Update README.md

README.md

@@ -475,7 +475,13 @@ We recommend using LIBRA Mini as the primary evaluation suite for model comparis
 
 ## Dataset Structure
 
-The datasets are divided into subsets based on context lengths. The table below shows the number of examples per context length for each dataset. Datasets included in **LIBRA Mini** are highlighted in bold. Note that not all datasets cover the full range of context lengths — some are designed for specific length ranges that best suit their task type.
+The datasets are divided into subsets based on context lengths. 
+
+<p align="center">
+  <img src="libra_structure.svg" width="600" />
+</p>
+
+The table below shows the number of examples per context length for each dataset. Datasets included in **LIBRA Mini** are highlighted in bold. Note that not all datasets cover the full range of context lengths — some are designed for specific length ranges that best suit their task type.
 
 | **Task** | **4k** | **8k** | **16k** | **32k** | **64k** | **128k** | **256k** | **512k** | | **Total** |
 |:---|---:|---:|---:|---:|---:|---:|---:|---:|---|---:|