Digital Commons Network^™

The Promises And Perils Of Using Llms For Effective Public Services, Erina Seh-Young Moon, Matthew Tamura, Angelina Zhai, Nuzaira Habib, Behnaz Shirazi, Altaf Kassam, Devansh Saxena, Shion Guha

Health Services and Informatics Research

Governments are the primary providers of essential public services and are responsible for delivering them effectively. In high-stakes decision-making domains such as child welfare (CW), agencies must protect children without unnecessarily prolonging a family’s engagement with the system. With growing optimism around AI, governments are pushing for its integration but concerns regarding feasibility and harms remain. Through collaborations with a large Canadian CW agency, we examined how LocalLLM and BERTopic models can track CW case progress. We demonstrate how the tools can potentially assist workers in opportunistically addressing gaps in their work by signaling case progress/deviations. And yet, we also …

Characterizing User-Reported Risks Across Llm Chatbots, Lingyao Li, Renkai Ma, Zhaoqian Xue, Junjie Xiong

Computer Science Faculty Research & Creative Works

As Large Language Models (LLMs) become increasingly integral to daily life, users are engaging with multiple LLM chatbots for various needs; however, prior research on LLM risks often remains lab-based or focuses on single LLMs like ChatGPT or singular risks like privacy. To gain a multi-risk, cross-chatbot understanding of user experiences, we analyze Reddit discussions around seven major LLM chatbots using the NIST AI Risk Management Framework. We find that user-reported risks are unevenly distributed and chatbot-specific: ChatGPT is associated with safety and fairness concerns, Gemini with privacy, and Claude with security and resilience. Less frequent risks, such as explainability …

Pre-Experiential Constraint Reconstruction: A Structural Account Of The Prior Layer In Dialogue With Jung, Griselda Poe

Publications and Research

Experiences commonly described as pre-experiential memory—such as immediate recognition, familiarity without prior exposure, and the sense of "already knowing"—are typically interpreted as the retrieval of stored content. However, this storage-based account does not provide a structurally consistent explanation of how such content is preserved prior to experience or reactivated in a form that aligns with present input.

This paper extends the three-layer cognitive architecture developed in prior work in this series (Poe, 2026n), in which cognition operates across Core processing, Modulation, and an operationally inaccessible Prior layer that supplies constraints to all processing.

This paper proposes an alternative account in …

Does Culture Translate? A Case Study On Artificial Intelligencegenerated Text Concerning Japanese-Style Employment, Kanji Kitamura

Languages, Literatures, and Cultures Faculty Publications

Translation matters to international business when it has a discernible impact on firm performance or critical business processes. Current developments include machine translation, which is evolving in tandem with generative artificial intelligence (AI). On one hand, useful translation apps have become available to help the public communicate internationally. On the other hand, recent research concludes that even machine translation's best output still requires post-editing by a human. This argument implies that a machine-translated, grammatically accurate text may not always make sense to a human audience, possibly because of a semantic-pragmatic issue beyond the lexical level. This paper regards it as …

The Core-Modulation Architecture (Cma): A Structural Overview Of Hallucination As Structural Mismatch, Griselda Poe

Publications and Research

Hallucination is defined not as factual error but as a structural failure of alignment across target, layer, and constraint.

Within the Core-Modulation Architecture (CMA), cognition proceeds through layered processing and requires layer-specific termination conditions. Hallucination arises when Modulation-level termination is registered as completion while Core-level resolution has not occurred, producing structurally ungrounded but locally coherent outputs.

Detection is therefore structural rather than content-based, focusing on layer mismatch and termination failure.

This document presents a minimal structural account of hallucination within the CMA framework.

Advice For Incorporating Ai Tools Into Your Legal Practice, Celia Bigoness, Robert A. Mackenzie, David J. Reiss

Cornell Law Faculty Publications

We have been speaking with many lawyers and law students about using generative artificial intelligence (AI) tools in their legal practice. We are struck by the fact that many of them have not been experimenting much, if at all, with the tools that are available to them - although many acknowledge that their clients are increasingly integrating generative AI into their businesses. We have been integrating a lot of these tools into our own professional lives, and here are some tips to help lawyers and law students get comfortable with AI tools that can help them, in big ways and …

Aiw26s: Applied Llms, Chengjie Zheng

Paul English Applied Artificial Intelligence (AI) Institute Publications

This workshop introduces the concept of applied large language models (LLMs), focusing on how users can move from simple prompt-based interaction to building structured, repeatable AI-driven workflows. Participants explore how AI enables faster prototyping, lowers barriers to entry, and expands who can participate in building technology. Through a hands-on demonstration, attendees learn how to transform raw inputs into meaningful outputs such as summaries, key concepts, and actionable steps. The session emphasizes the importance of clear problem definition, iterative refinement, and critical evaluation when working with AI systems.

