Update README.md

README.md

@@ -10,7 +10,7 @@ pretty_name: GPTDynamics
 
 
 ## Dataset Summary
-**GPTDynamics** is a dataset designed for training and evaluating GPT simulators using structured training curriculums. It supports both fine-tuning and command-based fine-tuning scenarios and provides comprehensive test metrics (such as loss, BLEU, and ROUGE scores) for each test sample at various training steps. This dataset was introduced in the [EMNLP'24 paper](https://arxiv.org/pdf/2404.07840), and detailed instructions on how to use it can be found on the [GitHub page](https://github.com/ernie-research/gptfluence).
+**GPTDynamics** is a dataset designed for training and evaluating GPT simulators using structured training curriculums. It supports both fine-tuning and instruction-tuning scenarios and provides comprehensive test metrics (such as loss, BLEU, and ROUGE scores) for each test sample at various training steps. This dataset was introduced in the [EMNLP'24 paper](https://arxiv.org/pdf/2404.07840), and detailed instructions on how to use it can be found on the [GitHub page](https://github.com/ernie-research/gptfluence).
 
 ## Dataset Structure
 - **id**: ID of test examples