UMBC Student Collection

Permanent URI for this collectionhttp://hdl.handle.net/11603/33

Browse

Now showing 1 - 20 of 3661

Memorization Over Reasoning? Exposing and Mitigating Verbatim Memorization in Large Language Models' Character Understanding Evaluation
(2024-12-30) Jiang, Yuxuan; Ferraro, Francis
Recently, Large Language Models (LLMs) have shown impressive performance in character understanding tasks, such as analyzing the roles, personalities, and relationships of fictional characters. However, the extensive pre-training corpora used by LLMs raise concerns that they may rely on memorizing popular fictional works rather than genuinely understanding and reasoning about them. In this work, we argue that 'gist memory'-capturing essential meaning - should be the primary mechanism for character understanding tasks, as opposed to 'verbatim memory' - exact match of a string. We introduce a simple yet effective method to mitigate mechanized memorization in character understanding evaluations while preserving the essential implicit cues needed for comprehension and reasoning. Our approach reduces memorization-driven performance on popular fictional works from 96% accuracy to 72% and results in up to an 18% drop in accuracy across various character understanding tasks. These findings underscore the issue of data contamination in existing benchmarks, which often measure memorization rather than true character understanding.
MedReg-KG: KnowledgeGraph for Streamlining Medical Device Regulatory Compliance
(2024-12-15) Chattoraj, Subhankar; Joshi, Karuna
Healthcare providers are deploying a large number of AI-driven Medical devices to help monitor and medicate patients. For patients with chronic ailments, like diabetes or gastric diseases, usage of these devices becomes part of their daily lifestyle. These medical devices often capture personally identifiable information (PII) and hence are strictly regulated by the Food and Drug Administration (FDA) to ensure the safety and efficacy of the medical device. Medical device regulations are currently available as large textual documents, called Code of Federal Regulations (CFR) Title 21, that cross-reference other documents and so require substantial human effort and cost to parse and comprehend. We have developed a semantically rich framework MedReg-KG to extract the knowledge from the rules and policies for Medical devices and translate it into a machine-processable format that can be reasoned over. By applying Deontic Logic over the policies, we are able to identify the permissions and prohibitions in the regulation policies. This framework was developed using AI/Knowledge extraction techniques and Semantic Web technologies like OWL/RDF and SPARQL. This paper presents our Ontology/Knowledge graph and the Deontic rules integrated into the design. We include the results of our validation against the dataset of Gastroenterology Urology devices and demonstrate the efficiency gained by using our system.
An Efficient Computational Algorithm for Modeling Slow Soliton Interactions in Microresonators
(Optica, 2024) Akter, Sanzida; Shandilya, Pradyoth; Courtright, Logan; D’Aguanno, Giuseppe; Leshem, Amir; Gat, Omri; Menyuk, Curtis
Standard simulations of microresonator waveforms are limited by the photon lifetime. We describe a computational method that enables simulations on a laboratory timescale and apply this approach to study two-soliton interactions.
Leveraging the Human Factors Discipline for Better Cybersecurity Outcomes: A Roundtable Discussion
(IEEE, 2024-11) Cunningham, Margaret; Nobles, Calvin; Robinson, Nikki; Haney, Julie
Three human factors experts get to the bottom of what the human factors discipline actually is, how the cybersecurity community and organizations can benefit from it, and how to create a pipeline of professionals with human factors and cybersecurity expertise.
Increasing Visual Literacy With Collaborative Foraging, Annotation, Curation, and Critique
(ACM, 2024-12-05) Williams, Rebecca M.; Syed, Afrin Unnisa; Kurumaddali, Krishna Vamsi
Students today are facing information overload, contamination, and bloat from dubious sources: AI-generated content, masqueraded influencer opinions, context-less listicles, and consumer manipulation - frequently heralded by graphs and charts to bolster the argument. Because this information firehose presents as technical visual communications, the overload is both cognitive and perceptual, potentially causing more insidious misperceptions than text alone. In addition to consuming such media, students in computing fields work with data to produce graphs and charts themselves, including assignments, academic research, and personal projects/blog posts/tweets. Depending on visual literacy (VL) and prior data analysis instruction, many students inadvertently code misleading, unethical, or biased visualizations, potentially contributing to the dark corpus already festering online. Prior research on misconceptions in visualization pedagogy suggests students benefit from repeated opportunities to forage, curate and critique examples, discussing and debating with peers and instructors. Inspired by these findings, we incorporated a visual curation + annotation platform into a Data Visualization Computer Science course, enabling students to participate in processes of searching for and curating found examples of misleading visualizations, collaborative annotation + critique of examples, and structured self-evaluation of misleading elements in their own work. We assess our interventions with pre-/post-course Visualization Literacy Assessment Tests, qualitative evaluation of student reflections, taxonomic evaluation of formative student-produced visualizations, and post-course exit surveys. Post-course, students' VL increased significantly, and the number and severity of misleading visualizations they created decreased. Students also reflected that they gained increased confidence in spotting visual disinformation online, and in avoiding its creation in software.
