Introducing s-1: Sievly’s new document reading model
All posts
Research
Research note: reading order in complex layouts
Multi-column pages, sidebars and callouts confuse simple reading order. We share what we learned teaching s-1 to follow the page the way a person would.
Rafael Ström
·
Machine Learning Engineer
sievly
Reading order
Reading order is easy to ignore until it is wrong. A model that reads a two-column page straight across produces sentences that never existed.
What makes it hard
Magazine layouts, academic papers and financial reports mix columns, sidebars, footnotes and pull quotes. Each needs a different path through the page.
Our approach
s-1 predicts reading order from the layout tree rather than raw coordinates, and treats sidebars and callouts as separate threads.
What is next
We are working on better handling for pages that change layout halfway down, such as a single-column introduction above a two-column body.
More from the blog
sievly
Introducing s-1
Announcements
Introducing s-1, our new document reading model
s-1 reads PDFs, scans and spreadsheets the way a careful analyst does, and returns structured data with a citation for every value. It is available in preview today.
sievly
Quarryline × Sievly
Case Studies
How Quarryline turns research notes into a searchable knowledge base
Quarryline’s data team uses Parse and Extract to turn thousands of analyst notes, filings and spreadsheets into clean, cited records their AI assistant can trust.
sievly
Layout-first chunking
Engineering
Layout-first chunking: why we split by structure, not tokens
Fixed-size chunks cut tables in half and separate headers from their rows. Here is how Sievly decides where one idea ends and the next begins.