Recent LLM-based data agents aim to automate data science tasks ranging from data analysis to deep learning. However, the open-ended nature of real-world data science problems, which often span multiple taxonomies and lack standard answers, poses a significant challenge for evaluation. To address this, we introduce DSAEval, a benchmark comprising 641 real-world data science problems grounded in 285 diverse datasets, covering both structured and unstructured data (e.g., vision and text). DSAEval incorporates three distinctive features: (1) Multimodal Environment Perception, which enables agents to interpret observations from multiple modalities including text and vision; (2) Multi-Query Interactions, which mirror the iterative and cumulative nature of real-world data science projects; and (3) Multi-Dimensional Evaluation, which provides a holistic assessment across reasoning, code, and results. We systematically evaluate 11 advanced agentic LLMs using DSAEval. Our results show that Claude-Sonnet-4.5 achieves the strongest overall performance, GPT-5.2 is the most efficient, and MiMo-V2-Flash is the most cost-effective. We further demonstrate that multimodal perception consistently improves performance on vision-related tasks, with gains ranging from 2.04% to 11.30%. Overall, while current data science agents perform well on structured data and routine data anlysis workflows, substantial challenges remain in unstructured domains. Finally, we offer critical insights and outline future research directions to advance the development of data science agents.
Kai-Ho Cheng, Chih-Hsun Chang, Yi-Chung Yang
et al.
The relationship between wind and waves has been extensively studied over time. However, understanding the local wind and wave relationship remains crucial for advancing renewable energy development and optimizing ocean management strategies. This study used wind and wave data collected by the ten weather buoys in the waters surrounding Taiwan to analyze regional sea states. The relationship between wind speed and significant wave height (SWH) was examined using regression analysis. Additionally, machine learning techniques were employed to assess the relative importance of features contributing to SWH growth. The regression analysis revealed that SWH in the waters surrounding Taiwan was not fully developed, with notable discrepancies observed between the waters east and west of Taiwan. According to the power law formula describing the relationship between wind speed and SWH, the eastern waters exhibited a larger prefactor coupled with a smaller scaling exponent, while the western waters manifested a converse parametric configuration. Through an evaluation of four machine learning algorithms, it was determined that wind speed is the most influential factor driving these regional differences, especially in the waters west of Taiwan. Beyond wind speed, air pressure or temperature emerged as the secondary feature factor governing wind–wave interactions in the waters east of Taiwan.
The design of ships has changed dramatically since the 1970s. We have shifted from manual drafting to digital tools and computers, mostly because computer technology has greatly improved. Nowadays, with the growth of smart digitalization in Industry 4.0, using modern digital software and tools makes ship design more efficient and enhances its quality throughout a ship's entire lifespan. However, this shift has also made operations more complex and requires users of the software to have more specialized training. Today, technologies like automated optimization, simulation-based design, managing the entire product lifecycle, digital twins, and artificial intelligence are commonly used in the shipping industry. These technologies are applied during both the design and construction phases, as well as in preparing and inspecting ships. This paper reviews major advances in these areas and discusses how the industry can address current and future challenges.
ZONG Haoxiang, ZHANG Chen, BAO Yanhong, WU Feng, CAI Xu
Aimed at the small-signal synchronization instability of grid-following (GFL) and grid-forming (GFM) converter system, a synchronization perspective frequency-domain modeling and analysis method is proposed, which can intuitively reveal mechanism and accurately judge multi-machine stability. Specifically, a node admittance matrix considering GFL, GFM converters, and the transmission network is established. Then, the frequency domain modal analysis (FMA) method is adopted to evaluate system instability characteristics. Afterwards, synchronization forward and feedback paths are partitioned at the oscillation source to formulate a synchronization perspective stability model incorporating dynamics of each converter and transmission network. Finally, the proposed method is validated by using a typical two-machine GFL-GFM system. With such method, the stability judgment failure caused by the feedback path aggregation is addressed, and the interaction mechanism between GFL and GFM synchronization dynamics as well as their parameter influences are revealed.
Engineering (General). Civil engineering (General), Chemical engineering
Sergii Sagin, Oleksandr Haichenia, Sergey Karianskyi
et al.
This paper aims to consider the issue of increasing the environmental friendliness of shipping by using alternative fuels in marine diesel engines. It has been determined that marine diesel engines are not only the main heat engines used on ships of sea and inland waterway transport, but are also sources of emissions of toxic components with exhaust gases. The main compounds whose emissions are controlled and regulated by international organizations are sulfur oxides (SO<sub>X</sub>) and nitrogen oxides (NO<sub>X</sub>), as well as carbon dioxide (CO<sub>2</sub>). Reducing NO<sub>X</sub> and CO<sub>2</sub> emissions while simultaneously increasing the environmental friendliness of shipping is possible by using fuel mixtures in marine diesel engines that include biodiesel fuel. During the research carried out on Wartsila 6L32 marine diesel engines (Shanghai Wartsila Qiyao Diesel Co. Ltd., Shanghai, China), RMG500 and DMA10 petroleum fuels were used, as well as their mixtures with biodiesel fuel FAME. It was found that when using mixtures containing 10–30% of FAME biodiesel, NO<sub>X</sub> emissions are reduced by 11.20–27.10%; under the same conditions, CO<sub>2</sub> emissions are reduced by 5.31–19.47%. The use of alternative fuels in marine diesel engines (one of which is biodiesel and fuel mixtures containing it) is one of the ways to increase the level of environmental sustainability of seagoing vessels and promote ecological shipping. This is of particular relevance when operating vessels in special ecological areas of the World Ocean. The relatively low energy intensity of the method of creating and using such fuel mixtures contributes to the spread of its use on many means of maritime transport.
This entry provides an overview of Human-centered Geospatial Data Science, highlighting the gaps it aims to bridge, its significance, and its key topics and research. Geospatial Data Science, which derives geographic knowledge and insights from large volumes of geospatial big data using advanced Geospatial Artificial Intelligence (GeoAI), has been widely used to tackle a wide range of geographic problems. However, it often overlooks the subjective human experiences that fundamentally influence human-environment interactions, and few strategies have been developed to ensure that these technologies follow ethical guidelines and prioritize human values. Human-centered Geospatial Data Science advocates for two primary focuses. First, it advances our understanding of human-environment interactions by leveraging Geospatial Data Science to measure and analyze human subjective experiences at place including emotion, perception, cognition, and creativity. Second, it advocates for the development of responsible and ethical Geospatial Data Science methods that protect geoprivacy, enhance fairness and reduce bias, and improve the explainability and transparency of geospatial technologies. With these two missions, Human-centered Geospatial Data Sciences brings a fresh perspective to develop and utilize geospatial technologies that positively impact society and benefit human well-being and the humanities.
The Science Consultant Agent is a web-based Artificial Intelligence (AI) tool that helps practitioners select and implement the most effective modeling strategy for AI-based solutions. It operates through four core components: Questionnaire, Smart Fill, Research-Guided Recommendation, and Prototype Builder. By combining structured questionnaires, literature-backed solution recommendations, and prototype generation, the Science Consultant Agent accelerates development for everyone from Product Managers and Software Developers to Researchers. The full pipeline is illustrated in Figure 1.
A fundamental challenge in the cognitive sciences is discovering the dynamics that govern behaviour. Take the example of spoken language, which is characterised by a highly variable and complex set of physical movements that map onto the small set of cognitive units that comprise language. What are the fundamental dynamical principles behind the movements that structure speech production? In this study, we discover models in the form of symbolic equations that govern articulatory gestures during speech. A sparse symbolic regression algorithm is used to discover models from kinematic data on the tongue and lips. We explore these candidate models using analytical techniques and numerical simulations, and find that a second-order linear model achieves high levels of accuracy, but a nonlinear force is required to properly model articulatory dynamics in approximately one third of cases. This supports the proposal that an autonomous, nonlinear, second-order differential equation is a viable dynamical law for articulatory gestures in speech. We conclude by identifying future opportunities and obstacles in data-driven model discovery and outline prospects for discovering the dynamical principles that govern language, brain and behaviour.
As oil and gas exploration gradually advances into deep waters, the combined effects of various types of gas kick and the accurate calculation of the gas-kick volume have gained increasing attention. This study focused on gas kicks from permeable gas-bearing formations, considering the mass transfer of gas in the filtration region of the drilling fluids and revealed the mechanisms of seepage-driven and diffusion-driven gas kicks. Based on seepage mechanics and diffusion theory, a comprehensive model for calculating gas-kick volume was established, considering the synergistic effect of gas-concentration-diffusion and negative-differential-pressure, as well as mass transfer in both the filtrate zone and the filter-cake zone. The new model showed high calculation accuracy. The sensitivity analysis showed that both the seepage-driven and diffusion-driven gas-kick volumes in the wellbore increased with increasing formation porosity and open-hole length, while the thickness of the filter cake had a strong inhibitory effect on both. Additionally, a “seepage–diffusion ratio” was introduced to reveal the gas-kick evolution pattern under a seepage–diffusion mechanism. Under specific case conditions, when the seepage–diffusion ratio was less than approximately 1%, diffusion-driven gas kick contributed more than seepage-driven gas kick; when the seepage–diffusion ratio exceeded 1%, seepage-driven gas kick contributed more than diffusion-driven gas kick. The research can provide crucial parameters for wellbore multiphase flow calculation and wellbore pressure prediction.
Francisco Asensio-Montesinos, Rosa Molina, Giorgio Anfuso
et al.
Coasts are the most densely populated regions in the world and are vulnerable to different natural and human factors, e.g., sea-level rise, coastal accretion and erosion processes, the intensification of sea storms and hurricanes, the presence of marine litter, chronic pollution and beach oil spill accidents, etc. Although coastal zones have been affected by local anthropic activities for decades, their impacts on coastal ecosystems is often unclear. Several papers are presented in this Special Issue detailing the interactions between natural processes and human impacts in coastal ecosystems all around the world. A better understanding of such natural and human impacts is therefore of great relevance to confidently predict their negative effects on coastal areas and thus promote different conservation strategies. The implementation of adequate management measures will help coastal communities adapt to future scenarios in the short and long term and prevent damage due to different pollution types, e.g., beach oil spill accidents, through the establishment of Environmental Sensitivity Maps.
Advanced analytics science methods have enabled combining the power of artificial and human intelligence, creating \textit{centaurs} that allow superior decision-making. Centaurs are hybrid human-algorithm models that combine both formal analytics and human intuition in a symbiotic manner within their learning and reasoning process. We argue that the future of AI development and use in many domains needs to focus more on centaurs as opposed to other AI approaches. This paradigm shift towards centaur-based AI methods raises some fundamental questions: How are centaurs different from other human-in-the-loop methods? What are the most effective methods for creating centaurs? When should centaurs be used, and when should the lead be given to pure AI models? Doesn't the incorporation of human intuition -- which at times can be misleading -- in centaurs' decision-making process degrade its performance compared to pure AI methods? This work aims to address these fundamental questions, focusing on recent advancements in generative AI, and especially in Large Language Models (LLMs), as a main case study to illustrate centaurs' critical essentiality to future AI endeavors.
AbstractIn multi‐agent search planning for a randomly moving and camouflaging target, we examine heterogeneous searchers that differ in terms of their endurance level, travel speed, and detection ability. This leads to a convex mixed‐integer nonlinear program, which we reformulate using three linearization techniques. We develop preprocessing steps, outer approximations via lazy constraints, and bundle‐based cutting plane methods to address large‐scale instances. Further specializations emerge when the target moves according to a Markov chain. We carry out an extensive numerical study to show the computational efficiency of our methods and to derive insights regarding which approach should be favored for which type of problem instance.
Similar to other industries, the maritime industry is also facing increasing restrictions on ships regarding pollution control. The research presented in this paper is aimed at studying the pros and cons of alternative fuels followed by a detailed analysis on hydrogen fuel cells (PEMFC) for a particular ship operating in Lisbon, Portugal. Dynamic forces acting on the ship have been studied for a year. Assessing various scenarios based on these results aids ship operators in making informed decisions regarding the future course of action for their existing vessels. These different cases are first: business as usual (diesel engine), second: replacing the diesel engine with a hydrogen hybrid system and, third: replacement of the ship with a new hydrogen hybrid ship. The study is based on the simulation of numerical equations and CFD simulation results. As the result, the second scenario is best suited in both aspects; namely, environmental and economic.
Compared to traditional vehicles, the underwater bionic manta ray vehicle (UBMRV) is highly maneuverable, has strong concealment, and is an emerging research field in underwater vehicles. Based on the completion of the single-body research, it is crucial to research the swarm of UBMRVs for the implementation of complex tasks, such as large-scale underwater detection. The relative positioning capability of the UBMRV is the key to realizing a swarm, especially when underwater acoustic communications are delayed. To solve the real-time relative positioning problem between individuals in the UBMRV swarm, this study proposes a relative positioning method based on the combination of the improved object detection algorithm and binocular distance measurement. To increase the precision of underwater object detection in small samples, this paper improves the original YOLOx algorithm. It increases the network’s interest in the object area by adding an attention mechanism module to the network model, thereby improving its detection accuracy. Further, the output of the object detection result is used as the input of the binocular distance measurement module. We use the ORB algorithm to extract and match features in the object-bounding box and obtain the disparity of the features. The relative distance and bearing information of the target are output and shown on the image. We conducted pool experiments to verify the proposed algorithm on the UBMRV platform, proved the method’s feasibility, and analyzed the results.
The influence of the maritime environment such as water currents, water depth, and traffic separation rules should be considered when conducting ship path planning. Additionally, the maneuverability constraints of the ship play a crucial role in navigation. Addressing the limitations of the traditional A-star algorithm in ship path planning, this paper proposes an improved A-star algorithm. Specifically, this paper examines the factors influencing ship navigation safety, and develops a risk model that takes into account water currents, water depth, and obstacles. The goal is to mitigate the total risk of ship collisions and grounding. Secondly, a traffic model is designed to ensure that the planned path adheres to the traffic separation rules and reduces the risk of collision with incoming ships. Then, a turning model and smoothing method are designed to make the generated path easy to track and control for the ship. To validate the effectiveness of the proposed A-star ship path-planning algorithm, three cases are studied in simulations and representative operational scenarios. The results of the cases demonstrate that the proposed A-star ship path-planning algorithm can better control the distance to obstacles, effectively avoid shallow water areas, and comply with traffic separation rules. The safety level of the path is effectively improved.
AbstractInternal waves (IW) are crucial contributors to the transport of sediment, heat, and nutrients in coastal areas. While IW have been extensively studied using point measurements, their spatial variability is less well understood. In this paper we present a unique set of high‐resolution infrared imagery collected from a helicopter, hovering over very energetic shoaling and breaking IW. We compute surface velocities by tracking the evolution of thermal structures at the ocean surface and find horizontal velocity gradients with magnitudes that are more than 100 times the Coriolis frequency. Under the assumption of no vertical shear we determine vertical velocities from the obtained horizontal divergence estimates and identify areas of the wave undergoing breaking. The spatial variability of the internal wave occurs on scales from a few to a few hundred meters. These results highlight the need to collect spatio‐temporal observations of the evolution of IW in coastal areas.
In this paper, the optimization of perfect optical vortex (POV) parameter for underwater wireless optical communication link under M-QAM by average bit-error rate (ABER) and the effect of seawtaer turbulence on link information capacity are investigated. The link is absorbent, weakly turbulent, and bandwidth-limited. In investigating, we use the spectral absorption coefficient to describe the wavelength effect of seawater absorption. Specifically, under the paraxial approximation and Rytov approximation conditions, we define the average signal-to-noise-crosstalk ratio including the system bandwidth factor and derive the bandwidth-limited ABER of the OAM carrier link. Capitalizing on the defined average signal-to-noise crosstalk ratio and the derived bandwidth-limited ABER of link, the novel closed-form expression for the average information capacity of the perfect optical vortex link under M-QAM modulation is proposed. Through the numerical analysis of the ABER and the average information capacity, the POV optimization parameters in specific communication links are obtained and new conclusions are drawn that the average information capacity is restricted by both signal wavelength and the seawater absorption coefficient.
Being a powerful strategy to preclude drag and achieve hydrodynamic invisibility, flow field manipulation is attracting widespread attention. In this investigation, we introduce a systematic set of arbitrary-space divide-and-conquer transformation strategies to design complex hydrodynamic cloaks. This theory removes the difficulties associated with the analytic description of complicated and irregular structures to construct hydrodynamic cloaks by adopting the divide-and-conquer algorithm and reconfiguring strategies. It also provides an approach for redistributing the flow field energy and guiding the fluid flow as desired. The proposed theory not only opens up new ideas for improving the speed and concealment of marine vehicles but also provides a new strategy for ensuring the safety of aquatic and underwater structure operations.