Search

This brief summarizes the published Cochrane Review, "Healthy Eating Interventions Delivered in Early Childhood Education and Care Settings for Improving the Diet of Children Aged Six Months to Six Years," by S. L. Yoong, M. Lum, L. Wolfenden, J. Jackson, C. Barnes, A. E. Hall, S. McCrabb, N. Pearson, C. Lane, J. Z. Jones, E. Nolan, L. Dinour, T. McDonnell, D. Booth, and A. Grady (https://doi.org/10.1002/14651858.CD013862.pub3). The Cochrane Review examines evidence for effectiveness of healthy eating interventions delivered in early childhood education and care (ECEC) settings for improving child dietary quality and health among children aged between 6 months and 6 years. Review authors conclude that ECEC-based healthy eating interventions may improve child diet quality slightly, potentially increasing fruit consumption (moderate-certainty evidence) and may have favorable effects on child weight (high-certainty evidence) and risk of being overweight or obese (moderate-certainty evidence). Overall, the review supports healthy eating interventions delivered in ECEC settings on several outcome measures; however, the certainty of evidence is moderate to very low because of considerable heterogeneity, potential publication bias, and high/unclear risks of bias. The review and original commentary acknowledge the limitations of the evidence, suggesting that future trials examine the impact of specific intervention components and focus on populations in low- and lower-middle-income countries. The original commentary also describes other relevant research focused on healthy eating interventions in ECEC settings that have shown an impact.

Ensuring that Large Language Models (LLMs) return just responses which adhere to societal values is crucial for their broader application. Prior research has shown that LLMs often fail to perform satisfactorily on tasks requiring moral cognizance, such as ethics-based judgments. While current approaches have focused on fine-tuning LLMs with curated datasets to improve their capabilities on such tasks, choosing the optimal learning paradigm to enhance the ethical responses of LLMs remains an open research debate. In this work, we aim to address this fundamental question: can current learning paradigms enable LLMs to acquire sufficient moral reasoning capabilities? Drawing from distributional semantics theory and the pragmatic nature of moral discourse, our analysis indicates that performance improvements follow a mechanism similar to that of semantic-level tasks, and therefore remain affected by the pragmatic nature of morals latent in discourse, a phenomenon we name the pragmatic dilemma. We conclude that this pragmatic dilemma imposes significant limitations on the generalization ability of current learning paradigms, making it the primary bottleneck for moral reasoning acquisition in LLMs.

We measure radial stellar age gradients in the relatively isolated gas-rich dwarf irregular WLM, combining JWST NIRCam and NIRISS imaging with six archival Hubble fields over semi-major axis equivalent distances of 0$\lesssim$R$_{SMA}$$\lesssim$4 kpc ($\lesssim$3R$_{hl}$). Fitting lifetime star formation histories (SFHs) to resolved color-magnitude diagrams (CMDs), radial age gradients are quantified using $\tau_{90}$ and $\tau_{50}$, the lookback times to form 90\% and 50\% of the cumulative stellar mass. We find that globally, the outskirts of WLM are older on average, with ($\delta$$\tau_{90}$, $\delta$$\tau_{50}$)/$\delta$R$_{SMA}=$(0.82$^{+0.10}_{-0.10}$, 1.60$^{+0.23}_{-0.22}$) Gyr/kpc (stat.), in good agreement with simulations. However, we also detect an azimuthal dependence of radial stellar age gradients, finding that stars on the leading edge of WLM (relative to its proper motion) are both younger and have a flatter age gradient compared to the trailing edge. This difference persists over 0.6$\lesssim$R$_{SMA}$$\lesssim$3.2 kpc ($\sim$0.5$-$2.5R$_{hl}$) and lookback times up to $\sim$8 Gyr, and is robust to assumed stellar evolutionary model. Our results are consistent with star formation triggered by ram pressure stripping from a circumgalactic and/or intergalactic medium, suggested by recent HI observations. If confirmed, processes typifying dense environments, such as ram pressure stripping, may be more relevant to the evolution of isolated galaxies than previously thought., Comment: ApJ in press. 23 pages, 8 figures, 3 tables

