Search

Digital Twins have emerged as a disruptive technology with great potential; they can enhance WDS by offering real-time monitoring, predictive maintenance, and optimization capabilities. This paper describes the development of a state-of-the-art DT platform for WDS, introducing advanced technologies such as the Internet of Things, Artificial Intelligence, and Machine Learning models. This paper provides insight into the architecture of the proposed platform-CAUCCES-that, informed by both historical and meteorological data, effectively deploys AI/ML models like LSTM networks, Prophet, LightGBM, and XGBoost in trying to predict water consumption patterns. Furthermore, we delve into how optimization in the maintenance of WDS can be achieved by formulating a Constraint Programming problem for scheduling, hence minimizing the operational cost efficiently with reduced environmental impacts. It also focuses on cybersecurity and protection to ensure the integrity and reliability of the DT platform. In this view, the system will contribute to improvements in decision-making capabilities, operational efficiency, and system reliability, with reassurance being drawn from the important role it can play toward sustainable management of water resources., Comment: 78 pages, 18 figures

The capabilities and limitations of Large Language Models have been sketched out in great detail in recent years, providing an intriguing yet conflicting picture. On the one hand, LLMs demonstrate a general ability to solve problems. On the other hand, they show surprising reasoning gaps when compared to humans, casting doubt on the robustness of their generalisation strategies. The sheer volume of data used in the design of LLMs has precluded us from applying the method traditionally used to measure generalisation: train-test set separation. To overcome this, we study what kind of generalisation strategies LLMs employ when performing reasoning tasks by investigating the pretraining data they rely on. For two models of different sizes (7B and 35B) and 2.5B of their pretraining tokens, we identify what documents influence the model outputs for three simple mathematical reasoning tasks and contrast this to the data that are influential for answering factual questions. We find that, while the models rely on mostly distinct sets of data for each factual question, a document often has a similar influence across different reasoning questions within the same task, indicating the presence of procedural knowledge. We further find that the answers to factual questions often show up in the most influential data. However, for reasoning questions the answers usually do not show up as highly influential, nor do the answers to the intermediate reasoning steps. When we characterise the top ranked documents for the reasoning questions qualitatively, we confirm that the influential documents often contain procedural knowledge, like demonstrating how to obtain a solution using formulae or code. Our findings indicate that the approach to reasoning the models use is unlike retrieval, and more like a generalisable strategy that synthesises procedural knowledge from documents doing a similar form of reasoning.

As object detection techniques continue to evolve, understanding their relationships with complementary visual tasks becomes crucial for optimising model architectures and computational resources. This paper investigates the correlations between object detection accuracy and two fundamental visual tasks: depth prediction and visual saliency prediction. Through comprehensive experiments using state-of-the-art models (DeepGaze IIE, Depth Anything, DPT-Large, and Itti's model) on COCO and Pascal VOC datasets, we find that visual saliency shows consistently stronger correlations with object detection accuracy (mA$\rho$ up to 0.459 on Pascal VOC) compared to depth prediction (mA$\rho$ up to 0.283). Our analysis reveals significant variations in these correlations across object categories, with larger objects showing correlation values up to three times higher than smaller objects. These findings suggest incorporating visual saliency features into object detection architectures could be more beneficial than depth information, particularly for specific object categories. The observed category-specific variations also provide insights for targeted feature engineering and dataset design improvements, potentially leading to more efficient and accurate object detection systems., Comment: Code Available at: https://github.com/mbar0075/Object-Detection-Correlation-Saliency-vs-Depth

Future cosmic microwave background (CMB) experiments are primarily targeting a detection of the primordial $B$-mode polarisation. The faintness of this signal requires exquisite control of systematic effects which may bias the measurements. In this work, we derive requirements on the relative calibration accuracy of the overall polarisation gain ($\Delta g_\nu$) for LiteBIRD experiment, through the application of the blind Needlet Internal Linear Combination (NILC) foreground-cleaning method. We find that minimum variance techniques, as NILC, are less affected by gain calibration uncertainties than a parametric approach, which requires a proper modelling of these instrumental effects. The tightest constraints are obtained for frequency channels where the CMB signal is relatively brighter (166 GHz channel, $\Delta {g}_\nu \approx 0.16 \%$), while, with a parametric approach, the strictest requirements were on foreground-dominated channels. We then propagate gain calibration uncertainties, corresponding to the derived requirements, into all frequency channels simultaneously. We find that the overall impact on the estimated $r$ is lower than the required budget for LiteBIRD by almost a factor $5$. The adopted procedure to derive requirements assumes a simple Galactic model. We therefore assess the robustness of obtained results against more realistic scenarios by injecting the gain calibration uncertainties, according to the requirements, into LiteBIRD simulated maps and assuming intermediate- and high-complexity sky models. In this case, we employ the so-called Multi-Clustering NILC (MC-NILC) foreground-cleaning pipeline and obtain that the impact of gain calibration uncertainties on $r$ is lower than the LiteBIRD gain systematics budget for the intermediate-complexity sky model. For the high-complexity case, instead, it would be necessary to tighten the requirements by a factor $1.8$., Comment: 29 pages, 11 figures

The nature of the gravitational wave background (GWB) is a key question in modern astrophysics and cosmology, with significant implications for understanding of the structure and evolution of the Universe. We demonstrate how cross-correlating large-scale structure (LSS) tracers with the GWB spatial anisotropies can extract a clear astrophysical imprint from the GWB signal. Focusing on the unresolved population of supermassive black hole binaries (SMBHBs) as the primary source for the GWB at nanohertz frequencies, we construct full-sky maps of galaxy distributions and characteristic strain of the GWB to explore the relationship between GWB anisotropies and the LSS. We find that at current pulsar timing array (PTA) sensitivities, very few loud SMBHBs act as Poisson-like noise. This results in anisotropies dominated by a small number of sources, making GWB maps where SMBHBs trace the LSS indistinguishable from a GWBs from a uniform distribution of SMBHBs. In contrast, we find that the bulk of the unresolved SMBHBs produce anisotropies which mirror the spatial distribution of galaxies, and thus trace the LSS. Importantly, we show that cross-correlations are required to retrieve a clear LSS imprint in the GWB. Specifically, we find this LSS signature can me measured at a $3\sigma$ level in near-future PTA experiments that probe angular scales of $\ell_{\text{max}} \geq 42$, and $5\sigma$ for $\ell_{\text{max}} \geq 72$. Our approach opens new avenues to employ the GWB as an LSS tracer, providing unique insights into SMBHB population models and the nature of the GWB itself. Our results motivate further exploration of potential synergies between next-generation PTA experiments and cosmological tracers of the LSS., Comment: 9 pages, 7 figures. Submitted

By confining colloidal active fluids in microchannel networks, we demonstrate that their degenerate flows corresponds to the configurations of the six-vertex model. We use this quantitative correspondence to control and explain the active flows that emerge in square grid networks. In particular, we show that the Lagrangian trajectories of active particles realize the Baxter-Kelland-Wu mapping and form completely packed loops, whose geometry can be exactly predicted and explained. We then go beyond the square-grid geometry and introduce a general framework for predicting the geometry of active-hydraulic flows in arbitrary networks.

Recently, possible hints of parity violation have been observed in the connected galaxy four-point correlation function. Although the true origin of the signal from the analysis has been debated, should they have a physical origin, they might point to primordial non-Gaussianity and would be evidence of new physics. In this work, we examine the single-field slow-roll model of inflation within chiral scalar-tensor theories of modified gravity. These theories extend the Chern-Simons one by including parity-violating operators containing first and second derivatives of the non-minimally coupled scalar (inflaton) field. This model is capable of imprinting parity-violating signatures in late-time observables, such as the galaxy four-point correlation function. We perform an analysis of the graviton-mediated scalar trispectrum of the gauge-invariant curvature perturbation $\zeta(t,\mathbf{x})$. We estimate that for a set of parameters of the theory it is possible to produce a signal-to-noise ratio for the parity-violating part of the trispectrum of order one without introducing modifications to the single-field slow-roll setup. Even if the signal found in the analysis turns out to be spurious or if no parity violation is ever detected in the galaxy four-point correlation function, our analysis can be used to constrain the free parameters of these theories., Comment: 30 pages, 4 figures

