Search

Recent data from Chicago Public Schools (CPS) show that active English Learners were less likely to graduate from high school (77% did so in 2023) than their peers (85%) and less likely to enroll in college (56% of 2022 English Learners high school graduates enrolled in college compared to 66 percent of their non-English Learners high school graduate peers), drawing attention to what supports these students need and what barriers they face on the road to college. To understand the academic performance in high school, college enrollment, and college persistence of English Learners, this study highlights metrics for former English Learners and different groups of active English Learners (at the time of enrollment in ninth grade). The focus of this study is on understanding different indicators of academic performance in high school, college enrollment, and college persistence of English Learners--including variation in attainment among active and former English Learners--to support their path to high school graduation and access to college. Using three cohorts of students who were first-time ninth-graders in the fall of 2014, 2015, and 2016 where 33% of students were English Learners, the authors conclude that English Learners are not perpetually struggling, as publicly reported numbers suggest, but specific groups of English Learners need more support. There are some commonalities among English Learners, but also some key differences, suggesting that there are different strategies that would be effective in supporting different English Learners groups. [This report was co-produced by the University of Chicago Crown Family School of Social Work, Policy, and Practice.]

We report the successful resolution of the multiplet structure of the K{\alpha}1 x-ray emission in manganese (Mn) complexes through seeded stimulated X-ray emission spectroscopy (seeded S-XES). By employing a femtosecond pump pulse above the Mn K edge to generate simultaneous 1s core-holes, and a second-color tunable seed pulse to initiate the stimulated emission process, we were able to enhance individual lines within the K{\alpha}1 emission. This approach allows to resolve the fine multiplet features that are obscured by the life-time broadening in conventional Mn K{\alpha} XES. The work builds on our previous observation that S-XES from Mn(II) and Mn(VII) complexes pumped at high intensities can exhibit stimulated emission without sacrificing the chemical sensitivity to oxidation states. This technique opens the door to controlled high-resolution electronic structure spectroscopy in transition metal complexes beyond core hole life time broadening with potential applications in catalysis, inorganic chemistry, and materials science., Comment: 6 pager, 5 figures

One of the most prolific methods of studying exoplanet atmospheres is transmission spectroscopy, which measures the difference between the depth of an exoplanet's transit signal at various wavelengths and attempts to correlate the depth changes to potential features in the exoplanet's atmosphere. Here we present reconnaissance observations of 21 exoplanet atmospheres measured with the Exoplanet Transmission Spectroscopy Imager (ETSI), a recently deployed spectro-photometer on the McDonald Observatory Otto Struve 2.1 m telescope. ETSI measurements are mostly free of systematics through the use of a novel observing technique called common-path multi-band imaging (CMI), which has been shown to achieve photometric color precision on-par with space-based observations (300ppm or 0.03%). This work also describes the various statistical tests performed on the data to evaluate the efficacy of the CMI method and the ETSI instrument in combination. We find that none of the 8 comparisons of exoplanet atmospheres measured with ETSI and other observatories (including the Hubble Space Telescope) provide evidence that the spectra are statistically dissimilar. These results suggest that ETSI can provide initial transmission spectroscopy observations for a fraction of the observational and monetary overhead previously required to detect an exoplanet's atmosphere. Ultimately these reconnaissance observations increase the number of planets with transmission spectroscopy measurements by ~10% and provide an immediate prioritization of 21 exoplanets for future follow-up with more precious observatories, such as the James Webb Space Telescope. The reconnaissance spectra are available through the Filtergraph visualization portal at the URL: https://filtergraph.com/etsi/., Comment: 28 pages, 8 figures, 11 tables, article published in AJ. ETSI data is available on request and spectra can be found at the URL https://filtergraph.com/etsi/

Multimodal representation learning has demonstrated remarkable potential in enabling models to process and integrate diverse data modalities, such as text and images, for improved understanding and performance. While the medical domain can benefit significantly from this paradigm, the scarcity of paired multimodal data and reliance on proprietary or pretrained encoders pose significant challenges. In this work, we present a shared encoder framework for multimodal representation learning tailored to the medical domain. Our approach employs a single set of encoder parameters shared across modalities, augmented with learnable modality features. Empirical results demonstrate that our shared encoder idea achieves superior performance compared to separate modality-specific encoders, demonstrating improved generalization in data-constrained settings. Notably, the performance gains are more pronounced with fewer training examples, underscoring the efficiency of our shared encoder framework for real-world medical applications with limited data. Our code and experiment setup are available at https://github.com/VectorInstitute/shared_encoder.

The rapid development of advanced AI agents and the imminent deployment of many instances of these agents will give rise to multi-agent systems of unprecedented complexity. These systems pose novel and under-explored risks. In this report, we provide a structured taxonomy of these risks by identifying three key failure modes (miscoordination, conflict, and collusion) based on agents' incentives, as well as seven key risk factors (information asymmetries, network effects, selection pressures, destabilising dynamics, commitment problems, emergent agency, and multi-agent security) that can underpin them. We highlight several important instances of each risk, as well as promising directions to help mitigate them. By anchoring our analysis in a range of real-world examples and experimental evidence, we illustrate the distinct challenges posed by multi-agent systems and their implications for the safety, governance, and ethics of advanced AI., Comment: Cooperative AI Foundation, Technical Report #1

Anomaly detection (AD) is a fundamental task for time-series analytics with important implications for the downstream performance of many applications. In contrast to other domains where AD mainly focuses on point-based anomalies (i.e., outliers in standalone observations), AD for time series is also concerned with range-based anomalies (i.e., outliers spanning multiple observations). Nevertheless, it is common to use traditional point-based information retrieval measures, such as Precision, Recall, and F-score, to assess the quality of methods by thresholding the anomaly score to mark each point as an anomaly or not. However, mapping discrete labels into continuous data introduces unavoidable shortcomings, complicating the evaluation of range-based anomalies. Notably, the choice of evaluation measure may significantly bias the experimental outcome. Despite over six decades of attention, there has never been a large-scale systematic quantitative and qualitative analysis of time-series AD evaluation measures. This paper extensively evaluates quality measures for time-series AD to assess their robustness under noise, misalignments, and different anomaly cardinality ratios. Our results indicate that measures producing quality values independently of a threshold (i.e., AUC-ROC and AUC-PR) are more suitable for time-series AD. Motivated by this observation, we first extend the AUC-based measures to account for range-based anomalies. Then, we introduce a new family of parameter-free and threshold-independent measures, Volume Under the Surface (VUS), to evaluate methods while varying parameters. We also introduce two optimized implementations for VUS that reduce significantly the execution time of the initial implementation. Our findings demonstrate that our four measures are significantly more robust in assessing the quality of time-series AD methods.

Fingerspelling is a critical part of American Sign Language (ASL) recognition and has become an accessible optional text entry method for Deaf and Hard of Hearing (DHH) individuals. In this paper, we introduce SpellRing, a single smart ring worn on the thumb that recognizes words continuously fingerspelled in ASL. SpellRing uses active acoustic sensing (via a microphone and speaker) and an inertial measurement unit (IMU) to track handshape and movement, which are processed through a deep learning algorithm using Connectionist Temporal Classification (CTC) loss. We evaluated the system with 20 ASL signers (13 fluent and 7 learners), using the MacKenzie-Soukoref Phrase Set of 1,164 words and 100 phrases. Offline evaluation yielded top-1 and top-5 word recognition accuracies of 82.45% (9.67%) and 92.42% (5.70%), respectively. In real-time, the system achieved a word error rate (WER) of 0.099 (0.039) on the phrases. Based on these results, we discuss key lessons and design implications for future minimally obtrusive ASL recognition wearables.

As blind and low-vision (BLV) players engage more deeply with games, accessibility features have become essential. While some research has explored tools and strategies to enhance game accessibility, the specific experiences of these players with mobile games remain underexamined. This study addresses this gap by investigating how BLV users experience mobile games with varying accessibility levels. Through interviews with 32 experienced BLV mobile players, we explore their perceptions, challenges, and strategies for engaging with mobile games. Our findings reveal that BLV players turn to mobile games to alleviate boredom, achieve a sense of accomplishment, and build social connections, but face barriers depending on the game's accessibility level. We also compare mobile games to other forms of gaming, highlighting the relative advantages of mobile games, such as the inherent accessibility of smartphones. This study contributes to understanding BLV mobile gaming experiences and provides insights for enhancing accessible mobile game design., Comment: 18 pages, 3 figures, Accepted by CHI '25

The Pandora Software Development Kit and algorithm libraries perform reconstruction of neutrino interactions in liquid argon time projection chamber detectors. Pandora is the primary event reconstruction software used at the Deep Underground Neutrino Experiment, which will operate four large-scale liquid argon time projection chambers at the far detector site in South Dakota, producing high-resolution images of charged particles emerging from neutrino interactions. While these high-resolution images provide excellent opportunities for physics, the complex topologies require sophisticated pattern recognition capabilities to interpret signals from the detectors as physically meaningful objects that form the inputs to physics analyses. A critical component is the identification of the neutrino interaction vertex. Subsequent reconstruction algorithms use this location to identify the individual primary particles and ensure they each result in a separate reconstructed particle. A new vertex-finding procedure described in this article integrates a U-ResNet neural network performing hit-level classification into the multi-algorithm approach used by Pandora to identify the neutrino interaction vertex. The machine learning solution is seamlessly integrated into a chain of pattern-recognition algorithms. The technique substantially outperforms the previous BDT-based solution, with a more than 20\% increase in the efficiency of sub-1\,cm vertex reconstruction across all neutrino flavours., Comment: 32 pages, 18 figures

We present an outcome-driven fine-tuning framework that enhances the forecasting capabilities of large language models (LLMs) without relying on human-curated reasoning samples. Our method leverages model self-play to generate pairs of diverse reasoning trajectories and probabilistic forecasts for a set of diverse questions that resolve after the models' knowledge cutoff date. We then rank pairs of these reasoning traces by their distance to the actual outcomes before fine-tuning the model via Direct Preference Optimization (DPO). On a separate test set, our approach increases prediction accuracy of Phi-4 14B and DeepSeek-R1 14B by between 7--10\% over a base model and a DPO fine-tuned control model with randomized labels, bringing them on par with forecasting capabilities of much larger frontier models like GPT-4o.

Recent work showed Best-of-N (BoN) jailbreaking using repeated use of random augmentations (such as capitalization, punctuation, etc) is effective against all major large language models (LLMs). We have found that $100\%$ of the BoN paper's successful jailbreaks (confidence interval $[99.65\%, 100.00\%]$) and $99.8\%$ of successful jailbreaks in our replication (confidence interval $[99.28\%, 99.98\%]$) were blocked with our Defense Against The Dark Prompts (DATDP) method. The DATDP algorithm works by repeatedly utilizing an evaluation LLM to evaluate a prompt for dangerous or manipulative behaviors--unlike some other approaches, DATDP also explicitly looks for jailbreaking attempts--until a robust safety rating is generated. This success persisted even when utilizing smaller LLMs to power the evaluation (Claude and LLaMa-3-8B-instruct proved almost equally capable). These results show that, though language models are sensitive to seemingly innocuous changes to inputs, they seem also capable of successfully evaluating the dangers of these inputs. Versions of DATDP can therefore be added cheaply to generative AI systems to produce an immediate significant increase in safety.

This paper argues that model-free reinforcement learning (RL) agents, while lacking explicit planning mechanisms, exhibit behaviours that can be analogised to System 1 ("thinking fast") processes in human cognition. Unlike model-based RL agents, which operate akin to System 2 ("thinking slow") reasoning by leveraging internal representations for planning, model-free agents react to environmental stimuli without anticipatory modelling. We propose a novel framework linking the dichotomy of System 1 and System 2 to the distinction between model-free and model-based RL. This framing challenges the prevailing assumption that intentionality and purposeful behaviour require planning, suggesting instead that intentionality can manifest in the structured, reactive behaviours of model-free agents. By drawing on interdisciplinary insights from cognitive psychology, legal theory, and experimental jurisprudence, we explore the implications of this perspective for attributing responsibility and ensuring AI safety. These insights advocate for a broader, contextually informed interpretation of intentionality in RL systems, with implications for their ethical deployment and regulation.

Cochlear implants (CIs) are devices that restore the sense of hearing in people with severe sensorineural hearing loss. An electrode array inserted in the cochlea bypasses the natural transducer mechanism that transforms mechanical sound waves into neural activity by artificially stimulating the auditory nerve fibers with electrical pulses. The perception of sounds is possible because the brain extracts features from this neural activity, and loudness is among the most fundamental perceptual features. A computational model that uses a three-dimensional (3D) representation of the peripheral auditory system of CI users was developed to predict categorical loudness from the simulated peripheral neural activity. In contrast, current state-of-the-art computational loudness models predict loudness from the electrical pulses with minimal parametrization of the electrode-nerve interface. In the proposed model, the spikes produced in a population of auditory nerve fibers were grouped by cochlear places, a physiological representation of the auditory filters in psychoacoustics, to be transformed into loudness contribution. Then, a loudness index was obtained with a spatiotemporal integration over this loudness contribution. This index served to define the simulated threshold of hearing (THL) and most comfortable loudness (MCL) levels resembling the growth function in CI users. The performance of real CI users in loudness summation experiments was also used to validate the computational model. These experiments studied the effect of stimulation rate, electrode separation and amplitude modulation. The proposed model provides a new set of perceptual features that can be used in computational frameworks for CIs and narrows the gap between simulations and the human peripheral neural activity., Comment: Preprint

This paper argues that market governance mechanisms should be considered a key approach in the governance of artificial intelligence (AI), alongside traditional regulatory frameworks. While current governance approaches have predominantly focused on regulation, we contend that market-based mechanisms offer effective incentives for responsible AI development. We examine four emerging vectors of market governance: insurance, auditing, procurement, and due diligence, demonstrating how these mechanisms can affirm the relationship between AI risk and financial risk while addressing capital allocation inefficiencies. While we do not claim that market forces alone can adequately protect societal interests, we maintain that standardised AI disclosures and market mechanisms can create powerful incentives for safe and responsible AI development. This paper urges regulators, economists, and machine learning researchers to investigate and implement market-based approaches to AI governance.

The modern Everett interpretation of quantum mechanics describes an emergent multiverse. The goal of this paper is to provide a perspicuous characterisation of how the multiverse emerges making use of a recent account of (weak) ontological emergence. This will be cashed out with a case study that identifies decoherence as the mechanism for emergence. The greater metaphysical clarity enables the rebuttal of critiques due to Baker (2007) and Dawid and Th\'ebault (2015) that cast the emergent multiverse ontology as incoherent; responses are also offered to challenges to the Everettian approach from Maudlin (2010) and Monton (2013)., Comment: 22 pages, 1 figure

Dunes are bedforms found on sandy terrains shaped by fluid flow on Earth, Mars, and other celestial bodies. Despite their prevalence, understanding dune dynamics at the grain scale is challenging due to the vast number of grains involved. In this study, we demonstrate a novel approach to estimate the forces acting on individual dune grains using images. By combining subaqueous experiments, high-speed camera recordings, discrete numerical simulations, and a specially trained convolutional neural network, we can quantify these forces with high accuracy. This method represents a breakthrough in studying granular dynamics, offering a new way to measure forces not only on dune grains but also on smaller objects, such as rocks, boulders, rovers, and man-made structures, observed in satellite images of both Earth and Mars. This technique expands our ability to analyze and understand fluid-grain interactions in diverse environments.

In online educational platforms, adaptive experiment designs play a critical role in personalizing learning pathways, instructional sequencing, and content recommendations. Traditional adaptive policies, such as Thompson Sampling, struggle with scalability in high-dimensional and sparse settings such as when there are large amount of treatments (arms) and limited resources such as funding and time to conduct to a classroom constraint student size. Furthermore, the issue of under-exploration in large-scale educational interventions can lead to suboptimal learning recommendations. To address these challenges, we build upon the concept of lenient regret, which tolerates limited suboptimal selections to enhance exploratory learning, and propose a framework for determining the feasible number of treatments given a sample size. We illustrate these ideas with a case study in online educational learnersourcing examples, where adaptive algorithms dynamically allocate peer-crafted interventions to other students under active recall exercise. Our proposed Weighted Allocation Probability Adjusted Thompson Sampling (WAPTS) algorithm enhances the efficiency of treatment allocation by adjusting sampling weights to balance exploration and exploitation in data-sparse environments. We present comparative evaluations of WAPTS across various sample sizes (N=50, 300, 1000) and treatment conditions, demonstrating its ability to mitigate under-exploration while optimizing learning outcomes.

Vision generation remains a challenging frontier in artificial intelligence, requiring seamless integration of visual understanding and generative capabilities. In this paper, we propose a novel framework, Vision-Driven Prompt Optimization (VDPO), that leverages Large Language Models (LLMs) to dynamically generate textual prompts from visual inputs, guiding high-fidelity image synthesis. VDPO combines a visual embedding prompt tuner, a textual instruction generator, and a vision generation module to achieve state-of-the-art performance in diverse vision generation tasks. Extensive experiments on benchmarks such as COCO and Sketchy demonstrate that VDPO consistently outperforms existing methods, achieving significant improvements in FID, LPIPS, and BLEU/CIDEr scores. Additional analyses reveal the scalability, robustness, and generalization capabilities of VDPO, making it a versatile solution for in-domain and out-of-domain tasks. Human evaluations further validate the practical superiority of VDPO in generating visually appealing and semantically coherent outputs.

National context: Schools across the United States have long grappled with the role and impact of school-based police officers, often referred to as school resource officers (SROs). Proponents for school-based policing believe that SROs contribute to school safety by preventing or addressing crime and violence in schools. Opponents of SROs in schools argue that the presence of SROs criminalizes students and increases the likelihood of school-based arrest, particularly for students of color. Policies around SROs vary in districts across the country. Chicago context: In the wake of George Floyd's murder in 2020, the Chicago Board of Education (CBOE) asked the district to develop a plan to phase-out the SRO program, which assigned two SROs to most Chicago Public Schools (CPS) high schools. In February 2024, the CBOE voted to remove all remaining SROs from schools starting in 2024-25. In May 2024, CPS proposed a Whole School Safety Policy that did not include SROs and focused on supports for physical safety, emotional safety, and relational trust in schools. The research: This brief examines what happened when Chicago Public Schools (CPS) began the process of removing School Resource Officers (SROs) from its high schools during the 2020-21 school year. These findings can inform conversations in Chicago, and across the country, about SROs and whole school safety practices and policies--while recognizing that questions about the presence of police and the experience of safety in schools are complex and reflect differences in lived experiences and perspectives across policymakers, practitioners, young people, families, and communities. The findings are part of a larger, ongoing study by researchers from the UChicago Consortium; University of Illinois, Urbana-Champaign; and Lurie Children's Hospital's Center for Childhood Resilience. [This research brief received additional support from the Pritzker-Pucker Family Foundation and the Consortium Investor Council.]

We present a novel method that extends the self-attention mechanism of a vision transformer (ViT) for more accurate object detection across diverse datasets. ViTs show strong capability for image understanding tasks such as object detection, segmentation, and classification. This is due in part to their ability to leverage global information from interactions among visual tokens. However, the self-attention mechanism in ViTs are limited because they do not allow visual tokens to exchange local or global information with neighboring features before computing global attention. This is problematic because tokens are treated in isolation when attending (matching) to other tokens, and valuable spatial relationships are overlooked. This isolation is further compounded by dot-product similarity operations that make tokens from different semantic classes appear visually similar. To address these limitations, we introduce two modifications to the traditional self-attention framework; a novel aggressive convolution pooling strategy for local feature mixing, and a new conceptual attention transformation to facilitate interaction and feature exchange between semantic concepts. Experimental results demonstrate that local and global information exchange among visual features before self-attention significantly improves performance on challenging object detection tasks and generalizes across multiple benchmark datasets and challenging medical datasets. We publish source code and a novel dataset of cancerous tumors (chimeric cell clusters)., Comment: 20 Pages, 24 figures

The integration of Large Language Models (LLMs) and Vision-Language Models (VLMs) opens new avenues for addressing complex challenges in multimodal content analysis, particularly in biased news detection. This study introduces VLBias, a framework that leverages state-of-the-art LLMs and VLMs to detect linguistic and visual biases in news content. We present a multimodal dataset comprising textual content and corresponding images from diverse news sources. We propose a hybrid annotation framework that combines LLM-based annotations with human review to ensure high-quality labeling while reducing costs and enhancing scalability. Our evaluation compares the performance of state-of-the-art SLMs and LLMs for both modalities (text and images) and the results reveal that while SLMs are computationally efficient, LLMs demonstrate superior accuracy in identifying subtle framing and text-visual inconsistencies. Furthermore, empirical analysis shows that incorporating visual cues alongside textual data improves bias detection accuracy by 3 to 5%. This study provides a comprehensive exploration of LLMs, SLMs, and VLMs as tools for detecting multimodal biases in news content and highlights their respective strengths, limitations, and potential for future applications, Comment: Under review

Monitoring animal populations is crucial for assessing the health of ecosystems. Traditional methods, which require extensive fieldwork, are increasingly being supplemented by time-lapse camera-trap imagery combined with an automatic analysis of the image data. The latter usually involves some object detector aimed at detecting relevant targets (commonly animals) in each image, followed by some postprocessing to gather activity and population data. In this paper, we show that the performance of an object detector in a single frame of a time-lapse sequence can be improved by including spatio-temporal features from the prior frames. We propose a method that leverages temporal information by integrating two additional spatial feature channels which capture stationary and non-stationary elements of the scene and consequently improve scene understanding and reduce the number of stationary false positives. The proposed technique achieves a significant improvement of 24\% in mean average precision (mAP@0.05:0.95) over the baseline (temporal feature-free, single frame) object detector on a large dataset of breeding tropical seabirds. We envisage our method will be widely applicable to other wildlife monitoring applications that use time-lapse imaging., Comment: 18 pages, 13 figures

As language models become integral to critical workflows, assessing their behavior remains a fundamental challenge -- human evaluation is costly and noisy, while automated metrics provide only coarse, difficult-to-interpret signals. We introduce natural language unit tests, a paradigm that decomposes response quality into explicit, testable criteria, along with a unified scoring model, LMUnit, which combines multi-objective training across preferences, direct ratings, and natural language rationales. Through controlled human studies, we show this paradigm significantly improves inter-annotator agreement and enables more effective LLM development workflows. LMUnit achieves state-of-the-art performance on evaluation benchmarks (FLASK, BigGenBench) and competitive results on RewardBench. These results validate both our proposed paradigm and scoring model, suggesting a promising path forward for language model evaluation and development.

Emotional voice communication plays a crucial role in effective daily interactions. Deaf and hard-of-hearing (DHH) individuals often rely on facial expressions to supplement sign language to convey emotions, as the use of voice is limited. However, in American Sign Language (ASL), these facial expressions serve not only emotional purposes but also as linguistic markers, altering sign meanings and often confusing non-signers when interpreting a signer's emotional state. Most existing ASL translation technologies focus solely on signs, neglecting the role of emotional facial expressions in the translated output (e.g., text, voice). This paper present studies which 1) confirmed the challenges for non-signers of interpreting emotions from facial expressions in ASL communication, of facial expressions, and 2) how integrating emotional voices into translation systems can enhance hearing individuals' comprehension of a signer's emotions. An online survey conducted with 45 hearing participants (Non-ASL Signers) revealed that they frequently misinterpret signers' emotions when emotional and linguistic facial expressions are used simultaneously. The findings indicate that incorporating emotional voice into translation systems significantly improves the recognition of signers' emotions by 32%. Additionally, further research involving 6 DHH participants discusses design considerations for the emotional voice feature from both perspectives, emphasizing the importance of integrating emotional voices in translation systems to bridge communication gaps between DHH and hearing communities.

Achieving high-throughput, comprehensive analysis of single nanoparticles to determine their size, shape, and composition is essential for understanding particle heterogeneity with applications ranging from drug delivery to environmental monitoring. Existing techniques are hindered by low throughput, lengthy trapping times, irreversible particle adsorption, or limited characterization capabilities. Here, we introduce Interferometric Electrohydrodynamic Tweezers (IET), an integrated platform that rapidly traps single nanoparticles in parallel within three seconds. IET enables label-free characterization of particle size and shape via interferometric imaging and identifies molecular composition through Raman spectroscopy, all without the need for fluorescent labeling. We demonstrate the platform's capabilities by trapping and imaging colloidal polymer beads, nanoscale extracellular vesicles (EVs), and newly discovered extracellular nanoparticles known as supermeres. By monitoring their interferometric contrast images while trapped, we accurately determine the sizes of EVs and supermeres. Our IET represents a powerful optofluidics platform for comprehensive characterization of nanoscale objects, opening new avenues in nanomedicine, environmental monitoring, and beyond.

The meeting patronized by Ernest Solvay in 1911, the first Solvay Council, marked the beginning of what can be called the first quantum revolution. Maurice de Broglie was one of the secretaries of the Council. He participated in the two following conferences and contributed to the third Council with his work. Louis de Broglie collaborated and discussed with his elder brother. Departing from Maurice's achievements, Louis developed his approach to the wave-particle duality of light and extended it to other particles. He presented it at the fifth meeting in 1927. This paper investigates the connection between the two scientists and the Solvay Councils, arguing that this link was essential in developing the first quantum revolution, a process that ended, according to Werner Heisenberg, at the fifth Solvay Physics Council., Comment: 46 pages - no figures

Pretrained foundation models and transformer architectures have driven the success of large language models (LLMs) and other modern AI breakthroughs. However, similar advancements in health data modeling remain limited due to the need for innovative adaptations. Wearable movement data offers a valuable avenue for exploration, as it's a core feature in nearly all commercial smartwatches, well established in clinical and mental health research, and the sequential nature of the data shares similarities to language. We introduce the Pretrained Actigraphy Transformer (PAT), the first open source foundation model designed for time-series wearable movement data. Leveraging transformer-based architectures and novel techniques, such as patch embeddings, and pretraining on data from 29,307 participants in a national U.S. sample, PAT achieves state-of-the-art performance in several mental health prediction tasks. PAT is also lightweight and easily interpretable, making it a robust tool for mental health research. GitHub: https://github.com/njacobsonlab/Pretrained-Actigraphy-Transformer/

When several intruders move in a granular medium, coupling effects are observed, the motion of one intruder affecting that of others. In this paper, we investigate experimentally how the drag forces acting on a pair of spherical intruders moving amid grains at constant velocity vary with the transverse separation between them and their depth. When intruders are sufficiently far apart, they do not influence each other, and the average drag felt by each of them matches that of a single intruder. However, for small distances between intruders and at a given depth, the average drag per intruder decreases, highlighting a collaborative effect that facilitates motion. This collaboration effect is amplified when the depth of the intruders increases. We propose a model for the drag reduction of a pair of intruders based on the breakup of contact chains, caused by the perturbation generated by the neighbor intruder. Our findings provide new insights into the interaction effects on the motion of solids in sand, such as those observed in animal locomotion, root growth, and soil survey.

In the realm of Large Language Models (LLMs), prompt optimization is crucial for model performance. Although previous research has explored aspects like rephrasing prompt contexts, using various prompting techniques (like in-context learning and chain-of-thought), and ordering few-shot examples, our understanding of LLM sensitivity to prompt templates remains limited. Therefore, this paper examines the impact of different prompt templates on LLM performance. We formatted the same contexts into various human-readable templates, including plain text, Markdown, JSON, and YAML, and evaluated their impact across tasks like natural language reasoning, code generation, and translation using OpenAI's GPT models. Experiments show that GPT-3.5-turbo's performance varies by up to 40\% in a code translation task depending on the prompt template, while larger models like GPT-4 are more robust to these variations. Our analysis highlights the need to reconsider the use of fixed prompt templates, as different formats can significantly affect model performance., Comment: Submitted to NAACL 2025

In a recent paper (Gonzalez et al., 2023), we investigated the motion of grains within a granular bed sheared by a viscous fluid, and showed how segregation and hardening occur in the fluid- (bedload) and solid-like (creep) regions. In this paper, we inquire further into the mechanisms leading to grain segregation in a bidisperse bed, and how the forces are distributed. For that, we carried out numerical simulations at the grain scale by using CFD-DEM (computational fluid dynamics-discrete element method), with which we were able to track the positions, velocities, forces, and solid contacts underwent by each grain. We show that during the upward motion of large grains the direct action of fluid forces is significant in the middle and upper parts of the bedload layer, while only contact forces are significant in the creep layer and lower part of the bedload layer. We also show that in all cases the particles experience a moment about a -45 degrees contact point (with respect to the horizontal plane) when migrating upward, whether entrained by other contacts or directly by the fluid. In addition, we show the variations in the average solid-solid contacts, and how forces caused either by solid-solid contacts or directly by the fluid are distributed within the bed. Our results provide the relationship between force propagation and reorganization of grains in sheared beds, explaining mechanisms found, for example, in river beds and landslides.

We inquire into the morphodynamics of barchans under seasonal flows. For that, we carried out grain-scale numerical computations of a subaqueous barchan exposed to two-directional flows, and we varied the angle and frequency of oscillations. We show that when the frequency is lower than the inverse of the characteristic time for barchan formation, the dune adapts to the new flow direction and recovers the barchan shape while losing less grains than under one-directional flow. For higher frequencies, the dune has not enough time for adaptation and becomes more round while losing more grains. For both cases, we show, for the first time, the typical dynamics of grains (trajectories and forces). In particular, the round barchans are similar to the so-called occluded dunes observed on Mars, where seasons have very high frequencies compared to the dune timescale, different from Earth. Our results represent a possible explanation for that shape., Comment: Geomorphology (2024). This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/

Multimodal models leverage large-scale pre-training to achieve strong but still imperfect performance on tasks such as image captioning, visual question answering, and cross-modal retrieval. In this paper, we present a simple and efficient method for correcting errors in trained contrastive image-text retrieval models with no additional training, called Nearest Neighbor Normalization (NNN). We show an improvement on retrieval metrics in both text retrieval and image retrieval for all of the contrastive models that we tested (CLIP, BLIP, ALBEF, SigLIP, BEiT) and for both of the datasets that we used (MS-COCO and Flickr30k). NNN requires a reference database, but does not require any training on this database, and can even increase the retrieval accuracy of a model after finetuning.

In response to numerous programs seeking to advance quantum education and workforce development in the United States, experts from academia, industry, government, and professional societies convened for a National Science Foundation-sponsored workshop in February 2024 to explore the benefits and challenges of establishing a national center for quantum education. Broadly, such a center would foster collaboration and build the infrastructure required to develop a diverse and quantum-ready workforce. The workshop discussions centered around how a center could uniquely address gaps in public, K-12, and undergraduate quantum information science and engineering (QISE) education. Specifically, the community identified activities that, through a center, could lead to an increase in student awareness of quantum careers, boost the number of educators trained in quantum-related subjects, strengthen pathways into quantum careers, enhance the understanding of the U.S. quantum workforce, and elevate public engagement with QISE. Core proposed activities for the center include professional development for educators, coordinated curriculum development and curation, expanded access to educational laboratory equipment, robust evaluation and assessment practices, network building, and enhanced public engagement with quantum science.

Offline paired preference optimization algorithms have become a popular approach for fine-tuning on preference data, outperforming traditional supervised fine-tuning in various tasks. However, traditional implementations often involve redundant computations, especially for tasks with long shared prompts. We introduce prefix sharing for preference tuning, a novel technique that processes chosen and rejected responses as one sequence with a shared prefix. To prevent cross-response contamination, we use a custom block-sparse attention mask. Our method achieves $1.1$-$1.5\times$ improvement in training throughput on popular DPO datasets, without any effect on convergence. When combined with sequence packing, we observe consistent $1.3$-$1.6\times$ speedups, benefiting even datasets with smaller sequence lengths. While we focus on Direct Preference Optimization (DPO), our approach is applicable to other paired preference tuning methods. By enhancing computational efficiency, our work contributes to making preference-based fine-tuning more accessible for a wider range of applications and model sizes. We open-source our code at https://github.com/frankxwang/dpo-prefix-sharing., Comment: To appear in NeurIPS 2024 in the Fine-Tuning in Machine Learning Workshop

This review explores the most recent advancements in transformer-based data fusion techniques in agricultural remote sensing (RS), with a particular focus on soil analysis. Utilizing a systematic, data-driven approach, we demonstrate that transformers have significantly outperformed conventional deep learning and machine learning methods since 2022, achieving prediction performance between 92% and 97%. The review is specifically focused on soil analysis, due to the importance of soil condition in optimizing crop productivity and ensuring sustainable farming practices. Transformer-based models have shown remarkable capabilities in handling complex multivariate soil data, improving the accuracy of soil moisture prediction, soil element analysis, and other soil-related applications. This systematic review primarily focuses on 1) analysing research trends and patterns in the literature, both chronologically and technically, and 2) conducting a comparative analysis of data fusion approaches, considering factors such as data types, techniques, and RS applications. Finally, we propose a roadmap for implementing data fusion methods in agricultural RS., Comment: 14 pages, 9 figures, 4 tables, Journal

Permutations are usually enumerated by size, but new results can be found by enumerating them by inversions instead, in which case one must restrict one's attention to indecomposable permutations. In the style of the seminal paper by Simion and Schmidt, we investigate all combinations of permutation patterns of length at most 3.

Quantum search algorithms are crucial for exploring large solution spaces, but their robustness to environmental perturbations, such as noise or disorder, remains a critical challenge. We examine the impact of biased disorder potentials modeled by a bivariate Gaussian distribution function on the dynamics of quantum walks in spatial search problems. Building on the Ambainis-Kempe-Rivosh (AKR) model for searching on a two-dimensional grid, we incorporate potential fields to investigate how changes in standard deviation and normalization of the bivariate Gaussian function impact the performance of the search algorithm. Our results show that the quantum walk closely mirrors the AKR algorithm when the standard deviation is small but exhibits a rapid decay in success probability as the standard deviation increases. This behavior demonstrates how the bivariate Gaussian can effectively model a noisy oracle within the AKR algorithm. Additionally, we compare the AKR-based model with an alternative quantum walk model using a Hadamard coin and standard shift. These findings contribute to understanding the robustness of quantum walk search algorithms, and provide insights into how quantum walks can be applied to optimization algorithms., Comment: 15 pages, 8 figures. Improved text, and added references