Maryland Shared Open Access Repository

MD-SOAR is a shared digital repository platform for twelve colleges and universities in Maryland. It is currently funded by the University System of Maryland and Affiliated Institutions (USMAI) Library Consortium (usmai.org) and other participating partner institutions. MD-SOAR is jointly governed by all participating libraries, who have agreed to share policies and practices that are necessary and appropriate for the shared platform. Within this broad framework, each library provides customized repository services and collections that meet local institutional needs. Please follow the links below to learn more about each library's repository services and collections.

Browse

Submit

Submit item to MD-SOAR

Institutions in MD-SOAR

Select a community to browse its collections.

Bowie State University

Frostburg State University

Goucher College

Hood College

Loyola Notre Dame Library

Morgan State University

Salisbury University

Stevenson University

St. Mary's College of Maryland

Towson University

University of Baltimore

University of Maryland, Baltimore County (UMBC)

University of Maryland Eastern Shore

Recent Submissions

Beyond Math: Stories as a Testbed for Memorization-Constrained Reasoning in LLMs
(Association for Computational Linguistics, 2026-03) Jiang, Yuxuan; Ferraro, Francis
Memorization has been shown to greatly inflate Large Language Models' (LLMs) performance on domains such as math and logic, where success should primarily rely on applying generalizable reasoning rules. In many real-world applications, however, memorization is not meant to be eliminated but selectively constrained—for example, in story understanding, where background knowledge must be integrated with narrative context. Drawing on the cognitive science distinction between “verbatim” (exact recall) and “gist” (semantic abstraction) memorization, we propose a two-tier framework for analyzing how LLMs reason under different degrees of memory access. The Inductive (prompt-guided) Setting softly steers models to reason through selective, context-relevant recall, while the Restrictive Setting imposes stronger constraints by limiting verbatim memory access. Evaluating GPT-4o, LLaMA3.3-70B, and DeepSeek V3 on six character-centric story understanding benchmarks, we find up to a 45.2% accuracy drop under the Restrictive Setting, revealing strong dependence on surface recall. By contrast, the Inductive Setting maintains performance, indicating that prompting can align LLMs toward memorization-constrained reasoning.
GRB 241030A: a bright afterglow challenging forward shock emission
(2026-03-19) Ducoin, J.-G.; Pellouin, C.; Aivazyan, V.; Akl, D.; Alvarez, F.; Andrade, C.; Angulo, C.; Antier, S.; Atteia, J.-L.; Basa, S.; Becerra, R. L.; Benkhaldoun, Z.; Bissaldi, E.; Breeveld, A.; Bruin, E. de; Burns, E.; Butler, N. R.; Coughlin, M. W.; Daigne, F.; Dietrich, T.; Dornic, D.; Douzet, C.; Laz, T. du; Duverne, P.-A.; Eggenstein, H. B.; Elhosseiny, E.; Esamdin, A.; Evans, P. A.; Fernández, J. F. Agüí; Ferro, M.; Fortin, F.; Freeberg, M.; García-García, L.; Gill, R.; Globus, N.; Guessoum, N.; Hamed, G. M.; Hello, P.; Airasca, A. Holzmann; Hu, D. F.; Hussenot-Desenonges, T.; Inasaridze, R.; Iskandar, A.; Jiang, S. Q.; Jin, C. C.; Kaeouach, A.; Karpov, S.; Klingler, Noel; Klotz, A.; Kochiashvili, N.; Koehn, H.; Kneip, R.; Kvernadze, T.; Calloch, A. Le; Lee, W. H.; Lekic, A.; Liang, Y. F.; Limonta, C.; Liu, J.; López, K. Ocelotl C.; López-Cámara, D.; Mabrouk, R. H.; Magnani, F.; Mao, J.; Mašek, M.; Méndez, E. Moreno; Mihov, B. M.; Molham, M.; Noysena, K.; Odeh, M.; Omodei, N.; Peng, H.; Pereyra, M.; Pillas, M.; Pillera, R.; Pradier, T.; Rajabov, Y.; Rakotondrainibe, N. A.; Schneider, B.; Serrau, M.; Slavcheva-Mihova, L.; Sokoliuk, O.; Sun, H.; Takey, A.; Tanasan, M.; Tinyanont, K. S.; Turpin, D.; Postigo, A. de Ugarte; Wang, B. T.; Wang, L. T.; Wang, X. F.; Wang, Z. M.; Watson, A. M.; Wu, H. Z.; Wu, Q. Y.; Xu, J. J.; Yan, Y. S.; Yang, H. N.; Yuan, W.; Zhao, H. S.
Gamma-Ray Burst GRB 241030A (z = 1.411) exhibited a bright afterglow (similar to GRB 221009A), detected across gamma-ray, X-ray, UV, and optical bands, providing a probe of GRB afterglow physics. We compiled multi-wavelength observations spanning from a minute to a week after the prompt emission, processing the data through a unified photometry pipeline. We analysed the observations both analytically and using Bayesian inference with two independent models. Our models assume that the afterglow emission arises from the strong forward shock of a laterally structured jet, with possible contributions from synchrotron self-Compton (SSC) scatterings. Our models reproduce X-ray to optical data, favouring a jet propagating into a constant-density interstellar medium, with a viewing angle within the jet core. However, both analyses require parameter values that are extreme compared to expectations from standard theory. In particular, our results imply extremely energetic jets despite regular prompt energy, leading to a very inefficient prompt emission. Furthermore, the jets are inefficient at accelerating particles, with low electron and magnetic energy fractions, leading to significant SSC emission. Our analyses indicate that the jets have large opening angles and propagate in high-density media. If the afterglow is indeed powered by radiation emitted behind a strong forward shock, our results place GRB 241030A within a sub-class of GRBs characterised by extreme kinetic energies, large jet opening angles, and very low prompt emission efficiencies, with strong SSC radiation. These predictions are difficult to reconcile with typical expectations from other GRBs. We therefore suggest that the afterglow of GRB 241030A is not solely powered by forward shock emission.
