This dataset is a transformed version of mlabonne/FineTome-100k, created by applying the following steps:
Extracted the first turn of each conversation: optional system message + user message + assistant message.
Performed deduplication: using MinHash to remove duplicate entries.
Converted the format from ShareGPT to OpenAI
