TeXpad

Academic Papers

arXiv preprint LaTeX template

A compact technical note: Setup, Main Result with a proof, and Discussion. The plain article class compiles cleanly with arXiv's TeX Live, and the math packages are already loaded.

Open in TeXpad, no signup → Download example PDF

Free · real pdfTeX in your browser · works offline · your files stay on your device

Page 1 of the arXiv Preprint template, compiled with pdfTeX
Page 1, compiled with pdfTeX. This is the actual output.

Who it's for

For a short, math-forward note you want on arXiv quickly. For a full paper, use the research paper template.

How to customize this template

  1. State the main result in the abstract in one sentence, including the bound or rate.
  2. Define notation once in Setup, then use it consistently.
  3. Download the .tex from Export to upload the source to arXiv. arXiv prefers source over PDF.
  4. Keep figures as PDF or PNG next to the .tex. Teleport bundles them for you.

When it looks right, press Download PDF. To take the .tex to Overleaf or a local TeX install, choose Export → LaTeX source.

The first lines of the source

\documentclass{article}
\usepackage{amsmath}
\usepackage{amssymb}

\title{A Note on the Approximation Error of Top-$k$ Attention Routing}
\author{Your Name}
\date{\today}

\begin{document}
\maketitle

\begin{abstract}
We give a short bound on the approximation error introduced by restricting attention to the top-$k$ highest-scoring keys, showing the error decays exponentially in the score gap $\gamma$ between selected and discarded keys.
\end{abstract}

\section{Setup}
Let $q \in \mathbb{R}^d$ be a query and $k_1, \dots, k_n$ the key vectors for a sequence of length $n$. Define attention weights
\[
…

Open the full template in the editor →

Related templates

Free LaTeX tools