Profoundly shallow. Funambulista, sobreviviente, cambiante, suertudote y casi permanentemente triste por puro gusto. Aborrezco los fanatismos y los mosquitos.
Joined April 2010
- Tweets7.9K
- Following1.6K
- Followers204
- Likes12.6K
Edher Jimenez retweeted
Decision models can also improve your OCR pipelines.
You can also use Jev-like models for verification.
Here is a great example of this:
MiniCPM-OCR from OpenBMB and ModelBest pairs a small open document parser with a second model that scores the reliability of the OCR output.
The parser has only 1.3B parameters. It handles Chinese and English documents, including scans, photos, curved pages, and tilted inputs, and it supports key information extraction.
The second model, MiniCPM-OCR Jev, checks extracted text, formulas, and tables. It also works on outputs from other OCR models, so you can run it on top of the parser you already use.
More details below.
Agents now read PDFs, invoices, and papers all day. One bad table extraction carries into every step that follows.
A simple setup looks like this.
> Parse the document with MiniCPM-OCR
> Score the extracted text, formulas, and tables with Jev
> Send low-scoring output to a human or a correction step
Jev estimates reliability. It doesn't fix errors itself.
The weights are open, with a fine-tuning framework and inference support for llama.cpp, vLLM, SGLang, and Ollama. There are also Agent Skills integrations.
If you build document agents, test the parser and Jev together on your hardest pages, like dense tables, formulas, and phone photos.
Edher Jimenez retweeted
We (actually) KILLED Palantir this time!!!
New updates, better ontology generation, cloud service, and smarter agents!
You can deploy it locally here: github.com/HelixDB/helix-fou…
or use the cloud: foundry.helix-db.com
Edher Jimenez retweeted
Introducing, 100% free charts kit for you guys.
Spent whole damn month working on this 62 chart components. Available in Figma, code, or connect your MCP.
Download here: mebius.design/freebies/chart…
Edher Jimenez retweeted
🎉 Introducing OpenDuet
Self-hostable, proactive AI agents that you own and works with ANY agent harness.
Includes:
- Build, generate, and edit live assets
- Human-in-the-loop approvals
- Computer use
- Mobile, Web, and Channels
Repo → github.com/CopilotKit/OpenDu…
Powered by @CopilotKit and AG-UI.
Clone this template and customize it however you want. Enterprise-ready.
Edher Jimenez retweeted
T-minus 6.4 seconds.
That's how long our speaker diarization model, Who, took to label every speaker when we tested it on a 57-minute NASA podcast.
It ran at 530× real time, on a MacBook, with nothing leaving the device.
Edher Jimenez retweeted
📦 NVIDIA did it first, now we are releasing 2,600+ free game-ready 3D models (+200 each week). For good.
🎛️ Each model is a program, with parameters you can tweak to customize look, state, season, color, wear, etc.
All thanks to the $500,000 in funding we received from @xprize and @GeminiApp and all of our early individual funders, thank you all!
Go build.
polyfork.dev
NVIDIA just put 2,575 simulation-ready 3D assets on Hugging Face 🤯
A 472GB collection covering everyday objects, tools and industrial equipment.
It’s called NVIDIA SimReady Assets. The models include physics settings and semantic labels for use in simulated environments.
- Train robots to pick up and move objects
- Build virtual factories and warehouses
- Generate labelled training images
- Populate 3D scenes with ready-made objects
Useful for robotics research, digital twins and anyone building detailed virtual environments.
Download the collection below 👇
Edher Jimenez retweeted
Oh my. Best open OCR model by a mile.
... and Baidu took the KV cache for hundred-page OCR from 47GB → 1.77GB.
Unlimited OCR 2 is only ~3.3B parameters and it's designed to parse documents 100's of pages long without memory and decode latency growing as the document gets bigger.
The secret sauce is SR-SWA.
Instead of every new token repeatedly attending across an ever-growing document, it keeps a fixed-size set of useful text/visual references.
According to Baidu ...
🧠 Peak KV cache → 47.0GB → 1.77GB
📉 ~96% reduction
⚡ per-step decode → up to 7× faster
🚀 16-doc throughput → 360 → 1,315 tps 👈👀
Now look at Baidu's 100–300-page benchmark MarathonDocBench:
Unlimited OCR 2 → 77.1
Kimi-K3 → 33.7
Gemini 3.8 Flash → 28.6
Qwen3.8-Max → 22.9
Of course, this is vendor-reported, so independent reproduction matters.
Baidu also reports that shorter documents get even more accurate
Unlimited OCR 2 → 84.6
Qwen3.8-Max → 78.3
Model Stats
🧠 ~3.34B parameters
💾 ~6.7GB BF16 weights
🔓 MIT licensed
📄 text + tables + formulas + layout
⚙️ Transformers + SGLang
Can we squeeze this onto a 16GB GPU?