Ai Method For Classification Of Diagnosis Of Near-Infrared Breast Lesion Images, Kaiquan Chen, Fangyang Shen, Honggang Wang, Zhengchao Dong, Jizhong Xiao, Ming Ma, Afroza Aktar, Christopher Chow, Wenxiong Zhang

Publications and Research

In near-infrared optical breast lesion screening and diagnosis systems, high-speed four-dimensional scanners can dynamically acquire tens of thousands of lesion images within a five-minute period. Currently, manual computer annotation is required to generate standard samples from these scanned breast lesion images, a process that depends heavily on physicians with clinical expertise. On average, a single physician can annotate only approximately ten samples per working day. As a result, this process is time-consuming and labor-intensive, and the collected samples often suffer from low accuracy, large variability, and limited diagnostic reliability. Several AI-based annotation tools, such as QuPath, HALO AI™, and X-AnyLabeling, …

Innovations And Applications Of Virtual Private Networks And Sustainable Security In Society 5.0 Libraries, Stella Chinnaya Nduka Dr., Adeyinka Tella Prof, Petros Dlamini Dr

Journal of Cybersecurity Education, Research and Practice

In order to improve digital resilience, privacy, and access equity in contemporary library environments, this study investigates the role of Virtual Private Networks (VPNs) in fostering sustainable cybersecurity within the framework of Society 5.0 libraries. It does this by looking at the latest developments, applications, difficulties, moral dilemmas, and tactical methods associated with VPN deployment. Using peer-reviewed journal articles, conference proceedings, white papers, and policy documents published between 2010 and 2024, a methodical approach to literature review was used. The literature that bridges the fields of cybersecurity, library science, and Society 5.0 concepts was the main focus of the review. …

Spatial Future Ahead! Augmented Reality And Anticipated Life Consequences, Sergio Barta, Reto Felix, Chris Hinsch, Mahdokht Kalantari, Nina Krey, Philipp A. Rauschnabel

Marketing Faculty Publications

Purpose: This study explores how initial exposure to immersive spatial computing experiences using AR headsets generates lasting inspiration and shapes consumers expected long-term life consequences (i.e., enhancement of reality, perceived substitutability and social impact).

Design/methodology/approach: The study uses a time-lagged research design based on 148 first-time users of spatial computing devices (AR headsets). Respondents were interviewed once shortly after being exposed to AR and a few days later. Data is analyzed using partial least squares structural equation modeling (PLS-SEM).

Findings: Users' immediate “inspired-by” experiences predict increased “inspired-to” intentions days later. Such inspiration translates into anticipated consequences such as virtually customizing …

Adaptive Parallel Downloader For Large Genomic Datasets, Rasman Mubtasim Swargo

Miners Solving for Tomorrow Research Conference

Modern next-generation sequencing (NGS) projects routinely generate terabytes of data that researchers download from public repositories such as SRA and ENA. Existing download tools typically employ static concurrency settings, leading to inefficient bandwidth utilization and prolonged download times under dynamic network conditions. We introduce FastBioDL, a parallel downloader for large biological datasets with an adaptive concurrency controller. FastBioDL frames downloading as an online optimization problem, using a utility function and gradient descent to adjust the number of concurrent socket streams during runtime. This approach maximizes throughput while minimizing resource overhead. Evaluations on public genomic datasets show that FastBioDL achieves up …

Introducing Gridtrees For Streaming, Scalable Hierarchical Data Visualization, Nathan Tibbetts

Miners Solving for Tomorrow Research Conference

File browsing in the consumer sphere has seen very little advancement in recent years, although attempts have been made to improve upon it. With the goal of multi-level visual file browsing in mind, we present a prototype GridTree, a dynamic, recursive, stable, spatial layout data structure based on subdividing grids, represented as a hierarchy of maps whose granularity increases with depth. We motivate this work with characteristics we have identified as important for viability of a multi-level file-browser, which have become our design goals. We touch on the underlying logic of a GridTree and identify its complexity. We briefly discuss …

Thermal Mirage: Towards Robust Thermal Perception Via Gan-Guided Diffusion, Nuzaer Omar

