diff --git a/README.md b/README.md
index 2abf04b0dbe7c1f69e9f589df82749c65e7e1e17..10ec80cadabab9b2d1cc09d034be0bce048ed697 100644
--- a/README.md
+++ b/README.md
@@ -212,8 +212,8 @@ The save/load and predict operations of a multi_head_trainer is the same as a tr
 
 For more implementation details with `multi_head_trainer`, see
 
-- [Joint training of dialogue intent recognition and slot filling for ATIS]()
-- [Learning reading comprehension auxilarized with mask language model for MRQA]() (初次发版先不用加)
+- [ATIS: joint training of dialogue intent recognition and slot filling]()
+- [MRQA: learning reading comprehension auxilarized with mask language model]() (初次发版先不用加)
 
 
 ## License