Search

The increasing capabilities of large generative models and their ever more widespread deployment have raised concerns about their reliability, safety, and potential misuse. To address these issues, recent works have proposed to control model generation by steering model activations in order to effectively induce or prevent the emergence of concepts or behaviors in the generated output. In this paper we introduce Activation Transport (AcT), a general framework to steer activations guided by optimal transport theory that generalizes many previous activation-steering works. AcT is modality-agnostic and provides fine-grained control over the model behavior with negligible computational overhead, while minimally impacting model abilities. We experimentally show the effectiveness and versatility of our approach by addressing key challenges in large language models (LLMs) and text-to-image diffusion models (T2Is). For LLMs, we show that AcT can effectively mitigate toxicity, induce arbitrary concepts, and increase their truthfulness. In T2Is, we show how AcT enables fine-grained style control and concept negation.

Despite the increasing demand for safer machine learning practices, the use of Uncertainty Quantification (UQ) methods in production remains limited. This limitation is exacerbated by the challenge of validating UQ methods in absence of UQ ground truth. In classification tasks, when only a usual set of test data is at hand, several authors suggested different metrics that can be computed from such test points while assessing the quality of quantified uncertainties. This paper investigates such metrics and proves that they are theoretically well-behaved and actually tied to some uncertainty ground truth which is easily interpretable in terms of model prediction trustworthiness ranking. Equipped with those new results, and given the applicability of those metrics in the usual supervised paradigm, we argue that our contributions will help promoting a broader use of UQ in deep learning.

Optical clocks have improved their frequency stability and estimated accuracy by more than two orders of magnitude over the best caesium microwave clocks that realise the SI second. Accordingly, an optical redefinition of the second has been widely discussed, prompting a need for the consistency of optical clocks to be verified worldwide. While satellite frequency links are sufficient to compare microwave clocks, a suitable method for comparing high-performance optical clocks over intercontinental distances is missing. Furthermore, remote comparisons over frequency links face fractional uncertainties of a few $10^{-18}$ due to imprecise knowledge of each clock's relativistic redshift, which stems from uncertainty in the geopotential determined at each distant location. Here, we report a landmark campaign towards the era of optical clocks, where, for the first time, state-of-the-art transportable optical clocks from Japan and Europe are brought together to demonstrate international comparisons that require neither a high-performance frequency link nor information on the geopotential difference between remote sites. Conversely, the reproducibility of the clocks after being transported between countries was sufficient to determine geopotential height offsets at the level of 4 cm. Our campaign paves the way for redefining the SI second and has a significant impact on various applications, including tests of general relativity, geodetic sensing for geosciences, precise navigation, and future timing networks., Comment: 29 pages, 5 figures

Hyperparameters are configuration variables controlling the behavior of machine learning algorithms. They are ubiquitous in machine learning and artificial intelligence and the choice of their values determine the effectiveness of systems based on these technologies. Manual hyperparameter search is often unsatisfactory and becomes unfeasible when the number of hyperparameters is large. Automating the search is an important step towards automating machine learning, freeing researchers and practitioners alike from the burden of finding a good set of hyperparameters by trial and error. In this survey, we present a unified treatment of hyperparameter optimization, providing the reader with examples and insights into the state-of-the-art. We cover the main families of techniques to automate hyperparameter search, often referred to as hyperparameter optimization or tuning, including random and quasi-random search, bandit-, model- and gradient- based approaches. We further discuss extensions, including online, constrained, and multi-objective formulations, touch upon connections with other fields such as meta-learning and neural architecture search, and conclude with open questions and future research directions., Comment: Preprint

Electrons in solids owe their properties to the periodic potential landscapes they experience. The advent of moir\'e lattices has revolutionized our ability to engineer such landscapes on nanometer scales, leading to numerous groundbreaking discoveries. Despite this progress, direct imaging of these electrostatic potential landscapes remains elusive. In this work, we introduce the Atomic Single Electron Transistor (SET), a novel scanning probe utilizing a single atomic defect in a van der Waals (vdW) material, which serves as an ultrasensitive, high-resolution potential imaging sensor. Built upon the quantum twisting microscope (QTM) platform, this probe leverages the QTM's distinctive capability to form a pristine, scannable 2D interface between vdW heterostructures. Using the Atomic SET, we present the first direct images of the electrostatic potential in one of the most canonical moir\'e interfaces: graphene aligned to hexagonal boron nitride. Our results reveal that this potential exhibits an approximate C6 symmetry, has minimal dependence on the carrier density, and has a substantial magnitude of ~60 mV even in the absence of carriers. Theoretically, the observed symmetry can only be explained by a delicate interplay of physical mechanisms with competing symmetries. Intriguingly, the magnitude of the measured potential significantly exceeds theoretical predictions, suggesting that current understanding may be incomplete. With a spatial resolution of 1 nm and a sensitivity to detect the potential of even a few millionths of an electron charge, the Atomic SET opens the door for ultrasensitive imaging of charge order and thermodynamic properties for a range of quantum phenomena, including various symmetry-broken phases, quantum crystals, vortex charges, and fractionalized quasiparticles.

