Search

We present a comprehensive analysis of the region containing the large smooth patch on comet Tempel 1, focusing on its spectral and morphological characteristics and those of its surroundings. Utilizing observational data from the Deep Impact and Stardust-NExT missions, an updated stereophotoclinometry-based shape model, and numerical simulations, we aim to investigate the origin and evolution of this feature. Our study characterizes the morphological changes between the two mission visits, determining that the smooth patch has a thickness of approximately 25 meters. This patch is embedded in a cliff with an average height of 50 meters and exhibits a lobate U-shape morphology. Our findings support the previously suggested idea that an ice flow phenomenon is compatible with the observations. Moreover, our simulations indicate that a single phenomenon could link both the large and secondary smooth patches and the mass wasting feature observed on the comet's opposite northern face. We propose that a cryovolcano-like event may be responsible for this smooth feature, although more evidence is needed to support this hypothesis.

Wearable sensors have become ubiquitous thanks to a variety of health tracking features. The resulting continuous and longitudinal measurements from everyday life generate large volumes of data; however, making sense of these observations for scientific and actionable insights is non-trivial. Inspired by the empirical success of generative modeling, where large neural networks learn powerful representations from vast amounts of text, image, video, or audio data, we investigate the scaling properties of sensor foundation models across compute, data, and model size. Using a dataset of up to 40 million hours of in-situ heart rate, heart rate variability, electrodermal activity, accelerometer, skin temperature, and altimeter per-minute data from over 165,000 people, we create LSM, a multimodal foundation model built on the largest wearable-signals dataset with the most extensive range of sensor modalities to date. Our results establish the scaling laws of LSM for tasks such as imputation, interpolation and extrapolation, both across time and sensor modalities. Moreover, we highlight how LSM enables sample-efficient downstream learning for tasks like exercise and activity recognition.

Modular neural networks outperform nonmodular neural networks on tasks ranging from visual question answering to robotics. These performance improvements are thought to be due to modular networks' superior ability to model the compositional and combinatorial structure of real-world problems. However, a theoretical explanation of how modularity improves generalizability, and how to leverage task modularity while training networks remains elusive. Using recent theoretical progress in explaining neural network generalization, we investigate how the amount of training data required to generalize on a task varies with the intrinsic dimensionality of a task's input. We show theoretically that when applied to modularly structured tasks, while nonmodular networks require an exponential number of samples with task dimensionality, modular networks' sample complexity is independent of task dimensionality: modular networks can generalize in high dimensions. We then develop a novel learning rule for modular networks to exploit this advantage and empirically show the improved generalization of the rule, both in- and out-of-distribution, on high-dimensional, modular tasks.

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.

Language models (LM) are capable of remarkably complex linguistic tasks; however, numerical reasoning is an area in which they frequently struggle. An important but rarely evaluated form of reasoning is understanding probability distributions. In this paper, we focus on evaluating the probabilistic reasoning capabilities of LMs using idealized and real-world statistical distributions. We perform a systematic evaluation of state-of-the-art LMs on three tasks: estimating percentiles, drawing samples, and calculating probabilities. We evaluate three ways to provide context to LMs 1) anchoring examples from within a distribution or family of distributions, 2) real-world context, 3) summary statistics on which to base a Normal approximation. Models can make inferences about distributions, and can be further aided by the incorporation of real-world context, example shots and simplified assumptions, even if these assumptions are incorrect or misspecified. To conduct this work, we developed a comprehensive benchmark distribution dataset with associated question-answer pairs that we have released publicly., Comment: EMNLP 2024 (Main), 21 pages, 9 figures, 2 tables

Despite the proliferation of wearable health trackers and the importance of sleep and exercise to health, deriving actionable personalized insights from wearable data remains a challenge because doing so requires non-trivial open-ended analysis of these data. The recent rise of large language model (LLM) agents, which can use tools to reason about and interact with the world, presents a promising opportunity to enable such personalized analysis at scale. Yet, the application of LLM agents in analyzing personal health is still largely untapped. In this paper, we introduce the Personal Health Insights Agent (PHIA), an agent system that leverages state-of-the-art code generation and information retrieval tools to analyze and interpret behavioral health data from wearables. We curate two benchmark question-answering datasets of over 4000 health insights questions. Based on 650 hours of human and expert evaluation we find that PHIA can accurately address over 84% of factual numerical questions and more than 83% of crowd-sourced open-ended questions. This work has implications for advancing behavioral health across the population, potentially enabling individuals to interpret their own wearable data, and paving the way for a new era of accessible, personalized wellness regimens that are informed by data-driven insights., Comment: 38 pages

Morphological mapping is a fundamental step in studying the processes that shaped an asteroid surface. Yet, it is challenging and often requires multiple independent assessments by trained experts. Here, we present fast methods to detect and characterize meaningful terrains from the topographic roughness: entropy of information, and local mean surface orientation. We apply our techniques to Didymos and Dimorphos, the target asteroids of NASA's DART mission: first attempt to deflect an asteroid. Our methods reliably identify morphological units at multiple scales. The comparative study reveals various terrain types, signatures of processes that transformed Didymos and Dimorphos. Didymos shows the most heterogeneity and morphology that indicate recent resurfacing events. Dimorphos is comparatively rougher than Didymos, which may result from the formation process of the binary pair and past interaction between the two bodies. Our methods can be readily applied to other bodies and data sets., Comment: submitted to PSJ

Developers rely on the static safety guarantees of the Rust programming language to write secure and performant applications. However, Rust is frequently used to interoperate with other languages which allow design patterns that conflict with Rust's evolving aliasing models. Miri is currently the only dynamic analysis tool that can validate applications against these models, but it does not support foreign functions, indicating that there may be a critical correctness gap across the Rust ecosystem. We conducted a large-scale evaluation of Rust libraries that call foreign functions to determine whether Miri's dynamic analyses remain useful in this context. We used Miri and an LLVM interpreter to jointly execute applications that call foreign functions, where we found 47 instances of undefined or undesired behavior from 37 libraries. Three bugs were found in libraries that had more than 10,000 daily downloads on average during our observation period, and one was found in a library maintained by the Rust Project. Many of these bugs were violations of Rust's aliasing models, but the latest Tree Borrows model was significantly more permissive than the earlier Stacked Borrows model. The Rust community must invest in new, production-ready tooling for multi-language applications to ensure that developers can detect these errors., Comment: 13 pages, preprint

The Rust programming language restricts aliasing to provide static safety guarantees. However, in certain situations, developers need to bypass these guarantees by using a set of unsafe features. If they are used incorrectly, these features can reintroduce the types of safety issues that Rust was designed to prevent. We seek to understand how current development tools can be improved to better assist developers who find it necessary to interact with unsafe code. To that end, we study how developers reason about foreign function calls, the limitations of the tools that they currently use, their motivations for using unsafe code, and how they reason about encapsulating it. We conducted a mixed-methods investigation consisting of semi-structured interviews with 19 developers, followed by a survey that reached an additional 160 developers. Our participants were motivated to use unsafe code when they perceived that there was no alternative, and most avoided using it. However, limited tooling support for foreign function calls made participants uncertain about their design choices, and certain foreign aliasing and concurrency patterns were difficult to encapsulate. To overcome these challenges, Rust developers need verification tools that can provide guarantees of soundness within multi-language applications., Comment: 33 pages with references, preprint

Reverse-mode automatic differentiation (autodiff) has been popularized by deep learning, but its ability to compute gradients is also valuable for interactive use cases such as bidirectional computer-aided design, embedded physics simulations, visualizing causal inference, and more. Unfortunately, the web is ill-served by existing autodiff frameworks, which use autodiff strategies that perform poorly on dynamic scalar programs, and pull in heavy dependencies that would result in unacceptable webpage sizes. This work introduces Rose, a lightweight autodiff framework for the web using a new hybrid approach to reverse-mode autodiff, blending conventional tracing and transformation techniques in a way that uses the host language for metaprogramming while also allowing the programmer to explicitly define reusable functions that comprise a larger differentiable computation. We demonstrate the value of the Rose design by porting two differentiable physics simulations, and evaluate its performance on an optimization-based diagramming application, showing Rose outperforming the state-of-the-art in web-based autodiff by multiple orders of magnitude.

