Large language models (LLMs) already play an influential role in how humans access information. However, their behavior varies depending on their design, training, and use. We prompt a diverse panel of 19 popular LLMs to describe 3,991 prominent persons with political relevance, and then judge how positively they portray each person. When comparing these assessments, we find disparities in ideological positions between LLMs across different geopolitical regions (Arabic countries, China, Russia, and Western countries), and across different languages (the United Nations’ six official languages). Moreover, among only models from the United States, we find significant normative differences related to progressive values. Among Chinese models, we characterize division between internationally- and domestically-focused models. Our results suggest that the ideological stance of an LLM reflects the worldview of its creators. This poses the risk of political instrumentalization and raises concerns around technological and regulatory efforts aiming to make LLMs ideologically ‘unbiased’.
This study aimed to evaluate a methodology based on multispectral time series acquired by UAV for detecting irrigation anomalies in maize at the plot level across different crop development stages. The research addressed how irrigation blockage duration, spatial resolution (pixel size), and topographic conditions influence anomaly detectability. Experiments were conducted during the 2024 summer growing season on two maize plots with contrasting topography: one flat and one sloped. In each plot, three sprinklers per irrigation event were intentionally blocked for periods ranging from 15 to 25 days, depending on crop phenological stage. Multispectral images were acquired every five days using a DJI Mavic 3 M UAV equipped with RGB and multispectral sensors and a GNSS RTK module. Flights were performed during peak solar irradiance to ensure radiometric consistency and generate homogeneous time series for analysis. Detectability of irrigation-induced vegetation responses was influenced by both pixel size and irrigation blockage duration. Higher spatial resolution improved detection performance, particularly during prolonged water deficit periods. Precipitation events and topographic variability attenuated the effect of irrigation blockages by increasing water availability, which delayed their differentiation in the NDVI time series. The proposed methodology integrates multispectral UAV time series with statistical testing based on p-values and mean differences between irrigated and non-irrigated zones to detect irrigation-induced vegetation responses at the plot scale. It demonstrates robustness under varying agronomic and topographic conditions and provides a transferable framework for irrigation monitoring, with potential application to other crops and integration with UAV and satellite-based precision agriculture systems.
Geometrical optimisation studies of thermoelectric modules often overlook the significant influence of system-level components, particularly heat exchangers. This research addresses this critical gap by developing an integrated system-level optimisation approach, applied to a geothermal passive thermoelectric generator installed in Antarctica. The methodology is based on the premise that the thermal and electrical phenomena are interdependent, demanding a holistic design. For the first time, a comprehensive multi-objective, multi-parameter computational model for geometric optimisation of modules is presented, which incorporates all thermoelectric effects, temperature-dependent properties, and all system components. The framework simultaneously maximises electrical power output and minimises semiconductor material volume by adjusting the thermocouple length and cross-sectional area to their optimal values. Application to the Antarctic installation demonstrated that geometric optimisation yields energy production improvements from 3.7% to 12.6% with respect to the commercial modules used in the installation. These improvements correspond to configurations with leg lengths of 0.11cm and 0.27cm and cross-sectional areas of 0.023cm2 and 0.06cm2, respectively. Crucially, an extended analysis revealed that while the source temperature difference does not affect optimal geometric parameters, the thermal resistances are paramount. Reducing the sum of heat exchanger thermal resistances by 26% increases optimisation potential by 60%, while also reducing semiconductor volume by 18.5%. These results demonstrate a system-level optimisation approach is essential. Findings indicate efforts must first prioritise optimising heat exchangers for the lowest thermal resistance. Subsequently, the optimal module geometry is determined by maximising the thermocouple cross-sectional area before adjusting the length for optimal thermal matching.
Credal sets, i.e., closed convex sets of probability measures, provide a natural framework to represent aleatoric and epistemic uncertainty in machine learning. Yet how to \emph{quantify} these two types of uncertainty for a given credal set, particularly in multiclass classification, remains underexplored. In this paper, we propose a distance-based approach to quantify total, aleatoric, and epistemic uncertainty for credal sets. Concretely, we introduce a family of such measures within the framework of Integral Probability Metrics (IPMs). The resulting quantities admit clear semantic interpretations, satisfy natural theoretical desiderata, and remain computationally tractable for common choices of IPMs. We instantiate the framework with the total variation distance and obtain simple, efficient uncertainty measures for multiclass classification. In the binary case, this choice recovers established uncertainty measures, for which a principled multiclass generalization has so far been missing. Empirical results confirm practical usefulness, with favorable performance at low computational cost.
This paper presents a compact 3 & times; 3 beamforming Nolen-matrix feeding network (BNMFN) based on groove gapwaveguide (GGW) technology for Ka-band satellite and space-multiplexing applications. The proposed design eliminates conventional components such as power dividers, crossovers, switches, and terminations by utilizing only directional couplers and phase-shift compensators, thereby reducing size, loss, and complexity. The network employs three 90 degrees couplers (3.14 dB and 4.84 dB) and two-phase shifters (-38 degrees and -63 degrees) to achieve output phase differences of 270 degrees, 30 degrees, and 150 degrees at 38 GHz. Integrated with a GGW-based feeding transition and a threeelement slotted antenna array, the system demonstrates multi-beam radiation with gains exceeding 17 dBi over 36-39 GHz. Experimental results validate low loss, wide bandwidth, and accurate phase performance, making the design a promising solution for 6 G millimeter-wave satellite communications and IoT connectivity.
