Maziyar PANAHI Profile picture
Jan 15 10 tweets 4 min read Read on X
🚨 Another day, another drop!!!

I am releasing 8 new medical RL environments on Prime Hub!

From differential diagnosis to drug safety, these environments cover the full spectrum of clinical AI challenges.

Thread 🧵👇Image
1/ MedXpertQA - Expert-Level Medical Reasoning

Expert-level medical reasoning that's significantly harder than MedMCQA.

This env challenges models with complex clinical scenarios that require deep medical knowledge and multi-step reasoning. Image
2/ DDI - Drug-Drug Interaction Detection

Critical for pharmacovigilance and patient safety.

Models learn to identify dangerous drug combinations from clinical text using the DDI Corpus.

Essential for building AI that keeps patients safe from harmful medication interactions. Image
3/ ADE - Adverse Drug Event Detection

Detect adverse drug events from clinical narratives using ADE Corpus V2.

Models learn to identify when medications cause harmful side effects—a crucial capability for pharmacovigilance systems and patient monitoring. Image
4/ BioASQ - Biomedical Question Answering

Evidence-based medicine powered by PubMedQA.

Models learn to answer biomedical questions using peer-reviewed literature, bridging the gap between research and clinical practice.

The foundation for AI-assisted medical research. Image
5/ HealthFact - Medical Fact-Checking

Combat medical misinformation with the PubHealth dataset.

Models learn to verify health claims against scientific evidence, critical for public health in the age of viral misinformation.

AI as a defense against falsehoods. Image
6/ DDXPlus - Differential Diagnosis

Train models to predict diseases from patient symptoms and medical histories.

Emulates the diagnostic reasoning process that physicians use every day.

From symptoms to diagnosis: teaching AI to think like a doctor. Image
7/ RadQA - Diagnostic Radiology

Radiology-focused medical QA using MedQA-USMLE.

Models learn to interpret imaging findings and answer diagnostic questions—essential for AI-assisted radiology.

Building expertise in diagnostic imaging, one question at a time. Image
@PrimeIntellect 8/ ICD10 - Medical Coding

Train models to assign accurate ICD-10 diagnostic codes to clinical notes.

Critical for healthcare billing, EHR systems, and clinical documentation.

Automating the complex task of translating clinical notes into standardized codes. Image
@PrimeIntellect Closing Post:

All 8 environments are now live on Prime Intellect! 🚀

Each includes:
✅ Built-in verifiers
✅ Automated reward functions
✅ Train/eval splits
✅ Single-turn format with thinking support

Start training medical AI models today.
app.primeintellect.ai/dashboard/user…Image

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Maziyar PANAHI

Maziyar PANAHI Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @MaziyarPanahi

Jan 13
🚨 OpenMed just mass-released 35 state-of-the-art PII detection models to the open-source community!

All Apache 2.0. All free. Forever. 🍀

Here's what @OpenMed_AI built and why it matters for healthcare AI safety. Supporting HIPAA, GDPR, and beyond.

Thread 🧵👇Image
The results speak for themselves:

- 96.08% F1 on our best model (DeBERTa-v3-large)
- Top 10 models all above 95.7% F1
- 35 models ranging from 33M to 600M parameters
- Built on top of Nemotron-PII by Nvidia

This isn't a single model release. It's an entire ecosystem.Image
OpenMed's PII models detect 54 types of sensitive information:

Identifiers: SSN, passport, medical record numbers, credit cards, API keys, passwords...

Personal: Names, DOB, age, gender, occupation, blood type, biometrics...

Location: Addresses, cities, countries, GPS coordinates, zip codes...

Digital: IP addresses, MAC addresses, emails, phone numbers, URLs...

Built for real-world de-identification.Image
Read 9 tweets
Dec 11, 2025
🚨 New open dataset for real medical AI.

⚕️ 200K step-by-step clinical reasoning chats (537M+ tokens)

🤖 Generated with OpenAI’s gpt-oss-120B, reasoning effort set to “high”

🏥 Built for training medical reasoning LLMs

Available on @huggingface 🧵👇Image
Medical-Reasoning-SFT-GPT-OSS-120B

- features step-by-step clinical reasoning
- differential diagnosis, treatment planning, and medical education
- clinical cases, physiology, pharmacology, diagnostics

📊 Stats:
- 537M+ tokens,
- avg 3K tokens/conversation Image
gpt-oss-120B with high reasonin over Qwen3-235B-A22B

gpt-oss-120B excels in:
• complex multi-step clinical reasoning
• accurate medical terminology & protocols
• consistent diagnostic accuracy
• superior handling of rare conditionsImage
Read 6 tweets
Oct 23, 2025
what happened this week with OCR and VLMs?

* deepseek-ocr
* chandra-ocr
* nanonets-ocr2
* paddleocr-vl
* qwen3-vl (2B, 32B, Instruct and Thinking)
* dots.ocr
* olmOCR 2 (based on Qwen2.5-VL)
* LightOnOCR (smallies)

top 5 trending models on @huggingface are still OCR/VLM! Image
@huggingface DeepSeek-OCR:

huggingface.co/deepseek-ai/De…
@huggingface chandra (9B):

huggingface.co/datalab-to/cha…
Read 15 tweets
Sep 8, 2025
Introducing MultiCaRe, open-source, multimodal clinical case datasets on @HuggingFace by @OpenMed_AI Community. Public and ready for load_dataset.

Images: 160K+ figures/subimages

Cases: 85K de-identified narratives + demographics

Articles: 85K metadata + abstracts

🧵 (1/7)Image
Why MultiCaRe?

- One place to connect Images ↔ Cases ↔ Articles
- Stable IDs and clean joins
- Built for classification, retrieval, grounding, VQA/doc-QA, and multimodal modeling

(2/7) Image
Start in seconds (no token):

(3/7) Image
Read 7 tweets
Sep 17, 2024
Self-Harmonized Chain of Thought (ECHO) 🧵

• @vllm_project for local inference on Llama-3.1-70B Instruct
• @GroqInc for blazing fast end-to-end testing
• @huggingface Inference Endpoints for Cohere Command R+ comparisons
• @Gradio for an intuitive, responsive UI

🚀🔥
The implementation process was fascinating. I started with question generation, moved to clustering, and then dove into the iterative demonstration unification. Each step presented unique challenges and opportunities for optimization.Image
One of the most interesting aspects was seeing how different models handled the ECHO process. The performance improvements were notable, especially in complex reasoning tasks.Image
Image
Read 7 tweets
Jun 25, 2024
1/4 🚀 Exciting news for AI enthusiasts! Check out NuExtract, a cutting-edge LLM designed for structured extraction tasks. It transforms any text into a structured output with just a template!

🤗 Open-source and available on @huggingface !
🌟 More info: numind.ai/blog/nuextract…
Image
2/4 🤖 NuExtract comes in three versions:

• NuExtract-tiny (0.5B) -
• NuExtract (3.8B) -
• NuExtract-large (7B) -

Get started with the one that suits your needs best! 💡huggingface.co/numind/NuExtra…
huggingface.co/numind/NuExtra…
huggingface.co/numind/NuExtra…
3/4 🛠️ This project utilized Llama-3 70B to annotate 50k documents, fine-tuning Phi-3-mini, Phi-3-small, and Qwen1.5-0.5B. Overcoming challenges like reducing hallucinations was key.

Dive into the journey here: numind.ai/blog/nuextract…
Read 5 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(