Direct Alignment Algorithms (DAAs), such as Direct Preference Optimisation (DPO) and Identity Preference Optimisation (IPO), have emerged as alternatives to online Reinforcement Learning from Human Feedback (RLHF) algorithms such as Proximal Policy Optimisation (PPO) for aligning language models to human preferences, without the need for explicit reward modelling. These methods generally aim to increase the likelihood of generating better (preferred) completions while discouraging worse (non-preferred) ones, while staying close to the original model's behaviour. In this work, we explore the relationship between completion likelihood and model performance in state-of-the-art DAAs, and identify a critical issue of likelihood over-optimisation. Contrary to expectations, we find that higher likelihood of better completions and larger margins between better and worse completion likelihoods do not necessarily lead to better performance, and may even degrade it. Our analysis reveals that while higher likelihood correlates with better memorisation of factual knowledge patterns, a slightly lower completion likelihood tends to improve output diversity, thus leading to better generalisation to unseen scenarios. Moreover, we identify two key indicators that signal when over-optimised output diversity begins to harm performance: Decreasing Entropy over Top-k Tokens and Diminishing Top-k Probability Mass. Our experimental results validate that these indicators are reliable signs of declining performance under different regularisations, helping prevent over-optimisation and improve alignment with human preferences., Comment: Preprint Version

Reduced power series in two variables with coefficients in a field of characteristic zero satisfy a well-known formula that relates a codimension related to the normalization of a ring and the jacobian ideal. In the general case Deligne proved that this formula is only an inequality; Garc\'ia Barroso and P{\l}oski stated a conjecture for irreducible power series. In this work we generalize Kouchnirenko's formula for any degenerated power series and also generalize Garc\'ia Barroso and P{\l}oski's conjecture. We prove the conjecture in some cases using in particular Greuel and Nguyen.

Optical Character Recognition (OCR) continues to face accuracy challenges that impact subsequent applications. To address these errors, we explore the utility of OCR confidence scores for enhancing post-OCR error detection. Our study involves analyzing the correlation between confidence scores and error rates across different OCR systems. We develop ConfBERT, a BERT-based model that incorporates OCR confidence scores into token embeddings and offers an optional pre-training phase for noise adjustment. Our experimental results demonstrate that integrating OCR confidence scores can enhance error detection capabilities. This work underscores the importance of OCR confidence scores in improving detection accuracy and reveals substantial disparities in performance between commercial and open-source OCR technologies.

In the fields of security systems, forensic investigations, and personalized services, the importance of speech as a fundamental human input outweighs text-based interactions. This research delves deeply into the complex field of Speaker Identification (SID), examining its essential components and emphasising Mel Spectrogram and Mel Frequency Cepstral Coefficients (MFCC) for feature extraction. Moreover, this study evaluates six slightly distinct model architectures using extensive analysis to evaluate their performance, with hyperparameter tuning applied to the best-performing model. This work performs a linguistic analysis to verify accent and gender accuracy, in addition to bias evaluation within the AB-1 Corpus dataset., Comment: Resultant work from Assignment, Department of AI, University of Malta. Code available at: https://github.com/mbar0075/Speech-Technology

With the ever-growing variety of object detection approaches, this study explores a series of experiments that combine reinforcement learning (RL)-based visual attention methods with saliency ranking techniques to investigate transparent and sustainable solutions. By integrating saliency ranking for initial bounding box prediction and subsequently applying RL techniques to refine these predictions through a finite set of actions over multiple time steps, this study aims to enhance RL object detection accuracy. Presented as a series of experiments, this research investigates the use of various image feature extraction methods and explores diverse Deep Q-Network (DQN) architectural variations for deep reinforcement learning-based localisation agent training. Additionally, we focus on optimising the detection pipeline at every step by prioritising lightweight and faster models, while also incorporating the capability to classify detected objects, a feature absent in previous RL approaches. We show that by evaluating the performance of these trained agents using the Pascal VOC 2007 dataset, faster and more optimised models were developed. Notably, the best mean Average Precision (mAP) achieved in this study was 51.4, surpassing benchmarks set by RL-based single object detectors in the literature., Comment: Resultant work from Dissertation, Department of AI, University of Malta. Code available at: https://github.com/mbar0075/SaRLVision

