diff --git a/policy/feudalgainRL/README.md b/policy/feudalgainRL/README.md
index 59bc4080af35e0b046b12aa29639c51636da7581..c3b89e4ab2af9e9d52fdd4a9d83251929d5e0bb8 100644
--- a/policy/feudalgainRL/README.md
+++ b/policy/feudalgainRL/README.md
@@ -2,7 +2,7 @@
 
 This is the implementation to the work **What Does The User Want? Information Gain for Hierarchical Dialogue Policy Optimisation**, published at ASRU2021.
 
-Reference: 
+Reference: https://arxiv.org/pdf/2109.07129.pdf
 
 #### Train a FeudalGain policy
 
@@ -44,4 +44,4 @@ js_threshold = 1.0
 architecture = duel
 ```
 
-Log files and policies will be saved in the directories specified in the config file in section **[exec_config]**.
\ No newline at end of file
+Log files and policies will be saved in the directories specified in the config file in section **[exec_config]**.