diff --git a/src/llamafactory/hparams/parser.py b/src/llamafactory/hparams/parser.py
index c22c1e375..5602bd26a 100644
--- a/src/llamafactory/hparams/parser.py
+++ b/src/llamafactory/hparams/parser.py
@@ -180,6 +180,13 @@ def _verify_trackio_args(training_args: "TrainingArguments") -> None:
         logger.warning("Consider setting --run_name for better experiment tracking clarity.")
 
 
+def _normalize_swanlab_args(training_args: "TrainingArguments", finetuning_args: "FinetuningArguments") -> None:
+    r"""Route SwanLab reporting through the native SwanLab callback."""
+    if "swanlab" in training_args.report_to:
+        training_args.report_to = [reporter for reporter in training_args.report_to if reporter != "swanlab"]
+        finetuning_args.use_swanlab = True
+
+
 def _set_transformers_logging() -> None:
     if os.getenv("LLAMAFACTORY_VERBOSITY", "INFO") in ["DEBUG", "INFO"]:
         transformers.utils.logging.set_verbosity_info()
@@ -373,6 +380,8 @@ def get_train_args(args: dict[str, Any] | list[str] | None = None) -> _TRAIN_CLS
         finetuning_args.use_mca = False
         finetuning_args.use_megatron_bridge = False
 
+    _normalize_swanlab_args(training_args, finetuning_args)
+
     # Setup logging
     if training_args.should_log:
         _set_transformers_logging()
@@ -553,9 +562,6 @@ def get_train_args(args: dict[str, Any] | list[str] | None = None) -> _TRAIN_CLS
         # https://github.com/huggingface/transformers/blob/v4.50.0/src/transformers/trainer.py#L782
         training_args.label_names = training_args.label_names or ["labels"]
 
-    if "swanlab" in training_args.report_to and finetuning_args.use_swanlab:
-        training_args.report_to.remove("swanlab")
-
     if (
         training_args.parallel_mode == ParallelMode.DISTRIBUTED
         and training_args.ddp_find_unused_parameters is None
diff --git a/tests/hparams/test_parser.py b/tests/hparams/test_parser.py
new file mode 100644
index 000000000..5397166c7
--- /dev/null
+++ b/tests/hparams/test_parser.py
@@ -0,0 +1,56 @@
+# Copyright 2026 the LlamaFactory team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from types import SimpleNamespace
+
+import pytest
+from transformers.training_args import ParallelMode
+
+from llamafactory.hparams import parser
+from llamafactory.hparams.parser import _normalize_swanlab_args
+
+
+@pytest.mark.parametrize(
+    ("report_to", "use_swanlab", "expected_report_to", "expected_use_swanlab"),
+    [
+        (["swanlab"], False, [], True),
+        (["tensorboard", "swanlab", "wandb"], False, ["tensorboard", "wandb"], True),
+        (["swanlab"], True, [], True),
+        (["tensorboard"], False, ["tensorboard"], False),
+    ],
+)
+def test_normalize_swanlab_args(report_to, use_swanlab, expected_report_to, expected_use_swanlab):
+    training_args = SimpleNamespace(report_to=report_to)
+    finetuning_args = SimpleNamespace(use_swanlab=use_swanlab)
+
+    _normalize_swanlab_args(training_args, finetuning_args)
+
+    assert training_args.report_to == expected_report_to
+    assert finetuning_args.use_swanlab is expected_use_swanlab
+
+
+def test_report_to_swanlab_uses_native_callback(monkeypatch, tmp_path):
+    monkeypatch.setattr(parser.TrainingArguments, "parallel_mode", property(lambda self: ParallelMode.DISTRIBUTED))
+    monkeypatch.setattr(parser, "check_version", lambda *args, **kwargs: None)
+
+    _, _, training_args, finetuning_args, _ = parser.get_train_args(
+        {
+            "model_name_or_path": "dummy",
+            "output_dir": str(tmp_path),
+            "report_to": "swanlab",
+        }
+    )
+
+    assert training_args.report_to == []
+    assert finetuning_args.use_swanlab is True