Recent work has highlighted the risks of LLM-generated content for a wide range of harmful behaviors, including incorrect and harmful code. In this work, we extend this by studying whether LLM-generated web design contains dark patterns. This work evaluated designs of ecommerce web components generated by four popular LLMs: Claude, GPT, Gemini, and Llama. We tested 13 commonly used ecommerce components (e.g., search, product reviews) and used them as prompts to generate a total of 312 components across all models. Over one-third of generated components contain at least one dark pattern. The majority of dark pattern strategies involve hiding crucial information, limiting users' actions, and manipulating them into making decisions through a sense of urgency. Dark patterns are also more frequently produced in components that are related to company interests. These findings highlight the need for interventions to prevent dark patterns during front-end code generation with LLMs and emphasize the importance of expanding ethical design education to a broader audience., Comment: 15 pages

The search for image compression optimization techniques is a topic of constant interest both in and out of academic circles. One method that shows promise toward future improvements in this field is image colorization since image colorization algorithms can reduce the amount of color data that needs to be stored for an image. Our work focuses on optimizing a color grid based approach to fully-automated image color information retention with regard to convolutional colorization network architecture for the purposes of image compression. More generally, using a convolutional neural network for image re-colorization, we want to minimize the amount of color information that is stored while still being able to faithfully re-color images. Our results yielded a promising image compression ratio, while still allowing for successful image recolorization reaching high CSIM values.

Modern science emerged from reasoning over repeatedly-observed planetary motions. We present Gravity-Bench-v1, an environment-based benchmark that challenges AI agents on tasks that parallel this historical development. Gravity-Bench-v1 evaluates agents on the discovery of physics concealed within a dynamic environment, using rigorous gravitational dynamics simulations. Gravity-Bench includes out-of-distribution cases, i.e. with physics that deviates from the real world, to evaluate true scientific generalization capabilities. Agents must plan to collect data within an experimental budget and must perform a dynamic form of data analysis and reasoning to solve tasks efficiently. Our benchmark admits an open-ended space of solutions. PhD-level solutions for each task are provided, to calibrate AI performance against human expertise. Technically at an upper-undergraduate level, our benchmark proves challenging to baseline AI agents. Gravity-Bench-v1 and planned extensions should help map out AI progress towards scientific discovery capabilities., Comment: Technical report - Work in progress

We present new multi-wavelength observations of two ultraluminous X-ray sources (ULXs) hosted by globular clusters (GCs) in the giant elliptical NGC 1399, focusing on CXO J0338318-352604 (GCU7), only the second GC ULX known to have luminous optical emission lines. Notably, only NII and OIII emission is observed in the optical spectra, suggesting H-poor material. Previous work suggested the possibility that the properties of GCU7 could be explained by the tidal disruption of a horizontal branch star by an intermediate-mass black hole. We use new data to show that the lack of evolution in the X-ray or optical properties of the source over the last 20 years rules out this scenario. Instead, we use CLOUDY simulations to demonstrate that the optical emission lines are consistent with an outflow from an ultra-compact X-ray binary where a compact object - likely a neutron star (NS) - is accreting above the Eddington limit from a helium white dwarf (He WD). This binary would have dynamically formed from a direct collision between a NS and a red giant, or else via an exchange interaction. The ULX is predicted to evolve to lower mass transfer rates over time and eventually become a doppelganger to the well-studied ultra-compact X-ray binaries in Galactic GCs such as 4U 1820-30. These results show the utility of using extragalactic GCs to study short-lived phases in dynamical binary evolution that occur too rarely to be observed in Galactic clusters., Comment: 11 pages, 9 figures, 4 tables. Accepted for publication in MNRAS

Deep reinforcement learning (DRL) has gained widespread adoption in control and decision-making tasks due to its strong performance in dynamic environments. However, DRL agents are vulnerable to noisy observations and adversarial attacks, and concerns about the adversarial robustness of DRL systems have emerged. Recent efforts have focused on addressing these robustness issues by establishing rigorous theoretical guarantees for the returns achieved by DRL agents in adversarial settings. Among these approaches, policy smoothing has proven to be an effective and scalable method for certifying the robustness of DRL agents. Nevertheless, existing certifiably robust DRL relies on policies trained with simple Gaussian augmentations, resulting in a suboptimal trade-off between certified robustness and certified return. To address this issue, we introduce a novel paradigm dubbed \texttt{C}ertified-r\texttt{A}dius-\texttt{M}aximizing \texttt{P}olicy (\texttt{CAMP}) training. \texttt{CAMP} is designed to enhance DRL policies, achieving better utility without compromising provable robustness. By leveraging the insight that the global certified radius can be derived from local certified radii based on training-time statistics, \texttt{CAMP} formulates a surrogate loss related to the local certified radius and optimizes the policy guided by this surrogate loss. We also introduce \textit{policy imitation} as a novel technique to stabilize \texttt{CAMP} training. Experimental results demonstrate that \texttt{CAMP} significantly improves the robustness-return trade-off across various tasks. Based on the results, \texttt{CAMP} can achieve up to twice the certified expected return compared to that of baselines. Our code is available at https://github.com/NeuralSec/camp-robust-rl., Comment: Accepted to USENIX Security Symposium 2025, Seattle, WA, USA. Source code is available at Github (https://github.com/NeuralSec/camp-robust-rl) and Zenodo (https://zenodo.org/records/14729675)

