Search for: All records

The practice of soliciting self-explanations from students is widely recognized for its pedagogical benefits. However, the labor-intensive effort required to manually assess students’ explanations makes it impractical for classroom settings. As a result, many current solutions to gauge students’ understanding during class are often limited to multiple choice or fill-in-the-blank questions, which are less effective at exposing misconceptions or helping students to understand and integrate new concepts. Recent advances in large language models (LLMs) present an opportunity to assess student explanations in real-time, making explanation-based classroom response systems feasible for implementation. In this work, we investigate LLM-based approaches for assessing the correctness of students’ explanations in response to undergraduate computer science questions. We investigate alternative prompting approaches for multiple LLMs (i.e., Llama 2, GPT-3.5, and GPT-4) and compare their performance to FLAN-T5 models trained in a fine-tuning manner. The results suggest that the highest accuracy and weighted F1 score were achieved by fine-tuning FLAN-T5, while an in-context learning approach with GPT-4 attains the highest macro F1 score. 

Science learning reforms require shifting epistemic and power structures so students and teachers build knowledge together within the context of meaningful questions and problems. Curricular customization allows teachers to preserve these reform-oriented goals while adapting for their specific contexts and students. This paper presents three cases from professional learning communities (PLCs) who followed the same curriculum customization model, but with different equity goals: supporting bi/multilingual learners’ ownership of learning, increasing the relevance of curriculum for students, and encouraging more student voices and multiple perspectives. Together, these cases highlight how the collaborative customization model facilitated productive tensions that lead to teacher learning. 

Calibration is a crucial step for model validity, yet its representation is often disregarded. This paper proposes a two-stage approach to calibrate a model that represents target data by identifying multiple diverse parameter sets while remaining computationally efficient. The first stage employs a black-box optimization algorithm to generate near-optimal parameter sets, the second stage clusters the generated parameter sets. Five black-box optimization algorithms, namely, Latin Hypercube Sampling (LHS), Sequential Model-based Algorithm Configuration (SMAC), Optuna, Simulated Annealing (SA), and Genetic Algorithm (GA), are tested and compared using a disease-opinion compartmental model with predicted health outcomes. Results show that LHS and Optuna allow more exploration and capture more variety in possible future health outcomes. SMAC, SA, and GA, are better at finding the best parameter set but their sampling approach generates less diverse model outcomes. This two-stage approach can reduce computation time while producing robust and representative calibration. 

We use the three‐dimensional (3‐D) global hybrid code ANGIE3D to simulate the interaction of four solar wind tangential discontinuities (TDs) observed by ARTEMIS P1 from 0740 UT to 0800 UT on 28 December 2019 with the bow shock, magnetosheath, and magnetosphere. We demonstrate how the four discontinuities produce foreshock transients, a magnetosheath cavity‐like structure, and a brief magnetopause crossing observed by THEMIS and MMS spacecraft from 0800 UT to 0830 UT. THEMIS D observed entries into foreshock transients exhibiting low density, low magnetic field strength, and high temperature cores bounded by compressional regions with high densities and high magnetic field strengths. The MMS spacecraft observed cavities with strongly depressed magnetic field strengths and highly deflected velocity in the magnetosheath downstream from the foreshock. Dawnside THEMIS A magnetosheath observations indicate a brief magnetosphere entry exhibiting enhanced magnetic field strength, low density, and decreased and deflected velocity (sunward flow). The solar wind inputs into the 3‐D hybrid simulations resemble those seen by ARTEMIS. We simulate the interaction of four oblique TDs with properties similar to those in the observation. We place virtual spacecraft at the locations where observations were made. The hybrid simulations predict similar characteristics of the foreshock transients, a magnetosheath cavity, and a magnetopause crossing with characteristics similar to those observed by the multi‐spacecraft observations. The detailed and successful comparison of the interaction involving multiple TDs will be presented.