Semiconducting CrSBr is a layered A-type antiferromagnet, with individual layers antiferromagnetically coupled along the stacking direction. Due to its unique orthorhombic crystal structure, CrSBr exhibits highly anisotropic mechanical and optoelectronic properties acting itself as a quasi-1D material. CrSBr demonstrates complex coupling phenomena involving phonons, excitons, magnons, and polaritons. Here we show through polarization-resolved resonant Raman scattering the intricate interaction between the vibrational and electronic properties of CrSBr. For samples spanning from few-layer to bulk thickness, we observe that the polarization of the A$_g^2$ Raman mode can be rotated by 90 degrees, shifting from alignment with the crystallographic a (intermediate magnetic) axis to the b (easy magnetic) axis, depending on the excitation energy. In contrast, the A$_g^1$ and A$_g^3$ modes consistently remain polarized along the b axis, regardless of the laser energy used. We access real and imaginary parts of the Raman tensor in our analysis, uncovering resonant electron-phonon coupling., Comment: Main text + Supplementary Information

We consider nonlinear dispersive equations of Schr\"odinger-type involving fractional powers $0

It has been claimed that graphite hosts superconductivity at room temperature, although all efforts to isolate it have been vain. Here we report a separation method that uses magnetic field gradients to sort the superconducting from normal grains out of industrial graphite powders. We have obtained a concentrate of above room temperature superconducting particles. Electrical resistance measurements on agglomerates of sorted grains of three types of graphite show transition temperatures up to T$_{c{_{onset}}} \sim$ 700K with zero resistance up to $\sim$ 500K. Magnetization measurements confirm these values through jumps at \textit{T$_c$} in the zero field cooled curves, and by the occurrence diamagnetic hysteretic cycles shrinking with temperature. Our results open the door towards the study of above room temperature superconducting ill-stacked graphite phases., Comment: 21 pages, 9 figures

Recent advances in memory technologies, devices and materials have shown great potential for integration into neuromorphic electronic systems. However, a significant gap remains between the development of these materials and the realization of large-scale, fully functional systems. One key challenge is determining which devices and materials are best suited for specific functions and how they can be paired with CMOS circuitry. To address this, we introduce TEXEL, a mixed-signal neuromorphic architecture designed to explore the integration of on-chip learning circuits and novel two- and three-terminal devices. TEXEL serves as an accessible platform to bridge the gap between CMOS-based neuromorphic computation and the latest advancements in emerging devices. In this paper, we demonstrate the readiness of TEXEL for device integration through comprehensive chip measurements and simulations. TEXEL provides a practical system for testing bio-inspired learning algorithms alongside emerging devices, establishing a tangible link between brain-inspired computation and cutting-edge device research., Comment: 17 pages, 7 figures. Supplementary material: 8 pages, 4 figures

Self-supervised learning on tabular data seeks to apply advances from natural language and image domains to the diverse domain of tables. However, current techniques often struggle with integrating multi-domain data and require data cleaning or specific structural requirements, limiting the scalability of pre-training datasets. We introduce PORTAL (Pretraining One-Row-at-a-Time for All tabLes), a framework that handles various data modalities without the need for cleaning or preprocessing. This simple yet powerful approach can be effectively pre-trained on online-collected datasets and fine-tuned to match state-of-the-art methods on complex classification and regression tasks. This work offers a practical advancement in self-supervised learning for large-scale tabular data., Comment: Accepted at Table Representation Learning Workshop at NeurIPS 2024

Stellar active regions like spots and faculae can distort the shapes of spectral lines, inducing variations in the radial velocities that are often orders of magnitude larger than the signals from Earth-like planets. Efforts to mitigate these activity signals have hitherto focused on either the time or the velocity (wavelength) domains. We present a physics-driven Gaussian process (GP) framework to model activity signals directly in time series of line profiles or Cross-Correlation Functions (CCFs). Unlike existing methods which correct activity signals in line profile time series, our approach exploits the time correlation between velocity (wavelength) bins in the line profile variations, and is based on a simplified but physically motivated model for the origin of these variations. When tested on both synthetic and real data sets with signal-to-noise ratios down to $\sim$ 100, our method was able to separate the planetary signal from the activity signal, even when their periods were identical. We also conducted injection/recovery tests using two years of realistically sampled HARPS-N solar data, demonstrating the ability of the method to accurately recover a signal induced by a 1.5-Earth mass planet with a semi-amplitude of 0.3 m/s and a period of 33 days during high solar activity., Comment: Accepted for publication in MNRAS

Gyroscopes are inertial sensors that measure the angular velocity of the platforms to which they are attached. To estimate the gyroscope deterministic error terms prior mission start, a calibration procedure is performed. When considering low-cost gyroscopes, the calibration requires a turntable as the gyros are incapable of sensing the Earth turn rate. In this paper, we propose a data-driven framework to estimate the scale factor and bias of a gyroscope. To train and validate our approach, a dataset of 56 minutes was recorded using a turntable. We demonstrated that our proposed approach outperforms the model-based approach, in terms of accuracy and convergence time. Specifically, we improved the scale factor and bias estimation by an average of 72% during six seconds of calibration time, demonstrating an average of 75% calibration time improvement. That is, instead of minutes, our approach requires only several seconds for the calibration., Comment: 19 Pages, 5 Figures, 3 Tables

