diff --git a/paddle/operators/positive_negative_pair_op.cc b/paddle/operators/positive_negative_pair_op.cc
index b234e9c0de13a074787c15d8e2493dc22825b684..f740af1859c5d18c6f50899e07319ef7c5a6c1e0 100644
--- a/paddle/operators/positive_negative_pair_op.cc
+++ b/paddle/operators/positive_negative_pair_op.cc
@@ -129,7 +129,10 @@ class PositiveNegativePairOpMaker : public framework::OpProtoAndCheckerMaker {
         .AsDispensable();
     AddInput("Weight",
              "(float) Optional. Weight of current item. If specified, its "
-             "shape should be the same as Label.")
+             "shape should be the same as Label, and the meaning of the output "
+             "changes from numbers of pairs to the total sum of pairs' "
+             "weights. Weight of a pair of items is the average of their "
+             "weights.")
         .AsDispensable();
     AddOutput("PositivePair",
               "(float) Number of positive pairs, i.e. the pairs of "
@@ -150,9 +153,13 @@ class PositiveNegativePairOpMaker : public framework::OpProtoAndCheckerMaker {
         "Noting that reducing on the first dim will make the LoD info lost.")
         .SetDefault(0);
     AddComment(R"DOC(
-        PositiveNegativePairOp can be used to evaluate Learning To Rank(LTR) model performance. 
-        Within some context, e.g. the "query", a LTR model generates scores for a list of items, which gives a partial order of the items. 
-        PositiveNegativePairOp takes a list of reference rank order (Input("Label")) and the model generated scores (Input(Score)) as inputs and counts the pairs that ranked correctly and incorrectly.
+        PositiveNegativePairOp can be used to evaluate Learning To Rank(LTR) 
+        model performance. 
+        Within some context, e.g. the "query", a LTR model generates scores
+        for a list of items, which gives a partial order of the items.
+        PositiveNegativePairOp takes a list of reference rank order 
+        (Input("Label")) and the model generated scores (Input(Score)) as 
+        inputs and counts the pairs that ranked correctly and incorrectly.
 )DOC");
   }
 };
diff --git a/python/paddle/v2/framework/tests/test_positive_negative_pair_op.py b/python/paddle/v2/framework/tests/test_positive_negative_pair_op.py
index 64438c09a65a32b4ce16e432643b6f1f452aa0f0..cbd05a4f51f913a82496345ce270deff6cfa7070 100644
--- a/python/paddle/v2/framework/tests/test_positive_negative_pair_op.py
+++ b/python/paddle/v2/framework/tests/test_positive_negative_pair_op.py
@@ -12,9 +12,6 @@ def py_pnpair_op(score, label, query, column=-1, weight=None):
     if weight is None:
         weight = np.ones(shape=(batch_size, 1)).astype('float32')
     for s, l, q, w in zip(score, label, query, weight):
-        # s = s[column]
-        # q = q[0]
-        # w = w[0]
         s, l, q, w = s[column], l[0], q[0], w[0]
         if q not in predictions:
             predictions[q] = []