update

example/README.md

@@ -54,6 +54,25 @@ One-click pipeline (prepare -> train -> export -> eval -> plot):
 python example/run_pipeline.py --device auto
 ```
 
+## Ablation: Feature Split Variants
+Generate alternative continuous/discrete splits (baseline/strict/loose):
+```
+python example/ablation_splits.py --data-glob "../../dataset/hai/hai-21.03/train*.csv.gz"
+```
+
+Then run prepare/train with a chosen split:
+```
+python example/prepare_data.py --split-path example/results/ablation_splits/split_strict.json
+python example/train.py --config example/config.json --device cuda
+```
+
+Update `example/config.json` to point `split_path` at the chosen split file.
+
+One-click ablation (runs baseline/strict/loose end-to-end):
+```
+python example/run_ablation.py --device cuda
+```
+
 ## Notes
 - Heuristic: integer-like values with low cardinality (<=10) are treated as
   discrete. All other numeric columns are continuous.