diff --git a/backend/apps/chat/curd/chat.py b/backend/apps/chat/curd/chat.py
index 1f8befdf..eaddc671 100644
--- a/backend/apps/chat/curd/chat.py
+++ b/backend/apps/chat/curd/chat.py
@@ -211,7 +211,7 @@ def format_json_list_data(origin_data: list[dict]):
                         if len(decimal_str) > 15:
                             value = str(value)
             _row[key] = value
-        data.append(_row)
+        data.append(DataFormat.normalize_qualified_sql_column_keys(_row))
 
     return data
 
@@ -253,6 +253,7 @@ def get_chart_data_ds(session: SessionDep,ds_id,sql):
         else:
             result = exec_sql(ds=datasource,sql=sql, origin_column=False)
             _data = DataFormat.convert_large_numbers_in_object_array(result.get('data'))
+            _data = DataFormat.normalize_qualified_sql_column_keys_in_object_array(_data)
             json_result['data'] = _data
             return json_result
     except Exception as e:
diff --git a/backend/apps/chat/task/llm.py b/backend/apps/chat/task/llm.py
index 7953b5e3..cf469b37 100644
--- a/backend/apps/chat/task/llm.py
+++ b/backend/apps/chat/task/llm.py
@@ -1318,6 +1318,7 @@ class LLMService:
                                                                                  'count': len(result.get('data'))})
 
             _data = DataFormat.convert_large_numbers_in_object_array(result.get('data'))
+            _data = DataFormat.normalize_qualified_sql_column_keys_in_object_array(_data)
             result["data"] = _data
 
             self.save_sql_data(session=_session, data_obj=result)
diff --git a/backend/common/utils/data_format.py b/backend/common/utils/data_format.py
index 1991fb3e..bfa9e88b 100644
--- a/backend/common/utils/data_format.py
+++ b/backend/common/utils/data_format.py
@@ -17,6 +17,34 @@ class DataFormat:
 
         return df_copy
 
+    @staticmethod
+    def normalize_qualified_sql_column_keys(row: dict) -> dict:
+        """Add unqualified keys for names like ``alias.column`` (Hive/MySQL return shape).
+
+        Chart bindings use the bare column name (``table_name``) while drivers may return
+        ``_u2.table_name``. Only adds ``short`` when absent to avoid clobbering real duplicates.
+        """
+        if not row:
+            return row
+        out = dict(row)
+        for k, v in row.items():
+            ks = str(k)
+            if "." not in ks:
+                continue
+            short = ks.rsplit(".", 1)[-1]
+            if short not in out:
+                out[short] = v
+        return out
+
+    @staticmethod
+    def normalize_qualified_sql_column_keys_in_object_array(obj_array: list) -> list:
+        if not obj_array:
+            return obj_array
+        return [
+            DataFormat.normalize_qualified_sql_column_keys(obj) if isinstance(obj, dict) else obj
+            for obj in obj_array
+        ]
+
     @staticmethod
     def convert_large_numbers_in_object_array(obj_array, int_threshold=1e15, float_threshold=1e10):
         """处理对象数组，将每个对象中的大数字转换为字符串"""
diff --git a/backend/templates/sql_examples/Hive.yaml b/backend/templates/sql_examples/Hive.yaml
index d4ef6e0e..813f6ab5 100644
--- a/backend/templates/sql_examples/Hive.yaml
+++ b/backend/templates/sql_examples/Hive.yaml
@@ -5,6 +5,7 @@ template:
       <note>
         1. 点号（.）不能包含在引号内，必须写成 `database`.`table`
         2. 即使标识符不含特殊字符或非关键字，也需强制加反引号
+        3. 在多表关联（JOIN）/ 多子查询的 SQL 中，只要多个表 / 子查询存在 同名字段，所有引用该字段的位置，必须显式指定 表别名
       </note>
     </rule>