The primordial power spectrum of curvature perturbations has been well-measured on large scales but remains fairly unconstrained at smaller scales, where significant deviations from $\Lambda$CDM may occur. Measurements of 21-cm intensity mapping in the dark ages promise to access very small scales that have yet to be probed, extending beyond the reach of CMB and galaxy surveys. In this paper, we investigate how small-scale power-law enhancements -- or blue tilts -- of the primordial power spectrum affect the 21-cm power spectrum. We consider generic enhancements due to curvature modes, isocurvature modes, and runnings of the spectral tilt. We present forecasts for Earth- and lunar-based instruments to detect a blue-tilted primordial spectrum. We find that an Earth-based instrument capable of reaching the dark ages could detect any enhancements of power on nearly all the scales it can observe, which depends on the baseline of the interferometer. The smallest scales observed by such an instrument can only detect a very strong enhancement. However, an instrument on the far side of the Moon of the same size would be able to probe shallower slopes with higher precision. We forecast results for instruments with $100 \, {\rm km} \, (3000 \, {\rm km})$ baselines and find that they can probe up to scales of order $k_{\rm max} \sim 8 \, {\rm Mpc}^{-1} \, (k_{\rm max} \sim 250 \, {\rm Mpc}^{-1})$, thereby providing invaluable information on exotic physics and testing inflationary models on scales not otherwise accessible., Comment: 12 pages, 6 figures

Large angular scale surveys in the absence of atmosphere are essential for measuring the primordial $B$-mode power spectrum of the Cosmic Microwave Background (CMB). Since this proposed measurement is about three to four orders of magnitude fainter than the temperature anisotropies of the CMB, in-flight calibration of the instruments and active suppression of systematic effects are crucial. We investigate the effect of changing the parameters of the scanning strategy on the in-flight calibration effectiveness, the suppression of the systematic effects themselves, and the ability to distinguish systematic effects by null-tests. Next-generation missions such as LiteBIRD, modulated by a Half-Wave Plate (HWP), will be able to observe polarisation using a single detector, eliminating the need to combine several detectors to measure polarisation, as done in many previous experiments and hence avoiding the consequent systematic effects. While the HWP is expected to suppress many systematic effects, some of them will remain. We use an analytical approach to comprehensively address the mitigation of these systematic effects and identify the characteristics of scanning strategies that are the most effective for implementing a variety of calibration strategies in the multi-dimensional space of common spacecraft scan parameters. We also present Falcons, a fast spacecraft scanning simulator that we developed to investigate this scanning parameter space.

The theory of inflation provides a mechanism to explain the structures we observe today in the Universe, starting from quantum-mechanically generated fluctuations. However, this leaves the question of: how did the quantum-to-classical transition, occur? During inflation, tensor perturbations interact (at least gravitationally) with other fields, meaning that we need to view these perturbations as an open system that interacts with an environment. In this paper, the evolution of the system is described using a Lindblad equation, which describes the quantum decoherence of the system. This is a possible mechanism for explaining the quantum-to-classical transition. We show that this quantum decoherence leads to a scale-dependent increase of the gravitational wave power spectrum, depending on the strength and time dependence of the interaction between the system and the environment. By using current upper bounds on the gravitational wave power spectrum from inflation, obtained from CMB and the LIGO-Virgo-KAGRA constraints, we find an upper bound on the interaction strength. Furthermore, we compute the decoherence criterion, which indicates the minimal interaction strength needed for a specific scale to have successfully decohered by the end of inflation. Assuming that the CMB modes have completely decohered, we indicate a lower bound on the interaction strength. In addition, this decoherence criterion allows us to look at which scales might not have fully decohered and could still show some relic quantum signatures. Lastly, we use sensitivity forecasts to study how future gravitational-wave detectors, such as LISA and ET, could constrain the decoherence parameter space. Due to the scale-dependence of the power spectrum, LISA could only have a very small impact. However, ET will be able to significantly improve our current constraints for specific decoherence scenarios., Comment: 34 pages, 9 figures. Matches version published in JCAP. arXiv admin note: text overlap with arXiv:1801.09949 by other authors

