CLAP: Coreference-Linked Augmentation for Passage Retrieval

TL;DR

CLAP is a novel LLM-based passage augmentation method that improves dense retrieval by maintaining coreference coherence and aligning pseudo-queries with retriever representations, leading to significant performance gains across domains.

Contribution

Introduces CLAP, a lightweight, domain-agnostic passage expansion framework that resolves coreference and generates localized pseudo-queries for enhanced retrieval.

Findings

Up to 20.68% absolute nDCG@10 improvement in retrieval performance.

Consistent gains across different retriever strengths and domains.

Outperforms traditional LLM-based expansion methods in out-of-domain settings.

Abstract

Large Language Model (LLM)-based passage expansion has shown promise for enhancing first-stage retrieval, but often underperforms with dense retrievers due to semantic drift and misalignment with their pretrained semantic space. Beyond this, only a portion of a passage is typically relevant to a query, while the rest introduces noise--an issue compounded by chunking techniques that break coreference continuity. We propose Coreference-Linked Augmentation for Passage Retrieval (CLAP), a lightweight LLM-based expansion framework that segments passages into coherent chunks, resolves coreference chains, and generates localized pseudo-queries aligned with dense retriever representations. A simple fusion of global topical signals and fine-grained subtopic signals achieves robust performance across domains. CLAP yields consistent gains even as retriever strength increases, enabling dense retrievers to match or surpass second-stage rankers such as BM25 + MonoT5-3B, with up to 20.68% absolute nDCG@10 improvement. These improvements are especially notable in out-of-domain settings, where conventional LLM-based expansion methods relying on domain knowledge often falter. CLAP instead adopts a logic-centric pipeline that enables robust, domain-agnostic generalization.