Researchers have developed a new method called SSR-GRPO to improve dense retrieval in e-commerce search. This approach integrates supervised learning and semantic identifiers with reinforcement learning to address limitations in existing methods, such as noisy candidates and biased relevance assessments. SSR-GRPO utilizes a dual-perspective framework for relevance scoring and mines hard negative samples to refine the model's ability to distinguish fine-grained semantic differences. Extensive experiments have confirmed the effectiveness of SSR-GRPO, leading to its deployment on a large-scale e-commerce platform. AI
IMPACT Enhances e-commerce search capabilities by improving the accuracy and relevance of product recommendations.
RANK_REASON The cluster describes a new academic paper detailing a novel method for information retrieval. [lever_c_demoted from research: ic=1 ai=1.0]
Read on arXiv cs.IR (Information Retrieval) →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
