Use Cases

How different parts are used to build and to auto-optimize various LLM applications.

We will build use cases end-to-end, ranging from classification (classical NLP tasks) to question answering, retrieval-augmented generation (RAG), and multi-generator pipelines.

Optimization

Part	Description
Question Answering	Question Answering with bhh_hard_object_count dataset, including textual-gradient descent and few-shot boostrap optimization.
Classification Optimization	Classification with gpt-3.5-turbo. The optimized task pipeline performs on-par with gpt-4o.
RAG optimization	Different from previous tasks where we only used one generator component, in this of hotpotqa dataset, we will demonstrates how to optimize a pipeline composed of multiple GradComponent`(`Retriever & Generator) with a standard RAG architectures