diff --git a/CITATION.cff b/CITATION.cff
index 79804c7..140233f 100644
--- a/CITATION.cff
+++ b/CITATION.cff
@@ -23,7 +23,28 @@ authors:
   - family-names: "Xie"
     given-names: "Zejian"
 date-released: "2025-06-30"
-identifiers:
-  - type: arXiv
-    value: "2506.23667"
-url: "https://arxiv.org/abs/2506.23667"
+preferred-citation:
+  type: generic
+  url: "https://arxiv.org/abs/2506.23667"
+  title: "L0: Reinforcement Learning to Become General Agents"
+  authors:
+    - family-names: "Zhang"
+      given-names: "Junjie"
+    - family-names: "Xi"
+      given-names: "Jingyi"
+    - family-names: "Song"
+      given-names: "Zhuoyang"
+    - family-names: "Lu"
+      given-names: "Junyu"
+    - family-names: "Ke"
+      given-names: "Yuhua"
+    - family-names: "Sun"
+      given-names: "Ting"
+    - family-names: "Yang"
+      given-names: "Yukun"
+    - family-names: "Zhang"
+      given-names: "Jiaxing"
+    - family-names: "Zhang"
+      given-names: "Songxin"
+    - family-names: "Xie"
+      given-names: "Zejian"
diff --git a/README.md b/README.md
index 2f0f99f..00a1677 100644
--- a/README.md
+++ b/README.md
@@ -47,7 +47,6 @@ L0 is a scalable, end-to-end training pipeline for general-purpose agents. We pr
     - [For developing and testing](#for-developing-and-testing)
   - [⚠️ Known Issues](#️-known-issues)
   - [🙏 Acknowledgments](#-acknowledgments)
-  - [📖 Citation](#-citation)