Optimizing Retrieval Components For A Shared Backbone Via Component-wise Multi-stage Training

Abstract

Recent advances in embedding-based retrieval have enabled dense retrievers to serve as core infrastructure in many industrial systems, where a single retrieval backbone is often shared across multiple downstream applications. In such settings, retrieval quality directly constrains system performance and extensibility, while coupling model selection, deployment, and rollback decisions across applications. In this paper, we present empirical findings and a system-level solution for optimizing retrieval components deployed as a shared backbone in production legal retrieval systems. We adopt a multi-stage optimization framework for dense retrievers and rerankers, and show that different retrieval components exhibit stage-dependent trade-offs. These observations motivate a component-wise, mixed-stage configuration rather than relying on a single uniformly optimal checkpoint. The resulting backbone is validated through end-to-end evaluation and deployed as a shared retrieval service suppor

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keyli2026optimizing

Related papers

• Efficient And Effective Tail Latency Minimization In Multi-stage Retrieval Systems (2017)11.76
• Optimizing Compound Retrieval Systems (2025)0.00
• L^2R: Lifelong Learning For First-stage Retrieval With Backward-compatible Representations (2023)5.24
• Taxonomy Of The Retrieval System Framework: Pitfalls And Paradigms (2026)0.00
• Dynamic Trade-off Prediction In Multi-stage Retrieval Systems (2016)11.93
• Align Then Train: Efficient Retrieval Adapter Learning (2026)0.00
• Ultron: An Ultimate Retriever On Corpus With A Model-based Indexer (2022)0.00
• Revisiting Neural Retrieval On Accelerators (2023)9.41