diff --git a/src/llamafactory/chat/hf_engine.py b/src/llamafactory/chat/hf_engine.py
index 1e670b92c9..478cd6a0f0 100644
--- a/src/llamafactory/chat/hf_engine.py
+++ b/src/llamafactory/chat/hf_engine.py
@@ -82,8 +82,9 @@ def _process_args(
         images: Optional[list["ImageInput"]] = None,
         videos: Optional[list["VideoInput"]] = None,
         audios: Optional[list["AudioInput"]] = None,
-        input_kwargs: Optional[dict[str, Any]] = {},
+        input_kwargs: Optional[dict[str, Any]] = None,
     ) -> tuple[dict[str, Any], int]:
+        input_kwargs = {} if input_kwargs is None else dict(input_kwargs)
         mm_input_dict = {"images": [], "videos": [], "audios": [], "imglens": [0], "vidlens": [0], "audlens": [0]}
         if images is not None:
             mm_input_dict.update({"images": images, "imglens": [len(images)]})
@@ -221,8 +222,9 @@ def _chat(
         images: Optional[list["ImageInput"]] = None,
         videos: Optional[list["VideoInput"]] = None,
         audios: Optional[list["AudioInput"]] = None,
-        input_kwargs: Optional[dict[str, Any]] = {},
+        input_kwargs: Optional[dict[str, Any]] = None,
     ) -> list["Response"]:
+        input_kwargs = {} if input_kwargs is None else dict(input_kwargs)
         gen_kwargs, prompt_length = HuggingfaceEngine._process_args(
             model,
             tokenizer,
@@ -276,8 +278,9 @@ def _stream_chat(
         images: Optional[list["ImageInput"]] = None,
         videos: Optional[list["VideoInput"]] = None,
         audios: Optional[list["AudioInput"]] = None,
-        input_kwargs: Optional[dict[str, Any]] = {},
+        input_kwargs: Optional[dict[str, Any]] = None,
     ) -> Callable[[], str]:
+        input_kwargs = {} if input_kwargs is None else dict(input_kwargs)
         gen_kwargs, _ = HuggingfaceEngine._process_args(
             model,
             tokenizer,
@@ -315,8 +318,9 @@ def _get_scores(
         model: "PreTrainedModelWrapper",
         tokenizer: "PreTrainedTokenizer",
         batch_input: list[str],
-        input_kwargs: Optional[dict[str, Any]] = {},
+        input_kwargs: Optional[dict[str, Any]] = None,
     ) -> list[float]:
+        input_kwargs = {} if input_kwargs is None else dict(input_kwargs)
         max_length: Optional[int] = input_kwargs.pop("max_length", None)
         device = getattr(model.pretrained_model, "device", "cuda")
         inputs: dict[str, torch.Tensor] = tokenizer(
diff --git a/tests/chat/test_hf_engine.py b/tests/chat/test_hf_engine.py
new file mode 100644
index 0000000000..2bcd0978bf
--- /dev/null
+++ b/tests/chat/test_hf_engine.py
@@ -0,0 +1,69 @@
+# Copyright 2025 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.
+
+import ast
+from pathlib import Path
+
+
+def test_huggingface_engine_input_kwargs_defaults_are_not_mutable():
+    module_path = Path(__file__).parents[2] / "src" / "llamafactory" / "chat" / "hf_engine.py"
+    module = ast.parse(module_path.read_text(encoding="utf-8"))
+
+    expected_methods = {"_process_args", "_chat", "_stream_chat", "_get_scores"}
+    methods = {
+        node.name: node
+        for node in ast.walk(module)
+        if isinstance(node, ast.FunctionDef) and node.name in expected_methods
+    }
+
+    assert set(methods) == expected_methods
+
+    for method in methods.values():
+        arguments = method.args.args
+        defaults = method.args.defaults
+        default_by_arg = dict(zip((arg.arg for arg in arguments[-len(defaults) :]), defaults))
+
+        default = default_by_arg["input_kwargs"]
+        assert isinstance(default, ast.Constant)
+        assert default.value is None
+
+
+def test_huggingface_engine_input_kwargs_are_copied_before_use():
+    module_path = Path(__file__).parents[2] / "src" / "llamafactory" / "chat" / "hf_engine.py"
+    module = ast.parse(module_path.read_text(encoding="utf-8"))
+
+    expected_methods = {"_process_args", "_chat", "_stream_chat", "_get_scores"}
+    methods = {
+        node.name: node
+        for node in ast.walk(module)
+        if isinstance(node, ast.FunctionDef) and node.name in expected_methods
+    }
+
+    for method in methods.values():
+        first_statement = method.body[0]
+        assert isinstance(first_statement, ast.Assign)
+        assert len(first_statement.targets) == 1
+        assert isinstance(first_statement.targets[0], ast.Name)
+        assert first_statement.targets[0].id == "input_kwargs"
+
+        value = first_statement.value
+        assert isinstance(value, ast.IfExp)
+        assert isinstance(value.test, ast.Compare)
+        assert isinstance(value.body, ast.Dict)
+        assert isinstance(value.orelse, ast.Call)
+        assert isinstance(value.orelse.func, ast.Name)
+        assert value.orelse.func.id == "dict"
+        assert len(value.orelse.args) == 1
+        assert isinstance(value.orelse.args[0], ast.Name)
+        assert value.orelse.args[0].id == "input_kwargs"