Search

Rapid search and rescue is critical to maximizing survival rates following natural disasters. However, these efforts are challenged by the need to search large disaster zones, lack of reliability in the communications infrastructure, and a priori unknown numbers of objects of interest (OOIs), such as injured survivors. Aerial robots are increasingly being deployed for search and rescue due to their high mobility, but there remains a gap in deploying multi-robot autonomous aerial systems for methodical search of large environments. Prior works have relied on preprogrammed paths from human operators or are evaluated only in simulation. We bridge these gaps in the state of the art by developing and demonstrating a decentralized active search system, which biases its trajectories to take additional views of uncertain OOIs. The methodology leverages stochasticity for rapid coverage in communication denied scenarios. When communications are available, robots share poses, goals, and OOI information to accelerate the rate of search. Extensive simulations and hardware experiments in Bloomingdale, OH, are conducted to validate the approach. The results demonstrate the active search approach outperforms greedy coverage-based planning in communication-denied scenarios while maintaining comparable performance in communication-enabled scenarios.

We propose a new predictive theory for the analysis of common envelope (CE) events which incorporates the effects of relevant hydrodynamical processes into a simple analytical framework. We introduce the ejection and dynamical parameters $\xi$ and $\beta$, which define if envelope ejection is energetically or hydrodynamically favorable, respectively, during CE inspiral. When combined, these parameters offer a detailed narrative of how inspiral begins, proceeds, and ends that is consistent with preliminary comparisons to 3D hydrodynamical models. This physically-motivated framework impacts predictions for CE outcomes, especially for systems that have energy excess, and offers promise as a potential alternative for the treatment of CE in binary population synthesis., Comment: 11 pages, 3 figures, submitted to ApJL

Monocular depth estimation is a critical task for autonomous driving and many other computer vision applications. While significant progress has been made in this field, the effects of viewpoint shifts on depth estimation models remain largely underexplored. This paper introduces a novel dataset and evaluation methodology to quantify the impact of different camera positions and orientations on monocular depth estimation performance. We propose a ground truth strategy based on homography estimation and object detection, eliminating the need for expensive lidar sensors. We collect a diverse dataset of road scenes from multiple viewpoints and use it to assess the robustness of a modern depth estimation model to geometric shifts. After assessing the validity of our strategy on a public dataset, we provide valuable insights into the limitations of current models and highlight the importance of considering viewpoint variations in real-world applications., Comment: 17 pages, 5 figures. Accepted at ECCV 2024 2nd Workshop on Vision-Centric Autonomous Driving (VCAD)

Observations of gamma-ray burst afterglows have begun to readily reveal contamination from a kilonova or a supernova. This contamination presents significant challenges towards traditional methods of inferring the properties of these phenomena from observations. Given current knowledge of kilonova and afterglow modelling, observations (as expected) with near-infrared bands and at early observing times provide the greatest diagnostic power for both observing the presence of a kilonova and inferences on its properties in gamma-ray burst afterglows. However, contemporaneous observations in radio and X-ray are critical for reducing the afterglow parameter space and for more efficient parameter estimation. We compare different methods for fitting joint kilonova and afterglow observations under different scenarios. We find that ignoring the contribution of one source (even in scenarios where the source is sub-dominant) can lead to significantly biased estimated parameters but could still produce great light curve fits that do not raise suspicion. This bias is also present for analyses that fit data where one source is "subtracted". In most scenarios, the bias is smaller than the systematic uncertainty inherent to kilonova models but significant for afterglow parameters, particularly in the absence of high-quality radio and X-ray observations. Instead, we show that the most reliable method for inference in any scenario where contamination can not be confidently dismissed is to jointly fit for both an afterglow and kilonova/supernova, and showcase a Bayesian framework to make this joint analysis computationally feasible., Comment: Will be submitted to MNRAS after a few days for comments. 14 pages and 12 figures. Model and framework available via Redback at https://github.com/nikhil-sarin/redback

Ads Content Safety at Google requires classifying billions of ads for Google Ads content policies. Consistent and accurate policy enforcement is important for advertiser experience and user safety and it is a challenging problem, so there is a lot of value for improving it for advertisers and users. Inconsistent policy enforcement causes increased policy friction and poor experience with good advertisers, and bad advertisers exploit the inconsistency by creating multiple similar ads in the hope that some will get through our defenses. This study proposes a method to understand advertiser's intent for content policy violations, using Large Language Models (LLMs). We focus on identifying good advertisers to reduce content over-flagging and improve advertiser experience, though the approach can easily be extended to classify bad advertisers too. We generate advertiser's content profile based on multiple signals from their ads, domains, targeting info, etc. We then use LLMs to classify the advertiser content profile, along with relying on any knowledge the LLM has of the advertiser, their products or brand, to understand whether they are likely to violate a certain policy or not. After minimal prompt tuning our method was able to reach 95\% accuracy on a small test set.

This work presents a detailed case study on using Generative AI (GenAI) to develop AI surrogates for simulation models in fusion energy research. The scope includes the methodology, implementation, and results of using GenAI to assist in model development and optimization, comparing these results with previous manually developed models.

