Large spatial datasets with non-Gaussian responses are increasingly common in environmental monitoring, ecology, and remote sensing, yet scalable Bayesian inference for such data remains challenging. Markov chain Monte Carlo (MCMC) methods are often prohibitive for large datasets, and existing variational Bayes methods rely on conjugacy or strong approximations that limit their applicability and can underestimate posterior variances. A scalable variational framework that incorporates semi-implicit variational inference (SIVI) with basis representations of spatial generalized linear mixed models (SGLMMs), which may not have conjugacy, is proposed. The proposed framework accommodates gamma, negative binomial, Poisson, Bernoulli, and Gaussian responses on continuous spatial domains. Across 20 simulation scenarios with 50,000 locations, SIVI achieves predictive accuracy and posterior distributions comparable to Metropolis-Hastings and Hamiltonian Monte Carlo while providing notable computational speedups. Applications to MODIS land surface temperature and Blue Jay abundance further demonstrate the utility of the approach for large non-Gaussian spatial datasets.
Abstract Progressive public–private partnerships (P3s) have emerged as alternatives to fixed-price P3 models in response to concerns about excessive risk transfer and the challenges of delivering complex transportation projects. Through a predevelopment agreement (PDA), progressive P3s introduce the development entity earlier, enabling collaborative scope refinement, iterative risk allocation, and a clearer pricing basis before committing to a long-term agreement. This study investigates progressive P3s through comparative case analysis of the Pennsylvania DOT (PennDOT) Major Bridge P3 Project (MBP3) and Maryland DOT’s Op Lanes Maryland Project, two initiatives that followed contrasting development trajectories. Using an exploratory case study design structured around four analytical dimensions and one inductively derived dimension—owner readiness and project preparation—the analysis examines how progressive P3 structures shape collaborative development and the conditions under which projects converge toward implementation or diverge during development. The MBP3 case illustrates how deliberate predevelopment program design, including an availability-payment structure and scope configured for adaptive delivery, created conditions for convergence despite significant institutional setbacks, although advancement was partial, with three of nine candidate bridges not advancing under the program. The Op Lanes case illustrates how unresolved institutional dependencies and a financial structure lacking adaptive capacity created conditions that collaborative development could not overcome, ultimately leading to orderly termination of the PDA. The comparison indicates that development-phase outcomes depend substantially on conditions established before collaborative development begins, that unresolved third-party dependencies become shared challenges for public agencies and developers once the PDA is executed because both parties are invested in advancing the project, and that governance in progressive P3s depends not only on contractual design but also on the public agency’s capacity to manage collaborative development. Five propositions are derived to guide future empirical research on collaborative development processes and outcomes in progressive P3s.
Abstract To examine emergency communication on Twitter across different phases of Hurricane Irma, we developed a Python script to access the Twitter Premium application programming interface (API). We applied a stratified sampling method to build a representative sample of 1,168,484 Irma-related tweets posted from August 30, 2017 to September 28, 2017. Using natural language processing (NLP) algorithms, we categorized tweet topics and traced the evolution of communication patterns in affected Florida communities. Our analysis validated established temporal markers for hurricane phases, such as the issuance of watches and the timing of landfall. We further compared emergency communication patterns between counties that were more directly impacted by Hurricane Irma and those outside its path, highlighting significant geographic variations. The results underscore the importance of phase- and location-specific messaging and suggest that emergency management agencies should adopt a more proactive communication approach during the pre-event stage and immediately following landfall.
Words sound different every time they are said, both by the same person and across people. This variability has been found to have mixed effects on learning, sometimes benefitting learning and other times making it more challenging. However, all prior studies investigating effects of talker variability have used passive exposure, presenting participants with predetermined orders or fixed types of variability. Here, we used a novel active learning paradigm wherein 3- and 4-year old preschoolers could choose who to hear words from during word learning. Then, we tested whether participants seek out variability during word learning, how these talker variability choices influence word learning and whether actively making these choices results in different outcomes than passively receiving them. We find that preschoolers choose to hear words from multiple talkers, but that talker variability does not predict word learning. Taken together, our results suggest that preschool aged children do seek out variability, but that word learning outcomes are most predicted by their relative vocabulary sizes and not talker variability choices. These results inform theories on the roles of both talker variability and active learning in language learning.
Large language models (LLMs) have significantly advanced the field of natural language processing (NLP), providing a highly useful, task-agnostic foundation for a wide range of applications. However, directly applying LLMs to solve sophisticated problems in specific domains meets many hurdles, caused by the heterogeneity of domain data, the sophistication of domain knowledge, the uniqueness of domain objectives, and the diversity of the constraints (e.g., various social norms, cultural conformity, religious beliefs, and ethical standards in the domain applications). Domain specification techniques are key to make large language models disruptive in many applications. Specifically, to solve these hurdles, there has been a notable increase in research and practices conducted in recent years on the domain specialization of LLMs. This emerging field of study, with its substantial potential for impact, necessitates a comprehensive and systematic review to better summarize and guide ongoing work in this area. In this article, we present a comprehensive survey on domain specification techniques for large language models, an emerging direction critical for large language model applications. First, we propose a systematic taxonomy that categorizes the LLM domain-specialization techniques based on the accessibility to LLMs and summarizes the framework for all the subcategories as well as their relations and differences to each other. Second, we present an extensive taxonomy of critical application domains that can benefit dramatically from specialized LLMs, discussing their practical significance and open challenges. Last, we offer our insights into the current research status and future trends in this area.
