PAR$^2$-RAG: Planned Active Retrieval and Reasoning for Multi-Hop Question Answering

TL;DR

PAR$^2$-RAG introduces a two-stage retrieval and reasoning framework that improves multi-hop question answering by balancing coverage and commitment, leading to significant accuracy gains.

Contribution

It presents a novel planned active retrieval and reasoning method that outperforms existing approaches on multiple benchmarks by separating evidence coverage from reasoning commitment.

Findings

Achieves up to 23.5% higher accuracy on MHQA benchmarks.

Provides up to 10.5% retrieval gains in NDCG.

Outperforms state-of-the-art baselines across four benchmarks.

Abstract

Large language models (LLMs) remain brittle on multi-hop question answering (MHQA), where answering requires combining evidence across documents through retrieval and reasoning. Iterative retrieval systems can fail by locking onto an early low-recall trajectory and amplifying downstream errors, while planning-only approaches may produce static query sets that cannot adapt when intermediate evidence changes. We propose \textbf{Planned Active Retrieval and Reasoning RAG (PAR$^2$-RAG)}, a two-stage framework that separates \emph{coverage} from \emph{commitment}. PAR$^2$-RAG first performs breadth-first anchoring to build a high-recall evidence frontier, then applies depth-first refinement with evidence sufficiency control in an iterative loop. Across four MHQA benchmarks, PAR$^2$-RAG consistently outperforms existing state-of-the-art baselines, compared with IRCoT, PAR$^2$-RAG achieves up to \textbf{23.5\%} higher accuracy, with retrieval gains of up to \textbf{10.5\%} in NDCG.