In some beyond-Einstein theories of gravity, gravitational waves can contain up to six polarizations, which are allowed to propagate at different speeds faster than light. These different propagation speeds imply that polarizations generated by the same source will not arrive simultaneously at the detector. Current constraints on the speed of propagation of transverse-traceless polarizations, however, indicate that any additional polarizations must arrive with or before the transverse-traceless ones. We propose a new technique to test for the existence of superluminal, non-transverse-traceless polarizations that arrive in the data before a gravitational-wave observation of transverse-traceless modes. We discuss the circumstances in which these non-transverse-traceless polarizations would be detectable and what constraints could be placed if they are not detected. To determine whether this new test of general relativity with gravitational wave observations is practical, we outline and address many of the challenges it might face. Our arguments lead us to conclude that this new test is not only physically well-motivated but also feasible with current detectors., Comment: 15 pages, 10 figures

We present a 5 month NICER X-ray monitoring campaign for two low luminosity active galactic nuclei (LLAGNs) -- NGC 4594 and IC 1459 -- with complementary Swift and NuSTAR observations. Utilizing an absorbed power law and thermal source model combined with NICER's SCORPEON background model, we demonstrate the effectiveness of joint source/background modeling for constraining emission from faint, background-dominated targets. Both sources are dominated by nuclear power law emission with photon indices $\Gamma \sim 1.5 - 2$, with NGC 4594 being slightly harder than IC 1459. The thermal contribution in both sources is fainter, but constant, with $kT \sim 0.5$ keV ($\sim 5 \times 10^6$ K). The power law flux and $\Gamma$ are strongly anti-correlated in both sources, as has been seen for other LLAGNs with radiatively inefficient accretion flows. NGC 4594 is the brighter source and exhibits significant aperiodic variability. Its variability timescale with an upper limit of $5 - 7$ days indicates emission originating from $< 100 R_{g}$, at the scale of the inner accretion flow. A spectral break found at $\sim 6$ keV, while tentative, could arise from synchrotron/inverse compton emission. This high-cadence LLAGN X-ray monitoring campaign underlines the importance of multi-wavelength variability studies for a sample of LLAGNs to truly understand their accretion and outflow physics., Comment: 18 pages, 5 figures, 5 tables. Accepted to ApJ

Detecting the first generation of stars, Population III (PopIII), has been a long-standing goal in astrophysics, yet they remain elusive even in the JWST era. Here we present a novel NIRCam-based selection method for PopIII galaxies, and carefully validate it through completeness and contamination simulations. We systematically search ~500 arcmin$^{2}$ across JWST legacy fields for PopIII candidates, including GLIMPSE which, assisted by gravitational lensing, has produced JWST's deepest NIRCam imaging thus far. We discover one promising PopIII galaxy candidate (GLIMPSE-16043) at $z=6.50^{+0.03}_{-0.24}$, a moderately lensed galaxy (mu=2.9) with an intrinsic UV magnitude of $M_{UV}$=-15.89. It exhibits key PopIII features: strong H$\alpha$ emission (rest-frame EW $2810\pm550$\AA); a Balmer jump; no dust (UV slope $\beta=-2.34\pm0.36$); and undetectable metal lines (e.g., [OIII]; [OIII]/H$\beta$<0.44) implying a gas-phase metallicity of Zgas/Zsun<0.5%. These properties indicate the presence of a nascent, metal-deficient young stellar population (<5Myr) with a stellar mass of $\simeq10^{5}M_{\odot}$. Intriguingly, this source deviates significantly from the extrapolated UV-metallicity relation derived from recent JWST observations at $z=4-10$, consistent with UV enhancement by a top-heavy PopIII initial mass function or the presence of an extremely metal-poor AGN. We also derive the first observational constraints on the PopIII UV luminosity function at z~6-7. The volume density of GLIMPSE-16043 ($\approx10^{-4}$ cMpc$^{-3}$) is in excellent agreement with theoretical predictions, independently reinforcing its plausibility. This study demonstrates the power of our novel NIRCam method to finally reveal distant galaxies even more pristine than the Milky Way's most metal-poor satellites, thereby promising to bring us closer to the first generation of stars than we have ever been before., Comment: 41 pages, 20 figures, and 10 Tables. Submitted to ApJ. Comments are welcome!

