Search

This study was undertaken to determine whether the grade point averages of students at Santa Ana College (SAC) have been increasing over time, and, with an increasing number of work-experience students, whether the work experience grade distribution has had any effect on the grade distribution of the entire college. Results indicate that SAC's overall student grade point average has been increasing during the past four years at an average annual rate of 3.2 percent through the spring term, 1974. The absolute increase since fall 1970 has been 8.9 percent. There has been an 18 percent average increase in "A's" given and a corresponding 18 percent average decrease in "C's" given during the past four years. The number of withdrawals has increased 13 percent on the average during this time period. The percentages of "B,""D," and "F" grades have remained fairly constant. This trend has skewed the college grade distribution to the left, away from the traditional "normal" distribution. Mean grade point average for all students has increased from 2.57 in fall 1970 to 2.80 in spring 1974. Work experience grades were found to have little effect on the rate of increase of the overall college grade point average. (Author/DC)

Building pluralistic AI requires designing models that are able to be shaped to represent a wide range of value systems and cultures. Achieving this requires first being able to evaluate the degree to which a given model is capable of reflecting various personas. To this end, we propose a benchmark for evaluating the steerability of model personas as a function of prompting. Our design is based on a formal definition of prompt steerability, which analyzes the degree to which a model's joint behavioral distribution can be shifted from its baseline behavior. By defining steerability indices and inspecting how these indices change as a function of steering effort, we can estimate the steerability of a model across various persona dimensions and directions. Our benchmark reveals that the steerability of many current models is limited -- due to both a skew in their baseline behavior and an asymmetry in their steerability across many persona dimensions. We release an implementation of our benchmark at https://github.com/IBM/prompt-steering.

In this paper we provide a means of certifying infinitesimal projective rigidity relative to the cusp for hyperbolic once punctured torus bundles in terms of twisted Alexander polynomials of representations associated to the holonomy. We also relate this polynomial to an induced action on the tangent space of the character variety of the free group of rank 2 into PGL(4,R) that arises from the holonomy of a hyperbolic once-punctured torus bundle. We prove the induced action on the tangent space of the character variety is the same as the group theoretic action that arises in the Lyndon Hochschild Serre spectral sequence on cohomology., Comment: 25 pages

Large language models (LLMs) offer powerful capabilities but also introduce significant risks. One way to mitigate these risks is through comprehensive pre-deployment evaluations using benchmarks designed to test for specific vulnerabilities. However, the rapidly expanding body of LLM benchmark literature lacks a standardized method for documenting crucial benchmark details, hindering consistent use and informed selection. BenchmarkCards addresses this gap by providing a structured framework specifically for documenting LLM benchmark properties rather than defining the entire evaluation process itself. BenchmarkCards do not prescribe how to measure or interpret benchmark results (e.g., defining ``correctness'') but instead offer a standardized way to capture and report critical characteristics like targeted risks and evaluation methodologies, including properties such as bias and fairness. This structured metadata facilitates informed benchmark selection, enabling researchers to choose appropriate benchmarks and promoting transparency and reproducibility in LLM evaluation.

LLM-as-a-Judge is a widely used method for evaluating the performance of Large Language Models (LLMs) across various tasks. We address the challenge of quantifying the uncertainty of LLM-as-a-Judge evaluations. While uncertainty quantification has been well-studied in other domains, applying it effectively to LLMs poses unique challenges due to their complex decision-making capabilities and computational demands. In this paper, we introduce a novel method for quantifying uncertainty designed to enhance the trustworthiness of LLM-as-a-Judge evaluations. The method quantifies uncertainty by analyzing the relationships between generated assessments and possible ratings. By cross-evaluating these relationships and constructing a confusion matrix based on token probabilities, the method derives labels of high or low uncertainty. We evaluate our method across multiple benchmarks, demonstrating a strong correlation between the accuracy of LLM evaluations and the derived uncertainty scores. Our findings suggest that this method can significantly improve the reliability and consistency of LLM-as-a-Judge evaluations.

Rising breast cancer (BC) occurrence and mortality are major global concerns for women. Deep learning (DL) has demonstrated superior diagnostic performance in BC classification compared to human expert readers. However, the predominant use of unimodal (digital mammography) features may limit the current performance of diagnostic models. To address this, we collected a novel multimodal dataset comprising both imaging and textual data. This study proposes a multimodal DL architecture for BC classification, utilising images (mammograms; four views) and textual data (radiological reports) from our new in-house dataset. Various augmentation techniques were applied to enhance the training data size for both imaging and textual data. We explored the performance of eleven SOTA DL architectures (VGG16, VGG19, ResNet34, ResNet50, MobileNet-v3, EffNet-b0, EffNet-b1, EffNet-b2, EffNet-b3, EffNet-b7, and Vision Transformer (ViT)) as imaging feature extractors. For textual feature extraction, we utilised either artificial neural networks (ANNs) or long short-term memory (LSTM) networks. The combined imaging and textual features were then inputted into an ANN classifier for BC classification, using the late fusion technique. We evaluated different feature extractor and classifier arrangements. The VGG19 and ANN combinations achieved the highest accuracy of 0.951. For precision, the VGG19 and ANN combination again surpassed other CNN and LSTM, ANN based architectures by achieving a score of 0.95. The best sensitivity score of 0.903 was achieved by the VGG16+LSTM. The highest F1 score of 0.931 was achieved by VGG19+LSTM. Only the VGG16+LSTM achieved the best area under the curve (AUC) of 0.937, with VGG16+LSTM closely following with a 0.929 AUC score., Comment: The paper was accepted and presented in 3rd Workshop on Cancer Prevention, detection, and intervenTion (CaPTion @ MICCAI 2024)

Federated Learning (FL) is a machine learning technique that enables multiple entities to collaboratively learn a shared model without exchanging their local data. Over the past decade, FL systems have achieved substantial progress, scaling to millions of devices across various learning domains while offering meaningful differential privacy (DP) guarantees. Production systems from organizations like Google, Apple, and Meta demonstrate the real-world applicability of FL. However, key challenges remain, including verifying server-side DP guarantees and coordinating training across heterogeneous devices, limiting broader adoption. Additionally, emerging trends such as large (multi-modal) models and blurred lines between training, inference, and personalization challenge traditional FL frameworks. In response, we propose a redefined FL framework that prioritizes privacy principles rather than rigid definitions. We also chart a path forward by leveraging trusted execution environments and open-source ecosystems to address these challenges and facilitate future advancements in FL.

Evaluation of large language model (LLM) outputs requires users to make critical judgments about the best outputs across various configurations. This process is costly and takes time given the large amounts of data. LLMs are increasingly used as evaluators to filter training data, evaluate model performance or assist human evaluators with detailed assessments. To support this process, effective front-end tools are critical for evaluation. Two common approaches for using LLMs as evaluators are direct assessment and pairwise comparison. In our study with machine learning practitioners (n=15), each completing 6 tasks yielding 131 evaluations, we explore how task-related factors and assessment strategies influence criteria refinement and user perceptions. Findings show that users performed more evaluations with direct assessment by making criteria task-specific, modifying judgments, and changing the evaluator model. We conclude with recommendations for how systems can better support interactions in LLM-assisted evaluations.

Participatory Design -- an iterative, flexible design process that uses the close involvement of stakeholders, most often end users -- is growing in use across design disciplines. As an increasing number of practitioners turn to Participatory Design (PD), it has become less rigidly defined, with stakeholders engaged to varying degrees through the use of disjointed techniques. This ambiguous understanding can be counterproductive when discussing PD processes. Our findings synthesize key decisions and approaches from design peers that can support others in engaging in PD practice. We investigated how scholars report the use of Participatory Design in the field through a systematic literature review. We found that a majority of PD literature examined specific case studies of PD (53 of 88 articles), with the design of intangible systems representing the most common design context (61 of 88 articles). Stakeholders most often participated throughout multiple stages of a design process (65 of 88 articles), recruited in a variety of ways and engaged in several of the 14 specific participatory techniques identified. This systematic review provides today's practitioners synthesized learnings from past Participatory Design processes to inform and improve future use of PD, attempting to remedy inequitable design by engaging directly with stakeholders and users.

As generative AI, particularly large language models (LLMs), become increasingly integrated into production applications, new attack surfaces and vulnerabilities emerge and put a focus on adversarial threats in natural language and multi-modal systems. Red-teaming has gained importance in proactively identifying weaknesses in these systems, while blue-teaming works to protect against such adversarial attacks. Despite growing academic interest in adversarial risks for generative AI, there is limited guidance tailored for practitioners to assess and mitigate these challenges in real-world environments. To address this, our contributions include: (1) a practical examination of red- and blue-teaming strategies for securing generative AI, (2) identification of key challenges and open questions in defense development and evaluation, and (3) the Attack Atlas, an intuitive framework that brings a practical approach to analyzing single-turn input attacks, placing it at the forefront for practitioners. This work aims to bridge the gap between academic insights and practical security measures for the protection of generative AI systems.

The emergent dynamics in spacetime diagrams of cellular automata (CAs) is often organised by means of a number of behavioural classes. Whilst classification of elementary CAs is feasible and well-studied, non-elementary CAs are generally too diverse and numerous to exhaustively classify manually. In this chapter we treat the spacetime diagram as a digital image, and implement simple computer vision techniques to perform an automated classification of elementary cellular automata into the five Li-Packard classes. In particular, we present a supervised learning task to a convolutional neural network, in such a way that it may be generalised to non-elementary CAs. If we want to do so, we must divert the algorithm's focus away from the underlying 'microscopic' local updates. We first show that previously developed deep learning approaches have in fact been trained to identify the local update rule, rather than directly focus on the mesoscopic patterns that are associated with the particular behavioural classes. By means of a well-argued neural network design, as well as a number of data augmentation techniques, we then present a convolutional neural network that performs nearly perfectly at identifying the behavioural class, without necessarily first identifying the underlying microscopic dynamics., Comment: 19 pages, 12 figures, book chapter

Cellular automata (CAs) and convolutional neural networks (CNNs) are closely related due to the local nature of information processing. The connection between these topics is beneficial to both related fields, for conceptual as well as practical reasons. Our contribution solidifies this connection in the case of non-uniform CAs (nuCAs), simulating a global update in the architecture of the Python package TensorFlow. Additionally, we demonstrate how the highly optimised out-of-the-box multiprocessing in TensorFlow offers interesting computational benefits, especially when simulating large numbers of nuCAs with many cells., Comment: 10 pages, 5 figures, Conference Proceedings Article

This review paper describes the energy-upgraded CEBAF accelerator. This superconducting linac has achieved 12 GeV beam energy by adding 11 new high-performance cryomodules containing eighty-eight superconducting cavities that have operated CW at an average accelerating gradient of 20 MV/m. After reviewing the attributes and performance of the previous 6 GeV CEBAF accelerator, we discuss the upgraded CEBAF accelerator system in detail with particular attention paid to the new beam acceleration systems. In addition to doubling the acceleration in each linac, the upgrade included improving the beam recirculation magnets, adding more helium cooling capacity to allow the newly installed modules to run cold, adding a new experimental hall, and improving numerous other accelerator components. We review several of the techniques deployed to operate and analyze the accelerator performance, and document system operating experience and performance. In the final portion of the document, we present much of the current planning regarding projects to improve accelerator performance and enhance operating margins, and our plans for ensuring CEBAF operates reliably into the future. For the benefit of potential users of CEBAF, the performance and quality measures for beam delivered to each of the experimental halls is summarized in the appendix., Comment: 66 pages, 73 figures, 21 tables

Higher-order interactions (HOIs) have the potential to greatly increase our understanding of ecological interaction networks beyond what is possible with established models that usually consider only pairwise interactions between organisms. While equilibrium values of such HOI-based models have been studied, the dynamics of these models and the stability of their equilibria remain underexplored. Here we present a novel investigation on the effect of the onset speed of a higher-order interaction. In particular, we study the stability of the equilibrium of all configurations of a three-species interaction network, including transitive as well as intransitive ones. We show that the HOI onset speed has a dramatic effect on the evolution and stability of the ecological network, with significant structural changes compared to commonly used HOI extensions or pairwise networks. Changes in the HOI onset speed from fast to slow can reverse the stability of the interaction network. The evolution of the system also affects the equilibrium that will be reached, influenced by the HOI onset speed. This implies that the HOI onset speed is an important determinant in the dynamics of ecological systems, and including it in models of ecological networks can improve our understanding thereof., Comment: Main: 9 pages, supplementary: 9 pages

We present X-ray scattering studies resolving structural twinning and phase separation in the charge density wave (CDW) state of the kagome superconductor CsV$_3$Sb$_5$. The three-dimensional CDW state in CsV$_3$Sb$_5$ is reported to form a complex superposition of Star of David (SoD) or Tri-Hexagonal (TrH) patterns of distortion within its kagome planes, but the out-of-plane stacking is marked by metastability. In order to resolve the impact of this metastability, we present reciprocal space mapping and real-space images of CsV$_3$Sb$_5$ collected across multiple length scales using temperature-dependent high-dynamic range mapping (HDRM) and dark-field X-ray microscopy (DFXM). The experimental data provide evidence for a rich microstructure that forms in the CDW state. Data evidence metastability in the formation of $2\times 2\times 4$ and $2\times 2\times 2$ CDW supercells dependent on thermal history and mechanical deformation. We further directly resolve the real space phase segregation of both supercells as well as a real-space, structural twinning driven by the broken rotational symmetry of the CDW state. Our combined results provide insights into the role of microstructure and twinning in experiments probing the electronic properties of CsV$_3$Sb$_5$ where rotational symmetry is broken by the three-dimensional charge density wave order but locally preserved for any single kagome layer., Comment: 12 pages, 9 figures

In this paper we provide a computer assisted proof that about two thousand surgeries far away from the ideal point in the hyperbolic Dehn filling space of the figure-eight knot complement are infinitesimally projectively rigid. We also prove that for projective deformations of the figure-eight knot complement sufficiently close to the complete hyperbolic structure, the induced map on the first cohomology of the longitude of the boundary torus is non-zero. This paper provides a complementary piece to the results of Heusener and Porti who showed that for each k in Z, there is a sufficiently large Nk for which every k/n-Dehn filling on the figure-eight knot complement for n larger than Nk is infinitesimally projectively rigid. In the process of the proof, we provide explicit representations of the figure-eight knot complement in PSO(3,1) which are rational in the real and imaginary parts of the shapes of the ideal tetrahedra used to glue the knot complement together.

The reconstruction of electrical current densities from magnetic field measurements is an important technique with applications in materials science, circuit design, quality control, plasma physics, and biology. Analytic reconstruction methods exist for planar currents, but break down in the presence of high spatial frequency noise or large standoff distance, restricting the types of systems that can be studied. Here, we demonstrate the use of a deep convolutional neural network for current density reconstruction from two-dimensional (2D) images of vector magnetic fields acquired by a quantum diamond microscope (QDM) utilizing a surface layer of Nitrogen Vacancy (NV) centers in diamond. Trained network performance significantly exceeds analytic reconstruction for data with high noise or large standoff distances. This machine learning technique can perform quality inversions on lower SNR data, reducing the data collection time by a factor of about 400 and permitting reconstructions of weaker and three-dimensional current sources., Comment: 17 pages, 10 figures. Includes Supplemental Information

Background/Objective: Vitamin D deficiency (serum 25-hydroxyvitamin D [25(OH)D] concentration <50 nmol/L) is prevalent among Aboriginal and Torres Strait Islander peoples in Australia. Alternative to sun exposure (the primary source of vitamin D), vitamin D can also be obtained from food (e.g., fish, eggs, and meat) and supplements. However, vitamin D intake among Aboriginal and Torres Strait Islander peoples is currently unknown. We aimed to provide the first quantification of vitamin D intake using nationally representative data from Aboriginal and Torres Strait Islander peoples. Methods: We used food consumption data collected in the 2012-2013 National Aboriginal and Torres Strait Islander Nutrition and Physical Activity Survey (n = 4,109) and vitamin D food composition data to quantify mean absolute vitamin D intake by sex, age group, and remoteness of location. Differences in mean vitamin D intake between sexes and between remoteness of location were assessed using the 95% confidence interval (95% CI). Results: The mean (standard deviation (SD)) vitamin D intake among Aboriginal and Torres Strait Islander peoples was 2.9 (3.0) {\mu}g/day. Males had a statistically significantly higher mean (SD) [95% CI] vitamin D intake (3.2 (3.1) [3.0-3.4] {\mu}g/day) than females (2.6 (2.7) [2.4-2.7] {\mu}g/day). There were no statistically significant differences between mean (SD) [95% CI] vitamin D intake in non-remote (2.9 (2.2) [2.7-3.1] {\mu}g/day) and remote areas (2.8 (4.8) [2.6-3.0] {\mu}g/day). Conclusions: Vitamin D intake among Aboriginal and Torres Strait Islander peoples is low. Food-based public health strategies could be developed to promote higher vitamin D intake among this population.

Traditional reference-based metrics, such as BLEU and ROUGE, are less effective for assessing outputs from Large Language Models (LLMs) that produce highly creative or superior-quality text, or in situations where reference outputs are unavailable. While human evaluation remains an option, it is costly and difficult to scale. Recent work using LLMs as evaluators (LLM-as-a-judge) is promising, but trust and reliability remain a significant concern. Integrating human input is crucial to ensure criteria used to evaluate are aligned with the human's intent, and evaluations are robust and consistent. This paper presents a user study of a design exploration called EvaluLLM, that enables users to leverage LLMs as customizable judges, promoting human involvement to balance trust and cost-saving potential with caution. Through interviews with eight domain experts, we identified the need for assistance in developing effective evaluation criteria aligning the LLM-as-a-judge with practitioners' preferences and expectations. We offer findings and design recommendations to optimize human-assisted LLM-as-judge systems., Comment: 14 pages, 6 figures, Accepted for publication in ACL 2024 Workshop HuCLLM

We explore the spatial distribution of organics on Ceres using the visible and near-infrared data collected by the Dawn mission. We employ a spectral mixture analysis (SMA) approach to map organic materials within the Ernutet crater at the highest available spatial resolution revealing a discontinuous, granular distribution and a possible correlation with an ancient crater on which Ernutet has been superimposed. The SMA technique also helps us identify 11 new areas as potential sites for organics. These regions are predominantly located within craters or along their walls, resembling the distribution pattern observed in Ernutet, which implies a possible geological link with materials exposed from beneath the surface. In one of these candidate regions situated in the Yalode quadrangle, we detected the characteristic 3.4-micron absorption band in the infrared spectrum, indicative of organics and carbonates. By combining the spatial resolution of the Framing Camera data with the spectral resolution of the Visual and Infrared Imaging Spectrometer using SMA, we investigated the distribution of the 3.4-micron band in this quadrangle. The absorption pattern correlates with the Yalode/Urvara smooth material unit, which formed after significant impacts on Ceres. The association of organic-rich materials with complex and multiple large-impact events supports for an endogenous origin for the organics on Ceres.

Retrieval-augmented generation (RAG) has emerged as a promising solution to mitigate the limitations of large language models (LLMs), such as hallucinations and outdated information. However, it remains unclear how LLMs handle knowledge conflicts arising from different augmented retrieved passages, especially when these passages originate from the same source and have equal trustworthiness. In this work, we conduct a comprehensive evaluation of LLM-generated answers to questions that have varying answers based on contradictory passages from Wikipedia, a dataset widely regarded as a high-quality pre-training resource for most LLMs. Specifically, we introduce WikiContradict, a benchmark consisting of 253 high-quality, human-annotated instances designed to assess LLM performance when augmented with retrieved passages containing real-world knowledge conflicts. We benchmark a diverse range of both closed and open-source LLMs under different QA scenarios, including RAG with a single passage, and RAG with 2 contradictory passages. Through rigorous human evaluations on a subset of WikiContradict instances involving 5 LLMs and over 3,500 judgements, we shed light on the behaviour and limitations of these models. For instance, when provided with two passages containing contradictory facts, all models struggle to generate answers that accurately reflect the conflicting nature of the context, especially for implicit conflicts requiring reasoning. Since human evaluation is costly, we also introduce an automated model that estimates LLM performance using a strong open-source language model, achieving an F-score of 0.8. Using this automated metric, we evaluate more than 1,500 answers from seven LLMs across all WikiContradict instances. To facilitate future work, we release WikiContradict on: https://ibm.biz/wikicontradict.

We explore the ability of GPT-4 to perform ad-hoc schema based information extraction from scientific literature. We assess specifically whether it can, with a basic prompting approach, replicate two existing material science datasets, given the manuscripts from which they were originally manually extracted. We employ materials scientists to perform a detailed manual error analysis to assess where the model struggles to faithfully extract the desired information, and draw on their insights to suggest research directions to address this broadly important task.

As hydrogen is touted as a key player in the decarbonization of modern society, it is critical to enable quantitative H analysis at high spatial resolution, if possible at the atomic scale. Indeed, H has a known deleterious impact on the mechanical properties (strength, ductility, toughness) of most materials that can hinder their use as part of the infrastructure of a hydrogen-based economy. Enabling H mapping, including local hydrogen concentration analyses at specific microstructural features, is essential for understanding the multiple ways that H affect the properties of materials, including for instance embrittlement mechanisms and their synergies, but also spatial mapping and quantification of hydrogen isotopes is essential to accurately predict tritium inventory of future fusion power plants, ensuring their safe and efficient operation for example. Atom probe tomography (APT) has the intrinsic capabilities for detecting hydrogen (H), and deuterium (D), and in principle the capacity for performing quantitative mapping of H within a material's microstructure. Yet the accuracy and precision of H analysis by APT remain affected by the influence of residual hydrogen from the ultra-high vacuum chamber that can obscure the signal of H from within the material, along with a complex field evaporation behavior. The present article reports the essence of discussions at a focused workshop held at the Max-Planck Institute for Sustainable Materials in April 2024. The workshop was organized to pave the way to establishing best practices in reporting APT data for the analysis of H. We first summarize the key aspects of the intricacies of H analysis by APT and propose a path for better reporting of the relevant data to support interpretation of APT-based H analysis in materials.

Several long-period binaries with a carbon-rich Wolf-Rayet star and an O star produce dust in their wind collisions. In eccentric binaries, this is seen most strongly near periastron passage. The exact conditions leading to dust creation require orbital properties to be determined, which is difficult owing to their long periods. Recently, the binary system WR 125 (WC7+O9III) began a dust creation episode seen through an infrared outburst first detected by NEOWISE-R, which was the first outburst detected since 1991. We present new near- and mid-infrared photometry, which we use to show consistency between the two outbursts and derive an orbital period of 28.12$^{+0.10}_{-0.05}$ yr. We use a long time-series of optical spectra to place the first constraints on its orbital elements, on the assumption that this system will produce dust near periastron. The orbit has a mild eccentricity of 0.29$\pm$0.12 and is only derived for the Wolf-Rayet component, as the O star's radial velocities have noise that is likely larger than the expected semi-amplitude of the orbit. We also present SOFIA/FORCAST grism spectroscopy to examine the infrared spectral energy distribution (SED) of the dust during this outburst, comparing its properties to other WCd binaries, deriving a dust temperature of 580 K in 2021. This collection of observations will allow us to plan future observations of this system and place the system in the context of dust-creating Wolf-Rayet binaries., Comment: accepted to ApJ

Compiling programs to an instruction set architecture (ISA) requires a set of rewrite rules that map patterns consisting of compiler instructions to patterns consisting of ISA instructions. We synthesize such rules by constructing SMT queries, whose solutions represent two functionally equivalent programs. These two programs are interpreted as an instruction selection rewrite rule. Existing work is limited to single-instruction ISA patterns, whereas our solution does not have that restriction. Furthermore, we address inefficiencies of existing work by developing two optimized algorithms. The first only generates unique rules by preventing synthesis of duplicate and composite rules. The second only generates lowest-cost rules by preventing synthesis of higher-cost rules. We evaluate our algorithms on multiple ISAs. Without our optimizations, the vast majority of synthesized rewrite rules are either duplicates, composites, or higher cost. Our optimizations result in synthesis speed-ups of up to 768x and 4004x for the two algorithms., Comment: 9 pages, 2 figures, in submission to FMCAD24, fixed small bugs in results generation, improved baseline comparison, improved background and writing clarity

Generative Artificial Intelligence (AI) is integrated into everyday technology, including news, education, and social media. AI has further pervaded private conversations as conversational partners, auto-completion, and response suggestions. As social media becomes young people's main method of peer support exchange, we need to understand when and how AI can facilitate and assist in such exchanges in a beneficial, safe, and socially appropriate way. We asked 622 young people to complete an online survey and evaluate blinded human- and AI-generated responses to help-seeking messages. We found that participants preferred the AI-generated response to situations about relationships, self-expression, and physical health. However, when addressing a sensitive topic, like suicidal thoughts, young people preferred the human response. We also discuss the role of training in online peer support exchange and its implications for supporting young people's well-being. Disclaimer: This paper includes sensitive topics, including suicide ideation. Reader discretion is advised.

Objective: Diet quality is important for people with multiple sclerosis (MS), but conflicting online information causes them confusion. People with MS want evidence-based MS-specific information to help them make healthy dietary changes, and we co-designed an asynchronous, online nutrition education program (Eating Well with MS) with the MS community. Our aim was to determine the feasibility of Eating Well with MS. Methods: We used a single-arm pre-post design. The feasibility trial was a nine-week intervention with adults with confirmed MS. Feasibility outcomes: 1) demand (recruitment); 2) practicality (completion); 3) acceptability (Intrinsic Motivation Inventory: interest/enjoyment and value/usefulness subscales); and 4) limited efficacy testing (Diet Habits Questionnaire (DHQ); Critical Nutrition Literacy Tool (CNLT); Food Literacy Behaviour Checklist (FLBC)). Results: The recruitment target (n=70) was reached. 87% completed at least one module and 57% completed the full program (five modules). The median interest/enjoyment rating was 5 out of 7 and median value/usefulness rating was 6 out of 7 (where 7 = very true). Compared to pre-program, participants who completed any of the program had statistically significantly improved DHQ, CNLT, and FLBC scores. Conclusion: Eating Well with MS was well received by the MS community and improved their dietary behaviours; demonstrating feasibility. Our findings support the use of co-design methods when developing resources to improve dietary behaviours.

Federated Learning and Analytics (FLA) have seen widespread adoption by technology platforms for processing sensitive on-device data. However, basic FLA systems have privacy limitations: they do not necessarily require anonymization mechanisms like differential privacy (DP), and provide limited protections against a potentially malicious service provider. Adding DP to a basic FLA system currently requires either adding excessive noise to each device's updates, or assuming an honest service provider that correctly implements the mechanism and only uses the privatized outputs. Secure multiparty computation (SMPC) -based oblivious aggregations can limit the service provider's access to individual user updates and improve DP tradeoffs, but the tradeoffs are still suboptimal, and they suffer from scalability challenges and susceptibility to Sybil attacks. This paper introduces a novel system architecture that leverages trusted execution environments (TEEs) and open-sourcing to both ensure confidentiality of server-side computations and provide externally verifiable privacy properties, bolstering the robustness and trustworthiness of private federated computations.

