Search

The language technology moonshot moment of Generative, Large Language Models (GLLMs) was not limited to English: These models brought a surge of technological applications, investments and hype to low-resource languages as well. However, the capabilities of these models in languages such as Danish were until recently difficult to verify beyond qualitative demonstrations due to a lack of applicable evaluation corpora. We present a GLLM benchmark to evaluate Danoliteracy, a measure of Danish language and cultural competency, across eight diverse scenarios such Danish citizenship tests and abstractive social media question answering. This limited-size benchmark is found to produce a robust ranking that correlates to human feedback at $\rho \sim 0.8$ with GPT-4 and Claude Opus models achieving the highest rankings. Analyzing these model results across scenarios, we find one strong underlying factor explaining $95\%$ of scenario performance variance for GLLMs in Danish, suggesting a $g$ factor of model consistency in language adaption., Comment: 16 pages, 13 figures, submitted to: NoDaLiDa/Baltic-HLT 2025

Multimodal systems have great potential to assist humans in procedural activities, where people follow instructions to achieve their goals. Despite diverse application scenarios, systems are typically evaluated on traditional classification tasks, e.g., action recognition or temporal action segmentation. In this paper, we present a novel evaluation dataset, ProMQA, to measure system advancements in application-oriented scenarios. ProMQA consists of 401 multimodal procedural QA pairs on user recording of procedural activities coupled with their corresponding instruction. For QA annotation, we take a cost-effective human-LLM collaborative approach, where the existing annotation is augmented with LLM-generated QA pairs that are later verified by humans. We then provide the benchmark results to set the baseline performance on ProMQA. Our experiment reveals a significant gap between human performance and that of current systems, including competitive proprietary multimodal models. We hope our dataset sheds light on new aspects of models' multimodal understanding capabilities., Comment: 18 pages, 11 figures

This paper studies a non-commutative generalisation of Coxeter friezes due to Berenstein and Retakh. It generalises several earlier results to this situation: A formula for frieze determinants, a $T$-path formula expressing the Laurent phenomenon, and results on gluing friezes together. One of our tools is a non-commutative version of the weak friezes introduced by Canakci and Jorgensen., Comment: Updated references. 26 pages and 16 figures

This work presents a novel means for understanding learning dynamics and scaling relations in neural networks. We show that certain measures on the spectrum of the empirical neural tangent kernel, specifically entropy and trace, yield insight into the representations learned by a neural network and how these can be improved through architecture scaling. These results are demonstrated first on test cases before being shown on more complex networks, including transformers, auto-encoders, graph neural networks, and reinforcement learning studies. In testing on a wide range of architectures, we highlight the universal nature of training dynamics and further discuss how it can be used to understand the mechanisms behind learning in neural networks. We identify two such dominant mechanisms present throughout machine learning training. The first, information compression, is seen through a reduction in the entropy of the NTK spectrum during training, and occurs predominantly in small neural networks. The second, coined structure formation, is seen through an increasing entropy and thus, the creation of structure in the neural network representations beyond the prior established by the network at initialization. Due to the ubiquity of the latter in deep neural network architectures and its flexibility in the creation of feature-rich representations, we argue that this form of evolution of the network's entropy be considered the onset of a deep learning regime., Comment: 11 pages, 3 figures

A novel balanced air-biased coherent detection scheme for capturing ultrabroadband terahertz (THz) waveforms is implemented. The balanced detection scheme allows for coherent detection at the full repetition rate of the laser system without requiring bias modulation, signal generators, or lock-in amplifiers while doubling the dynamic range and quadrupling the signal-to-noise ratio compared to conventional air-biased coherent detection. These advantages are achieved by rotating the bias electrodes by 90{\deg} relative to the conventional scheme. With a 1 kHz driving laser, the scheme enables sub-second, high-fidelity waveform acquisition with a continuously moving delay stage, demonstrated by collecting 200 waveforms in 100 s. The balanced detection scheme paves the way for much faster and higher quality 2D ultrabroadband terahertz spectroscopy., Comment: 4 pages, 4 figures

