Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

We’re excited to officially launch LlamaParse, the first genAI-native document parsing solution. Not only is it better at parsing out images/tables/charts 📊📈 than virtually every other parser, it is now steerable through natural language instructions - output the document in whatever format you desire! It is also the only...

llama_index's profile picture

LlamaIndex 🦙

119,080 subscribers

143,178 Aufrufe • vor 2 Jahren •via X (Twitter)

Wissenschaft & Technologie
LIVE

Anya Rossi• Live Now

Private livecam show

10 Kommentare

Profilbild von Manish Sinha
Manish Sinhavor 2 Jahren

Charts ( bar graph) and images ( of math formulae’s) don’t work still. Text and tables work well ( especially tables accuracy and markdown format). Will test more elaborately later.

Profilbild von DataStax
DataStaxvor 2 Jahren

Congrats @llama_index team! LlamaParse is 🔥🔥🔥

Profilbild von matthew muller
matthew mullervor 2 Jahren

Is there a data privacy statement for llamacloud? I cant seem to find that... thx!

Profilbild von Duc Nguyen Huu
Duc Nguyen Huuvor 2 Jahren

I have a question on information security when using llama-parse. For example, I have pdf files and I want to use llama-parse for parsing. My concern is whether utilizing llama-parse through API access poses any potential risks of information leakage.

Profilbild von Andrew Batutin ☸️/acc
Andrew Batutin ☸️/accvor 2 Jahren

Is this price on top of Claude 3 pricing? Or it’s for all including llm calls?

Profilbild von LlamaIndex 🦙
LlamaIndex 🦙vor 2 Jahren

That's the full price!

Profilbild von nohrt
nohrtvor 2 Jahren

Great stuff!! good job :) Now if you could fix the issue of auto rotating pages i would be able to use llama parse :P

Profilbild von RM
RMvor 2 Jahren

Here's my article. Please upvote!

Profilbild von Art Intelligence
Art Intelligencevor 2 Jahren

It doesn’t seem to work well with chinese documents and papers. It is also lacking the feature of translating text.

Profilbild von LLL
LLLvor 2 Jahren

mathpix is excellent at parsing academic pdfs, did you have a comparison with them when you mentioned "better virtually every other parsers"?

Ähnliche Videos

An underrated issue with document parsing for RAG / agent use cases is dealing with multi-page tables - sometimes a big table spills over into multiple pages. This breaks chunking algorithms that generally operate at the page-level or smaller, and causes LLMs to lose the full view of the data. With LlamaParse Continuous Mode (in beta), you can now parse a document with multi-page tables and join them into a single table! This means you can now: 💡 Do contiguous chunking for RAG use cases OR 💡 Parse the table for text-to-SQL Check out our blog post highlighting this feature. Huge shoutout to Pierre-Loic Doulcet and Sacha Bron : Signup here: It's in beta, let us know your feedback!
2:22

An underrated issue with document parsing for RAG / agent use cases is dealing with multi-page tables - sometimes a big table spills over into multiple pages. This breaks chunking algorithms that generally operate at the page-level or smaller, and causes LLMs to lose the full view of the data. With LlamaParse Continuous Mode (in beta), you can now parse a document with multi-page tables and join them into a single table! This means you can now: 💡 Do contiguous chunking for RAG use cases OR 💡 Parse the table for text-to-SQL Check out our blog post highlighting this feature. Huge shoutout to Pierre-Loic Doulcet and Sacha Bron : Signup here: It's in beta, let us know your feedback!

Jerry Liu

24,245 Aufrufe • vor 1 Jahr

Introducing LlamaCloud 🦙🌤️ Today we’re thrilled to introduce LlamaCloud, a managed service designed to bring production-grade data for your LLM and RAG app. Spend less time data wrangling and more time on application logic. Launching with the following components: 1️⃣ LlamaParse 📑: a proprietary parser designed to be really really good at complex documents with embedded tables. Build advanced RAG over semi-structured PDFs, and ask questions that simply aren’t possible with the naive stack. Available publicly day 1 🔥 2️⃣ Managed Ingestion/Retrieval API ⚙️: An API letting you easily ingest/retrieve data from data sources. Opening up in private beta to select enterprises. We’re excited to be joined by launch users, partners, and collaborators: Mendable @DataStax MongoDB Qdrant NVIDIA + some awesome hackathon projects at the LlamaIndex 🦙 hackathon Check out our FULL blog post on LlamaCloud and LlamaParse: LlamaParse Client Repo: Signup for a LlamaCloud account to use LlamaParse: Interested in the broader LlamaCloud offering? Come talk to us: Also we have a slick new website 🌐:
2:59

