FineWebEdu-Conversation is a dataset containing excerpts from HuggingFaceFW/fineweb-edu, enriched with generated conversations between an ordinary person and an expert. The format simulates a natural interview-style dialogue.
Source: FineWebEdu excerpts.
Methodology:
Chunking: Segmented into chunks of approximately 1950 Llama 3 tokens.
Conversation Generation: Used agentlans/Llama3.1-LexiHermes-SuperStorm
and a model distilled… See the full description on the dataset page:
https://huggingface.co/datasets/agentlans/finewebedu-conversation.
