cs.IR提交于 2026-10-31已译

DuQUAD：面向多文档问答的双视图框架，带质量感知证据剪枝

DuQUAD: A Dual-View Framework with Quality-Aware Evidence Pruning for Multi-Document Question Answering

Yeonjun Choi · Seyeon Jeong · Jong Wook Kim · Beakcheol Jang

中文摘要

具备检索增强生成能力的大语言模型在多文档上下文中会系统性地低估中间位置证据，这一现象被称为 Lost in the Middle。文档重排序与注意力引导等现有缓解策略能降低边界偏置，但在噪声场景下依然脆弱：不准确的相关性估计与虚假文档仍会掩盖中心位置的有用证据。DuQUAD 是一个多文档问答框架，通过双视图推理与质量感知证据剪枝缓解位置偏置。它结合由 Structural Fusion Score 引导的 Local Agent 与对全文档集推理的 Global Agent，兼顾定向证据恢复与全局上下文理解。框架进一步用显式质量评分过滤候选证据，并在句子级精炼以降低噪声。DuQUAD 稳定优于强 Lost-in-the-Middle 缓解基线，包括混合检索与重排序、多智能体推理和上下文优化方法，相较最强竞争基线平均提升准确率 3.0%，金标证据召回率提升 9.4%，同时相比现有多智能体框架大幅降低计算成本。

关键要点

01问题：多文档问答中 Lost in the Middle 导致中间证据被系统性低估，噪声场景下现有缓解策略脆弱。
02方法：Local Agent 依据 Structural Fusion Score 定向恢复证据，Global Agent 对全文档集推理，形成双视图。
03方法：用显式质量评分过滤候选证据，并在句子级精炼，降低虚假文档与噪声干扰。
04结果：平均准确率较最强基线提升 3.0%，金标证据召回率提升 9.4%。
05结果：相比现有多智能体框架显著降低计算成本。

解读

尚无解读。

原始英文摘要

arXiv:2610.37701v1 Announce Type: cross Abstract: Large language models with retrieval-augmented generation systematically underutilize evidence in the middle of multi-document contexts, a phenomenon known as Lost in the Middle. Existing mitigation strategies, such as document reordering and attention steering, can reduce boundary bias but remain brittle in noisy settings, where inaccurate relevance estimation and spurious documents still obscure useful evidence in central positions. We propose DuQUAD, a multi-document QA framework that mitigates positional bias through dual-view reasoning and quality-aware evidence pruning. DuQUAD combines a Local Agent guided by a Structural Fusion Score with a Global Agent that reasons over the full document set, providing both targeted evidence recovery and broad contextual understanding. The framework further filters candidate evidence using explicit quality scoring and refines it at the sentence level to reduce noise. DuQUAD consistently outperforms strong Lost-in-the-Middle mitigation baselines, including hybrid retrieval and reranking, multi-agent reasoning, and context optimization methods, improving accuracy by 3.0% on average over the strongest competing baseline and gold evidence recall by 9.4%, while substantially reducing computational cost compared to existing multi-agent frameworks.