Mine And Refine: Optimizing Graded Relevance In E-commerce Search Retrieval

Abstract

We propose a two-stage "Mine and Refine" contrastive training framework for semantic text embeddings to enhance multi-category e-commerce search retrieval. Large scale e-commerce search demands embeddings that generalize to long tail, noisy queries while adhering to scalable supervision compatible with product and policy constraints. A practical challenge is that relevance is often graded: users accept substitutes or complements beyond exact matches, and production systems benefit from clear separation of similarity scores across these relevance strata for stable hybrid blending and thresholding. To obtain scalable policy consistent supervision, we fine-tune a lightweight LLM on human annotations under a three-level relevance guideline and further reduce residual noise via engagement driven auditing. In Stage 1, we train a multilingual Siamese two-tower retriever with a label aware supervised contrastive objective that shapes a robust global semantic space. In Stage 2, we mine hard sam

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keyxi2026mine

Related papers

• Retrieval-grpo: A Multi-objective Reinforcement Learning Framework For Dense Retrieval In Taobao Search (2025)0.00
• Enhancing Relevance Of Embedding-based Retrieval At Walmart (2024)7.16
• Unified Embedding Based Personalized Retrieval In Etsy Search (2023)2.26
• MRSE: An Efficient Multi-modality Retrieval System For Large Scale E-commerce (2024)0.00
• Unified Supervision For Walmart's Sponsored Search Retrieval Via Joint Semantic Relevance And Behavioral Engagement Modeling (2026)0.00
• Query-life: Query-aware Language Image Fusion Embedding For E-commerce Relevance (2023)0.00
• Embedding-based Product Retrieval In Taobao Search (2021)13.70
• Large Reasoning Embedding Models: Towards Next-generation Dense Retrieval Paradigm (2025)0.00