Miners Solving for Tomorrow Research Conference

Thermal object detection systems are critical for safety sensitive applications due to their reliability under adverse conditions. However, existing robustness evaluations in thermal domain primarily focuses on physical or sensor-level perturbations, overlooking vulnerabilities from semantically realistic scene and object manipulations. We introduce Thermal Mirage, a generative framework that leverages GAN-guided diffusion to expose weaknesses in thermal detectors through controlled object and context level perturbations. Our approach learns class-conditional thermal priors via a GAN and uses diffusion to transform object appearances into ambiguous or low-saliency patterns. Simultaneously, a context module degrades scene conditions by simulating harsher night environments. Integrated with a …

Robust Federated Learning With Strategic Adversaries, Manoj Twarakavi

Miners Solving for Tomorrow Research Conference

Federated Learning (FL) leverages the intelligence of untrusted distributed devices through collaborative training. This makes the training process susceptible to malicious behavior. Existing defense mechanisms largely consider an adversary who attacks a proactive FL server without any adaptability. However, they overlook the presence of a strategic adversary. To address this challenge, our work proposes a Robust Game-theoretic framework where the adversary is both strategic and is equipped with the capability of performing large-scale poisoning attacks.

Distributed Control Plane For Cross-Silo Federated Learning, Rabin Pandey

Miners Solving for Tomorrow Research Conference

Cross-silo federated learning (FL) trains shared models across geographically distributed institutions without centralizing raw data, but its reliance on wide-area networks makes round completion time highly sensitive to heterogeneous link conditions, congestion, and straggler clients. Existing SDN-based FL frameworks address this through centralized traffic engineering, an approach that breaks down when silos span independent administrative domains where no single entity can maintain complete, timely network knowledge. We replace the centralized control plane with a fully distributed overlay in which each silo gateway operates as an equal peer, continuously probing local links, exchanging EWMA-smoothed metrics via bounded gossip, and computing least-cost …

Robustness Of Fuzzy Artmap To Adversarial Attacks And Progressive Adversarial Training For Streaming Learning, Shane Cairns

Miners Solving for Tomorrow Research Conference

Incremental learners deployed on streaming data must remain robust to evolving adversarial perturbations, yet most adversarial-robustness studies assume offline multi-epoch training with repeated access to historical data. We investigate adversarial robustness in Fuzzy ARTMAP, a prototype-based Adaptive Resonance Theory model that supports single-pass learning without replay. We propose WB-Softmax, a differentiable relaxation that aggregates category-level activations into class-level scores for gradient-based attacks. WB-Softmax PGD achieves 89–100% attack success on vanilla models, exceeding transfer and query-based baselines. We then study adversarial training under true streaming constraints by comparing offline versus online adversarial example generation and standard versus selective updates. Offline adversarial …

Elucidating Complex H Behavior In Amorphous Oxide Semiconductors By Machine Learning, Lucas Ethington

Miners Solving for Tomorrow Research Conference

In this project, the disordered oxide structures will be mapped by a machine-learning algorithm (e.g., HDBSCAN) to identify characteristic behaviors of the proton across different material densities and defect types. This should help identify the under-coordinated, highly distorted, weakly-bonded, and dynamically unstable atoms to predict the most probable H locations. This fast and accurate prediction of energetically favorable H distribution will enable a reliable and fast screening of a large number of AOSs with variable cation and/or anion compositions. The approach will help find AOSs with suppressed numbers of M-OH defects (that form deep electron traps, limiting the number of …

Wheel-Spoke Encoding: A Product Quantization-Compatible Radial Encoding Scheme For Convex Polygonal Data, Maris Reinkemeyer

Miners Solving for Tomorrow Research Conference

Product quantization has historically been unapplied to GIS datasets, likely due to a mismatch between quantization’s input precondition of fixed, equal length vectors and GIS data’s inherent variability in the number of data points. Therefore, any quantization-compatible encoding method must operate independently of the data points within GIS records. The challenge is to balance the guarantee of producing fixed, equal length vectors with the preservation of semantic meaning present in the raw data. Here, we develop a radial polygon encoding method that achieves this balance while providing acceptable recall in a time complexity of O(nrv), where n is the number …

Dynamical Transition From A Two-Dimensional Soliton To A Rogue Wave In Quantum Droplets, Punit Sesha Sai Turlapati

