🤏 DeepSeek-OCR: Contexts Optical Compression

episode

Build Wiz AI Show not yet transcribed

▲ 0

Transcript

This episode hasn't been transcribed yet Upvotes decide what gets transcribed next — 0 so far. Sign in to upvote

Related

recent transcribed audio

Finance and Budget Committee - Sep 15, 2026 1h 43m · 15 Sep 2026 · Prince William County Finance and Budget Committee

The Art of True Happiness | Dr. Arthur Brooks 2h 49m · 14 Sep 2026 · Huberman Lab

The Chris Wolfe Podcast: The Power of Quality Supplements 48 min · 14 Sep 2026 · EyeCode Media

1208: TM Europe win the '26 Solheim Cup 1h 49m · 13 Sep 2026 · No Laying Up - Golf Podcast

MORNING TEA: Harry & Meghan's Brutal SNL Roast, Lady Gaga's Baby News & Dakota Johnson's New Man 8 min · 13 Sep 2026 · The Spill

"I don't think they have enough match winners."| First Thoughts - Brisbane vs Adelaide (14.09.26) 15 min · 13 Sep 2026 · SEN Breakfast

Report any issue

Source

Recorded21 Oct 2025

Publisheranchor.fm ↗

Handlerec:12512874

Show notes

Welcome to the show, where we discuss DeepSeek-OCR and its investigation into using optical 2D mapping for contexts compression, addressing the computational challenges of quadratic scaling faced by Large Language Models. We explore the DeepEncoder, the core engine designed to achieve high compression ratios, delivering near-lossless OCR precision (approximately 97%) even at a 10× token reduction. This groundbreaking work demonstrates strong practical value, achieving state-of-the-art document parsing performance on OmniDocBench while using the fewest vision tokens, offering a promising direction for future memory systems.

Ask this episode

“What did the council decide on item 42?” — answered with the speaker and the moment.

Sign in