Fastchat's AI-enhanced evaluation pipeline is based on GPT-4, which is used to generate answers and reviews from different models.
Data for Fastchat's Vicuna model is gathered from around 70K user-shared conversations on ShareGPT.com, ensuring quality and filtering out inappropriate content.
Key terms:
Fastchat: A platform with an AI-enhanced evaluation pipeline based on GPT-4
Vicuna: The fine-tuned model created using Fastchat's evaluation pipeline
ShareGPT.com: The source of approximately 70K user-shared conversations used for fine-tuning the Vicuna model
LLaMA base model: The base model used for fine-tuning Vicuna in Fastchat's evaluation pipeline