CS Conference Acceptance Rates: What 64 Venues Actually Show
Acceptance rate is the number everyone quotes and almost nobody contextualises. It goes in job talks, on CVs, in the first line of grant reports. It is also, taken on its own, close to meaningless.
Here are the figures we hold for 64 major computer science conferences, and the pattern that shows up once you put them side by side.
The short version
Rates across our data run from 14.0% to 50.0%. Both ends are CORE A* or A venues. A*-ranked ICRA accepts 43% of submissions; A*-ranked SIGCOMM accepts 14%. Neither is a mistake.
Your subfield sets the number far more than the venue’s prestige does. That is the single most useful thing to know before you compare two rates, or before someone compares yours.
The most selective venues
| Conference | Rate | Rank | Field |
|---|---|---|---|
| SIGCOMM | 14.0% | A* | Networking |
| IJCAI | 14.0% | A* | AI |
| CCS | 14.5% | A* | Security |
| NSDI | 14.7% | A* | Networking |
| IEEE S&P | 14.8% | A* | Security |
| OSDI | 15.6% | A* | Systems |
| SOSP | 16.0% | A* | Systems |
| NDSS | 16.1% | A* | Security |
| EuroSys | 16.5% | A | Systems |
| USENIX Security | 17.0% | A | Security |
Systems, networking and security take the whole top ten. That is not a coincidence, and it is the clue to everything below.
Rates cluster by subfield, not by rank
Group the same data by area and the structure is obvious.
Systems and networking: 14 to 22% SIGCOMM 14.0, NSDI 14.7, OSDI 15.6, SOSP 16.0, EuroSys 16.5, MobiCom 18.0, SenSys 19.0, ISCA 19.6, MICRO 22.4.
Security and privacy: 14 to 26% CCS 14.5, IEEE S&P 14.8, NDSS 16.1, USENIX Security 17.0, SOUPS 19.1, ACSAC 20.7, RAID 25.0, PETS 26.0.
Databases and data mining: 17 to 20% WSDM 17.3, CIKM 19.0, VLDB 19.0, SIGMOD 19.0, ICDM 20.0, KDD 20.0, SIGIR 20.0. A remarkably tight band.
Software engineering: 21 to 23% FSE 21.0, ASE 22.0, ICSE 22.3.
Machine learning: 26 to 32% NeurIPS 25.8, AISTATS 27.0, ICML 27.5, UAI 30.7, ICLR 32.1.
Computer vision: 24 to 28% CVPR 23.6, ICCV 26.2, ECCV 27.9.
Theory: 25 to 30% FOCS 25.1, SODA 29.3, STOC 29.8.
Cryptography: 21 to 29% Eurocrypt 21.0, CRYPTO 24.3, Asiacrypt 29.3.
HCI: 22 to 25% UIST 22.0, CHI 24.9, CSCW 25.0.
Robotics and speech: 24 to 50% HRI 24.0, ICRA 43.0, IROS 47.5, Interspeech 50.0.
What this actually means
NeurIPS is less selective than CCS. 25.8% against 14.5%. Nobody sensible thinks a NeurIPS paper is worth less than a CCS paper, which tells you the rate is measuring something other than quality.
What it mostly measures is submission culture. Machine learning venues receive enormous volumes and have expanded to absorb them: ICLR reviewed over eleven thousand submissions in its 2025 edition. Systems conferences stayed small, single-track and deliberately brutal. Robotics venues like ICRA and IROS function partly as community gatherings where presenting is the point, so a 43% rate is a feature of the format rather than a lack of standards.
Comparing across subfields is the mistake. A 24% acceptance in robotics and a 24% acceptance in systems are not the same achievement, and a hiring committee that treats them identically is reading the wrong signal. Compare within a field, or compare against the venue’s own history.
Rank and rate are close to independent. Our A* venues span 14.0% to 43.0%. If you want a sense of standing, CORE rank tells you more than the percentage does, and neither is a substitute for knowing the field.
Caveats worth stating
These figures are drawn from different years, mostly 2024 and 2025, and we record the year alongside each. Rates move: a venue that jumped 4 points last cycle may move back this one.
They are also self-reported and inconsistently defined. Some conferences count full research-track papers only; others fold in short papers, industry tracks or posters, which can shift a headline number by several points. Multi-cycle venues like VLDB and EuroSys report differently again. Treat a single decimal place as false precision.
Finally, this is 64 venues out of the 348 editions we track, weighted toward established CORE-ranked conferences. It is not everything, and the venues that do not publish a rate are missing for a reason rather than at random.
Using it
If you are choosing where to submit, use the rate as one input among several, well behind fit and behind whether the community you want to reach actually attends. A 27% venue in your area beats a 15% venue that will not read your paper properly.
If you are writing a CV or a report, quote the rate with the year and the venue’s field. “22.3% (ICSE 2024)” is informative. “22.3%” alone invites the reader to compare it against a number from a different discipline.
And if you are early in a PhD deciding where to aim, the useful takeaway is that the top of your field is probably harder or easier than the headline numbers from another field suggest. Find out what normal looks like where you actually work.
We keep these figures alongside dates, venues and deadlines for every conference we track, and update them as new editions publish. If you spot one that is wrong or out of date, tell us.