Search Results (126)

Genetic programming (GP) has been widely used in quantitative finance for discovering formulaic alpha factors that can predict asset returns. However, GP often produces overgrown expressions that are difficult to interpret and expensive to evaluate. This paper proposes a large language model (LLM)-assisted pruning framework that reviews expression trees generated by GP, with the LLM acting as a semantic reviewer that flags redundant or financially implausible branches based on structural complexity and contextual reasoning. The proposed method is formalized as a closed-loop Trigger–Evaluate–Decide–Execute (TEDE) process. We present mathematical formulations, algorithmic design, and examples showing how redundant nested functions can be simplified while monitoring predictive performance. Experiments with high-frequency cryptocurrency market data, using DeepSeek-V4-Flash as the semantic engine, show lower expression complexity and higher rubric-based interpretability scores for the pruned symbolic factors. Under the reported test setup, the LLM-pruned configuration has higher Information Ratio (IR) values than the listed baselines and more compact expression trees than the GP baselines. Full article

This study examines the psychometric functioning of the Basic Learning Ability Assessment (BLAA), a high-stakes placement assessment used in Taiwan’s Adaptive Guidance Placement System (AGPS) for junior high school students with intellectual disabilities (IDs). The sample comprised 203 ninth-grade students with ID from 47 public junior high schools in Taiwan, all of whom completed three operational multiple-choice forms of the BLAA. Using classical test theory (CTT), we examined item difficulty using proportion-correct indices, item discrimination using upper–lower group discrimination indices, distractor functioning by comparing response patterns between higher- and lower-performing examinees, and internal consistency reliability using the Kuder–Richardson Formula 20 (KR-20). The results show that most items fell within the average-to-easy range and demonstrated acceptable to strong discrimination. Distractor functioning was generally satisfactory, with most items containing no nonfunctioning distractors. KR-20 coefficients ranged from 0.904 to 0.926, indicating high internal consistency within each form. Functional Language and Social Adaptation showed relatively stable psychometric patterns, whereas Mathematical Skills displayed greater variability in item difficulty, discrimination, and distractor functioning. Overall, the findings provide initial CTT-based internal psychometric evidence regarding the item functioning and form-level reliability of the BLAA, while highlighting the need for ongoing item refinement, particularly in the Mathematical Skills domain. Full article

Spectrum-based fault localization (SBFL), one of the typical types of software fault localization techniques, has been widely adopted to assist developers in identifying faulty program elements. However, conventional SBFL techniques rely solely on test coverage statistics and overlook intrinsic characteristics of the source code itself. To fill this gap, this study proposes an enhanced SBFL approach, code-naturalness-based fault localization (CNFL), which incorporates code naturalness evaluated by a large language model (LLM) in the pipeline of SBFL. By weighting program statements according to their naturalness scores, CNFL prioritizes statements that deviate from typical coding patterns and therefore optimizes the coverage matrix for effective fault localization. Comprehensive experiments are conducted on the Defects4J dataset with five representative SBFL formulas and five LLMs for naturalness evaluation. The results demonstrate that CNFL significantly outperforms conventional SBFL techniques. Specifically, it boosts the Top-1 fault localization hit rate by up to 60.8% and 56.8% when applied to classic SBFL formulas like Jaccard and Ochiai, respectively. Moreover, CNFL is further confirmed to consistently surpass both standalone LLM methods and representative fault localization approaches that primarily optimize the coverage matrix. Full article

With conditional probability as a primitive notion rather than a ratio of classical probability, we extend the language of epistemic logic by introducing conditional probability operators. We propose a conditional probabilistic epistemic logic (CPEL) based on the general frame, which enables the assignment of conditional probabilities to any formula in the language defined in this paper. Furthermore, we discuss the relationship between knowledge and conditional probability in CPEL, as well as the connection between indicative conditionals and conditional probability. Finally, we present a sound and weakly complete axiomatization for our framework and demonstrate its application in analyzing the lottery paradox. Full article

