At the extreme, this can be as much as 20 million data points, with millions of rows of volatile formulas such as indirect functions, index match, etc running off of this data, interconnected across dozens of tabs. Each one of those probably deserves its own blog post and I only briefly touch upon them here.I work at an investing firm and we often receive significant amounts of data on companies that we need to crunch in excel. Each one has different requirements in terms of these pillars and therefore require a different approach.īeyond these 6 pillars there are also other aspects that are often not as straightforward when considering RAG vs finetuning such as latency, maintenance, robustness, integration with existing systems, UX, cost, and complexity. I use these 6 pillars to assess three very popular LLM use cases: (1) Summarisation, (2) Q&A System, (3) Customer support automation. □ Is transparency/interpretability required? □ Do we need to minimise hallucinations? There are at least 6 pillars customers should consider which approach is right for them: In this blog post I aim to provide and extensive lens through which customers can assess the right approach(es) for their use case(s). Spoiler alert: It’s the wrong question! RAG and finetuning are very different approaches and serve very different needs - they are □□□□□□□□□□! □□□ □□ □□□□□□□□□□ - that is the question many of our customers ask themselves at the moment to get the most out of their LLM applications.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |