diff --git a/scripts/assess_belgium_building_training_iteration.py b/scripts/assess_belgium_building_training_iteration.py
index 5063576f..7fed75c1 100644
--- a/scripts/assess_belgium_building_training_iteration.py
+++ b/scripts/assess_belgium_building_training_iteration.py
@@ -39,7 +39,7 @@ INFERENCE_CONFIG_FIELDS = (
     "max_detections_per_tile", "nms_iou", "containment_nms", "box_scale",
     "box_offset_x", "box_offset_y", "additional_model", "ensemble_mode",
     "ensemble_match_iou", "proposal_classifier", "proposal_classifier_threshold",
-    "proposal_crop_scale", "proposal_classifier_batch",
+    "proposal_crop_scale", "proposal_classifier_batch", "regional_models",
 )
 
 
diff --git a/tests/test_assess_belgium_building_training_iteration.py b/tests/test_assess_belgium_building_training_iteration.py
new file mode 100644
index 00000000..74125e28
--- /dev/null
+++ b/tests/test_assess_belgium_building_training_iteration.py
@@ -0,0 +1,26 @@
+import pytest
+
+from scripts.assess_belgium_building_training_iteration import (
+    INFERENCE_CONFIG_FIELDS,
+    assert_same_inference_config,
+)
+
+
+def test_regional_model_routes_are_frozen_in_inference_config() -> None:
+    assert "regional_models" in INFERENCE_CONFIG_FIELDS
+    calibration = {field: None for field in INFERENCE_CONFIG_FIELDS}
+    calibration["regional_models"] = {"flanders": "/models/flanders.pt"}
+    test = dict(calibration)
+    test["regional_models"] = {"flanders": "/models/other.pt"}
+
+    with pytest.raises(ValueError, match="regional_models"):
+        assert_same_inference_config(calibration, test, "test")
+
+
+def test_regional_tile_counts_may_differ_between_splits() -> None:
+    calibration = {field: None for field in INFERENCE_CONFIG_FIELDS}
+    test = dict(calibration)
+    calibration["regional_model_tile_counts"] = {"flanders": 48}
+    test["regional_model_tile_counts"] = {"flanders": 64}
+
+    assert_same_inference_config(calibration, test, "test")