🤖AI❤️Data enjoyer , building robots to helps folks learn things quicker.
Paris, France
Joined April 2009
- Tweets4.3K
- Following6.3K
- Followers2K
- Likes40.7K
👋hi there folks ,
I was inspired by this Jev-based ocr corrector concept and wanted to try it for myself ! it's actually free to deploy and run on @huggingface , and doesnt use any GPU quota .
shoutout to @OpenBMB for the implementations 🫡
demo : huggingface.co/spaces/Tonic/…
Your OCR says it’s done—but did it get everything right? 🤔 In high-stakes workflows, even a small error can be costly.
🚀 Meet the MiniCPM-OCR suite: an open-source solution to parse fast and guard the results.
Two new models:
🔹 MiniCPM-OCR 1.0 — SOTA OCR performance and speed at 1B scale. Compact, accurate, and fast.
🔹 MiniCPM-OCR Jev — A 2B decision model that helps assess parsing accuracy, with performance approaching that of Qwen3.5-397B-A17B on JevOCRBench.
⚡ What’s new?
Small model, SOTA results: MiniCPM-OCR 1.0 scores 93.68 on Real5-OmniDocBench and 54.7 on OCRBench v2.
A sharp goalkeeper 🛡️: MiniCPM-OCR Jev is model-agnostic, checking text, formulas, and tables from different OCR parsers. It achieves 93.6 on JevOCRBench, outperforming JEV-35B-A3B (90.1) and approaching Qwen3.5-397B-A17B (93.8).
Up to 128 pages per minute 🤯: With DSpark, MiniCPM-OCR runs at 1.49× the speed of PaddleOCR-VL-1.6 and 3.62× that of MinerU2.5 Pro. The SDK’s concurrency optimizations further boost throughput to as much as 128 pages per minute.
Deploy where you work: Day-0 support for Arm, Houmo AI, Intel, and NVIDIA, plus 10 more chip vendors through FlagOS.
Ready for your stack: vLLM, SGLang, llama.cpp, and Ollama, with quantized models, fine-tuning support, and Agent Skills.
🔗 GitHub: github.com/OpenBMB/MiniCPM-O…
🤗 Hugging Face:
MiniCPM-OCR 1.0：huggingface.co/openbmb/MiniC…
MiniCPM-OCR Jev：huggingface.co/openbmb/MiniC…
👉 Try the Demo：platform.modelbest.ai/consol…
Use the API：platform.modelbest.ai/consol…
Joseph Pollack #Ï 🎗️ retweeted
Yesterday, a russian Geran-4/5 struck the Hospital and Institute of Cardiology in Solom'yanka, Kyiv
50.420375, 30.459610
@GeoConfirmed
t.me/gunpKyiv/17593
t.me/kievinfo_kyiv/104064
Meanwhile criminals in the @KremlinRussia continue to deliberately target health infrastructure in a genocidal bid to cause mass population movements and death to steal their belongings .
Public health is #notatarget
Help stop the war by donating to: u24.gov.ua
Another ballistic missile in the 3am wave impacted right outside Shuliavska Polyclinic, next to the mall struck in the QT.
50.451698, 30.445697
@GeoConfirmed
Reference: maps.app.goo.gl/gbr1kGyQ1uni…
t.me/kyivtv/155328
Joseph Pollack #Ï 🎗️ retweeted
We release SlideOn and mSlideOn at @LightOnIO today, two sota sparse retrievers, models designed by @paulomouraj and I.
Those models are really strong for first stage retrieval, you can add then a multi-vector model as a reranker and then a decision model 👇
We surfed the dense and late-interaction waves… now it's time to SlideOn sparse retrieval 🐕
🚀 Meet SlideOn and mSlideOn, fully open SOTA sparse retrievers at 174M and 339M parameters, trained by @LightOnIO R&D
Models:
huggingface.co/collections/l…
Blog: huggingface.co/blog/lightona…
Joseph Pollack #Ï 🎗️ retweeted
Ukraine’s 21st KRAKEN Unmanned Systems Regiment says its MURAMASA operators scored their first hit on a Russian TOS-3 Drakon heavy flamethrower system during Operation Vivaldi. The TOS-3 is one of Russia’s newest weapons systems. #Ukraine
Meanwhile every European country has recorded dozens of terrorist attacks sponsored by the @KremlinRussia .
You can help save lives by writing to your representative demanding a response or donate to u24.gov.ua
Wake up hun new programming language just dropped - follow @Chi_Wang_ 😉👇🏻check out vxlang.com
I’ll take that bet : mandarin is the superior concept-per-token language .
🤝
This quoted post is unavailable.
Joseph Pollack #Ï 🎗️ retweeted
Direct strike on a hospital. Savages.
This morning, a russian Shahed struck a hospital in Zaporizhzhia
47.844764, 35.140323
@GeoConfirmed
t.me/dsns_telegram/75650
t.me/riamelitopolua/94865
🤝 @99Dominik_
Joseph Pollack #Ï 🎗️ retweeted
Russian drones are hitting book warehouses all across Ukraine, becuase the destruction of Ukrainian language and culture is one their war's central goals.
nytimes.com/2026/09/25/world…
the Readmes are really well done , and a lot of depth is shared here , which is nice.
github.com/featherless-ai/si…
We are jev-ifying open source models @FeatherlessAI
- which surpasses jev in evals
- and includes image/vision
With a starting price of [check notes]... Free?
simple-jev-demo-api.featherl…
This is great for lerobot datasets
New JS package: @huggingface/lerobot 🤖
Read LeRobot datasets on the Hub straight from the browser. No download.
Point your coding agent at it and build custom viewers fast. Example 3D camera-frustum replay streams 1.1 MB of a 793 MB dataset:
hf.co/spaces/mishig/lerobot-…
Xiaomi did a really nice job with CodeMidas paper sharing knowledge and experience in massively distributed RL using agents.
The one thing I do like in this timeline is seeing decades of promoting Chinese open source collaboration finally coming to bear and it’s better than exp
The backwards American citizen will be very glad to receive low value dirt from a bloodthirsty dictatorship . They might even do a traditional dance when the container ships roll on in like they did for the urea shipments in South Africa
This is actually cool for folks that expand base and instruct models with new/better tokenizers
Tokenizers v1's first release candidate is out 🔥 absolutely huge release across the board:
Up to 30x faster tokenization, rivaling the best in the ecosystem. Scaling across threads, much better latency, less memory consumed, and microscopic crate size.
huggingface.co/blog/tokenize…
Meanwhile , criminals in the @KremlinRussia gleefully continue to target public health facilities in their genocidal bid to murder civilians and thieve as much as they can .
stop this by writing by to your rep to demand air defense or by donating here u24.gov.ua