Organ transplantation serves as the primary therapeutic strategy for end-stage organ failures. However, allograft rejection is a common complication of organ transplantation. Histological assessment is essential for the timely detection and diagnosis of transplant rejection and remains the gold standard. Nevertheless, the traditional histochemical staining process is time-consuming, costly, and labor-intensive. Here, we present a panel of virtual staining neural networks for lung and heart transplant biopsies, which digitally convert autofluorescence microscopic images of label-free tissue sections into their brightfield histologically stained counterparts, bypassing the traditional histochemical staining process. Specifically, we virtually generated Hematoxylin and Eosin (H&E), Masson's Trichrome (MT), and Elastic Verhoeff-Van Gieson (EVG) stains for label-free transplant lung tissue, along with H&E and MT stains for label-free transplant heart tissue. Subsequent blind evaluations conducted by three board-certified pathologists have confirmed that the virtual staining networks consistently produce high-quality histology images with high color uniformity, closely resembling their well-stained histochemical counterparts across various tissue features. The use of virtually stained images for the evaluation of transplant biopsies achieved comparable diagnostic outcomes to those obtained via traditional histochemical staining, with a concordance rate of 82.4% for lung samples and 91.7% for heart samples. Moreover, virtual staining models create multiple stains from the same autofluorescence input, eliminating structural mismatches observed between adjacent sections stained in the traditional workflow, while also saving tissue, expert time, and staining costs., Comment: 21 Pages, 5 Figures

The Zernike wavefront sensor (ZWFS) stands out as one of the most sensitive optical systems for measuring the phase of an incoming wavefront, reaching photon efficiencies close to the fundamental limit. This quality, combined with the fact that it can easily measure phase discontinuities, has led to its widespread adoption in various wavefront control applications, both on the ground but also for future space-based instruments. Despite its advantages, the ZWFS faces a significant challenge due to its extremely limited dynamic range, making it particularly challenging for ground-based operations. To address this limitation, one approach is to use the ZWFS after a general adaptive optics (AO) system; however, even in this scenario, the dynamic range remains a concern. This paper investigates two optical configurations of the ZWFS: the conventional setup and its phase-shifted counterpart, which generates two distinct images of the telescope pupil. We assess the performance of various reconstruction techniques for both configurations, spanning from traditional linear reconstructors to gradient-descent-based methods. The evaluation encompasses simulations and experimental tests conducted on the Santa cruz Extreme Adaptive optics Lab (SEAL) bench at UCSC. Our findings demonstrate that certain innovative reconstruction techniques introduced in this study significantly enhance the dynamic range of the ZWFS, particularly when utilizing the phase-shifted version.

After shipment to the Daresbury Lab and return to Fermilab, the prototype HB650 cryomodule underwent another phase of 2K RF testing to ascertain any performance issues that may have arisen from the transport of the cryomodule. While measurements taken at room temperature after the conclusion of shipment indicated that there were no negative impacts on cavity alignment, beamline vacuum, or cavity frequency, testing at 2K was required to validate other aspects such as tuner operation, cavity coupling, cryogenic system integrity, and cavity performance. Results of this latest round of limited 2K testing will be presented., Comment: 32nd Linear Accelerator Conference (LINAC 2024)

Human pose estimation involves detecting and tracking the positions of various body parts using input data from sources such as images, videos, or motion and inertial sensors. This paper presents a novel approach to human pose estimation using machine learning algorithms to predict human posture and translate them into robot motion commands using ultra-wideband (UWB) nodes, as an alternative to motion sensors. The study utilizes five UWB sensors implemented on the human body to enable the classification of still poses and more robust posture recognition. This approach ensures effective posture recognition across a variety of subjects. These range measurements serve as input features for posture prediction models, which are implemented and compared for accuracy. For this purpose, machine learning algorithms including K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and deep Multi-Layer Perceptron (MLP) neural network are employed and compared in predicting corresponding postures. We demonstrate the proposed approach for real-time control of different mobile/aerial robots with inference implemented in a ROS 2 node. Experimental results demonstrate the efficacy of the approach, showcasing successful prediction of human posture and corresponding robot movements with high accuracy.

Colorectal cancer (CRC) is the third most common cause of cancer diagnosed in the United States and the second leading cause of cancer-related death among both genders. Notably, CRC is the leading cause of cancer in younger men less than 50 years old. Colonoscopy is considered the gold standard for the early diagnosis of CRC. Skills vary significantly among endoscopists, and a high miss rate is reported. Automated polyp segmentation can reduce the missed rates, and timely treatment is possible in the early stage. To address this challenge, we introduce \textit{\textbf{\ac{FANetv2}}}, an advanced encoder-decoder network designed to accurately segment polyps from colonoscopy images. Leveraging an initial input mask generated by Otsu thresholding, FANetv2 iteratively refines its binary segmentation masks through a novel feedback attention mechanism informed by the mask predictions of previous epochs. Additionally, it employs a text-guided approach that integrates essential information about the number (one or many) and size (small, medium, large) of polyps to further enhance its feature representation capabilities. This dual-task approach facilitates accurate polyp segmentation and aids in the auxiliary classification of polyp attributes, significantly boosting the model's performance. Our comprehensive evaluations on the publicly available BKAI-IGH and CVC-ClinicDB datasets demonstrate the superior performance of FANetv2, evidenced by high dice similarity coefficients (DSC) of 0.9186 and 0.9481, along with low Hausdorff distances of 2.83 and 3.19, respectively. The source code for FANetv2 is available at https://github.com/xxxxx/FANetv2.

