Spatial cognition is fundamental to real-world multimodal intelligence, allowing models to effectively interact with the physical environment. While multimodal large language models (MLLMs) have made significant strides, existing benchmarks often oversimplify spatial cognition, reducing it to a single-dimensional metric, which fails to capture the hierarchical structure and interdependence of spatial abilities. T o address this gap, we propose a hierarchical spatial cognition framework that decomposes spatial intelligence into five progressively complex levels from basic observation to high-level planning. Building upon this taxonomy, we construct SpatialBench, a large-scale, fine-grained benchmark covering 15 tasks aligned with these cognitive levels. T o provide a unified evaluation across heterogeneous tasks, we further introduce a high-level capability-oriented metric that reliably assesses a model's overall spatial reasoning ability. Extensive experiments over massive MLLMs reveal distinct performance stratification across cognitive levels: models exhibit strong perceptual grounding yet remain limited in symbolic reasoning, causal inference, and planning. Additional human tests demonstrate that humans perform selective, goal-directed abstraction, while MLLMs tend to over-attend to surface details without coherent spatial intent. Our work establishes the first systematic framework for measuring hierarchical spatial cognition in MLLMs, laying the foundation for future spatially intelligent systems.

Neural Radiance Fields (NeRFs) have demonstrated remarkable capabilities in 3D reconstruction and novel view synthesis. However, most existing NeRF frameworks require complete retraining when new views are introduced incrementally, limiting their applicability in domains where data arrives sequentially. This limitation is particularly problematic in satellite-based terrain analysis, where regions are repeatedly observed over time. Incremental refinement of NeRFs remains underexplored, and naive approaches suffer from catastrophic forgetting when past data is unavailable. We propose $Δ$-NeRF, a unique modular residual framework for incremental NeRF refinement. $Δ$-NeRF introduces several novel techniques including: (1) a residual controller that injects per-layer corrections into a frozen base NeRF, enabling refinement without access to past data; (2) an uncertainty-aware gating mechanism that prevents overcorrection by adaptively combining base and refined predictions; and (3) a view selection strategy that reduces training data by up to 47\% while maintaining performance. Additionally, we employ knowledge distillation to compress the enhanced model into a compact student network (20\% of original size). Experiments on satellite imagery demonstrate that $Δ$-NeRF achieves performance comparable to joint training while reducing training time by 30-42\%. $Δ$-NeRF consistently outperforms existing baselines, achieving an improvement of up to 43.5\% in PSNR over naive fine-tuning and surpassing joint training on some metrics.

The planning, management, and resource scheduling of cellular mobile networks require joint estimation of mobile traffic across different layers and nodes. Mobile traffic generation can proactively anticipate user demands and capture the dynamics of network load. However, existing methods mainly focus on generating traffic at a single spatiotemporal resolution, making it difficult to jointly model multi-scale traffic patterns. In this paper, we propose ZoomDiff, a diffusion-based model for multi-scale mobile traffic generation. ZoomDiff maps urban environmental context into mobile traffic with multiple spatial and temporal resolutions through a set of customized Denoising Refinement Diffusion Models (DRDM). DRDM employs a multi-stage noise-adding and denoising mechanism, enabling different stages to generate traffic at distinct spatiotemporal resolutions. This design aligns the progressive denoising process with hierarchical network layers, including base stations, cells, and grids of varying granularities. Experiments on real-world mobile traffic datasets show that ZoomDiff achieves at least an 18.4% improvement over state-of-the-art baselines in multi-scale traffic generation tasks. Moreover, ZoomDiff demonstrates strong efficiency and cross-city generalization, highlighting its potential as a powerful generative framework for modeling multi-scale mobile network dynamics.

Multi-object tracking (MOT) is one of the most challenging tasks in computer vision, where it is important to correctly detect objects and associate these detections across frames. Current approaches mainly focus on tracking objects in each frame of a video stream, making it almost impossible to run the model under conditions of limited computing resources. T o address this issue, we propose Stable-Track, a novel approach that stabilizes the quality of tracking on low-frequency detections. Our method introduces a new two-stage matching strategy to improve the cross-frame association between low-frequency detections. W e propose a novel Bbox-Based Distance instead of the conventional Mahalanobis distance, which allows us to effectively match objects using the Re-ID model. Furthermore, we integrate visual tracking into the Kalman Filter and the overall tracking pipeline. Our method outperforms current state-of-the-art trackers in the case of low-frequency detections, achieving 11.6% HOTA improvement at 1 Hz on MOT17-val, while keeping up with the best approaches on the standard MOT17, MOT20, and DanceTrack benchmarks with full-frequency detections.