We assess the capabilities of the LiteBIRD mission to map the hot gas distribution in the Universe through the thermal Sunyaev-Zeldovich (SZ) effect. Our analysis relies on comprehensive simulations incorporating various sources of Galactic and extragalactic foreground emission, while accounting for specific instrumental characteristics of LiteBIRD, such as detector sensitivities, frequency-dependent beam convolution, inhomogeneous sky scanning, and $1/f$ noise. We implement a tailored component-separation pipeline to map the thermal SZ Compton $y$-parameter over 98% of the sky. Despite lower angular resolution for galaxy cluster science, LiteBIRD provides full-sky coverage and, compared to the Planck satellite, enhanced sensitivity, as well as more frequency bands to enable the construction of an all-sky $y$-map, with reduced foreground contamination at large and intermediate angular scales. By combining LiteBIRD and Planck channels in the component-separation pipeline, we obtain an optimal $y$-map that leverages the advantages of both experiments, with the higher angular resolution of the Planck channels enabling the recovery of compact clusters beyond the LiteBIRD beam limitations, and the numerous sensitive LiteBIRD channels further mitigating foregrounds. The added value of LiteBIRD is highlighted through the examination of maps, power spectra, and one-point statistics of the various sky components. After component separation, the $1/f$ noise from LiteBIRD is effectively mitigated below the thermal SZ signal at all multipoles. Cosmological constraints on $S_8=\sigma_8\left(\Omega_{\rm m}/0.3\right)^{0.5}$ obtained from the LiteBIRD-Planck combined $y$-map power spectrum exhibits a 15% reduction in uncertainty compared to constraints from Planck alone. This improvement can be attributed to the increased portion of uncontaminated sky available in the LiteBIRD-Planck combined $y$-map., Comment: 38 pages, 13 figures, abstract shortened. Updated to match version accepted by JCAP

Unveiling the dark sector of the Universe is one of the leading efforts in theoretical physics. Among the many models proposed, axions and axion-like particles stand out due to their solid theoretical foundation, capacity to contribute significantly to both dark matter and dark energy, and potential to address the small-scale crisis of $\Lambda$CDM. Moreover, these pseudo-scalar fields couple to the electromagnetic sector through a Chern-Simons parity-violating term, leading to a rotation of the plane of linearly polarized waves, namely cosmic birefringence. We explore the impact of the axion-parameters on anisotropic birefringence and study, for the first time, its cross-correlation with the spatial distribution of galaxies, focusing on ultralight axions with masses $10^{-33}\,{\rm eV}\le m_\phi\le10^{-28}\,{\rm eV}$. Through this novel approach, we investigate the axion-parameter space in the mass $m_\phi$ and initial misalignment angle $\theta_i$, within the framework of early dark energy models, and constrain the axion-photon coupling $g_{\phi\gamma}$ required to achieve unity in the signal-to-noise ratio of the underlying cross-correlation, computed with the instrument specifications of Euclid and forthcoming CMB-polarization data. Our findings reveal that for masses below $10^{-32}\,{\rm eV}$ and initial misalignment angles greater in absolute value than $\pi/4$, the signal-to-noise ratio not only exceeds unity but also surpasses that achievable from the auto-correlation of birefringence alone (up to a factor 7), highlighting the informative potential of this new probe. Additionally, given the late-time evolution of these low-mass axions, the signal stems from the epoch of reionization, providing an excellent tool to single out the birefringence generated during this period., Comment: 19 + 5 pages + references, 15 figures