Introducing LlamaCloud 🦙🌤️ Today we’re thrilled to introduce LlamaCloud, a managed service designed to bring production-grade data for your LLM and RAG app. Spend less time data wrangling and more time on application logic. Launching with the following components: 1️⃣ LlamaParse 📑: a proprietary parser designed to be really really good at complex documents with embedded tables. Build advanced RAG over semi-structured PDFs, and ask questions that simply aren’t possible with the naive stack. Available publicly day 1 🔥 2️⃣ Managed Ingestion/Retrieval API ⚙️: An API letting you easily ingest/retrieve data from data sources. Opening up in private beta to select enterprises. We’re excited to be joined by launch users, partners, and collaborators: Mendable @DataStax MongoDB Qdrant NVIDIA + some awesome hackathon projects at the LlamaIndex 🦙 hackathon Check out our FULL blog post on LlamaCloud and LlamaParse: LlamaParse Client Repo: Signup for a LlamaCloud account to use LlamaParse: Interested in the broader LlamaCloud offering? Come talk to us: Also we have a slick new website 🌐:

LlamaIndex 🦙

141,258 Aufrufe • vor 2 Jahren

LiteParse Samples 📄🧑‍🏫 LiteParse is our free and fast document parser that can parse text with bounding boxes from any document. I’ve created a new repository that contains some demos on how you can make use of its outputs! ✅ Comparison against PyPDF and PyMuPDF ✅ Visual Citations: search for any keyword and see joined bounding boxes pop up in the source images It should be a default tool to help AI agents OCR any document type (supports PDF, Word, Powerpoint + dozens of other formats), before heavier-weight VLM-based document parsing solutions (like our own LlamaParse). LiteParse Samples repo: LiteParse:
0:33

LiteParse Samples 📄🧑‍🏫 LiteParse is our free and fast document parser that can parse text with bounding boxes from any document. I’ve created a new repository that contains some demos on how you can make use of its outputs! ✅ Comparison against PyPDF and PyMuPDF ✅ Visual Citations: search for any keyword and see joined bounding boxes pop up in the source images It should be a default tool to help AI agents OCR any document type (supports PDF, Word, Powerpoint + dozens of other formats), before heavier-weight VLM-based document parsing solutions (like our own LlamaParse). LiteParse Samples repo: LiteParse:

Jerry Liu

13,196 Aufrufe • vor 4 Monaten

Introducing LiteParse - the best model-free document parsing tool for AI agents 💫 ✅ It’s completely open-source and free. ✅ No GPU required, will process ~500 pages in 2 seconds on commodity hardware ✅ More accurate than PyPDF, PyMuPDF, Markdown. Also way more readable - see below for how we parse tables!! ✅ Supports 50+ file formats, from PDFs to Office docs to images ✅ Is designed to plug and play with Claude Code, OpenClaw, and any other AI agent with a one-line skills install. Supports native screenshotting capabilities. We spent years building up LlamaParse by orchestrating state-of-the-art VLMs over the most complex documents. Along the way we realized that you could get quite far on most docs through fast and cheap text parsing. Take a look at the video below. For really complex tables within PDFs, we output them in a spatial grid that’s both AI and human-interpretable. Any other free/light parser light PyPDF will destroy the representation of this table and output a sequential list. This is not a replacement for a VLM-based OCR tool (it requires 0 GPUs and doesn’t use models), but it is shocking how good it is to parse most documents. Huge shoutout to Logan Markewich and Clelia Bertelli (🦀/acc) for all the work here. Come check it out: Repo:
0:31

Introducing LiteParse - the best model-free document parsing tool for AI agents 💫 ✅ It’s completely open-source and free. ✅ No GPU required, will process ~500 pages in 2 seconds on commodity hardware ✅ More accurate than PyPDF, PyMuPDF, Markdown. Also way more readable - see below for how we parse tables!! ✅ Supports 50+ file formats, from PDFs to Office docs to images ✅ Is designed to plug and play with Claude Code, OpenClaw, and any other AI agent with a one-line skills install. Supports native screenshotting capabilities. We spent years building up LlamaParse by orchestrating state-of-the-art VLMs over the most complex documents. Along the way we realized that you could get quite far on most docs through fast and cheap text parsing. Take a look at the video below. For really complex tables within PDFs, we output them in a spatial grid that’s both AI and human-interpretable. Any other free/light parser light PyPDF will destroy the representation of this table and output a sequential list. This is not a replacement for a VLM-based OCR tool (it requires 0 GPUs and doesn’t use models), but it is shocking how good it is to parse most documents. Huge shoutout to Logan Markewich and Clelia Bertelli (🦀/acc) for all the work here. Come check it out: Repo:

Jerry Liu

257,050 Aufrufe • vor 5 Monaten

We built an AI agent that lets you vibe-code document extraction - high accuracy and citations over the most complex documents. Our latest release lets you upload documents as context. All you then have to do is describe what you want extracted in natural language. 💡 Our agent will then read the document with file tools to infer the right schema, validation rules, and other pre/postprocessing logic. ✅ It will give you back a workflow that can extract over thousands/millions of documents at scale. You can still of course review and edit every output before approving. Stop handling paperwork manually; just upload files, describe your task, and let our agent handle the rest. Our vision for LlamaAgents is to provide the most advanced and easy-to-use way for you to orchestrate document work. Walkthrough: Check it out: If you’re interested in reducing the operational burden of document extraction (invoices, claims, onboarding forms), come talk to us!
0:58

We built an AI agent that lets you vibe-code document extraction - high accuracy and citations over the most complex documents. Our latest release lets you upload documents as context. All you then have to do is describe what you want extracted in natural language. 💡 Our agent will then read the document with file tools to infer the right schema, validation rules, and other pre/postprocessing logic. ✅ It will give you back a workflow that can extract over thousands/millions of documents at scale. You can still of course review and edit every output before approving. Stop handling paperwork manually; just upload files, describe your task, and let our agent handle the rest. Our vision for LlamaAgents is to provide the most advanced and easy-to-use way for you to orchestrate document work. Walkthrough: Check it out: If you’re interested in reducing the operational burden of document extraction (invoices, claims, onboarding forms), come talk to us!

Jerry Liu

20,857 Aufrufe • vor 6 Monaten

Introducing NotebookLlama - an open-source version of NotebookLM! 📓🦙 NotebookLlama is a full implementation of NotebookLM that includes all the capabilities that makes it so great for researchers+business users: ✅ Create a knowledge repository of documents. Has likely higher accuracy than NotebookLM since it’s using LlamaCloud under the hood for high-quality parsing/extraction over complex docs. ✅ Generate summaries and knowledge graph mind-maps 🤯 ✅ Generate podcasts thanks to ElevenLabs 🗣️ ✅ Agentic chat with docs and view metrics with OpenTelemetry This all lives within an open-source repo so you can clone/modify at will to swap in your own components! Huge shoutout to Clelia Bertelli (🦙/acc) for leading this. Repo: LlamaCloud helps power the parsing/ingestion. You can always use your own stuff too, but in the meantime you can check out LlamaCloud here:
0:56

Introducing NotebookLlama - an open-source version of NotebookLM! 📓🦙 NotebookLlama is a full implementation of NotebookLM that includes all the capabilities that makes it so great for researchers+business users: ✅ Create a knowledge repository of documents. Has likely higher accuracy than NotebookLM since it’s using LlamaCloud under the hood for high-quality parsing/extraction over complex docs. ✅ Generate summaries and knowledge graph mind-maps 🤯 ✅ Generate podcasts thanks to ElevenLabs 🗣️ ✅ Agentic chat with docs and view metrics with OpenTelemetry This all lives within an open-source repo so you can clone/modify at will to swap in your own components! Huge shoutout to Clelia Bertelli (🦙/acc) for leading this. Repo: LlamaCloud helps power the parsing/ingestion. You can always use your own stuff too, but in the meantime you can check out LlamaCloud here:

Jerry Liu

137,793 Aufrufe • vor 1 Jahr

We’re open sourcing the first document OCR benchmark for the agentic era, ParseBench. Document parsing is the foundation of every AI agent that works with real-world files. ParseBench is a benchmark that measures parsing quality specifically for agent knowledge work: ✅ It optimizes for semantic correctness (instead of exact similarity) ✅ It has the most comprehensive distribution of real-world enterprise documents It contains ~2,000 human-verified enterprise document pages with 167,000+ test rules across five dimensions that matter most: tables, charts, content faithfulness, semantic formatting, and visual grounding. We benchmarked 14 known document parsers on ParseBench, from frontier/OSS VLMs to specialized parsers to LlamaParse. Here are some of our findings: 💡 Increasing compute budget yields diminishing returns - Gemini/gpt-5-mini/haiku gain 3-5 points from minimal to high thinking, at 4x the cost. 💡 Charts are the most polarizing dimension for evaluation. Most specialized parsers score below 6%, while some VLM-based parsers do a bit better. 💡 VLMs are great at visual understanding but terrible at layout extraction. GPT-5-mini/haiku score below 10% on our visual grounding task, all specialized parsers do much better. 💡 No method crushes all 5 dimensions at once, but LlamaParse achieves the highest overall score at 84.9%, and is the leader in 4 out of the 5 dimensions. This is by far the deepest technical work that we’ve published as a company. I would encourage you to start with our blog and explore our links to Hugging Face to GitHub. All the details are in our full 35-page (!!) ArXiv whitepaper. 🌐: Blog: 📄 Paper: 💻 Code: 📊 Dataset: 🎥 YouTube:
1:08

