跳转至

文章背景与核心概要

检索增强生成（RAG）在处理多跳问答（QA）时常常面临挑战，因为支撑性段落往往通过中间实体和关系相互关联。传统方法依赖于单一的检索意图或单次查询扩展，容易引入噪声或冗余信号。

本文介绍了一种无需训练的多跳检索框架，其核心特点包括：条件证据探索、段落特定的对比精炼以及覆盖度感知的高级排序。通过在离线索引阶段构建段落特定的对比切面，并在推理阶段针对未解决的信息需求迭代生成探针，该框架显著提升了MuSiQue、HotpotQA和2WikiMultihopQA等基准测试中的检索质量与下游问答性能。

Beyond One-Shot Expansion: Contrastive Evidence Exploration for Multi-Hop Retrieval

Beyond One-Shot Expansion: Contrastive Evidence Exploration for Multi-Hop Retrieval

Summary

Summary

Retrieval-augmented generation (RAG) often struggles with multi-hop question answering (QA), where supporting passages are linked through intermediate entities and relations. Traditional approaches rely on a single retrieval intent or one-shot query expansion, leading to noisy or redundant signals.

This paper introduces a training-free multi-hop retrieval framework featuring: - Evidence-conditioned exploration - Passage-specific contrastive refinement - Coverage-aware final ranking

By constructing passage-specific contrastive facets during offline indexing and iteratively generating probes for unresolved information needs at inference time, the framework significantly improves retrieval quality and downstream QA performance across benchmarks like MuSiQue, HotpotQA, and 2WikiMultihopQA.

Paper Metadata

Paper Metadata

  • arXiv ID: arXiv:2609.07050 [cs.AI]
  • Authors: JungMin Yun, YoungBin Kim
  • Submitted Date: September 7, 2026
  • Primary Subject: Artificial Intelligence (cs.AI)
  • Secondary Subject: Information Retrieval (cs.IR)
  • Comments: Accepted to CIKM 2026
  • Related DOI: 10.1145/3799682.3840072

Abstract

Abstract

Retrieval-augmented generation (RAG) critically depends on retrieving the evidence necessary for effective reasoning. However, this remains particularly challenging in multi-hop question answering (QA), where supporting passages are often linked through intermediate entities and relations that must be progressively uncovered.

Existing retrieval approaches typically rely on a single retrieval intent or one-shot query expansion, limiting their ability to adapt to newly retrieved evidence and potentially introducing noisy or redundant retrieval signals. To address these limitations, we propose a training-free multi-hop retrieval framework that integrates evidence-conditioned exploration, passage-specific contrastive refinement, and coverage-aware final ranking.

During offline indexing, the framework constructs passage-specific contrastive facets that characterize each passage relative to its semantically similar neighbors, providing fine-grained signals to distinguish closely related candidates. At inference time, the framework iteratively retrieves evidence, generates probes targeting unresolved information needs, refines candidate relevance using the contrastive facets, and selects a complementary set of passages that collectively cover diverse evidence-seeking intents. Experiments on MuSiQue, HotpotQA, and 2WikiMultihopQA demonstrate consistent improvements in retrieval quality and downstream QA performance over baselines.

Full-Text and Resources

Full-Text and Resources

External Citations & Tools

External Citations & Tools