The multiplicities of positive and negative pions, kaons and unidentified hadrons produced in deep-inelastic scattering are measured in bins of the Bjorken scaling variable $x$, the relative virtual-photon energy $y$ and the fraction of the virtual-photon energy transferred to the final-state hadron $z$. Data were obtained by the COMPASS Collaboration using a 160 GeV muon beam of both electric charges and a liquid hydrogen target. These measurements cover the kinematic domain with photon virtuality $Q^2 > 1$ (GeV/$c)^2$, $0.004 < x < 0.4$, $0.1 < y < 0.7$ and $0.2 < z < 0.85$, in accordance with the kinematic domain used in earlier published COMPASS multiplicity measurements with an isoscalar target. The calculation of radiative corrections was improved by using the Monte Carlo generator DJANGOH, which results in up to 12\% larger corrections in the low-$x$ region., Comment: 19 pages, 29 figures

Quantitative evaluation of echocardiography is essential for precise assessment of cardiac condition, monitoring disease progression, and guiding treatment decisions. The diverse nature of echo images, including variations in probe types, manufacturers, and pathologies, poses challenges for developing artificial intelligent models that can generalize across different clinical practice. We introduce EchoApex, the first general-purpose vision foundation model echocardiography with applications on a variety of clinical practice. Leveraging self-supervised learning, EchoApex is pretrained on over 20 million echo images from 11 clinical centres. By incorporating task-specific decoders and adapter modules, we demonstrate the effectiveness of EchoApex on 4 different kind of clinical applications with 28 sub-tasks, including view classification, interactive structure segmentation, left ventricle hypertrophy detection and automated ejection fraction estimation from view sequences. Compared to state-of-the-art task-specific models, EchoApex attains improved performance with a unified image encoding architecture, demonstrating the benefits of model pretraining at scale with in-domain data. Furthermore, EchoApex illustrates the potential for developing a general-purpose vision foundation model tailored specifically for echocardiography, capable of addressing a diverse range of clinical applications with high efficiency and efficacy.

We propose a universal framework for classifying quantum states based on their scale-resolved correlation structure. Using the recently introduced information lattice, which provides an operational definition of the total amount of correlations at each scale, we define intrinsic characteristic length scales of quantum states. We analyze ground and midspectrum eigenstates of the disordered interacting Kitaev chain, showing that our framework provides a novel unbiased approach to quantum matter., Comment: 7 pages, 3 figures

The use of Large Language Models (LLMs) in automated test generation is gaining popularity, with much of the research focusing on metrics like compilability rate, code coverage and bug detection. However, an equally important quality metric is the presence of test smells design flaws or anti patterns in test code that hinder maintainability and readability. In this study, we explore the diffusion of test smells in LLM generated unit test suites and compare them to those found in human written ones. We analyze a benchmark of 20,500 LLM-generated test suites produced by four models (GPT-3.5, GPT-4, Mistral 7B, and Mixtral 8x7B) across five prompt engineering techniques, alongside a dataset of 780,144 human written test suites from 34,637 projects. Leveraging TsDetect, a state of the art tool capable of detecting 21 different types of test smells, we identify and analyze the prevalence and co-occurrence of various test smells in both human written and LLM-generated test suites. Our findings reveal new insights into the strengths and limitations of LLMs in test generation. First, regarding prevalence, we observe that LLMs frequently generate tests with common test smells, such as Magic Number Test and Assertion Roulette. Second, in terms of co occurrence, certain smells, like Long Test and Useless Test, tend to co occur in LLM-generated suites, influenced by specific prompt techniques. Third, we find that project complexity and LLM specific factors, including model size and context length, significantly affect the prevalence of test smells. Finally, the patterns of test smells in LLM-generated tests often mirror those in human-written tests, suggesting potential data leakage from training datasets. These insights underscore the need to refine LLM-based test generation for cleaner code and suggest improvements in both LLM capabilities and software testing practices.

Early detection of intrapartum risk enables interventions to potentially prevent or mitigate adverse labor outcomes such as cerebral palsy. Currently, there is no accurate automated system to predict such events to assist with clinical decision-making. To fill this gap, we propose "Artificial Intelligence (AI) for Modeling and Explaining Neonatal Health" (AIMEN), a deep learning framework that not only predicts adverse labor outcomes from maternal, fetal, obstetrical, and intrapartum risk factors but also provides the model's reasoning behind the predictions made. The latter can provide insights into what modifications in the input variables of the model could have changed the predicted outcome. We address the challenges of imbalance and small datasets by synthesizing additional training data using Adaptive Synthetic Sampling (ADASYN) and Conditional Tabular Generative Adversarial Networks (CTGAN). AIMEN uses an ensemble of fully-connected neural networks as the backbone for its classification with the data augmentation supported by either ADASYN or CTGAN. AIMEN, supported by CTGAN, outperforms AIMEN supported by ADASYN in classification. AIMEN can predict a high risk for adverse labor outcomes with an average F1 score of 0.784. It also provides counterfactual explanations that can be achieved by changing 2 to 3 attributes on average. Resources available: https://github.com/ab9mamun/AIMEN., Comment: 17 pages, 8 figures

