Search

Nanoscale silicate dust particles are the most abundant refractory component observed in the interstellar medium and thought to play a key role in catalysing the formation of complex organic molecules in the star forming regions of space. We present a method to synthesise a laboratory analogue of nanoscale silicate dust particles on highly oriented pyrolytic graphite (HOPG) substrates by co-deposition of the atomic constituents. The resulting nanoparticulate films are sufficiently thin and conducting to allow for surface science investigations, and are characterised here, in situ under UHV, using X-ray photoelectron spectroscopy, near-edge X-ray absorption atomic fine spectroscopy and scanning tunnelling microscopy, and, ex situ, using scanning electron microscopy. We compare SiO$_{x}$ film growth with and without the use of atomic O beams during synthesis and conclude that exposure of the sample to atomic O leads to homogeneous films of interconnected nanoparticle networks. The networks covers the graphite substrate and demonstrate superior thermal stability, up to 1073 K, when compared to oxides produced without exposure to atomic O. In addition, control over the flux of atomic O during growth allows for control of the average oxidation state of the film produced. Photoelectron spectroscopy measurements demonstrate that fully oxidised films have an SiO$_{2}$ stoichiometry very close to bulk SiO$_{2}$ and scanning tunnelling microscopy images show the basic cluster building unit to have a radius of approximately 2.5 nm. The synthesis of SiO$_{x}$ films with adjustable stoichiometry and suitable for surface science experiments that require conducting substrates will be of great interest to the astrochemistry community, and will allow for nanoscale-investigation of the chemical processes thought to be catalysed at the surface of dust grains in space.

Token representations in high-dimensional latent spaces often exhibit redundancy, limiting computational efficiency and reducing structural coherence across model layers. Hierarchical latent space folding introduces a structured transformation mechanism that enforces a multi-scale organization within learned embeddings, refining representational compactness while preserving essential contextual distinctions. The proposed approach incorporates dynamic folding operations that iteratively adjust token embeddings through structured transformations, influencing both short-range and long-range dependencies in sequential processing tasks. Empirical evaluation demonstrates a reduction in representational variance across layers, contributing to more stable perplexity distributions and enhancing predictive confidence in text generation. The structured redistribution of attention head utilization leads to more efficient allocation of computational resources, particularly in deeper layers, where hierarchical refinements improve contextual abstraction. Comparative analysis of activation sparsity patterns suggests that hierarchical adjustments selectively reinforce critical pathways while reducing computational overhead in non-essential regions of the model. Statistical assessments of token reordering frequencies reveal that hierarchical modifications introduce subtle shifts in sequential dependencies, improving contextual alignment while maintaining syntactic correctness. Computational trade-offs associated with hierarchical folding introduce marginal increases in training time per epoch, yet empirical findings indicate that inference efficiency benefits from the structured representation adjustments. The results highlight the impact of hierarchical latent space folding on optimizing model performance through improved representation structuring and computational efficiency.

The origin of nebular HeII-emission in both local and high-redshift galaxies remains an unsolved problem. Various theories have been proposed to explain it, including HeII-ionization by high mass X-ray binaries, ultra-luminous X-ray sources, or "stripped" He stars, shock ionization, and hidden AGNs. All these theories have shortcomings, however, leaving the cause of nebular HeII emission unclear. We investigate the hypothesis that the photons responsible for driving nebular HeII emissions are produced by the evolution of single massive stars and/or WR stars. We combine models of stellar evolution with population synthesis and nebular models to identify the most favorable scenarios for producing nebular HeII via this channel. We find that, if WR winds are clumpy enough to become close to optically thin, stellar populations with a wide range of metallicities and rotation rates can produce HeII ionizing photons at rates sufficient to explain the observed nebular $I(HeII)/I(\mathrm{H}\beta)$ ratio $\sim 0.004-0.07$ found in HeII-emitting galaxies. Metal-poor, rapidly rotating stellar populations ($[\mathrm{Fe}/\mathrm{H}]=-2.0$, $v/v_\mathrm{crit}=0.4$) also reach these levels of HeII production even for partially clumpy winds. These scenarios also yield HeII, H$\beta$, and "Blue-Bump" line equivalent widths comparable to those observed in HeII emitters. Only for laminar, non-clumpy winds, do we fail to find combinations of metallicity and stellar rotation rate that yield $I(HeII)/I(\mathrm{H}\beta)$ values as high as those observed in HeII-emitters. Contrary to previous findings, we conclude that single WR stars can be a strong source for nebular HeII emission if their winds are sufficiently clumpy allowing significant escape of hard ionizing photons., Comment: submitted to A&A, 16 pages, 16 figures

