Beyond Function-Level Search: Repository-Aware Dual-Encoder Code Retrieval with Adversarial Verification

TL;DR

This paper introduces a new repository-level code retrieval benchmark and a dual-encoder model, ReflectCode, that significantly improves retrieval accuracy by understanding cross-component change intents in complex codebases.

Contribution

The paper presents RepoAlign-Bench, a novel benchmark for repository-level code retrieval, and ReflectCode, an adversarial reflection-based dual-encoder architecture for enhanced context-aware code retrieval.

Findings

ReflectCode improves Top-5 Accuracy by 12.2%.

ReflectCode enhances Recall by 7.1%.

RepoAlign-Bench provides 52k annotated instances for repository-level retrieval.

Abstract

The escalating complexity of modern codebases has intensified the need for retrieval systems capable of interpreting cross-component change intents, a capability fundamentally absent in conventional function-level search paradigms. While recent studies have improved the alignment between natural language queries and code snippets, retrieving contextually relevant code for specific change requests remains largely underexplored. To address this gap, we introduce RepoAlign-Bench, the first benchmark specifically designed to evaluate repository-level code retrieval under change request driven scenarios, encompassing 52k annotated instances. This benchmark shifts the retrieval paradigm from function-centric matching to holistic repository-level reasoning. Furthermore, we propose ReflectCode, an adversarial reflection augmented dual-tower architecture featuring disentangled code_encoder and doc_encoder components. ReflectCode dynamically integrates syntactic patterns, function dependencies, and semantic expansion intents through large language model guided reflection. Comprehensive experiments demonstrate that ReflectCode achieves 12.2% improvement in Top-5 Accuracy and 7.1% in Recall over state-of-the-art baselines, establishing a new direction for context-aware code retrieval.