Visconde: Multi-document QA with GPT-3 and Neural Reranking

TL;DR

Visconde is a multi-document question-answering system that decomposes questions, retrieves relevant passages, and aggregates information using GPT-3, achieving near-human performance when relevant evidence is provided.

Contribution

The paper introduces a novel three-step pipeline combining question decomposition, neural retrieval, and LLM-based aggregation for multi-document QA.

Findings

Retrievers are the main bottleneck in system performance.

Readers perform at human level with relevant passages.

Explanation prompting improves answer quality.

Abstract

This paper proposes a question-answering system that can answer questions whose supporting evidence is spread over multiple (potentially long) documents. The system, called Visconde, uses a three-step pipeline to perform the task: decompose, retrieve, and aggregate. The first step decomposes the question into simpler questions using a few-shot large language model (LLM). Then, a state-of-the-art search engine is used to retrieve candidate passages from a large collection for each decomposed question. In the final step, we use the LLM in a few-shot setting to aggregate the contents of the passages into the final answer. The system is evaluated on three datasets: IIRC, Qasper, and StrategyQA. Results suggest that current retrievers are the main bottleneck and that readers are already performing at the human level as long as relevant passages are provided. The system is also shown to be more effective when the model is induced to give explanations before answering a question. Code is available at \url{https://github.com/neuralmind-ai/visconde}.