We describe the development of a one-credit course to promote AI literacy at The University of Texas at Austin. In response to a call for the rapid deployment of class to serve a broad audience in Fall of 2023, we designed a 14-week seminar-style course that incorporated an interdisciplinary group of speakers who lectured on topics ranging from the fundamentals of AI to societal concerns including disinformation and employment. University students, faculty, and staff, and even community members outside of the University, were invited to enroll in this online offering: The Essentials of AI for Life and Society. We collected feedback from course participants through weekly reflections and a final survey. Satisfyingly, we found that attendees reported gains in their AI literacy. We sought critical feedback through quantitative and qualitative analysis, which uncovered challenges in designing a course for this general audience. We utilized the course feedback to design a three-credit version of the course that is being offered in Fall of 2024. The lessons we learned and our plans for this new iteration may serve as a guide to instructors designing AI courses for a broad audience., Comment: Accepted to EAAI-25: The 15th Symposium on Educational Advances in Artificial Intelligence, collocated with AAAI-25

We present the discovery of PSR J1947-1120, a new huntsman millisecond pulsar with a red giant companion star in a 10.3 d orbit. This pulsar was found via optical, X-ray, and radio follow-up of the previously unassociated gamma-ray source 4FGL J1947.6-1121. PSR J1947-1120 is the second confirmed pulsar in the huntsman class and establishes this as a bona fide subclass of millisecond pulsar. We use MESA models to show that huntsman pulsars can be naturally explained as neutron star binaries whose secondaries are currently in the "red bump" region of the red giant branch, temporarily underfilling their Roche lobes and hence halting mass transfer. Huntsman pulsars offer a new view of the formation of typical millisecond pulsars, allowing novel constraints on the efficiency of mass transfer and recycling at an intermediate stage in the process., Comment: ApJ in press

Recent advancements in audio deepfake detection have leveraged graph neural networks (GNNs) to model frequency and temporal interdependencies in audio data, effectively identifying deepfake artifacts. However, the reliance of GNN-based methods on substantial labeled data for graph construction and robust performance limits their applicability in scenarios with limited labeled data. Although vast amounts of audio data exist, the process of labeling samples as genuine or fake remains labor-intensive and costly. To address this challenge, we propose SIGNL (Spatio-temporal vIsion Graph Non-contrastive Learning), a novel framework that maintains high GNN performance in low-label settings. SIGNL constructs spatio-temporal graphs by representing patches from the audio's visual spectrogram as nodes. These graph structures are modeled using vision graph convolutional (GC) encoders pre-trained through graph non-contrastive learning, a label-free that maximizes the similarity between positive pairs. The pre-trained encoders are then fine-tuned for audio deepfake detection, reducing reliance on labeled data. Experiments demonstrate that SIGNL outperforms state-of-the-art baselines across multiple audio deepfake detection datasets, achieving the lowest Equal Error Rate (EER) with as little as 5% labeled data. Additionally, SIGNL exhibits strong cross-domain generalization, achieving the lowest EER in evaluations involving diverse attack types and languages in the In-The-Wild dataset.

Beginner musicians often struggle to identify specific errors in their performances, such as playing incorrect notes or rhythms. There are two limitations in existing tools for music error detection: (1) Existing approaches rely on automatic alignment; therefore, they are prone to errors caused by small deviations between alignment targets.; (2) There is a lack of sufficient data to train music error detection models, resulting in over-reliance on heuristics. To address (1), we propose a novel transformer model, Polytune, that takes audio inputs and outputs annotated music scores. This model can be trained end-to-end to implicitly align and compare performance audio with music scores through latent space representations. To address (2), we present a novel data generation technique capable of creating large-scale synthetic music error datasets. Our approach achieves a 64.1% average Error Detection F1 score, improving upon prior work by 40 percentage points across 14 instruments. Additionally, compared with existing transcription methods repurposed for music error detection, our model can handle multiple instruments. Our source code and datasets are available at https://github.com/ben2002chou/Polytune., Comment: AAAI 2025