The Mittelman-di Cicco-Walker (MDW) H$\alpha$ Sky Survey is an autonomously-operated and ongoing all-sky imaging survey in the narrowband H$\alpha$ wavelength. The survey was founded by amateur astronomers, and is presented here in its first stage of refinement for rigorous scientific use. Each field is exposed through an H$\alpha$ filter with a 3nm bandwidth for a total of four hours, with a pixel scale of 3.2 arcsec. Here, we introduce the first Data Release of the MDW H$\alpha$ Survey (Data Release 0, or DR0), spanning 238 fields in the region of Orion (~3100 deg$^2$). DR0 includes: calibrated mean fields, star-removed mean fields, a point source catalog matched to Data Release 1 of the Panoramic Survey Telescope and Rapid Response System (Pan-STARRS1) and the INT Galactic Plane Survey (IGAPS), and mosaics.

We study the influence of airborne CO2 on the charge state of carboxylate stabilized polymer latex particles suspended in aqueous electrolytes. We combine conductometric experiments interpreted in terms of Hessinger's conductivity model with Poisson-Boltzmann cell (PBC) model calculations with charge regulation boundary conditions. Without CO2, a minority of the weakly acidic surface groups are dissociated and only a fraction of the total number of counter-ions actually contribute to conductivity. The remaining counter-ions exchange freely with added other ions like Na+, K+ or Cs+. From the PBC-calculations we infer a corresponding pKa of 4.26 as well as a renormalized charge in reasonably good agreement with the number of freely mobile counter-ions. Equilibration of salt- and CO2-free suspensions against ambient air leads to a drastic de-charging, which exceeds by far the expected effects of to dissolved CO2 and its dissociation products. Further, no counter-ion-exchange is observed. To reproduce the experimental findings, we have to assume an effective pKa of 6.48. This direct influence of CO2 on the state of surface group dissociation explains our recent finding of a CO2-induced decrease of the {\zeta}-potential and supports the suggestion of an additional charge regulation caused by molecular CO2. Given the importance of charged surfaces in contact with aqueous electrolytes, we anticipate that our observations bear substantial theoretical challenges and important implications for applications ranging from desalination to bio-membranes., Comment: Submitted to Soft Matter, regular article, 22 pages, 6 figures, 65 references, supplementary material of 8 pages with 7 figures

Protein characterization using nanopore-based devices promises to be a breakthrough method in basic research, diagnostics, and analytics. Current research includes the use of machine learning to achieve this task. In this work, a comprehensive statistical analysis of nanopore current signals is performed and demonstrated to be sufficient for classifying up to 42 peptides with over 70 % accuracy. Two sets of features, the statistical moments and the catch22 set, are compared both in their representations and after training small classifier neural networks. We demonstrate that complex features of the events, captured in both the catch22 set and the central moments, are key in classifying peptides with otherwise similar mean currents. These results highlight the efficacy of purely statistical analysis of nanopore data and suggest a path forward for more sophisticated classification techniques.

In this work, we investigate the shape evolution of rotated, embedded, initially cylindrical grains (with [001] cylinder axis) in Ni under an applied synthetic driving force via molecular dynamics simulations and a continuum, disconnection-based grain boundary migration model. For some initial misorientations, the expanding grains form well-defined, faceted shapes, while for others the shapes remain cylindrical. The embedded grains tend to rotate during grain boundary migration, with the direction of rotation dependent on initial misorientation and direction of growth (expand/shrink). The kinetic shapes, which are bounded by low mobility grain boundary planes, differ from equilibrium shapes (bounded by low energy grain boundaries). The multi-mode disconnection model-based predictions are consistent with the molecular dynamics results for faceting tendency, kinetic grain shape, and grain rotation as a function of misorientation and whether the grains are expanding or contracting. This demonstrates that grain boundary migration and associated grain rotation are mediated by disconnection flow along grain boundaries., Comment: 40 pages, 18 figures

