An Empirical Evaluation of Prompt Injection Vulnerabilities in Large Language Models Across Multilingual and Obfuscated Attack Scenarios

Large Language Models (LLMs) have rapidly evolved, transforming industries by automating complex tasks and generating human-like content. However, as their adoption accelerates, prompt injection vulnerabilities have become increasingly apparent. Malicious actors exploit these weaknesses to generate phishing emails, deceptive websites, nd malware, posing serious security risks. This paper presents an empirical evaluation of six state-of-the-art LLMs (DeepSeek, GPT, Gemini, Grok, Llama, and Qwen) under diverse adversarial prompt scenarios, including direct and multi-stage obfuscated attacks across multiple languages and character encodings. The proposed framework measures how effectively current LLMs resist manipulation into performing harmful actions. Our findings reveal systematic vulnerabilities across all tested models. Even direct prompt injections frequently induce the generation of phishing content, websites, and malware, while elaborate prompts achieve even higher malicious compliance rates, particularly for phishing. Models such as DeepSeek, Gemini, and Grok show especially high susceptibility under complex instructions. Notably, non-English languages consistently exhibit …

Comparison of multimetric D index with keratometric, pachymetric, and posterior elevation parameters in diagnosing subclinical keratoconus in fellow eyes of asymmetric keratoconus patients.

To compare the multimetric D index and other keratoconus-screening parameters in patients with clinical keratoconus in 1 eye and subclinical keratoconus in the fellow eye.

Gravitational Wave Solutions to Linearized Jordan-Brans-Dicke Theory on a Cosmological Background

Approximate vacuum solutions of Jordan-Brans-Dicke theory for perturbed scalar field and perturbed Robertson-Walker metric, are found. Solutions for the scale factor and the scalar field in unperturbed JBD theory are dependent on the $ω$ parameter which determines how the scalar field is coupled to geometry of space-time. After adding a metric perturbation to Robertson-Walker metric and a perturbation to the scalar field, we solved the linearized JBD equations and found the scale factor and the scalar field as $a\propto t$ and $φ\propto t^{-2}$ with $ω=$-3/2. The results are necessary conditions for ordinary and scalar gravitational waves to exist in the vacuum case. Despite omega is a large positive number for current solar system environment observations, this value of omega makes JBD theory conformally invariant and fits recent supernovae type Ia data. We also looked for the value of omega for the case which has nonzero spatial curvature parameter.

[The struggle against malaria in the Ottoman Empire during World War I and the legal regulations made to this end].

One of the most important disadvantages of war environmental is infectious diseases. The Ottoman Empire combated infectious diseases in addition to the war because of Balkan wars and afterwards first world war. Because of increasing migrations to Anatolia after Balkan wars spread some epidemic diseases, such as cholera, typhoid fever, plaque, dysentery, syphilis. With the start of the First World War, malaria began to spread within civilian population as well as the military. The population fell from power because of illness and therefore could not process the land tax failed to pay taxes. Founded in 1914 with the fight against epidemic diseases was initiated by the Sıhhiye ministry. Quinine was formed as tablets which was imported from Germany by legal regulation and was distributed to the public by Ziraat Bank. However, malaria epidemic could not be prevented because of long war years, lack of population, insufficiency of the preventive methods and lack of quinine, and about three quarters of the population caught malaria and in four years 412.000 soldiers had malaria and 20.000 of them died despite of measures.

Core-elements Subsampling for Alternating Least Squares

In this paper, we propose a novel element-wise subset selection method for the alternating least squares (ALS) algorithm, focusing on low-rank matrix factorization involving matrices with missing values, as commonly encountered in recommender systems. While ALS is widely used for providing personalized recommendations based on user-item interaction data, its high computational cost, stemming from repeated regression operations, poses significant challenges for large-scale datasets. To enhance the efficiency of ALS, we propose a core-elements subsampling method that selects a representative subset of data and leverages sparse matrix operations to approximate ALS estimations efficiently. We establish theoretical guarantees for the approximation and convergence of the proposed approach, showing that it achieves similar accuracy with significantly reduced computational time compared to full-data ALS. Extensive simulations and real-world applications demonstrate the effectiveness of our method in various scenarios, emphasizing its potential in large-scale recommendation systems.

[The city of Canik and provisioning during World War I, 1914-18].

Beyond Truncation: Rethinking LLM Decoding as Ensemble Pruning

We introduce Mahalanobis-Ensemble Decoding (ME-Decoding), a novel Large Language Model (LLM) decoding framework that frames candidate token selection as ensemble pruning. Existing selection strategies rely predominantly on scalar probabilities, ignoring geometric semantic relationships and causing candidate redundancy. Meanwhile, current geometry-aware methods often require complex optimization or directly reweighting the original token probabilities, leading to significant computational overhead or inference instability. To address this, we formulate decoding as a subset optimization problem using a Mahalanobis distance-driven objective to enhance semantic diversity while preserving high probabilities. Specifically, we dynamically discount redundant generation paths using a token similarity matrix, constructed via an adaptive-bandwidth kernel over token embeddings. We further devise an efficient greedy selection algorithm with near-linear complexity in the candidate size under early stopping, while establishing its theoretical approximation guarantees. This renders ME-Decoding a robust, plug-and-play module with negligible inference overhead. Extensive experiments across diverse …

