Search

We measure the dynamical behavior of colloidal singlets and dumbbells on an inclined magnetic moir\'e pattern, subject to a precessing external homogeneous magnetic field. At low external field strength single colloidal particles and dumbbells move everywhere on the pattern: At stronger external field strengths colloidal singlets and dumbbells are localized in generic locations. There are however nongeneric locations of flat channels that cross the moir\'e Wigner Seitz cell. In the flat channels we find gravitational driven translational and non-translational dynamic phase behavior of the colloidal singlets and dumbbells depending on the external field strength and the precession angle of the external homogeneous magnetic field.

Vision-language models (VLMs), such as CLIP and SigLIP, have found remarkable success in classification, retrieval, and generative tasks. For this, VLMs deterministically map images and text descriptions to a joint latent space in which their similarity is assessed using the cosine similarity. However, a deterministic mapping of inputs fails to capture uncertainties over concepts arising from domain shifts when used in downstream tasks. In this work, we propose post-hoc uncertainty estimation in VLMs that does not require additional training. Our method leverages a Bayesian posterior approximation over the last layers in VLMs and analytically quantifies uncertainties over cosine similarities. We demonstrate its effectiveness for uncertainty quantification and support set selection in active learning. Compared to baselines, we obtain improved and well-calibrated predictive uncertainties, interpretable uncertainty estimates, and sample-efficient active learning. Our results show promise for safety-critical applications of large-scale models., Comment: Project page: https://aaltoml.github.io/BayesVLM/

Reconstructing complex structures from planar cross-sections is a challenging problem, with wide-reaching applications in medical imaging, manufacturing, and topography. Out-of-the-box point cloud reconstruction methods can often fail due to the data sparsity between slicing planes, while current bespoke methods struggle to reconstruct thin geometric structures and preserve topological continuity. This is important for medical applications where thin vessel structures are present in CT and MRI scans. This paper introduces CrossSDF, a novel approach for extracting a 3D signed distance field from 2D signed distances generated from planar contours. Our approach makes the training of neural SDFs contour-aware by using losses designed for the case where geometry is known within 2D slices. Our results demonstrate a significant improvement over existing methods, effectively reconstructing thin structures and producing accurate 3D models without the interpolation artifacts or over-smoothing of prior approaches.

Matrix valued orthogonal polynomials (MVOP) appear in the study of doubly periodic tiling models. Of particular interest is their limiting behavior as the degree tends to infinity. In recent years, MVOP associated with doubly periodic domino tilings of the Aztec diamond have been successfully analyzed. The MVOP related to doubly periodic lozenge tilings of a hexagon are more complicated. In this paper we focus on a special subclass of hexagon tilings with 3x3 periodicity. The special subclass leads to a genus one spectral curve with additional symmetries that allow us to find an equilibrium measure in an external field explicitly. The equilibrium measure gives the asymptotic distribution for the zeros of the determinant of the MVOP. The associated g-functions appear in the strong asymptotic formula for the MVOP that we obtain from a steepest descent analysis of the Riemann-Hilbert problem for MVOP., Comment: 60 pages, 14 figures

We introduce the $\Sigma^{\text{BSE}}@L^{\text{BSE}}$ self-energy in the quasi-particle self-consistent $GW$ (qs$GW$) framework (qs$\Sigma^{\text{BSE}}@L^{\text{BSE}}$). Here, $L$ is the two-particle response function which we calculate by solving the Bethe-Salpeter equation with the static, first-order $GW$ kernel. The same kernel is added to $\Sigma$ directly. For a set of medium organic molecules, we show that including the vertex both in $L$ and $\Sigma$ is crucial. This approach retains the good performance of qs$GW$ for predicting first ionization potentials and fundamental gaps, while it greatly improves the description of electron affinities. Its good performance places qs$\Sigma^{\text{BSE}}@L^{\text{BSE}}$ among the best-performing electron propagator methods for charged excitations. Adding the vertex in $L$ only, as commonly done in the solid state community, leads to devastating results for electron affinities and fundamental gaps. We also test the performance of BSE@qs$GW$ and qs$\Sigma^{\text{BSE}}@L^{\text{BSE}}$ for neutral charge-transfer excitation and find both methods to perform similar. We conclude that $\Sigma^{\text{BSE}}@L^{\text{BSE}}$ is a promising approximation to the electronic self-energy beyond $GW$. We hope that future research on dynamical vertex effects, second-order vertex corrections, and full self-consistency will improve the accuracy of this method, both for charged and neutral excitation energies., Comment: 5 figures