These lecture notes in geometric mechanics are meant to convey insight through clear definitions and workable examples. The lecture format adopted here is intended to convey the immediacy of the taught course and to be useful as a basis for other courses. The lecture notes comprise: AP = Applications of Pure maths, e.g., Noether's theorem: Lie group symmetry of Hamilton's variational principle implies conservation laws for its equations of motion.\smallskip PA = Purifications of Applied maths, e.g., Euler fluid dynamics describes geodesic flow on the manifold of smooth invertible maps acting on the domain of flow. \smallskip Both AP and PA appear here, though the difference is not mentioned. It is left to the reader to decide whether it was AP or PA in each of the lectures containing well over sixty solved exercises. An aspect of modern applications emphasised here is the use of the composition of evolutionary maps for multi-physics, multi-timescale interactions including waves interacting with flows in the Euler--Poincar\'e framework in geophysical fluid dynamics (GFD) for ocean and atmosphere dynamics, and in magnetohydrodynamics (MHD) for applications in plasma physics such as magnetic confinement fusion (MFC) and astrophysical processes such as Alfv\'en waves and gravity waves propagating on the Solar tachocline. The topics covered in each lecture can also be gleaned from its table of contents listed at the onset of each lecture., Comment: Textbook in the form of 31 consecutive lectures. arXiv admin note: substantial text overlap with arXiv:0708.1585

This paper investigates the effects of stochastic variations in bathymetry on the solutions of the thermal quasigeostrophic (TQG) equations. These stochastic perturbations generate a variety of different types of ensemble spread in the solution behaviour whilst also preserving the deterministic Lie Poisson structure and Casimir conservation laws. We numerically compare the solution sensitivity, to another type of structure-preserving stochastic perturbation where instead of bathymetry, the velocity is stochastically perturbed.

Superconductivity and magnetism are antagonistic quantum matter, while their intertwining has long been considered in frustrated-lattice systems1-3. In this work, we utilize scanning tunneling microscopy and muon spin resonance to discover time-reversal symmetry-breaking superconductivity in kagome metal Cs(V,Ta)3Sb5, where the Cooper pairing exhibits magnetism and is modulated by it. In the magnetic channel, we observe spontaneous internal magnetism in a full-gap superconducting state. Under perturbations of inverse magnetic fields, we detect a time-reversal asymmetrical interference of Bogoliubov quasi-particles at a circular vector. At this vector, the pairing gap spontaneously modulates, which is distinct from pair density waves occurring at a point vector and consistent with the theoretical proposal of unusual interference effect under time-reversal symmetry-breaking. The correlation between internal magnetism, Bogoliubov quasi-particles, and pairing modulation provides a chain of experimental clues for time-reversal symmetry-breaking kagome superconductivity., Comment: To appear in Nature Materials (2024)

Graph-based holistic scene representations facilitate surgical workflow understanding and have recently demonstrated significant success. However, this task is often hindered by the limited availability of densely annotated surgical scene data. In this work, we introduce an end-to-end framework for the generation and optimization of surgical scene graphs on a downstream task. Our approach leverages the flexibility of graph-based spectral clustering and the generalization capability of foundation models to generate unsupervised scene graphs with learnable properties. We reinforce the initial spatial graph with sparse temporal connections using local matches between consecutive frames to predict temporally consistent clusters across a temporal neighborhood. By jointly optimizing the spatiotemporal relations and node features of the dynamic scene graph with the downstream task of phase segmentation, we address the costly and annotation-burdensome task of semantic scene comprehension and scene graph generation in surgical videos using only weak surgical phase labels. Further, by incorporating effective intermediate scene representation disentanglement steps within the pipeline, our solution outperforms the SOTA on the CATARACTS dataset by 8% accuracy and 10% F1 score in surgical workflow recognition, Comment: 9 pages, 3 figures, 3 tables, MICCAI GRAIL Workshop paper

Absorption of elastic waves in complex media often depends on frequency in a linear way for both longitudinal and shear waves. This universal property occurs in media such as rocks, unconsolidated sediments, and human tissue. Absorption is due to relaxation processes at atomic scale up to the sub-micron scale of biological materials, and we argue that these processes are thermally activated. Unusual for ultrasonics and seismics, we can therefore express absorption as an integral over an activation energy landscape weighted by an energy distribution. The universal power-law property corresponds to a flat activation energy landscape corresponding to maximal randomness., Comment: 6 pages + 2 in supplement, 4 figures

