Urban scene reconstruction is crucial for real-world autonomous driving simulators. Although existing methods have achieved photorealistic reconstruction, they mostly focus on pinhole cameras and neglect fisheye cameras. In fact, how to effectively simulate fisheye cameras in driving scene remains an unsolved problem. In this work, we propose UniGaussian, a novel approach that learns a unified 3D Gaussian representation from multiple camera models for urban scene reconstruction in autonomous driving. Our contributions are two-fold. First, we propose a new differentiable rendering method that distorts 3D Gaussians using a series of affine transformations tailored to fisheye camera models. This addresses the compatibility issue of 3D Gaussian splatting with fisheye cameras, which is hindered by light ray distortion caused by lenses or mirrors. Besides, our method maintains real-time rendering while ensuring differentiability. Second, built on the differentiable rendering method, we design a new framework that learns a unified Gaussian representation from multiple camera models. By applying affine transformations to adapt different camera models and regularizing the shared Gaussians with supervision from different modalities, our framework learns a unified 3D Gaussian representation with input data from multiple sources and achieves holistic driving scene understanding. As a result, our approach models multiple sensors (pinhole and fisheye cameras) and modalities (depth, semantic, normal and LiDAR point clouds). Our experiments show that our method achieves superior rendering quality and fast rendering speed for driving scene simulation.

We introduce MMSciCode, a comprehensive expert-level, multilingual multi-discipline benchmark for evaluating foundation models in scientific code generation.It includes 624 expert-annotated research coding problems spanning six core scientific disciplines.Compared to prior benchmarks, MMSciCode features three key advancements.First, it challenges models to integrate domain-specific knowledge with algorithmic reasoning to implement core functions from research papers.Second, each problem is meticulously annotated by domain experts through a rigorous papergrounded process, with strict quality controls implemented to ensure dataset integrity and authenticity.Finally, each problem is equipped with comprehensive unit test suites and containerized environments, enabling reproducible and diagnostic evaluation of both functional correctness and domain validity.We conduct an extensive evaluation of 23 state-of-the-art foundation models and 2 coding agents on MMSciCode.We identify substantial performance gaps between models and human experts, providing actionable insights for advancing expert-level scientific code generation.

The integration of aerial reconfigurable intelligent surfaces (ARISs) into IoT networks offers transformative potential for enhancing secure and energy-efficient communication in the presence of blockages and eavesdropping threats. This paper proposes to integrate ARIS into Internet of Things (IoT) networks to simultaneously improve communication reliability, enforce information freshness, and defend against eavesdropping. We formulate a joint optimization problem to minimize the average total transmit energy of IoT devices through the coordinated design of unmanned aerial vehicle (UAV) trajectory, transmit power allocation, ARIS phase shifts, and device scheduling, subject to rigorous constraints on age of information (AoI), UAV energy budget, and secrecy rate guarantees. The optimization problem is formulated as a dynamic programming problem. To address the complexity of long-term dynamic optimization, we employ Lyapunov optimization to decompose it into a per-slot deterministic optimization problem, which can be solved without requiring future state information. However, the per-slot problem is a mixed-integer non-convex optimization problem, making it inherently challenging to solve optimally. To address this, we propose an efficient algorithm that effectively balances the tradeoff between minimizing average total energy consumption and stabilizing average total queue backlogs. Simulation results demonstrate that our algorithm reduces average transmit energy by 46% compared to the round-robin comparison scheme while strictly adhering to information freshness and UAV energy constraints.

Recently, the electroencephalogram (EEG) has been widely adopted as a quantitative indicator for monitoring emotional modulation during virtual reality (VR) experiences. Although VR emotion-induction materials are continuously being developed, few methods have been proposed for constructing VR scenarios through psychophysiological calibration. To achieve precise emotional regulation, we propose an integrated psychophysiological oriented EEG-based VR scenario modeling approach for emotion induction. This methodology constructs 3D VR scenarios by extending 2D elements through core processes: (1) deconstruction of emotion-annotated 2D images to extract visual-audio emotional patterns, (2) development of immersive environments with dynamic camera trajectories, and (3) integration of audio stimuli and real-world physical configurations to form multi-sensory emotional stimuli. During emotioninduction experiments, EEG signals reflecting emotional states were captured and analyzed for each scenario. We implemented multidimensional calibration of emotion-induction materials by: (a) calculating emotion indicators from EEG signals, (b) combining these with traditional assessments using the Self-Assessment Manikin (SAM) scale, and (c) calibrating scenarios for distinct emotion categories. This yields psychophysiologically calibrated VR scenarios for emotion induction, delivering standardised affective stimuli with synchronised EEG datasets to advance affective computing research.

