MD-SOAR Maryland Shared Open Access Repository
Not a member yet
34521 research outputs found
Sort by
Review: Getting Under Our Skin: The Cultural and Social History of Vermin, by Lisa T. Sarasohn
No full textThe inaugural issue of Animal History included a “Tracks in the Field” piece by Christopher Blakley that outlined the historiography of relationships among racism, enslavement, and human-animal studies. A comparable literature is accumulating that relates racism to creatures labeled as pests—some of it about the ecological and political conditions that expose racialized people to pests and pesticides, some of it about the very idea of pests as connected with abjection and disgust, and some interweaving the material and the discursive. Indeed, Blakley mentions some contributions to the slavery literature that address enslaved people’s role in pest control. Lisa Sarasohn’s book Getting Under Our Skin mostly focuses on human cultural representations of infestation with some of humankind’s most intimate other-than-human associates. The book shines important light on how “vermin” became associated with human hierarchy, dehumanization, and violent oppression.https://online.ucpress.edu/ah/article/doi/10.1525/ah.2024.110/21214
Sequentially Acquiring Concept Knowledge to Guide Continual Learning
No full textThe goal of continual learning (CL) is to adapt to new data (plasticity) while retaining the knowledge acquired from old data (stability). Existing methods focus on balancing stability and plasticity to mitigate the challenge of catastrophic forgetting while promoting learning. However, the impactof order and nature of new samples that the network is trained on remains an underexplored factor. A CL algorithm should ideally also have the ability to rank incoming samples in terms of their relationship with prior data and their effect on the learning process. In this work, we investigate if scoring and prioritizing incoming data based on their semantic relationships with the model’s current knowledge can boost CL performance. We propose SACK, short for Sequentially Acquiring Concept Knowledge, a scalable and model-agnostic two-step technique for continual learning. SACK dissects categorical knowledge of the model into fine-grained concepts, computes the relationships between previously learned concepts and new concepts in each experience, and uses this relationship knowledge for prioritizing new samples. Experiments across several types of CL methods (regularization, replay, and prompt-based) in class-incremental and task-incremental settings demonstrate that our approach consistently results in higher accuracy, reduces forgetting, and enhances plasticity. Code: https://github.com/abcxyz709/SACKSK was supported by the UMBC Cybersecurity Graduate Fellows program. TG was supported by the SURFF award from UMBC ORCA. Computing support was provided by UMBC HPCF. KT’s work was performed under the auspices of the U.S. Department of Energy by the Lawrence Livermore National Laboratory under Contract No. DE-AC52-07NA27344, Lawrence Livermore National Security, LLC and was supported by the LLNL-LDRD Program under Project No. 25-SI-001. The views and opinions of the authors expressed herein do not necessarily state or reflect those of the funding agencies and employers.https://openreview.net/forum?id=U4vcWks22
The Shadow of the Almighty: A Black Family’s Story of Liberation and the Mirage of The American Dream
No full textI Hate the News Jul 1
No full textThe weekly news analysis from I Hate Politics: Breaking down two of the most consequential Supreme Court rulings that came down last week: the ban on universal injunction and the expansion of religious rights of parents in public schools. The Trump Administration’s case against all 16 judges in the federal District of Maryland. The second Trump Administration has built a sophisticated legal team that seemed impossible in the first term. How? Lessons for Ranked Choice Voting from Zohran Mamdani’s victory in the New York Democratic primary. Sunil Dasgupta talks with Cato legal scholar and RCV advocate Walter Olson. Music by Washington DC art-pop rock band, Catscan!https://open.spotify.com/episode/4EO4rgU2jBoLeVqFp2tVF
Exoplanet Atmospheric Escape Observations with the Habitable Worlds Observatory
No full textTowards the Habitable Worlds Observatory: Visionary Science and Transformational Technology (HWO25), Washington DC, July 28-31, 2025. Endorsed by: Katherine Bennett (Johns Hopkins University), Abby Boehm (Cornell University), Aarynn Carter (Space Telescope Science Institute), Miguel Chavez Dagostino (National Institute of Astrophysics, Optics and Electronics), Caleb Harada (UC Berkeley), James Kirk (Imperial College London), Adam Langeveld (Johns Hopkins University), Eunjeong Lee (EisKosmos (CROASAEN), Inc.), Drew Miles (California Institute of Technology), Faraz Nasir Saleem (Egypt Space Agency), Gaetano Scandariato (INAF), Jessica Spake (Carnegie Observatories), Christopher Stark (NASA/GSFC), Antoine Strugarek (CEA Paris-Saclay, DAp-AIM), Johanna Teske (Carnegie Earth and Planets Laboratory), Daniel Valentine (University of Bristol), Austin Ware (Arizona State University), Peter Wheatley (University of Warwick)The Decadal Survey on Astronomy and Astrophysics 2020 highlights the importance of advancing research focused on discovering and characterizing habitable worlds. In line with this priority, our goal is to investigate how planetary systems evolve through atmospheric escape and to develop methods for identifying potentially Earth-like planets. By leveraging the ultraviolet (UV) capabilities of the Habitable Worlds Observatory (HWO), we can use transit spectroscopy to observe atmospheric escape in exoplanets and explore the processes that shape their evolution, assess the ability of small planets to retain their atmospheres, and search for signs of Earth-like atmospheres. To achieve this, we support the development of a UV spectrograph with moderate- to high-resolution capabilities for point-source observations, coverage of key spectral features in the 100-300 nm range, and detectors that can register high count rates reliably. This article is an adaptation of a science case document developed for the Characterizing Exoplanets Steering Committee within HWO's Solar Systems in Context Working Group.L.A.D.S. acknowledges the oftenoverlooked labor of the custodial, facilities, information technology and security staff at STScI – this research would not be possible without them. This research is based on observations made with the NASA/ESA Hubble Space Telescope. The data are openly available in the Mikulski Archive for Space Telescopes (MAST), which is maintained by the Space Telescope Science Institute (STScI). STScI is operated by the Association of Universities for Research in Astronomy, Inc. under NASA contract NAS 5-26555. This research made use of the NASA Exoplanet Archive, which is operated by the California Institute of Technology, under contract with the National Aeronautics and Space Administration under the Exoplanet Exploration Program. R.E. carried out research at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration (80NM0018D0004).http://arxiv.org/abs/2507.0712
A new high-resolution hydrodynamic model for the coastal Beaufort Sea in the Arctic Ocean: model construction and evaluation
No full textThe aquatic environment of the coastal Arctic is rapidly changing, and understanding how this change will affect the coastal ocean is critical across sectors. To address this, a three-dimensional (3-D) hydrodynamic model was constructed, spanning the coastal Beaufort Sea from −153° to −142° W, explicitly including river delta channels and lagoons, and extending to the continental shelf. The Finite Volume Community Ocean Model (FVCOM) was used to predict ocean physical properties from January 2018 to September 2022, including dynamic sea ice and landfast ice. Model calibration and validation were conducted using a variety of data sources, including in situ hydrodynamic data from oceanographic cruises and moorings. Overall, the model captured interannual temperature variation at Prudhoe Bay from 2018 to 2022 with a model efficiency (MEF) score > 0 (better than the average) for all years (MEF = 0.59, 0.63, 0.23, 0.46, and 0.55). The seasonal temperatures in 2018 and 2019 at bottom-mounted moorings were also well captured (R² = 0.80–0.90), and sea surface height (SSH) was compared to hourly observations at Prudhoe Bay, with both the low-frequency (R² = 0.42) and diurnal (R² = 0.71) variations validated over the model period. Modeled salinity and water current velocity had mixed results compared to the observations: seasonal trends in salinity were generally captured well, but hypersaline lagoon conditions in the winter were not replicated. Measured bottom water velocity proved difficult to recreate within the model for any given point in time from 2018 to 2019. Covariance analyses of the surface wind velocity, SSH, and current velocity indicated that wind forcing significantly correlated to errors in local SSH predictions. Current velocity covaried substantially less with SSH and wind velocity, with large differences across the three moorings: this suggests that local factors such as bathymetry and shielding by islands are likely important. Future work building on this system will include analyses of the drivers of landfast ice and sea ice breakup; the potential for erosion via waves, large storms, and elevated surface temperatures; and the linkage to an ecosystem model that represents processes from carbon cycling to higher trophic levels.The author(s) declare that financial support was received for the research and/or publication of this article. This research was supported by the National Aeronautics and Space Administration Ocean Biology and Biogeochemistry program #NNH20ZDA001NOBB. A portion of J. Shermans’ work was performed and funded under ST13301CQ0050/1332KP22FNEED0042.https://www.frontiersin.org/journals/marine-science/articles/10.3389/fmars.2025.1580690/ful
Southeast Asian nations look to hedge their way out of troubled waters in the South China Sea