An accurate differential diagnosis (DDx) is a cornerstone of medical care, often reached through an iterative process of interpretation that combines clinical history, physical examination, investigations and procedures. Interactive interfaces powered by Large Language Models (LLMs) present new opportunities to both assist and automate aspects of this process. In this study, we introduce an LLM optimized for diagnostic reasoning, and evaluate its ability to generate a DDx alone or as an aid to clinicians. 20 clinicians evaluated 302 challenging, real-world medical cases sourced from the New England Journal of Medicine (NEJM) case reports. Each case report was read by two clinicians, who were randomized to one of two assistive conditions: either assistance from search engines and standard medical resources, or LLM assistance in addition to these tools. All clinicians provided a baseline, unassisted DDx prior to using the respective assistive tools. Our LLM for DDx exhibited standalone performance that exceeded that of unassisted clinicians (top-10 accuracy 59.1% vs 33.6%, [p = 0.04]). Comparing the two assisted study arms, the DDx quality score was higher for clinicians assisted by our LLM (top-10 accuracy 51.7%) compared to clinicians without its assistance (36.1%) (McNemar's Test: 45.7, p < 0.01) and clinicians with search (44.4%) (4.75, p = 0.03). Further, clinicians assisted by our LLM arrived at more comprehensive differential lists than those without its assistance. Our study suggests that our LLM for DDx has potential to improve clinicians' diagnostic reasoning and accuracy in challenging cases, meriting further real-world evaluation for its ability to empower physicians and widen patients' access to specialist-level expertise.

We present observations obtained with the Near Infrared Spectrograph on JWST of the five Jupiter Trojans that will be visited by the Lucy spacecraft -- the Patroclus-Menoetius binary, Eurybates, Orus, Leucus, and Polymele. The measured 1.7-5.3 $\mu$m reflectance spectra, which provide increased wavelength coverage, spatial resolution, and signal-to-noise ratio over previous ground-based spectroscopy, reveal several distinct absorption features. We detect a broad OH band centered at 3 $\mu$m that is most prominent on the less-red objects Eurybates, Patroclus-Menoetius, and Polymele. An additional absorption feature at 3.3-3.6 $\mu$m, indicative of aliphatic organics, is systematically deeper on the red objects Orus and Leucus. The collisional fragment Eurybates is unique in displaying an absorption band at 4.25 $\mu$m that we attribute to bound or trapped CO$_2$. Comparisons with other solar system small bodies reveal broad similarities in the 2.7-3.6 $\mu$m bands with analogous features on Centaurs, Kuiper belt objects (KBOs), and the active asteroid 238P. In the context of recent solar system evolution models, which posit that the Trojans initially formed in the outer solar system, the significant attenuation of the 2.7-3.6 $\mu$m absorption features on Trojans relative to KBOs may be the result of secondary thermal processing of the Trojans' surfaces at the higher temperatures of the Jupiter region. The CO$_2$ band manifested on the surface of Eurybates suggests that CO$_2$ may be a major constituent in the bulk composition of Trojans, but resides in the subsurface or deeper interior and is largely obscured by refractory material that formed from the thermophysical processes that were activated during their inward migration., Comment: Published in PSJ. 13 pages, 6 figures

The Didymos binary asteroid was the target of the Double Asteroid Redirection Test (DART) mission, which intentionally impacted Dimorphos, the smaller member of the binary system. We used the Near-Infrared Spectrograph (NIRSpec) and Mid-Infrared Instrument (MIRI) instruments on JWST to measure the 0.6-5 $\mu$m and 5-20 $\mu$m spectra of Didymos approximately two months after the DART impact. These observations confirm that Didymos belongs to the S asteroid class and is most consistent with LL chondrite composition as was previously determined from its 0.6-2.5-$\mu$m reflectance spectrum. Measurements at wavelengths $>$ 2.5 $\mu$m show Didymos to have thermal properties typical for an S-complex asteroid of its size and to be lacking absorptions deeper than $\sim$2\% due to OH or H2O. Didymos' mid-infrared emissivity spectrum is within the range of what has been observed on S-complex asteroids observed with Spitzer Space Telescope and is most consistent with emission from small ($<$ 25 $\mu$m) surface particles. We conclude that the observed reflectance and physical properties make the Didymos system a good proxy for the type of ordinary chondrite asteroids that cross near-Earth space, and a good representative of likely future impactors., Comment: Accepted by the Planetary Science Journal, 16 October 2023 Paper has 29 pages, 13 figures, 1 table

According to density functional theory, any chemical property can be inferred from the electron density, making it the most informative attribute of an atomic structure. In this work, we demonstrate the use of established physical methods to obtain important chemical properties from model-predicted electron densities. We introduce graph neural network architectural choices that provide physically relevant and useful electron density predictions. Despite not training to predict atomic charges, the model is able to predict atomic charges with an order of magnitude lower error than a sum of atomic charge densities. Similarly, the model predicts dipole moments with half the error of the sum of atomic charge densities method. We demonstrate that larger data sets lead to more useful predictions in these tasks. These results pave the way for an alternative path in atomistic machine learning, where data-driven approaches and existing physical methods are used in tandem to obtain a variety of chemical properties in an explainable and self-consistent manner.

Property-based testing (PBT), while an established technique in the software testing research community, is still relatively underused in real-world software. Pain points in writing property-based tests include implementing diverse random input generators and thinking of meaningful properties to test. Developers, however, are more amenable to writing documentation; plenty of library API documentation is available and can be used as natural language specifications for PBTs. As large language models (LLMs) have recently shown promise in a variety of coding tasks, we investigate using modern LLMs to automatically synthesize PBTs using two prompting techniques. A key challenge is to rigorously evaluate the LLM-synthesized PBTs. We propose a methodology to do so considering several properties of the generated tests: (1) validity, (2) soundness, and (3) property coverage, a novel metric that measures the ability of the PBT to detect property violations through generation of property mutants. In our evaluation on 40 Python library API methods across three models (GPT-4, Gemini-1.5-Pro, Claude-3-Opus), we find that with the best model and prompting approach, a valid and sound PBT can be synthesized in 2.4 samples on average. We additionally find that our metric for determining soundness of a PBT is aligned with human judgment of property assertions, achieving a precision of 100% and recall of 97%. Finally, we evaluate the property coverage of LLMs across all API methods and find that the best model (GPT-4) is able to automatically synthesize correct PBTs for 21% of properties extractable from API documentation.

Large language models (LLMs) can capture rich representations of concepts that are useful for real-world tasks. However, language alone is limited. While existing LLMs excel at text-based inferences, health applications require that models be grounded in numerical data (e.g., vital signs, laboratory values in clinical domains; steps, movement in the wellness domain) that is not easily or readily expressed as text in existing training corpus. We demonstrate that with only few-shot tuning, a large language model is capable of grounding various physiological and behavioral time-series data and making meaningful inferences on numerous health tasks for both clinical and wellness contexts. Using data from wearable and medical sensor recordings, we evaluate these capabilities on the tasks of cardiac signal analysis, physical activity recognition, metabolic calculation (e.g., calories burned), and estimation of stress reports and mental health screeners.