Gaussian splatting enables fast novel view synthesis in static 3D environments. However, reconstructing real-world environments remains challenging as distractors or occluders break the multi-view consistency assumption required for accurate 3D reconstruction. Most existing methods rely on external semantic information from pre-trained models, introducing additional computational overhead as pre-processing steps or during optimization. In this work, we propose a novel method, DeSplat, that directly separates distractors and static scene elements purely based on volume rendering of Gaussian primitives. We initialize Gaussians within each camera view for reconstructing the view-specific distractors to separately model the static 3D scene and distractors in the alpha compositing stages. DeSplat yields an explicit scene separation of static elements and distractors, achieving comparable results to prior distractor-free approaches without sacrificing rendering speed. We demonstrate DeSplat's effectiveness on three benchmark data sets for distractor-free novel view synthesis. See the project website at https://aaltoml.github.io/desplat/.

The rising interest in Bayesian deep learning (BDL) has led to a plethora of methods for estimating the posterior distribution. However, efficient computation of inferences, such as predictions, has been largely overlooked with Monte Carlo integration remaining the standard. In this work we examine streamlining prediction in BDL through a single forward pass without sampling. For this we use local linearisation on activation functions and local Gaussian approximations at linear layers. Thus allowing us to analytically compute an approximation to the posterior predictive distribution. We showcase our approach for both MLP and transformers, such as ViT and GPT-2, and assess its performance on regression and classification tasks.

In this paper, the first in a series, we present a new theoretical model for the global structure and dissipation of relativistically magnetized collisionless shock waves. Quite remarkably, we find that in contrast to unmagnetized shocks, energy dissipation does not involve collective plasma interactions. Rather, it is a consequence of nonlinear particle dynamics. We demonstrate that the kinetic-scale shock transition can be modeled as a stationary system consisting of a large set of cold beams coupled through the magnetic field. The fundamental mechanism governing shock dissipation relies on the onset of chaos in orbital dynamics within quasiperiodic solitonic structures. We discuss the impact of upstream temperature and magnetization on the shock profile, recovering the magnetic field compression, downstream velocities, and heating expected from the Rankine-Hugoniot jump conditions. We deduce a rate of entropy generation from the spectrum of Lyapunov exponents and discuss the thermalization of the beam distribution. Our model provides a general framework to study magnetized collisionless shock structures., Comment: 11 pages, 8 figures

I present Astro-HEP-BERT, a transformer-based language model specifically designed for generating contextualized word embeddings (CWEs) to study the meanings of concepts in astrophysics and high-energy physics. Built on a general pretrained BERT model, Astro-HEP-BERT underwent further training over three epochs using the Astro-HEP Corpus, a dataset I curated from 21.84 million paragraphs extracted from more than 600,000 scholarly articles on arXiv, all belonging to at least one of these two scientific domains. The project demonstrates both the effectiveness and feasibility of adapting a bidirectional transformer for applications in the history, philosophy, and sociology of science (HPSS). The entire training process was conducted using freely available code, pretrained weights, and text inputs, completed on a single MacBook Pro Laptop (M2/96GB). Preliminary evaluations indicate that Astro-HEP-BERT's CWEs perform comparably to domain-adapted BERT models trained from scratch on larger datasets for domain-specific word sense disambiguation and induction and related semantic change analyses. This suggests that retraining general language models for specific scientific domains can be a cost-effective and efficient strategy for HPSS researchers, enabling high performance without the need for extensive training from scratch., Comment: 7 pages, 4 figures, 1 table

