# RAG based Question-Answering for Contextual Response Prediction System

**Authors:** Sriram Veturi, Saurabh Vaichal, Reshma Lal Jagadheesh, Nafis Irtiza Tripto, Nian Yan  
**arXiv:** [2409.03708](https://arxiv.org/abs/2409.03708)  
**Date:** 2024-09-05  
**Summary page:** https://www.summarizepaper.com/en/paper/2409.03708v2/

## Key points

- Researchers presented their study on enhancing question-answering systems using Retrieval Augmented Generation (RAG) techniques
- The study focused on improving factual accuracy and reducing hallucinations in Large Language Models (LLMs) by incorporating Reason+Act (ReAct) prompting
- Leveraging external knowledge databases through RAG architecture aimed to address issues like outdated information and lack of transparency in LLMs
- Significant improvements in response accuracy and relevance were showcased compared to existing systems, demonstrating the potential of RAG-based LLMs for automating customer care processes
- Recent advancements have led to more sophisticated approaches like Advanced and Modular RAG frameworks, aiming to enhance LLMs' ability to generate contextually appropriate responses

## Summary

, , , , 

In the 1st Workshop on GenAI and RAG Systems for Enterprise, held in October 24, 2024, in Boise, Idaho, USA, Sriram Veturi, Saurabh Vaichal, Reshma Lal Jagadheesh, Nafis Irtiza Tripto, and Nian Yan presented their research on enhancing question-answering systems using Retrieval Augmented Generation (RAG) techniques. Their study focused on addressing the challenge of providing precise and relevant information to customer queries in real-time settings without hallucinations. The researchers explored the effectiveness of incorporating Reason+Act (ReAct) prompting to improve factual accuracy and reduce hallucinations in Large Language Models (LLMs). By leveraging external knowledge databases through RAG architecture, they aimed to overcome issues such as outdated information and lack of transparency in LLMs. The study showcased significant improvements in response accuracy and relevance compared to existing systems, highlighting the potential of RAG-based LLMs for automating customer care processes. The RAG architecture discussed in the workshop builds upon traditional methods by integrating advanced indexing, retrieval, and generation processes. While ChatGPT popularized the initial concept of RAG, recent advancements have led to more sophisticated approaches like Advanced and Modular RAG frameworks. These developments aim to enhance the ability of LLMs to generate contextually appropriate responses by leveraging external knowledge sources. Overall, the workshop findings suggest that incorporating RAG capabilities into LLMs can significantly enhance their performance in real-world applications such as customer service automation. By utilizing a combination of advanced techniques and external knowledge bases, researchers are paving the way for more accurate and efficient question-answering systems that can support human agents in handling customer queries effectively.