We’re open sourcing the first document OCR benchmark for the agentic era, ParseBench. Document parsing is the foundation of every AI agent that works with real-world files. ParseBench is a benchmark that measures parsing quality specifically for agent knowledge work: ✅ It optimizes for semantic correctness (instead of exact similarity) ✅ It has the most comprehensive distribution of real-world enterprise documents It contains ~2,000 human-verified enterprise document pages with 167,000+ test rules across five dimensions that matter most: tables, charts, content faithfulness, semantic formatting, and visual grounding. We benchmarked 14 known document parsers on ParseBench, from frontier/OSS VLMs to specialized parsers to LlamaParse. Here are some of our findings: 💡 Increasing compute budget yields diminishing returns - Gemini/gpt-5-mini/haiku gain 3-5 points from minimal to high thinking, at 4x the cost. 💡 Charts are the most polarizing dimension for evaluation. Most specialized parsers score below 6%, while some VLM-based parsers do a bit better. 💡 VLMs are great at visual understanding but terrible at layout extraction. GPT-5-mini/haiku score below 10% on our visual grounding task, all specialized parsers do much better. 💡 No method crushes all 5 dimensions at once, but LlamaParse achieves the highest overall score at 84.9%, and is the leader in 4 out of the 5 dimensions. This is by far the deepest technical work that we’ve published as a company. I would encourage you to start with our blog and explore our links to Hugging Face to GitHub. All the details are in our full 35-page (!!) ArXiv whitepaper. 🌐: Blog: 📄 Paper: 💻 Code: 📊 Dataset: 🎥 YouTube:

Jerry Liu

108,093 Aufrufe • vor 4 Monaten

As frontier models (e.g. Fable 5) continue to push the task horizon of knowledge work automation, it becomes ever more important for humans to be able to audit decisions back to the source context. It is extremely easy for agents to cite an entire document or document page, but much harder for them to trace back to the exact numbers/words/figures within a page. Today we've launched granular bounding boxes within LlamaParse, which allows you to obtain visual citations of every single word in the document. This allows human users to audit exact words and figures - not just general document regions or entire pages! Come check it out:
0:51

As frontier models (e.g. Fable 5) continue to push the task horizon of knowledge work automation, it becomes ever more important for humans to be able to audit decisions back to the source context. It is extremely easy for agents to cite an entire document or document page, but much harder for them to trace back to the exact numbers/words/figures within a page. Today we've launched granular bounding boxes within LlamaParse, which allows you to obtain visual citations of every single word in the document. This allows human users to audit exact words and figures - not just general document regions or entire pages! Come check it out:

Jerry Liu

23,491 Aufrufe • vor 2 Monaten

I’ve excited to announce a brand-new website and documentation hub 💫 that solidifies our evolution towards automating knowledge work over your documents. You might’ve followed us since the “RAG framework” days. Even then, the biggest challenge users faced was figuring out how to actually ingest an entire collection of unstructured docs (.pdf, .pptx, .docx, and more) for chatbot/agentic workflow use cases. Over the past year we’ve progressively built up incredibly deep tech around document parsing, extraction, and indexing - while teaching developers how to build various workflows on top. We’re now going all in on documents, and we’re the only company that has both 1) SOTA document processing and file management 📈, and 2) agentic orchestration on top to solve use cases like deep research, report generation, and document workflows end-to-end. Our llamas will continue to love all sorts of data (we have 600+ integrations on the open-source framework!), but they now especially love automating paperwork 🦙📄. If you would also love to automate paperwork, come check out our new website and come talk to us! Site: Developer Hub:
3:07

I’ve excited to announce a brand-new website and documentation hub 💫 that solidifies our evolution towards automating knowledge work over your documents. You might’ve followed us since the “RAG framework” days. Even then, the biggest challenge users faced was figuring out how to actually ingest an entire collection of unstructured docs (.pdf, .pptx, .docx, and more) for chatbot/agentic workflow use cases. Over the past year we’ve progressively built up incredibly deep tech around document parsing, extraction, and indexing - while teaching developers how to build various workflows on top. We’re now going all in on documents, and we’re the only company that has both 1) SOTA document processing and file management 📈, and 2) agentic orchestration on top to solve use cases like deep research, report generation, and document workflows end-to-end. Our llamas will continue to love all sorts of data (we have 600+ integrations on the open-source framework!), but they now especially love automating paperwork 🦙📄. If you would also love to automate paperwork, come check out our new website and come talk to us! Site: Developer Hub:

Jerry Liu

25,710 Aufrufe • vor 11 Monaten

I used Claude Code Skills + souped up PDF parsing to create an M&A deal comp agent 🤖 Given a directory of public M&A filings (DEF 14A), it parses and analyzes each pdf, and generates an Excel sheet with deal terms and comparables. 1️⃣ The native parsing is pypdf which sucks, so I swapped it out with LlamaIndex semtools. This uses LlamaCloud for parsing - letting it handle deeply complex financial tables, charts, and anything else you throw at it. 2️⃣ Claude has native skills to write Excel spreadsheets! Full prompt is in the video below 👇 LlamaCloud + Semtools access also takes < 5 mins to setup, come check it out! Some notes: - disclaimer: Some of the values in the Excel spreadsheet are 100x bigger because they’re formatted as percentages not raw values - We’re working on native Claude skill integrations too! Semtools: LlamaCloud:
1:09

I used Claude Code Skills + souped up PDF parsing to create an M&A deal comp agent 🤖 Given a directory of public M&A filings (DEF 14A), it parses and analyzes each pdf, and generates an Excel sheet with deal terms and comparables. 1️⃣ The native parsing is pypdf which sucks, so I swapped it out with LlamaIndex semtools. This uses LlamaCloud for parsing - letting it handle deeply complex financial tables, charts, and anything else you throw at it. 2️⃣ Claude has native skills to write Excel spreadsheets! Full prompt is in the video below 👇 LlamaCloud + Semtools access also takes < 5 mins to setup, come check it out! Some notes: - disclaimer: Some of the values in the Excel spreadsheet are 100x bigger because they’re formatted as percentages not raw values - We’re working on native Claude skill integrations too! Semtools: LlamaCloud:

Jerry Liu

46,748 Aufrufe • vor 10 Monaten

🦙 Since you loved NotebookLlaMa, our open-source, LlamaCloud-backed NotebookLM alternative, we decided to give you more features! In the latest release you can: 📊 Extract and download images and tables from your files, and interactively visualize all tabular data ✅ Chat with your document and verify immediately everything that AI is saying about it, using the new claim verification and citation finding functionality 👇 Watch the video below for a full demo of the new functionalities 👩‍💻🧑‍💻 There is also a bonus: thanks to the community effort and contribution, you can now customize the embedding model you want to use to build the LlamaCloud index, through a textualize.io CLI! ⭐ Star NotebookLlaMa on GitHub: 🦙☁️ Sign-up for LlamaCloud and test our awesome products: 📚 Learn more about LlamaCloud:
1:13

🦙 Since you loved NotebookLlaMa, our open-source, LlamaCloud-backed NotebookLM alternative, we decided to give you more features! In the latest release you can: 📊 Extract and download images and tables from your files, and interactively visualize all tabular data ✅ Chat with your document and verify immediately everything that AI is saying about it, using the new claim verification and citation finding functionality 👇 Watch the video below for a full demo of the new functionalities 👩‍💻🧑‍💻 There is also a bonus: thanks to the community effort and contribution, you can now customize the embedding model you want to use to build the LlamaCloud index, through a textualize.io CLI! ⭐ Star NotebookLlaMa on GitHub: 🦙☁️ Sign-up for LlamaCloud and test our awesome products: 📚 Learn more about LlamaCloud:

LlamaIndex 🦙

11,466 Aufrufe • vor 1 Jahr

Finally, a RAG solution that works with complex documents! Real-world documents can be messy, filled with text, tables, images, and intricate flow charts. Traditional parsing and chunking methods struggle to handle these. So, what’s the solution? We need smart techniques that can intuitively chunk relevant context and understand what’s inside each chunk, whether it's text, images, or diagrams. In this video, I’ll walk you through a breakthrough technique for extracting structured information from complex documents. It's unlike any other technique you've seen before.✨ It takes any unstructured (text, tables, images, flow-charts) input and parses it into a JSON format that LLMs can easily process. I used EyeLevel.AI's GroundX platform for this – a powerful tool that allows you to build a RAG application in just 3 steps. It also comes with a nice Python SDK and can be easily deployed on-premise (K8s cluster)! Try it yourself:
6:45

Finally, a RAG solution that works with complex documents! Real-world documents can be messy, filled with text, tables, images, and intricate flow charts. Traditional parsing and chunking methods struggle to handle these. So, what’s the solution? We need smart techniques that can intuitively chunk relevant context and understand what’s inside each chunk, whether it's text, images, or diagrams. In this video, I’ll walk you through a breakthrough technique for extracting structured information from complex documents. It's unlike any other technique you've seen before.✨ It takes any unstructured (text, tables, images, flow-charts) input and parses it into a JSON format that LLMs can easily process. I used EyeLevel.AI's GroundX platform for this – a powerful tool that allows you to build a RAG application in just 3 steps. It also comes with a nice Python SDK and can be easily deployed on-premise (K8s cluster)! Try it yourself:

