diff --git a/convlab/util/custom_util.py b/convlab/util/custom_util.py
index cc49d733d25566d502b10bac71189e05cfbc3fef..5c6b0d33f1755cf443521307024d6e17280f060d 100644
--- a/convlab/util/custom_util.py
+++ b/convlab/util/custom_util.py
@@ -389,6 +389,7 @@ def evaluate(sess, num_dialogues=400, sys_semantic_to_usr=False, save_flag=False
                 complete = sess.evaluator.complete
                 task_succ = sess.evaluator.success
                 task_succ_strict = sess.evaluator.success_strict
+                break
         else:
             complete = 0
             task_succ = 0