liveBook · Manning

Generalist model

As explained in Rearchitecting LLMs

A generalist model is a model designed for broad capabilities across many tasks and domains rather than a narrow specialization.

A generalist model supports broad capabilities, but relying on the same model and similar data as competitors can produce similar outputs. In investment software, that may mean identical stock recommendations and little value for clients. section 1.1.2

Distinguish the available adaptation paths:

  • Prompt engineering can shape responses, but does not change the model’s underlying expertise. section 1.1.2
  • Retrieval-augmented generation (RAG) supplies external information without transforming the model; it can complement rearchitecting that improves how the model uses that information. section 1.1.2
  • Fine-tuning adapts a model to a specific dataset and task; fine-tuning closed models through provider tools can be extremely expensive. section 1.1.2

How Rearchitecting LLMs covers it

How deep it goes
Contrasts prompting and RAG with model adaptation, noting that prompts shape responses and RAG adds information without changing expertise.
Examples
Uses investment software issuing similar buy recommendations for Nasdaq stocks to illustrate generic outputs.
What you can do after
Evaluate whether shared-model outputs differentiate a product, then consider proprietary data and model specialization.
How it connects
Links generalist models to prompt engineering, retrieval-augmented generation, fine-tuning, and specialized models.

Also explained in

Compare how 2 books explain generalist model →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Rearchitecting LLMs. The section links point to the parts of the book it draws on.