diff --git a/backend/app/storage/plan_session_storage.py b/backend/app/storage/plan_session_storage.py
index c6c0f3da..6b99cde2 100644
--- a/backend/app/storage/plan_session_storage.py
+++ b/backend/app/storage/plan_session_storage.py
@@ -84,7 +84,7 @@ def _deserialize(self, data: Dict[str, Any]) -> PlanSession:
 
     def create(self, session: PlanSession) -> PlanSession:
         path = self._path(session.id)
-        with open(path, 'w') as f:
+        with open(path, 'w', encoding='utf-8') as f:
             json.dump(self._serialize(session), f, indent=2, default=str)
         return session
 
@@ -92,7 +92,7 @@ def get(self, session_id: str) -> Optional[PlanSession]:
         path = self._path(session_id)
         if not path.exists():
             return None
-        with open(path, 'r') as f:
+        with open(path, 'r', encoding='utf-8') as f:
             return self._deserialize(json.load(f))
 
     def update(self, session: PlanSession) -> PlanSession:
@@ -102,7 +102,7 @@ def list_all(self, status: Optional[PlanSessionStatus] = None) -> List[PlanSessi
         sessions = []
         for p in sorted(self.base_path.glob("psess_*.json"), reverse=True):
             try:
-                with open(p, 'r') as f:
+                with open(p, 'r', encoding='utf-8') as f:
                     s = self._deserialize(json.load(f))
                 if status is None or s.status == status:
                     sessions.append(s)
@@ -136,7 +136,7 @@ def _deserialize(self, data: Dict[str, Any]) -> CandidatePlan:
 
     def create(self, candidate: CandidatePlan) -> CandidatePlan:
         path = self._path(candidate.id)
-        with open(path, 'w') as f:
+        with open(path, 'w', encoding='utf-8') as f:
             json.dump(self._serialize(candidate), f, indent=2, default=str)
         return candidate
 
@@ -144,14 +144,14 @@ def get(self, candidate_id: str) -> Optional[CandidatePlan]:
         path = self._path(candidate_id)
         if not path.exists():
             return None
-        with open(path, 'r') as f:
+        with open(path, 'r', encoding='utf-8') as f:
             return self._deserialize(json.load(f))
 
     def list_by_session(self, session_id: str) -> List[CandidatePlan]:
         candidates = []
         for p in sorted(self.base_path.glob("cplan_*.json")):
             try:
-                with open(p, 'r') as f:
+                with open(p, 'r', encoding='utf-8') as f:
                     c = self._deserialize(json.load(f))
                 if c.sessionId == session_id:
                     candidates.append(c)
diff --git a/backend/tests/test_pr_04_plan_session_encoding.py b/backend/tests/test_pr_04_plan_session_encoding.py
new file mode 100644
index 00000000..ed76b52c
--- /dev/null
+++ b/backend/tests/test_pr_04_plan_session_encoding.py
@@ -0,0 +1,81 @@
+"""Test that plan_session_storage.py handles non-ASCII content via encoding=utf-8."""
+
+from datetime import datetime
+
+import pytest
+
+from app.models.plan_session import (
+    PlanSession,
+    PlanSessionConfig,
+    PlanSessionStatus,
+    CandidatePlan,
+)
+from app.storage.plan_session_storage import PlanSessionStorage, CandidatePlanStorage
+
+
+@pytest.fixture
+def tmp_data_dir(tmp_path):
+    return str(tmp_path)
+
+
+def test_session_roundtrip_non_ascii(tmp_data_dir):
+    """Create a PlanSession with non-ASCII config and verify round-trip."""
+    storage = PlanSessionStorage(tmp_data_dir)
+    config = PlanSessionConfig(
+        userNotes="优化模型性能并提高代码覆盖率",
+        paperType="algorithmic_method",
+    )
+    session = PlanSession(
+        id="psess_test001",
+        config=config,
+        status=PlanSessionStatus.PENDING,
+        createdAt=datetime.now(),
+    )
+
+    storage.create(session)
+    fetched = storage.get("psess_test001")
+
+    assert fetched is not None
+    assert fetched.config.userNotes == "优化模型性能并提高代码覆盖率"
+
+
+def test_list_all_non_ascii(tmp_data_dir):
+    """Verify list_all returns sessions with non-ASCII content correctly."""
+    storage = PlanSessionStorage(tmp_data_dir)
+
+    for i, note in enumerate(["First note", "Deuxième note", "第三笔记"]):
+        config = PlanSessionConfig(userNotes=note, paperType="algorithmic_method")
+        session = PlanSession(
+            id=f"psess_test{i:03d}",
+            config=config,
+            status=PlanSessionStatus.PENDING,
+            createdAt=datetime.now(),
+        )
+        storage.create(session)
+
+    sessions = storage.list_all()
+    assert len(sessions) == 3
+    notes = {s.config.userNotes for s in sessions}
+    assert "First note" in notes
+    assert "Deuxième note" in notes
+    assert "第三笔记" in notes
+
+
+def test_candidate_roundtrip_non_ascii(tmp_data_dir):
+    """Create a CandidatePlan with non-ASCII content and verify round-trip."""
+    storage = CandidatePlanStorage(tmp_data_dir)
+    candidate = CandidatePlan(
+        id="cplan_test001",
+        sessionId="psess_001",
+        indexNumber=1,
+        title="使用遗传算法优化超参数配置",
+        planAbstract="本文提出了一种新的方法",
+        createdAt=datetime.now(),
+    )
+
+    storage.create(candidate)
+    fetched = storage.get("cplan_test001")
+
+    assert fetched is not None
+    assert fetched.title == "使用遗传算法优化超参数配置"
+    assert fetched.planAbstract == "本文提出了一种新的方法"