Vision-Language-Action (VLA) models inherit strong priors from pretrained Vision-Language Models (VLMs), but naive fine-tuning often disrupts these representations and harms generalization. Existing fixes -- freezing modules or applying uniform regularization -- either overconstrain adaptation or ignore the differing roles of VLA components. We present MAPS (Module-Wise Proximity Scheduling), the first robust fine-tuning framework for VLAs. Through systematic analysis, we uncover an empirical order in which proximity constraints should be relaxed to balance stability and flexibility. MAPS linearly schedules this relaxation, enabling visual encoders to stay close to their pretrained priors while action-oriented language layers adapt more freely. MAPS introduces no additional parameters or data, and can be seamlessly integrated into existing VLAs. Across MiniVLA-VQ, MiniVLA-OFT, OpenVLA-OFT, and challenging benchmarks such as SimplerEnv, CALVIN, LIBERO, as well as real-world evaluations on the Franka Emika Panda platform, MAPS consistently boosts both in-distribution and out-of-distribution performance (up to +30%). Our findings highlight empirically guided proximity to pretrained VLMs as a simple yet powerful principle for preserving broad generalization in VLM-to-VLA transfer.

Road network representation learning (RNRL) has attracted increasing attention from both researchers and practitioners as various spatiotemporal tasks are emerging. Recent advanced methods leverage Graph Neural Networks (GNNs) and contrastive learning to characterize the spatial structure of road segments in a self-supervised paradigm. However, spatial heterogeneity and temporal dynamics of road networks raise severe challenges to the neighborhood smoothing mechanism of self-supervised GNNs. To address these issues, we propose a $\textbf{D}$ual-branch $\textbf{S}$patial-$\textbf{T}$emporal self-supervised representation framework for enhanced road representations, termed as DST. On one hand, DST designs a mix-hop transition matrix for graph convolution to incorporate dynamic relations of roads from trajectories. Besides, DST contrasts road representations of the vanilla road network against that of the hypergraph in a spatial self-supervised way. The hypergraph is newly built based on three types of hyperedges to capture long-range relations. On the other hand, DST performs next token prediction as the temporal self-supervised task on the sequences of traffic dynamics based on a causal Transformer, which is further regularized by differentiating traffic modes of weekdays from those of weekends. Extensive experiments against state-of-the-art methods verify the superiority of our proposed framework. Moreover, the comprehensive spatiotemporal modeling facilitates DST to excel in zero-shot learning scenarios.

Spatial understanding remains a weakness of Large Vision-Language Models (LVLMs). Existing supervised fine-tuning (SFT) and recent reinforcement learning with verifiable rewards (RLVR) pipelines depend on costly supervision, specialized tools, or constrained environments that limit scale. We introduce Spatial-SSRL, a self-supervised RL paradigm that derives verifiable signals directly from ordinary RGB or RGB-D images. Spatial-SSRL automatically formulates five pretext tasks that capture 2D and 3D spatial structure: shuffled patch reordering, flipped patch recognition, cropped patch inpainting, regional depth ordering, and relative 3D position prediction. These tasks provide ground-truth answers that are easy to verify and require no human or LVLM annotation. Training on our tasks substantially improves spatial reasoning while preserving general visual capabilities. On seven spatial understanding benchmarks in both image and video settings, Spatial-SSRL delivers average accuracy gains of 4.63% (3B) and 3.89% (7B) over the Qwen2.5-VL baselines. Our results show that simple, intrinsic supervision enables RLVR at scale and provides a practical route to stronger spatial intelligence in LVLMs.