This paper treats nonlinear wave current interactions in their simplest form, as an overtaking collision. In one spatial dimension, the paper investigates the collision interaction formulated as an initial value problem of a Burgers bore overtaking solutions of two types of nonlinear wave equations, Korteweg de Vries (KdV) and nonlinear Schrodinger (NLS). The bore wave state arising after the overtaking Burgers-KdV collision in numerical simulations is found to depend qualitatively on the balance between nonlinearity and dispersion in the KdV equation. The Burgers-KdV system is also made stochastic by following the stochastic advection by Lie transport approach (SALT)., Comment: 16 pages, 4 figures, 1st version

Cosmological emulators of observables such as the Cosmic Microwave Background (CMB) spectra and matter power spectra commonly use training data sampled from a Latin hypercube. This method often incurs high computational costs by covering less relevant parts of the parameter space, especially in high dimensions where only a small fraction of the parameter space yields a significant likelihood. In this paper, we introduce hypersphere sampling, which instead concentrates sample points in regions with higher likelihoods, significantly enhancing the efficiency and accuracy of emulators. A novel algorithm for sampling within a high-dimensional hyperellipsoid aligned with axes of correlation in the cosmological parameters is presented. This method focuses the distribution of training data points on areas of the parameter space that are most relevant to the models being tested, thereby avoiding the computational redundancies common in Latin hypercube approaches. Comparative analysis using the \textsc{connect} emulation tool demonstrates that hypersphere sampling can achieve similar or improved emulation precision with more than an order of magnitude fewer data points and thus less computational effort than traditional methods. This was tested for both the $\Lambda$CDM model and a 5-parameter extension including Early Dark Energy, massive neutrinos, and additional ultra-relativistic degrees of freedom. Our results suggest that hypersphere sampling holds potential as a more efficient approach for cosmological emulation, particularly suitable for complex, high-dimensional models., Comment: 22 pages, 10 figures

Ammonia is expected to play an important role in the green transition, both as a hydrogen carrier and a zero-emission fuel. The use of refrigerated ammonia is attractive due to its relatively high volumetric energy density and increased safety compared to pressurized solutions. Ammonia is highly toxic, and with new applications and increased global demand come stricter requirements for safe handling. Cold gaseous ammonia following a spill of refrigerated ammonia will in contact with humid air cause fog formation. In an environment rich in ammonia, these droplets will due to ammonia's strong hygroscopicity consist of considerable amounts of liquid ammonia as well as water. Fog formation affects the ammonia-air density and thus influences the dispersion dynamics, with a potentially significant impact on hazardous zones. In this work, we present a CFD model including an ammonia-water fog formation model based on accurate thermodynamics. This includes modeling the vapor-liquid equilibrium and accounting for the exothermic mixing of ammonia and water. We apply this CFD model to relevant cases and demonstrate the significant impact of the fog. We analyze the effect of varying relative humidity, fog visibility, influence of wind, and pool evaporation rate. Finally, we model the experimental Red Squirrel test 1F., Comment: 15 pages, 18 figures

This work introduces SwarmRL, a Python package designed to study intelligent active particles. SwarmRL provides an easy-to-use interface for developing models to control microscopic colloids using classical control and deep reinforcement learning approaches. These models may be deployed in simulations or real-world environments under a common framework. We explain the structure of the software and its key features and demonstrate how it can be used to accelerate research. With SwarmRL, we aim to streamline research into micro-robotic control while bridging the gap between experimental and simulation-driven sciences. SwarmRL is available open-source on GitHub at https://github.com/SwarmRL/SwarmRL., Comment: 16 pages, 3 figures