DNA-Encoded Libraries (DEL) are combinatorial small molecule libraries that offer an efficient way to characterize diverse chemical spaces. Selection experiments using DELs are pivotal to drug discovery efforts, enabling high-throughput screens for hit finding. However, limited availability of public DEL datasets hinders the advancement of computational techniques designed to process such data. To bridge this gap, we present KinDEL, one of the first large, publicly available DEL datasets on two kinases: Mitogen-Activated Protein Kinase 14 (MAPK14) and Discoidin Domain Receptor Tyrosine Kinase 1 (DDR1). Interest in this data modality is growing due to its ability to generate extensive supervised chemical data that densely samples around select molecular structures. Demonstrating one such application of the data, we benchmark different machine learning techniques to develop predictive models for hit identification; in particular, we highlight recent structure-based probabilistic approaches. Finally, we provide biophysical assay data, both on- and off-DNA, to validate our models on a smaller subset of molecules. Data and code for our benchmarks can be found at: https://github.com/insitro/kindel.

Autonomous underwater vehicles (AUVs) are underwater robotic platforms used in a variety of applications. An AUV's navigation solution relies heavily on the fusion of inertial sensors and Doppler velocity logs (DVL), where the latter delivers accurate velocity updates. To ensure accurate navigation, a DVL calibration is undertaken before the mission begins to estimate its error terms. During calibration, the AUV follows a complex trajectory and employs nonlinear estimation filters to estimate error terms. In this paper, we introduce DCNet, a data-driven framework that utilizes a two-dimensional convolution kernel in an innovative way. Using DCNet and our proposed DVL error model, we offer a rapid calibration procedure. This can be applied to a trajectory with a nearly constant velocity. To train and test our proposed approach a dataset of 276 minutes long with real DVL recorded measurements was used. We demonstrated an average improvement of 70% in accuracy and 80% improvement in calibration time, compared to the baseline approach, with a low-performance DVL. As a result of those improvements, an AUV employing a low-cost DVL, can achieve higher accuracy, shorter calibration time, and apply a simple nearly constant velocity calibration trajectory. Our results also open up new applications for marine robotics utilizing low-cost, high-accurate DVLs., Comment: 10 Pages, 9 Figures, 5 Tables

A standard method to reduce the linewidth of semiconductor lasers involves the use of external optical feedback (EOF). However, feedback powers less than 1 % usually trigger coherence collapse (CC), leading to chaotic laser dynamics and linewidth broadening. This paper explores a method to mitigate CC through precise tuning of the feedback polarization depending on the feedback power. We report a semiconductor laser with a sub-kHz linewidth, achieved via EOF. The laser features a U-shaped cavity with two sampled grating distributed Bragg reflectors (SG-DBRs), enabling broad tunability across a 42 nm wavelength range (1513-1555 nm). By injecting optical feedback into both sides of the laser cavity via an external fiber-based cavity, we reduce the linewidth by more than three orders of magnitude, from MHz to sub-kHz across the laser's tuning range. Our approach achieves significant linewidth reduction while maintaining coherence at high feedback levels, marking an improvement over prior studies where CC limited performance. These results pave the way for ultra-narrow linewidth diode lasers with wide tunability, which would benefit fields like coherent optical communications and spectroscopy.

Efficient physics simulations are essential for numerous applications, ranging from realistic cloth animations or smoke effects in video games, to analyzing pollutant dispersion in environmental sciences, to calculating vehicle drag coefficients in engineering applications. Unfortunately, analytical solutions to the underlying physical equations are rarely available, and numerical solutions require high computational resources. Latest developments in the field of physics-based Deep Learning have led to promising efficiency improvements but still suffer from limited generalization capabilities and low accuracy compared to numerical solvers. In this work, we introduce Metamizer, a novel neural optimizer that iteratively solves a wide range of physical systems with high accuracy by minimizing a physics-based loss function. To this end, our approach leverages a scale-invariant architecture that enhances gradient descent updates to accelerate convergence. Since the neural network itself acts as an optimizer, training this neural optimizer falls into the category of meta-optimization approaches. We demonstrate that Metamizer achieves unprecedented accuracy for deep learning based approaches - sometimes approaching machine precision - across multiple PDEs after training on the Laplace, advection-diffusion and incompressible Navier-Stokes equation as well as on cloth simulations. Remarkably, the model also generalizes to PDEs that were not covered during training such as the Poisson, wave and Burgers equation. Our results suggest that Metamizer could have a profound impact on future numerical solvers, paving the way for fast and accurate neural physics simulations without the need for retraining.