This paper explores the potential of contextualized word embeddings (CWEs) as a new tool in the history, philosophy, and sociology of science (HPSS) for studying contextual and evolving meanings of scientific concepts. Using the term "Planck" as a test case, I evaluate five BERT-based models with varying degrees of domain-specific pretraining, including my custom model Astro-HEP-BERT, trained on the Astro-HEP Corpus, a dataset containing 21.84 million paragraphs from 600,000 articles in astrophysics and high-energy physics. For this analysis, I compiled two labeled datasets: (1) the Astro-HEP-Planck Corpus, consisting of 2,900 labeled occurrences of "Planck" sampled from 1,500 paragraphs in the Astro-HEP Corpus, and (2) a physics-related Wikipedia dataset comprising 1,186 labeled occurrences of "Planck" across 885 paragraphs. Results demonstrate that the domain-adapted models outperform the general-purpose ones in disambiguating the target term, predicting its known meanings, and generating high-quality sense clusters, as measured by a novel purity indicator I developed. Additionally, this approach reveals semantic shifts in the target term over three decades in the unlabeled Astro-HEP Corpus, highlighting the emergence of the Planck space mission as a dominant sense. The study underscores the importance of domain-specific pretraining for analyzing scientific language and demonstrates the cost-effectiveness of adapting pretrained models for HPSS research. By offering a scalable and transferable method for modeling the meanings of scientific concepts, CWEs open up new avenues for investigating the socio-historical dynamics of scientific discourses., Comment: 18 pages, 7 figures (1 in the Supplement)

A Dirac quantum spin liquid hosts Dirac spinons, which are low-energy fractionalized neutral quasiparticles with spin 1/2 that obey the Dirac equation. Recent studies have revealed cone spin continuum in YCu$_3$(OD)$_6$Br$_2$[Br$_{x}$(OD)$_{1-x}$], consistent with the convolution of two Dirac spinons. In this work, we further studied spin excitations using the inelastic neutron scattering technique. The width of low-energy spin excitations shows a linear temperature dependence, which can be explained by spinon-spinon interactions with a Dirac dispersion. Polarized neutron scattering measurements reveal that in-plane magnetic fluctuations are about 1.5 times stronger than the out-of-plane ones, suggesting the presence of out-of-plane Dzyaloshinskii-Moriya interaction. Moreover, the high-energy spin excitations around 14 meV agree with the one-pair spinon-antispinon excitations in Raman studies. The real part of the dynamical susceptibility derived from the Kramers-Kronig relationship also accords with the Knight shift measured by nuclear magnetic resonance. These results provide further insights for the possible Dirac quantum spin liquid in this system., Comment: 7 pages, 4 figures

Supporting the health and well-being of dynamic populations around the world requires governmental agencies, organizations and researchers to understand and reason over complex relationships between human behavior and local contexts in order to identify high-risk groups and strategically allocate limited resources. Traditional approaches to these classes of problems often entail developing manually curated, task-specific features and models to represent human behavior and the natural and built environment, which can be challenging to adapt to new, or even, related tasks. To address this, we introduce a Population Dynamics Foundation Model (PDFM) that aims to capture the relationships between diverse data modalities and is applicable to a broad range of geospatial tasks. We first construct a geo-indexed dataset for postal codes and counties across the United States, capturing rich aggregated information on human behavior from maps, busyness, and aggregated search trends, and environmental factors such as weather and air quality. We then model this data and the complex relationships between locations using a graph neural network, producing embeddings that can be adapted to a wide range of downstream tasks using relatively simple models. We evaluate the effectiveness of our approach by benchmarking it on 27 downstream tasks spanning three distinct domains: health indicators, socioeconomic factors, and environmental measurements. The approach achieves state-of-the-art performance on all 27 geospatial interpolation tasks, and on 25 out of the 27 extrapolation and super-resolution tasks. We combined the PDFM with a state-of-the-art forecasting foundation model, TimesFM, to predict unemployment and poverty, achieving performance that surpasses fully supervised forecasting. The full set of embeddings and sample code are publicly available for researchers., Comment: 28 pages, 16 figures, preprint; v2: updated github url

This work explores the intersection of continual learning (CL) and differential privacy (DP). Crucially, continual learning models must retain knowledge across tasks, but this conflicts with the differential privacy requirement of restricting individual samples to be memorised in the model. We propose using pre-trained models to address the trade-offs between privacy and performance in a continual learning setting. More specifically, we present necessary assumptions to enable privacy-preservation and propose combining pre-trained models with parameter-free classifiers and parameter-efficient adapters that are learned under differential privacy. Our experiments demonstrate their effectiveness and provide insights into balancing the competing demands of continual learning and privacy., Comment: 15 pages, 3 figures, Accepted at Scalable Continual Learning for Lifelong Foundation Models Workshop at 38th Conference on Neural Information Processing Systems (NeurIPS 2024)

Open-weight large language model (LLM) zoos allow users to quickly integrate state-of-the-art models into systems. Despite increasing availability, selecting the most appropriate model for a given task still largely relies on public benchmark leaderboards and educated guesses. This can be unsatisfactory for both inference service providers and end users, where the providers usually prioritize cost efficiency, while the end users usually prioritize model output quality for their inference requests. In commercial settings, these two priorities are often brought together in Service Level Agreements (SLA). We present MESS+, an online stochastic optimization algorithm for energy-optimal model selection from a model zoo, which works on a per-inference-request basis. For a given SLA that requires high accuracy, we are up to 2.5x more energy efficient with MESS+ than with randomly selecting an LLM from the zoo while maintaining SLA quality constraints., Comment: Accepted at the 2024 Workshop on Adaptive Foundation Models in conjunction with NeurIPS 2024

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.

We introduce Planning-guided Retrieval Augmented Generation (Plan$\times$RAG), a novel framework that augments the \emph{retrieve-then-reason} paradigm of existing RAG frameworks to \emph{plan-then-retrieve}. Plan$\times$RAG formulates a reasoning plan as a directed acyclic graph (DAG), decomposing queries into interrelated atomic sub-queries. Answer generation follows the DAG structure, allowing significant gains in efficiency through parallelized retrieval and generation. While state-of-the-art RAG solutions require extensive data generation and fine-tuning of language models (LMs), Plan$\times$RAG incorporates frozen LMs as plug-and-play experts to generate high-quality answers. Compared to existing RAG solutions, Plan$\times$RAG demonstrates significant improvements in reducing hallucinations and bolstering attribution due to its structured sub-query decomposition. Overall, Plan$\times$RAG offers a new perspective on integrating external knowledge in LMs while ensuring attribution by design, contributing towards more reliable LM-based systems., Comment: 22 pages, preprint

We model and investigate the collective nonlinear optical response of an ensemble of two-level emitters that are weakly coupled to a single-mode waveguide. Our approach generalizes the insight that photon-photon correlations in the light scattered by a single two-level emitter result from two-photon interference to the case of many emitters. Using our model, we study different configurations for probing the nonlinear response of the ensemble, e.g., through the waveguide or via external illumination, and derive analytical expressions for the second-order quantum coherence function, $g^{(2)}(\tau)$, as well as for the squeezing spectrum of the output light in the waveguide, $S_\theta(\omega)$. For the transmission of resonant guided light, we recover the same predictions as previously made with far more involved theoretical models when analyzing experimental results regarding $g^{(2)}(\tau)$ (Prasad et al. [1]) and $S_\theta(\omega)$ (Hinney et al. [2]). We also study the transmission of light that is detuned from the transition of the two-level emitter, a situation that we recently studied experimentally (Cordier et al. [3]). Our model predictions show how the collectively enhanced nonlinear response of weakly coupled emitters can be harnessed to generate non-classical states of light using ensembles ranging from a few to many emitters.

We consider robustness to distribution shifts in the context of diagnostic models in healthcare, where the prediction target $Y$, e.g., the presence of a disease, is causally upstream of the observations $X$, e.g., a biomarker. Distribution shifts may occur, for instance, when the training data is collected in a domain with patients having particular demographic characteristics while the model is deployed on patients from a different demographic group. In the domain of applied ML for health, it is common to predict $Y$ from $X$ without considering further information about the patient. However, beyond the direct influence of the disease $Y$ on biomarker $X$, a predictive model may learn to exploit confounding dependencies (or shortcuts) between $X$ and $Y$ that are unstable under certain distribution shifts. In this work, we highlight a data generating mechanism common to healthcare settings and discuss how recent theoretical results from the causality literature can be applied to build robust predictive models. We theoretically show why ignoring covariates as well as common invariant learning approaches will in general not yield robust predictors in the studied setting, while including certain covariates into the prediction model will. In an extensive simulation study, we showcase the robustness (or lack thereof) of different predictors under various data generating processes. Lastly, we analyze the performance of the different approaches using the PTB-XL dataset, a public dataset of annotated ECG recordings.

Hedin's $GW$ approximation to the electronic self-energy has been impressively successful to calculate quasiparticle energies, such as ionization potentials, electron affinities, or electronic band structures. The success of this fairly simple approximation has been ascribed to the cancellation of the so-called vertex corrections that go beyond $GW$. This claim is mostly based on past calculations using vertex corrections within the crude local-density approximation. Here, we explore a wide variety of non-local vertex corrections in the polarizability and the self-energy, using first-order approximations or infinite summations to all orders. In particular, we use vertices based on statically screened interactions like in the Bethe-Salpeter equation. We demonstrate on realistic molecular systems that the two vertices in Hedin's equation essentially compensate. We further show that consistency between the two vertices is crucial to obtain realistic electronic properties. We finally consider increasingly large clusters and extrapolate that our conclusions would hold for extended systems., Comment: Update of original version: More detailed results including an additional figure, additional references, and another figure with diagrams

This whitepaper offers an overview of the ethical considerations surrounding research into or with large language models (LLMs). As LLMs become more integrated into widely used applications, their societal impact increases, bringing important ethical questions to the forefront. With a growing body of work examining the ethical development, deployment, and use of LLMs, this whitepaper provides a comprehensive and practical guide to best practices, designed to help those in research and in industry to uphold the highest ethical standards in their work., Comment: 47 pages

The covariance for clean data given a noisy observation is an important quantity in many conditional generation methods for diffusion models. Current methods require heavy test-time computation, altering the standard diffusion training process or denoiser architecture, or making heavy approximations. We propose a new framework that sidesteps these issues by using covariance information that is available for free from training data and the curvature of the generative trajectory, which is linked to the covariance through the second-order Tweedie's formula. We integrate these sources of information using {\em (i)} a novel method to transfer covariance estimates across noise levels and (ii) low-rank updates in a given noise level. We validate the method on linear inverse problems, where it outperforms recent baselines, especially with fewer diffusion steps., Comment: 24 pages, 11 figures

The origins of the magnetic fields that power gamma-ray burst (GRB) afterglow emission are not fully understood. One possible channel for generating these fields involves the pre-conditioning of the circumburst medium: in the early afterglow phase, prompt photons streaming ahead of the GRB external shock can pair produce, seeding the upstream with drifting electron-positron pairs and triggering electromagnetic microinstabilities. To study this process, we employ 2D periodic particle-in-cell simulations in which a cold electron-proton plasma is gradually enriched with warm electron-positron pairs injected at mildly relativistic speeds. We find that continuous pair injection drives the growth of large-scale magnetic fields via filamentation-like instabilities; the temporal evolution of the field is self-similar and depends on a single parameter, $\left[\alpha/(t_f \omega_{pi})\right]^{1/2} t\omega_{pi}$, where $\alpha$ is the ratio of final pair beam density to background plasma density, $t_f$ is the duration of pair injection, and $\omega_{pi}$ is the plasma frequency of background protons. Extrapolating our results to parameter regimes realistic for long GRBs, we find that upstream pair enrichment generates weak magnetic fields on scales much larger than the proton skin depth; for bright bursts, the extrapolated coherence scale at a shock radius of $R \sim 10^{17}$ cm is $\left<\lambda_y\right> \sim 100 c/\omega_{pi}$ and the corresponding magnetization is $\sigma \sim 10^{-8}$ for typical circumburst parameters. These results may help explain the persistence of magnetic fields at large distances behind GRB shocks., Comment: 21 pages, 21 figures. Submitted to ApJ

We present two multilingual LLMs designed to embrace Europe's linguistic diversity by supporting all 24 official languages of the European Union. Trained on a dataset comprising around 60% non-English data and utilizing a custom multilingual tokenizer, our models address the limitations of existing LLMs that predominantly focus on English or a few high-resource languages. We detail the models' development principles, i.e., data composition, tokenizer optimization, and training methodologies. The models demonstrate competitive performance across multilingual benchmarks, as evidenced by their performance on European versions of ARC, HellaSwag, MMLU, and TruthfulQA.

One of the central goals of causal machine learning is the accurate estimation of heterogeneous treatment effects from observational data. In recent years, meta-learning has emerged as a flexible, model-agnostic paradigm for estimating conditional average treatment effects (CATE) using any supervised model. This paper examines the performance of meta-learners when the confounding variables are expressed in text. Through synthetic data experiments, we show that learners using pre-trained text representations of confounders, in addition to tabular background variables, achieve improved CATE estimates compared to those relying solely on the tabular variables, particularly when sufficient data is available. However, due to the entangled nature of the text embeddings, these models do not fully match the performance of meta-learners with perfect confounder knowledge. These findings highlight both the potential and the limitations of pre-trained text representations for causal inference and open up interesting avenues for future research., Comment: Presented at the NeurIPS 2024 Workshop on Causal Representation Learning

Differential equations are important mechanistic models that are integral to many scientific and engineering applications. With the abundance of available data there has been a growing interest in data-driven physics-informed models. Gaussian processes (GPs) are particularly suited to this task as they can model complex, non-linear phenomena whilst incorporating prior knowledge and quantifying uncertainty. Current approaches have found some success but are limited as they either achieve poor computational scalings or focus only on the temporal setting. This work addresses these issues by introducing a variational spatio-temporal state-space GP that handles linear and non-linear physical constraints while achieving efficient linear-in-time computation costs. We demonstrate our methods in a range of synthetic and real-world settings and outperform the current state-of-the-art in both predictive and computational performance.