Injective resolutions of modules are key objects of homological algebra, which are used for the computation of derived functors. Semiinjective resolutions of chain complexes are more general objects, which are used for the computation of $\operatorname{Hom}$ spaces in the derived category $\mathscr{D}( A )$ of a ring $A$. Minimal semiinjective resolutions have the additional property of being unique. The $Q$-shaped derived category $\mathscr{D}_Q( A )$ consists of $Q$-shaped diagrams for a suitable preadditive category $Q$, and it generalises $\mathscr{D}( A )$. Some special cases of $\mathscr{D}_Q( A )$ are the derived categories of differential modules, $m$-periodic chain complexes, and $N$-complexes, and there are many other possibilities. The category $\mathscr{D}_Q( A )$ shares some key properties of $\mathscr{D}( A )$; for instance, it is triangulated and compactly generated. This paper establishes a theory of minimal semiinjective resolutions in $\mathscr{D}_Q( A )$. As a sample application, it generalises a theorem by Ringel--Zhang on differential modules., Comment: 21 pages

Gravitational clustering in our cosmic vicinity is expected to lead to an enhancement of the local density of relic neutrinos. We derive expressions for the neutrino density, using a perturbative approach to kinetic field theory and perturbative solutions of the Vlasov equation up to second order. Our work reveals that both formalisms give exactly the same results and can thus be considered equivalent. Numerical evaluation of the local relic neutrino density at first and second order provides some fundamental insights into the frequently applied approach of linear response to neutrino clustering (also known as the Gilbert equation). Against the naive expectation, including the second-order contribution does not lead to an improvement of the prediction for the local relic neutrino density but to a dramatic overestimation. This is because perturbation theory breaks down in a momentum-dependent fashion and in particular for densities well below unity., Comment: 34 pages, 7 figures

We demonstrate a novel approach to reservoir computer measurements using random matrices. We do so to motivate how atomic-scale devices might be used for real-world computing applications. Our approach uses random matrices to construct reservoir measurements, introducing a simple, scalable means for producing state descriptions. In our studies, two reservoirs, a five-atom Heisenberg spin chain, and a five-qubit quantum circuit, perform time series prediction and data interpolation. The performance of the measurement technique and current limitations are discussed in detail alongside an exploration of the diversity of measurements yielded by the random matrices. Additionally, we explore the role of the parameters of the reservoirs, adjusting coupling strength and the measurement dimension, yielding insights into how these learning machines might be automatically tuned for different problems. This research highlights using random matrices to measure simple quantum reservoirs for natural learning devices and outlines a path forward for improving their performance and experimental realization., Comment: 12 pages, 5 figures

We derive new models of stochastic Hall magnetohydrodynamics (MHD) by using a symmetry-reduced stochastic Euler-Poincar\'e variational principle. The new stochastic Hall MHD theory has potential applications for uncertainty quantification and data assimilation in space plasma (space weather) and solar physics. The stochastic geometric mechanics approach we take here produces coordinate-free results which may then be applied in a variety of spatial configurations., Comment: 24 pages, 0 figures, 1st version

Polysomnographic recordings are essential for diagnosing many sleep disorders, yet their detailed analysis presents considerable challenges. With the rise of machine learning methodologies, researchers have created various algorithms to automatically score and extract clinically relevant features from polysomnography, but less research has been devoted to how exactly the algorithms should be incorporated into the workflow of sleep technologists. This paper presents a sophisticated data collection platform developed under the Sleep Revolution project, to harness polysomnographic data from multiple European centers. A tripartite platform is presented: a user-friendly web platform for uploading three-night polysomnographic recordings, a dedicated splitter that segments these into individual one-night recordings, and an advanced processor that enhances the one-night polysomnography with contemporary automatic scoring algorithms. The platform is evaluated using real-life data and human scorers, whereby scoring time, accuracy and trust are quantified. Additionally, the scorers were interviewed about their trust in the platform, along with the impact of its integration into their workflow., Comment: 19 pages