This paper introduces the novel Droop-e grid-forming power electronic converter control strategy, which establishes a non-linear, active power--frequency droop relationship based on an exponential function of the power output. A primary advantage of Droop-e is an increased utilization of available power headroom that directly mitigates system frequency excursions and reduces the rate of change of frequency. The motivation for Droop-e as compared to a linear grid-forming control is first established, and then the full controller is described, including the mirrored inversion at the origin, the linearization at a parameterized limit, and the auxiliary autonomous power sharing controller. The analytic stability of the controller, including synchronization criteria and a small signal stability analysis, is assessed. Electromagnetic transient time domain simulations of the Droop-e controller with full order power electronic converters and accompanying DC-side dynamics, connected in parallel with synchronous generators, are executed at a range of dispatches on a simple 3-bus system. Finally, IEEE 39-bus system simulations highlight the improved frequency stability of the system with multiple, Droop-e controlled grid-forming inverters., Comment: 10 pages, 11 figures

GQ Lup B is one of the few substellar companions with a detected cicumplanetary disk, or CPD. Observations of the CPD suggest the presence of a cavity, possibly formed by an exo-satellite. Using the Keck Planet Imager and Characterizer (KPIC), a high contrast imaging suite that feeds a high resolution spectrograph (1.9-2.5 microns, R$\sim$35,000), we present the first dedicated radial velocity (RV) observations around a high-contrast, directly imaged substellar companion, GQ Lup B, to search for exo-satellites. Over 11 epochs, we find a best and median RV error of 400-1000 m/s, most likely limited by systematic fringing in the spectra due to transmissive optics within KPIC. With this RV precision, KPIC is sensitive to exomoons 0.6-2.8% the mass of GQ Lup B ($\sim 30 M_{\text{Jup}}$) at separations between the Roche limit and $65 R_{\text{Jup}}$, or the extent of the cavity inferred within the CPD detected around GQ Lup B. Using simulations of HISPEC, a high resolution infrared spectrograph planned to debut at W.M. Keck Observatory in 2026, we estimate future exomoon sensitivity to increase by over an order of magnitude, providing sensitivity to less massive satellites potentially formed within the CPD itself. Additionally, we run simulations to estimate the amount of material that different masses of satellites could clear in a CPD to create the observed cavity. We find satellite-to-planet mass ratios of $q > 2 \times 10^{-4}$ can create observable cavities and report a maximum cavity size of $\sim 51 \, R_{\text{Jup}}$ carved from a satellite., Comment: 15 pages, 5 figures

The Keck Planet Imager and Characterizer (KPIC) combines high contrast imaging with high resolution spectroscopy (R$\sim$35,000 in K band) to study directly imaged exoplanets and brown dwarfs in unprecedented detail. KPIC aims to spectrally characterize substellar companions through measurements of planetary radial velocities, spins, and atmospheric composition. Currently, the dominant source of systematic noise for KPIC is fringing, or oscillations in the spectrum as a function of wavelength. The fringing signal can dominate residuals by up to 10% of the continuum for high S/N exposures, preventing accurate wavelength calibration, retrieval of atmospheric parameters, and detection of planets with flux ratios less than 1% of the host star. To combat contamination from fringing, we first identify its three unique sources and adopt a physically informed model of Fabry-P\'{e}rot cavities to apply to post-processed data. We find this strategy can effectively model the fringing in observations of A0V/F0V stars, reducing the residual systematics caused by fringing by a factor of 2. Next, we wedge two of the transmissive optics internal to KPIC to eliminate two sources of fringing and confirm the third source as the entrance window to the spectrograph. Finally, we apply our previous model of the Fabry-P\'{e}rot cavity to new data taken with the wedged optics to reduce the amplitude of the residuals by a factor of 10., Comment: 13 pages, 6 figures

We use data from the ALMA Evolutionary Study of High Mass Protocluster Formation in the Galaxy (ALMAGAL) survey to study 100 ALMAGAL regions at $\sim$ 1 arsecond resolution located between $\sim$ 2 and 6 kpc distance. Using ALMAGAL $\sim$ 1.3mm line and continuum data we estimate flow rates onto individual cores. We focus specifically on flow rates along filamentary structures associated with these cores. Our primary analysis is centered around position velocity cuts in H$_2$CO (3$_{0,3}$ - 2$_{0,2}$) which allow us to measure the velocity fields, surrounding these cores. Combining this work with column density estimates we derive the flow rates along the extended filamentary structures associated with cores in these regions. We select a sample of 100 ALMAGAL regions covering four evolutionary stages from quiescent to protostellar, Young Stellar Objects (YSOs), and HII regions (25 each). Using dendrogram and line analysis, we identify a final sample of 182 cores in 87 regions. In this paper, we present 728 flow rates for our sample (4 per core), analysed in the context of evolutionary stage, distance from the core, and core mass. On average, for the whole sample, we derive flow rates on the order of $\sim$10$^{-4}$ M$_{sun}$yr$^{-1}$ with estimated uncertainties of $\pm$50%. We see increasing differences in the values among evolutionary stages, most notably between the less evolved (quiescent/protostellar) and more evolved (YSO/HII region) sources. We also see an increasing trend as we move further away from the centre of these cores. We also find a clear relationship between the flow rates and core masses $\sim$M$^{2/3}$ which is in line with the result expected from the tidal-lobe accretion mechanism. Overall, we see increasing trends in the relationships between the flow rate and the three investigated parameters; evolutionary stage, distance from the core, and core mass., Comment: 11 pages, 11 figures, accepted for publication in A&A

We present results from a search for X-ray/gamma-ray counterparts of gravitational-wave (GW) candidates from the third observing run (O3) of the LIGO-Virgo-KAGRA (LVK) network using the Swift Burst Alert Telescope (Swift-BAT). The search includes 636 GW candidates received in low latency, 86 of which have been confirmed by the offline analysis and included in the third cumulative Gravitational-Wave Transient Catalogs (GWTC-3). Targeted searches were carried out on the entire GW sample using the maximum--likelihood NITRATES pipeline on the BAT data made available via the GUANO infrastructure. We do not detect any significant electromagnetic emission that is temporally and spatially coincident with any of the GW candidates. We report flux upper limits in the 15-350 keV band as a function of sky position for all the catalog candidates. For GW candidates where the Swift-BAT false alarm rate is less than 10$^{-3}$ Hz, we compute the GW--BAT joint false alarm rate. Finally, the derived Swift-BAT upper limits are used to infer constraints on the putative electromagnetic emission associated with binary black hole mergers., Comment: 50 pages, 10 figures, 4 tables

Young, self-luminous super-Jovian companions discovered by direct imaging provide a challenging test of planet formation and evolution theories. By spectroscopically characterizing the atmospheric compositions of these super-Jupiters, we can constrain their formation histories. Here we present studies of the recently discovered HIP 99770 b, a 16 MJup high-contrast companion on a 17 au orbit, using the fiber-fed high-resolution spectrograph KPIC (R~35,000) on the Keck II telescope. Our K-band observations led to detections of H2O and CO in the atmosphere of HIP 99770 b. We carried out free retrieval analyses using petitRADTRANS to measure its chemical abundances, including the metallicity and C/O ratio, projected rotation velocity (vsini), and radial velocity (RV). We found that the companion's atmosphere has C/O=0.55(-0.04/+0.06) and [M/H]=0.26(-0.23/+0.24) (1{\sigma} confidence intervals), values consistent with those of the Sun and with a companion formation via gravitational instability or core accretion. The projected rotation velocity < 7.8 km/s is small relative to other directly imaged companions with similar masses and ages. This may imply a near pole-on orientation or effective magnetic braking by a circumplanetary disk. In addition, we added the companion-to-primary relative RV measurement to the orbital fitting and obtained updated constraints on orbital parameters. Detailed characterization of super-Jovian companions within 20 au like HIP 99770 b is critical for understanding the formation histories of this population., Comment: 18 pages, 5 figures, accepted to AJ

Pancreatic cancer remains one of the leading causes of cancer-related mortality worldwide. Precise segmentation of pancreatic tumors from medical images is a bottleneck for effective clinical decision-making. However, achieving a high accuracy is often limited by the small size and availability of real patient data for training deep learning models. Recent approaches have employed synthetic data generation to augment training datasets. While promising, these methods may not yet meet the performance benchmarks required for real-world clinical use. This study critically evaluates the limitations of existing generative-AI based frameworks for pancreatic tumor segmentation. We conduct a series of experiments to investigate the impact of synthetic \textit{tumor size} and \textit{boundary definition} precision on model performance. Our findings demonstrate that: (1) strategically selecting a combination of synthetic tumor sizes is crucial for optimal segmentation outcomes, and (2) generating synthetic tumors with precise boundaries significantly improves model accuracy. These insights highlight the importance of utilizing refined synthetic data augmentation for enhancing the clinical utility of segmentation models in pancreatic cancer decision making including diagnosis, prognosis, and treatment plans. Our code will be available at https://github.com/lkpengcs/SynTumorAnalyzer., Comment: MICCAI Workshop AIPAD 2024

We extend our previous demonstration of the first on-sky primary mirror segment closed-loop control on Keck using a vector-Zernike wavefront sensor (vZWFS), which improved the Strehl ratio on the NIRC2 science camera by up to 10 percentage points. Segment co-phasing errors contribute to Keck contrast limits and will be necessary to correct for the segmented Extremely Large Telescopes and future space missions. The goal of the post-AO vZWFS on Keck is to monitor and correct segment co-phasing errors in parallel with science observations. The ZWFS is ideal for measuring phase discontinuities and is one of the most sensitive WFSs, but has limited dynamic range. The Keck vZWFS consists of a metasurface mask imposing two different phase shifts to orthogonal polarizations, split into two pupil images, extending its dynamic range. We report on the vZWFS closed-loop co-phasing performance and early work towards understanding the interactions between the AO system and segment phasing. We discuss a comparison of the AO performance when co-phasing by aligning segment edges, as is currently done at Keck, compared with aligning to the average phase over the segments, as is done by the vZWFS., Comment: Proceedings of SPIE, 13097-61, 8 pages, 3 figures, 3 tables

The enormous scale of state-of-the-art foundation models has limited their accessibility to scientists, because customized experiments at large model sizes require costly hardware and complex engineering that is impractical for most researchers. To alleviate these problems, we introduce NNsight, an open-source Python package with a simple, flexible API that can express interventions on any PyTorch model by building computation graphs. We also introduce NDIF, a collaborative research platform providing researchers access to foundation-scale LLMs via the NNsight API. Code, documentation, and tutorials are available at https://www.nnsight.net., Comment: Code at https://nnsight.net

I expound and defend the ``bare probabilism'' reading of Gibbsian (i.e. mainstream) statistical mechanics, responding to Frigg and Werndl's recent (BJPS 72 (2021), 105-129) plea: ``can somebody please say what Gibbsian statistical mechanics says?'', Comment: 17 pages

Instruction-tuned Large Language Models (LLMs) can perform a wide range of tasks given natural language instructions to do so, but they are sensitive to how such instructions are phrased. This issue is especially concerning in healthcare, as clinicians are unlikely to be experienced prompt engineers and the potential consequences of inaccurate outputs are heightened in this domain. This raises a practical question: How robust are instruction-tuned LLMs to natural variations in the instructions provided for clinical NLP tasks? We collect prompts from medical doctors across a range of tasks and quantify the sensitivity of seven LLMs -- some general, others specialized -- to natural (i.e., non-adversarial) instruction phrasings. We find that performance varies substantially across all models, and that -- perhaps surprisingly -- domain-specific models explicitly trained on clinical data are especially brittle, compared to their general domain counterparts. Further, arbitrary phrasing differences can affect fairness, e.g., valid but distinct instructions for mortality prediction yield a range both in overall performance, and in terms of differences between demographic groups., Comment: To appear at BioNLP, ACL 2024

The Solar eruptioN Integral Field Spectrograph (SNIFS) is a solar-gazing spectrograph scheduled to fly in the summer of 2025 on a NASA sounding rocket. Its goal is to view the solar chromosphere and transition region at a high cadence (1s) both spatially (0.5") and spectrally (33 m{\AA}) viewing wavelengths around Lyman Alpha (1216 {\AA}), Si iii (1206 {\AA}) and O v (1218 {\AA}) to observe spicules, nanoflares, and possibly a solar flare. This time cadence will provide yet-unobserved detail about fast-changing features of the Sun. The instrument is comprised of a Gregorian-style reflecting telescope combined with a spectrograph via a specialized mirrorlet array that focuses the light from each spatial location in the image so that it may be spectrally dispersed without overlap from neighboring locations. This paper discusses the driving science, detailed instrument and subsystem design, and pre-integration testing of the SNIFS instrument., Comment: 22 pages (not including references), 7 figures, submitting to Solar Physics

The production of wind energy is a crucial part of sustainable development and reducing the reliance on fossil fuels. Maintaining the integrity of wind turbines to produce this energy is a costly and time-consuming task requiring repeated inspection and maintenance. While autonomous drones have proven to make this process more efficient, the algorithms for detecting anomalies to prevent catastrophic damage to turbine blades have fallen behind due to some dangerous defects, such as hairline cracks, being barely-visible. Existing datasets and literature are lacking and tend towards detecting obvious and visible defects in addition to not being geographically diverse. In this paper we introduce a novel and diverse dataset of barely-visible hairline cracks collected from numerous wind turbine inspections. To prove the efficacy of our dataset, we detail our end-to-end deployed turbine crack detection pipeline from the image acquisition stage to the use of predictions in providing automated maintenance recommendations to extend the life and efficiency of wind turbines.

This paper introduces Helios, the first extremely low-power, real-time, event-based hand gesture recognition system designed for all-day on smart eyewear. As augmented reality (AR) evolves, current smart glasses like the Meta Ray-Bans prioritize visual and wearable comfort at the expense of functionality. Existing human-machine interfaces (HMIs) in these devices, such as capacitive touch and voice controls, present limitations in ergonomics, privacy and power consumption. Helios addresses these challenges by leveraging natural hand interactions for a more intuitive and comfortable user experience. Our system utilizes a extremely low-power and compact 3mmx4mm/20mW event camera to perform natural hand-based gesture recognition for always-on smart eyewear. The camera's output is processed by a convolutional neural network (CNN) running on a NXP Nano UltraLite compute platform, consuming less than 350mW. Helios can recognize seven classes of gestures, including subtle microgestures like swipes and pinches, with 91% accuracy. We also demonstrate real-time performance across 20 users at a remarkably low latency of 60ms. Our user testing results align with the positive feedback we received during our recent successful demo at AWE-USA-2024., Comment: Accepted at ECCV-Integrating Computer Vision in Smart Eyewear, 2024. 18 pages, 10 figures. First three authors contributed equally to this paper

Recent work on evaluating the diversity of text generated by LLMs has focused on word-level features. Here we offer an analysis of syntactic features to characterize general repetition in models, beyond frequent n-grams. Specifically, we define syntactic templates and show that models tend to produce templated text in downstream tasks at a higher rate than what is found in human-reference texts. We find that most (76%) templates in model-generated text can be found in pre-training data (compared to only 35% of human-authored text), and are not overwritten during fine-tuning processes such as RLHF. This connection to the pre-training data allows us to analyze syntactic templates in models where we do not have the pre-training data. We also find that templates as features are able to differentiate between models, tasks, and domains, and are useful for qualitatively evaluating common model constructions. Finally, we demonstrate the use of templates as a useful tool for analyzing style memorization of training data in LLMs., Comment: EMNLP 2024

LLMs process text as sequences of tokens that roughly correspond to words, where less common words are represented by multiple tokens. However, individual tokens are often semantically unrelated to the meanings of the words/concepts they comprise. For example, Llama-2-7b's tokenizer splits the word "northeastern" into the tokens ['_n', 'ort', 'he', 'astern'], none of which correspond to semantically meaningful units like "north" or "east." Similarly, the overall meanings of named entities like "Neil Young" and multi-word expressions like "break a leg" cannot be directly inferred from their constituent tokens. Mechanistically, how do LLMs convert such arbitrary groups of tokens into useful higher-level representations? In this work, we find that last token representations of named entities and multi-token words exhibit a pronounced "erasure" effect, where information about previous and current tokens is rapidly forgotten in early layers. Using this observation, we propose a method to "read out" the implicit vocabulary of an autoregressive LLM by examining differences in token representations across layers, and present results of this method for Llama-2-7b and Llama-3-8B. To our knowledge, this is the first attempt to probe the implicit vocabulary of an LLM., Comment: 13 pages, 14 figures. Code and data at https://footprints.baulab.info/

Black-box finetuning is an emerging interface for adapting state-of-the-art language models to user needs. However, such access may also let malicious actors undermine model safety. To demonstrate the challenge of defending finetuning interfaces, we introduce covert malicious finetuning, a method to compromise model safety via finetuning while evading detection. Our method constructs a malicious dataset where every individual datapoint appears innocuous, but finetuning on the dataset teaches the model to respond to encoded harmful requests with encoded harmful responses. Applied to GPT-4, our method produces a finetuned model that acts on harmful instructions 99% of the time and avoids detection by defense mechanisms such as dataset inspection, safety evaluations, and input/output classifiers. Our findings question whether black-box finetuning access can be secured against sophisticated adversaries., Comment: 22 pages

The Keck Planet Imager and Characterizer (KPIC), a series of upgrades to the Keck II Adaptive Optics System and Instrument Suite, aims to demonstrate high-resolution spectroscopy of faint exoplanets that are spatially resolved from their host stars. In this paper, we measure KPIC's sensitivity to companions as a function of separation (i.e., the contrast curve) using on-sky data collected over four years of operation. We show that KPIC is able to reach contrasts of $1.3 \times 10^{-4}$ at 90 mas and $9.2 \times 10^{-6}$ at 420 mas separation from the star, and that KPIC can reach planet-level sensitivities at angular separations within the inner working angle of coronagraphic instruments such as GPI and SPHERE. KPIC is also able to achieve more extreme contrasts than other medium-/high-resolution spectrographs that are not as optimized for high-contrast performance. We decompose the KPIC performance budget into individual noise terms and discuss limiting factors. The fringing that results from combining a high-contrast imaging system with a high-resolution spectrograph is identified as an important source of systematic noise. After mitigation and correction, KPIC is able to reach within a factor of 2 of the photon noise limit at separations < 200 mas. At large separations, KPIC is limited by the background noise performance of NIRSPEC., Comment: 16 pages, 6 figures, submitted to the proceedings of SPIE Astronomical Telescopes + Instrumentation 2024, 13096-69

Eliciting "chain of thought" (CoT) rationales -- sequences of token that convey a "reasoning" process -- has been shown to consistently improve LLM performance on tasks like question answering. More recent efforts have shown that such rationales can also be used for model distillation: Including CoT sequences (elicited from a large "teacher" model) in addition to target labels when fine-tuning a small student model yields (often substantial) improvements. In this work we ask: Why and how does this additional training signal help in model distillation? We perform ablations to interrogate this, and report some potentially surprising results. Specifically: (1) Placing CoT sequences after labels (rather than before) realizes consistently better downstream performance -- this means that no student "reasoning" is necessary at test time to realize gains. (2) When rationales are appended in this way, they need not be coherent reasoning sequences to yield improvements; performance increases are robust to permutations of CoT tokens, for example. In fact, (3) a small number of key tokens are sufficient to achieve improvements equivalent to those observed when full rationales are used in model distillation., Comment: Accepted to EMNLP 2024

How should we interpret physical theories, and especially quantum theory, if we drop the assumption that we should treat it as an exact description of the whole Universe? I expound and develop the claim that physics is about the study of autonomous, but not necessarily isolated, dynamical systems, and that when applied to quantum mechanics this entails that in general we should take quantum systems as having mixed states and non-unitary dynamics. I argue that nonetheless unitary dynamics continues to have a special place in physics, via the empirically-well-supported reductionist principles that non-unitarity is to be explained by restriction to a subsystem of a larger unitary system and that microscopic physics is governed by unitary and largely known dynamics. I contrast this position with the `Open Systems View' advocated recently by Michael Cuffaro and Stephan Hartmann.

Entity matching is the task of linking records from different sources that refer to the same real-world entity. Past work has primarily treated entity linking as a standard supervised learning problem. However, supervised entity matching models often do not generalize well to new data, and collecting exhaustive labeled training data is often cost prohibitive. Further, recent efforts have adopted LLMs for this task in few/zero-shot settings, exploiting their general knowledge. But LLMs are prohibitively expensive for performing inference at scale for real-world entity matching tasks. As an efficient alternative, we re-cast entity matching as a conditional generation task as opposed to binary classification. This enables us to "distill" LLM reasoning into smaller entity matching models via natural language explanations. This approach achieves strong performance, especially on out-of-domain generalization tests (10.85% F-1) where standalone generative methods struggle. We perform ablations that highlight the importance of explanations, both for performance and model robustness., Comment: Accepted to EMNLP 2024

Transparent conducting oxides are highly doped semiconductors that exhibit favourable characteristics when compared to metals, including reduced material losses, tuneable electronic and optical properties, and enhanced damage thresholds. Recently, the photonic community has renewed its attention towards these materials, recognizing their remarkable nonlinear optical properties in the near-infrared spectrum, a feature previously overlooked despite their long-standing application in photovoltaics and touchscreens. The exceptionally large and ultra-fast change of the refractive index, which can be optically induced in these compounds, extends beyond the boundaries of conventional perturbative analysis and makes this class of materials the closest approximation to a time-varying system, and a unique playground for studying a variety of novel phenomena within the domain of photon acceleration. Here we report the spatio-spectral fission of an ultra-fast pulse trespassing a thin film of aluminium zinc oxide with a non-stationary refractive index. By applying phase conservation to this time-varying layer, our model can account for both space and time refraction and explain in quantitative terms, the spatial separation of both the spectrum and energy. Our findings represent an example of extreme nonlinear phenomena on subwavelength propagation distances and shed light on the nature of several nonlinear effects recently reported not accounting for the full optical field distribution. Our work also provides new important insights into transparent conducting oxides transient optical properties which are critical for the ongoing research in photonic time crystals, on-chip generation of nonclassical states of light, integrated optical neural networks as well as ultra-fast beam steering and frequency division multiplexing

On the path towards high-gain inertial confinement fusion ignition, foams are being considered to tamp the hohlraum wall-motion, and mitigate laser backscattering from Stimulated Raman Scattering (SRS) and Stimulated Brillouin Scattering (SBS). Here we present the results from an experimental campaign on foam-filled hohlraums conducted at the OMEGA laser facility. SiO2 foam-fills, with densities as low as 1 mg/cm3, successfully reduce the gold wall expansion, with laser backscattering comparable to gas-fill.

