GeoGR: A Generative Retrieval Framework for Spatio-Temporal Aware POI Recommendation

TL;DR

GeoGR tackles next-POI prediction in large-scale, navigation-centric LBS by learning geo-aware Semantic IDs (SIDs) and aligning LLMs through a two-stage process: geo-aware SID construction via spatio-temporal co-visited POI pairs and RQ-Kmeans quantization, followed by CPT and SFT to enable autoregressive POI generation conditioned on real-time context. An EM-style SID refinement further improves SID-LLM alignment, while CPT training aligns SID tokens with domain prompts before instruction-tuning for next-POI generation. Offline results on NYC/TKY and a large AMAP dataset show substantial gains over baselines, and online deployment on AMAP yields measurable engagement improvements with low latency. Overall, GeoGR advances generative retrieval for location-based services by marrying spatial semantics with scalable, end-to-end LLM fine-tuning in production.

Abstract

Next Point-of-Interest (POI) prediction is a fundamental task in location-based services, especially critical for large-scale navigation platforms like AMAP that serve billions of users across diverse lifestyle scenarios. While recent POI recommendation approaches based on SIDs have achieved promising, they struggle in complex, sparse real-world environments due to two key limitations: (1) inadequate modeling of high-quality SIDs that capture cross-category spatio-temporal collaborative relationships, and (2) poor alignment between large language models (LLMs) and the POI recommendation task. To this end, we propose GeoGR, a geographic generative recommendation framework tailored for navigation-based LBS like AMAP, which perceives users' contextual state changes and enables intent-aware POI recommendation. GeoGR features a two-stage design: (i) a geo-aware SID tokenization pipeline that explicitly learns spatio-temporal collaborative semantic representations via geographically constrained co-visited POI pairs, contrastive learning, and iterative refinement; and (ii) a multi-stage LLM training strategy that aligns non-native SID tokens through multiple template-based continued pre-training(CPT) and enables autoregressive POI generation via supervised fine-tuning(SFT). Extensive experiments on multiple real-world datasets demonstrate GeoGR's superiority over state-of-the-art baselines. Moreover, deployment on the AMAP platform, serving millions of users with multiple online metrics boosting, confirms its practical effectiveness and scalability in production.

Figures (3)

• Figure 1: (A) SID-based POI retrieval with GeoGR. (B) The spatio-temporal aware POI Recommendation.
• Figure 2: Two stages for obtaining GeoGR: (A) Construct the POI semantic ID; (B) Train the generative POI recommendation.
• Figure 3: Visualization of POI representations before and after applying the spatio-temporal collaborative optimization.