The Third Generation Partnership Project (3GPP) has successfully introduced standards for global mobility. However, the volume and complexity of these standards has increased over time, thus complicating access to relevant information for vendors and service providers. Use of Generative Artificial Intelligence (AI) and in particular Large Language Models (LLMs), may provide faster access to relevant information. In this paper, we evaluate the capability of state-of-art LLMs to be used as Question Answering (QA) assistants for 3GPP document reference. Our contribution is threefold. First, we provide a benchmark and measuring methods for evaluating performance of LLMs. Second, we do data preprocessing and fine-tuning for one of these LLMs and provide guidelines to increase accuracy of the responses that apply to all LLMs. Third, we provide a model of our own, TeleRoBERTa, that performs on-par with foundation LLMs but with an order of magnitude less number of parameters. Results show that LLMs can be used as a credible reference tool on telecom technical documents, and thus have potential for a number of different applications from troubleshooting and maintenance, to network operations and software product development., Comment: Accepted to ICMLCN 2024, Stockholm, May 2024. Updating typo in authors list

Reinforcement learning (RL) is a flexible and efficient method for programming micro-robots in complex environments. Here we investigate whether reinforcement learning can provide insights into biological systems when trained to perform chemotaxis. Namely, whether we can learn about how intelligent agents process given information in order to swim towards a target. We run simulations covering a range of agent shapes, sizes, and swim speeds to determine if the physical constraints on biological swimmers, namely Brownian motion, lead to regions where reinforcement learners' training fails. We find that the RL agents can perform chemotaxis as soon as it is physically possible and, in some cases, even before the active swimming overpowers the stochastic environment. We study the efficiency of the emergent policy and identify convergence in agent size and swim speeds. Finally, we study the strategy adopted by the reinforcement learning algorithm to explain how the agents perform their tasks. To this end, we identify three emerging dominant strategies and several rare approaches taken. These strategies, whilst producing almost identical trajectories in simulation, are distinct and give insight into the possible mechanisms behind which biological agents explore their environment and respond to changing conditions., Comment: 12 pages, 6 figures

Recently, cosmology has seen a surge in alternative models that purport to solve the discrepancy between the values of the Hubble constant $H_0$ as measured by cosmological microwave background anisotropies and local supernovae, respectively. In particular, many of the most successful approaches have involved varying fundamental constants, such as an alternative value of the fine structure constant and time-varying values of the electron mass, the latter of which showed particular promise as the strongest candidate in several earlier studies. Inspired by these approaches, in this paper, we investigate a cosmological model where the value of the geometric constant $\pi$ is taken to be a free model parameter. Using the latest CMB data from Planck as well as baryon-acoustic oscillation data, we constrain the parameters of the model and find a strong correlation between $\pi$ and $H_0$, with the final constraint $H_0 = 71.3 \pm 1.1 \ \mathrm{ km/s/Mpc}$, equivalent to a mere $1.5\sigma$ discrepancy with the value measured by the SH0ES collaboration. Furthermore, our results show that $\pi = 3.206 \pm 0.038$ at $95 \%$ C.L., which is in good agreement with several external measurements discussed in the paper. Hence, we conclude that the $\pi \Lambda$CDM model presented in this paper, which has only a single extra parameter, currently stands as the perhaps strongest solution to the Hubble tension., Comment: 8 pages, 6 figures, 0 unicorns, 1 divine vision. Comments welcome

This paper investigates a robust joint power allocation and beamforming scheme for in-band full-duplex multi-cell multi-user (IBFD-MCMU) networks. A mean-squared error (MSE) minimization problem is formulated with constraints on the power budgets and residual self-interference (RSI) power. The problem is not convex, so we decompose it into two sub-problems: interference management beamforming and power allocation, and give closed-form solutions to the sub-problems. Then we propose an iterative algorithm to yield an overall solution. The computational complexity and convergence behavior of the algorithm are analyzed. Our method can enhance the analog self-interference cancellation (ASIC) depth provided by the precoder with less effect on the downlink communication than the existing null-space projection method, inspiring a low-cost but efficient IBFD transceiver design. It can achieve 42.9% of IBFD gain in terms of spectral efficiency with only antenna isolation, while this value increases to 60.9% with further digital self-interference cancellation (DSIC). Numerical results illustrate that our algorithm is robust to hardware impairments and channel uncertainty. With sufficient ASIC depth, our method reduces the computation time by at least 20% than the existing scheme due to its faster convergence speed at the cost of < 12.5% sum rate loss. The benefit is much more significant with single-antenna users that our algorithm saves at least 40% of the computation time at the cost of < 10% sum rate reduction.

