Mason Journals (George Mason Univ.)
Not a member yet
3256 research outputs found
Sort by
An Analysis of Generative AI Agents in Automatic Task Completion in Interactive Web Environments
No full textAutonomous AI agents are software programs that can independently complete multi-step tasks given a prompt (e.g. “Tell me the full address of the nearest airport to George Mason University”). They are expected to complete mundane daily tasks that do not require human input in the near future. Autonomous agents have become more sophisticated in recent years owing to advancements in machine learning and natural language processing. However, these agents are still susceptible to mistakes in comprehension, reasoning, and action execution. WebArena is a standalone web environment that can host autonomous agents. In this project, we conducted an analysis of WebArena and its provided agent using the procedural guidelines and task prompts provided by the WebArena team. We ran WebArena using 100 unique prompts ranging from shopping and store management to mapping and navigation. We categorized errors and discovered which errors occurred most frequently. The autonomous agent successfully completed the task in 23% of runs. Many runs suffered from an incorrect, missing, or unnecessary action. Common reasons for errors included a lack of world knowledge or human-like visualization, ignoring part of a prompt, and hallucinations. In the future, training should ensure that autonomous agents have sufficient world knowledge and experience to navigate the web successfully. Developers can also add weights to prompts so that agents can focus more on relevant criteria and introduce memory so that agents can learn from past actions. Hopefully, with more time and research, autonomous agents will soon improve the lives of many
Training Deep Learning Models on Virtual Reality Spatial Motion Data for Implicit User Identification
No full textVirtual reality (VR) is a rapidly growing technology that attracts millions of users. Implicit user identification has many applications in VR, such as removing the possibility for external threats to observe password entry for authentication or maintaining the high degree of immersion that a smooth VR experience requires. Past studies have utilized behavioral biometrics to implicitly identify users in VR by distinguishing patterns between users’ movements. An existing study by Liebers et al. conducted a user study to record spatial motion data and trained two deep learning models to classify user data, reaching an identification accuracy of up to 90%. It remains to be seen if the findings reported are robust upon replication or if the results are generalizable to other deep learning models. As such, we aim to both replicate the data analysis portion of the study as reported and expand it to include two additional deep learning models. Using the dataset published by Liebers et al., we train MLP, KNN, SVM, and LSTM models across four feature sets, two task scenarios, and four types of body normalization, and evaluate the accuracies of the models. Preliminary data show discrepancies between evaluated accuracies and reported accuracies in Liebers et al. For example, the MLP model reached up to a difference of 25 accuracy points or 79% error. These discrepancies highlight the challenges faced when replicating studies without source code. Additional analysis is needed to evaluate the generalizability of the study to other deep learning models
Detection Tools in Smart Contracts: Reevaluating the Effectiveness of Reentrant Detection Methods Through Replication
No full textSmart contracts are programs that facilitate financial exchanges through cryptocurrency, making it crucial that they are secure and resistant to exploitation. One prominent vulnerability in smart contracts is reentrancy, which has led to the theft of millions of dollars in cryptocurrency. To address this issue, many reentrancy detection tools have been developed and tested over the years. This study aims to validate the findings of a previous study by Zheng, et. al, which concluded that five widely used detection tools were largely ineffective at identifying reentrancy vulnerabilities. The tools examined were Oyente, Mythril, Sailfish, Smartian, and Securify. To replicate the original study, we set up and ran each detection tool within Docker containers. While three of the detection tools have Docker images to pull straight from Docker hub, the other two requires us to build our own images which comes with some issues in version compatibility. We then analyzed 139,424 smart contracts pulled from the original study's git repository of test data by creating python scripts to feed the data into each detection tool. The original study found that 99.8% of the detected vulnerabilities were false positives. We anticipate that our results will align with Zheng et. al's findings, although there may be some improvement in detection accuracy due to updates to Mythril since the initial study. Our findings will highlight the ineffectiveness of modern tools in detecting reentrancy vulnerabilities in smart contracts
Goal Attainment and Quality of Life through Inclusive College: Three Years of Progress
No full textWith growing opportunity for students with intellectual and developmental disability to access a variety of inclusive higher education programs comes an increased need for program implementers to evaluate practices and outcomes alongside participants. This mixed method, exploratory study examines self-determined goal setting, goal attainment, and quality of life within an inclusive college program as a measure of participant outcome and program evaluation. Furthermore, it provides implications for the importance of self-determined learning and participant voice within program planning, revision, and implementation.