A powerful approach to computing Feynman integrals or cosmological correlators is to consider them as solution to systems of differential equations. Often these can be chosen to be Gelfand-Kapranov-Zelevinsky (GKZ) systems. However, their naive construction introduces a significant amount of unnecessary complexity. In this paper we present an algorithm which allows for reducing these GKZ systems to smaller subsystems if a parameter associated to the GKZ systems is resonant. These simpler subsystems can then be solved separately resulting in solutions for the full system. The algorithm makes it possible to check when reductions happen and allows for finding the associated simpler solutions. While originating in the mathematical theory of D-modules analyzed via exact sequences of Euler-Koszul homologies, the algorithm can be used without knowledge of this framework. We motivate the need for such reduction techniques by considering cosmological correlators on an FRW space-time and solve the tree-level single-exchange correlator in this way. It turns out that this integral exemplifies an interesting relation between locality and the reduction of the differential equations., Comment: 45 pages plus appendices (16 pages), 4 figures, v2: Fixed typos, including an important one in the abstract, added references

Graph neural networks (GNNs) are a type of neural network capable of learning on graph-structured data. However, training GNNs on large-scale graphs is challenging due to iterative aggregations of high-dimensional features from neighboring vertices within sparse graph structures combined with neural network operations. The sparsity of graphs frequently results in suboptimal memory access patterns and longer training time. Graph reordering is an optimization strategy aiming to improve the graph data layout. It has shown to be effective to speed up graph analytics workloads, but its effect on the performance of GNN training has not been investigated yet. The generalization of reordering to GNN performance is nontrivial, as multiple aspects must be considered: GNN hyper-parameters such as the number of layers, the number of hidden dimensions, and the feature size used in the GNN model, neural network operations, large intermediate vertex states, and GPU acceleration. In our work, we close this gap by performing an empirical evaluation of 12 reordering strategies in two state-of-the-art GNN systems, PyTorch Geometric and Deep Graph Library. Our results show that graph reordering is effective in reducing training time for CPU- and GPU-based training, respectively. Further, we find that GNN hyper-parameters influence the effectiveness of reordering, that reordering metrics play an important role in selecting a reordering strategy, that lightweight reordering performs better for GPU-based than for CPU-based training, and that invested reordering time can in many cases be amortized., Comment: To be published in proceedings of the 51st International Conference on Very Large Data Bases (VLDB), September 1-5, 2025

We present the SynSUM benchmark, a synthetic dataset linking unstructured clinical notes to structured background variables. The dataset consists of 10,000 artificial patient records containing tabular variables (like symptoms, diagnoses and underlying conditions) and related notes describing the fictional patient encounter in the domain of respiratory diseases. The tabular portion of the data is generated through a Bayesian network, where both the causal structure between the variables and the conditional probabilities are proposed by an expert based on domain knowledge. We then prompt a large language model (GPT-4o) to generate a clinical note related to this patient encounter, describing the patient symptoms and additional context. The SynSUM dataset is primarily designed to facilitate research on clinical information extraction in the presence of tabular background variables, which can be linked through domain knowledge to concepts of interest to be extracted from the text - the symptoms, in the case of SynSUM. Secondary uses include research on the automation of clinical reasoning over both tabular data and text, causal effect estimation in the presence of tabular and/or textual confounders, and multi-modal synthetic data generation. The dataset can be downloaded from https://github.com/prabaey/SynSUM.

