Update README.md

README.md

@@ -55,6 +55,8 @@ It achieves the following results on the evaluation set:
 
 This is just a base model/example, and there is likely even better performance with larger models. 
 
+Additionally, this was trained on a dataset of **only** instructions+outputs with the `inputs` filtered out/dropped. This means that text of  *1) cookies and cream 2) chocolate chip 3) mint chip 4) oreo* will **not** get you *"Rank the following icecream flavors: oreo, mint chip, chocolate chip, cookies and cream"*
+
 ## Training and evaluation data
 
 See the linked dataset `pszemraj/fleece2instructions` - it is a filtered/formatted version of `tatsu-lab/alpaca` to generate instructions for arbitrary text.