Increasing product complexity, shorter development cycles and cross-domain integration demands pose significant challenges for modern race car engineering teams. In Formula Student teams, heterogeneous toolchains, manual data exchange, late system integration, and high personnel turnover hinder efficient collaborative development and lead to repeated knowledge loss. This paper presents an integrated digital-engineering framework combining graph-based design languages (GBDL), model-to-text transformations, natural-language interactions via Large Language Models (LLMs), and Git-based version control to address these issues. By formalizing design knowledge and storing it in a centralized design graph, the framework ensures digital consistency of data and models, supports automated vehicle design variant generation, and enables seamless cross-domain integration. Through case studies of three Formula Student teams, the methodology demonstrates quantifiable reductions in design iteration time, enabling the evaluation of more than 10⁴ suspension variants within days instead of a few dozen manually created variants, while reducing hands-on engineering effort from minutes per variant to a largely unattended optimization process. The results indicate that the approach not only enhances efficiency and collaboration but also preserves design knowledge for long-term knowledge management and reuse. Looking forward, this methodology provides a scalable route toward further engineering automation, systematic variant-driven development, and early-stage design optimization supported by design languages and integrated downstream toolchains. Full article

This study addresses the issue of missing basic data and insufficient accuracy in predicting runoff and non-point-source pollution in the Heilongjiang region of China using the Soil and Water Assessment Tool (SWAT) model. Based on the China Ground Climate Data Daily Dataset (V3.0) and SPAW soil characteristic calculation formula, and assisted by the Python V3.0 language for data processing and computation, new high-precision weather generators and soil attribute databases suitable for the Heilongjiang region of China were established. The weather generator is based on daily data and contains detailed meteorological parameters such as temperature, humidity, wind speed, rainfall, etc., used to characterize the periodic changes in meteorological elements. And the differences and fluctuations outside this change curve were also retained in the basic construction of the weather generator. The soil database covers various parameters, such as soil type, texture, structure, nutrient content, organic matter content, etc., enabling the SWAT model to better simulate hydrological and pollutant transport processes in the soil. Additionally, point-source input data, including various industrial and domestic wastewater discharge situations, were collected and organized to improve data quality. Furthermore, a series of agricultural management measures were developed based on the use of fertilizers and pesticides for simulation, providing an important basis for analyzing non-point-source pollution using the SWAT model. By comparing the different results of the simulation using optimized databases, it is shown that the above work improved the simulation accuracy of the SWAT model in predicting runoff and pollution load in Heilongjiang, China. The NSE of runoff simulation increased from 0.923 to 0.988, and the NSE of ammonia nitrogen and CBOD simulation increased from 0.852 and 0.758 to 0.930 and 0.902, respectively. It is expected that these efforts will provide strong data support for subsequent research and provide a theoretical basis for government decision-makers to build scientifically rigorous and effective pollution control strategies. Full article

As a contribution to automated set-theoretic inferencing, a translation is proposed of conjunctions of literals of the forms $x=y\setminus z$, $x\ne y\setminus z$, and $z=\left\{x\right\}$, where $x,y,z$ stand for variables ranging over the von Neumann universe of sets, into quantifier-free Boolean formulae of a rather simple conjunctive normal form. The formulae in the target language involve variables ranging over a Boolean ring of sets, along with a difference operator and relators designating equality, non-disjointness, and inclusion. Moreover, the result of each translation is a conjunction of literals of the forms $x=y\setminus z$ and $x\ne y\setminus z$ and of implications whose antecedents are isolated literals and whose consequents are either inclusions (strict or non-strict) between variables, or equalities between variables. Besides reflecting a simple and natural semantics, which ensures satisfiability preservation, the proposed translation has quadratic algorithmic time complexity and bridges two languages, both of which are known to have an $\mathsf{NP}$-complete satisfiability problem. Full article

The development of modern vehicles faces an increase in complexity, as well as a need for shorter development cycles and a seamless cross-domain integration. In order to meet these challenges, a graph-based design language which formalizes and automates engineering workflows is presented and applied in a design case study to a Formula Student race car suspension system. The proposed method uses an ontology-based vocabulary definition and executable model transformations to compile design knowledge into a central and consistent design graph. This graph enables the automatic generation of consistent 3D CAD models, domain-specific simulations and suspension kinematic analyses, replacing manual and error-prone tool and data handover processes. The design language captures both the structural and dynamic behavior of the suspension, supports variant exploration and allows for integrated validation, such as 3D collision detection. The study illustrates how graph-based design languages can serve as ‘digital DNA’ for knowledge-based product development, offering a scalable, reusable platform for engineering automation. This approach enhances the digital consistency of data, the digital continuity of processes and the digital interoperability of tools across all relevant engineering disciplines in order to support the validation of early-stage designs and the optimization of complex systems. Full article