In this work, we present the first study to explore inference-time scaling on table reasoning tasks.We develop and evaluate two post-training strategies to enable inferencetime scaling: distillation from frontier model reasoning traces and reinforcement learning with verifiable rewards (RLVR).For distillation, we introduce a large-scale dataset of reasoning traces generated by DeepSeek-R1, which we use to fine-tune LLMs into the Table-R1-SFT model.For RLVR, we propose taskspecific verifiable reward functions and apply the GRPO algorithm to obtain the Table-R1-Zero model.We evaluate our Table-R1series models across diverse table reasoning tasks, including short-form QA, fact verification, and free-form QA.Notably, the Table-R1-Zero model matches or exceeds the performance of GPT-4.1 and DeepSeek-R1, while using only a 7B-parameter LLM.It also demonstrates strong generalization to out-of-domain datasets.Extensive ablation and qualitative analyses reveal the benefits of instruction tuning, model architecture choices, and cross-task generalization, as well as emergence of essential table reasoning skills during RL training.Model huggingface.co/Table-R1Code github.com/Table-R1

Virtual reality (VR) mindfulness meditation has emerged as a prominent emotion regulation strategy in recent years. Current research often seeks to enhance meditation effectiveness through biofeedback and overlooks the trajectory of emotional changes and the changing needs during regulation. In this study, we propose an electroencephalography (EEG)–based positive feedback mechanism for VR mindfulness meditation aimed at optimizing the effects of emotion regulation. This mechanism consists of three modules: 1) EEG-based emotional state computation; 2) process-based relaxation assessment; and 3) adaptive positive decision feedback. Collectively, these components form a computation-assessment-feedback closed-loop system that objectively quantifies emotions while enabling real-time decision adjustments based on emotional trends, thereby enhancing user engagement and emotion regulation efficacy through personalized feedback. The contribution of the proposed feedback mechanism was evaluated through a randomized controlled trial (N= 36). The results indicated that both physiological measures and self-reported relaxation significantly increased when compared to interventions without feedback. These findings validate that the EEG-based positive feedback mechanism effectively enhances emotion regulation while providing additional insights into improving both the engagement and effectiveness within digital mental health interventions.

In future B5G/6G wideband communication systems, non-linear signal distortion caused by the impairment of transmit power amplifier (PA) can severely degrade the communication performance. The performance impact is especially significant when uplink users share the wireless medium using Non-orthogonal Multiple Access (NOMA) scheme. This is because the successive interference cancellation (SIC) information decoding technique of NOMA cannot eliminate the interference caused by the PA non-linear distortion, such that the decoding of each user will suffer from the aggregate distortion noise of all the uplink users. In this paper, we study the impact of PA non-linear distortion on the performance of uplink NOMA. In particular, we first establish a new PA distortion signal model based on real-world measurements, where the distortion noise power is a polynomial function of PA transmit power, instead of a simplified linear function in most existing studies. Under the proposed signal model, we then accurately characterize the capacity region of a two-user uplink NOMA by optimizing the user transmit power. We show that the polynomial distortion noise power significantly shrinks the achievable capacity region of NOMA. This indicates that existing studies may have overestimated the communication performance of NOMA in practical wideband systems. Besides, the non-linear noise power also leads to a rather different optimal power allocation strategy to attain maximum throughput. Simulation results show that, for a PA following the polynomial distortion noise power model, the proposed optimal power allocation method achieves on average 12.2% higher sum throughput than that obtained from ideal PA model. Overall, our results demonstrate the importance of accurate PA distortion modeling to the performance of NOMA and provide an efficient power allocation method to attain the optimal performance.

