$\pi$-CoT: Prolog-Initialized Chain-of-Thought Prompting for Multi-Hop Question-Answering

TL;DR

This paper introduces $$-CoT, a prompting method that uses Prolog-based decomposition of multi-hop questions to improve reasoning accuracy in large language models, especially in retrieval-augmented settings.

Contribution

It proposes a novel Prolog-Initialized Chain-of-Thought prompting strategy that enhances multi-hop reasoning by decomposing questions into logical sub-queries for better model performance.

Findings

$$-CoT outperforms standard RAG and in-context CoT on multiple benchmarks.

The method reduces reasoning errors caused by circular logic.

Sequential sub-query resolution improves multi-hop question-answering accuracy.

Abstract

Chain-of-Thought (CoT) prompting significantly enhances large language models' (LLMs) problem-solving capabilities, but still struggles with complex multi-hop questions, often falling into circular reasoning patterns or deviating from the logical path entirely. This limitation is particularly acute in retrieval-augmented generation (RAG) settings, where obtaining the right context is critical. We introduce Prolog-Initialized Chain-of-Thought ($\pi$-CoT), a novel prompting strategy that combines logic programming's structural rigor with language models' flexibility. $\pi$-CoT reformulates multi-hop questions into Prolog queries decomposed as single-hop sub-queries. These are resolved sequentially, producing intermediate artifacts, with which we initialize the subsequent CoT reasoning procedure. Extensive experiments demonstrate that $\pi$-CoT significantly outperforms standard RAG and in-context CoT on multi-hop question-answering benchmarks.