Most active learning research has focused on methods which perform well when many labels are available, but can be dramatically worse than random selection when label budgets are small. Other methods have focused on the low-budget regime, but do poorly as label budgets increase. As the line between "low" and "high" budgets varies by problem, this is a serious issue in practice. We propose uncertainty coverage, an objective which generalizes a variety of low- and high-budget objectives, as well as natural, hyperparameter-light methods to smoothly interpolate between low- and high-budget regimes. We call greedy optimization of the estimate Uncertainty Herding; this simple method is computationally fast, and we prove that it nearly optimizes the distribution-level coverage. In experimental validation across a variety of active learning tasks, our proposal matches or beats state-of-the-art performance in essentially all cases; it is the only method of which we are aware that reliably works well in both low- and high-budget settings.

Although entangled state vectors cannot be described in terms of classically realistic variables, localized in space and time, any given entanglement experiment can be built from basic quantum circuit components with well-defined locations. By analyzing the (local) weak values for any given run of a quantum circuit, we present evidence for a localized account of any circuit's behavior. Specifically, even if the state is massively entangled, the weak values are found to evolve only when they pass through a local circuit element. They otherwise remain constant and do not evolve when other qubits pass through their circuit elements. A further surprise is found when two qubits are brought together in an exchange interaction, as their weak values then evolve according to a simple classical equation. The weak values are subject to both past and future constraints, so they can only be determined by considering the entire circuit "all-at-once", as in action principles. In the context of a few basic quantum gates, we show how an all-at-once model of a complete circuit could generate weak values without using state vectors as an intermediate step. Since these gates comprise a universal quantum gate set, this lends support to the claim that any quantum circuit can plausibly be underpinned by localized variables, providing a realistic, lower-level account of generic quantum systems., Comment: 17 pages, 6 figures, Comments and References welcome

Singlet fission (SF), the spin-allowed conversion of one singlet exciton into two triplet excitons, offers a promising strategy for enhancing the efficiency of photovoltaic devices. However, realising this potential necessitates materials capable of ultrafast (sub-picosecond) SF and the generation of long-lived (> microsecond) triplet excitons, a synthetic challenge. Some photosynthetic organisms have evolved sophisticated molecular architectures that demonstrate these criteria, but despite 40 years of study, the underlying SF mechanisms and its functional significance in these organisms remain unclear. Here, we use a suite of ultrafast and magneto-optical spectroscopic techniques to understand the mechanism of SF within light-harvesting 1 (LH1) complexes from wild-type and genetically modified photosynthetic bacteria. Our findings reveal a SF process, termed "heterofission", wherein singlet excitons are transformed into triplet excitons localised on adjacent carotenoid (Crt) and bacteriochlorophyll (BChl) molecules. We also uncover an unexpected functional role for SF in augmenting Crt-to-BChl photosynthetic energy transfer efficiency. By transiently storing electronic excitation within the SF-generated triplet pair, the system circumvents rapid thermalisation of Crt excitations, thereby enhancing energy transfer efficiency to the BChl Qy state, and enabling the organism to usefully harvest more sunlight., Comment: Main Text: 33 pages and 5 figures; Supplementary Information: 31 pages and 18 figures