In emergency scenarios, unmanned aerial vehicles (UAVs) can be deployed to assist localization and communication services for ground terminals. In this paper, we propose a new integrated air-ground networking paradigm that uses dual-functional UAVs to assist the ground networks for improving both communication and localization performance. We investigate the optimization problem of deploying the minimal number of UAVs to satisfy the communication and localization requirements of ground users. The problem has several technical difficulties including the cardinality minimization, the non-convexity of localization performance metric regarding UAV location, and the association between user and communication terminal. To tackle the difficulties, we adopt$D$-optimality as the localization performance metric, and derive the geometric characteristics of the feasible UAV hovering regions in 2D and 3D based on accurate approximation values. We solve the simplified 2D projection deployment problem by transforming the problem into a minimum hitting set problem, and propose a low-complexity algorithm to solve it. Through numerical simulations, we compare our proposed algorithm with benchmark methods. The number of UAVs required by the proposed algorithm is close to the optimal solution, while other benchmark methods require much more UAVs to accomplish the same task.

The recent development of unmanned aerial vehicle (UAV) and mobile edge computing (MEC) technologies provides flexible and resilient computation services to mobile users out of the terrestrial computing service coverage. In this paper, we consider a UAV-enabled MEC platform that serves multiple mobile ground users with random movements and task arrivals. We aim to minimize the average weighted energy consumption of all users subject to the average UAV energy consumption and data queue stability constraints. We formulate the problem as a multi-stage stochastic optimization, and adopt Lyapunov optimization to convert it into per-slot deterministic problems with fewer optimizing variables. We design two reduced-complexity methods that solve the resource allocation and the UAV movement either in two sequential steps or jointly in one step. Both methods can guarantee to satisfy the average UAV energy and queue stability constraints, meanwhile achieving a tradeoff between the user energy consumption and the length of queue backlog. Simulation results show that the two methods significantly outperform the other benchmark methods including a learning-based method in reducing the energy consumption of ground users. In between, the proposed joint optimization method achieves better performance than the two-stage method at the cost of higher computational complexity.

Unmanned aerial vehicle (UAV)-enabled mobile edge computing (MEC) has recently emerged as a cost-effective solution to provide computation service to distributed devices in the absence of terrestrial infrastructure. In this paper, we consider a UAV-enabled MEC system serving multiple energy harvesting (EH) devices, where the energy and task data arrive at the users stochastically. Without any future knowledge of task data and energy arrivals, our objective is to design an online algorithm to jointly optimize the UAV energy and task processing rate, meanwhile satisfying the long-term data queue stability. We formulate the problem as a multi-stage stochastic programming and propose an online algorithm, named PLOT, based on perturbed Lyapunov optimization technique. In particular, PLOT resolves the coupling effect of sequential control actions, and converts the stochastic problem into per-slot deterministic optimization problem. For each per-slot problem, we design a low-complexity algorithm to solve it. We show that the PLOT algorithm can derive a feasible solution to the original problem and achieve an$[O(1/V),O(V)]$trade-off between the system cost and the data queue length. Simulation results justify our analysis and demonstrate that the PLOT algorithm achieves better performance in terms of system utility and maintains queue stability that is not achieved by other benchmark methods.

In this paper, we study an unmanned aerial vehicle (UAV)-enabled mobile edge computing (MEC) system with multiple energy harvesting (EH) devices. Considering the stochastic energy and data arrivals in sequential time slots, we formulate the UAV propulsion energy minimization problem with long-term data queue stability and battery causality constraints as a multi-stage stochastic optimization programming. To facilitate online control without any prior knowledge of future information, we adopt the perturbed Lyapunov optimization method that decouples the control decisions made in sequential time slots and determines the real-time control decisions by solving a deterministic problem in each time slot. For the per-slot deterministic problem, we decouple it into three sub-problems: the optimal energy harvesting, the computation resource allocation and the UAV trajectory control, and propose a reduced-complexity method to solve them sepa-rately. Simulation results demonstrate that the proposed algorithm guarantees the data queue stability that is not achievable by the benchmark method when the two methods consume identical UAV propulsion energy.