ArXiv Endorsement Protocol

I receive many arXiv endorsement requests (and more due to Gen AI). I follow a structured protocol with three steps before endorsing. All three must pass before I can endorse.

Why a Protocol?

ArXiv endorsements carry real scientific weight. An endorsement signals that the endorsed author is a legitimate researcher submitting work appropriate for a given subject area. a bad endorsement hurts arXiv, the scientific community, and ultimately the integrity of science.

At the same time, I want to be helpful to genuine researchers who lack an endorser. The protocol below is designed to be fast, transparent, and fair.

---

Gate 1: Verified LinkedIn Profile

What I need: A link to your LinkedIn profile, and it must be verified (LinkedIn's identity verification badge).

Why: ArXiv endorsement requests can come from anyone. Without identity verification, there is no way to distinguish a genuine researcher from a fake account or a bot. A verified LinkedIn profile provides a basic but meaningful assurance that you are who you say you are. If you do not have a LinkedIn account or have not completed LinkedIn's identity verification, please do so before reaching out.

---

Gate 2: The Paper

What I need: A complete draft of the paper you intend to submit, in PDF form.

Why: I cannot endorse a researcher in a vacuum. I need to skim the paper to assess whether it is a genuine scientific contribution appropriate for arXiv. Please send the full draft, not just an abstract or a title.

---

Gate 3: Automated AI Peer Review

What I need: Nothing extra from you — I run this step myself.

Why: Reading every paper in depth is time-consuming. To ensure consistent and thorough assessment, I pass the paper through an automated AI peer review. This checks for basic scientific soundness, clarity, and appropriateness for the claimed subject area.

This is not a replacement for human peer review and does not guarantee the paper will be accepted at a journal or conference. It is a fast, reproducible filter to catch papers that are clearly not ready or not appropriate for arXiv.

---

How to Request an Endorsement

Submit a pull request to this repository — do not send requests by email. All endorsement requests are handled publicly via pull requests for transparency: anyone can see what was submitted and what decision was made.

To submit a request, open a pull request that adds exactly one new .txt file under requests/.

Use this exact four-line format, with one field per line:

LinkedIn: https://www.linkedin.com/in/your-profile
Paper: https://example.org/your-paper.pdf
Subject: cs.LG
EndorsementCode: A1B2C3

Rules:

1. The file must contain exactly these four fields, in this order: LinkedIn, Paper, Subject, EndorsementCode.
2. Each field must appear exactly once, on a single line, as Field: value.
3. LinkedIn must be an https://www.linkedin.com/in/... or https://www.linkedin.com/pub/... profile URL.
4. Paper must be a public https:// URL to the paper PDF or preprint page.
5. Subject must be a valid arXiv category ID such as cs.LG, stat.ML, or math.OC.
6. EndorsementCode must be the six-character alphanumeric code generated by arXiv for the endorsement request.
7. LinkedIn verification is still checked manually. The automated check only validates that the URL has the right shape.

You can find the current arXiv category taxonomy here: https://arxiv.org/category_taxonomy

arXiv describes the endorsement code here: https://info.arxiv.org/help/endorsement.html

I will run the AI review on my end and post my decision as a comment on the pull request.