The Higgs Field and the Jordan Brans Dicke Cosmology

We investigate a field theoretical approach to the Jordan-Brans- Dicke (JBD) theory extended with a particular potential term on a cosmological background by starting with the motivation that the Higgs field and the scale factor of the universe are related. Based on this relation, it is possible to come up with mathematically equivalent but two different interpretations. From one point of view while the universe is static, the masses of the elementary particles change with time. The other one, which we stick with throughout the manuscript, is that while the universe is expanding, particle masses are constant. Thus, a coupled Lagrangian density of the JBD field and the scale factor (the Higgs field), which exhibit a massive particle and a linearly expanding space in zeroth order respectively, is obtained. By performing a coordinate transformation in the field space for the reduced JBD action whose kinetic part is nonlinear sigma model, the Lagrangian of two scalar fields can be written as uncoupled for the Higgs mechanism. After this transformation, as a result of spontaneous symmetry breaking, the time dependent vacuum expectation value (vev) of the Higgs field and the Higgs …

NeuroChat: A Neuroadaptive AI Chatbot for Customizing Learning Experiences

Generative AI is transforming education by enabling personalized, on-demand learning experiences. However, current AI systems lack awareness of the learner's cognitive state, limiting their adaptability. Meanwhile, electroencephalography (EEG)-based neuroadaptive systems have shown promise in enhancing engagement through real-time physiological feedback. This paper presents NeuroChat, a neuroadaptive AI tutor that integrates real-time EEG-based engagement tracking with generative AI to adapt its responses. NeuroChat continuously monitors a learner's cognitive engagement and dynamically adjusts content complexity, tone, and response style in a closed-loop interaction. In a within-subjects study (n=24), NeuroChat significantly increased both EEG-measured and self-reported engagement compared to a non-adaptive chatbot. However, no significant differences in short-term learning outcomes were observed. These findings demonstrate the feasibility of real-time cognitive feedback in LLMs, highlighting new directions for adaptive learning, AI tutoring, and deeper personalization in human-AI interaction.

A New Approach to Flatness, Horizon and Late-time Accelerating Expansion Problems on the basis of Mach Principle

Based on the idea that the components of a cosmological metric may be determined by the total gravitational potential of the universe, the scalar field $φ=1/G$ in the Jordan-Brans-Dicke (JBD) theory is introduced as evolving with the inverse square of the scale factor. Since the gravitational potential is related to the field $φ$ resulting from Mach's principle and depends on time due to the expansion of space, the temporal evolution of the field should be in accord with the evolution of time and space intervals in the metric tensor. For the same reason, the time dependence of the field makes these comoving intervals relative for different points on the time axis. Thus, it is shown that introduction of the cosmic gravitational potential as a time dependent scalar field proportional to $1/a^2$ may resolve the flatness, the horizon and the late-time accelerating expansion problems of the standard model of cosmology. The luminosity distance vs redshift data of Type Ia supernovae is in agreement with this approach.

A Multiple Linear Regression Approach For Estimating the Market Value of Football Players in Forward Position

In this paper, market values of the football players in the forward positions are estimated using multiple linear regression by including the physical and performance factors in 2017-2018 season. Players from 4 major leagues of Europe are examined, and by applying the test for homoscedasticity, a reasonable regression model within 0.10 significance level is built, and the most and the least affecting factors are explained in detail.

Transfer-free fabrication of graphene transistors

We invented a method to fabricate graphene transistors on oxidized silicon wafers without the need to transfer graphene layers. To stimulate the growth of graphene layers on oxidized silicon a catalyst system of nanometer thin aluminum/nickel double layer is used. This catalyst system is structured via liftoff before the wafer enters the catalytic chemical vapor deposition (CCVD) chamber. In the subsequent methane based growth process monolayer graphene field-effect transistors (MoLGFETs) and bilayer graphene transistors (BiLGFETs) are realized directly on oxidized silicon substrate, whereby the number of stacked graphene layers is determined by the selected CCVD process parameters, e.g. temperature and gas mixture. Subsequently, Raman spectroscopy is performed within the channel region in between the catalytic areas and the Raman spectra of fivelayer, bilayer and monolayer graphene confirm the existence of graphene grown by this silicon-compatible, transfer-free and in-situ fabrication approach. These graphene FETs will allow a simple and low-cost integration of graphene devices for nanoelectronic applications in a hybrid silicon CMOS environment.

Transfer-free Grown Bilayer Graphene Transistors for Digital Applications

We invented a novel method to fabricate graphene transistors on oxidized silicon wafers without the need to transfer graphene layers. By means of catalytic chemical vapor deposition (CCVD) the in-situ grown bilayer graphene transistors (BiLGFETs) are realized directly on oxidized silicon substrate, whereby the number of stacked graphene layers is determined by the selected CCVD process parameters, e.g. temperature and gas mixture. BiLGFETs exhibit ultra-high on/off-current ratios of 107 at room temperature, exceeding previously reported values by several orders of magnitude. This will allow a simple and low-cost integration of graphene devices for digital nanoelectronic applications in a hybrid silicon CMOS environment for the first time.