Graph Transformers excel in long-range dependency modeling, but generally require quadratic memory complexity in the number of nodes in an input graph, and hence have trouble scaling to large graphs. Sparse attention variants such as Exphormer can help, but may require high-degree augmentations to the input graph for good performance, and do not attempt to sparsify an already-dense input graph. As the learned attention mechanisms tend to use few of these edges, such high-degree connections may be unnecessary. We show (empirically and with theoretical backing) that attention scores on graphs are usually quite consistent across network widths, and use this observation to propose a two-stage procedure, which we call Spexphormer: first, train a narrow network on the full augmented graph. Next, use only the active connections to train a wider network on a much sparser graph. We establish theoretical conditions when a narrow network's attention scores can match those of a wide network, and show that Spexphormer achieves good performance with drastically reduced memory requirements on various graph datasets.

Camera traps offer enormous new opportunities in ecological studies, but current automated image analysis methods often lack the contextual richness needed to support impactful conservation outcomes. Here we present an integrated approach that combines deep learning-based vision and language models to improve ecological reporting using data from camera traps. We introduce a two-stage system: YOLOv10-X to localise and classify species (mammals and birds) within images, and a Phi-3.5-vision-instruct model to read YOLOv10-X binding box labels to identify species, overcoming its limitation with hard to classify objects in images. Additionally, Phi-3.5 detects broader variables, such as vegetation type, and time of day, providing rich ecological and environmental context to YOLO's species detection output. When combined, this output is processed by the model's natural language system to answer complex queries, and retrieval-augmented generation (RAG) is employed to enrich responses with external information, like species weight and IUCN status (information that cannot be obtained through direct visual analysis). This information is used to automatically generate structured reports, providing biodiversity stakeholders with deeper insights into, for example, species abundance, distribution, animal behaviour, and habitat selection. Our approach delivers contextually rich narratives that aid in wildlife management decisions. By providing contextually rich insights, our approach not only reduces manual effort but also supports timely decision-making in conservation, potentially shifting efforts from reactive to proactive management., Comment: 32 Pages, 22 images

Transductive tasks on graphs differ fundamentally from typical supervised machine learning tasks, as the independent and identically distributed (i.i.d.) assumption does not hold among samples. Instead, all train/test/validation samples are present during training, making them more akin to a semi-supervised task. These differences make the analysis of the models substantially different from other models. Recently, Graph Transformers have significantly improved results on these datasets by overcoming long-range dependency problems. However, the quadratic complexity of full Transformers has driven the community to explore more efficient variants, such as those with sparser attention patterns. While the attention matrix has been extensively discussed, the hidden dimension or width of the network has received less attention. In this work, we establish some theoretical bounds on how and under what conditions the hidden dimension of these networks can be compressed. Our results apply to both sparse and dense variants of Graph Transformers.

Gas metallicity, ionization parameter, and gas pressure can affect the observed ratios of specific strong emission lines within galaxies. While the theoretical strong lines diagnostics for gas metallicity, ionization parameters, and gas pressure in star-forming regions are well-established, theoretical diagnostics for active galactic nuclei (AGNs) narrow line regions are still lacking. In Zhu et al. (2023), we presented a new AGN model that provides the best predictions for observations spanning the UV, optical, and infrared wavelengths. This paper presents a suite of theoretical diagnostics for the gas metallicity, ionization parameter, gas pressure, and the peak energy in AGN ionizing radiation field $E_{peak}$ for AGN narrow-line regions spanning the UV and optical wavelengths. We investigate the model dependency on the ionization parameter, gas pressure, $E_{peak}$, and the nitrogen scaling relation and make recommendations on metallicity diagnostics that are most robust against these parameters. We test our new AGN metallicity diagnostics using optical galaxy spectra from Sloan Digital Sky Survey DR16. These tests show that the metallicities measured from different diagnostics in this paper are consistent within $\sim0.3$ dex. We compare consistent HII and AGN diagnostics and demonstrate that HII and AGN diagnostics should not be used interchangeably. With a wide wavelength coverage, we anticipate that these AGN diagnostics will enable new metallicity studies of galaxies dominated by AGN., Comment: 33 pages, 22 figures, 7 tables, Accepted for publication in ApJ