Improving Gamma Imaging in Proton Therapy by Sanitizing Compton Camera Simulated Patient Data using Neural Networks through the BRIDE Pipeline
(UMBC High Performance Computing Facility, 2024) Chen, Michael O.; Hodge, Julian; Jin, Peter L.; Protz, Ella; Wong, Elizabeth; Cham, Mostafa; Gobbert, Matthias; Barajas, Carlos A.
Precision medicine in cancer treatment increasingly relies on advanced radiotherapies, such as proton beam radiotherapy, to enhance efficacy of the treatment. When the proton beam in this treatment interacts with patient matter, the excited nuclei may emit prompt gamma ray interactions that can be captured by a Compton camera. The image reconstruction from this captured data faces the issue of mischaracterizing the sequences of incoming scattering events, leading to excessive background noise. To address this problem, several machine learning models such as Feedfoward Neural Networks (FNN) and Recurrent Neural Networks (RNN) were developed in PyTorch to properly characterize the scattering sequences on simulated datasets, including newly-created patient medium data, which were generated by using a pipeline comprised of the GEANT4 and Monte-Carlo Detector Effects (MCDE) softwares. These models were implemented using the novel ‘Big-data REU Integrated Development and Experimentation’ (BRIDE) platform, a modular pipeline that streamlines preprocessing, feature engineering, and model development and evaluation on parallelized GPU processors. Hyperparameter studies were done on the novel patient data as well as on water phantom datasets used during previous research. Patient data was more difficult than water phantom data to classify for both FNN and RNN models. FNN models had higher accuracy on patient medium data but lower accuracy on water phantom data when compared to RNN models. Previous results on several different datasets were reproduced on BRIDE and multiple new models achieved greater performance than in previous research.
A New Semi-Discretization of the Fully Clamped Euler-Bernoulli Beam Preserving Boundary Observability Uniformly
(IEEE, 2024-12-17) Aydin, Ahmet Kaan; Haider, Md Zulfiqur; Özkan Özer, Ahmet
This letter extends a Finite Difference model reduction method to the Euler-Bernoulli beam equation with fully clamped boundary conditions. The corresponding partial differential equation (PDE) is exactly observable in the energy space with a single boundary observer in arbitrarily short observation times. However, standard Finite Difference spatial discretization fails to achieve uniform exact observability as the mesh parameter approaches zero, with minimal observation time potentially depending on the filtering parameter. To address this, we propose a Finite Difference algorithm incorporating an averaging operator and discrete multipliers, leveraging Haraux’s theorem on the spectral gap to ensure uniform observability. This approach eliminates the need for artificial viscosity or Fourier filtering. Our method achieves uniform observability for arbitrarily small times with dual observers-the tip moment and average tip velocity-mirroring results from mixed Finite Elements applied to the wave equation with homogeneous Dirichlet boundary conditions, where dual controllers converge to the single controller of the PDE model [Castro, Micu-Numerische Mathematiik’06]. Our reduced model is applicable to more complex systems involving Euler-Bernoulli beam equations.
