Preprint
Aug 2026
SSR-GRPO: Integrating Supervision and Semantic IDs into Reinforcement Learning for Dense Retrieval in E-commerce
This paper proposes Supervised Retrieval-GRPO with Semantic Identifiers (SSR-GRPO), a dual-perspective framework for relevance assessment that leverages both Semantic Identifiers produced by quantization learning and dense representation vectors to generate more unbiased relevance scores.
Guangxin Song, Xingye Fang, Ming-Min Jin et al.
· 0 citations