Abstract:Large language models (LLMs) have shown promising potential for next Point-of-Interest (POI) recommendation. However, existing methods only perform direct zero-shot prompting, leading to ineffective extraction of user preferences, insufficient injection of collaborative signals, and a lack of user privacy protection. As such, we propose a novel Multitask Reflective Large Language Model for Privacy-preserving Next POI Recommendation (MRP-LLM), aiming to exploit LLMs for better next POI recommendation while preserving user privacy. Specifically, the Multitask Reflective Preference Extraction Module first utilizes LLMs to distill each user's fine-grained (i.e., categorical, temporal, and spatial) preferences into a knowledge base (KB). The Neighbor Preference Retrieval Module retrieves and summarizes the preferences of similar users from the KB to obtain collaborative signals. Subsequently, aggregating the user's preferences with those of similar users, the Multitask Next POI Recommendation Module generates the next POI recommendations via multitask prompting. Meanwhile, during data collection, a Privacy Transmission Module is specifically devised to preserve sensitive POI data. Extensive experiments on three real-world datasets demonstrate the efficacy of our proposed MRP-LLM in providing more accurate next POI recommendations with user privacy preserved.

What problem does this paper attempt to address?

The problems that this paper attempts to solve mainly focus on the following aspects: 1. **Inadequate extraction of user preferences**: Existing Point - of - Interest (POI) recommendation systems based on large language models (LLMs) perform poorly in extracting fine - grained preferences (such as category, time, and spatial preferences) from users' historical check - in records. This leads to insufficient understanding of user preferences, which in turn affects the accuracy of the recommendation system. 2. **Insufficient injection of collaborative signals**: Traditional POI recommendation systems obtain global behavior patterns by analyzing the behaviors of similar users, thereby improving accuracy and generalization ability. However, existing methods based on LLMs only rely on users' own check - in histories and fail to fully utilize the collaborative information of other users, resulting in limited recommendation effects. 3. **Lack of privacy protection**: Existing methods directly expose users' check - in histories to LLMs, and these data may contain sensitive information (such as users' locations, home addresses, and behavior habits). This practice may lead to the leakage of sensitive information and damage users' trust in the recommendation system. To solve the above problems, the author proposes a Multi - task Reflective - type large language model (MRP - LLM), aiming to improve the Point - of - Interest recommendation system in the following ways: - **Multi - task Reflective - type Preference Extraction Module**: Use LLMs to distill fine - grained preferences (including category, time, and spatial preferences) from users' check - in records and store them in the knowledge base (KB). - **Neighbor Preference Retrieval Module**: Retrieve and summarize the preferences of similar users from KB to introduce collaborative signals and enhance the recommendation effect. - **Multi - task Next - Point - of - Interest Recommendation Module**: Combine the preferences of users and their similar users to generate more accurate Next - Point - of - Interest recommendations. - **Privacy Transmission Module**: Adopt techniques such as differential privacy to protect the data uploaded by users and ensure user privacy and security. Through these improvements, MRP - LLM not only improves the accuracy of the recommendation system but also provides a powerful user data privacy protection mechanism.

MRP-LLM: Multitask Reflective Large Language Models for Privacy-Preserving Next POI Recommendation

Large Language Models for Next Point-of-Interest Recommendation

Unleashing the Power of Large Language Models for Group POI Recommendations

Explainable next POI recommendation based on spatial-temporal disentanglement representation and pseudo profile generation

Privacy in LLM-based Recommendation: Recent Advances and Future Directions

Where to Move Next: Zero-shot Generalization of LLMs for Next POI Recommendation

Hierarchical and Multi-resolution Preference Modeling for Next POI Recommendation

An efficient privacy-preserving point-of-interest recommendation model based on local differential privacy

Aligning Large Language Model with Direct Multi-Preference Optimization for Recommendation

Re2LLM: Reflective Reinforcement Large Language Model for Session-based Recommendation

Decentralized Collaborative Learning Framework for Next POI Recommendation

PALR: Personalization Aware LLMs for Recommendation

Embracing LLMs for Point-of-Interest Recommendations

LLMs for User Interest Exploration in Large-scale Recommendation Systems

LLMRG: Improving Recommendations through Large Language Model Reasoning Graphs

Interpretable Embeddings for Next Point-of-Interest Recommendation via Large Language Model Question–Answering

Bridging the Information Gap Between Domain-Specific Model and General LLM for Personalized Recommendation

Transparent and Scrutable Recommendations Using Natural Language User Profiles

PrivacyRestore: Privacy-Preserving Inference in Large Language Models via Privacy Removal and Restoration

Personalized Long- and Short-term Preference Learning for Next POI Recommendation

Harnessing Multimodal Large Language Models for Multimodal Sequential Recommendation