Chain-of-Retrieval Augmented Generation

TL;DR

This paper presents CoRAG, a novel retrieval-augmented generation method that iteratively retrieves and reasons over information, significantly improving performance on complex knowledge-intensive tasks.

Contribution

It introduces a dynamic, multi-step retrieval process with rejection sampling for training, advancing beyond single-step retrieval in RAG models.

Findings

Over 10 points improvement in EM score on multi-hop QA tasks

State-of-the-art performance on KILT benchmark

Effective scaling behavior analysis of CoRAG

Abstract

This paper introduces an approach for training o1-like RAG models that retrieve and reason over relevant information step by step before generating the final answer. Conventional RAG methods usually perform a single retrieval step before the generation process, which limits their effectiveness in addressing complex queries due to imperfect retrieval results. In contrast, our proposed method, CoRAG (Chain-of-Retrieval Augmented Generation), allows the model to dynamically reformulate the query based on the evolving state. To train CoRAG effectively, we utilize rejection sampling to automatically generate intermediate retrieval chains, thereby augmenting existing RAG datasets that only provide the correct final answer. At test time, we propose various decoding strategies to scale the model's test-time compute by controlling the length and number of sampled retrieval chains. Experimental results across multiple benchmarks validate the efficacy of CoRAG, particularly in multi-hop question answering tasks, where we observe more than 10 points improvement in EM score compared to strong baselines. On the KILT benchmark, CoRAG establishes a new state-of-the-art performance across a diverse range of knowledge-intensive tasks. Furthermore, we offer comprehensive analyses to understand the scaling behavior of CoRAG, laying the groundwork for future research aimed at developing factual and grounded foundation models.