Language identity and sense of belonging among 1.5-generation Asian and Asian American immigrant young adults in the United States
(APA, 2026-03-19) Park, Chulwoo; Bang, Janet Y.; Edberg, Mark
Language identity has been defined in the research as an individual’s connection to their sense of self using language with others. Understanding language identity construction in a new environment among Asian immigrants, a racial/ethnic minority population, is important for addressing health disparities and inequities. This study examined how language identity could be defined by 1.5-generation Asian and Asian American immigrant young adults and how using multiple languages influenced their sense of belonging in the United States. Additionally, we investigated how languages influenced the way they interact with others and see themselves and what language use contexts and characteristics helped them establish their language identity. Participants were defined as individuals who arrived in the United States from Asia with their first-generation parents when they were 5–17 years old; had lived in the United States for at least 12 months; were residing in the San Francisco Bay Area, California; and were aged 18–29 at the time of the study. We conducted eight focus group discussions with 24 participants (two to four participants in each group) and analyzed verbatim transcriptions using Dedoose. We report on three themes identified in the analysis: (a) languages played a key role in forming personal and professional relationships that impacted identities, (b) language identity was distinct from language proficiency, and (c) multilingualism shaped their ethno-racial identity and sense of belonging. Future research will pursue one-on-one in-depth interviews and longitudinal studies to subsequently understand individualized experiences and expand the scope of the target population to provide generalizability to other 1.5-generation Asian and Asian American immigrants.
PUF-based Robust Over-the-Air Software Update Protocol for IoT Devices
(IEEE, 2026-03-19) Sanjana Mehjabin, Suhee; Younis, Mohamed
Secure software/firmware updates continue to be a critical challenge, particularly for resource-constrained IoT devices, where traditional cryptographic mechanisms are deemed unsuitable given their significant computational and energy overhead. This article promotes PROUD, a novel approach that is both robust and lightweight. Specifically, hardware-embedded Physical Unclonable Functions (PUFs) are leveraged for device-specific authentication and software/firmware integrity verification. PROUD employs PUFs as tamper-proof identifiers, eliminating the vulnerabilities of conventional key storage-based methods. The design also incorporates a streamlined verification process that validates the authenticity of the update during over-the-air transfers, ensuring resilience against unauthorized modifications. Prototype implementation on a Xilinx Artix-7 FPGA shows that PROUD imposes minimal overhead, in terms of logic (790 LUT, 373 FF), latency (1.4ms) and power (6mW). PROUD marks a step towards more autonomous, hardware-rooted trust models, offering a scalable and secure path for Software/firmware updates in IoT ecosystems.
Unmasking Algorithmic Bias in Predictive Policing: A GAN-Based Simulation Framework with Multi-City Temporal Analysis
(2026-03-20) Barman, Pronob Kumar; Barman, Pronoy Kumar
Predictive policing systems that direct patrol resources based on algorithmically generated crime forecasts have been widely deployed across US cities, yet their tendency to encode and amplify racial disparities remains poorly understood in quantitative terms. We present a reproducible simulation framework that couples a Generative Adversarial Network GAN with a Noisy OR patrol detection model to measure how racial bias propagates through the full enforcement pipeline from crime occurrence to police contact. Using 145000 plus Part 1 crime records from Baltimore 2017 to 2019 and 233000 plus records from Chicago 2022, augmented with US Census ACS demographic data, we compute four monthly bias metrics across 264 city year mode observations: the Disparate Impact Ratio DIR, Demographic Parity Gap, Gini Coefficient, and a composite Bias Amplification Score. Our experiments reveal extreme and year variant bias in Baltimores detected mode, with mean annual DIR up to 15714 in 2019, moderate under detection of Black residents in Chicago DIR equals 0.22, and persistent Gini coefficients of 0.43 to 0.62 across all conditions. We further demonstrate that a Conditional Tabular GAN CTGAN debiasing approach partially redistributes detection rates but cannot eliminate structural disparity without accompanying policy intervention. Socioeconomic regression analysis confirms strong correlations between neighborhood racial composition and detection likelihood Pearson r equals 0.83 for percent White and r equals negative 0.81 for percent Black. A sensitivity analysis over patrol radius, officer count, and citizen reporting probability reveals that outcomes are most sensitive to officer deployment levels. The code and data are publicly available at this repository.