Update README.md

README.md

@@ -25,4 +25,8 @@ This llama model was trained 2x faster with [Unsloth](https://github.com/unsloth
 
 # Description
 
-This model is fine-tuned on 10k sample points from Cendol V2 dataset.
+This model is fine-tuned on 10k sample points from Cendol V2 dataset.
+
+# Impact of Fine-tune
+
+This model is more straightforward and more natural in Indonesian (idk, it's vibe-based fine-tuning, need benchmark)