Current benchmarks for evaluating large language models (LLMs) in medicine are primarily focused on question-answering involving domain knowledge and descriptive reasoning. While such qualitative capabilities are vital to medical diagnosis, in real-world scenarios, doctors frequently use clinical calculators that follow quantitative equations and rule-based reasoning paradigms for evidence-based decision support. To this end, we propose MedCalc-Bench, a first-of-its-kind dataset focused on evaluating the medical calculation capability of LLMs. MedCalc-Bench contains an evaluation set of over 1000 manually reviewed instances from 55 different medical calculation tasks. Each instance in MedCalc-Bench consists of a patient note, a question requesting to compute a specific medical value, a ground truth answer, and a step-by-step explanation showing how the answer is obtained. While our evaluation results show the potential of LLMs in this area, none of them are effective enough for clinical settings. Common issues include extracting the incorrect entities, not using the correct equation or rules for a calculation task, or incorrectly performing the arithmetic for the computation. We hope our study highlights the quantitative knowledge and reasoning gaps in LLMs within medical settings, encouraging future improvements of LLMs for various clinical calculation tasks. MedCalc-Bench is publicly available at: https://github.com/ncbi-nlp/MedCalc-Bench.

In the challenging field of introductory programming, high enrolments and failure rates drive us to explore tools and systems to enhance student outcomes, especially automated tools that scale to large cohorts. This paper presents and evaluates the dcc --help tool, an integration of a Large Language Model (LLM) into the Debugging C Compiler (DCC) to generate unique, novice-focused explanations tailored to each error. dcc --help prompts an LLM with contextual information of compile- and run-time error occurrences, including the source code, error location and standard compiler error message. The LLM is instructed to generate novice-focused, actionable error explanations and guidance, designed to help students understand and resolve problems without providing solutions. dcc --help was deployed to our CS1 and CS2 courses, with 2,565 students using the tool over 64,000 times in ten weeks. We analysed a subset of these error/explanation pairs to evaluate their properties, including conceptual correctness, relevancy, and overall quality. We found that the LLM-generated explanations were conceptually accurate in 90% of compile-time and 75% of run-time cases, but often disregarded the instruction not to provide solutions in code. Our findings, observations and reflections following deployment indicate that dcc --help provides novel opportunities for scaffolding students' introduction to programming.

A fundamental challenge for system engineers is to capture a problem with an effective model or framework and then facilitate transferring the information of that captured problem to practical systems engineering tools and methods. The early problem definition phase requires an application of systems thinking with adequate modeling tools and methods. Then, the later problem definition phase and early system architecting phase requires transferring the captured problem to systems engineering tools and methods through emerging techniques such as model-based systems engineering (MBSE) using SysML (MBSE is the practice of using a modeling tools to capture systems engineering diagrams). This paper presents a method for capturing a problem through systemigrams and the Boardman soft systems methodology and then directly translating the systemigrams into SysML diagrams. With MBSE increasing in usage, this method could provide a time savings opportunity during model development along with the possibility of lowering information distortion or loss that can occur during transformation of systems thinking to systems engineering activities. This paper includes a case study which demonstrates how the proposed approach was applied on a problem being considered by the U.S. Army-Contingency Basing for Small Combat Units. Finally, this paper will provide the conclusion on the development of the method and describe future research directions that can allow systems thinking and MBSE to function in a congruent methodology.

This paper is concerned with the modeling of a plastic film manufacturing process and the development and implementation of a model-based Cross-Directional (CD) controller. The model is derived from first-principles and some empirical relationships. The final validated nonlinear model could provide a useful offline platform for developing control and monitoring algorithms. A new controller is designed which has a similar structure to that of Internal Model Control (IMC) with the addition of an observer whose gain is designed to minimize process and model mismatch. The observer gain is obtained by solving a multiobjective optimization problem through the application of a genetic algorithm. The controller is applied to the nonlinear model and simulation results are presented demonstrating improvements that can be achieved by the proposed controller over two existing CD controllers.