Extracting the Optical Constants of Partially Absorbing TiO2 ALD Films
(MDPI, 2024-12-12) Chowdhary, Nimarta Kaur; Gougousi, Theodosia
Typical titanium oxide (TiO₂) films are transparent in the visible range, allowing for their index of refraction and thickness to be extracted by single-angle spectroscopic ellipsometry (SE) using a Cauchy model. However, TiO₂ films grown by atomic layer deposition (ALD) from tetrakis(dimethylamino)titanium (IV) (TDMAT) and H₂O at 350 °C absorb in the visible range due to the formation of Ti-O-N/Ti-N bonds. Single-angle SE is inadequate for extracting the optical constants of these films, as there are more unknowns (n, k, d) than measurable parameters (ψ, Δ). To overcome this limitation, we combined SE with transmission (T) measurements, a method known as SE + T. In the process, we developed an approach to prevent backside deposition on quartz substrates during ALD deposition. When applying a B-spline model to SE + T data, the film thicknesses on the quartz substrates closely matched those on companion Si samples measured via standard lithography. The resulting optical constants indicate a reduced refractive index, n, and increased extinction coefficient, k, when compared to purer TiO₂ thin films deposited via a physical vapor deposition (PVD) method, reflecting the influence of nitrogen incorporation on the optical properties.
Demand Modeling for Advanced Air Mobility
(2024-11-25) Acharya, Kamal; Lad, Mehul; Sun, Liang; Song, Houbing
In recent years, the rapid pace of urbanization has posed profound challenges globally, exacerbating environmental concerns and escalating traffic congestion in metropolitan areas. To mitigate these issues, Advanced Air Mobility (AAM) has emerged as a promising transportation alternative. However, the effective implementation of AAM requires robust demand modeling. This study delves into the demand dynamics of AAM by analyzing employment based trip data across Tennessee's census tracts, employing statistical techniques and machine learning models to enhance accuracy in demand forecasting. Drawing on datasets from the Bureau of Transportation Statistics (BTS), the Internal Revenue Service (IRS), the Federal Aviation Administration (FAA), and additional sources, we perform cost, time, and risk assessments to compute the Generalized Cost of Trip (GCT). Our findings indicate that trips are more likely to be viable for AAM if air transportation accounts for over 70\% of the GCT and the journey spans more than 250 miles. The study not only refines the understanding of AAM demand but also guides strategic planning and policy formulation for sustainable urban mobility solutions. The data and code can be accessed on GitHub.{https://github.com/lotussavy/IEEEBigData-2024.git }
Current Strategies and Future Directions of Wearable Biosensors for Measuring Stress Biochemical Markers for Neuropsychiatric Applications
(Wiley, 2024-12-17) Sheffield, Zach; Paul, Priyanka; Krishnakumar, Shraddha; Pan, Prof Dipanjan
Most wearable biosensors aimed at capturing psychological state target stress biomarkers in the form of physical symptoms that can correlate with dysfunction in the central nervous system (CNS). However, such markers lack the specificity needed for diagnostic or preventative applications. Wearable biochemical sensors (WBSs) have the potential to fill this gap, however, the technology is still in its infancy. Most WBSs proposed thus far target cortisol. Although cortisol detection is demonstrated as a viable method for approximating the extent and severity of psychological stress, the hormone also lacks specificity. Multiplex WBSs that simultaneously target cortisol alongside other viable stress-related biochemical markers (SBMs) can prove to be indispensable for understanding how psychological stress contributes to the pathophysiology of neuropsychiatric illnesses (NPIs) and, thus, lead to the discovery of new biomarkers and more objective clinical tools. However, none target more than one SBM implicated in NPIs. Till this review, cortisol's connection to dysfunctions in the CNS, to other SBMs, and their implication in various NPIs has not been discussed in the context of developing WBS technology. As such, this review is meant to inform the biosensing and neuropsychiatric communities of viable future directions and possible challenges for WBS technology for neuropsychiatric applications.
Time transfer and clock synchronization with ghost frequency comb
(AIP, 2024-12-10) Joshi, Binod; Smith, Thomas A.; Shih, Yanhua
We report an experimental demonstration of a time transfer and distant clock synchronization scheme based on what we have labeled as a ghost frequency comb, observed from the nonlocal correlation measurements of a laser beam. Unlike a conventional frequency comb, the laser beam used in this work does not consist of a pulse train but instead it is in a continuous-wave operation. The laser beam, consisting of half a million longitudinal cavity modes from a fiber ring laser, is split into two beams, each sent to a distant observer. In their local measurements, both observers observe constant intensity with no pulse structure present. Surprisingly, a pulse train of comb-like, ultra-narrow peaks is observed from their nonlocal correlation function measurement. This observation makes an important contribution to the field of precision spectroscopy, as we show in optical correlation-based nonlocal timing and positioning.
