diff --git a/src/graphnet/data/dataset/dataset.py b/src/graphnet/data/dataset/dataset.py
index 49329e4ac..838ea40b7 100644
--- a/src/graphnet/data/dataset/dataset.py
+++ b/src/graphnet/data/dataset/dataset.py
@@ -141,7 +141,16 @@ def from_config(  # type: ignore[override]
             cfg["graph_definition"] = parse_graph_definition(cfg)
         if cfg["labels"] is not None:
             cfg["labels"] = parse_labels(cfg)
-        return source._dataset_class(**cfg)
+
+        if isinstance(cfg["path"], list):
+            sources = []
+            for path in cfg["path"]:
+                cfg["path"] = path
+                sources.append(source._dataset_class(**cfg))
+            source = EnsembleDataset(sources)
+            return source
+        else:
+            return source._dataset_class(**cfg)
 
     @classmethod
     def concatenate(
diff --git a/tests/utilities/test_dataset_config.py b/tests/utilities/test_dataset_config.py
index f54e37840..b3d6aec11 100644
--- a/tests/utilities/test_dataset_config.py
+++ b/tests/utilities/test_dataset_config.py
@@ -269,3 +269,22 @@ def test_dataset_config_files(backend: str) -> None:
         )
         == 0
     )
+
+
+@pytest.mark.order(6)
+@pytest.mark.parametrize("backend", ["sqlite"])
+def test_multiple_dataset_config_dict_selection(backend: str) -> None:
+    """Test constructing Dataset with multiple data paths."""
+    # Arrange
+    config_path = CONFIG_PATHS[backend]
+
+    # Single dataset
+    config = DatasetConfig.load(config_path)
+    dataset = Dataset.from_config(config)
+    # Construct multiple datasets
+    config_ensemble = DatasetConfig.load(config_path)
+    config_ensemble.path = [config_ensemble.path, config_ensemble.path]
+
+    ensemble_dataset = Dataset.from_config(config_ensemble)
+
+    assert len(dataset) * 2 == len(ensemble_dataset)