CORE vs CCF: how computer science venue rankings work

At the University of Chinese Academy of Sciences, one route to a computer science PhD defense requires at least two papers, one of them at CCF B level or an equivalent. In Spain, the national computing society recommends ICORE as the reference ranking for conference papers from 2024 on. Venue rankings decide who graduates and who gets hired, so it is worth knowing who makes them, what the letters mean, and where the two lists that computer scientists quote most often disagree.

Those two lists are CORE, now run internationally as ICORE, and the CCF list from the China Computer Federation. For the flagship conferences most researchers aim at, they mostly agree. The exceptions are more instructive than the agreement.

CORE: an Australian list that went international

CORE began as a research-assessment tool. In late 2005, NICTA and the Australian National University funded work on a ranking of the conferences Australian computing academics attended, at a time when the national research assessment leaned on metrics. The list takes its name from CORE, the Computing Research and Education Association of Australasia, which started it.

In 2024 it became ICORE, managed jointly with two Italian bodies (GII and GRIN) and the Spanish Computing Society (SCIE). The latest round, ICORE2026, was published in January 2026, and the next is expected to run in 2028 with results in 2029.

The four ranks are A*, A, B and C. CORE's own descriptions are short:

• A* is "Exceptional (Flagship conferences)": the top, or equal top, conference for its area.
• A is "Excellent".
• B is "Good to Very Good".
• C is "Sound and Satisfactory".

A* is selective. ICORE2026 ranks 825 venues, and 62 of them, 7.52%, are A*. Nearly half, 46.18%, are C.

Venues apply to be ranked. Panels of five or six senior researchers per area, about 50 academics in all, assess each one on three things: what share of its papers have substantial impact relative to other venues in the same field, measured by citations; how strong the researchers who publish there are, measured by their h-index; and how strong its program committee is. Google Scholar's h5-index feeds in too. The ICORE management committee signs off on the result.

CCF: a recommendation that became a requirement

The China Computer Federation publishes the 中国计算机学会推荐国际学术会议和期刊目录, its Recommended List of International Academic Conferences and Periodicals. The seventh edition came out on 31 March 2026 and replaced the 2022 edition. The committee reviewed 169 proposals and added 23 conferences and 14 journals.

Each venue gets a class, A, B or C, within one of ten areas: computer architecture, networks, security, software engineering and systems, databases and data mining, theory, graphics and multimedia, artificial intelligence, human-computer interaction, and an interdisciplinary area.

The list carries more weight than its name suggests. An ACM SIGMM report observed back in 2013 that it is "typically consulted by most academic institutions in China as a quality metric for PhD promotions and tenure track jobs", and graduation rules like the UCAS one above name its classes directly. CCF itself says the opposite should happen: its announcement calls the list a recommendation and advises institutions not to use it as a simple basis for academic evaluation.

Two changes in the 2026 edition stand out for machine learning researchers. ICLR, absent from the 2022 list, entered directly at A. IJCAI dropped from A to B.

Where the two lists agree, and where they do not

Lune indexes 23 conferences. Here is each one in both lists, from the ICORE2026 portal and the official CCF 2026 PDF.

Twenty-one of the 23 sit at the top of both lists. Two do not, and they disagree in opposite directions.

EMNLP is A* in CORE and B in CCF, the only venue here that CORE puts in its top tier and CCF does not. CORE moved it from A to A* in 2023, while CCF kept it at B in 2026, below ACL. An EMNLP main-conference paper can therefore count as top-tier in Sydney or Madrid and as CCF B, the second tier, in Beijing.

ISSTA runs the other way: A in CCF, its top class, but A in CORE, which is CORE's second tier. That is also the trap in the letters themselves. "A-ranked" on a CV means top class under CCF and one step below the top under CORE, so always say which list, and which edition.

ICLR shows the other weakness, which is timing. CORE has ranked it A* since 2021, yet the 2022 CCF list did not include it at all. A list revised every few years can lag a field that moves quickly.

Other signals worth knowing

• Google Scholar Metrics reports each venue's h5-index, the h-index of its papers from the last five complete years, currently 2020 to 2024. It is easy to check, and because it is an h-index it favors large venues.
• CSRankings ranks institutions, not venues, by counting papers at a short, hand-picked list of top conferences per area. The list is still a useful signal, because its rule is that the venues in an area must be "roughly equivalent in terms of number of submissions, selectivity and impact".
• The GII-GRIN-SCIE rating grades conferences from A++ to C by combining CORE with bibliometric data. Its last edition is from 2021, and SCIE now points researchers to ICORE instead.

Use a rank without misusing it

Every organization behind these lists warns against the use they are most often put to. The San Francisco Declaration on Research Assessment asks institutions not to use journal-based metrics "as a surrogate measure of the quality of individual research articles". CORE's own guidance says "the quality of the venue is only a proxy for the quality of an individual paper", and lists judging hiring or promotion "solely on counts of papers in top-ranked venues" under "What Not to Do". CCF says the influence of a venue is not directly tied to the influence of any single paper in it.

Rankings are good at two jobs: deciding where to submit, and deciding what to read first. For judging a paper, read the paper.

Why Lune indexes only ranked venues

An AI agent's answer is only as good as the sources it retrieves, and it cannot peer-review what it reads. So Lune chooses its venues from the top of both lists: every one of the 23 is ranked by CORE and by CCF, and all but one are A* in CORE. A venue's rank still says nothing certain about any single paper, which is why Lune also returns the full text and the exact passages where it has them, so you can judge each paper on what it says.

The same choice has a cost. Journals, workshops, preprints and fast-rising new venues are outside the corpus, and an agent should say so rather than pretend otherwise. You can browse every indexed venue, with its papers, on the venues page.

Sources

• ICORE, conference rankings portal, ICORE2026 round, and its description of conference ranks
• ICORE, details of data used in CORE rankings, revised June 2025
• ICORE, history of CORE rankings and advice on use of the CORE conference rankings
• SCIE, GGS and ICORE conference classification
• China Computer Federation, recommended list of international academic conferences and periodicals, seventh edition (2026), PDF
• ACM SIGMM Records, CCF ranking of conferences (2013)
• Google Scholar, Scholar Metrics help
• CSRankings, FAQ
• DORA, San Francisco Declaration on Research Assessment