Next poI recommendation improves by selecting task-focused agent groups
Learning Complementary Agent Groups via Task Hypergraphs for Next-POI Recommendation with Missing Modalities
Information Retrieval
Summary
Point-of-interest (POI) recommendation systems help suggest places to visit based on your past visits and descriptions of those places. However, these systems often lack complete information like text or images about the places. The authors propose a way to use multiple specialized AI agents that work together, each handling different parts of the task to understand and recommend POIs better even when some data is missing. Their approach learns which group of agents to pick for each recommendation task, improving accuracy over existing methods.
What this means in practice
- •For location-based service developers: Improve recommendation accuracy in location apps despite missing text or image data about places.
- •For mobile app engineers: Design recommendation features that dynamically assign AI agent groups based on available information per user request.
Authors
Jinze Wang, Yuze Liu, Liang Yuan, Gaoyuan Du, Xiaohang Xu, Yong Xiang, Jiong Jin
Abstract
Next point-of-interest (POI) recommendation uses visit histories and POI content, but text and images are often missing. Large language model agents have been applied to recommender systems for preference modeling and evidence acquisition. The challenge is to select agents that can jointly analyze the available evidence for each request. We propose \textsc{CATH} (\emph{Complementary Agents via Task Hypergraphs}), which learns to select agent groups for tasks such as comparing candidates or analyzing a POI with missing content. Agents with distinct processing roles are the task hypergraph nodes; each hyperedge connects the selected group and records its analysis goal and target POIs. The policy selects members conditioned on the request, task, and agents already selected. During execution, members share candidate-linked results that guide their subsequent operations. Completed tasks update candidate states and agent memories within the request. A shared ranking head converts candidate states into corrections to a fixed base ranking. Supervised ranking trains the executor, followed by policy optimization that accounts for computation cost. Experiments under natural missingness on Gowalla-CA, Foursquare-NYC, and Foursquare-TKY show 8.4--9.1\% relative NDCG@20 gains over HyperAgent4POI, the strongest prior baseline evaluated. Additional analyses compare group-selection and execution designs, test ranking under additional masking, and characterize quality--cost tradeoffs.