We present $[OIII]/H_{\rm \beta}$ emision line flux ratio predictions for galaxies at $z \sim 7-9$ using the MAPPINGS V v5.2.0 photoionization modelling code combined with an analytic galaxy formation model. Properties such as pressure and ionization parameter that determine emission line properties are thought to evolve towards high redshift. In order to determine the range of expected interstellar conditions we extend previous modelling of the Star Formation Rate Density (SFRD) function to calculate the metallicity and ionization parameter, and incorporate the potential impact of turbulence on the density of the ISM. To validate our emission line predictions we calculate the [OIII] line luminosity and its dependence on UV luminosity, as well as the flux ratio $[OIII]/H_{\rm \beta}$ and its variation with the line luminosity, finding that both reproduce recent JWST observations from the FRESCO survey. We also use our model to predict the number counts of emission line galaxies across a range of redshift as well as the dependence of $[OIII]/H_{\rm \beta}$ on ionization parameter and metallicity. Finally, we show that the dependence of flux ratio on luminosity may provide a diagnostic of turbulent motion in galactic discs., Comment: 12 pages, 17 figures, Submitted to MNRAS

We discuss a general technique for using quantum logic spectroscopy to perform quantum non-demolition (QND) measurements that determine which of two subspaces a logic ion is in. We then show how to use the scheme to perform high fidelity state preparation and measurement (SPAM) and non-destructive leakage detection, as well as how this would reduce the engineering task associated with building larger quantum computers. Using a magnetic field gradient, we can apply a geometric phase interaction to map the magnetic sensitivity of the logic into onto the spin-flip probability of a readout ion. In the `low quantization field' regime, where the Zeeman splitting of the sublevels in the hyperfine ground state manifold are approximately integer multiples of each other, we show how to we can use the technique to distinguish between $2^{K}$ hyperfine sublevels in only $K$ logic/readout measurement sequences--opening the door for efficient state preparation of the logic ion. Finally, we show how the binary nature of the scheme, as well as its potential for high QND purity, let us improve SPAM fidelities by detecting and correcting errors rather than preventing them.

High-redshift protoclusters consisting of dusty starbursts are thought to play an important role in galaxy evolution. Their dusty nature makes them bright in the FIR/submm but difficult to find in optical/NIR surveys. Radio observations are an excellent way to study these dusty starbursts, as dust is transparent in the radio and there is a tight correlation between the FIR and radio emission of a galaxy. Here, we present MeerKAT 1.28 GHz radio imaging of 3 Herschel candidate protoclusters, with a synthesised beam size of ~$7.5'' \times 6.6''$ and a central thermal noise down to $4.35~\mu$Jy/beam. Our source counts are consistent with other radio counts with no evidence of overdensities. Around 95% of the Herschel sources have 1.28 GHz IDs. Using the Herschel 250 micron primary beam size as the searching radius, we find 54.2% Herschel sources have multiple 1.28 GHz IDs. Our average FIR-radio correlation coefficient $q_{250\mu m}$ is $2.33\pm 0.26$. Adding $q_{250\mu m}$ as a new constraint, the probability of finding chance-aligned sources is reduced by a factor of ~6, but with the risk of discarding true identifications of radio-loud/quiet sources. With accurate MeerKAT positions, we cross-match our Herschel sources to optical/NIR data followed by photometric redshift estimations. By removing $z<1$ sources, the density contrasts of two of the candidate protoclusters increase, suggestive of them being real protoclusters at $z>1$. There is also potentially a $0.9

We consider a class of BSM models where a generic scalar electroweak multiplet obtains a significant fraction of its mass from a coupling to the Higgs. Such models are non-decoupling: their new states are necessarily at the TeV scale or below, they can significantly alter the electroweak phase transition, and they have a pattern of low energy effects that are distinct from those predicted by SMEFT. Using their minimal gauge and Higgs couplings, we show that a future precision lepton collider (such as FCC-ee, CEPC, ILC, or CLIC) can probe all the non-decoupling parameter space of scalar electroweak multiplets, providing fundamental information on the mechanism of electroweak symmetry breaking., Comment: 27 pages, 4 figures

Obtaining compositional mappings is important for the model to generalize well compositionally. To better understand when and how to encourage the model to learn such mappings, we study their uniqueness through different perspectives. Specifically, we first show that the compositional mappings are the simplest bijections through the lens of coding length (i.e., an upper bound of their Kolmogorov complexity). This property explains why models having such mappings can generalize well. We further show that the simplicity bias is usually an intrinsic property of neural network training via gradient descent. That partially explains why some models spontaneously generalize well when they are trained appropriately., Comment: 4 pages

The Hilbert-Schmidt Independence Criterion (HSIC) is a powerful tool for nonparametric detection of dependence between random variables. It crucially depends, however, on the selection of reasonable kernels; commonly-used choices like the Gaussian kernel, or the kernel that yields the distance covariance, are sufficient only for amply sized samples from data distributions with relatively simple forms of dependence. We propose a scheme for selecting the kernels used in an HSIC-based independence test, based on maximizing an estimate of the asymptotic test power. We prove that maximizing this estimate indeed approximately maximizes the true power of the test, and demonstrate that our learned kernels can identify forms of structured dependence between random variables in various experiments.

This study evaluates the productivity improvements achieved using a generative artificial intelligence personal assistant tool (PAT) developed by Trane Technologies. The PAT, based on OpenAI's GPT 3.5 model, was deployed on Microsoft Azure to ensure secure access and protection of intellectual property. To assess the tool's productivity effectiveness, an experiment was conducted comparing the completion times and content quality of four common office tasks: writing an email, summarizing an article, creating instructions for a simple task, and preparing a presentation outline. Sixty-three (63) participants were randomly divided into a test group using the PAT and a control group performing the tasks manually. Results indicated significant productivity enhancements, particularly for tasks involving summarization and instruction creation, with improvements ranging from 3.3% to 69%. The study further analyzed factors such as the age of users, response word counts, and quality of responses, revealing that the PAT users generated more verbose and higher-quality content. An 'LLM-as-a-judge' method employing GPT-4 was used to grade the quality of responses, which effectively distinguished between high and low-quality outputs. The findings underscore the potential of PATs in enhancing workplace productivity and highlight areas for further research and optimization., Comment: 18 pages, 9 figures

Previous measurements of neutron energy using fast plastic scintillators while operating the Fusion Z Pinch Experiment (FuZE) constrained the energy of any yield-producing deuteron beams to less than $4.65 keV$. FuZE has since been operated at increasingly higher input power, resulting in increased plasma current and larger fusion neutron yields. A detailed experimental study of the neutron energy isotropy in these regimes applies more stringent limits to possible contributions from beam-target fusion. The FuZE device operated at $-25~kV$ charge voltage has resulted in average plasma currents of $370~kA$ and D-D fusion neutron yields of $4\times10^7$ neutrons per discharge. Measurements of the neutron energy isotropy under these operating conditions demonstrates the energy of deuteron beams is less than $7.4 \pm 5.6^\mathrm{(stat)} \pm 3.7^\mathrm{(syst)}~keV$. Characterization of the detector response has reduced the number of free parameters in the fit of the neutron energy distribution, improving the confidence in the forward-fit method. Gamma backgrounds have been measured and the impact of these contributions on the isotropy results have been studied. Additionally, a time dependent measurement of the isotropy has been resolved for the first time, indicating increases to possible deuteron beam energies at late times. This suggests the possible growth of $m$=0 instabilities at the end of the main radiation event but confirms that the majority of the neutron production exhibits isotropy consistent with thermonuclear origin., Comment: 16 pages, 11 figures, submitted to Journal of Nuclear Fusion

