Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher.
Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?
Some links on this page may take you to non-federal websites. Their policies may differ from this site.
-
Generative artificial intelligence (GenAI) is becoming part of everyday knowledge work, where it can accelerate drafting, coding, summarization, ideation, customer communication, and other language-intensive tasks. Yet the organizational problem is conversion: local GenAI gains create organizational value only when they survive the full work system, including verification, coordination, learning, ownership, risk, and accountability. Drawing on 251 academic and practice-facing sources across information systems (IS), management, human–computer interaction, human factors, organizational theory, and responsible AI, this article synthesizes why GenAI-enabled work often produces local augmentation but less often yields governable organizational synergy. The review develops three findings. First, GenAI shifts the relevant unit of collaboration from a one-step task interaction to a recurrent production-and-revision workflow in which human framing, prompting, verification, correction, and reuse become part of the value-producing system. Second, the augmentation-to-synergy boundary is net work-system value: visible task gains and reallocation benefits must exceed work displaced across roles and time into verification, rework, coordination, learning, ownership, risk, and accountability. Third, durable synergy requires behavioral embedding and multi-outcome governance: prompts, memory, feedback traces, review routines, and escalation rules must improve local fit while keeping reliance, learning, ownership, risk, and accountability observable. The article contributes a work-system conversion account that distinguishes substitution, augmentation, and synergy as different organizational states; a trace-centered behavioral-embedding view of how recurring GenAI use becomes a persistent, history-dependent workflow state within established sociotechnical IS theory; and a governance-instrumentation account that evaluates visible gains together with hidden labor, learning, ownership, risk, reviewability, and accountability.more » « lessFree, publicly-accessible full text available February 1, 2028
-
Sequence data, such as DNA, RNA and protein sequences, exhibit intricate, multi-scale structures that pose significant challenges for conventional analysis methods, particularly those relying on alignment or purely statistical representations. In this work, we introduce category-based topological sequence analysis (CTSA) of genomes. CTSA models a sequence as a resolution category, capturing its hierarchical structure through a categorical construction. Substructure complexes are then derived from this categorical representation, and their persistent homology is computed to extract multi-scale topological features. Our models depart from traditional alignment-free approaches by incorporating structured mathematical formalisms rooted in sequence topology. The resulting topological signatures provide informative representations across a variety of tasks, including the phylogenetic analysis of SARS-CoV-2 variants and the prediction of protein–nucleic acid binding affinities. Comparative studies were carried out against six state-of-the-art methods. Experimental results demonstrate that CTSA achieves excellent and consistent performance in these tasks, suggesting its general applicability and robustness. Beyond sequence analysis, the proposed framework opens new directions for the integration of categorical and homological theories for biological sequence analysis.more » « lessFree, publicly-accessible full text available December 1, 2027
-
Published conference paper in Computer Vision - ECCV 2026, Lecture Notes in Computer Science, volume 17007, pages 38-56. Published September 25, 2026. DOI: 10.1007/978-3-032-37010-5_3. DiffUE studies semantic-space protective perturbations using diffusion autoencoders to preserve visual utility while resisting unauthorized learning.more » « lessFree, publicly-accessible full text available September 25, 2027
-
The maintenance of carboxysome distribution (Mcd) system comprises the proteins McdA and McdB, which spatially organize carboxysomes to ensure their inheritance during cell division and promote efficient carbon fixation. McdA, a member of the ParA/MinD family of ATPases, forms dynamic gradients on the nucleoid that position McdB-bound carboxysomes. McdB belongs to a widespread class of ParA/MinD partner proteins. In most studied cases, partner protein interactions with their cognate ParA/MinD ATPase have been proposed to be mainly driven by key arginine or lysine residues near the N-terminus. The molecular basis of McdB interaction with McdA remains unknown. Here, we provide evidence showing that the N-terminal 20 residues of H. neapolitanus McdB are important for interaction with McdA. Within this region, we identify critical residues whose individual substitution modulates McdA binding and leads to distinct carboxysome organization phenotypes. Counter to prevailing models that highlight basic residues as key to partner protein interaction, we found that the hydrophobic residue, leucine 8, was critical for McdB binding to McdA and its mutation phenocopied an mcdB deletion - where carboxysomes cluster and become nucleoid excluded to the cell poles. Lysine 7 (K7) of McdB was also critical for McdA interaction, but substitutions at this site, or the interacting residue on McdA (E188), resulted in the dynamic positioning of a single carboxysome cluster at mid-nucleoid. This ‘positioning without partitioning’ phenotype was not unique to K7 - mutating hydrophobic residues predicted to weakly interact with McdA, or simply decreasing the cellular levels of the McdAB system, also produced this intermediate phenotype. The findings suggest that down-tuning McdA–McdB interaction strength and/or frequency are sufficient to position carboxysome clusters on the nucleoid but inadequate for partitioning individual carboxysomes. We propose that, within the ParA/MinD family, cargo positioning and partitioning are separable – weakened and/or low-frequency interactions with the cognate partner can mediate positioning, but effective partitioning requires stronger and/or higher-frequency interactions capable of overcoming the self-association forces of the cargo itself.more » « lessFree, publicly-accessible full text available September 3, 2027