Miners Solving for Tomorrow Research Conference

We investigate the nonequilibrium dynamics of two-dimensional quantum droplets: ultracold self-bound many-body states stabilized by the interplay of mean-field attractive interactions and repulsive quantum fluctuations. Flat-top ground state droplets are subject to an external potential, an attractive well and a repulsive barrier. Under the influence of the attractive well, we observe signatures of a Townes soliton formation, which for increasing strength of the well transitions into a two-dimensional rogue wave structure, a time-periodic highly localized configuration with amplitude three times larger than the background. The barrier instead favors a dynamical splitting of the droplet. We have developed a parallelized simulation …

Lay Summarization For Medical Patents, Manav Raja Vinotha

Miners Solving for Tomorrow Research Conference

Technical documents, such as medical patents, are often inaccessible to non-expert audiences due to specialized terminology. This paper presents a multi-agent system for automated lay summarization that optimizes both quality and computational efficiency through strategic task decomposition. The proposed pipeline utilizes four specialized stages: autonomous medical entity retrieval from the Unified Medical Language System (UMLS), context distillation into lay conceptualizations, and a structured writer-critic loop for iterative refinement. By distributing cognitive load across specialized agents, this architecture effectively leverages smaller, cost-effective language models while maintaining the performance of SOTA LLMs. Evaluated against single-agent baselines using an LLM as a Judge …

Analysis Of Autonomous Vehicle Survivability To 5g Communication Errors, Sydney Clark

Miners Solving for Tomorrow Research Conference

Autonomous vehicles rely on low-latency, high-reliability data exchange for real-time perception and control. Disruptions such as packet loss, latency variation, protocol-level errors, and malicious interference can pose significant safety risks to both passengers and surrounding environments. This project aims to evaluate, quantify, and predict the survivability of autonomous vehicle systems to communication errors, with focus on 5G network environments. The impact of these communication impairments on vehicle stability and control will be investigated through high-fidelity cyber-physical simulation of the vehicle and its surrounding environment. Experiments designed to capture varying network conditions will be used to assess a broad range of …

A Wearable Mxene-Based Sweat Sensor For Real-Time Monitoring Of Inflammatory Biomarkers, Ariel Pilger

Miners Solving for Tomorrow Research Conference

Many conventional biosensing approaches rely on invasive sampling or bulky benchtop instrumentation, limiting their use in continuous and portable applications. This project focuses on the development of wearable sweat-based biosensors that enable non-invasive, continuous, and portable monitoring of physical, chemical, and biological markers. The system will be designed to target markers present in sweat and transduce the biochemical interactions into measurable electrical signals. These signals will be processed through integrated electronics to produce clear, interpretable outputs for users and medical professionals. Supporting circuitry including filters, amplifiers, and an independent power supply will be implemented as necessary to ensure signal accuracy, …

A.I.R.E., Laurene Robinson

Presentations - 2026

•Cybersecurity analysts rely on reverse engineering to understand suspicious software. •Ghidra can surface decompiled code, but it does not fully explain function purpose, behavioral meaning, or analyst priority. •When symbols are stripped and context is weak, analysts must still reconstruct intent manually from low-level output. •That process is Time-consuming , complex and , operationally costly

Be Responsible In Your Answers! Monitoring Out-Of-Domain Behaviors In Domain-Specific Llms, Boquan Li, Chenzhe Lou, Zhe Ren, Peixin Zhang, Zirui Fu, Jun Sun, Yaowen Zheng

Research Collection School Of Computing and Information Systems

Large Language Models (LLMs) have accelerated the rapid development of chatbot web applications in various domains, such as coding, biomedicine and psychology. Compared to general LLMs like ChatGPT, domain-specific LLMs require a greater sense of responsibility. For instance, if a programming LLM casually answers medical or psychological questions, it not only misleads the public but also poses legal risks. This highlights new demands for monitoring and preventing such irresponsible behaviors. Existing efforts attempt to monitor LLMs from multiple aspects, such as lying, jailbreaks, and toxic content, while overlooking out-of-domain behaviors. In this work, we propose an innovative LLM domain monitoring …

Architecture-Agnostic Test-Time Adaptation Via Backprop-Free Embedding Alignment, Xiao Ma, Young D. Kwon, Pan Zhou, Dong Ma

PhD Student’s Publications Collection

Test-Time Adaptation (TTA) adapts a deployed model during online inference to mitigate the impact of domain shift. While achieving strong accuracy, most existing methods rely on backpropagation, which is memory and computation intensive, making them unsuitable for resource-constrained devices. Recent attempts to reduce this overhead often suffer from high latency or are tied to specific architectures such as ViT-only or CNN-only. In this work, we revisit domain shift from an embedding perspective. Our analysis reveals that domain shift induces three distinct structural changes in the embedding space: translation (mean shift), scaling (variance shift), and rotation (covariance shift). Based on this …

