the paper is not a form the roster knows.
- When an official list is not complete yet, a paper from a personal page is counted although the list does not have it. This applies to a list published before its conference, which today means NeurIPS 2026. Such a paper is unofficial: it appears as +N after the official number and carries the tag “unofficial”, with the page it comes from, until the conference publishes it. A paper in that list is unofficial too until the professor’s authorship can be confirmed on OpenReview, where the submissions become public after the conference.
- A paper on a colleague’s page that names a listed professor as co-author is treated the same way for that professor when it is not in an official list yet. If it is in an official list, it waits for a yes or no by hand.
For a year whose official list is complete, a paper that only a personal page mentions is not counted. It is usually a workshop paper, another track or another year, and it is written to the build report.
A page that only announces a number (“five papers accepted”) adds nothing to the counts. The professor’s page says so when the number is higher than what could be identified.
How a paper is matched to a professor
- OpenReview records are matched by profile ID, which is exact once the ID has been confirmed.
- In the accepted lists, a paper is matched by author name. It counts automatically only when the affiliation printed on the paper is the professor’s institution.
- If the paper names another Vietnamese institution, or the list has no affiliations, the match waits for a yes or no by hand.
- If the paper names only a foreign institution, the author is treated as a different person with the same name.
Limits
- A paper published under a name form that is not on the roster is missed.
- For ACL, EMNLP, AAAI, IJCAI, and CVPR before 2023 and in 2026, the lists have no affiliations, so a match by name needs a decision by hand; the affiliations are then read from the PDF of the paper.
- The Vietnam rule works by year. A paper that appeared in the year of the move counts, even if it was submitted before the move.
- For ICLR 2020 and 2021, OpenReview keeps the decision in a separate record. Where that record has not been read, the outcome is inferred from the accepted list, and the paper is marked.
- A paper with two listed professors counts once for each of them, and once in the totals of an institution or a venue.
- Counts say nothing about author order, contribution or the quality of a paper.
Updating
The data is rebuilt with make data in the project folder. make openreview fetches the OpenReview records and needs a login. The README lists the main commands.
Visitors
The public site counts visits with GoatCounter, which sets no cookies and keeps no personal data about visitors: only the page, the referring site, the browser, the screen size and the country derived from the address, which is not stored.
Author
This site was built by an AI, Claude, under the direction of Thanh Le (thanhle.xyz). The code and data are public on GitHub. I built this site out of personal curiosity, in the hope that it also serves the public interest: the Vietnamese academic community, and anyone who wishes to follow the current state of AI research in Vietnam. If you find this site useful, please give it a ⭐ on GitHub. Corrections are warmly welcome on the Feedback page.
Acknowledgements
Thanks to Dr. Nam Vo for his comments, and for suggesting that the Venues page separate papers from before 2026 and from 2026 on, to show the numbers before and after AI became really strong.
Thanks to Prof. Huynh Thi Thanh Binh for her comments, for suggesting that the site also track EMNLP, ICCV and ECCV, and for suggesting that it track AAAI and IJCAI and group the conferences as CSRankings does.
Useful links
- CS Picks: helps students choose a university and a PhD advisor in the US. It compares computer science professors and universities by research strengths, publication trends and NSF funding, from open academic data.
- CS conferences: dates and submission deadlines of computer science conferences.
- VietProfs: a directory of Vietnamese professors worldwide, across universities, disciplines, countries and ranks.
These links, made by ThanhVu Nguyen, inspired me to build this site.