Update README.md

README.md

@@ -252,7 +252,7 @@ By selecting either `cls_emb` or `channel_emb`, you leverage the pre-trained mod
 ---
 
 ### 10. **Generate Labels if Necessary**
-If your dataset requires labels, you can easily generate them using DeepMIMO data. Here's an example to create labels for either LoS/NLoS classification or beam prediction, depending on the scenario selected:You can generate labels for your dataset using the DeepMIMO data. Below is an example for generating labels for the LoS/NLoS classification and beam prediction tasks based on the dataset scenario you chose earlier:
+If your dataset requires labels, you can easily generate them using DeepMIMO data. Here's an example to create labels for either LoS/NLoS classification or beam prediction, depending on the scenario selected:
 ```python
 from input_preprocess import create_labels
 tasks = ['LoS/NLoS Prediction', 'Beam Prediction']
@@ -262,7 +262,13 @@ labels = create_labels(task, selected_scenario_names, n_beams=64)
 
 ---
 
-### 11. **Explore the Interactive Demo**
+### 11. **Leverage the Dataset for Downstream Tasks**
+
+LWM, pre-trained on a vast and diverse dataset using self-supervised learning, does not rely on labeled data. During inference, it transforms raw channels into rich embeddings in real time, capturing both general and intricate patterns within the wireless channels. These embeddings can be directly applied to various downstream tasks, offering a more powerful alternative to using the original channel data.
+
+---
+
+### 12. **Explore the Interactive Demo**
 
 To experience **LWM** interactively, visit our demo hosted on Hugging Face Spaces: