diff --git a/src/operator/tensor/sparse_retain.cc b/src/operator/tensor/sparse_retain.cc
index 007541b00b37..b1f38c6d13b1 100644
--- a/src/operator/tensor/sparse_retain.cc
+++ b/src/operator/tensor/sparse_retain.cc
@@ -31,7 +31,7 @@ namespace op {
 // accepts row-sparse format ndarrays. It will be registered
 // under mxnet.ndarray.sparse with name retain.
 NNVM_REGISTER_OP(_sparse_retain)
-.describe(R"code(pick rows specified by user input index array from a row sparse matrix
+.describe(R"code(Pick rows specified by user input index array from a row sparse matrix
 and save them in the output sparse matrix.
 
 Example::
@@ -39,10 +39,10 @@ Example::
   data = [[1, 2], [3, 4], [5, 6]]
   indices = [0, 1, 3]
   shape = (4, 2)
-  rsp_in = row_sparse(data, indices)
+  rsp_in = row_sparse_array(data, indices)
   to_retain = [0, 3]
   rsp_out = retain(rsp_in, to_retain)
-  rsp_out.values = [[1, 2], [5, 6]]
+  rsp_out.data = [[1, 2], [5, 6]]
   rsp_out.indices = [0, 3]
 
 The storage type of ``retain`` output depends on storage types of inputs