Clarify datasets are trained separately

README.md

@@ -17,7 +17,7 @@ The files are provided to make reproduction easier, since small differences in p
 
 ## Contents
 
-The repository contains processed files for three datasets:
+The repository contains processed files for three separate datasets. These datasets are not mixed together; DIGER trains and evaluates them independently.
 
 - `beauty/`
 - `instruments/`
@@ -36,11 +36,11 @@ The LLaMA embeddings follow the generation procedure described in:
 
 https://github.com/honghuibao2000/letter
 
-We include the processed embeddings here so that downstream users can reproduce the released DIGER artifacts without depending on small preprocessing or embedding-generation differences.
+We include the processed embeddings here so that downstream users can reproduce the released DIGER artifacts without depending on small preprocessing or embedding-generation differences. Each dataset uses its own embedding matrix and is trained independently.
 
 ## Models
 
-The corresponding released RQ-VAE checkpoints are available at:
+The corresponding released RQ-VAE checkpoints are trained separately for each dataset and are available at:
 
 - Beauty: https://huggingface.co/junchenfu/diger-rqvae-beauty
 - Instruments: https://huggingface.co/junchenfu/diger-rqvae-instruments