Links in ALT
ALT https://github.com/baidu/Unlimited-OCR https://huggingface.co/baidu/Unlimited-OCR-2
Edher Jimenez retweeted
Typed Decision Engine for Agent Workflows
Forget slow autoregressive generation and fragile JSON parsing.
This fine-tune turns EmbeddingGemma 2 into a fast decision engine that evaluates workflow states (like agent traces or support tickets) against a dynamic schema of questions in a single forward pass. It returns calibrated choice, score, and bool outputs without retraining.
Model and post by @mrm8488: nitter.cf/mrm8488/status/2109406…
Hugging Face: huggingface.co/mrm8488/embed…
🧭 New model: Typed Decision, built on EmbeddingGemma 2
Give it a state + any schema of questions → typed, calibrated answers (choice · score · noul). No text generation, no output parsing.
76.4% on typed-decisions, ~6 pts behind a 27B LLM with ~100× fewer params.
🤗 huggingface.co/mrm8488/embed…
Edher Jimenez retweeted
AI doesn't hear the world like people do. In fact, AI doesn't hear the world like other AI.
Across 24 speech recognizers, phonemes were easiest to read out at 21% of encoder depth in Conformers on average, vs 49% in Transformers.
Catch my presentation tomorrow at EMNLP.
Edher Jimenez retweeted
Introducing Anvaya, the world's lightest agentic harness.
Built in Rust. Designed to run locally.
• 17x lighter than Claude Code
• Just ~5 MB memory at idle
• Supports up to 256 concurrent agents
• Works with Ollama, Anthropic and OpenAI
• Learns from previous sessions instead of starting from scratch
Less memory. Less wasted context. Smarter agents.
Meet Anvaya.
Edher Jimenez retweeted
Is the end of LLMs near?
arxiv.org/html/2610.14162v1
Edher Jimenez retweeted
▲ Microsoft just open-sourced Agensh, a multi-agent harness that runs up to 1,024 coding agents with no lead agent in charge.
The key idea is self-organization. Every worker runs the same loop: gather context, claim a sub-task, build, verify, merge. They coordinate through a shared Gitea repo, a Mattermost channel and a shared findings board, instead of an orchestrator.
On the 5 hardest ProgramBench tasks (GPT-5.6-sol, 6 hours, offline), the mean test-pass rate rose from 19.31% with 1 agent to 28.78% with 128 agents. On pandoc, 1,024 agents hit 55.06%, up from 33.89%. With GPT-6.1-sol, 32 agents reached 79.59%. It is a trade-off though: going from 1 to 8 agents added only 1.37 points on the 5-task mean, and token cost is not disclosed.
The full harness, launch scripts and reproduction guide are on GitHub under the MIT license. It runs on Linux with Docker Compose......
Full analysis: marktechpost.com/2026/10/24/…
Repo: github.com/microsoft/Agensh
Paper: arxiv.org/abs/2609.26781
Project: aka.ms/Agensh
@MSFTResearch @msdev @msdev
Edher Jimenez retweeted
Usually, exciting news from @Google is about geospatial (and many other things), but this one piece I caught in the Guardian is rather sad.
In a recent paper of mine, I figured that in the US, most of the large data center constructions seem to happen around former farmland and industrial sites - perhaps time for them to read it:
𝐐𝐮𝐚𝐧𝐭𝐢𝐟𝐲𝐢𝐧𝐠 𝐕𝐞𝐠𝐞𝐭𝐚𝐭𝐢𝐨𝐧 𝐋𝐨𝐬𝐬 𝐀𝐫𝐨𝐮𝐧𝐝 𝐭𝐡𝐞 𝐋𝐚𝐫𝐠𝐞𝐬𝐭 𝐔.𝐒. 𝐃𝐚𝐭𝐚 𝐂𝐞𝐧𝐭𝐞𝐫𝐬 - researchgate.net/publication…
The reporting gave a size and "about 400 m from the Oulujoki" - but no coordinates. So I did not go looking for the site. I differenced two summers of Sentinel-2 NDVI across the whole district and let the largest fresh clearing identify itself: 64.836 N, 26.113 E.
Reported as 330 hectares, current activity measured from the satellite comes out between 175 and 311 (almost 400 soccer pitches), depending on where you put the threshold - because a construction site is not one clean edge. Haul roads, stripped soil and stockpiles each read differently from felled trees.
On the Guardian: theguardian.com/environment/…
𝐓𝐡𝐞 𝐍𝐞𝐰 𝐒𝐜𝐢𝐞𝐧𝐜𝐞 𝐨𝐟 𝐌𝐚𝐩𝐬 - thenewscienceofmaps.com/
#satellite #remotesensing #gis #dataviz #finland
Edher Jimenez retweeted
We wrote a 37-page technical report on OCRia-1, our SOTA model for document parsing.
