Source-free domain adaptive semantic segmentation aims at adapting a model trained on the source domain to the target domain without requiring access to the source data. Self-training has emerged as a leading approach to address this challenging problem. However, without robust denoising mechanisms to reduce the noise in pseudo labels, it still easily fall into biased estimates. Most existing methods address this issue by introducing novel architectures, but often at the cost of increased model complexity or reliance on additional input modalities. Different from previous studies, this article introduces UniSFDA, a unified multi-stage self-training framework that integrates cross-model transfer learning, uncertainty-aware pseudo label fusion, and intra-domain style augmentation, thereby enhancing both segmentation accuracy and test-time efficiency. Our proposed framework offers exceptional flexibility, with each component being independent and ready to be integrated into any existing self-training framework. Additionally, we investigate the performance of various representative segmentation models, including DeepLabv2, SegFormer, DFormer, and ViT-Adapter, within our framework. It is worth noting that UniSFDA is model-agnostic, allowing both source and target networks to be instantiated with arbitrary segmentation architectures, and thus readily benefiting from future advances in segmentation models. Experiments on the GTA5-Cityscapes and SYNTHIA-Cityscapes benchmarks demonstrate the effectiveness of our framework. With DeepLabv2 (SegFormer) as the source model, UniSFDA establishes new state-of-the-art performance, achieving mIoU scores of 61.8% (65.4%) and 57.9% (59.6%) on the two benchmarks, respectively.

Sensory Temporal Action Detection (STAD) aims to localize and classify human actions within long, untrimmed sequences captured by non-visual sensors such as WiFi or inertial measurement units (IMUs). Unlike video-based TAD, STAD poses unique challenges due to the low-dimensional, noisy, and heterogeneous nature of sensory data, as well as the real-time and resource constraints on edge devices. While recent STAD models have improved detection performance, their high computational cost hampers practical deployment. In this paper, we propose SlimSTAD, a simple yet effective framework that achieves both high accuracy and low latency for STAD. SlimSTAD features a novel Decoupled Channel Modeling (DCM) encoder, which preserves modality-specific temporal features and enables efficient inter-channel aggregation via lightweight graph attention. An anchor-free cascade predictor then refines action boundaries and class predictions in a two-stage design without dense proposals. Experiments on two real-world datasets demonstrate that SlimSTAD outperforms strong video-derived and sensory baselines by an average of 2.1 mAP, while significantly reducing GFLOPs, parameters, and latency, validating its effectiveness for real-world, edge-aware STAD deployment.

Large Language Models (LLMs) have recently shown promise in addressing combinatorial optimization problems (COPs) through prompt-based strategies. However, their scalability and generalization remain limited, and their effectiveness diminishes as problem size increases, particularly in routing problems involving more than 30 nodes. We propose DRAGON, which stands for Decomposition and Reconstruction Agents Guided OptimizatioN, a novel framework that combines the strengths of metaheuristic design and LLM reasoning. Starting from an initial global solution, DRAGON autonomously identifies regions with high optimization potential and strategically decompose large-scale COPs into manageable subproblems. Each subproblem is then reformulated as a concise, localized optimization task and solved through targeted LLM prompting guided by accumulated experiences. Finally, the locally optimized solutions are systematically reintegrated into the original global context to yield a significantly improved overall outcome. By continuously interacting with the optimization environment and leveraging an adaptive experience memory, the agents iteratively learn from feedback, effectively coupling symbolic reasoning with heuristic search. Empirical results show that, unlike existing LLM-based solvers limited to small-scale instances, DRAGON consistently produces feasible solutions on TSPLIB, CVRPLIB, and Weibull-5k bin packing benchmarks, and achieves near-optimal results (0.16

