Update README.md

README.md

@@ -24,7 +24,7 @@ Each sample contains:
 * perturbation_type (original / lexical_rephrase / symptom_removed / contradictory_signal)
 
 
-### Task Definition
+## Task Definition
 
 Given a vignette, models must predict:
 * Risk tier
@@ -34,7 +34,7 @@ Given a vignette, models must predict:
 
 
 
-### Intended Use
+## Intended Use
 
 PulmoBench is designed for evaluating:
 * Clinical reasoning robustness
@@ -43,6 +43,31 @@ PulmoBench is designed for evaluating:
 * Escalation reliability
 Not intended for real-world medical deployment.
 
+## Test loading:
+```python
+  from datasets import load_dataset
+  dataset = load_dataset("saibhossain/PulmoBench")
+  print(dataset)
+```
+
+output :
+
+```json
+  DatasetDict({
+      train: Dataset({
+          features: ['case_id', 'age', 'sex', 'risk_tier', 'risk_score', 'urgency', 'gold_differential', 'escalation_required', 'vignette', 'perturbation_type'],
+          num_rows: 2544
+      })
+      validation: Dataset({
+          features: ['case_id', 'age', 'sex', 'risk_tier', 'risk_score', 'urgency', 'gold_differential', 'escalation_required', 'vignette', 'perturbation_type'],
+          num_rows: 544
+      })
+      test: Dataset({
+          features: ['case_id', 'age', 'sex', 'risk_tier', 'risk_score', 'urgency', 'gold_differential', 'escalation_required', 'vignette', 'perturbation_type'],
+          num_rows: 548
+      })
+  })
+```
 --
 
 ### Source Dataset