Add link to paper

README.md

@@ -1,20 +1,20 @@
 ---
-license: other
-license_name: tencent
 language:
 - en
 - zh
+license: other
+size_categories:
+- n<1K
 task_categories:
 - text-generation
+pretty_name: Tencent WorkBuddy Bench
+license_name: tencent
 tags:
 - code
 - agent
 - benchmark
 - swe
 - security
-pretty_name: Tencent WorkBuddy Bench
-size_categories:
-- n<1K
 ---
 
 # WorkBuddy Bench — Datasets
@@ -25,6 +25,8 @@ for evaluating coding agents on real-world developer, PM, algo, QA, ops, and
 security work. This repository hosts **only the task data**; the evaluation
 framework, setup, and usage all live in the GitHub repository above.
 
+Paper: https://huggingface.co/papers/2607.20911
+
 ### Dataset Summary
 
 WorkBuddy Bench is built from real-world work tasks. Given a task and a
@@ -62,4 +64,4 @@ extracts it into `/workspace` inside the container.
 
 These archives are consumed by the WorkBuddy Bench framework, not loaded
 directly. See the [GitHub repository](https://github.com/Tencent/workbuddy-bench)
-for how to download, extract, and run evaluations.
+for how to download, extract, and run evaluations.