Satellite-to-radar translation is a critical yet underexplored task in modern weather forecasting. While satellites provide near-global coverage and high-frequency updates, their multichannel radiance data is difficult to interpret directly by forecasters or end users, and not readily compatible with existing radar-based forecasting models. By converting satellite observations into radar-equivalent representations, we can: 1) extend radar-like availability to regions without ground-based radar coverage; 2) provide a more compact and operationally meaningful representation of storms; and 3) enable the reuse of a large ecosystem of radar-based nowcasting models without retraining. Despite its importance, methods tailored for this task remain limited, and the performance of general-purpose generative models has not been systematically benchmarked. To fill this gap, we propose SRDiff, a cross-modal, sequence-aware diffusion model specifically designed for satellite-to-radar translation. SRDiff employs a cross-modal conditional adapter (CMCA) to align heterogeneous satellite channels with radar reflectivity, and a Diffusion Transformer (DiT) backbone to capture sequential dependencies, producing stable and coherent radar predictions. Extensive experiments on the SEVIR and Sat2Rdr datasets show that SRDiff outperforms existing deterministic and diffusion-based methods on the satellite-to-radar translation task across all key metrics. As a downstream application, we integrate SRDiff with off-the-shelf radar-based nowcasting models, demonstrating three key advantages: 1) operational efficiency, since no new training is required; 2) strong forecasting accuracy, often surpassing models trained from scratch; and 3) practical validation that effective satellite-to-radar translation directly improves downstream nowcasting. The code is available at https://github.com/42xingxing/SRDiff

Audio-visual segmentation aims to separate sounding objects from videos by predicting pixel-level masks based on audio signals. Existing methods primarily concentrate on closed-set scenarios and direct audio-visual alignment and fusion, which limits their capability to generalize to new, unseen situations. In this paper, we propose OpenAVS, a novel training-free language-based approach that, for the first time, effectively aligns audio and visual modalities using text as a proxy for open-vocabulary Audio-Visual Segmentation (AVS). Equipped with multimedia foundation models, OpenAVS directly infers masks through 1) audio-to-text prompt generation, 2) LLM-guided prompt translation, and 3) text-to-visual sounding object segmentation. The objective of OpenAVS is to establish a simple yet flexible architecture that relies on the most appropriate foundation models by fully leveraging their capabilities to enable more effective knowledge transfer to the downstream AVS task. Moreover, we present a model-agnostic framework OpenAVS-ST that enables the integration of OpenAVS with any advanced supervised AVS model via pseudo-label based self-training. This approach enhances performance by effectively utilizing large-scale unlabeled data when available. Comprehensive experiments on three benchmark datasets demonstrate the superior performance of OpenAVS. It surpasses existing unsupervised, zero-shot, and few-shot AVS methods by a significant margin, achieving absolute performance gains of approximately 9.4

Traffic prediction is a crucial component of data management systems, leveraging historical data to learn spatio-temporal dynamics for forecasting future traffic and enabling efficient decision-making and resource allocation. Despite efforts to develop increasingly complex architectures, existing traffic prediction models often struggle to generalize across diverse datasets and contexts, limiting their adaptability in real-world applications. In contrast to existing traffic prediction models, large language models (LLMs) progress mainly through parameter expansion and extensive pre-training while maintaining their fundamental structures. In this paper, we propose ST-LLM+, the graph enhanced spatio-temporal large language models for traffic prediction. Through incorporating a proximity-based adjacency matrix derived from the traffic network into the calibrated LLMs, ST-LLM+ captures complex spatio-temporal dependencies within the traffic network. The Partially Frozen Graph Attention (PFGA) module is designed to retain global dependencies learned during LLMs pre-training while modeling localized dependencies specific to the traffic domain. To reduce computational overhead, ST-LLM+ adopts the LoRA-augmented training strategy, allowing attention layers to be fine-tuned with fewer learnable parameters. Comprehensive experiments on real-world traffic datasets demonstrate that ST-LLM+ outperforms state-of-the-art models. In particular, ST-LLM+ also exhibits robust performance in both few-shot and zero-shot prediction scenarios. Additionally, our case study demonstrates that ST-LLM+ captures global and localized dependencies between stations, verifying its effectiveness for traffic prediction tasks.

Spatio-temporal data mining plays a pivotal role in informed decision making across diverse domains. However, existing models are often restricted to narrow tasks, lacking the capacity for multi-task inference and complex long-form reasoning that require generation of in-depth, explanatory outputs. These limitations restrict their applicability to real-world, multi-faceted decision scenarios. In this work, we introduce STReason, a novel framework that integrates the reasoning strengths of large language models (LLMs) with the analytical capabilities of spatio-temporal models for multi-task inference and execution. Without requiring task-specific finetuning, STReason leverages in-context learning to decompose complex natural language queries into modular, interpretable programs, which are then systematically executed to generate both solutions and detailed rationales. To facilitate rigorous evaluation, we construct a new benchmark dataset and propose a unified evaluation framework with metrics specifically designed for long-form spatio-temporal reasoning. Experimental results show that STReason significantly outperforms advanced LLM baselines across all metrics, particularly excelling in complex, reasoning-intensive spatio-temporal scenarios. Human evaluations further validate STReason's credibility and practical utility, demonstrating its potential to reduce expert workload and broaden the applicability to real-world spatio-temporal tasks. We believe STReason provides a promising direction for developing more capable and generalizable spatio-temporal reasoning systems.

