HierLoc: Hyperbolic Entity Embeddings for Hierarchical Visual Geolocation

Visual geolocalization, the task of predicting where an image was taken, remains challenging due to global scale, visual ambiguity, and the inherently hierarchical structure of geography. Existing paradigms rely on either large-scale retrieval, which requires storing a large number of image embeddings, grid-based classifiers that ignore geographic continuity, or generative models that diffuse over space but struggle with fine detail. We introduce an entity-centric formulation of geolocation that replaces image-to-image retrieval with a compact hierarchy of geographic entities embedded in Hyperbolic space. Images are aligned directly to country, region, subregion, and city entities through Geo-Weighted Hyperbolic contrastive learning by directly incorporating haversine distance into the contrastive objective. This hierarchical design enables interpretable predictions and efficient inference with 240k entity embeddings instead of over 5 million image embeddings on the OSV5M benchmark, on which our method establishes a new state-of-the-art performance. Compared to the current methods in the literature, it reduces mean geodesic error by 19.5\%, while improving the fine-grained subregion accuracy by 43%. These results demonstrate that geometry-aware hierarchical embeddings provide a scalable and conceptually new alternative for global image geolocation.

Robust Energy-Efficient Sleep-Mode Strategy for Multi-RIS-Aided Cell-Free Massive MIMO

With the explosive growth of data traffic and the ubiquitous connectivity of wireless devices, the energy demands of wireless networks have inevitably escalated. Reconfigurable intelligent surface (RIS) has emerged as a promising solution for 6G networks due to its energy efficiency (EE) and low cost, while cell-free massive multiple-input multiple-output (CF-mMIMO) was proposed as an innovative network architecture without fixed cell boundaries to enhance these measures even further. However, existing studies often assume consistently high traffic loads, neglecting the dynamic nature of user demand. This can result in underutilized access points (APs) and unnecessary energy expenditure during low-demand periods. To tackle the challenge of EE in CF-mMIMO systems during low load periods, this paper proposes a novel energy-efficient transmission scheme that jointly coordinates active APs and multiple passive RISs. Specifically, a dynamic AP sleep-mode strategy is designed, where certain APs are selectively deactivated while nearby RISs assist in maintaining coverage. We formulate the EE maximization objective as a fractional programming problem and adopt the Dinkelbach method in conjunction with alternating optimization (AO) to iteratively solve the three coupled subproblems: (i) AP selection via a hybrid branch-and-bound (BnB) and greedy algorithm, (ii) transmit power optimization using a sequential convex approximation (SCA) method, initialized by a heuristic zero-forcing strategy, and (iii) RIS phase shift optimization using gradient projection. Simulation results show that the proposed scheme achieves significantly higher EE than existing methods in both low and moderate user scenarios.