mlsanityData to 7 Oct 2026

Reasoning Over Space: Enabling Geographic Reasoning for LLM-Based Generative Next POI Recommendation

arXiv 2601.04562 ↗First posted 8 Jan 2026cs.AIcited by 2

On the map: LLM reasoning → Spatial and embodied reasoning

Abstract (arXiv)

Generative recommendation with large language models (LLMs) reframes prediction as sequence generation, yet existing LLM-based recommenders remain limited in leveraging geographic signals that are crucial in mobility and local-services scenarios. Here, we present Reasoning Over Space (ROS), a framework that utilizes geography as a vital decision variable within the reasoning process. ROS introduces a Hierarchical Spatial Semantic ID (SID) that discretizes coarse-to-fine locality and POI semantics into compositional tokens, and endows LLM with a three-stage Mobility Chain-of-Thought (CoT) paradigm that models user personality, constructs an intent-aligned candidate space, and performs locality informed pruning. We further align the model with real world geography via spatial-guided Reinforcement Learning (RL). Experiments on three widely used location-based social network (LBSN) datasets show that ROS achieves over 10% relative gains in hit rate over strongest LLM-based baselines and improves cross-city transfer, despite using a smaller backbone model.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning. Builds on 11 of its 49 references that are in the map. First posted on arXiv 8 Jan 2026.

Builds on

11 of 49 references.

References · 11

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  2. 2025Qwen3 Technical Reportcited by 8,195
  3. 2015Session-based Recommendations with Recurrent Neural Networkscited by 685
  4. 2023Plan-and-Solve Prompting: Improving Zero-Shot Chain-of-Thought Reasoning by Large Language Modelscited by 550
  5. 2025OneRec: Unifying Retrieve and Rank with Generative Recommender and Iterative Preference Alignmentcited by 167
  6. 2024HLLM: Enhancing Sequential Recommendations via Hierarchical Large Language Models for Item and User Modelingcited by 63
  7. 2025OneLoc: Geo-Aware Generative Recommender Systems for Local Life Servicecited by 5
  8. 2025LLM-Aligned Geographic Item Tokenization for Local-Life Recommendationcited by 3
  9. 2025Geography-Aware Large Language Models for Next POI Recommendationcited by 2
  10. 2025Spacetime-GR: A Spacetime-Aware Generative Model for Large Scale Online POI Recommendationcited by 2
  11. 2025Cognitive-Aligned Spatio-Temporal Large Language Models For Next Point-of-Interest Predictioncited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026Think Before You Act: Intention-Guided Reasoning for LLM-Based Location Predictioncited by 0
  2. 2026Interpretable Representation via LLM-Driven Generative Disentanglement for Local-Life Service Recommendationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Why Users Go There: World Knowledge-Augmented Generative Next POI Recommendation
  2. 2026Think2Go: Generative Next POI Recommendation with LLM Reasoning
  3. 2026SemPOI-RL: Aligning LLM Semantic Reasoning for Interpretable Out-of-Town POI Sequential Generation
  4. 2026An Empirical Evaluation of Cross-City POI Recommendation on a Large-Scale Benchmark
  5. 2026Generative Spatiotemporal Intent Sequence Recommendation via Implicit Reasoning in Amap
  6. 2026Enhancing Human Mobility Prediction with Spatially Aware LLM-based Multi-Agent Systems
  7. 2025TOOL4POI: A Tool-Augmented LLM Framework for Next POI Recommendation
  8. 2025AskNearby: An LLM-Based Application for Neighborhood Information Retrieval and Personalized Cognitive-Map Recommendations