Researchers have developed MARS, a multi-agent re-ranking framework designed for repeat-order food delivery recommendations. This modular system integrates collaborative filtering signals with large language models (LLMs) for a two-stage recommendation process: cuisine prediction and vendor ranking. MARS combines global preference signals from LightGCN, local peer evidence, geospatial filtering, and LLM reasoning over various contexts. The framework was evaluated on real-world benchmarks from Delivery Hero, demonstrating competitive performance when pre-trained LLMs are paired with lightweight collaborative retrieval. AI
IMPACT This framework could improve the efficiency and personalization of food delivery recommendations by integrating LLMs into existing recommendation pipelines.
RANK_REASON The item describes a research paper presenting a new framework for recommender systems. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