One of the primary sources of gravitational waves (GWs) anticipated to be detected by the Laser Interferometer Space Antenna (LISA) are Galactic double white dwarf binaries (DWDs). However, most of these binaries will be unresolved, and their GWs will overlap incoherently, creating a stochastic noise known as the Galactic foreground. Similarly, the population of unresolved systems in the Milky Way's (MW) satellites is expected to contribute to a stochastic gravitational wave background (SGWB). Due to their anisotropy and the annual motion of the LISA constellation, both the Galactic foreground and the satellite SGWB fall into the category of cyclostationary processes. Leveraging this property, we develop a purely frequency-based method to study LISA's capability to detect the MW foreground and SGWBs from the most promising MW satellites. We analyze both mock data generated by an astrophysically motivated SGWB spectrum, and realistic ones from a DWD population generated via binary population synthesis. We are able to recover or put constrains on the candidate foregrounds, reconstructing -- in the presence of noise uncertainties -- their sky distribution and spectrum. Our findings highlight the significance of the interplay between the astrophysical spectrum and LISA's sensitivity to detect the satellites' SGWB. Considering an astrophysically motivated prior on the satellite positions improves their detectability, which becomes otherwise challenging in the presence of the Galactic foreground. Furthermore, we explore the potential to observe a hypothetical satellite located behind the Galactic disk. Our results suggest that a Large Magellanic Cloud-like satellite could indeed be observable by LISA., Comment: 19 pages, 13 figures, 2 tables

Upcoming space-based gravitational-wave detectors will be sensitive to millions and resolve tens of thousands of stellar-mass binary systems at mHz frequencies. The vast majority of these will be double white dwarfs in our Galaxy. The greatest part will remain unresolved, forming an incoherent stochastic foreground signal. Using state-of-the-art Galactic models for the formation and evolution of binary white dwarfs and accurate LISA simulated signals, we introduce a test for foreground Gaussianity and stationarity. We explain the former with a new analytical modulation induced by the LISA constellation motion and the intrinsic anisotropy of the source distribution. By demodulating the foreground signal, we reveal a deviation from Gaussianity in the 2-10 mHz frequency band. Our finding is crucial to design faithful data models, i.e. unbiased likelihoods for both individual sources and astrophysical foregrounds parameter estimation, and ultimately for an accurate interpretation of the LISA data., Comment: 17 pages, 8 figures

The effect of surprisal on processing difficulty has been a central topic of investigation in psycholinguistics. Here, we use eyetracking data to examine three language processing regimes that are common in daily life but have not been addressed with respect to this question: information seeking, repeated processing, and the combination of the two. Using standard regime-agnostic surprisal estimates we find that the prediction of surprisal theory regarding the presence of a linear effect of surprisal on processing times, extends to these regimes. However, when using surprisal estimates from regime-specific contexts that match the contexts and tasks given to humans, we find that in information seeking, such estimates do not improve the predictive power of processing times compared to standard surprisals. Further, regime-specific contexts yield near zero surprisal estimates with no predictive power for processing times in repeated reading. These findings point to misalignments of task and memory representations between humans and current language models, and question the extent to which such models can be used for estimating cognitively relevant quantities. We further discuss theoretical challenges posed by these results., Comment: Accepted to CoNLL

Diffusion and flow-matching models achieve remarkable generative performance but at the cost of many sampling steps, this slows inference and limits applicability to time-critical tasks. The ReFlow procedure can accelerate sampling by straightening generation trajectories. However, ReFlow is an iterative procedure, typically requiring training on simulated data, and results in reduced sample quality. To mitigate sample deterioration, we examine the design space of ReFlow and highlight potential pitfalls in prior heuristic practices. We then propose seven improvements for training dynamics, learning and inference, which are verified with thorough ablation studies on CIFAR10 $32 \times 32$, AFHQv2 $64 \times 64$, and FFHQ $64 \times 64$. Combining all our techniques, we achieve state-of-the-art FID scores (without / with guidance, resp.) for fast generation via neural ODEs: $2.23$ / $1.98$ on CIFAR10, $2.30$ / $1.91$ on AFHQv2, $2.84$ / $2.67$ on FFHQ, and $3.49$ / $1.74$ on ImageNet-64, all with merely $9$ neural function evaluations.

Bimodal software analysis initially appeared to be within reach with the advent of large language models. Unfortunately, the complex interplay of natural language text and code in software engineering, presents unique challenges that prevent pretrained models to generalize to a variety of tasks. We postulate that in-context learning for the code-text bimodality is a promising avenue. This paper thus introduces a comprehensive study of in-context code-text learning, focusing on leveraging pretrained CodeLLAMA models. We consider a diverse dataset encompassing 23 software engineering tasks, which we transform in an in-context learning format. To effectively extract informative features, we propose a configurable prompt template. Our proposed pipeline, InCTRL, then unifies prompt learning across various software engineering tasks. Extensive evaluation on the study datasets demonstrates the superiority of INCTRL-models in few-shot performance, surpassing state-of-the-art models including the support model, CodeLLAMA. Typically, we observe that applied to the CodeLLAMA model, INCTRL brings improvements in terms of precision (at least about 12\%) and recall (up to 93.88\%) on various tasks. For example, on the task of program repair, INCTRL improves the BLEU score of CodeLLAMA by 85 points, while for clone detection, INCTRL achieves an improvement of 69 percentage points. Moreover, INCTRL-models offer state-of-the-art performance when using retrieval-augmented generation on individual downstream tasks. Finally, we qualitatively analyze the benefits of INCTRL over CodeLLAMA and open-source all models for broader impact. We make our code and dataset publicly available at: \begin{center} {\url{https://anonymous.4open.science/r/inctrl-B65B}} \end{center}

