Search: “jerry-liu--x”

Agent filesystems are the new RAG It seems like this pattern is around to stay and will only get more robust over time. Agents need tools to not only read and search over documents, but an entire infr

It really isn't an exaggeration! LiteParse clocks in at an average of 3ms per page for a reason: it skips the heavy AI processing and cloud overhead entirely. Here is exactly how it pulls off that kin

We're excited to introduce the Retrieval Harness in LlamaParse - which is the 2026 version of RAG over documents Generalized agents need the right set of tools to scalably search and read through an a

Happy Canada Day! 🇨🇦 AI, ML, RAG, and LLMs are a pretty niche space overall. But, I'm consistently impressed with the talent that Canada ships out, from colleagues to competitors. Gives me the patriot

3 years ago I gave a talk at the first @aiDotEngineer conference on 'Advanced RAG' techniques in order to work around the limitations of naive RAG. It's insane how much the world has changed since the

We've created a comprehensive Retrieval Harness for modern agentic retrieval in 2026. The harness provides a persistent data pipeline that can connect to a data source, index and update a large knowle

Most agentic retrieval demos assume clean, well-structured documents. Enterprise reality is often different, consisting of messy PDFs where critical information is buried across tables, figures, and c

We dropped some new LlamaDrip 🧢 Fear of Docs LlamaParse The team flew into SF for a week onsite. 🌉 2x'd in size since we last did this — first time this many of us have been in the same room. The reca

Simon @disiok proposed a novel IQ test for language models, asking whether an LLM becomes smarter or dumber after being fine‑tuned on a user’s stream of consciousness. The tweet was posted on 9 Aug 20

A downside with VLM-based parsing is that they’re generally slower than text-based heuristic approaches. As a result they add latency to any ad-hoc file processing *in-the agent loop* (e.g. if you upl

Introducing 𝗘𝘅𝘁𝗿𝗮𝗰𝘁𝗕𝗲𝗻𝗰𝗵: the most comprehensive benchmark for information extraction from complex enterprise documents. Our applied research team tested: 14 systems — frontier VLMs, coding agents, ex

Introducing ExtractBench, the most comprehensive benchmark for information extraction from complex enterprise documents. The latest models are pushing the frontier of coding and knowledge work, but su

ExtractBench is one of the most comprehensive benchmarks for real-world document extraction. ✅ It covers 4869 pages, across 67 document types, spanning 8 real-world domains: finance, energy, gov, auto

We wrote a 36-page ArXiv whitepaper on ExtractBench 🧑‍🔬 , our effort to create the most comprehensive, schema-guided, real-world document extraction benchmark. It’s extremely detailed and covers every

The most dangerous document extraction failure isn't a wrong value. It's a missing row that looks like nothing is wrong. We released ExtractBench yesterday: 370 enterprise docs, 14 systems. The hardes

A 36‑page ArXiv whitepaper titled **ExtractBench** was released by Jerry Liu (jerryjliu0), describing a large‑scale, schema‑guided benchmark for real‑world document extraction from complex enterprise

There is a massive opportunity for any startup to build specialized, domain-specific workflows that are at the Pareto frontier in accuracy, cost, and latency. It is clear that not every task requires

If I only went off X posts, I'd think Ramp was an AI lab We’re open-sourcing PorTAL, our framework for shared task representations and cross model LoRA adaptation. It now spans from hybrid attention m

Simon @disiok proposed a novel IQ test for language models, asking whether an LLM becomes smarter or dumber after being fine‑tuned on a user’s stream of consciousness. The tweet was posted on 9 Aug 20

How can you improve your agentic search pipeline? I just wrote a blog post with @tech_optimist from @lancedb to answer exactly that. TLDR: - Parse files and take page-level screenshots with LiteParse,

Common Failure Modes Break VLM-Powered OCR in Production. 🔁 Repetition Loops — model spirals into infinite whitespace, exhausts resources, cascades latency across your system 🛑 Recitation Errors — saf

Our company mission today is to give AI agents the highest-quality document context. The native open-source libs that agents have access to (e.g. PyPDF) do naive text extraction. But this is incomplet

We pit LlamaParse against frontier models (Opus 4.6, Gemini 3.1 Pro, GPT-5.4) in a live OCR arena. ICYMI: the full workshop is on Youtube! Frontier VLMs are getting quite good at visual understanding,

It really isn't an exaggeration! LiteParse clocks in at an average of 3ms per page for a reason: it skips the heavy AI processing and cloud overhead entirely. Here is exactly how it pulls off that kin

Most agentic retrieval demos assume clean, well-structured documents. Enterprise reality is often different, consisting of messy PDFs where critical information is buried across tables, figures, and c

If I only went off X posts, I'd think Ramp was an AI lab We’re open-sourcing PorTAL, our framework for shared task representations and cross model LoRA adaptation. It now spans from hybrid attention m

We're not Palantir, but we do think a lot about evals and hillclimbing w.r.t. document processing. If you have really hairy problems around large-scale extraction over complex, real-world document cor

Common Failure Modes Break VLM-Powered OCR in Production. 🔁 Repetition Loops — model spirals into infinite whitespace, exhausts resources, cascades latency across your system 🛑 Recitation Errors — saf

We pit LlamaParse against frontier models (Opus 4.6, Gemini 3.1 Pro, GPT-5.4) in a live OCR arena. ICYMI: the full workshop is on Youtube! Frontier VLMs are getting quite good at visual understanding,

LiteParse: our open-source, layout-aware PDF parser for AI agents. The secret? Grid projection. Instead of heavy ML layout models or flat text extraction, it projects text onto a monospace grid so ali

Ever wished your agent could read PDFs, images, and Office documents as easily as plain text? Or combine the safety of a secure sandbox with the full power of Bash access? We built exactly that. Meet

Agents + file sandboxes are all in the range in 2026 🤖🗃️ This is a nifty reference implementation by @itsclelia showing you how to run your agent over a collection of docs (PDFs, images, Office) with

This is a nice article (not sure how I stumbled upon it a month later) I directionally agree with it in that: ✅ I have a massive bias for slope, grit, and scrappiness in candidates vs. pure experience

Up until yesterday, our entire MTS team has operated under the philosophy of tokenmaxxing as much as possible on Claude Max plans. With Fable, this may no longer be possible: - One of our team members

Yesterday I cohosted a dinner with @dexhorthy with a wonderful group of founders, to talk about agent loops and loop engineering. Some interesting insights: * Most of our group was *not* actively usin