Visual-Language Navigation (VLN) is a fundamental challenge in robotic systems, with broad applications for the deployment of embodied agents in real-world environments. Despite recent advances, existing approaches are limited in long-range spatial reasoning, often exhibiting low success rates and high inference latency, particularly in long-range navigation tasks. To address these limitations, we propose FSR-VLN, a vision-language navigation system that combines a Hierarchical Multi-modal Scene Graph (HMSG) with Fast-to-Slow Navigation Reasoning (FSR). The HMSG provides a multi-modal map representation supporting progressive retrieval, from coarse room-level localization to fine-grained goal view and object identification. Building on HMSG, FSR first performs fast matching to efficiently select candidate rooms, views, and objects, then applies VLM-driven refinement for final goal selection. We evaluated FSR-VLN across four comprehensive indoor datasets collected by humanoid robots, utilizing 87 instructions that encompass a diverse range of object categories. FSR-VLN achieves state-of-the-art (SOTA) performance in all datasets, measured by the retrieval success rate (RSR), while reducing the response time by 82% compared to VLM-based methods on tour videos by activating slow reasoning only when fast intuition fails. Furthermore, we integrate FSR-VLN with speech interaction, planning, and control modules on a Unitree-G1 humanoid robot, enabling natural language interaction and real-time navigation.

Vision-language-action (VLA) models have shown remarkable capabilities in robotic manipulation, but their performance is sensitive to the $\textbf{action chunk length}$ used during training, termed $\textbf{horizon}$. Our empirical study reveals an inherent trade-off: longer horizons provide stronger global foresight but degrade fine-grained accuracy, while shorter ones sharpen local control yet struggle on long-term tasks, implying fixed choice of single horizons being suboptimal. To mitigate the trade-off, we propose a $\textbf{mixture of horizons (MoH)}$ strategy. MoH rearranges the action chunk into several segments with different horizons, processes them in parallel with a shared action transformer, and fuses outputs with a light linear gate. It has three appealing benefits. 1) MoH exploits long-term foresight and short-term precision jointly within a single model, improving both performance and generalizability to complex tasks. 2) MoH is plug-and-play for full-attention action modules with minimal training or inference overhead. 3) MoH enables dynamic inference with adaptive horizons, which selects stable actions through cross-horizon consensus, achieving 2.5$\times$ higher throughput than baselines while preserving superior performance. Extensive experiments over flow-based policies $π_0$, $π_{0.5}$, and one-step regression policy $π_{\text{reg}}$ demonstrate that MoH yields consistent and significant gains on both simulations and real-world tasks. Notably, under mixed-task setting, $π_{0.5}$ with MoH reaches a new state-of-the-art with 99$\%$ average success rate on LIBERO after only $30k$ training iterations. Project page: https://github.com/Timsty1/MixtureOfHorizons

Graph transformers have demonstrated remarkable capability on complex spatio-temporal tasks, yet their depth is often limited by oversmoothing and weak long-range dependency modeling. To address these challenges, we introduce GRIT -LP, a graph transformer explicitly designed for polar ice-layer thickness estimation from polar radar imagery. Accurately estimating ice layer thickness is critical for understanding snow accumulation, reconstructing past climate patterns and reducing uncertainties in projections of future ice sheet evolution and sea level rise. GRIT -LP combines an inductive geometric graph learning framework with self-attention mechanism, and introduces two major innovations that jointly address challenges in modeling the spatio-temporal patterns of ice layers: a partitioned spatial graph construction strategy that forms overlapping, fully connected local neighborhoods to preserve spatial coherence and suppress noise from irrelevant long-range links, and a long-range skip connection mechanism within the transformer that improves information flow and mitigates oversmooth-ing in deeper attention layers. We conducted extensive experiments, demonstrating that GRIT -LP outperforms current state-of-the-art methods with a 24.92% improvement in root mean squared error. These results highlight the effectiveness of graph transformers in modeling spatiotemporal patterns by capturing both localized structural features and long-range dependencies across internal ice layers, and demonstrate their potential to advance data-driven understanding of cryospheric processes. Introduction Graph transformers have proven to be highly effective for modeling complex graph-structured data, with wide-range of applications in real-world scenarios, particularly those involving spatiotemporal patterns. Their ability to capture intricate relationships and dependencies makes them highly valuable in domains such as pedestrian trajectory prediction [1] and traffic prediction [2]. Despite their success, current graph transformer architectures face notable limitations, including overfitting and over-smoothing--a phenomenon where node features become indistinguishable as layers deepen [3]. Additionally, many existing graph transformers are relatively shallow, limiting their ability to effectively capture the complex, long-range dependencies that often emerge in real-world datasets.