Based on Berenstein and Retakh's notion of noncommutative polygons we introduce and study noncommutative frieze patterns. We generalize several notions and fundamental properties from the classic (commutative) frieze patterns to noncommutative frieze patterns, e.g. propagation formulae and $\mu$-matrices, quiddity cycles and reduction formulae, and we show that local noncommutative exchange relations and local triangle relations imply all noncommutative exchange relations and triangle relations. Throughout, we allow coefficients, so we obtain generalizations of results from our earlier paper on frieze patterns with coefficients from the commutative to the noncommutative setting., Comment: 18 pages. v2: reference added to a paper by Bergeron and Reutenauer

We extend previous inelastic neutron scattering results on the geometrically frustrated antiferromagnet hexagonal-YMnO$_3$, which has been suggested to belong to the class of classical spin liquids. We extend the energy transfer coverage of the diffuse signal up to 6.9 meV within a wide temperature range around the ordering temperature, $T_\mathrm{N}$. The two distinct diffuse signals in the a-b plane, the signal localized at $\Gamma$' and the scattering intensity connecting $\Gamma$' points over the M', are shown to be only weakly energy dependent. In addition, an external magnetic field of up to 10.5 T applied along c is shown to have no effect on the diffuse signal. In the orthogonal scattering plane, the signals are shown to be dependent on l only through the magnetic form factor, showing that the correlations are purely two-dimensional, and supporting its origin to be the frustrated Mn$^{3+}$ triangles. This result is corroborated by atomistic spin dynamics simulations showing similar scattering vector and temperature behaviours. Lastly, data for the spin wave scattering in the (h, 0, l) plane allow for a discussion of the magnetic ground state where better agreement is found between the data and an ordered structure of the $\Gamma_1$ or $\Gamma_3$ symmetry, albeit crystal electric field arguments dismisses the $\Gamma_1$ as possibility.

Constructing a timeline requires identifying the chronological order of events in an article. In prior timeline construction datasets, temporal orders are typically annotated by either event-to-time anchoring or event-to-event pairwise ordering, both of which suffer from missing temporal information. To mitigate the issue, we develop a new evaluation dataset, TimeSET, consisting of single-document timelines with document-level order annotation. TimeSET features saliency-based event selection and partial ordering, which enable a practical annotation workload. Aiming to build better automatic timeline construction systems, we propose a novel evaluation framework to compare multiple task formulations with TimeSET by prompting open LLMs, i.e., Llama 2 and Flan-T5. Considering that identifying temporal orders of events is a core subtask in timeline construction, we further benchmark open LLMs on existing event temporal ordering datasets to gain a robust understanding of their capabilities. Our experiments show that (1) NLI formulation with Flan-T5 demonstrates a strong performance among others, while (2) timeline construction and event temporal ordering are still challenging tasks for few-shot LLMs. Our code and data are available at https://github.com/kimihiroh/timeset.

The quantum theory of atoms in molecules (QTAIM) gives access to well-defined local atomic energies. Due to their locality, these energies are potentially interesting in fitting atomistic machine learning models as they inform about physically relevant properties. However, computationally, quantum-mechanically accurate local energies are notoriously difficult to obtain for large systems. Here, we show that by employing semi-empirical correlations between different components of the total energy, we can obtain well-defined local energies at a moderate cost. We employ this methodology to investigate energetics in noble liquids or argon, krypton, and their mixture. Instead of using these local energies to fit atomistic models, we show how well these local energies are reproduced by machine-learned models trained on the total energies. The results of our investigation suggest that smaller neural networks, trained only on the total energy of an atomistic system, are more likely to reproduce the underlying local energy partitioning faithfully than larger networks. Furthermore, we demonstrate that networks more capable of this energy decomposition are, in turn, capable of transferring to previously unseen systems. Our results are a step towards understanding how much physics can be learned by neural networks and where this can be applied, particularly how a better understanding of physics aids in the transferability of these neural networks., Comment: 13 pages, 10 figures