Significance of Functional Status Scale in decannulation after pediatric tracheostomy: A single-center, retrospective study
(Wolters Kluwer, 2024) Teplitzky, Taylor B.; Randolph, Nicholas Paul; Li, Ji; Pereira, Kevin D.; Gopalakrishnan, Mathangi; Holloway, Adrian
Background:� Metrics to successfully predict pediatric decannulation have been ineffective. The Functional Status Scale (FSS) is a validated pediatric scoring system of functional outcomes. The objective of this study was to evaluate if the FSS over time predicts pediatric tracheostomy decannulation. Subjects and Methods:� Chart review of patients admitted to the pediatric intensive care unit (PICU) and underwent tracheostomy at a tertiary care children抯 hospital from 2010 to 2019. Baseline demographics, comorbidities, tracheostomy indication, decannulation status, and FSS scores were recorded at PICU discharge and 1 and 3 years after tracheostomy. Logistic regression was performed to assess the association of FSS components with decannulation status at 3 years. Results:� Fifty-three patients met the inclusion criteria. Forty (75.5%) patients had complete data. There were no decannulations at 1 year. Nine (22.5%) patients were decannulated at 3 years. An abnormal 3-year FSS score in the feeding domain was significantly associated with persistent tracheostomy at 3 years, with an odds ratio of 7.4 (95% confidence interval: 1.5�.6, P = 0.01). Conclusions:� FSS score can predict decannulation in children discharged from the PICU. This information could modify caregiver expectations and guide rehabilitative efforts.
Robust and Lightweight Challenge Obfuscation Mechanism for Anti-modeling Protection of Arbiter-PUFs
(Springer Nature, 2024-12-06) Ebrahimabadi, Mohammad; Younis, Mohamed; Sanjana Mehjabin, Suhee; Tekeoglu, Ali; Sookoor, Tamim I.; Karimi, Naghmeh
Physically unclonable functions (PUFs) are lightweight hardware security primitives that leverage the imperfection of the manufacturing process of integrated circuits to generate unique signatures (responses) when queried by various bit-strings (challenges). These signatures can be used not only to authenticate interconnected devices but also to generate cryptographic keys for preserving the integrity and confidentiality of data. Among the different designs, the arbiter-PUF and its variants have received the most attention due to the large cardinality of their challenge-response set. To prevent the PUF circuits from being modeled using machine learning techniques, challenge obfuscation is often being pursued. Particularly, the simplicity of bit scrambling makes it an attractive means to achieve such a goal without diminishing the low complexity advantages of PUFs. This paper first shows that the conventional, fixed pattern-based, bit scrambling scheme is vulnerable by developing a detailed attack scenario. Then, we propose a novel lightweight dynamic challenge scrambling (DCS) mechanism that predictably varies the bit-swapping pattern per packet and per node. Such variability severely degrades the PUF modeling accuracy. The results extracted from the FPGA implementation of DCS confirm its effectiveness in thwarting PUF modeling attacks.
Relaxation optimized heteronuclear experiments for extending the size limit of RNA nuclear magnetic resonance
(2024-12-11) Shah, Aarsh; Patel, Heer; Kanjarpane, Arjun; Summers, Michael; Marchant, Jan
The application of NMR to large RNAs has been limited by the inability to perform heteronuclear correlation experiments essential for resolving overlapping 1H NMR signals, determining inter-proton distance restraints and inter-helical orientations for structure calculations, and evaluating conformational dynamics. Approaches exploiting 1H-13C correlations that are routinely applied to proteins and small RNAs of ~50 nucleotides or fewer are impractical for larger RNAs due to rapid dipolar relaxation of protons by their attached carbons. Here we report a 2H-enhanced, 1H-15N correlation approach that enables atom-specific NMR characterization of much larger RNAs. Purine H8 transverse relaxation rates are reduced ~20-fold with ribose perdeuteration, enabling efficient magnetization transfer via two-bond 1H-15N couplings. We focus on H8-N9 correlation spectra which benefit from favorable N9 chemical shift anisotropy. Chemical shift assignment is enabled by retention of protons at the C1? position, which allow measurement of H8-H1? NOEs and two-bond H1?-N9 correlation strategies with only a minor effect on H8 relaxation. The approach is demonstrated for the 232 nucleotide HIV-1 Rev response element, where chemical shift assignments, 15N-edited nuclear Overhauser effects, and 1H-15N residual dipolar couplings are readily obtained from sensitive, high-resolution spectra. Heteronuclear correlated NMR methods that have been essential for the study of proteins can now be extended to RNAs of at least 78 kDa.
