Update README.md

README.md

@@ -5,4 +5,6 @@ size_categories:
 - 1M<n<10M
 ---
 
-TH 1.7M Arithmetic Tasks dataset inspired from [Goat: Fine-tuned LLaMA Outperforms GPT-4 on Arithmetic Tasks](https://arxiv.org/abs/2305.14201) & [Author's HF](https://huggingface.co/datasets/tiedong/goat)
+TH 1.7M Arithmetic Tasks dataset inspired from [Goat: Fine-tuned LLaMA Outperforms GPT-4 on Arithmetic Tasks](https://arxiv.org/abs/2305.14201) & [Author's HF](https://huggingface.co/datasets/tiedong/goat)
+
+**FYI:** Use columns "instruction" and "output" if you plan to instruct fine-tuning.