Analyzing the Wait: A Study of arXiv Submission Delays

Date: July 9, 2026
Source: fi-le.net

"Frustrated by my arXiv submission being moved into the 'on hold' queue, I did what any reasonable person would do in this situation: an analysis of arXiv submission delay times since 2015."

The Catalyst

Recently, I uploaded a preprint to arXiv. Based on my previous experiences, I expected the paper to be live by the next day. However, several days passed without an email notification. Upon checking the status, I discovered my submission was marked as "on hold."

I initially brainstormed a few potential reasons for this:

• Recent moderation policy shifts (e.g., the requirement that CS position papers or literature reviews be peer-reviewed before indexing).
• A general surge in "fake vibe-researched" submissions clogging the moderation pipeline.

Since the policy changes didn't actually apply to my specific paper, I suspected the latter. To verify if there was a systemic increase in wait times, I sought out data on the average duration between submission and announcement/indexing.

The Data Hunt

I discovered that neither the official arXiv statistics dashboard nor any third-party tools provided this specific metric. Consequently, I decided to scrape the data myself.

The Technical Challenge

I needed two specific data points: the submission date and the announcement date, segmented by time and category.

1. Submission Date: Easily accessible via the search feature.
2. Announcement Date: This is the tricky part. While search results can be sorted by this date, the date itself isn't explicitly displayed in the results.
3. Temporal Limits:
   • /new and /recent pages only cover a few days.
   • The /catchup page extends this to 90 days.

The Solution: I utilized the Internet Archive to access snapshots of announcement pages every three months, then cross-referenced these with the actual paper pages on arXiv to extract the submission dates.

The Calculation

To determine the delay, I used the following logic:

$\text{Delay} = \text{Announcement Date} - \text{Submission Date}$

Note: I excluded weekends and the specific 2026 arXiv holiday schedule to calculate the delay in working days.

# Conceptual logic for delay calculation
def calculate_working_days(sub_date, ann_date, holidays_2026):
    # Subtract weekends and specific holiday list
    return net_working_days(sub_date, ann_date, holidays_2026)

Findings and Analysis

Distribution of Delays

When plotting the working-day delays in a histogram, the results were striking. The distribution follows a $\text{zeta distribution}$ (or a power law).

Observation: On log-log axes, the height of the bars declines roughly linearly.

Temporal Trends

I aggregated the data by quarter for older samples and by week for the most recent 90 days. The results debunked my initial theory:

• There has been no recent increase in overall delays.
• There has never been a sustained increase in delay times over the sampled period.

Therefore, my "on hold" status was not part of a widespread systemic slowdown.

The Role of Categorization

To understand the nuance, I looked at how arXiv organizes papers. The hierarchy is structured as follows:

While most groups (excluding physics) have only one archive, there are 149 unique subject categories (plus 6 aliases).

The "High-Delay" Zones

After analyzing the mean (dot), median (triangle), and the 0.1/0.9 quantiles (whiskers) for each subject, I realized my paper fell under cs.CY (Computers and Society).

As it turns out, cs.CY has some of the longest delay times. This is likely because many authors—myself included—are unfamiliar with the specific categories and misclassify their work. Other "slow" categories include:

• cs.OH (Other Computer Science)
• math.GM (General Mathematics)
• physics.gen-ph (General Physics)
• physics.pop-ph (Popular Physics)

In these cases, moderators likely spend more time reading the papers to determine if they should be reclassified into a more specific subject.

Summary Data Table

If you are submitting to a new category, be aware that the announcement time may differ from what you expect.