Extract scientific papers, journal publications, and conference proceedings into clean Markdown with intact LaTeX math, 2-column de-interleaving, and citations.
100% In-Browser Conversion — Your Files Are Never Uploaded
Drag & Drop your PDF document here
or browse files on your device
See how our parser transforms an unstructured arXiv PDF into a beautifully formatted, KaTeX-ready Markdown document:
Attention Is All You Need Ashish Vaswani Noam Shazeer Niki Parmar Jakob Uszkoreit Llion Jones Aidan N. Gomez Lukasz Kaiser Illia Polosukhin
Abstract
The dominant sequence transduction models are based on complex recurrent or convolutional neural networks...
1 Introduction
Recurrent neural networks, long short-term memory [12] and gated recurrent [7] neural networks...# Attention Is All You Need
**Authors:** Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, Illia Polosukhin
**Published:** NeurIPS 2017
## Abstract
The dominant sequence transduction models are based on complex recurrent or convolutional neural networks. We propose the **Transformer**, a model architecture relying entirely on an attention mechanism to draw global dependencies between input and output.
### Scaled Dot-Product Attention Equation
$$
\text{Attention}(Q, K, V) = \text{softmax}\left(\frac{QK^T}{\sqrt{d_k}}\right)V
$$
## 1. Introduction
Recurrent neural networks, long short-term memory [12] and gated recurrent [7] architectures have been firmly established as state of the art...Academic papers represent the most complex document format in existence: combining two-column body text, full-width title blocks, embedded LaTeX equations, vector figures, and extensive bibliographies. Our spatial parser handles each component with specialized mathematical heuristics.
Everything you need to know about format extraction, privacy, and markdown compatibility.
Yes! Our parser incorporates a spatial column de-interleaving algorithm that separates Column 1 from Column 2, preventing sentences from crossing column gutters and preserving the true reading order of research papers.
Yes. Symbol font characters and vector fractions are translated into clean LaTeX math delimiters (`$...$` for inline, `$$...$$` for display equation blocks).
Numbered citations (e.g., `[1]`, `[12]`) and bibliography entries are structured as standard CommonMark footnote references and lists.
Yes. The generated Markdown includes YAML metadata and KaTeX blocks compatible with Obsidian vaults, Logseq graphs, and Zotero markdown export workflows.
Explore dedicated documentation for other document formats and developer pipelines.
Extract advanced calculus, Greek symbols, and matrices into KaTeX syntax.
Learn how the spatial parser prevents left and right columns from mixing.
Import arXiv papers directly into your personal knowledge management vault.