Agentic AutoRAG: RAG Pipeline Optimization through Reasoning-Driven Agents
What it is
This paper introduces Agentic AutoRAG, an LLM-agent framework that optimizes Retrieval-Augmented Generation (RAG) pipelines by treating configuration as a multi-objective problem. It uses a Diagnoser agent to attribute query failures to either retrieval or generation stages and a Proposer agent to select new configurations based on a knowledge base of model rankings and costs, aiming to trace a Pareto frontier between accuracy and expense.
The problem it solves
Configuring RAG pipelines is an expensive hyperparameter optimization problem where existing optimizers reduce trials to aggregate scores without modeling why specific configurations failed, leading to inefficient search and poor cost-accuracy trade-offs.
What you could use it for
- Automating the tuning of RAG pipelines for multi-hop question answering
- Optimizing RAG systems for cost-efficiency in production environments
- Diagnosing whether RAG performance issues stem from retrieval or generation components
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.