Skip to content
TheChatArchiveThe Chat Archive
koala-13b· UC BerkeleyHugging Face Arena Human Preference 55K

polish the paper bellow. Almost all models reach the same score after being sufficiently…

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with gpt-3.5-turbo-0314’s answer.
  1. User

    polish the paper bellow.

    Almost all models reach the same score after being sufficiently trained in the same setting, as shown in \ref{fig:Fine-tune}. The scores indicate the intrinsic difficulty of the tasks. As the difficulty of the task changes, the gap between fine-tuning and zero-shot also shifts. Given a limited number of samples in the training dataset, \ref{fig:few-shot} shows the performance of the models in few-shot scenarios. The model "GUR-FULL" is the best one across all tasks.

  2. Assistant

    Almost all models reach the same score after being sufficiently trained in the same setting, as demonstrated in \ref{fig:Fine-tune}. The scores serve as a measure of the intrinsic difficulty of the tasks. As the difficulty of the task changes, the gap between fine-tuning and zero-shot also shifts. Given a limited number of samples in the training dataset, \ref{fig:few-shot} displays the performance of the models in few-shot scenarios. In this study, the model "GUR-FULL" consistently outperforms other models across all tasks.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.