Regional Air Mobility Flight Demand Modeling in Tennessee State
(2024-12-11) Acharya, Kamal; Lad, Mehul; Song, Houbing; Sun, Liang
Advanced Air Mobility (AAM), encompassing Urban Air Mobility (UAM) and Regional Air Mobility (RAM), offers innovative solutions to mitigate the issues related to ground transportation like traffic congestion, environmental pollution etc. RAM addresses transportation inefficiencies over medium-distance trips (50-500 miles), which are often underserved by both traditional air and ground transportation systems. This study focuses on RAM in Tennessee, addressing the complexities of demand modeling as a critical aspect of effective RAM implementation. Leveraging datasets from the Bureau of Transportation Statistics (BTS), Internal Revenue Service (IRS), Federal Aviation Administration (FAA), and other sources, we assess trip data across Tennessee's Metropolitan Statistical Areas (MSAs) to develop a predictive framework for RAM demand. Through cost, time, and risk regression, we calculate a Generalized Travel Cost (GTC) that allows for comparative analysis between ground transportation and RAM, identifying factors that influence mode choice. When focusing on only five major airports (BNA, CHA, MEM, TRI, and TYS) as RAM hubs, the results reveal a mixed demand pattern due to varying travel distances to these central locations, which increases back-and-forth travel for some routes. However, by expanding the RAM network to include more regional airports, the GTC for RAM aligns more closely with traditional air travel, providing a smoother and more competitive option against ground transportation, particularly for trips exceeding 300 miles. The analysis shows that RAM demand is likely to be selected when air transportation accounts for more than 80\% of the total GTC, air travel time is more than 1 hour and when the ground GTC exceeds 300 for specific origin-destination pairs. The data and code can be accessed on GitHub. {https://github.com/lotussavy/AIAAScitecth-2025.git}
Psychological Ownership in Teleinstructive Augmented Reality Workspaces
(ACM, 2025-01-10) Mentis, Helena; Seo, Jwawon; Avellino, Ignacio
Psychological ownership over virtual and physical spaces in augmented reality can lead to tensions between collaborators, yet, there is still a significant challenge in understanding how psychological ownership manifests in shared AR and what that might mean for the inclusion of collaborative interaction mechanisms. Through an experimental instruction task with a teleAR system, we interviewed 16 participant pairs on their perceptions of ownership of virtual and physical spaces and how they thought their perceptions impacted their interaction within those spaces. Our findings indicate (1) how AR introduces new ideas around behavioral norms in spaces that are layered and (2) that the nature of the task itself, in this case one of instruction where collaborators have different levels of knowledge and the local worker is reliant on the remote expert, significantly affects the perceptions of ownership and therefore behavior norms.
Posttraumatic Stress Disorder and Emotion Regulation Difficulties among Sexual Minority Adults in Residential Substance Use Disorder Treatment
(Taylor & Francis, 2024-12-09) Meyer, Laurel; Wenzel, Kevin R.; Berg, Samantha; Mette, Meghan; Schacht, Rebecca
Background: PTSD rates are higher among lesbian, gay, bisexual, and other sexual minority individuals (LGB+), compared to heterosexual individuals. PTSD also frequently co-occurs with substance use disorders (SUDs). However, little is known about comorbid PTSD-SUD among LGB+ individuals. Further research is important given elevated rates of PTSD and SUD among LGB+ individuals and to inform culturally responsive practice. Objectives: This cross-sectional study examined trauma exposure, PTSD severity, and emotion regulation (ER) difficulties among LGB+ and heterosexual individuals in residential SUD treatment. We hypothesized that LGB+ individuals would report more trauma exposure and more severe PTSD and ER difficulties compared to heterosexual peers. We also hypothesized that adding ER difficulties to the hierarchical regression model would attenuate the contribution of sexual minority status to PTSD symptom severity. Results: Cross-sectional data were collected via questionnaires from 132 adults receiving residential SUD treatment (M age = 39.79 [SD = 12.26] years; 35% women, 65% men; 49% White, 40% Black, 11% multiracial/another race). Eighteen percent of the sample identified as LGB+ (29% gay or lesbian, 63% bisexual, and 8% other), and 82% identified as heterosexual. Consistent with hypotheses, LGB+ participants reported larger numbers of traumatic events (p < 0.01) and more severe PTSD symptoms (p < 0.01) and ER difficulties (p < 0.05). Controlling for trauma exposure, the association between sexual minority status and PTSD symptom severity became non-significant after adding ER difficulties to the model. Conclusion: This suggests that ER may play an important role in the relationship between sexual minority status and PTSD severity in individuals with SUD.
