Update README.md

README.md

@@ -7,7 +7,7 @@ Mistral-7B finetuned on a dataset of BTS fanfic.
 
 This model uses the `alpaca` format:
 ```
-{"instruction": "...", "input": "...", "output": "..."}
+{"instruction": "An interaction between a user providing instructions, and an imaginative assistant providing responses.", "input": "...", "output": "..."}
 ```
 
 Note RoPE scaling parameter 4.0, with RoPE scaling type `linear`