The process of 3D scene reconstruction can be affected by numerous uncertainty sources in real-world scenes. While Neural Radiance Fields (NeRFs) and 3D Gaussian Splatting (GS) achieve high-fidelity rendering, they lack built-in mechanisms to directly address or quantify uncertainties arising from the presence of noise, occlusions, confounding outliers, and imprecise camera pose inputs. In this paper, we introduce a taxonomy that categorizes different sources of uncertainty inherent in these methods. Moreover, we extend NeRF- and GS-based methods with uncertainty estimation techniques, including learning uncertainty outputs and ensembles, and perform an empirical study to assess their ability to capture the sensitivity of the reconstruction. Our study highlights the need for addressing various uncertainty aspects when designing NeRF/GS-based methods for uncertainty-aware 3D reconstruction., Comment: To appear in ECCV 2024 Workshop Proceedings. Project page at https://aaltoml.github.io/uncertainty-nerf-gs/

We present the detection and analysis of GRB 231115A, a candidate extragalactic magnetar giant flare (MGF) observed by Fermi/GBM and localized by INTEGRAL to the starburst galaxy M82. This burst exhibits distinctive temporal and spectral characteristics that align with known MGFs, including a short duration and a high peak energy. Gamma-ray analyses reveal significant insights into this burst, supporting conclusions already established in the literature: our time-resolved spectral studies provide further evidence that GRB 231115A is indeed a MGF. Significance calculations also suggest a robust association with M82, further supported by a high Bayes factor that minimizes the probability of chance alignment with a neutron star merger. Despite extensive follow-up efforts, no contemporaneous gravitational wave or radio emissions were detected. The lack of radio emission sets stringent upper limits on possible radio luminosity. Constraints from our analysis show no fast radio bursts (FRBs) associated with two MGFs. X-ray observations conducted post-burst by Swift/XRT and XMM/Newton provided additional data, though no persistent counterparts were identified. Our study underscores the importance of coordinated multi-wavelength follow-up and highlights the potential of MGFs to enhance our understanding of short GRBs and magnetar activities in the cosmos. Current MGF identification and follow-up implementation are insufficient for detecting expected counterparts; however, improvements in these areas may allow for the recovery of follow-up signals with existing instruments. Future advancements in observational technologies and methodologies will be crucial in furthering these studies.

We present a lightweight magnetic field simultaneous localisation and mapping (SLAM) approach for drift correction in odometry paths, where the interest is purely in the odometry and not in map building. We represent the past magnetic field readings as a one-dimensional trajectory against which the current magnetic field observations are matched. This approach boils down to sequential loop-closure detection and decision-making, based on the current pose state estimate and the magnetic field. We combine this setup with a path estimation framework using an extended Kalman smoother which fuses the odometry increments with the detected loop-closure timings. We demonstrate the practical applicability of the model with several different real-world examples from a handheld iPad moving in indoor scenes., Comment: To appear in International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI) 2024

Soft X-ray irradiation of molecules causes electronic core-level vacancies through photoelectronemission. In light elements, such as C, N, or O, which are abundant in the biosphere, these vacancies predominantly decay by Auger emission, leading inevitably to dissociative multiply charged states. It was recently demonstrated that an environment can prevent fragmentation of core-level-ionized small organic molecules through immediate non-local decay of the core hole, dissipating charge and energy to the environment. Here, we present an extended photoelectron-photoion-photoion coincidence (PEPIPICO) study of the biorelevant pyrimidine molecule embedded in a water cluster. It is observed and supported by theoretical calculations that the supposed protective effect of the environment is partially reversed if the vacancy is originally located at a water molecule. In this scenario, intermolecular energy or charge transfer from the core-ionized water environment to the pyrimidine molecule leads to ionization of the latter, however, presumably in non-dissociative cationic states. Our results contribute to a more comprehensive understanding of the complex interplay of protective and harmful effects of an environment in the photochemistry of microsolvated molecules exposed to X-rays.

We consider orthogonal polynomials with respect to the weight $|z^2+a^2|^{cN}e^{-N|z|^2}$ in the whole complex plane. We obtain strong asymptotics and the limiting normalized zero counting measure (mother body) of the orthogonal polynomials of degree $n$ in the scaling limit $n,N\to \infty$ such that $\frac{n}{N}\to t$. We restrict ourselves to the case $a^2\geq 2c$, $cN$ integer, and $t

Low-energy-electron emission from resonant Auger final states via intermolecular Coulombic decay (RA-ICD) has been previously described as a promising scenario for controlling radiation damage for medical purposes, but has so far only been observed in prototypical atomic and molecular van der Waals dimers and clusters. Here, we report the experimental observation of RA-ICD in aqueous solution. We show that for solvated Ca$^{2+}$ ions, the emission can be very efficiently controlled by tuning the photon energy of exciting X-rays to inner-shell resonances of the ions. Our results provide the next step from proving RA-ICD in relatively simple prototype systems to understanding the relevance and potential applications of ICD in real-life scenarios.

