Audio, notes and documents. The story starts somewhere.
A lab result. A discharge summary. The picture changes.
Records connected across visits and cohort deliveries.
Source datasets for your AI lab customers, with agreed cohorts, quality criteria and delivery cadence.
Open the report. Find the encounter. Put the right thing in the right place. We’ll check the chart.
727 public tasks · 5 families · 315 Hard · 126 image · 189 private
Scored on the final record state. Offline. No LLM judge.
14 models on 99 private tasks. Three attempts per task. November 2026.
Small models. Unreasonably specific interests.
We sent a 4B model to the Gym. It developed a taste for healthcare admin.
CRS-4B · Qwen3.5-4B base → fine-tuned
99 private tasks · 3 attempts per task. Frontier models: 0.76–0.89.
System One for speech. Still backstage. You’ll hear more when it’s ready.
CRS-4B is fine-tuned only on public synthetic Gym episodes and evaluated on the private split.
01
Source difficulty levels
CSV, HL7 v2, broken-table PDFs, superseded reports, non-standard units, ambiguous dates and same-name patients.
02
Difficulty bands
Difficulty bands for every task.
03
Clinical scores
Calculators implemented from the primary publications, including the correct refusal when inputs are missing.
04
Web EHR computer use
A synthetic web EHR with screenshots, coordinates, an accessibility tree and human approval steps, scored on the final FHIR state.
05
A larger public task set
Published as a versioned dataset.
Tell us what data you need. We’ll shape the cohort, format and delivery around it.