Datasets:

SwarmOS
/

profile

Name: profile
Creator: Swarm & Bee

The dataset is currently empty. Upload or create new data files. Then, you will be able to explore them in the Dataset Viewer.

YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/datasets-cards)

Swarm & Bee (SwarmOS)

Sovereign AI infrastructure. Specialized models. Your data never leaves your rack.

We build last-mile intelligence products — fine-tuned models, verified training data, and edge inference systems for regulated industries.

Product	Description
Medical Models	Specialized 7B-72B clinical models — pharmacology, cardiology, radiology, emergency medicine
Platinum Data	CoVe-verified QA training pairs. Every claim independently fact-checked by 235B parameter models
Sovereign Inference	Single-tenant GPU deployments. Your model, your rack, your data
CreditSniper	Litigation-grade credit dispute intelligence — FCRA/FDCPA violation analysis

Model	Base	Training Data	Key Result
SwarmMedQA-7B-v1	Qwen2.5-7B	124 gold clinical QA	MedQA 66% (+4%), PubMedQA 53% (+53%)
SwarmMed-7B-v5	Qwen2.5-7B	1,191 platinum (CoVe-verified)	52.5% concept coverage, 94% disclaimer compliance
SwarmPharm-7B-v2	Qwen2.5-7B	867 pharma gold	Loss 0.008, 93% disclaimer rate
SwarmMed-MRI-7B-v1	Qwen2.5-VL-7B	1,026 gold MRI pairs	First sovereign vision-language medical model

Dataset	Size	Quality	Description
SwarmMed-Platinum-1K	1,000 pairs	Platinum (CoVe-verified)	Free clinical QA — pharmacology + cardiology
SwarmMedQA	165 pairs	Gold	Multi-specialty clinical QA with chain-of-thought
SwarmPharm	867 pairs	Gold	Pharmacology-focused QA

Harvest (10+ sources) → Generate (7B-72B) → 4-Stage Audit → CoVe Verification (235B) → Platinum

CoVe (Chain-of-Verification): Each factual claim independently verified by 235B parameter model. No hallucinations. No slop. Platinum grade.

Conversion rates: ~45% raw to gold, ~60% gold to platinum. Only verified facts make it to training.