Logs are a first-hand source of information for software maintenance and failure diagnosis. Log parsing, which converts semi-structured log messages into structured templates, is a prerequisite for automated log analysis tasks such as anomaly detection, troubleshooting, and root cause analysis. However, existing log parsers fail in real-world systems for three main reasons. First, traditional heuristics-based parsers require handcrafted features and domain knowledge, which are difficult to generalize at scale. Second, existing large language model-based parsers rely on periodic offline processing, limiting their effectiveness in real-time use cases. Third, existing online parsing algorithms are susceptible to log drift, where slight log changes create false positives that drown out real anomalies. To address these challenges, we propose HELP, a Hierarchical Embeddings-based Log Parser. HELP is the first online semantic-based parser to leverage LLMs for performant and cost-effective log parsing. We achieve this through a novel hierarchical embeddings module, which fine-tunes a text embedding model to cluster logs before parsing, reducing querying costs by multiple orders of magnitude. To combat log drift, we also develop an iterative rebalancing module, which periodically updates existing log groupings. We evaluate HELP extensively on 14 public large-scale datasets, showing that HELP achieves significantly higher F1-weighted grouping and parsing accuracy than current state-of-the-art online log parsers. We also implement HELP into Iudex's production observability platform, confirming HELP's practicality in a production environment. Our results show that HELP is effective and efficient for high-throughput real-world log parsing.

Recent developments in integrated optics have made it possible to fabricate high-bandwidth electro-optic modulators (EOMs). Here we show cascaded operation of two of such modulators driven by custom-built electronics delivering on-demand picosecond pulses and application to a quantum dot cavity-QED single photon source. We implement an EOM-based correlation technique and demonstrate light pulses as short as 24$\pm$2 ps with a single EOM. The EOMs can be synchronized and operated in series, we then produce optical pulses down to 17 ps. To optimize the pulse contrast, we analyze for two different EOM devices the transmission as a function of EOM bias, wavelength and temperature, and we show that by temperature tuning and stabilization, a pulse contrast above 25 dB can be obtained at the targeted wavelength. With this high contrast we demonstrate resonant excitation of an InGaAs quantum dot - microcavity based single photon source, demonstrating a crucial technology for scalability and synchronization of large scale photonic quantum applications., Comment: 8 pages, 9 figures

In recent years, interest in quantum computing has increased due to technological advances in quantum hardware and algorithms. Despite the promises of quantum advantage, the applicability of quantum devices has been limited to few qubits on hardware that experiences decoherence due to noise. One proposed method to get around this challenge is distributed quantum computing (DQC). Like classical distributed computing, DQC aims at increasing compute power by spreading the compute processes across many devices, with the goal to minimize the noise and circuit depth required by quantum devices. In this paper, we cover the fundamental concepts of DQC and provide insight into where the field of DQC stands with respect to the field of chemistry -- a field which can potentially be used to demonstrate quantum advantage on noisy-intermediate scale quantum devices.

Attosecond X-ray pulses are the key to studying electron dynamics at their natural timescale in specifically targeted electronic states. They promise to build the conceptual bridge between physical and chemical photo-reaction processes. Free-electron lasers (FELs) have demonstrated their capability of generating intense attosecond X-ray pulses. The use of SASE-based FELs for time-resolving experiments and investigations of nonlinear X-ray absorption mechanisms, however, necessitates their full pulse-to-pulse characterisation which remains a cutting-edge challenge. We have characterised X-ray pulses with durations of down to 600 attoseconds and peak powers up to 200 GW at ~1 keV photon energy via angular streaking at the Small Quantum Systems instrument of the European XFEL in Germany. As a direct application, we present results of nonlinear X-ray--matter interaction via time-resolved electron spectroscopy on a transient system, observing single- and double-core-hole generation in neon atoms. Using the derived temporal information about each single X-ray pulse, we reveal an otherwise hidden peak-intensity dependence of the probability for formation of double-core vacancies in neon after primary K-shell ionisation. Our results advance the field of attosecond science with highly intense and fully characterised X-ray pulses to the state-specific investigation of electronic motion in non-stationary media.