Search for: All records

The development of generative language models that can create long and coherent textual outputs via autoregression has lead to a proliferation of uses and a corresponding sweep of analyses as researches work to determine the limitations of this new paradigm. Unlike humans, these ‘Large Language Models’ (LLMs) are highly sensitive to small changes in their inputs, leading to unwanted inconsistency in their behavior. One problematic inconsistency when LLMs are used to answer multiple-choice questions or analyze multiple inputs is order dependency: the output of an LLM can (and often does) change significantly when sub-sequences are swapped, despite both orderings being semantically identical. In this paper we present Set-Based Prompting, a technique that guarantees the output of an LLM will not have order dependence on a specified set of sub-sequences. We show that this method provably eliminates order dependency, and that it can be applied to any transformer-based LLM to enable text generation that is unaffected by re-orderings. Delving into the implications of our method, we show that, despite our inputs being out of distribution, the impact on expected accuracy is small, where the expectation is over the order of uniformly chosen shuffling of the candidate responses, and usually significantly less in practice. Thus, Set-Based Prompting can be used as a ‘dropped-in’ method on fully trained models. Finally, we discuss how our method’s success suggests that other strong guarantees can be obtained on LLM performance via modifying the input representations. Code is available at github.com/reidmcy/set-based-prompting. 

While smartphones are familiar in high schools, their intricate material composition and far-reaching geographical footprint remain largely unexplored. This research explores how educators can leverage ArcGIS, a geospatial software tool, to bridge the gap between theoretical understanding and real-world consequences. The paper details a collaborative project between researchers and teachers to integrate location-based learning into the curriculum. One such project involved a teacher using geographic information systems to guide students in exploring the global impact of smartphones. This approach fostered critical thinking and environmental awareness, equipping students with vital knowledge about the materials behind everyday technology. By harnessing the power of geospatial tools, educators can cultivate a generation of "geo-sustainable" thinkers prepared to navigate the complex life cycle of digital devices – from extraction to disposal – in an environmentally responsible manner. This paper delves into the project methodology, the application of geospatial technology, and the broader implications for future research in this domain. 

Excessive land application of poultry litter (PL) may lead to surface runoff of nitrogen (N) and phosphorus (P), which cause eutrophication, fish death, and water pollution that ultimately have negative effects on humans and animals. Increases in poultry production in the Delmarva Peninsula underscore the need for more efficient, cost-effective, and sustainable disposal technologies for processing PL instead of direct land application. The pyrolysis conversion process can potentially produce nutrient-rich poultry litter biochar (PLB), while the pyrolysis process can change the N and P to a more stable component, thus reducing its runoff. Pyrolysis also kills off any microorganisms that would otherwise trigger negative environmental health effects. This study is to apply an integrated method and investigate the effect of pyrolysis temperature (300 °C, 500 °C), poultry litter source (different feedstock composition), and bedding material mixture (10% pine shavings) on PLB qualities and quantities. Proximate and ultimate analysis showed PL sources and bedding material addition influenced the physicochemical properties of feedstock. The SEM and BET surface results indicate that pyrolysis temperature had a significant effect on changing the PLB morphology and structure, as well as the pH value (7.78 at 300 °C vs. 8.78 at 500 °C), extractable phosphorus (P) (18.73 ppm at 300 °C vs. 11.72 ppm at 500 °C), sulfur (S) (363 ppm at 300 °C vs. 344 ppm at 500 °C), and production yield of PLBs (47.65% at 300 °C vs. 60.62% at 500 °C). The results further suggest that adding a bedding material mixture (10% pine shavings) to PLs improved qualities by reducing the content of extractable P and S, as well as pH values of PLBs. This study also found the increment in both the pore volume and the area of Bethel Farm was higher than that of Sun Farm. Characterization and investigation of qualities and quantities of PLB using the integrated framework suggest that PL from Bethel Farm could produce better-quality PLB at a higher pyrolysis temperature and bedding material mixture to control N and P runoff problems. 

