Update README.md

README.md

@@ -1,3 +1,119 @@
 ---
+language:
+- en
 license: mit
+task_categories:
+- text-classification
+tags:
+- clinical
+- clinical-reasoning
+- constraint-pressure
+- treatment-boundary
+- reserve-capacity
+- stability
+- evidence-fidelity
+size_categories:
+- n<1K
+pretty_name: Clinical Constraint Pressure
 ---
+
+# What this dataset does
+
+This dataset tests whether a model can estimate how close a patient is to a treatment boundary.
+
+The task is not to predict diagnosis.
+
+The task is to classify pressure on the patient system.
+
+# Core stability idea
+
+A patient may be stable at the present moment while operating close to a boundary.
+
+Constraint pressure increases when support needs rise and reserve capacity falls.
+
+The model must classify whether pressure is low, medium, or high.
+
+# Prediction target
+
+The label column has three classes.
+
+Label 0 means low constraint pressure.
+
+Label 1 means medium constraint pressure.
+
+Label 2 means high constraint pressure.
+
+# Row structure
+
+Each row contains:
+
+- scenario_id
+- oxygen_requirement
+- vasopressor_requirement
+- renal_function
+- urine_output
+- mental_status
+- respiratory_reserve
+- circulatory_reserve
+- renal_reserve
+- label
+
+oxygen_requirement uses:
+
+- 0 = room air or minimal support
+- 1 = low oxygen requirement
+- 2 = high oxygen requirement
+- 3 = near respiratory boundary
+
+vasopressor_requirement uses:
+
+- 0 = none
+- low
+- moderate
+- high
+
+reserve fields use:
+
+- high
+- medium
+- low
+- very_low
+
+# Evaluation
+
+Submissions must contain:
+
+```csv
+scenario_id,prediction
+test_001,0
+test_002,0
+test_003,1
+
+Run:
+
+python scorer.py predictions.csv
+
+Optional truth path:
+
+python scorer.py predictions.csv data/test.csv
+
+The scorer reports:
+
+Accuracy
+Macro precision
+Macro recall
+Macro F1
+Confusion matrix
+Structural Note
+
+This dataset tests boundary awareness.
+
+It is designed to prevent shortcut logic based only on current vital signs.
+
+The intended reasoning requires integration across support requirements, organ reserve, and proximity to treatment limits.
+
+The dataset does not expose the hidden rationale behind each label.
+
+License
+
+MIT