Generative Recommendation Framework GRP Integrates Retrieval and Ranking

Researchers have introduced GRP, a generative recommendation framework that integrates retrieval, ranking, and reward modeling into a single encoder-decoder model. This approach aims to streamline industrial recommendation systems by enabling joint replacement of components. The framework generates multimodal Semantic IDs and uses a jointly trained ranking module to score candidates, with optimizations reducing retrieval latency by 69%. Early experiments show improvements in view time and shares when GRP is used as a retrieval source or replaces weaker sources. AI

IMPACT This framework could streamline industrial recommendation systems by integrating multiple components into a single model, potentially improving efficiency and performance.

RANK_REASON Academic paper detailing a new framework for recommendation systems. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.IR (Information Retrieval) →

• alphaXiv
• arXiv
• CatalyzeX
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Litmaps
• mGRPO
• ScienceCast
• scite Smart Citations

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →