Update README.md

README.md

@@ -2,32 +2,46 @@
 license: apache-2.0
 ---
 
-# Experimental pXRD Dataset (Subset of opXRD)
 
-This repository open-sources **labeled experimental powder X-ray diffraction (pXRD) data**, released as a **subset of the [opXRD]((ttps://zenodo.org/records/14254270）) database**, with data contributions from **HKUST (Guangzhou)**.
+# Experimental pXRD Dataset
+
+This repository open-sources **labeled experimental powder X-ray diffraction (pXRD) data**, released as a **curated subset of the [opXRD](https://zenodo.org/records/14254270) database**, with experimental data contributed by **HKUST (Guangzhou)**.
+
+The dataset is designed for **materials informatics, phase identification, and AI-driven XRD analysis**, and is made available for **research and educational purposes**.
+
 
-The dataset is intended for **materials informatics, phase identification, and AI-driven XRD analysis**, and is released for research and educational use.
 
----
 
 ## Dataset Composition
 
 The repository contains a total of **1,277 experimental pXRD patterns**, including:
 
-- **21 original pXRD patterns** from **[XRed](https://github.com/WPEM/XRED)**  
-  *(HKUST-A dataset)*
-
-- **499 labeled oxpXRD samples**  
+* **499 selected, labeled experimental opXRD samples (partial)**
   *(HKUST-B dataset)*
 
-- **757 processed pXRD patterns**, consisting of:
-  - **651 pXRD patterns** processed from the **opXRD database**
-  - **106 supplemental experimental pXRD patterns** from **HKUST (Guangzhou)**
+* **Additional experimental pXRD patterns retrieved from the scientific literature**
+
+
+## Data Processing
+
+All diffraction patterns have been **preprocessed** to ensure data quality, including:
+
+* Selection of **low-noise patterns** with **well-resolved and narrow diffraction peaks**
+* **Background fitting and subtraction** to reduce baseline effects
+
+
+
+## Notes on Data Reliability
+
+This dataset is **fully open for research use**. However, due to experimental variability and limitations in labeling, the **absolute ground-truth accuracy of all samples cannot be strictly guaranteed**.
+
+For **benchmarking, validation, or standard testing**, we recommend using well-established reference datasets such as **standard opXRD releases** or the **RRUFF database**.
 
 ---
 
 
 
+
 ## Usage and Citation
 
 If you use this dataset in your research, **please cite following paper**: