OPERA: Online Data Pruning for Efficient Retrieval Model Adaptation

2026-03-17

AI Summary

This work addresses the inefficiency and suboptimal performance of dense retrieval models in domain adaptation, where training data often exhibit redundancy and uneven contribution. To tackle this, we propose OPERA, a framework that integrates similarity-based static pruning with query-document-level dynamic sampling probability adjustment to efficiently select high-quality training samples during fine-tuning. OPERA explicitly reveals the trade-off between data quality and query coverage and introduces a two-stage dynamic pruning mechanism that enhances model performance while preserving data diversity. Evaluated across eight cross-domain datasets, OPERA achieves an average improvement of 1.9% in NDCG@10 and 0.7% in Recall@20, reduces training time by over 50%, and attains a strong overall ranking of 1.38, significantly outperforming existing approaches.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Data Mining & Knowledge Management: Intelligent Query Processing

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Domain-specific finetuning is essential for dense retrievers, yet not all training pairs contribute equally to the learning process. We introduce OPERA, a data pruning framework that exploits this heterogeneity to improve both the effectiveness and efficiency of retrieval model adaptation. We first investigate static pruning (SP), which retains only high-similarity query-document pairs, revealing an intrinsic quality-coverage tradeoff: ranking (NDCG) improves while retrieval (Recall) can degrade due to reduced query diversity. To resolve this tradeoff, we propose a two-stage dynamic pruning (DP) strategy that adaptively modulates sampling probabilities at both query and document levels throughout training, prioritizing high-quality examples while maintaining access to the full training set. Evaluations across eight datasets spanning six domains demonstrate the effectiveness of both approaches: SP improves ranking over standard finetuning (NDCG@10 +0.5\%), while DP achieves the strongest performance on both ranking (NDCG@10 +1.9\%) and retrieval (Recall@20 +0.7\%), with an average rank of 1.38 across all methods. These findings scale to Qwen3-Embedding, an LLM-based dense retriever, confirming architecture-agnostic benefits. Notably, DP reaches comparable performance in less than 50\% of the training time required by standard finetuning.

Problem

• data pruning
• dense retrieval
• model adaptation
• training efficiency
• domain-specific finetuning

Innovation

• data pruning
• dense retrieval
• dynamic sampling
• model adaptation
• efficiency