Precipitation nowcasting predicts future radar sequences based on current observations, which is a highly challenging task driven by the inherent complexity of the Earth system. Accurate nowcasting is of utmost importance for addressing various societal needs, including disaster management, agriculture, transportation, and energy optimization. As a complementary to existing non-autoregressive nowcasting approaches, we investigate the impact of prediction horizons on nowcasting models and propose SimCast, a novel training pipeline featuring a short-to-long term knowledge distillation technique coupled with a weighted MSE loss to prioritize heavy rainfall regions. Improved nowcasting predictions can be obtained without introducing additional overhead during inference. As SimCast generates deterministic predictions, we further integrate it into a diffusion-based framework named CasCast, leveraging the strengths from probabilistic models to overcome limitations such as blurriness and distribution shift in deterministic outputs. Extensive experimental results on three benchmark datasets validate the effectiveness of the proposed framework, achieving mean CSI scores of 0.452 on SEVIR, 0.474 on HKO-7, and 0.361 on MeteoNet, which outperforms existing approaches by a significant margin.

Uncertainty-aware weather nowcasting is critical for risk-sensitive decision-making in aviation, hazard preparedness, and public safety. This study improves short-term rainfall forecasting by integrating deep evidential learning into spatiotemporal modeling, enabling uncertainty-aware predictions for rainfall intensity regression and occurrence classification. Our framework treats radar observations as spatiotemporal evidence to parameterize Dirichlet prior for classification and Normal-Inverse Gamma prior for regression, building on data likelihood. With minimal added complexity, we extend a state-of-the-art model via evidential output layers. Results reveal that evidential regression struggles with rare high-intensity events due to data sparsity, while classification uncertainties lack intensity gradation. We establish operational thresholds for trustworthy classification forecasts and demonstrate a strong uncertaintyerror correlation in regression. Tropical rainfall case studies validate the model's reliability, with high-uncertainty regions spatially coinciding with higher model errors.

Aircraft trajectory prediction aims to estimate the future movements of aircraft in a scene, which is a crucial step for intelligent air traffic management such as capacity estimation and conflict detection. Current approaches primarily rely on inputting absolute locations, which improves the prediction accuracy but limits the model’s generalization ability to unseen environments. To bridge the gap, we propose to alternatively learn aircraft’s intentions from a repository of historical trajectories. Based on the observation that aircraft traveling through the same airspace may exhibit comparable behaviors, we utilize a location-adaptive threshold to identify nearby neighbors for a given query aircraft within the repository. The retrieved candidates are next filtered based on contextual information, such as landing time and landing direction, to eliminate less relevant components. The resulting set of nearby candidates are referred to as the local history, which emphasizes the modeling of aircraft’s local behavior. Moreover, an attention-based local history encoder is presented to aggregate information from all nearby candidates to generate a latent feature for capturing the aircraft’s intention. This latent feature is robust to normalized input trajectories, relative to the current location of the target aircraft, thus improving the model’s generalization capability to unseen areas. Our proposed intention modeling method is model-agnostic, which can be leveraged as an additional condition by any trajectory prediction model for improved robustness and accuracy. For evaluation, we integrate the intention modeling component into our previous diffusion-based aircraft trajectory prediction framework. We conduct experiments on two real-world aircraft trajectory datasets in both towered and non-towered terminal airspace. The experimental results show that our method captures various maneuvering patterns effectively, outperforming existing methods by a large margin in terms of both ADE and FDE.

Domain adaptation enables the reduction of distribution differences across domains, allowing for effective knowledge transfer from one domain to a different domain. In recent years, partial domain adaptation (PDA) has attracted growing interest due to its focus on a more realistic scenario, where the target label space is a subset of the source label space. As the source and target domains do not possess the same label space in the PDA setting, it is challenging but crucial to mitigate the domain gap without incurring negative transfer. In this paper, we propose a Reinforced Reweighting united with Self-supervised Adaptation (R2SA) method to address the challenges in PDA by leveraging the merits of deep reinforcement learning (DRL) and self-supervised learning (SSL) simultaneously in a cooperative way. Reinforced reweighting aims to learn a source reweighting policy automatically based on information provided by the PDA model, while self-supervised adaptation aims to boost the adaptability of the PDA model through an additional self-supervised objective on the target domain. Extensive experiments on several cross-domain benchmarks demonstrate that our method achieves state-of-the-art results, with larger performance gains on more challenging tasks.