No full textMalaysia, Vietnam and the Philippines have all attempted to counter Chinese claims to waters within their exclusive economic zones.John Rennie Short received funding from Fulbright Foundationhttp://theconversation.com/southeast-asian-nations-look-to-hedge-their-way-out-of-troubled-waters-in-the-south-china-sea-25709
Grok’s ‘white genocide’ responses show how generative AI can be weaponized
No full textThe tools that are meant to help make AI safer could actually make it much more dangerous.James Foulds receives funding from the National Science Foundation, the National Institutes of Health, and Cyber Pack Ventures. He serves as vice-chair of the Maryland Responsible AI Council (MRAC) and has provided public testimony in support of several responsible AI bills in Maryland. Shimei Pan receives funding from National Science Foundation (NSF), Defense Advanced Research Projects Agency (DARPA), US State Department Fulbright Program and Cyber Pack Ventureshttp://theconversation.com/groks-white-genocide-responses-show-how-generative-ai-can-be-weaponized-25788
GraphEdge: Dynamic Graph Partition and Task Scheduling for GNNs Computing in Edge Network
No full textWith the exponential growth of Internet of Things (IoT) devices, edge computing (EC) is gradually playing an important role in providing cost-effective services. However, existing approaches struggle to perform well in graph-structured scenarios where user data is correlated, such as traffic flow prediction and social relationship recommender systems. In particular, graph neural network (GNN)-based approaches lead to expensive server communication cost. To address this problem, we propose GraphEdge, an efficient GNN-based EC architecture. It considers the EC system of GNN tasks, where there are associations between users and it needs to take into account the task data of its neighbors when processing the tasks of a user. Specifically, the architecture first perceives the user topology and represents their data associations as a graph layout at each time step. Then the graph layout is optimized by calling our proposed hierarchical traversal graph cut algorithm (HiCut), which cuts the graph layout into multiple weakly associated subgraphs based on the aggregation characteristics of GNN, and the communication cost between different subgraphs during GNN inference is minimized. Finally, based on the optimized graph layout, our proposed deep reinforcement learning (DRL) based graph offloading algorithm (DRLGO) is executed to obtain the optimal offloading strategy for the tasks of users, the offloading strategy is subgraph-based, it tries to offload user tasks in a subgraph to the same edge server as possible while minimizing the task processing time and energy consumption of the EC system. Experimental results show the good effectiveness and dynamic adaptation of our proposed architecture and it also performs well even in dynamic scenarios.This work was supported in part by the National Natural Science Foundation of China 62462002 and partially supported by the Natural Science Foundation of Guangxi China Nos 2025GXNSFBA069394 2025GXNSFAA069958http://arxiv.org/abs/2504.1590
Q2E: Query-to-Event Decomposition for Zero-Shot Multilingual Text-to-Video Retrieval
No full textRecent approaches have shown impressive proficiency in extracting and leveraging parametric knowledge from Large-Language Models (LLMs) and Vision-Language Models (VLMs). In this work, we consider how we can improve the identification and retrieval of videos related to complex real-world events by automatically extracting latent parametric knowledge about those events. We present Q2E: a Query-to-Event decomposition method for zero-shot multilingual text-to-video retrieval, adaptable across datasets, domains, LLMs, or VLMs. Our approach demonstrates that we can enhance the understanding of otherwise overly simplified human queries by decomposing the query using the knowledge embedded in LLMs and VLMs. We additionally show how to apply our approach to both visual and speech-based inputs. To combine this varied multimodal knowledge, we adopt entropy-based fusion scoring for zero-shot fusion. Through evaluations on two diverse datasets and multiple retrieval metrics, we demonstrate that Q2E outperforms several state-of-the-art baselines. Our evaluation also shows that integrating audio information can significantly improve text-to-video retrieval. We have released code and data for future research.This material is based in part upon work supported by the National Science Foundation under Grant No. IIS-2024878. Some experiments were conducted on the UMBC HPCF, supported by the National Science Foundation under Grant No. CNS1920079. This material is also based on research that is in part supported by the Army Research Laboratory, Grant No. W911NF2120076, and by DARPA for the SciFy program under agreement number HR00112520301. The U.S. Government is authorized to reproduce and distribute reprints for Governmental purposes notwithstanding any copyright notation thereon. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either express or implied, of ARL, DARPA or the U.S. Government.http://arxiv.org/abs/2506.1020