-
Metal-organic frameworks (MOFs) and MOF-like porous materials exhibit vast structural diversity and support critical applications in gas storage, separations, and catalysis. Predictive modeling remains difficult because their structure-property relationships are multiscale and cage-like, governed by both local chemical environments and global pore-network topology. These challenges, together with sparse and unevenly distributed labeled data, hinder generalization across material families. We develop an interaction topology theory and propose the interaction topological transformer (ITT), a data-efficient framework that captures materials information across multiple scales and levels, including structural, elemental, atomic, and pairwise-elemental organization. ITT extracts scale-aware features reflecting both compositional and relational structures in complex porous frameworks and integrates them through a transformer architecture for joint reasoning across scales. Using self-supervised pretraining on more than 0.6 million unlabeled structures followed by supervised fine-tuning, ITT achieves accurate, transferable, state-of-the-art predictions for adsorption, transport, and stability properties across 17 tasks, providing a principled and scalable strategy for learning-guided discovery in diverse MOF-like materials.more » « lessFree, publicly-accessible full text available August 21, 2027
-
Free, publicly-accessible full text available June 1, 2027
-
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 (ART) model that supports single-pass learning without replay. To enable strong adaptive evaluation despite ARTMAP’s non-differentiable winner-take-all dynamics, we propose WB-Softmax, a differentiable softmax relaxation that aggregates category-level activations into class-level scores for gradient-based attacks. Across USPS, MNIST, and Fashion-MNIST, WB-Softmax Projected Gradient Descent (PGD) achieves 89–100% attack success on vanilla models, exceeding transfer and query-based baselines at matched budgets and satisfying anti-gradient-masking sanity checks. We then study adversarial training under true streaming constraints by comparing offline versus online adversarial example generation and standard versus selective updates. Offline adversarial training consistently collapses robustness despite substantial category growth, whereas online training is dataset-dependent. We introduce progressive two-stage selective training that combines selective filtering with ϵ scheduling, achieving the best overall robustness (AURAC) on all three datasets (28.2%, 64.5%, and 41.3%). Finally, leveraging ART’s explicit category geometry, we monitor incremental cluster validity indices to diagnose separation collapse and design a separation-aware absorption rule that improves high-ϵ robustness on USPS (11.0±1.4% at ϵ=0.30).more » « lessFree, publicly-accessible full text available June 1, 2027
-
Online learning models offer clear advantages over static alternatives when operating in dynamic settings. Unfortunately, this class of models is often too computationally demanding for the embedded and edge platforms where such adaptability is most valuable. This work introduces a new model architecture derived from the Adaptive Resonance Theory family that exploits binary data representations and exclusively hardware-efficient integer operations, eliminating the need for floating-point arithmetic or division. In contrast to existing ART architectures that rely on floating-point arithmetic, the proposed design enables fast, memory-efficient, and hardware-optimized online learning, making it suitable for deployment on resource-constrained devices. Through comparisons with analog counterparts, we demonstrate that the proposed model is capable of achieving faster training and inference while reducing memory usage by orders of magnitude, without any significant trade-offs in accuracy.more » « lessFree, publicly-accessible full text available June 1, 2027
-
Free, publicly-accessible full text available April 28, 2027
-
Adaptive Resonance Theory (ART) enables stable learning in non-stationary environments, but deploying ART models on embedded platforms is often limited by floating-point operations and the cost of repeatedly scanning many learned categories. We present BFAMod, a lightweight CPU–FPGA co-design that accelerates inference for Binary Fuzzy ART (BFA) on embedded MPSoCs. BFAMod streams magnitude-coded inputs and category prototypes via DMA, performs on-the-fly magnitude-to-thermometer expansion, and implements BFA matching and category selection using word-level logic and popcount accumulation with division-free cross-multiplication. An on-chip input cache reuses the current input across the full category scan, eliminating redundant host–device transfers. Implemented on a Kria KV260 platform at 100 MHz, BFAMod achieves more than an order-of-magnitude higher end-to-end inference rate than a quad-core ARM Cortex-A53 baseline on Fashion-MNIST, HAR, and Mini Speech Commands, and exhibits predictable scaling with category count and encoded dimensionality in controlled sweeps. The small resource footprint indicates feasibility as an embedded accelerator alongside other real-time pipelines, while optional online updates can remain host-managed when enabled.more » « lessFree, publicly-accessible full text available June 1, 2027
An official website of the United States government