Spatio-temporal graph neural networks have proven efficacy in capturing complex dependencies for urban computing tasks such as forecasting and kriging. Yet, their performance is constrained by the reliance on extensive data for training on a specific task, thereby limiting their adaptability to new urban domains with varied task demands. Although transfer learning has been proposed to remedy this problem by leveraging knowledge across domains, the cross-task generalization still remains under-explored in spatio-temporal graph transfer learning due to the lack of a unified framework. To bridge the gap, we propose Spatio-Temporal Graph Prompting (STGP), a prompt-based framework capable of adapting to multi-diverse tasks in a data-scarce domain. Specifically, we first unify different tasks into a single template and introduce a task-agnostic network architecture that aligns with this template. This approach enables capturing dependencies shared across tasks. Furthermore, we employ learnable prompts to achieve domain and task transfer in a two-stage prompting pipeline, facilitating the prompts to effectively capture domain knowledge and task-specific properties. Our extensive experiments demonstrate that STGP outperforms state-of-the-art baselines in three tasks-forecasting, kriging, and extrapolation-achieving an improvement of up to 10.7%.

Air quality prediction and modelling plays a pivotal role in public health and environment management, for individuals and authorities to make informed decisions. Although traditional data-driven models have shown promise in this domain, their long-term prediction accuracy can be limited, especially in scenarios with sparse or incomplete data and they often rely on black-box deep learning structures that lack solid physical foundation leading to reduced transparency and interpretability in predictions. To address these limitations, this paper presents a novel approach named Physics guided Neural Network for Air Quality Prediction (AirPhyNet). Specifically, we leverage two well-established physics principles of air particle movement (diffusion and advection) by representing them as differential equation networks. Then, we utilize a graph structure to integrate physics knowledge into a neural network architecture and exploit latent representations to capture spatio-temporal relationships within the air quality data. Experiments on two real-world benchmark datasets demonstrate that AirPhyNet outperforms state-of-the-art models for different testing scenarios including different lead time (24h, 48h, 72h), sparse data and sudden change prediction, achieving reduction in prediction errors up to 10%. Moreover, a case study further validates that our model captures underlying physical processes of particle movement and generates accurate predictions with real physical meaning.

Smart mobile apps have become an integral part of our daily lives, offering intelligent services anytime, anywhere. Various apps tap into mobility traces from different aspects of our daily activities, each managing these traces independently. Linking these isolated traces from different apps offers significant potential for gaining a deeper understanding of user behavior. Various solutions have been proposed to connect mobility traces of the same user across different apps, showing effectiveness in certain scenarios. However, we believe they have not effectively tackled the following four key challenges. First, a solution must accommodate a growing number of users. Second, mobility traces of newly joined users must be successfully linked. Third, auxiliary information, such as place of interests, might not always be accessible. Fourth, activity traces could be extremely sparse. In this paper, we introduce TLink, a deep learning framework designed to link mobility traces of the same user across different datasets, to address the above challenges. TLink employs a temporal convolutional network as its encoder and assesses the probability that a pair of traces, sourced from different datasets, originates from the same user. To train the model effectively, we adapted a multi-step training procedure that strategically modulates the complexity of the training objective. Furthermore, we introduce a new metric learning objective specifically crafted for the trace linking task. Our evaluation demonstrates a notable performance of TLink under challenging scenarios, evidencing a performance improvement of more than tenfold compared to an applicable baseline.

Aircraft trajectory prediction aims to estimate the future movements of the aircraft, which is a crucial step for air traffic management such as capacity estimation and conflict detection. In this paper, we present a context-aware trajectory prediction method, which generates the future movements based on both the aircraft's past status and the contextual information such as the pilot and controller intent and the environmental conditions. The proposed framework consists of 1) a Trajectory Encoder that captures the history behaviors and the social interactions of the aircraft, 2) a Context Encoder that extracts latent features from contextual information, and 3) a Transformer-based Decoder that generates future trajectories based on a diffusion model. Specifically, we model the trajectory prediction as the reverse diffusion process where we first gradually add noise to the ground-truth trajectory and then train a neural network to learn the reverse of this diffusion process conditioned on the output of the trajectory encoder and the context encoder. We conduct experiments on real-world aircraft trajectories collected at Singapore Changi Airport in December 2019, which correspond to one-week ADS-B data before the start of the COVID-19 pandemic. The experimental results show that our proposed approach outperforms existing methods by a significant margin.