Scalable Multi-Task Low-Rank Model Adaptation, Zichen Tian, Antoine Ledent, Qianru Sun

PhD Student’s Publications Collection

Scaling multi-task low-rank adaptation (LoRA) to a large number of tasks induces catastrophic performance degradation, such as an accuracy drop from 88.2% to 2.0% on DOTA when scaling from 5 to 15 tasks. This failure is due to parameter and representation misalignment. We find that existing solutions, like regularization and dynamic routing, fail at scale because they are constrained by a fundamental trade-off: strengthening regularization to reduce inter-task conflict inadvertently suppresses the essential feature discrimination required for effective routing. In this work, we identify two root causes for this trade-off. First, uniform regularization disrupts inter-task knowledge sharing: shared underlying knowledge …

Semantic Entanglement In Vector-Based Retrieval: A Formal Framework And Context-Conditioned Disentanglement Pipeline For Agentic Rag Systems, Nick Loghmani

iSchool - All Scholarship

Retrieval-Augmented Generation (RAG) systems deployed in agentic environments depend on the geometric properties of vector representations to retrieve contextually appropriate evidence for autonomous reasoning. When source documents conflate multiple topics within contiguous text regions, standard vectorization pipelines produce embedding spaces in which semantically distinct content occupies overlapping geometric neighborhoods — a condition we term semantic entanglement. This paper formalizes semantic entanglement as a model-relative measure of cross-topic overlap, defines an Entanglement Index (EI) as a quantitative proxy, and argues that higher EI is associated with reduced attainable Top-K retrieval precision under cosine similarity retrieval. We introduce the Semantic Disentanglement …

Penforge: On-The-Fly Expert Agent Construction For Automated Penetration Testing, Huihui Huang, Jieke Shi, Junkai Chen, Ting Zhang, Yikun Li, Chengran Yang, Eng Lieh Ouh, Lwin Khin Shar, David Lo

Research Collection School Of Computing and Information Systems

Penetration testing is essential for identifying vulnerabilities in web applications before real adversaries can exploit them. Recent work has explored automating this process with Large Language Model (LLM)-powered agents, but existing approaches either rely on a single generic agent that struggles in complex scenarios or narrowly specialized agents that cannot adapt to diverse vulnerability types. We therefore introduce PenForge, a framework that dynamically constructs expert agents during testing rather than relying on those prepared beforehand. By integrating automated reconnaissance of potential attack surfaces with agents instantiated on the fly for context-aware exploitation, PenForge achieves a 30.0% exploit success rate (12/40) …

Stacked From One: Multi-Scale Self-Injection For Context Window Extension, Wei Han, Pan Zhou, Shuicheng Yan

Research Collection School Of Computing and Information Systems

The limited context window of contemporary large language models (LLMs) remains a primary bottleneck for their broader application across diverse domains. Although continual pre-training on long-context data offers a straightforward solution, it incurs prohibitive data acquisition and computational costs. To address this challenge, we propose SHAREDLLM, a novel framework based on multi-grained context compression and query-aware information acquisition. SHAREDLLM comprises two stacked short-context LLMs: a lower model serving as a compressor and an upper model acting as a decoder. The lower model compresses long inputs into compact, multi-grained representations, which are then forwarded to the upper model for context-aware processing. …

Challenges In Synchronous And Remote Collaboration Around Visualization, Matthew Brehmer, Maxime Cordeil, Christophe Hurter, Takayuki Itoh, Wolfgang Büschel, Mahmood Jasim, Arnaud Prouzeau, David Saffo, Lyn Bartram, Sheelagh Carpendale, Chen Zhu-Tian, Andrew Cunningham, Anthony Tang, Samuel Huron, Masahiko Itoh, Arpit Joshi, Kiyoshi Kiyokawa, Hideaki Kuzuoka, Bongshin Lee, Guillermo Molina León

Research Collection School Of Computing and Information Systems

We characterize 16 challenges faced by those investigating and developing remote and synchronous collaborative experiences around visualization. Our work reflects the perspectives and prior research efforts of an international group of 29 experts from across human-computer interaction and visualization sub-communities. The challenges are anchored around five collaborative activities that exhibit a centrality of visualization and multimodal communication. These activities include exploratory data analysis, creative ideation, visualization-rich presentations, joint decision making grounded in data, and real-time data monitoring. The challenges also reflect the changing dynamics of these activities in the face of recent advances in extended reality (XR) and artificial intelligence …