The purpose of this study was to examine the relationship between students' engagement with Legends of Learning's Math Basecamp (MBC) and students' math achievement at the elementary school level in Rialto Unified School District (RUSD) in California. Researchers examined the association between MBC usage and student achievement scores on the Smarter Balanced Mathematics assessment for 3,273 students in grades 3-5 using ordinary least squares (OLS) regression. Additionally, researchers used propensity score matching to determine a causal relationship between MBC usage and math scores for 1,608 students. OLS regression results reveal that as the number of sessions of Legends of Learning's Math Basecamp increases, achievement on California's Smarter Balanced Mathematics assessment increases significantly. Further, propensity score matching results indicate that completing at least 24 sessions of MBC significantly increases student scores on the Smarter Balanced Mathematics assessment. Further research is needed to confirm a causal relationship between participation in Math Basecamp and math achievement, which could be done in a Randomized Controlled Trial (RCT) or a robust quasi-experimental design (QED). Expanding future studies to include additional populations and settings, as well as additional analytic approaches can confirm the robustness of these findings and the generalizability of Math Basecamp's impact. [This report was prepared for Legends of Learning.]

Muscles consume metabolic energy (ATP) to produce force. A mathematical model for energy expenditure can be useful in estimating real-time costs of movements or to predict energy optimal movements. Metabolic cost models developed so far have predominantly aimed at dynamic movement tasks, where mechanical work dominates. Further, while it is known that both force magnitude and rate of change of force (force rate) affect metabolic cost, it is not known how these terms interact, or if the force rate dependence can be a consequence of the force dependence. Here, we performed extensive human subject experiments, involving each subject over 5 hours of metabolic trials, which systematically changed the mean forces and forces rates so as to characterize a holistic relation for metabolic cost based on both force and force rate -- or analogously, torque and torque rate. Our experiments involved humans producing symmetric or asymmetric sinusoidal forces with different means, amplitudes, frequencies, and rise and fall periods. We showed that the metabolic cost can be well-approximated by a sum of power law functions of torque and torque rate. We found that the metabolic cost scales non-linearly with joint torque (with exponent = 1.36) and non-linearly with torque rate (with exponent = 2.5). Surprisingly, the data suggested that the cost was roughly four times higher for decreasing the torque than increasing, mirroring the analogous ratio between the cost of positive and negative work. Using these metabolic cost relations, we show that if the metabolic cost scales with particular exponents with muscle force and force rates, the same exponents will be observed in multi-joint tasks with multiple muscles. Our new metabolic cost model involving both force and force rate will potentially allow better predictions of energy optimal movements and thus inform wearable robot design and analysis.

We introduce Switch-a-View, a model that learns to automatically select the viewpoint to display at each timepoint when creating a how-to video. The key insight of our approach is how to train such a model from unlabeled--but human-edited--video samples. We pose a pretext task that pseudo-labels segments in the training videos for their primary viewpoint (egocentric or exocentric), and then discovers the patterns between those view-switch moments on the one hand and the visual and spoken content in the how-to video on the other hand. Armed with this predictor, our model then takes an unseen multi-view video as input and orchestrates which viewpoint should be displayed when. We further introduce a few-shot training setting that permits steering the model towards a new data domain. We demonstrate our idea on a variety of real-world video from HowTo100M and Ego-Exo4D and rigorously validate its advantages.

The security of Large Language Models (LLMs) has become an important research topic since the emergence of ChatGPT. Though there have been various effective methods to defend against jailbreak attacks, prefilling attacks remain an unsolved and popular threat against open-sourced LLMs. In-Context Learning (ICL) offers a computationally efficient defense against various jailbreak attacks, yet no effective ICL methods have been developed to counter prefilling attacks. In this paper, we: (1) show that ICL can effectively defend against prefilling jailbreak attacks by employing adversative sentence structures within demonstrations; (2) characterize the effectiveness of this defense through the lens of model size, number of demonstrations, over-defense, integration with other jailbreak attacks, and the presence of safety alignment. Given the experimental results and our analysis, we conclude that there is no free lunch for defending against prefilling jailbreak attacks with ICL. On the one hand, current safety alignment methods fail to mitigate prefilling jailbreak attacks, but adversative structures within ICL demonstrations provide robust defense across various model sizes and complex jailbreak attacks. On the other hand, LLMs exhibit similar over-defensiveness when utilizing ICL demonstrations with adversative structures, and this behavior appears to be independent of model size.