✅ It contains comprehensive evals against most other major open-weight OCR models (e.g. Infinity-Parser, LightOnOCR 3, Chandra 2, dots.mocr) on multiple OCR benchmarks: ParseBench, olmOCR-bench, PulseBench-Tab, fr-bench-pdf2md.
✅ It contains detailed visibility into our data and training pipeline, to give a glimpse into how we’re able to curate the right document datasets to push the evaluation metrics within dimensions that matter for doc parsing.
✅ It is blazing fast ⚡️ decoding 2x-4x+ faster than the latest open-weight models.
✅ If you’re building your own vision models, there are 24 pages worth of appendix details (!!) that showcase all of our experimental findings from training to evals!
Check it out: llamaindex.ai/ocria/ocria-1-…
Other links are in our blog! llamaindex.ai/blog/parse-ocr…
Today we’re releasing OCRia-1, a frontier VLM for document parsing that sets a new SOTA in accuracy, cost, and latency.
It not only sets a new high score on ParseBench (87.4), but also leads in three 3rd-party benchmarks: olmOCR-bench* (83.5 overall, 92.9 on tables!), PulseBench-Tab (84.1), and fr-bench-pdf2md (72.1).
OCRia-1 is really, really good at parsing complex tables: nested headers, merged cells, irregular grids, scanned tables, and more. It is also fantastic at parsing complex unlabeled charts (line, bar, pie, technical diagrams), and preserving complex formatting. Most real-world enterprise documents contain at least one of the these, whether it’s a big tax withholding table or a public economic report.
It is a 4B model and outperforms all other models within a similar parameter class: LightOnOCR-3, Chandra-OCR 2, Mistral OCR 4.1, and more. It is also extremely fast.
All credit goes to our research team, who spent countless hours on curating a diverse dataset, tuning our training pipeline, and squeezing every last drop out of model optimization so that it can process billions of pages at scale.
OCRia-1 as a base is powering all of our modes within LlamaParse (cost-effective, agentic, agentic plus), so you can access its capabilities today!
If you’re processing millions/billions of documents at scale, we’re offering enterprises that book a demo up to $3k in credits for a guided evaluation.**
*olmOCR-bench scores are adjusted based on unifying rendering at 1,288px for all models
**enterprise POC credits apply to eligible organizations that book a demo
Edher Jimenez retweeted
OlmoEarth v1.3 is now on @huggingface 🤗 🌎
It can create embeddings for the whole planet at 10-meter resolution, which can be used for:
🔸 Mapping crops and land cover
🔸 Tracking change over time (mining, deforestation)
🔸 Exploring places on Earth
Congrats to the @allen_ai team on this super exciting new release!
Edher Jimenez retweeted
オントロジーを使った 特許の「似ている理由」まで見える技術俯瞰マップ ― KareSanScapeベータ版を公開しました lexi2vent.com/posts/MnNYmh5r #lexi2vent
Edher Jimenez retweeted
Introducing Rothko, a foundation model of human behaviour designed from the ground up to simulate the world’s most consequential decisions.
Rothko is the result of over three years of frontier research at the intersection of behavioural science and machine learning.
We’re only just starting to uncover the model’s value across domains ranging from market research to AI safety and swarm alignment.
Edher Jimenez retweeted
AI agents are writing more code. How can developers ensure that code is correct?
Microsoft Research’s Proof-Driven Development (PDD) Kit explores a more rigorous approach: it translates a developer’s intent into a formal specification, then uses formal verification to prove that generated code meets it.
PDD Kit works with Verus for Rust and is integrated with @GitHubNext’s Chopin, bringing formal methods into an AI-assisted development workflow for teams.
See the work in action at @github Universe, Oct. 28–29, 2026.
Edher Jimenez retweeted
Frontier labs are advancing the capabilities in AI for biology but preventing the good guys (scientists, academia, non-profits, health agencies,...) to use, study and understand it with widespread but shallow guardrails that the bad guys are bypassing easily (blocked me to proofread this tweet for example 😅).
This will automatically create massive assymetry and risks in the coming years, just like in cybersecurity, especially now that frontier labs are building wet labs (HF cybersecurity scenario for biology when?).
This is why we created Carbon-X, the best open-source DNA model by midtraining Qwen3-14B on 300B tokens of DNA, English and interleaved data. We also used a hybrid tokenizer that tokenizes English with BPE and DNA with a 6-mer tokenizer.
As a result, Carbon-X outperforms all open language models we evaluated on five benchmarks of DNA understanding and generation. This is a 27-point improvement over Qwen3-14B, showing how much a language model can gain from DNA training. It is also API frontier level at 1,800x less cost than GPT 6.1 Sol!
We hope Carbon-X helps researchers explore DNA in new ways. Releasing it openly means anyone can study its capabilities, limitations, and risks. Let's make the AI biology frontier open-source to make it safe and impactful for the whole world (not just a few companies).
Apache 2.0 Model on @huggingface: huggingface.co/HuggingFaceBi…
