Static Pruning Across Sparse Retrieval Regimes: What Transfers, What Breaks, and What Still Helps

2026-08-17

AI Summary

This study addresses the lack of cross-engine portability validation in static pruning by presenting the first systematic evaluation of sparse retrieval pruning strategies across multiple engines and encoders. We propose a universal stopping criterion and a hybrid dynamic-static optimization method. Our analysis reveals the generalizability of index pruning versus the limitations of query pruning, supported theoretically by cache and TLB performance profiling. Experiments demonstrate that index pruning reduces latency by 1.2–6.6× and storage size by 18–82%. Furthermore, combined pruning achieves a 2.5× speedup with negligible NDCG@10 degradation (<0.003). These findings provide a robust empirical and theoretical foundation for optimizing modern sparse retrieval systems, confirming the efficacy of portable pruning strategies across diverse computational environments.

Abstract

Static pruning is widely used to accelerate sparse neural retrieval, yet existing studies each validate their conclusions within a single custom pipeline, leaving it unclear which findings transfer to modern engines with different index organizations and dynamic pruning mechanisms. We present the first cross-engine pruning portability study, evaluating static pruning strategies across three engines - a controlled C++ pipeline (exhaustive inverted index), BMP (block-max pruning), and SEISMIC (clustered inverted indexes) - on two benchmarks (MS MARCO, Natural Questions) with two encoders spanning opposite query-density regimes (SPLADE: 44 avg. query terms; V3-GTE: 7 avg. query terms), totaling 1,140 experimental configurations, with an additional deep-judgment validation on TREC DL 2019/2020. We find that index-side pruning (document and posting-list) is portable: it consistently reduces latency (1.2-6.6$\times$) and index size (18-82%) across all engines because sparse retrieval is memory-bound - a conclusion we support with cache-miss, TLB, and IPC profiling. In contrast, query pruning is already internalized by modern engines: it yields 4-11$\times$ speedup on the exhaustive pipeline but is subsumed by BMP's $β$ and SEISMIC's query_cut. Static pruning complements dynamic pruning: on BMP, combining document and query reduction yields 2.5$\times$ speedup with NDCG@10 within 0.003 of the exact baseline. Finally, NDCG@10 saturates while Recall@10 is still in the ${\sim}$85-95% range across all three engines, providing a portable stopping criterion: practitioners can push pruning to this knee without visible ranking degradation. Together, these findings answer what transfers (index-side pruning), what breaks (query pruning), and what still helps (static atop dynamic pruning).

Problem

• Static Pruning
• Sparse Retrieval
• Cross-engine Portability
• Dynamic Pruning

Innovation

• Static Pruning
• Sparse Retrieval
• Cross-engine Portability
• Dynamic Pruning
• Index-side Pruning