Update README.md

README.md

@@ -12,4 +12,8 @@ task_categories:
 
 ## Dataset Description:
 
-TaxaBench-8k is a multimodal dataset containing six modalities - image, text, satellite image, audio, geographic location, and environmental features for evaluating large ecological models.
+TaxaBench-8k is a multimodal dataset containing six modalities - image, text, satellite image, audio, geographic location, and environmental features for evaluating large ecological models.
+
+## Usage:
+
+Please use the `test_df.csv` for reading data which contains 8,813 samples. Optionally we have created `train_df.csv` and `val_df.csv` for potentially extending the dataset in the future.