diff --git a/python/scripts/nvbench_compare.py b/python/scripts/nvbench_compare.py
index 7dadcf5..1217bdc 100644
--- a/python/scripts/nvbench_compare.py
+++ b/python/scripts/nvbench_compare.py
@@ -2960,7 +2960,7 @@ def main() -> int:
     The number of detected regressions is reported in the summary output.
     """
     help_text = "%(prog)s [reference.json compare.json | reference_dir/ compare_dir/]"
-    parser = argparse.ArgumentParser(prog="nvbench_compare", usage=help_text)
+    parser = argparse.ArgumentParser(usage=help_text)
     parser.add_argument(
         "--ignore-devices",
         dest="ignore_devices",
@@ -3054,8 +3054,10 @@ def main() -> int:
         action=OrderedBenchmarkFilterAction,
         help="Filter by benchmark name (can repeat)",
     )
+    parser.add_argument("files_or_dirs", nargs="*")
 
-    args, files_or_dirs = parser.parse_known_args()
+    args = parser.parse_args()
+    files_or_dirs = args.files_or_dirs
     try:
         comparison_preset, comparison_thresholds = resolve_comparison_thresholds(
             args.preset, args.config
diff --git a/python/test/test_nvbench_compare.py b/python/test/test_nvbench_compare.py
index a9ea5b6..96e4a67 100644
--- a/python/test/test_nvbench_compare.py
+++ b/python/test/test_nvbench_compare.py
@@ -1969,6 +1969,19 @@ def test_main_dump_config_merges_config_and_cli_preset(
     assert "sample_coverage = 0.93" in output
 
 
+def test_main_rejects_unknown_options(monkeypatch, nvbench_compare):
+    monkeypatch.setattr(
+        sys,
+        "argv",
+        ["nvbench_compare", "--dispaly", "explain", "ref.json", "cmp.json"],
+    )
+
+    with pytest.raises(SystemExit) as exc_info:
+        nvbench_compare.main()
+
+    assert exc_info.value.code == 2
+
+
 def test_main_prints_bulk_debug_python_to_stdout(monkeypatch, capsys, nvbench_compare):
     devices = [{"id": 0, "name": "Test GPU"}]
     root = {