diff --git a/pipeline/io/export.py b/pipeline/io/export.py
index 17b0065..5d1f225 100644
--- a/pipeline/io/export.py
+++ b/pipeline/io/export.py
@@ -83,7 +83,7 @@ def cache_jsonl(
     processor: BaseProcessor,
     *,
     pack_size: int = -1,
-    pad_value: int = 1,
+    pad_value: int = 0,
 ) -> List[str]:
     """Tokenize JSONL files and pack them into HDF5 storage.
 
diff --git a/scripts/cache_h5.py b/scripts/cache_h5.py
index 6cbfed5..8dc8899 100644
--- a/scripts/cache_h5.py
+++ b/scripts/cache_h5.py
@@ -51,7 +51,7 @@ def main():
         help="Pack size, <=0 to disable (default: -1)",
     )
     parser.add_argument(
-        "--pad-value", type=int, default=1, help="Padding value (default: 1)"
+        "--pad-value", type=int, default=0, help="Padding value (default: 0)"
     )
     parser.add_argument(
         "--log-level",