Multi-Agent Retrieval Augmented Generation BOT for Attributed Question Answering

Artificial Intelligence

Multi-agent systems it provide a very effective way to build software that can handle difficult problems. Instead of relying on just one program to do everything, this architecture uses several independent agents that work together as a team to reach a single goal. Our project is a smart AI system called a “Multi-Agent RAG”. Think of it like a team of specialized AI workers instead of just one they work together to double-check facts, the answers they give are much more accurate, stay on topic, and are more dependable than what you would get from a standard AI. Our main goal is to make sure the AI gives answers you can actually trust. We focused on making the answers ”correct” by ensuring they fully cover what you asked and stay on topic. This system works like a team of editors. First, they go through all the information they found and remove the junk. The core of this system is a “hybrid” search strategy that uses two different methods at once. One method looks for specific keywords, while the other looks for the general meaning behind the words. This leads to answers that are more accurate and backed up 0bette evidence. Our work surely will outperform standard RAG baselines, achieving gains in correctness and faithfulness. Our project attempts to mitigate privacy issues by giving users control over whether they want their information processed locally (immediately) or remotely (via cloud). Furthermore, we are developing a unique solution that combines automated tasks with an intuitive conversational interface, capable of being used by anyone for daily activities

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.