Beyond Independent Passages: Adaptive Passage Combination Retrieval for Retrieval Augmented Open-Domain Question Answering

2025-07-05

AI Summary

Traditional RAG methods retrieve paragraphs independently, leading to redundancy, noise, and insufficient diversity—especially in multi-hop question answering and noisy corpora. To address this, we propose AdaPCR, the first framework to treat paragraph combinations—not individual paragraphs—as the fundamental unit for retrieval and reranking, explicitly modeling inter-paragraph dependencies. Its key contributions are: (1) an adaptive paragraph-combination retrieval mechanism that dynamically determines the optimal combination size without requiring an auxiliary stopping module; (2) a predictive reranking objective aligned with answer generation, jointly optimizing cross-paragraph reasoning capability; and (3) a hybrid reranker integrating context-aware query reformulation with black-box large language model–driven combination-level scoring. Evaluated on multiple open-domain QA benchmarks, AdaPCR significantly outperforms state-of-the-art baselines, with particularly pronounced gains on multi-hop tasks—demonstrating that modeling paragraph combinations fundamentally enhances retrieval quality.

Technology Category

• Natural Language Processing: Question Answering
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Search and Optimization: Metareasoning and Metaheuristics

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Retrieval-augmented generation (RAG) enhances large language models (LLMs) by incorporating external documents at inference time, enabling up-to-date knowledge access without costly retraining. However, conventional RAG methods retrieve passages independently, often leading to redundant, noisy, or insufficiently diverse context-particularly problematic - particularly problematic in noisy corpora and for multi-hop questions. To address this, we propose Adaptive Passage Combination Retrieval (AdaPCR), a novel framework for open-domain question answering with black-box LMs. AdaPCR explicitly models dependencies between passages by considering passage combinations as units for retrieval and reranking. It consists of a context-aware query reformulation using concatenated passages, and a reranking step trained with a predictive objective aligned with downstream answer likelihood. Crucially, AdaPCR adaptively selects the number of retrieved passages without additional stopping modules. Experiments across several QA benchmarks show that AdaPCR outperforms baselines, particularly in multi-hop reasoning, demonstrating the effectiveness of modeling inter-passage dependencies for improved retrieval.

Problem

• Retrieving redundant or noisy passages in RAG systems
• Handling multi-hop questions with insufficient context diversity
• Adaptively selecting optimal passage combinations without stopping modules

Innovation

• Adaptive retrieval of passage combinations
• Context-aware query reformulation with passages
• Reranking aligned with answer likelihood