 
Using Video-Based Instruction to Increase Employment-Related Social Behaviors for College Students with Intellectual and Developmental Disabilities
No full textPoor social skills is a leading factor why individuals with intellectual and developmental disabilities (IDD) lose their jobs. Fortunately, the use of technology has made learning and teaching social skills more seamless and integrated in employment contexts. We conducted a multiple-probe-across-participants single-case experimental design study to evaluate the effects of video-based instruction on the employment-related social behaviors of three college students with IDD enrolled in a comprehensive transition program at a large public university. Results indicated small to moderate effect sizes for all three students. Participants found the intervention to be helpful in improving their employment readiness skills. We discuss implications for research and practical ways technology can be used to support college students with IDD to strengthen their employment-related social behaviors
Navigating Risk? Enrollment in Inclusive Post-Secondary Education During COVID-19
No full textThis study explores challenges faced by young adults with intellectual and developmental disabilities (IDD) and their families enrolling in inclusive post-secondary education (IPSE) programs during COVID-19. The benefits of attending IPSE programs are well-documented, but this group is disadvantaged accessing post-secondary education and employment. The heightened risk of COVID-19 for people with IDD further complicates decision-making. Through interviews with 11 students with IDD and 10 parents, the study explores decisions about enrolling in IPSE, highlighting the importance of access to alternative options, expectations during the pandemic, and the ability of IPSE programs to adapt to future challenges, notably online options
Jonathan Reynolds, Sovereignty and Struggle: Africa and Africans in the Era of the Cold War, 1945–1994
No full textMolecular Profiling of Stromal Components in Prostate Cancer: Leveraging LCM and RPPA to examine Clinical Outcomes
No full textIntroduction:
Prostate cancer remains a significant health challenge, necessitating advanced methodologies for accurate diagnosis and understanding of its molecular makeup. The tumor microenvironment, particularly the stroma, plays a critical role in cancer progression and metastasis. In this study, we focus on characterizing the activation of the immune and DNA damage repair signaling pathways within the stromal compartment of prostate cancer tissue from a patient undergoing treatment with Pembrolizumab. By analyzing biopsies obtained at various stages of the treatment plan, we investigate the protein signaling changes in the microenvironment throughout the course of treatment.
Methodology:
We utilized Laser Capture Microdissection (LCM) technology to isolate stroma cells from prostate cancer tissue samples. A total of nine biopsies were obtained at various stages of the treatment plan. Following microdissection, the collected cells underwent analysis using Reverse Phase Protein Microarray (RPPA) technology, which allowed for the quantification of protein expression levels and the identification of signaling pathway alterations. Data from RPPA were analyzed using MicroVigene software, which facilitated data processing and interpretation. These methodologies collectively enabled an investigation of the stroma signaling pathways in prostate cancer, potentially identifying biomarkers for improved diagnosis.
Conclusion:
Our analysis of the stroma cells revealed several markers exhibiting elevated activation levels, including FoxP3, MSH2, mTOR S2448, Histone H2A.X S139, B7-H4, CD3zeta, HLA-DR, HLA-DR-DP-DQ-DX, Granzyme B, and CD31-PECAM1. These elevated values may indicate higher activity from the patient’s autoimmune response. Conversely, the markers Chk1_S345, STAT1_Y701, and PDL122C3 were not elevated, implying lower activity levels. Further research is needed to confirm our results