Machine learning techniques are an active area of research for speech enhancement for hearing aids, with one particular focus on improving the intelligibility of a noisy speech signal. Recent work has shown that feature encodings from self-supervised speech representation models can effectively capture speech intelligibility. In this work, it is shown that the distance between self-supervised speech representations of clean and noisy speech correlates more strongly with human intelligibility ratings than other signal-based metrics. Experiments show that training a speech enhancement model using this distance as part of a loss function improves the performance over using an SNR-based loss function, demonstrated by an increase in HASPI, STOI, PESQ and SI-SNR scores. This method takes inference of a high parameter count model only at training time, meaning the speech enhancement model can remain smaller, as is required for hearing aids., Comment: Accepted for EUSIPCO 2024

We present a theoretical explanation of the ``grokking'' phenomenon, where a model generalizes long after overfitting,for the originally-studied problem of modular addition. First, we show that early in gradient descent, when the ``kernel regime'' approximately holds, no permutation-equivariant model can achieve small population error on modular addition unless it sees at least a constant fraction of all possible data points. Eventually, however, models escape the kernel regime. We show that two-layer quadratic networks that achieve zero training loss with bounded $\ell_{\infty}$ norm generalize well with substantially fewer training points, and further show such networks exist and can be found by gradient descent with small $\ell_{\infty}$ regularization. We further provide empirical evidence that these networks as well as simple Transformers, leave the kernel regime only after initially overfitting. Taken together, our results strongly support the case for grokking as a consequence of the transition from kernel-like behavior to limiting behavior of gradient descent on deep networks., Comment: Accepted by ICML 2024

The ProbCover method of Yehuda et al. is a well-motivated algorithm for active learning in low-budget regimes, which attempts to "cover" the data distribution with balls of a given radius at selected data points. We demonstrate, however, that the performance of this algorithm is extremely sensitive to the choice of this radius hyper-parameter, and that tuning it is quite difficult, with the original heuristic frequently failing. We thus introduce (and theoretically motivate) a generalized notion of "coverage," including ProbCover's objective as a special case, but also allowing smoother notions that are far more robust to hyper-parameter choice. We propose an efficient greedy method to optimize this coverage, generalizing ProbCover's algorithm; due to its close connection to kernel herding, we call it "MaxHerding." The objective can also be optimized non-greedily through a variant of $k$-medoids, clarifying the relationship to other low-budget active learning methods. In comprehensive experiments, MaxHerding surpasses existing active learning methods across multiple low-budget image classification benchmarks, and does so with less computational cost than most competitive methods., Comment: Accepted to ECCV2024

Learning dynamics, which describes how the learning of specific training examples influences the model's predictions on other examples, gives us a powerful tool for understanding the behavior of deep learning systems. We study the learning dynamics of large language models during different types of finetuning, by analyzing the step-wise decomposition of how influence accumulates among different potential responses. Our framework allows a uniform interpretation of many interesting observations about the training of popular algorithms for both instruction tuning and preference tuning. In particular, we propose a hypothetical explanation of why specific types of hallucination are strengthened after finetuning, e.g., the model might use phrases or facts in the response for question B to answer question A, or the model might keep repeating similar simple phrases when generating responses. We also extend our framework and highlight a unique "squeezing effect" to explain a previously observed phenomenon in off-policy direct preference optimization (DPO), where running DPO for too long makes even the desired outputs less likely. This framework also provides insights into where the benefits of on-policy DPO and other variants come from. The analysis not only provides a novel perspective of understanding LLM's finetuning but also inspires a simple, effective method to improve alignment performance.

We discuss dressing trapped ions with the near field of a trap integrated wire. Ramping a dressing field on/off adiabatically before/after an operation changes its effective Hamiltonian. The amplitude and detuning of the dressing field act as tunable degrees of freedom we can use to `customize' the properties of any operation. We propose three use cases for this general tool. First, we can generate `artificial' clock states, where we eliminate the (assumed to be small) linear sensitivity of a qubit. Second, we can break the degeneracies that often complicate shelving at low quantization fields\textemdash allowing us to implement operations with linearly polarized microwaves that would, otherwise, require circular polarization. Finally, we can implement laser-free single qubit gates on a set of `target' ions using fields that are separated from the rest of the computer in frequency space.