Added a section on data quality aspect

README.md

@@ -99,6 +99,19 @@ Because answers are deterministic, evaluation can be performed via:
 
 ---
 
+## Data Quality Assurance Process
+
+To ensure scientific validity of the answer, all tasks are prepared and reviewed twice by PhD experts.
+
+Key quality rubrics include:
+
+* Prompt and answer accuracy 
+* Clarity of prompt and underlying reasoning 
+* Expert-verified model breaking cases due to model’s incorrect reasoning process
+* Google-proof originality validation.
+
+---
+
 ## Reinforcement Learning and Outcome Supervision
 
 This dataset is designed to support **outcome-based reinforcement learning** for reasoning models.