diff --git a/fe/fe-core/src/main/java/org/apache/doris/nereids/glue/translator/PhysicalPlanTranslator.java b/fe/fe-core/src/main/java/org/apache/doris/nereids/glue/translator/PhysicalPlanTranslator.java
index b51bb7a700..b7eebd4562 100644
--- a/fe/fe-core/src/main/java/org/apache/doris/nereids/glue/translator/PhysicalPlanTranslator.java
+++ b/fe/fe-core/src/main/java/org/apache/doris/nereids/glue/translator/PhysicalPlanTranslator.java
@@ -646,14 +646,10 @@ public class PhysicalPlanTranslator extends DefaultPlanVisitor<PlanFragment, Pla
         outputList.forEach(k -> {
             sortTupleOutputList.add(ExpressionTranslator.translate(k, context));
         });
-        // 2. Generate new Tuple
-        TupleDescriptor tupleDesc = generateTupleDesc(outputList, orderKeyList, context, null);
-        // 3. Get current slotRef
+        // 2. Generate new Tuple and get current slotRef for newOrderingExprList
         List<Expr> newOrderingExprList = Lists.newArrayList();
-        orderKeyList.forEach(k -> {
-            newOrderingExprList.add(ExpressionTranslator.translate(k.getExpr(), context));
-        });
-        // 4. fill in SortInfo members
+        TupleDescriptor tupleDesc = generateTupleDesc(outputList, orderKeyList, newOrderingExprList, context, null);
+        // 3. fill in SortInfo members
         SortInfo sortInfo = new SortInfo(newOrderingExprList, ascOrderList, nullsFirstParamList, tupleDesc);
         PlanNode childNode = childFragment.getPlanRoot();
         SortNode sortNode = new SortNode(context.nextPlanNodeId(), childNode, sortInfo, true);
@@ -1445,6 +1441,7 @@ public class PhysicalPlanTranslator extends DefaultPlanVisitor<PlanFragment, Pla
     }
 
     private TupleDescriptor generateTupleDesc(List<Slot> slotList, List<OrderKey> orderKeyList,
+            List<Expr> newOrderingExprList,
             PlanTranslatorContext context, Table table) {
         TupleDescriptor tupleDescriptor = context.generateTupleDesc();
         Set<ExprId> alreadyExists = Sets.newHashSet();
@@ -1458,9 +1455,11 @@ public class PhysicalPlanTranslator extends DefaultPlanVisitor<PlanFragment, Pla
             }
             // TODO: trick here, we need semanticEquals to remove redundant expression
             if (alreadyExists.contains(slotReference.getExprId())) {
+                newOrderingExprList.add(context.findSlotRef(slotReference.getExprId()));
                 continue;
             }
             context.createSlotDesc(tupleDescriptor, slotReference);
+            newOrderingExprList.add(context.findSlotRef(slotReference.getExprId()));
             alreadyExists.add(slotReference.getExprId());
         }
         for (Slot slot : slotList) {
diff --git a/regression-test/suites/nereids_syntax_p0/test_join_nereids.groovy b/regression-test/suites/nereids_syntax_p0/test_join_nereids.groovy
index 7b25521191..8a17fe34db 100644
--- a/regression-test/suites/nereids_syntax_p0/test_join_nereids.groovy
+++ b/regression-test/suites/nereids_syntax_p0/test_join_nereids.groovy
@@ -125,6 +125,7 @@ suite("test_join_nereids") {
     sql 'set enable_vectorized_engine=true;'
     sql 'set enable_fallback_to_original_planner=false;'
     sql 'set enable_nereids_planner=true;'
+    sql 'set parallel_fragment_exec_instance_num = 2;'
 
     qt_agg_sql1 """select /*+SET_VAR(disable_nereids_rules='TWO_PHASE_AGGREGATE_WITH_COUNT_DISTINCT_MULTI')*/ count(distinct k1, NULL) from test;"""
     qt_agg_sql2 """select /*+SET_VAR(disable_nereids_rules='TWO_PHASE_AGGREGATE_WITH_COUNT_DISTINCT_MULTI')*/ count(distinct k1, NULL), avg(k2) from baseall;"""
@@ -285,9 +286,11 @@ suite("test_join_nereids") {
             on a.k1 = b.k1 and a.k2 > 0 order by 1, isnull(b.k1), 2, 3, 4, 5 limit 65535"""
     qt_left_join2"""select ${i} from ${tbName1} a left join ${tbName2} b 
             on a.k1 = b.k1 and a.k2 > b.k2 order by 1, isnull(b.k1), 2, 3, 4, 5 limit 65535"""
+    sql 'set batch_size = 3;'
     qt_left_join3"""select ${i} from ${tbName1} a left join ${tbName2} b on a.k2 = b.k2 and a.k1 > 0 
             left join ${tbName3} c on a.k3 = c.k3 and b.k1 = c.k1 + 1 and c.k3 > 0 
             order by 1, isnull(b.k1), 2, 3, 4, 5 limit 65535"""
+    sql 'set batch_size = 4064;'
 
     for (s in selected) {
         qt_left_join4"""select ${s} from ${tbName1} a left join ${tbName2} b