This article investigates a wireless acoustic sensor network application—monitoring amphibian populations in the monsoonal woodlands of northern Australia. Our goal is to use automatic recognition of animal vocalizations to census the populations of native frogs and the invasive introduced species, the cane toad. This is a challenging application because it requires high frequency acoustic sampling, complex signal processing, wide area sensing coverage and long-lived unattended operation. We set up two prototypes of wireless sensor networks that recognize vocalizations of up to ninth frog species found in northern Australia. Our first prototype consists of only resource-rich Stargate devices. Our second prototype is more complex and consists of a hybrid mixture of Stargates and inexpensive, resource-poor Mica2 devices operating in concert. In the hybrid system, the Mica2s are used to collect acoustic samples, and expand the sensor network coverage. The Stargates are used for resource-intensive tasks such as fast Fourier transforms (FFTs) and machine learning. The hybrid system incorporates four algorithms designed to account for the sampling, processing, energy, and communication bottlenecks of the Mica2s (1) high frequency sampling, (2) thresholding and noise reduction, to reduce data transmission by up to 90%, (3) sampling scheduling, which exploits the sensor network redundancy to increase the effective sample processing rate, and (4) harvesting-aware energy management, which exploits sensor energy harvesting capabilities to extend the system lifetime. Our evaluation shows the performance of our systems over a range of scenarios, and demonstrate that the feasibility and benefits of a hybrid systems approach justify the additional system complexity.

This paper investigates a wireless, acoustic sensor network application-monitoring amphibian populations in the monsoonal woodlands of northern Australia. Our goal is to use automatic recognition of animal vocalizations to census the populations of native frogs and the invasive introduced species, the cane toad. This is a challenging application because it requires high frequency acoustic sampling, complex signal processing and wide area sensing coverage. We set up two prototypes of wireless sensor networks that recognize vocalizations of up to 9 frog species found in northern Australia. Our first prototype is simple and consists of only resource-rich Stargate devices. Our second prototype is more complex and consists of a hybrid mixture of Stargates and inexpensive, resource-poor Mica2 devices operating in concert. In the hybrid system, the Mica2s are used to collect acoustic samples, and expand the sensor network coverage. The Stargates are used for resource-intensive tasks such as fast Fourier transforms (FFTs) and machine learning. The hybrid system incorporates three algorithms designed to account for the sampling, processing and communication bottlenecks of the Mica2s (i) high frequency sampling, (ii) compression and noise reduction, to reduce data transmission by up to 90%, and (iii) sampling scheduling, which exploits the sensor network redundancy to increase the effective sample processing rate. We evaluate the performance of both systems over a range of scenarios, and demonstrate that the feasibility and benefits of a hybrid systems approach justify the additional system complexity.

This demonstration shows a wireless, acoustic sensor network application--- monitoring amphibian populations in the monsoonal woodlands of northern Australia. Our system uses automatic recognition of animal vocalizations to census the populations of native frogs and the invasive introduced species, the Cane Toad (see Fig. 1). This is a challenging application because it requires high frequency acoustic sampling, complex signal processing and wide area sensing coverage [2]. Our prototype consists of a hybrid mixture of Stargates and inexpensive, resource-poor Mica motes operating in concert. The Mica motes are used to collect acoustic samples, and expand the sensor network coverage. The Stargates are used for resource-intensive tasks.

Many non-speaking people use augmentative and alternative communication (AAC) systems to assist them to communicate with other people. Access to an AAC system is generally slow for its user, who may have other disabilities as well as being non-speaking. An AAC system can contain stored words, messages and stories for use in communication, and there can be a large quantity of such information for the user to search through and select from while he or she is trying to participate in conversation. New user interface techniques are required to assist a user to navigate through the information stored in an AAC system and select appropriate items for use during live conversation. Information visualisation techniques may be able to assist a user to overview and appraise the contents of an AAC system as part of that selection process, and therefore improve its efficiency of use. This paper outlines key AAC issues and explores visualisation techniques which may be of value in an AAC context.