RALLM-POI: Retrieval-Augmented LLM for Zero-shot Next POI Recommendation with Geographical Reranking

2025-09-21

AI Summary

To address the weak geospatial relevance and excessive generalization of large language models (LLMs) in zero-shot next Point-of-Interest (POI) recommendation, this paper proposes RAG-GeoAgent: a retrieval-augmented, geography-aware framework. It first retrieves historical trajectories (HTR) to capture user mobility patterns; then applies geographic distance-based re-ranking (GDR) to enhance spatial awareness; finally, it introduces an agent-like, training-free self-refinement mechanism (ALR) that enables reflective fine-tuning of LLM outputs. This work is the first to integrate retrieval-augmented generation (RAG) with geography-informed re-ranking for zero-shot POI recommendation, and it pioneers a zero-training-cost self-reflective correction paradigm. Extensive experiments on three real-world Foursquare datasets demonstrate significant improvements over conventional models and state-of-the-art LLM-based baselines, with notable gains in recommendation accuracy. The source code is publicly available.

Technology Category

• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Planning, Routing, and Scheduling: Planning with Language Models
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Next point-of-interest (POI) recommendation predicts a user's next destination from historical movements. Traditional models require intensive training, while LLMs offer flexible and generalizable zero-shot solutions but often generate generic or geographically irrelevant results due to missing trajectory and spatial context. To address these issues, we propose RALLM-POI, a framework that couples LLMs with retrieval-augmented generation and self-rectification. We first propose a Historical Trajectory Retriever (HTR) that retrieves relevant past trajectories to serve as contextual references, which are then reranked by a Geographical Distance Reranker (GDR) for prioritizing spatially relevant trajectories. Lastly, an Agentic LLM Rectifier (ALR) is designed to refine outputs through self-reflection. Without additional training, RALLM-POI achieves substantial accuracy gains across three real-world Foursquare datasets, outperforming both conventional and LLM-based baselines. Code is released at https://github.com/LKRcrocodile/RALLM-POI.

Problem

• Predicting user's next destination from historical movement data
• Addressing generic or geographically irrelevant LLM recommendations
• Improving zero-shot POI recommendation without intensive training

Innovation

• Retrieves relevant historical trajectories as context
• Reranks trajectories by geographical distance priority
• Refines outputs through agentic LLM self-reflection