This paper introduces a novel reconfigurable and power-efficient FPGA (Field-Programmable Gate Array) implementation of an operator splitting algorithm for Non-Terrestial Network's (NTN) relay satellites model predictive orientation control (MPC). Our approach ensures system stability and introduces an innovative reconfigurable bit-width FPGA-based optimization solver. To demonstrate its efficacy, we employ a real FPGA-In-the-Loop hardware setup to control simulated satellite dynamics. Furthermore, we conduct an in-depth comparative analysis, examining various fixed-point configurations to evaluate the combined system's closed-loop performance and power efficiency, providing a holistic understanding of the proposed implementation's advantages.

Abridged: Programmer attention represents the visual focus of programmers on parts of the source code in pursuit of programming tasks. We conducted an in-depth human study with XY Java programmers, where each programmer generated summaries for 40 methods from five large Java projects over five one-hour sessions. We used eye-tracking equipment to map the visual attention of programmers while they wrote the summaries. We also rate the quality of each summary. We found eye-gaze patterns and metrics that define common behaviors between programmer attention during context-aware code summarization. Specifically, we found that programmers need to read significantly (p<0.01) fewer words and make significantly fewer revisits to words (p\textless0.03) as they summarize more methods during a session, while maintaining the quality of summaries. We also found that the amount of source code a participant looks at correlates with a higher quality summary, but this trend follows a bell-shaped curve, such that after a threshold reading more source code leads to a significant decrease (p<0.01) in the quality of summaries. We also gathered insight into the type of methods in the project that provide the most contextual information for code summarization based on programmer attention. Specifically, we observed that programmers spent a majority of their time looking at methods inside the same class as the target method to be summarized. Surprisingly, we found that programmers spent significantly less time looking at methods in the call graph of the target method. We discuss how our empirical observations may aid future studies towards modeling programmer attention and improving context-aware automatic source code summarization., Comment: 10 pages, 4 figures, 4 tables. this is a pre-print submitted to IEEE Transactions on Software Engineering for review

Many observational studies feature irregular longitudinal data, where the observation times are not common across individuals in the study. Further, the observation times may be related to the longitudinal outcome. In this setting, failing to account for the informative observation process may result in biased causal estimates. This can be coupled with other sources of bias, including non-randomized treatment assignments and informative censoring. This paper provides an overview of a flexible weighting method used to adjust for informative observation processes and non-randomized treatment assignments. We investigate the sensitivity of the flexible weighting method to violations of the noninformative censoring assumption, examine variable selection for the observation process weighting model, known as inverse intensity weighting, and look at the impacts of weight trimming for the flexible weighting model. We show that the flexible weighting method is sensitive to violations of the noninformative censoring assumption and show that a previously proposed extension fails under such violations. We also show that variables confounding the observation and outcome processes should always be included in the observation intensity model. Finally, we show that weight trimming should be applied in the flexible weighting model when the treatment assignment process is highly informative and driving the extreme weights. We conclude with an application of the methodology to a real data set to examine the impacts of household water sources on malaria diagnoses., Comment: 20 pages and 2 figures in main document, 8 pages and 2 figures in Supplemental Material. Submitted to Statistical Methods in Medical Research

Guitar tablature transcription (GTT) aims at automatically generating symbolic representations from real solo guitar performances. Due to its applications in education and musicology, GTT has gained traction in recent years. However, GTT robustness has been limited due to the small size of available datasets. Researchers have recently used synthetic data that simulates guitar performances using pre-recorded or computer-generated tones and can be automatically generated at large scales. The present study complements these efforts by demonstrating that GTT robustness can be improved by including synthetic training data created using recordings of real guitar tones played with different audio effects. We evaluate our approach on a new evaluation dataset with professional solo guitar performances that we composed and collected, featuring a wide array of tones, chords, and scales.

Using Keck Planet Imager and Characterizer (KPIC) high-resolution ($R$~35000) spectroscopy from 2.29-2.49 $\mu$m, we present uniform atmospheric retrievals for eight young substellar companions with masses of ~10-30 $M_\textrm{Jup}$, orbital separations spanning ~50-360 au, and $T_\textrm{eff}$ between ~1500-2600 K. We find that all companions have solar C/O ratios, and metallicities, to within the 1-2$\sigma$ level, with the measurements clustered around solar composition. Stars in the same stellar associations as our systems have near-solar abundances, so these results indicate that this population of companions is consistent with formation via direct gravitational collapse. Alternatively, core accretion outside the CO snowline would be compatible with our measurements, though the high mass ratios of most systems would require rapid core assembly and gas accretion in massive disks. On a population level, our findings can be contrasted with abundance measurements for directly imaged planets with m<10 $M_\textrm{Jup}$, which show tentative atmospheric metal enrichment. In addition, the atmospheric compositions of our sample of companions are distinct from those of hot Jupiters, which most likely form via core accretion. For two companions with $T_\textrm{eff}$~1700-2000 K (kap And b and GSC 6214-210 b), our best-fit models prefer a non-gray cloud model with >3$\sigma$ significance. The cloudy models yield 2-3$\sigma$ lower $T_\textrm{eff}$ for these companions, though the C/O and [C/H] still agree between cloudy and clear models at the $1\sigma$ level. Finally, we constrain 12CO/13CO for three companions with the highest S/N data (GQ Lup b, HIP 79098 b, and DH Tau b), and report $v$sin($i$) and radial velocities for all companions., Comment: Accepted to ApJ, 52 pages, 20 figures