Atmospheric interaction with light has been an area of fascination for many researchers over the last century. Environmental conditions, such as temperature and wind speed, heavily influence the complex and rapidly varying optical distortions propagating optical fields experience. The continuous random phase fluctuations commonly make deciphering the exact origins of specific optical aberrations challenging. The generation of eddies is a major contributor to atmospheric turbulence, similar in geometric structure to optical vortices that sit at the centre of OAM beams. Decomposing the received optical fields into OAM provides a unique spatial similarity that can be used to analyse turbulent channels. In this work, we present a novel mode decomposition assisted machine learning approach that reveals trainable features in the distortions of vortex beams that allow for effective environmental monitoring. This novel technique can be used reliably with Support Vector Machine regression models to measure temperature variations of 0.49C and wind speed variations of 0.029 m/s over a 36m experimental turbulent free-space channels with controllable and verifiable temperature and wind speed with short 3s measurement. The predictable nature of these findings could indicate the presence of an underlying physical relationship between environmental conditions that lead to specific eddy formation and the OAM spiral spectra.

Modern language models, while sophisticated, exhibit some inherent shortcomings, particularly in conversational settings. We claim that many of the observed shortcomings can be attributed to violation of one or more conversational principles. By drawing upon extensive research from both the social science and AI communities, we propose a set of maxims -- quantity, quality, relevance, manner, benevolence, and transparency -- for describing effective human-AI conversation. We first justify the applicability of the first four maxims (from Grice) in the context of human-AI interactions. We then argue that two new maxims, benevolence (concerning the generation of, and engagement with, harmful content) and transparency (concerning recognition of one's knowledge boundaries, operational constraints, and intents), are necessary for addressing behavior unique to modern human-AI interactions. We evaluate the degree to which various language models are able to understand these maxims and find that models possess an internal prioritization of principles that can significantly impact their ability to interpret the maxims accurately.

Large language models (LLMs) are susceptible to a variety of risks, from non-faithful output to biased and toxic generations. Due to several limiting factors surrounding LLMs (training cost, API access, data availability, etc.), it may not always be feasible to impose direct safety constraints on a deployed model. Therefore, an efficient and reliable alternative is required. To this end, we present our ongoing efforts to create and deploy a library of detectors: compact and easy-to-build classification models that provide labels for various harms. In addition to the detectors themselves, we discuss a wide range of uses for these detector models - from acting as guardrails to enabling effective AI governance. We also deep dive into inherent challenges in their development and discuss future work aimed at making the detectors more reliable and broadening their scope.

On September 26, 2022, NASA's Double Asteroid Redirection Test (DART) mission successfully impacted Dimorphos, the natural satellite of the binary near-Earth asteroid (65803) Didymos. Numerical simulations of the impact provide a means to explore target surface material properties and structures, consistent with the observed momentum deflection efficiency, ejecta cone geometry, and ejected mass. Our simulation, which best matches observations, indicates that Dimorphos is weak, with a cohesive strength of less than a few pascals (Pa), similar to asteroids (162173) Ryugu and (101955) Bennu. We find that a bulk density of Dimorphos, rhoB, lower than 2400 kg/m3, and a low volume fraction of boulders (<40 vol%) on the surface and in the shallow subsurface, are consistent with measured data from the DART experiment. These findings suggest Dimorphos is a rubble pile that might have formed through rotational mass shedding and re-accumulation from Didymos. Our simulations indicate that the DART impact caused global deformation and resurfacing of Dimorphos. ESA's upcoming Hera mission may find a re-shaped asteroid, rather than a well-defined crater.