Search for: All records

Exploratory learning environments (ELEs), such as simulation-based platforms and open-ended science curricula, promote hands-on exploration and problem-solving but make it difficult for teachers to gain timely insights into students' conceptual understanding. This paper presents LearnLens, a generative AI (GenAI)-enhanced teacher-facing dashboard designed to support problem-based instruction in middle school science. LearnLens processes students' open-ended responses from digital assessments to provide various insights, including sample responses, word clouds, bar charts, and AI-generated summaries. These features elucidate students' thinking, enabling teachers to adjust their instruction based on emerging patterns of understanding. The dashboard was informed by teacher input during professional development sessions and implemented within a middle school Earth science curriculum. We report insights from teacher interviews that highlight the dashboard's usability and potential to guide teachers' instruction in the classroom. 

Computational modeling of scientific systems is a powerful approach for fostering science and computational thinking (CT) proficiencies. However, the role of programming activities for this synergistic learning remains unclear. This paper examines alternative ways to engage with computational models (CM) beyond programming. Students participated in an integrated Science, Engineering, and Computational Modeling unit through one of three distinct instructional versions: Construct a CM, Interpret-and-Evaluate a CM, and Explore-and-Evaluate a simulation. Analyzing 188 student responses to a science+CT embedded assessment task, we investigate how science proficiency and instructional versions related to pseudocode interpretation and debugging performances. We found that students in the Explore-and-Evaluate a simulation outperformed students in the programming-based versions on the CT assessment items. Additionally, science proficiency strongly predicted students’ CT performance, unlike prior programming experience. These results highlight the promise of diverse approaches for fostering CT practices with implications for STEM+C instruction and assessment design. 

Vision-language models are integral to computer vision research, yet many high-performing models remain closed-source, obscuring their data, design and training recipe. The research community has responded by using distillation from black-box models to label training data, achieving strong benchmark results, at the cost of measurable scientific progress. However, without knowing the details of the teacher model and its data sources, scientific progress remains difficult to measure. In this paper, we study building a Perception Language Model (PLM) in a fully open and reproducible framework for transparent research in image and video understanding. We analyze standard training pipelines without distillation from proprietary models and explore large-scale synthetic data to identify critical data gaps, particularly in detailed video understanding. To bridge these gaps, we release 2.8M human-labeled instances of fine-grained video question-answer pairs and spatio-temporally grounded video captions. Additionally, we introduce PLM-VideoBench, a suite for evaluating challenging video understanding tasks focusing on the ability to reason about "what", "where", "when", and "how" of a video. We make our work fully reproducible by providing data, training recipes, code & models. 

Collaborative dialogue offers rich insights into students’ learning and critical thinking, which is essential for personalizing pedagogical agent interactions in STEM+C settings. While large language models (LLMs) facilitate dynamic pedagogical interactions, hallucinations undermine confidence, trust, and instructional value. Retrieval-augmented generation (RAG) grounds LLM outputs in curated knowledge, but requires a clear semantic link between user input and a knowledge base, which is often weak in student dialogue. We propose log-contextualized RAG (LC-RAG), which enhances RAG retrieval by using the environment logs to contextualize collaborative discourse. Our findings show that LCRAG improves retrieval over a discourse-only baseline and allows our collaborative peer agent, Copa, to deliver relevant, personalized guidance that supports students’ critical thinking and epistemic decision-making in a collaborative computational modeling environment, C2STEM. 

The phase field method provides a simple mass conserving method for solving two-phase immiscible - incompressible Navier-Stokes Equations. The relative ease in implementing this method compared to other interface reconstruction methods, coupled with its conservativeness and boundedness makes it an attractive alternative. We implement the method in a parallel structured multi-block generalized coordinate finite volume solver using a collocated grid arrangement within the framework of the fractional-step method. The discretization uses a second-order central difference method for both the Navier-Stokes and the phase field equations. A TVD-based averaging technique is used for calculating density at cell faces in the pressure correction step to handle high-density ratios. The simulation framework is verified in standard test cases: Zalesak Disk, a droplet in shear flow, Solitary Wave Runup, Rayleigh Taylor Instability, and the Dam Break Problem. A second-order rate of convergence and excellent phase volume conservation is observed. 

The transport and deposition of firebrand particles is an important fire spread mechanism in wildland fires. These particles can be transported by wind over large distances and can ignite secondary fires upon landing. The transport of firebrands by wind is a complex, multiscale process that is largely controlled by interactions between the firebrand particles and the atmospheric wind. To account for the complex temporal evolution of atmospheric turbulence over large scales, the use of large-eddy simulation (LES) techniques is necessary. However, filtering of subgrid-scale (SGS) turbulence in LES hinders the accuracy of particle transport models. In this work, we employ a Lagrangian SGS model in an LES framework to investigate the effects of small-scale turbulence on the transport of mass- and size-changing firebrand particles. The impact of SGS turbulence was analyzed by comparing landing and trajectory statistics for firebrand and regular (fixed size and mass) particles under different Stokes numbers. It was found that the presence of SGS turbulence modifies the particle transport behavior, which is characterized by smaller spanwise dispersions but larger travel distances along the streamwise direction compared with particles under no SGS turbulence. As expected, the enhanced velocity field produced by the SGS model has larger influence on the statistics of firebrand particles compared with regular particles due to the time-evolving reduction in particle mass and size induced by pyrolysis. 

We report a study of the inelasticity distribution in the scattering of neutrinos of energy 80–560 GeV off nucleons. Using atmospheric muon neutrinos detected in IceCube’s sub-array DeepCore during 2012–2021, we fit the observed inelasticity in the data to a parameterized expectation and extract the values that describe it best. Finally, we compare the results to predictions from various combinations of perturbative QCD calculations and atmospheric neutrino flux models. Published by the American Physical Society2025