ReSCORE: Label-free Iterative Retriever Training for Multi-hop Question Answering with Relevance-Consistency Supervision
This addresses the challenge of high query variability in multi-hop question answering for researchers and practitioners, though it is an incremental advancement over existing dense retrieval methods.
The paper tackles the problem of training dense retrievers for multi-hop question answering without labeled query-document pairs by introducing ReSCORE, which uses large language models to supervise retrieval based on relevance and consistency, resulting in significant improvements in retrieval and state-of-the-art MHQA performance on three benchmarks.
Multi-hop question answering (MHQA) involves reasoning across multiple documents to answer complex questions. Dense retrievers typically outperform sparse methods like BM25 by leveraging semantic embeddings; however, they require labeled query-document pairs for fine-tuning. This poses a significant challenge in MHQA due to the high variability of queries (reformulated) questions throughout the reasoning steps. To overcome this limitation, we introduce Retriever Supervision with Consistency and Relevance (ReSCORE), a novel method for training dense retrievers for MHQA without labeled documents. ReSCORE leverages large language models to capture each documents relevance to the question and consistency with the correct answer and use them to train a retriever within an iterative question-answering framework. Experiments on three MHQA benchmarks demonstrate the effectiveness of ReSCORE, with significant improvements in retrieval, and in turn, the state-of-the-art MHQA performance. Our implementation is available at: https://leeds1219.github.io/ReSCORE.