Large language models (LLMs) exhibit remarkable reasoning abilities, allowing them to generalize across a wide range of downstream tasks, such as commonsense reasoning or instruction following. However, as LLMs scale, inference costs become increasingly prohibitive, accumulating significantly over their life cycle. This poses the question: Can we compress pre-trained LLMs to meet diverse size and latency requirements? We leverage Neural Architecture Search (NAS) to compress LLMs by pruning structural components, such as attention heads, neurons, and layers, aiming to achieve a Pareto-optimal balance between performance and efficiency. While NAS already achieved promising results on small language models in previous work, in this paper we propose various extensions that allow us to scale to LLMs. Compared to structural pruning baselines, we show that NAS improves performance up to 3.4% on MMLU with an on-device latency speedup.

The various limitations of Generative AI, such as hallucinations and model failures, have made it crucial to understand the role of different modalities in Visual Language Model (VLM) predictions. Our work investigates how the integration of information from image and text modalities influences the performance and behavior of VLMs in visual question answering (VQA) and reasoning tasks. We measure this effect through answer accuracy, reasoning quality, model uncertainty, and modality relevance. We study the interplay between text and image modalities in different configurations where visual content is essential for solving the VQA task. Our contributions include (1) the Semantic Interventions (SI)-VQA dataset, (2) a benchmark study of various VLM architectures under different modality configurations, and (3) the Interactive Semantic Interventions (ISI) tool. The SI-VQA dataset serves as the foundation for the benchmark, while the ISI tool provides an interface to test and apply semantic interventions in image and text inputs, enabling more fine-grained analysis. Our results show that complementary information between modalities improves answer and reasoning quality, while contradictory information harms model performance and confidence. Image text annotations have minimal impact on accuracy and uncertainty, slightly increasing image relevance. Attention analysis confirms the dominant role of image inputs over text in VQA tasks. In this study, we evaluate state-of-the-art VLMs that allow us to extract attention coefficients for each modality. A key finding is PaliGemma's harmful overconfidence, which poses a higher risk of silent failures compared to the LLaVA models. This work sets the foundation for rigorous analysis of modality integration, supported by datasets specifically designed for this purpose.

To give teenagers a new, everyday perspective on STEM topics, we use particle physics as our main topic to engage young students. Our strategy is designed to demystify particle physics, making it more accessible and attractive early in high school. In Germany, students usually decide whether or not to continue their physics education around the age of 15. That's why our project is aimed specifically at students aged 10 to 15 to give them a real insight into particle physics research before they have to make a final decision about continuing physics. Our efforts have focused on creating educational and engaging workshops for young learners. We have reached over 620 students across these age groups through more than 25 events in the last two years. The initial results are promising, indicating that our efforts are successfully igniting a motivation for physics, especially among girls. In this proceeding, we will present our workshops, the methodologies we use and a first evaluation., Comment: 6 pages, 4 figures, ICHEP 2024 parallel talk

This paper introduces the hypothetical track-length fitting algorithm, a novel method for measuring the kinetic energies of ionizing particles in liquid argon time projection chambers (LArTPCs). The algorithm finds the most probable offset in track length for a track-like object by comparing the measured ionization density as a function of position with a theoretical prediction of the energy loss as a function of the energy, including models of electron recombination and detector response. The algorithm can be used to measure the energies of particles that interact before they stop, such as charged pions that are absorbed by argon nuclei. The algorithm's energy measurement resolutions and fractional biases are presented as functions of particle kinetic energy and number of track hits using samples of stopping secondary charged pions in data collected by the ProtoDUNE-SP detector, and also in a detailed simulation. Additional studies describe impact of the dE/dx model on energy measurement performance. The method described in this paper to characterize the energy measurement performance can be repeated in any LArTPC experiment using stopping secondary charged pions.

Intra-cardiac Echocardiography (ICE) is a crucial imaging modality used in electrophysiology (EP) and structural heart disease (SHD) interventions, providing real-time, high-resolution views from within the heart. Despite its advantages, effective manipulation of the ICE catheter requires significant expertise, which can lead to inconsistent outcomes, particularly among less experienced operators. To address this challenge, we propose an AI-driven closed-loop view guidance system with human-in-the-loop feedback, designed to assist users in navigating ICE imaging without requiring specialized knowledge. Our method models the relative position and orientation vectors between arbitrary views and clinically defined ICE views in a spatial coordinate system, guiding users on how to manipulate the ICE catheter to transition from the current view to the desired view over time. Operating in a closed-loop configuration, the system continuously predicts and updates the necessary catheter manipulations, ensuring seamless integration into existing clinical workflows. The effectiveness of the proposed system is demonstrated through a simulation-based evaluation, achieving an 89% success rate with the 6532 test dataset, highlighting its potential to improve the accuracy and efficiency of ICE imaging procedures.