Letters of recommendation are a common tool used in graduate admissions. Most admissions systems require three letters for each applicant, burdening both letter writers and admissions committees with a heavy work load that may not be time well-spent. Most applicants do not have three research advisors who can comment meaningfully on research readiness, adding a large number of letters that are not useful. Ideally, letters of recommendation will showcase the students' promise for a research career, but in practice, the letters often do not fulfill this purpose. As a group of early and mid-career faculty who write dozens of letters every year for promising undergraduates, we are concerned and overburdened by the inefficiencies of the current system. In this open letter to the AAS Graduate Admissions Task Force, we offer an alternative to the current use of letters of recommendation: a portfolio submitted by the student, which highlights e.g., a paper, plot, or presentation that represents their past work and readiness for grad school, uploaded to a centralized system used by astronomy and astrophysics PhD programs. While we argue that we could eliminate letters in this new paradigm, it may instead be advisable to limit the number of letters of recommendation to one per applicant., Comment: Accepted for publication to the Bulletin of the AAS

Intermediate mass black holes (10^2 < M_BH< 10^5 Msun) are an open question in our understanding of black hole evolution and growth. They have long been linked to dense star cluster environments thanks to cluster dynamics, but there are a limited number of secure detections. We leverage existing X-ray observations from Chandra X-ray Observatory and optical catalogs from Hubble Space Telescope with new radio observations from the Karl G. Jansky Very Large Array to search for any evidence of accreting black holes in young massive clusters in the nearby galaxy M51. We find that of 43 bright ($L_X > 10^{38}$ erg/s) X-ray point sources in M51, 24 had probable matches to objects including possible associated star clusters in the HST Legacy Extragalactic UV Survey catalog, seven of which were classified as contaminants (background galaxies or foreground stars). We explore the optical properties of the remaining 17 sources, including cluster age and mass estimates, and search for radio counterparts in the 8-12 GHz band. The lack of radio counterparts to X-ray sources we know to be associated with young massive clusters in M51 suggests that we do not significantly detect hard-state IMBHs ~ 10^4 Msun or above. However, more sensitive radio facilities like the Square Kilometre Array and next generation Very Large Array may be able to provide evidence for IMBHs with masses down to ~ 10^3 Msun., Comment: 18 pages, 7 figures, accepted to ApJ

While image captioning provides isolated descriptions for individual images, and video captioning offers one single narrative for an entire video clip, our work explores an important middle ground: progress-aware video captioning at the frame level. This novel task aims to generate temporally fine-grained captions that not only accurately describe each frame but also capture the subtle progression of actions throughout a video sequence. Despite the strong capabilities of existing leading vision language models, they often struggle to discern the nuances of frame-wise differences. To address this, we propose ProgressCaptioner, a captioning model designed to capture the fine-grained temporal dynamics within an action sequence. Alongside, we develop the FrameCap dataset to support training and the FrameCapEval benchmark to assess caption quality. The results demonstrate that ProgressCaptioner significantly surpasses leading captioning models, producing precise captions that accurately capture action progression and set a new standard for temporal precision in video captioning. Finally, we showcase practical applications of our approach, specifically in aiding keyframe selection and advancing video understanding, highlighting its broad utility., Comment: Project website: https://vision.cs.utexas.edu/projects/ProgressCaptioner/

We investigate whether the pre-trained knowledge of vision-language models (VLMs), such as CLIP, can be retained or even enhanced during continual learning (CL) while absorbing knowledge from a data stream. Existing methods often rely on additional reference data, isolated components for distribution or domain predictions, leading to high training costs, increased inference complexity, and limited improvement potential for pre-trained models. To address these challenges, we first comprehensively analyze the effects of parameter update locations and ranks on downstream adaptation and knowledge retention. Based on these insights, we propose Dynamic Rank-Selective Low Rank Adaptation (LoRA), a universal and efficient CL approach that adaptively assigns ranks to LoRA modules based on their relevance to the current data. Unlike prior methods, our approach continually enhances the pre-trained VLM by retaining both the pre-trained knowledge and the knowledge acquired during CL. Our approach eliminates the need for explicit domain or distribution prediction and additional reference data, enabling seamless integration of new tasks while preserving pre-trained capabilities. It also maintains the original architecture and deployment pipeline of the pre-trained model without incurring any additional inference overhead. Extensive experiments and analyses demonstrate that our method outperforms state-of-the-art approaches in continually absorbing knowledge of downstream tasks while retaining pre-trained knowledge., Comment: Preprint

