← The rising topics of the week

In: All of arXiv. One to three words, as in a title; plurals count with their singular. Premium compares three terms, in any field or category, and exports the counts. Discover Premium

"unstructured mesh" is not clearly rising. 3.9 papers per 10,000 over the last 13 weeks (42 papers), against 2.3 over the 13 weeks before (21 papers). The difference is too small, or too uncertain, to say it rises. A year before, over the same weeks: 2.0 per 10,000.

Papers using the term per 10,000 papers of the week, averaged over 4 weeks
0246Apr 2025Jun 2025Aug 2025Oct 2025Dec 2025Feb 2026Apr 2026Jun 2026Aug 2026Oct 2026

Left axis: out of 10,000 papers submitted in a week, how many use the term. Each point averages that week with the three before, as the count of a single week jumps with holidays and conference deadlines. Hover over the curve to read each week and its number of papers.

"unstructured mesh": 113 papers in 80 weeks. Counted since the week of 7 April 2025.

The weekly values
Week ofunstructured mesh
12 Oct 202612 papers (12.6 per 10,000)
5 Oct 20262 papers (2.0 per 10,000)
28 Sep 20262 papers (1.7 per 10,000)
21 Sep 20264 papers (4.2 per 10,000)
14 Sep 20260 papers (0.0 per 10,000)
7 Sep 20262 papers (2.5 per 10,000)
31 Aug 20263 papers (3.8 per 10,000)
24 Aug 20263 papers (4.0 per 10,000)
17 Aug 20263 papers (4.5 per 10,000)
10 Aug 20262 papers (2.9 per 10,000)
3 Aug 20263 papers (4.0 per 10,000)
27 Jul 20263 papers (4.0 per 10,000)
20 Jul 20263 papers (4.7 per 10,000)
13 Jul 20260 papers (0.0 per 10,000)
6 Jul 20263 papers (4.6 per 10,000)
29 Jun 20264 papers (5.6 per 10,000)
22 Jun 20260 papers (0.0 per 10,000)
15 Jun 20260 papers (0.0 per 10,000)
8 Jun 20263 papers (4.2 per 10,000)
1 Jun 20260 papers (0.0 per 10,000)
25 May 20262 papers (2.5 per 10,000)
18 May 20260 papers (0.0 per 10,000)
11 May 20260 papers (0.0 per 10,000)
4 May 20264 papers (5.1 per 10,000)
27 Apr 20262 papers (3.1 per 10,000)
20 Apr 20263 papers (4.8 per 10,000)
13 Apr 20264 papers (6.0 per 10,000)
6 Apr 20260 papers (0.0 per 10,000)
30 Mar 20263 papers (4.6 per 10,000)
23 Mar 20262 papers (3.1 per 10,000)
16 Mar 20262 papers (3.0 per 10,000)
9 Mar 20260 papers (0.0 per 10,000)
2 Mar 20260 papers (0.0 per 10,000)
23 Feb 20262 papers (3.2 per 10,000)
16 Feb 20260 papers (0.0 per 10,000)
9 Feb 20260 papers (0.0 per 10,000)
2 Feb 20262 papers (3.0 per 10,000)
26 Jan 20260 papers (0.0 per 10,000)
19 Jan 20262 papers (3.6 per 10,000)
12 Jan 20260 papers (0.0 per 10,000)
5 Jan 20260 papers (0.0 per 10,000)
29 Dec 20253 papers (8.1 per 10,000)
22 Dec 20250 papers (0.0 per 10,000)
15 Dec 20252 papers (3.3 per 10,000)
8 Dec 20252 papers (3.4 per 10,000)
1 Dec 20250 papers (0.0 per 10,000)
24 Nov 20250 papers (0.0 per 10,000)
17 Nov 20250 papers (0.0 per 10,000)
10 Nov 20250 papers (0.0 per 10,000)
3 Nov 20250 papers (0.0 per 10,000)
27 Oct 20252 papers (3.3 per 10,000)
20 Oct 20252 papers (3.5 per 10,000)
13 Oct 20252 papers (3.3 per 10,000)
6 Oct 20250 papers (0.0 per 10,000)
29 Sep 20253 papers (4.4 per 10,000)
22 Sep 20252 papers (2.9 per 10,000)
15 Sep 20250 papers (0.0 per 10,000)
8 Sep 20252 papers (3.8 per 10,000)
1 Sep 20250 papers (0.0 per 10,000)
25 Aug 20250 papers (0.0 per 10,000)
18 Aug 20250 papers (0.0 per 10,000)
11 Aug 20252 papers (4.0 per 10,000)
4 Aug 20250 papers (0.0 per 10,000)
28 Jul 20252 papers (3.8 per 10,000)
21 Jul 20252 papers (3.7 per 10,000)
14 Jul 20252 papers (3.8 per 10,000)
7 Jul 20250 papers (0.0 per 10,000)
30 Jun 20250 papers (0.0 per 10,000)
23 Jun 20250 papers (0.0 per 10,000)
16 Jun 20250 papers (0.0 per 10,000)
9 Jun 20250 papers (0.0 per 10,000)
2 Jun 20253 papers (5.0 per 10,000)
26 May 20250 papers (0.0 per 10,000)
19 May 20250 papers (0.0 per 10,000)
12 May 20250 papers (0.0 per 10,000)
5 May 20250 papers (0.0 per 10,000)
28 Apr 20250 papers (0.0 per 10,000)
21 Apr 20250 papers (0.0 per 10,000)
14 Apr 20250 papers (0.0 per 10,000)
7 Apr 20252 papers (3.9 per 10,000)

Who works on it? Literature review Search and follow the term

How it works

  1. Every week, we read the title and abstract of every new paper on arXiv (first versions only, not updates) and note the terms of one to three words they use.
  2. For your term, we count the papers using it each week, and divide by all the papers of that week. A share rather than a number: arXiv grows every year, so the number of papers would rise for almost every topic, while the share rises only when the topic takes a larger place.
  3. The answer at the top compares the last 13 weeks with the 13 weeks before. The topic "rises" only when its share is at least ×1.2 and the gap is too large to come from chance, a test of the same kind as for the rising topics of the week; it "declines" in the same way the other way round.

How to read it

  • A curve that climbs over several months: more and more teams write about the topic. A peak that falls back after a few weeks: an event, a release, a data set or a conference deadline.
  • The height matters less than the trend: a broad term (graph neural network) is always higher than a precise one (equivariant message passing). To compare two topics, compare how their curves move.
  • The terms are counted as words, in English, plurals with their singular. A term with several meanings counts them all (transformer, in electrical engineering and in AI): then try a more specific term. A week where a single paper uses the term counts as zero.
  • The curve says how much the topic is written about, not whether the papers are good or cited, nor what is published outside arXiv.

Why it is useful

  • Check an impression: "everyone is working on it" — true, or only in the papers you happen to see?
  • Choose a thesis or project subject: a rising topic draws funding and competition; a declining one may be maturing, or being left aside.
  • Back up a "growing interest" in a grant proposal or an introduction with a number. When citing it, name the source and the period.
  • Compare competing methods or instruments week by week, to see which one is taking over. Premium

The rising topics of the week work the other way round: instead of checking a term you chose, they find by themselves the terms that jumped this week. The rising topics of the week