Explainable AI (XAI) is a rapidly growing domain with a myriad of proposed methods as well as metrics aiming to evaluate their efficacy. However, current studies are often of limited scope, examining only a handful of XAI methods and ignoring underlying design parameters for performance, such as the model architecture or the nature of input data. Moreover, they often rely on one or a few metrics and neglect thorough validation, increasing the risk of selection bias and ignoring discrepancies among metrics. These shortcomings leave practitioners confused about which method to choose for their problem. In response, we introduce LATEC, a large-scale benchmark that critically evaluates 17 prominent XAI methods using 20 distinct metrics. We systematically incorporate vital design parameters like varied architectures and diverse input modalities, resulting in 7,560 examined combinations. Through LATEC, we showcase the high risk of conflicting metrics leading to unreliable rankings and consequently propose a more robust evaluation scheme. Further, we comprehensively evaluate various XAI methods to assist practitioners in selecting appropriate methods aligning with their needs. Curiously, the emerging top-performing method, Expected Gradients, is not examined in any relevant related study. LATEC reinforces its role in future XAI research by publicly releasing all 326k saliency maps and 378k metric scores as a (meta-)evaluation dataset. The benchmark is hosted at: https://github.com/IML-DKFZ/latec., Comment: Accepted at NeurIPS 2024

Radio detections of stellar systems provide a window onto stellar magnetic activity and the space weather conditions of extrasolar planets, information that is difficult to attain at other wavelengths. There have been recent advances observing auroral emissions from radio-bright low-mass stars and exoplanets largely due to the maturation of low-frequency radio instruments and the plethora of wide-field radio surveys. To guide us in placing these recent results in context, we introduce the foremost local analogues for the field: Solar bursts and the aurorae found on Jupiter. We detail how radio bursts associated with stellar flares are foundational to the study of stellar coronae, and time-resolved radio dynamic spectra offers one of the best prospects of detecting and characterising coronal mass ejections from other stars. We highlight the prospects of directly detecting coherent radio emission from exoplanetary magnetospheres, and early tentative results. We bridge this discussion to the field of brown dwarf radio emission, in which their larger and stronger magnetospheres are amenable to detailed study with current instruments. Bright, coherent radio emission is also predicted from magnetic interactions between stars and close-in planets. We discuss the underlying physics of these interactions and implications of recent provisional detections for exoplanet characterisation. We conclude with an overview of outstanding questions in theory of stellar, star-planet interaction, and exoplanet radio emission, and the prospects of future facilities in answering them., Comment: Accepted to Nature Astronomy. The manuscript is designed to be a primer for new doctoral students and scholars to the field of radio stars and exoplanets. 36 pages, 3 figures

A number of machine learning models have been proposed with the goal of achieving systematic generalization: the ability to reason about new situations by combining aspects of previous experiences. These models leverage compositional architectures which aim to learn specialized modules dedicated to structures in a task that can be composed to solve novel problems with similar structures. While the compositionality of these architectures is guaranteed by design, the modules specializing is not. Here we theoretically study the ability of network modules to specialize to useful structures in a dataset and achieve systematic generalization. To this end we introduce a minimal space of datasets motivated by practical systematic generalization benchmarks. From this space of datasets we present a mathematical definition of systematicity and study the learning dynamics of linear neural modules when solving components of the task. Our results shed light on the difficulty of module specialization, what is required for modules to successfully specialize, and the necessity of modular architectures to achieve systematicity. Finally, we confirm that the theoretical results in our tractable setting generalize to more complex datasets and non-linear architectures., Comment: The Eleventh International Conference on Learning Representations 2023

We present constraints on the $f(R)$ gravity model using a sample of 1,005 galaxy clusters in the redshift range $0.25 - 1.78$ that have been selected through the thermal Sunyaev-Zel'dovich effect (tSZE) from South Pole Telescope (SPT) data and subjected to optical and near-infrared confirmation with the Multi-component Matched Filter (MCMF) algorithm. We employ weak gravitational lensing mass calibration from the Dark Energy Survey (DES) Year 3 data for 688 clusters at $z < 0.95$ and from the Hubble Space Telescope (HST) for 39 clusters with $0.6 < z < 1.7$. Our cluster sample is a powerful probe of $f(R)$ gravity, because this model predicts a scale-dependent enhancement in the growth of structure, which impacts the halo mass function (HMF) at cluster mass scales. To account for these modified gravity effects on the HMF, our analysis employs a semi-analytical approach calibrated with numerical simulations. Combining calibrated cluster counts with primary cosmic microwave background (CMB) temperature and polarization anisotropy measurements from the Planck2018 release, we derive robust constraints on the $f(R)$ parameter $f_{R0}$. Our results, $\log_{10} |f_{R0}| < -5.32$ at the 95 % credible level, are the tightest current constraints on $f(R)$ gravity from cosmological scales. This upper limit rules out $f(R)$-like deviations from general relativity that result in more than a $\sim$20 % enhancement of the cluster population on mass scales $M_\mathrm{200c}>3\times10^{14}M_\odot$., Comment: 21 pages, 6 figures, submitted to Phys. Rev. D

