The base Fileread model was trained on a large corpus of knowledge and information from the internet and public records. This training allows Fileread to refine its understanding of language as well as form a foundational language base. Although Fileread prioritizes its focus on the document universe at hand, it may rarely bring up “facts” that might not be based on knowledge.
If there's a conflict between the AI-generated answer and the answers derived from the actual document, always choose the green highlighted answers derived from the document.
Spreadsheets are difficult for LLMs to understand at the moment because its extracted text is hard for language models to examine in its raw form. LLMs are trained to consume human language which is mostly English.
Ultimately, this is cutting-edge technology and it's not perfect, but we are doing our best to ensure you with the best results and experience possible. We work alongside you to ensure the AI improves and gets accustomed to your workflow with each usage. Please feel free to contact us 24/7