Anticipating how a person will interact with objects in an environment is essential for activity understanding, but existing methods are limited to the 2D space of video frames-capturing physically ungrounded predictions of 'what' and ignoring the 'where' and 'how'. We introduce 4D future interaction prediction from videos. Given an input video of a human activity, the goal is to predict what objects at what 3D locations the person will interact with in the next time period (e.g., cabinet, fridge), and how they will execute that interaction (e.g., poses for bending, reaching, pulling). We propose a novel model FIction that fuses the past video observation of the person's actions and their environment to predict both the 'where' and 'how' of future interactions. Through comprehensive experiments on a variety of activities and real-world environments in Ego-Exo4D, we show that our proposed approach outperforms prior autoregressive and (lifted) 2D video models substantially, with more than 30% relative gains., Comment: Technical report

Multi-agent systems (MAS) utilizing multiple Large Language Model agents with Retrieval Augmented Generation and that can execute code locally may become beneficial in cosmological data analysis. Here, we illustrate a first small step towards AI-assisted analyses and a glimpse of the potential of MAS to automate and optimize scientific workflows in Cosmology. The system architecture of our example package, that builds upon the autogen/ag2 framework, can be applied to MAS in any area of quantitative scientific research. The particular task we apply our methods to is the cosmological parameter analysis of the Atacama Cosmology Telescope lensing power spectrum likelihood using Monte Carlo Markov Chains. Our work-in-progress code is open source and available at https://github.com/CMBAgents/cmbagent., Comment: Comments welcome!

Recently, a class of long-period radio transients (LPTs) has been discovered, exhibiting emission on timescales thousands of times longer than radio pulsars. Several models had been proposed implicating either a strong magnetic field neutron star, isolated white dwarf pulsar, or a white dwarf binary system with a low-mass companion. While several models for LPTs also predict X-ray emission, no LPTs have been detected in X-rays despite extensive searches. Here we report the discovery of an extremely bright LPT (10-20 Jy in radio), ASKAP J1832-0911, which has coincident radio and X-ray emission, both with a 44.2-minute period. The X-ray and radio luminosities are correlated and vary by several orders of magnitude. These properties are unique amongst known Galactic objects and require a new explanation. We consider a $\gtrsim0.5$ Myr old magnetar with a $\gtrsim 10^{13}$ G crustal field, or an extremely magnetised white dwarf in a binary system with a dwarf companion, to be plausible explanations for ASKAP J1832-0911, although both explanations pose significant challenges to formation and emission theories. The X-ray detection also establishes a new class of hour-scale periodic X-ray transients of luminosity $\sim10^{33}$ erg/s associated with exceptionally bright coherent radio emission., Comment: 52 pages, 17 figures, 3 tables

We report the discovery of five galaxy candidates at redshifts between $15.9

Transitioning from quantum computation on physical qubits to quantum computation on encoded, logical qubits can improve the error rate of operations, and will be essential for realizing valuable quantum computational advantages. Using a neutral atom quantum processor with 256 qubits, each an individual Ytterbium atom, we demonstrate the entanglement of 24 logical qubits using the distance-two [[4,2,2]] code, simultaneously detecting errors and correcting for lost qubits. We also implement the Bernstein-Vazirani algorithm with up to 28 logical qubits encoded in the [[4,1,2]] code, showing better-than-physical error rates. We demonstrate fault-tolerant quantum computation in our approach, guided by the proposal of Gottesman (2016), by performing repeated loss correction for both structured and random circuits encoded in the [[4,2,2]] code. Finally, since distance-two codes can correct qubit loss, but not other errors, we show repeated loss and error correction using the distance-three [[9,1,3]] Bacon-Shor code. These results begin to clear a path for achieving scientific quantum advantage with a programmable neutral atom quantum processor., Comment: 17 pages, 16 figures