Akshay 🚀

102,660 Aufrufe • vor 1 Jahr

We’re on a mission to parse the world’s hardest PDFs, and we’d love your help There are so many document types that introduce a million edge cases for current VLMs / OCR: handwritten forms, badly scanned/rotated pages, charts, diagrams, and more. We are running a contest right now for you to try to extract the hardest PDFs you can find. Come sign up on our agent builder, describe what you want to extract through natural language, upload your document, and show the results. If our platform doesn’t work, even better; this is great feedback for us to improve our service. Either way submit your project and we’d love to get your feedback! Check out LlamaCloud here:
0:34

We’re on a mission to parse the world’s hardest PDFs, and we’d love your help There are so many document types that introduce a million edge cases for current VLMs / OCR: handwritten forms, badly scanned/rotated pages, charts, diagrams, and more. We are running a contest right now for you to try to extract the hardest PDFs you can find. Come sign up on our agent builder, describe what you want to extract through natural language, upload your document, and show the results. If our platform doesn’t work, even better; this is great feedback for us to improve our service. Either way submit your project and we’d love to get your feedback! Check out LlamaCloud here:

Jerry Liu

21,005 Aufrufe • vor 6 Monaten

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 surprisingly they still struggle on complex doc extraction tasks in production. A well-tuned extractor must parse multi-page filings without dropping rows, emit exact spatial citations for auditability, and handle messy scans. Also they must do all of this at a viable per-page cost so that you can scale this to millions of docs in production (you can’t be paying upwards of $1 in tokens per page!) Existing extraction benchmarks fall short: they are not large/diverse enough in document domain (finance, energy, gov, auto), elements (long records, scans, grounding), and schemas. So our applied research team built ExtractBench. We evaluated 14 systems: frontier VLMs, coding agents, and specialized extraction APIs, against 370 enterprise documents: 4,869 pages, 67 document types. Our biggest finding 🧪: Short documents mask critical system flaws. On files past 50 pages, commercial VLMs collapse below 35% recall due to silent list truncation. They hold high precision, but lose output attention and drop most of the table rows. ExtractBench evaluates value accuracy, long-record completeness, spatial grounding, and per-page cost with zero LLM judges. It is 100% deterministic and reproducible. In tandem with ExtractBench, we’re also introducing 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗣𝗹𝘂𝘀, a new Extract tier in LlamaParse that debuts at #1 on the leaderboard: 95.6% value accuracy, at less than a third the cost of the closest peer. Explore the findings, download the dataset, or run the harness: Blog: GitHub: HuggingFace: We will be actively evolving both our extraction benchmark as well as our extraction harness over time. If you check out either ExtractBench or LlamaParse, let us know your feedback!
1:59

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 surprisingly they still struggle on complex doc extraction tasks in production. A well-tuned extractor must parse multi-page filings without dropping rows, emit exact spatial citations for auditability, and handle messy scans. Also they must do all of this at a viable per-page cost so that you can scale this to millions of docs in production (you can’t be paying upwards of $1 in tokens per page!) Existing extraction benchmarks fall short: they are not large/diverse enough in document domain (finance, energy, gov, auto), elements (long records, scans, grounding), and schemas. So our applied research team built ExtractBench. We evaluated 14 systems: frontier VLMs, coding agents, and specialized extraction APIs, against 370 enterprise documents: 4,869 pages, 67 document types. Our biggest finding 🧪: Short documents mask critical system flaws. On files past 50 pages, commercial VLMs collapse below 35% recall due to silent list truncation. They hold high precision, but lose output attention and drop most of the table rows. ExtractBench evaluates value accuracy, long-record completeness, spatial grounding, and per-page cost with zero LLM judges. It is 100% deterministic and reproducible. In tandem with ExtractBench, we’re also introducing 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗣𝗹𝘂𝘀, a new Extract tier in LlamaParse that debuts at #1 on the leaderboard: 95.6% value accuracy, at less than a third the cost of the closest peer. Explore the findings, download the dataset, or run the harness: Blog: GitHub: HuggingFace: We will be actively evolving both our extraction benchmark as well as our extraction harness over time. If you check out either ExtractBench or LlamaParse, let us know your feedback!