One of Wittgenstein’s most quoted passages from his Remarks on the Philosophy of Psychology concerns Turing’s “machines” and says verbatim: “These machines are humans who calculate. And one might express what he [Turing] says also in the form of games.” This passage not only captures the kernel of Turing’s conceptual argument for the adequacy of his definition of “computability”, as presented in his article On Computable Numbers (1936), but also helps clarify Turing’s idea of “mechanical intelligence.” Indeed, the notion of game provides an ideal means to focus on similarities and differences between Turing and Wittgenstein’s views of mechanical procedures, mathematical understanding, and thinking activity. The live encounter between Ludwig Wittgenstein and Alan Turing took place in Cambridge in 1939, when Wittgenstein’s Lectures on the Foundations of Mathematics were regularly attended by Turing. Interestingly, during the conversations between the two, Turing seems to play the role of the Wittgenstein of the Tractatus, to allow the present Wittgenstein to reassess what he deplores as mistaken or misleading in his early work. As for Turing himself, his reflection on thinking machines from the late 1940s demonstrates the significance of his dialogue with Wittgenstein. Full article

Road technical condition assessment and maintenance decision-making rely heavily on technical standards whose clauses, computational formulas, and decision logic are often expressed in unstructured formats, leading to fragmented knowledge representation, isolated indicator calculation procedures, and limited interpretability of decision outcomes. To address these challenges, a semantic framework with executable reasoning and computation components, Road Performance and Maintenance Ontology (RPMO), was developed, composed of a core ontology, an assessment ontology, and a maintenance ontology. The framework formalized clauses, computational formulas, and decision rules from standards and integrated semantic web rule language (SWRL) rules with external computational programs to automate distress identification and the computation and write-back of performance indicators. Validation through three use case scenarios conducted on eleven expressway asphalt pavement segments demonstrated that the framework produced distress severity inference, indicator computation, performance rating, and maintenance recommendations that were highly consistent with technical standards and expert judgment, with all reasoning results traceable to specific clauses and rule instances. This research established a methodological foundation for semantic transformation of road technical standards and automated execution of assessment and decision logic, enhancing the efficiency, transparency, and consistency of maintenance decision-making to support explicit, reliable, and knowledge-driven intelligent systems. Full article

Second-order derivative information, including mixed curvature, is central to Newton and trust-region optimization, uncertainty quantification, and simulation-based design. Classical finite differences (FD) remain popular but require delicate step-size tuning and can suffer from cancelation and noise amplification. Complex-step differentiation offers machine-precision gradients without subtractive cancelation, yet many second-derivative complex-step formulas reintroduce differencing. Hyper-dual numbers provide an algebraically principled alternative: by lifting real code to a four-component commutative nilpotent algebra, one obtains exact first and mixed second derivatives from a single evaluation, without finite differencing. This article gives a consolidated theoretical and experimental foundation for hyper-dual numbers. We formalize the algebra, prove exact Taylor truncation at second order, derive coefficient–extraction formulas for gradients and Hessians, and state assumptions for exactness, including limitations at non-smooth points and the need to branch on real parts. We present implementation patterns and language skeletons (C++, Python 3.11.5, MATLAB R2023b), and we provide fair numerical comparisons with FD, complex-step, and AD baselines. Stability tests under additive noise and ill-conditioning, together with runtime and memory profiling, demonstrate that hyper-dual coefficients are robust and reproducible in floating-point arithmetic, particularly for black-box codes where Hessian information is needed but differencing is fragile. Full article

