CORAL: Benchmarking Multi-turn Conversational Retrieval-Augmentation Generation

TL;DR

CORAL is a large-scale benchmark designed to evaluate multi-turn conversational retrieval-augmented generation systems, addressing a significant gap in existing research focused mainly on single-turn scenarios.

Contribution

The paper introduces CORAL, a comprehensive benchmark for multi-turn conversational RAG, and provides a unified framework for evaluation and comparison of different methods.

Findings

Existing methods show room for improvement on CORAL

CORAL reveals challenges in open-domain coverage and topic shifts

Evaluation highlights the need for more robust conversational RAG models

Abstract

Retrieval-Augmented Generation (RAG) has become a powerful paradigm for enhancing large language models (LLMs) through external knowledge retrieval. Despite its widespread attention, existing academic research predominantly focuses on single-turn RAG, leaving a significant gap in addressing the complexities of multi-turn conversations found in real-world applications. To bridge this gap, we introduce CORAL, a large-scale benchmark designed to assess RAG systems in realistic multi-turn conversational settings. CORAL includes diverse information-seeking conversations automatically derived from Wikipedia and tackles key challenges such as open-domain coverage, knowledge intensity, free-form responses, and topic shifts. It supports three core tasks of conversational RAG: passage retrieval, response generation, and citation labeling. We propose a unified framework to standardize various conversational RAG methods and conduct a comprehensive evaluation of these methods on CORAL, demonstrating substantial opportunities for improving existing approaches.