Abstract Marsh accretion models predict the resiliency of coastal wetlands and their ability to store carbon in the face of accelerating sea level rise. Most existing marsh accretion models are derived from two parent models: the Marsh Equilibrium Model, which formalizes the biophysical relationships between sea level rise, dominant macrophyte growth, and elevation change; and the Cohort Theory Model, which formalizes how carbon mass pools belowground contribute to soil volume expansion over time. While there are several existing marsh accretion models, the application of these models by a broader base of researchers and practitioners is hindered because of (a) limited descriptions of how empirically derived ecological mechanism informed the development of these models, (b) limitations in the ability to apply models to geographies with variable tidal regimes, and (c) a lack of open‐source code to apply models. Here, we provide for the first time an explicit description of a mathematical version of the Cohort Theory Model and a numerical version of a combined model: the Cohort Marsh Equilibrium Model (CMEM) with an accompanying open‐sourceRpackage,rCMEM. We show that, through this “depth‐aware” model, we can capture how tidal variation impacts broad patterns of marsh accretion and carbon sequestration across the United States. The application of this model will likely be imperative in predicting the fate and state of coastal wetlands and the ecosystem services they provide in an era of rapid environmental change. 

Across Texas, Washington, and Pennsylvania, three university teams worked with teachers at three high schools to integrate geographic information systems (GIS) and other geospatial tools into chemistry lessons as part of a larger, multi-disciplinary teacher professional development initiative. Each university followed a specific design model of socio-environmental science investigations (SESI) in their professional development and curriculum development processes. Each teacher’s work is presented as a case with distinct school contexts, professional development experiences, classroom implementation outcomes, and reflections after implementation. Cross-case findings include variability in teachers’ adoption processes, the importance of cross-site collaboration, and the ability of geospatial tools to bring chemistry topics “off the bench” and into students’ thinking about their world. These cases present an advance in the curricular reach of GIS, which to date has not been broadly used in high school chemistry instruction. Further, the cases illustrate examples of the teachers’ geospatial science pedagogical content knowledge. 

Across Texas, Washington, and Pennsylvania, three university teams worked with teachers at three high schools to integrate geographic information systems (GIS) and other geospatial tools into chemistry lessons as part of a larger, multi-disciplinary teacher professional development initiative. Each university followed a specific design model of socio-environmental science investigations (SESI) in their professional development and curriculum development processes. Each teacher’s work is presented as a case with distinct school contexts, professional development experiences, classroom implementation outcomes, and reflections after implementation. Cross-case findings include variability in teachers’ adoption processes, the importance of cross-site collaboration, and the ability of geospatial tools to bring chemistry topics “off the bench” and into students’ thinking about their world. These cases present an advance in the curricular reach of GIS, which to date has not been broadly used in high school chemistry instruction. Further, the cases illustrate examples of the teachers’ geospatial science pedagogical content knowledge. 

A critical outcome in social studies education is identity development, and an important component of this process is students establishing a sense of place in their communities, nation, and world. Using data from a southwestern city in the United States, researchers investigated the intersection of local history, identity development, and cultural heritage resources using GIS technology. The instructional unit in which students participated utilized a variety of geospatial technologies which facilitated the visualization of geographic concepts, field-based data collection of geocoded places, and creation of a digitally-mediated cultural heritage map, which allowed students to create a narrative around their cultural identity. The study followed an interpretive case study design. Based on the findings from this study, important implications emerged, which are valuable for both future research in this area, as well as for teachers who wish to replicate this pedagogical approach in their own teaching practice. The implications include the flexibility of geospatial technologies for addressing content-area concepts at all levels of the instructional unit, the potential of geospatial technologies for supporting student cultural identity development, and the value of school-university partnerships in promoting innovative teaching strategies in a high school classroom. 

The COVID-19 pandemic brought unexpected changes across the globe to nearly all aspects of life. Activities such as shopping, traveling, and school, all considered routine aspects of life, suddenly took on a new level of risk. Approximately 1.6 billion students across the world had their schooling experiences interrupted during this global event.1 Matthew Stroup, a high school AP Psychology teacher, was dealing with these unexpected changes, but he also decided to use this disruption as a teaching moment. Mr. Stoup understood that sheltering in place and the resulting social isolation increased stress and anxiety among many families.2 During the 2020-21 school year, approximately 71% of all students were receiving either all or some of their schooling virtually.3 Mr. Stroup observed the effect of this unpredictable learning environment on his students’ learning and the toll it was taking on their relationships with peers, teachers, and family. According to the American Psychological Association, 81% of teens experienced a degradation in their mental health between Spring of 2020 and into the end of the year.4 The idea for the activity described in this article sprang from a class discussion on the effect of social isolation and travel hesitancy on mental health.