We present recent laboratory results demonstrating high-contrast coronagraphy for the future space-based large IR/Optical/Ultraviolet telescope recommended by the Decadal Survey. The High-contrast Imager for Complex Aperture Telescopes (HiCAT) testbed aims to implement a system-level hardware demonstration for segmented aperture coronagraphs with wavefront control. The telescope hardware simulator employs a segmented deformable mirror with 37 hexagonal segments that can be controlled in piston, tip, and tilt. In addition, two continuous deformable mirrors are used for high-order wavefront sensing and control. The low-order sensing subsystem includes a dedicated tip-tilt stage, a coronagraphic target acquisition camera, and a Zernike wavefront sensor that is used to measure and correct low-order aberration drifts. We explore the performance of a segmented aperture coronagraph both in static operations (limited by natural drifts and instabilities) and in dynamic operations (in the presence of artificial wavefront drifts added to the deformable mirrors), and discuss the estimation and control strategies used to reach and maintain the dark-zone contrast using our low-order wavefront sensing and control. We summarize experimental results that quantify the performance of the testbed in terms of contrast, inner/outer working angle and bandpass, and analyze limiting factors., Comment: 17 pages, 14 figures, SPIE Astronomical Telescopes + Instrumentation, 2022, Montr\'eal, Qu\'ebec, Canada

Massive black-hole binaries will be the loudest sources detectable by LISA. These systems are predicted to form during the hierarchical assembly of cosmic structures and coalesce by interacting with the surrounding environment. The hardening phase of their orbit is driven by either stars or gas and encodes distinctive features into the binary black holes that can potentially be reconstructed with gravitational-wave observations. We present a Bayesian framework to assess the likelihood of massive mergers being hardened by either gaseous or stellar interactions. We use state-of-the-art astrophysical models tracking the cosmological evolution of massive black-hole binaries and construct a large number of simulated catalogs of sources detectable by LISA. From these, we select a representative catalog and run both parameter estimation assuming a realistic LISA response as well model comparison capturing selection effects. Our results suggest that, at least within the context of the adopted models, future LISA observations can confidently constrain whether stars or gas are responsible for the binary hardening. We stress that accurate astrophysical modeling of the black-hole spins and the inclusion of subdominant emission modes in the adopted signal might be crucial to avoid systematic biases., Comment: 10 pages, 6 figures (submitted to Physical Review D)

In this work, we significantly enhance masked particle modeling (MPM), a self-supervised learning scheme for constructing highly expressive representations of unordered sets relevant to developing foundation models for high-energy physics. In MPM, a model is trained to recover the missing elements of a set, a learning objective that requires no labels and can be applied directly to experimental data. We achieve significant performance improvements over previous work on MPM by addressing inefficiencies in the implementation and incorporating a more powerful decoder. We compare several pre-training tasks and introduce new reconstruction methods that utilize conditional generative models without data tokenization or discretization. We show that these new methods outperform the tokenized learning objective from the original MPM on a new test bed for foundation models for jets, which includes using a wide variety of downstream tasks relevant to jet physics, such as classification, secondary vertex finding, and track identification.

Calibration ensures that predicted uncertainties align with observed uncertainties. While there is an extensive literature on recalibration methods for univariate probabilistic forecasts, work on calibration for multivariate forecasts is much more limited. This paper introduces a novel post-hoc recalibration approach that addresses multivariate calibration for potentially misspecified models. Our method involves constructing local mappings between vectors of marginal probability integral transform values and the space of observations, providing a flexible and model free solution applicable to continuous, discrete, and mixed responses. We present two versions of our approach: one uses K-nearest neighbors, and the other uses normalizing flows. Each method has its own strengths in different situations. We demonstrate the effectiveness of our approach on two real data applications: recalibrating a deep neural network's currency exchange rate forecast and improving a regression model for childhood malnutrition in India for which the multivariate response has both discrete and continuous components.

Multi-agent collaboration involves multiple participants working together in a shared environment to achieve a common goal. These agents share information, divide tasks, and synchronize their actions. Key aspects of multi agent collaboration include coordination, communication, task allocation, cooperation, adaptation, and decentralization. On construction sites, surface grading is the process of leveling sand piles to increase a specific area's height. In this scenario, a bulldozer grades while a dumper allocates sand piles. Our work aims to utilize a multi-agent approach to enable these vehicles to collaborate effectively. To this end, we propose a decentralized and asymmetric multi-agent learning approach for construction sites (DAMALCS). We formulate DAMALCS to reduce expected collisions for operating vehicles. Therefore, we develop two heuristic experts capable of achieving their joint goal optimally by applying an innovative prioritization method. In this approach, the bulldozer's movements take precedence over the dumper's operations, enabling the bulldozer to clear the path for the dumper and ensure continuous operation of both vehicles. Since heuristics alone are insufficient in real-world scenarios, we utilize them to train AI agents, which proves to be highly effective. We simultaneously train the bulldozer and dumper agents to operate within the same environment, aiming to avoid collisions and optimize performance in terms of time efficiency and sand volume handling. Our trained agents and heuristics are evaluated in both simulation and real-world lab experiments, testing them under various conditions, such as visual noise and localization errors. The results demonstrate that our approach significantly reduces collision rates for these vehicles.