clarify

README.md

@@ -15,9 +15,7 @@ size_categories:
 
 # stacked samsum 1024
 
-Created with the `stacked-booksum` repo version v0.25. It contains
-
-To summarize the requested information in a cohesive numeric list:
+Created with the `stacked-booksum` repo version v0.25. It contains:
 
 1. Original Dataset: copy of the base dataset
 
@@ -25,7 +23,7 @@ To summarize the requested information in a cohesive numeric list:
    - Maximum Input Length: The maximum length for input sequences is 1024 tokens in the longt5 model tokenizer.
    - Maximum Output Length: The maximum length for output sequences is also 1024 tokens in the longt5 model tokenizer.
 
-3. Special Token: The dataset utilizes the `[NEXT_CONCEPT]` token to indicate a new topic **within** the same summary. It is essential to explicitly add this special token to your model's tokenizer before training, ensuring that it is recognized and processed correctly during downstream usage.
+3. Special Token: The dataset utilizes the `[NEXT_CONCEPT]` token to indicate a new topic **within** the same summary. It is recommended to explicitly add this special token to your model's tokenizer before training, ensuring that it is recognized and processed correctly during downstream usage.
 
 ## stats