This study aims to assess the capability of Large Language Models (LLMs), particularly GPT-4o, to evaluate and modify the complexity level of Russian school textbooks. We lay the groundwork for developing scalable, context-aware methods for readability assessment and text simplification in Russian educational materials, areas where traditional formulas often fall short. Using a corpus of 154 textbooks covering various subjects and grade levels, we evaluate the extent to which LLMs accurately predict the appropriate comprehension level of a text and how well they simplify texts by targeted grade reduction. Our evaluation framework employs GPT-4o as a multi-role agent in three distinct experiments. First, we prompt the model to estimate the target comprehension age for each segment and identify five key linguistic or conceptual features underpinning its assessment. Second, we simulate student comprehension by instructing the model to reason step-by-step through whether the text is understandable for a hypothetical student of the given grade. Third, we examine the model’s ability to simplify selected fragments by reducing their complexity by three grade levels. We further measure model perplexity and output token probabilities to probe the prediction confidence and coherence. Results indicate that while LLMs show considerable potential in complexity assessment (i.e., MAE of 1 grade level), they tend to overestimate text difficulty and face challenges in achieving precise simplification levels. Ease of understanding assessments generally align with human expectations, although texts with abstract, technical, or poetic content (e.g., Physics, History, and Literary Russian) pose challenges. Our study concludes that LLMs can substantially complement traditional readability metrics and assist teachers in developing suitable Russian educational materials. Full article

This paper presents accounts of preposition selection and agreement in English within Dynamic Syntax. To achieve this, I introduce two new, non-semantic, labels into the tree language: $Ph$ that takes as values phonological forms which are modelled as ordered sets of phonemes and $Md$ which takes as values sets of $Ph$ values, the phonological forms of certain words and forms with which a particular word can collocate. While these labels are not grounded in semantic concepts like type and formula, they are nevertheless grounded in phonological concepts and thus ultimately in phonetic phenomena. These labels are introduced through the parsing of words and are used to constrain the forms of other words they can felicitously appear with, such as nouns and certain determiners or verbs with selected prepositions or prepositional phrases, in a straightforward manner. It is shown how the remnant agreement and selection patterns in modern (standard) English can be captured without any recourse to traditional categories such as gender, person and number. Certain disagreement phenomena are discussed as are the broader implications of the approach. Full article

Financial report question answering (FRQA) presents unique challenges due to the need for precise numerical reasoning, complex table structures, and multi-table associations. Existing approaches often overlook the domain-specific complexities of financial reports and struggle with accurate numerical computation, leading to suboptimal performance in real-world financial intelligence applications. In this study, we propose FinQA-PKD, a framework designed to mitigate these challenges through a novel integration of progressive knowledge distillation and numerical reasoning enhancement. Our method introduces a difficulty-aware curriculum learning strategy that organizes training into two progressive stages, facilitating more effective and stable model learning. To address the limitations of large language models in numerical reasoning, we develop a numerical reasoning enhancement module that automatically decomposes calculation chains, augments numerical tokens, and validates results using a financial formula library. Furthermore, we implement a domain-adaptive selective knowledge distillation strategy, which evaluates teacher model outputs based on numerical accuracy, calculation correctness, and terminology precision, and selectively distills knowledge from high-quality samples. Experimental results in benchmark datasets demonstrate that FinQA-PKD improves numerical and calculation accuracy, achieving competitive performance with reduced computational resources. This framework provides a robust and efficient solution for answering financial report questions in practical financial analysis scenarios. Full article

Background/Objectives: Worldwide, an estimated 20–30% of infants suffer from infant colic (IC), with excessive crying and unsettled behavior, during the first three months of life. These infants are often referred for a medical evaluation, but the pathogenesis of IC remains poorly understood. The aim of this narrative review is to critically appraise the available literature regarding the relation between IC and cow’s milk allergy (CMA). Methods: A literature search using the search strings cow’s milk allergy [MeSH Terms] OR food allergy [MesH Terms] AND colic [MeSH Terms] OR crying [MeSH Terms], limited to the English language, from inception to 15 June 2025, resulted in the identification of 135 articles. Of these, 18 clinical trials assessed the effect of a cow’s milk elimination diet on IC. Results: The role of CMA in IC in the absence of other allergic manifestations remains uncertain. However, when standard treatment of infant colic has failed and when other allergic symptoms are present, CMA may be considered. A diagnostic elimination diet which includes a 2–4-week trial of maternal cow’s milk elimination in breastfed infants or an extensively hydrolyzed cow’s milk or hydrolyzed rice formula should be performed. If the elimination diet results in a significant decrease in symptoms, reintroduction of cow’s milk protein into the diet is mandatory to fulfill the diagnostic criteria of CMA. Conclusions: Considering the limited current evidence, future research should prioritize large well-designed clinical trials with a focus on investigating CMA in colicky breastfed and formula-fed infants. Full article