We used the Keck Planet Imager and Characterizer (KPIC) to obtain high-resolution (R$\sim$35,000) K-band spectra of kappa Andromedae b, a planetary-mass companion orbiting the B9V star, kappa Andromedae A. We characterized its spin, radial velocity, and bulk atmospheric parameters through use of a forward modeling framework to jointly fit planetary spectra and residual starlight speckles, obtaining likelihood-based posterior probabilities. We also detected H$_{2}$O and CO in its atmosphere via cross correlation. We measured a $v\sin(i)$ value for kappa And b of $38.42\pm{0.05}$ km/s, allowing us to extend our understanding of the population of close in bound companions at higher rotation rates. This rotation rate is one of the highest spins relative to breakup velocity measured to date, at close to $50\%$ of breakup velocity. We identify a radial velocity $-17.35_{-0.09}^{+0.05}$ km/s, which we use with existing astrometry and RV measurements to update the orbital fit. We also measure an effective temperature of $1700\pm{100}$ K and a $\log(g)$ of $4.7\pm{0.5}$ cgs dex., Comment: 14 pages, 7 figures, accepted for publication in AJ

The Super-Kamiokande and T2K collaborations present a joint measurement of neutrino oscillation parameters from their atmospheric and beam neutrino data. It uses a common interaction model for events overlapping in neutrino energy and correlated detector systematic uncertainties between the two datasets, which are found to be compatible. Using 3244.4 days of atmospheric data and a beam exposure of $19.7(16.3) \times 10^{20}$ protons on target in (anti)neutrino mode, the analysis finds a 1.9$\sigma$ exclusion of CP-conservation (defined as $J_{CP}=0$) and a preference for the normal mass ordering., Comment: 10 pages, 3 figures

Automated volumetric segmentation of the pancreas on cross-sectional imaging is needed for diagnosis and follow-up of pancreatic diseases. While CT-based pancreatic segmentation is more established, MRI-based segmentation methods are understudied, largely due to a lack of publicly available datasets, benchmarking research efforts, and domain-specific deep learning methods. In this retrospective study, we collected a large dataset (767 scans from 499 participants) of T1-weighted (T1W) and T2-weighted (T2W) abdominal MRI series from five centers between March 2004 and November 2022. We also collected CT scans of 1,350 patients from publicly available sources for benchmarking purposes. We developed a new pancreas segmentation method, called PanSegNet, combining the strengths of nnUNet and a Transformer network with a new linear attention module enabling volumetric computation. We tested PanSegNet's accuracy in cross-modality (a total of 2,117 scans) and cross-center settings with Dice and Hausdorff distance (HD95) evaluation metrics. We used Cohen's kappa statistics for intra and inter-rater agreement evaluation and paired t-tests for volume and Dice comparisons, respectively. For segmentation accuracy, we achieved Dice coefficients of 88.3% (std: 7.2%, at case level) with CT, 85.0% (std: 7.9%) with T1W MRI, and 86.3% (std: 6.4%) with T2W MRI. There was a high correlation for pancreas volume prediction with R^2 of 0.91, 0.84, and 0.85 for CT, T1W, and T2W, respectively. We found moderate inter-observer (0.624 and 0.638 for T1W and T2W MRI, respectively) and high intra-observer agreement scores. All MRI data is made available at https://osf.io/kysnj/. Our source code is available at https://github.com/NUBagciLab/PaNSegNet., Comment: under review version

We present the projected rotational velocity and molecular abundances for HD 33632 Ab obtained via Keck Planet Imager and Characterizer high-resolution spectroscopy. HD 33632 Ab is a nearby benchmark brown dwarf companion at a separation of $\sim$20 au that straddles the L/T transition. Using a forward-modeling framework with on-axis host star spectra, self-consistent substellar atmospheric and retrieval models for HD 33632 Ab, we derive a projected rotational velocity of 53 $\pm$ 3 km/s and carbon/water mass fractions of log CO = $-$2.3 $\pm$ 0.3 and log H$_2$O = $-$2.7 $\pm$ 0.2. The inferred carbon-to-oxygen ratio (C/O = 0.58 $\pm$ 0.14), molecular abundances, and metallicity ([C/H] = 0.0 $\pm$ 0.2 dex) of HD 33632 Ab are consistent with its host star. Although detectable methane opacities are expected in L/T transition objects, we did not recover methane in our KPIC spectra, partly due to the high $v\sin{i}$ and to disequilibrium chemistry at the pressures we are sensitive to. We parameterize the spin as the ratio of rotation over break-up velocity, and compare HD 33632 Ab to a compilation of >200 very low-mass objects (M$\lesssim$0.1 M$_{\odot}$) that have spin measurements in the literature. There appears to be no clear trend for the isolated field low-mass objects versus mass, but a tentative trend is identified for low-mass companions and directly imaged exoplanets, similar to previous findings. A larger sample of close-in gas giant exoplanets and brown dwarfs will critically examine our understanding of their formation and evolution through rotation and chemical abundance measurements., Comment: Accepted for publication in the Astrophysical Journal. 36 pages, 15 figures, 5 tables