Jerry Liu

74,199 Aufrufe • vor 17 Tagen

PDF parsing is still painful because LLMs reorder text in complex layouts, break tables across pages, and fail on graphs or images. 💡Testing the new open-source OCRFlux model, and here the results are really good for a change. So OCRFlux is a multimodal, LLM based toolkit for converting PDFs and images into clean, readable, plain Markdown text. Because the underlying VLM is only 3B param, it runs even on a 3090 GPU. The model is available on Hugging Face . The engine that powers the OCRFlux, teaches the model to rebuild every page and then stitch fragments across pages into one clean Markdown file. It bundles one vision language model with 3B parameters that was fine-tuned from Qwen 2.5-VL-3B-Instruct for both page parsing and cross-page merging. OCRFlux reads raw page images and, guided by task prompts, outputs Markdown for each page and merges split elements across pages. The evaluation shows Edit Distance Similarity (EDS) 0.967 and cross‑page table Tree Edit Distance 0.950, so the parser is both accurate and layout aware. How it works while parsing each page - Convert into text with a natural reading order, even in the presence of multi-column layouts, figures, and insets - Support for complicated tables and equations - Automatically removes headers and footers Cross-page table/paragraph merging - Cross-page table merging - Cross-page paragraph merging A compact vision‑language models can beat bigger models once cross‑page context is added. 🧵 1/n Read on 👇
0:44

PDF parsing is still painful because LLMs reorder text in complex layouts, break tables across pages, and fail on graphs or images. 💡Testing the new open-source OCRFlux model, and here the results are really good for a change. So OCRFlux is a multimodal, LLM based toolkit for converting PDFs and images into clean, readable, plain Markdown text. Because the underlying VLM is only 3B param, it runs even on a 3090 GPU. The model is available on Hugging Face . The engine that powers the OCRFlux, teaches the model to rebuild every page and then stitch fragments across pages into one clean Markdown file. It bundles one vision language model with 3B parameters that was fine-tuned from Qwen 2.5-VL-3B-Instruct for both page parsing and cross-page merging. OCRFlux reads raw page images and, guided by task prompts, outputs Markdown for each page and merges split elements across pages. The evaluation shows Edit Distance Similarity (EDS) 0.967 and cross‑page table Tree Edit Distance 0.950, so the parser is both accurate and layout aware. How it works while parsing each page - Convert into text with a natural reading order, even in the presence of multi-column layouts, figures, and insets - Support for complicated tables and equations - Automatically removes headers and footers Cross-page table/paragraph merging - Cross-page table merging - Cross-page paragraph merging A compact vision‑language models can beat bigger models once cross‑page context is added. 🧵 1/n Read on 👇

Rohan Paul

149,292 Aufrufe • vor 1 Jahr

Building a RAG system that works with real-life documents is crazy hard. Why is nobody talking about this? (I'll show you what a complex document looks like in the attached video. Good luck with the 10-line code demos if you want to deal with this.) All I see online are "how to talk to a PDF" demos that won't take you far. If you think that's all you need, you won't like what happens when you try. In the video below, I'll show you what it takes to build a system that processes huge numbers of documents without losing accuracy. There are a few tricks here that I'm sure will impress you. The magic is happening in three phases: • In the way we process the documents • In the way we chunk the content • In the way we augment and store the chunks This is all built-in in GroundX, EyeLevel.AI's out-of-the-box RAG system. You can use it as a SaaS or install it in your own Kubernetes cluster and run it locally. If you have seen any other RAG system doing what I do in the attached video, please, let me know. I'd love to check them out. Disclaimer: I work with the team at EyeLevel. My advice is to take any document you have lying around, go to their site, create an account, and try them out. Here is their website:
10:11

Building a RAG system that works with real-life documents is crazy hard. Why is nobody talking about this? (I'll show you what a complex document looks like in the attached video. Good luck with the 10-line code demos if you want to deal with this.) All I see online are "how to talk to a PDF" demos that won't take you far. If you think that's all you need, you won't like what happens when you try. In the video below, I'll show you what it takes to build a system that processes huge numbers of documents without losing accuracy. There are a few tricks here that I'm sure will impress you. The magic is happening in three phases: • In the way we process the documents • In the way we chunk the content • In the way we augment and store the chunks This is all built-in in GroundX, EyeLevel.AI's out-of-the-box RAG system. You can use it as a SaaS or install it in your own Kubernetes cluster and run it locally. If you have seen any other RAG system doing what I do in the attached video, please, let me know. I'd love to check them out. Disclaimer: I work with the team at EyeLevel. My advice is to take any document you have lying around, go to their site, create an account, and try them out. Here is their website:

Santiago

75,788 Aufrufe • vor 1 Jahr

I’ve created a full-stack financial analysis bot that can query both text + embedded tables across multiple SEC 10Ks 📊🤖 It’s made possible with `create-llama` scaffolding + LlamaIndex 🦙 advanced RAG, and I’m sharing the full template below for you to clone! It’s more sophisticated than a basic RAG setup: 🦾 Unstructured to parse embedded tables into a node graph 🦾 Recursive retriever to retrieve/query embedded tables + text 🦾 An agent to do chain of thought + document comparisons 🦾 Custom callback to stream intermediate function calls to the UI Our goal is to make building full-stack advanced RAG as easy as possible. Created a repo to host this template + future templates here: Want to submit a project? We’d love contributions! 🙌 One caveat: - Right now the index is lazily built/cached during the first query, we’re working to decouple the ingestion process - Check out `create-llama` if you haven’t already:
1:05

I’ve created a full-stack financial analysis bot that can query both text + embedded tables across multiple SEC 10Ks 📊🤖 It’s made possible with `create-llama` scaffolding + LlamaIndex 🦙 advanced RAG, and I’m sharing the full template below for you to clone! It’s more sophisticated than a basic RAG setup: 🦾 Unstructured to parse embedded tables into a node graph 🦾 Recursive retriever to retrieve/query embedded tables + text 🦾 An agent to do chain of thought + document comparisons 🦾 Custom callback to stream intermediate function calls to the UI Our goal is to make building full-stack advanced RAG as easy as possible. Created a repo to host this template + future templates here: Want to submit a project? We’d love contributions! 🙌 One caveat: - Right now the index is lazily built/cached during the first query, we’re working to decouple the ingestion process - Check out `create-llama` if you haven’t already:

Jerry Liu

131,442 Aufrufe • vor 2 Jahren

Web scraping will never be the same. (100% open-source visual search at scale) PixelRAG is a retrieval system that skips HTML parsing completely. Instead of scraping a page into text and embedding chunks, it screenshots the page and retrieves the image. A vision-language model reads the answer straight off the pixels. Why that matters: parsing is where web RAG quietly loses information. - A single HTML-to-text parser can drop 40%+ of a page. - Tables, charts, and layout get flattened or thrown out. - Swapping parsers alone can move accuracy ~10 points on the same docs. PixelRAG indexes the page a person actually sees. The team built a visual index of all of Wikipedia, 30M+ screenshots, and it still beats the strongest text RAG baseline by 18.1% on text-only QA. The repo also ships a Claude Code plugin that gives Claude eyes. It lets Claude screenshot any URL and read the rendered page instead of scraping the DOM. So you can hand it a live page, an arXiv paper, or your local site and ask what it actually looks like. One setup script. No MCP server, no backend. How the pipeline works: - Renders each document (web, PDF, image) to image tiles. - Embeds them with Qwen3-VL-Embedding, LoRA fine-tuned on screenshots. - Builds a FAISS index and serves a search API. A stronger reader model lifts accuracy with no re-indexing, since the index is just pixels. Everything is open-source under Apache-2.0. GitHub repo: Talking about RAG, I recently wrote an article on a new approach that makes retrieval much more efficient by cutting corpus size by 40x, reducing tokens per query by 3x, and improving vector search relevance by 2.3x. The article is quoted below.
0:55

Web scraping will never be the same. (100% open-source visual search at scale) PixelRAG is a retrieval system that skips HTML parsing completely. Instead of scraping a page into text and embedding chunks, it screenshots the page and retrieves the image. A vision-language model reads the answer straight off the pixels. Why that matters: parsing is where web RAG quietly loses information. - A single HTML-to-text parser can drop 40%+ of a page. - Tables, charts, and layout get flattened or thrown out. - Swapping parsers alone can move accuracy ~10 points on the same docs. PixelRAG indexes the page a person actually sees. The team built a visual index of all of Wikipedia, 30M+ screenshots, and it still beats the strongest text RAG baseline by 18.1% on text-only QA. The repo also ships a Claude Code plugin that gives Claude eyes. It lets Claude screenshot any URL and read the rendered page instead of scraping the DOM. So you can hand it a live page, an arXiv paper, or your local site and ask what it actually looks like. One setup script. No MCP server, no backend. How the pipeline works: - Renders each document (web, PDF, image) to image tiles. - Embeds them with Qwen3-VL-Embedding, LoRA fine-tuned on screenshots. - Builds a FAISS index and serves a search API. A stronger reader model lifts accuracy with no re-indexing, since the index is just pixels. Everything is open-source under Apache-2.0. GitHub repo: Talking about RAG, I recently wrote an article on a new approach that makes retrieval much more efficient by cutting corpus size by 40x, reducing tokens per query by 3x, and improving vector search relevance by 2.3x. The article is quoted below.

Akshay 🚀

945,628 Aufrufe • vor 2 Monaten