Source-free domain adaptive semantic segmentation has gained increasing attention recently. It eases the requirement of full access to the source domain by transferring knowledge only from a well-trained source model. However, reducing the uncertainty of the target pseudo labels becomes inevitably more challenging without the supervision of the labeled source data. In this work, we propose a novel asymmetric two-stream architecture that learns more robustly from noisy pseudo labels. Our approach simultaneously conducts dual-head pseudo label denoising and cross-modal consistency regularization. Towards the former, we introduce a multimodal auxiliary network during training (and discard it during inference), which effectively enhances the pseudo labels' correctness by leveraging the guidance from the depth information. Towards the latter, we enforce a new cross-modal pixel-wise consistency between the predictions of the two streams, encouraging our model to behave smoothly for both modality variance and image perturbations. It serves as an effective regularization to further reduce the impact of the inaccurate pseudo labels in source-free unsupervised domain adaptation. Experiments on GTA5 → Cityscapes and SYNTHIA → Cityscapes benchmarks demonstrate the superiority of our proposed method, obtaining the new state-of-the-art mIoU of 57.7% and 57.5%, respectively.

Policy adaptation remains one of the key challenges for reinforcement learning (RL). Thus, RL agents often fail to generalize to unseen scenarios. In this paper, we propose to improve the generalization of RL algorithms through multi-task self-supervised adaptation (MSSA). The proposed method is a general paradigm that can be implemented on top of any RL algorithm. It better extracts high-level feature representations from augmented observations through incorporating multiple self-supervised learning tasks with complementary objectives. The selected self-supervision tasks include rotation prediction, inverse dynamics prediction and contrastive learning. It then performs control actions based on the extracted features. The proposed MSSA method consistently outperforms all the baseline methods on diverse complex tasks in the DeepMind Control suite benchmark and sets new state-of-the-art results without incurring longer inference time. It is demonstrated that MSSA has superior generalization capability and is robust to environmental changes.

Tube internal erosion, which corresponds to its wall thinning process, is one of the major safety concerns for tubes. Many sensing technologies have been developed to detect a tube wall thinning process. Among them, fiber Bragg grating (FBG) sensors are the most popular ones due to their precise measurement properties. Most of the current works focus on how to design different types of FBG sensors according to certain physical laws and only test their sensors in controlled laboratory conditions. However, in practice, an industrial system usually suffers from harsh and dynamic environmental conditions, and FBG signals are affected by many unpredictable factors. Consequently, the FBG signals have more fluctuations and are polluted by noises. Hence, the signals no longer directly follow the assumed physical laws and their proposed thinning detection mechanisms no longer work. Targeting at this, this article develops a data-driven model for FBG signal feature extraction and tube wall thickness monitoring using data analytic techniques. In particular, we develop a spatiotemporal model to describe dynamic FBG signals and extract features related to thickness. By taking physical law as guideline, we trace the relationship between the extracted features and the tube wall thickness, based on which we construct an online statistical monitoring scheme for tube wall thinning process. We use both laboratory test and field trial experiment to demonstrate the efficacy and efficiency of the proposed scheme. Note to Practitioners—This article is motivated by the real industrial needs of inner erosion detection of tubes in harsh environment. Most of the current research works focus on designing various sensing apparatuses based on fiber Bragg grating (FBG) sensors for nondestructive erosion detection. These apparatuses prove to be able to collect signals reflecting tube wall thickness in static and controllable laboratory environment qualitatively. However, in reality, the industrial environment, which is impacted by many changing factors, is dynamic and uncontrollable. Consequently, the signals collected by these FBG apparatuses would have larger variations that mask the signals related to thickness. Furthermore, current methods have neither mentioned how to process their collected data to capture the unnoticeably slow but accumulative erosion information efficiently nor constructed online monitoring algorithms to detect the tube wall thinning process based on the collected signals quantitatively. Built upon their apparatuses but targeting at their unsolved challenges, we propose a novel data-driven approach for FBG signal analysis that can remove the environmental influence and extract features only related to tube wall thickness, and using the extracted features, we construct a statistical process control scheme to monitor tube wall thickness and detect erosion in real time efficiently.