We developed a shoe-mounted gait monitoring system capable of tracking up to 17 gait parameters, including gait length, step time, stride velocity, and others. The system employs a stereo camera mounted on one shoe to track a marker placed on the opposite shoe, enabling the estimation of spatial gait parameters. Additionally, a Force Sensitive Resistor (FSR) affixed to the heel of the shoe, combined with a custom-designed algorithm, is utilized to measure temporal gait parameters. Through testing on multiple participants and comparison with the gait mat, the proposed gait monitoring system exhibited notable performance, with the accuracy of all measured gait parameters exceeding 93.61%. The system also demonstrated a low drift of 4.89% during long-distance walking. A gait identification task conducted on participants using a trained Transformer model achieved 95.7% accuracy on the dataset collected by the proposed system, demonstrating that our hardware has the potential to collect long-sequence gait data suitable for integration with current Large Language Models (LLMs). The system is cost-effective, user-friendly, and well-suited for real-life measurements., Comment: 13 pages, 14 figures. This paper was submitted for publication to the IEEE Transactions on Instrumentation and Measurement

Given a multi-view video, which viewpoint is most informative for a human observer? Existing methods rely on heuristics or expensive "best-view" supervision to answer this question, limiting their applicability. We propose a weakly supervised approach that leverages language accompanying an instructional multi-view video as a means to recover its most informative viewpoint(s). Our key hypothesis is that the more accurately an individual view can predict a view-agnostic text summary, the more informative it is. To put this into action, we propose a framework that uses the relative accuracy of view-dependent caption predictions as a proxy for best view pseudo-labels. Then, those pseudo-labels are used to train a view selector, together with an auxiliary camera pose predictor that enhances view-sensitivity. During inference, our model takes as input only a multi-view video--no language or camera poses--and returns the best viewpoint to watch at each timestep. On two challenging datasets comprised of diverse multi-camera setups and how-to activities, our model consistently outperforms state-of-the-art baselines, both with quantitative metrics and human evaluation.

Expansion Microscopy is a super-resolution technique in which physically enlarging samples in an isotropic manner increases inter-molecular distances such that nano-scale structures can be resolved using light microscopy. This is particularly useful in neuroscience as many important structures are smaller than the diffraction limit. Since its invention in 2015, a variety of Expansion Microscopy protocols have been generated and applied to advance knowledge in many prominent organisms in neuroscience, including zebrafish, mice, Drosophila, and C. elegans. Here we review the last decade of Expansion Microscopy-enabled advances with a focus on neuroscience.

Self-correction is one of the most amazing emerging capabilities of Large Language Models (LLMs), enabling LLMs to self-modify an inappropriate output given a natural language feedback which describes the problems of that output. Moral self-correction is a post-hoc approach correcting unethical generations without requiring a gradient update, making it both computationally lightweight and capable of preserving the language modeling ability. Previous works have shown that LLMs can self-debias, and it has been reported that small models, i.e., those with less than 22B parameters, are not capable of moral self-correction. However, there is no direct proof as to why such smaller models fall short of moral self-correction, though previous research hypothesizes that larger models are skilled in following instructions and understanding abstract social norms. In this paper, we empirically validate this hypothesis in the context of social stereotyping, through meticulous prompting. Our experimental results indicate that (i) surprisingly, 3.8B LLMs with proper safety alignment fine-tuning can achieve very good moral self-correction performance, highlighting the significant effects of safety alignment; and (ii) small LLMs are indeed weaker than larger-scale models in terms of comprehending social norms and self-explanation through CoT, but all scales of LLMs show bad self-correction performance given unethical instructions.

Ultraluminous X-ray sources (ULXs) were once largely believed to be powered by super-Eddington accretion onto stellar-mass black holes, although in some rare cases, ULXs also serve as potential candidates for (sub-Eddington) intermediate mass black holes. However, a total of eight ULXs have now been confirmed to be powered by neutron stars, thanks to observed pulsations, and may act as contaminants for radio/X-ray selection of intermediate mass black holes. Here we present the first comprehensive radio study of seven known neutron star ULXs using new and archival data from the Karl G. Jansky Very Large Array and the Australia Telescope Compact Array, combined with the literature. Across this sample there is only one confident radio detection, from the Galactic neutron star ULX Swift J0243.6+6124. The other six objects in our sample are extragalactic, and only one has coincident radio emission, which we conclude is most likely contamination from a background HII region. We conclude that with current facilities, neutron star ULXs do not produce significant enough radio emission to cause them to be misidentified as radio/X-ray selected intermediate mass black hole candidates. Thus, if background star formation has been properly considered, the current study indicates that a ULX with a compact radio counterpart is not likely to be a neutron star., Comment: 14 pages, 2 figures, accepted to ApJ