Multi-modal Pre-silicon Evaluation of Hardware Masking Styles
(Springer, 2024-12-16) Anik, Md Toufiq Hasan; Reefat, Hasin Ishraq; Cheng, Wei; Danger, Jean-Luc; Guilley, Sylvain; Karimi, Naghmeh
Protecting sensitive logic functions in ASICs requires side-channel countermeasures. Many gate-level masking styles have been published, each with pros and cons. Some styles such as RSM, GLUT, and ISW are compact but can feature 1st-order leakage. Some other styles, such as TI, DOM, and HPC are secure at the 1st-order but incur significant overheads in terms of performance. Another requirement is that security shall be ensured even when the device is aged. Pre-silicon security evaluation is now a normatively approved method to characterize the expected resiliency against attacks ahead of time. However, in this regard, there is still a fragmentation in terms of leakage models, Points of Interest (PoI) selection, attack order, and distinguishers. Accordingly, in this paper we focus on such factors as they affect the success of side-channel analysis attacks and assess the resiliency of the state-of-the-art masking styles in various corners. Moreover, we investigate the impact of device aging as another factor and analyze its influence on the success of side-channel attacks targeting the state-of-the-art masking schemes. This pragmatic evaluation enables risk estimation in a complex PPA (Power, Performance, and Area) and security plane while also considering aging impacts into account. For instance, we explore the trade-off between low-cost secure styles attackable at 1st-order vs high-cost protection attackable only at 2nd-order.
Mode Coresets for Efficient, Interpretable Tensor Decompositions: An Application to Feature Selection in fMRI Analysis
(IEEE, 2024-12-13) Gabrielson, Ben; Yang, Hanlu; Vu, Trung; Calhoun, Vince; Adali, Tulay
Generalizations of matrix decompositions to multidimensional arrays, called tensor decompositions, are simple yet powerful methods for analyzing datasets in the form of tensors. These decompositions model a data tensor as a sum of rank-1 tensors, whose factors provide uses for a myriad of applications. Given the massive sizes of modern datasets, an important challenge is how well computational complexity scales with the data, balanced with how well decompositions approximate the data. Many efficient methods exploit a small subset of the tensor抯 elements, representing most of the tensor抯 variation via a basis over the subset. These methods� efficiencies are often due to their randomized natures; however, deterministic methods can provide better approximations, and can perform feature selection, highlighting a meaningful subset that well-represents the entire tensor. In this paper, we introduce an efficient subset-based form of the Tucker decomposition, by selecting coresets from the tensor modes such that the resulting core tensor can well-approximate the full tensor. Furthermore, our method enables a novel feature selection scheme unlike other methods for tensor data. We introduce methods for random and deterministic coresets, minimizing error via a measure of discrepancy between the coreset and full tensor. We perform the decompositions on simulated data, and perform on real-world fMRI data to demonstrate our method抯 feature selection ability. We demonstrate that compared with other similar decomposition methods, our methods can typically better approximate the tensor with comparably low computational complexities.
An Investigation of the Relationship Between Crime Rate and Police Compensation
(2024-11-21) Amarsingh, Jhancy; Appakondreddigari, Likhith Kumar Reddy; Nunna, Ashish; Tummala, Charishma Choudary; Winship, John; Zhou, Alex; Ashqar, Huthaifa
The goal of this paper is to assess whether there is any correlation between police salaries and crime rates. Using public data sources that contain Baltimore Crime Rates and Baltimore Police Department (BPD) salary information from 2011 to 2021, our research uses a variety of techniques to capture and measure any correlation between the two. Based on that correlation, the paper then uses established social theories to make recommendations on how this data can potentially be used by State Leadership. Our initial results show a negative correlation between salary/compensation levels and crime rates.

Browse

Recent Submissions