LiteBIRD, the next-generation cosmic microwave background (CMB) experiment, aims for a launch in Japan's fiscal year 2032, marking a major advancement in the exploration of primordial cosmology and fundamental physics. Orbiting the Sun-Earth Lagrangian point L2, this JAXA-led strategic L-class mission will conduct a comprehensive mapping of the CMB polarization across the entire sky. During its 3-year mission, LiteBIRD will employ three telescopes within 15 unique frequency bands (ranging from 34 through 448 GHz), targeting a sensitivity of 2.2\,$\mu$K-arcmin and a resolution of 0.5$^\circ$ at 100\,GHz. Its primary goal is to measure the tensor-to-scalar ratio $r$ with an uncertainty $\delta r = 0.001$, including systematic errors and margin. If $r \geq 0.01$, LiteBIRD expects to achieve a $>5\sigma$ detection in the $\ell=$2-10 and $\ell=$11-200 ranges separately, providing crucial insight into the early Universe. We describe LiteBIRD's scientific objectives, the application of systems engineering to mission requirements, the anticipated scientific impact, and the operations and scanning strategies vital to minimizing systematic effects. We will also highlight LiteBIRD's synergies with concurrent CMB projects., Comment: 23 pages, 9 figures, 1 table, SPIE Astronomical Telescopes + Instrumentation 2024

Reward models (RMs) play a critical role in aligning language models through the process of reinforcement learning from human feedback. RMs are trained to predict a score reflecting human preference, which requires significant time and cost for human annotation. Additionally, RMs tend to quickly overfit on superficial features in the training set, hindering their generalization performance on unseen distributions. We propose a novel approach using synthetic natural language critiques generated by large language models to provide additional feedback, evaluating aspects such as instruction following, correctness, and style. This offers richer signals and more robust features for RMs to assess and score on. We demonstrate that high-quality critiques improve the performance and data efficiency of RMs initialized from different pretrained models, reducing the reliance on costly human annotations. Furthermore, incorporating critiques improves both the interpretability and robustness of RM training.

When gravitational waves travel from their source to an observer, they interact with matter structures along their path, causing distinct deformations in their waveforms. In this study we introduce a novel theoretical framework for wave optics effects in gravitational lensing, addressing the limitations of existing approaches. We achieve this by incorporating the proper time technique, typically used in field theory studies, into gravitational lensing. This approach allows us to extend the standard formalism beyond the eikonal and paraxial approximations, which are traditionally assumed, and to account for polarization effects, which are typically neglected in the literature. We demonstrate that our method provides a robust generalization of conventional approaches, including them as special cases. Our findings enhance our understanding of gravitational wave propagation, which is crucial for accurately interpreting gravitational wave observations and extracting unbiased information about the lenses from the gravitational wave waveforms., Comment: 31 pages plus technical appendices, 4 figures

This technical report introduces Aya 23, a family of multilingual language models. Aya 23 builds on the recent release of the Aya model (\"Ust\"un et al., 2024), focusing on pairing a highly performant pre-trained model with the recently released Aya collection (Singh et al., 2024). The result is a powerful multilingual large language model serving 23 languages, expanding state-of-art language modeling capabilities to approximately half of the world's population. The Aya model covered 101 languages whereas Aya 23 is an experiment in depth vs breadth, exploring the impact of allocating more capacity to fewer languages that are included during pre-training. Aya 23 outperforms both previous massively multilingual models like Aya 101 for the languages it covers, as well as widely used models like Gemma, Mistral and Mixtral on an extensive range of discriminative and generative tasks. We release the open weights for both the 8B and 35B models as part of our continued commitment for expanding access to multilingual progress.

The current standard model of cosmology successfully describes a variety of measurements, but the nature of its main ingredients, dark matter and dark energy, remains unknown. Euclid is a medium-class mission in the Cosmic Vision 2015-2025 programme of the European Space Agency (ESA) that will provide high-resolution optical imaging, as well as near-infrared imaging and spectroscopy, over about 14,000 deg^2 of extragalactic sky. In addition to accurate weak lensing and clustering measurements that probe structure formation over half of the age of the Universe, its primary probes for cosmology, these exquisite data will enable a wide range of science. This paper provides a high-level overview of the mission, summarising the survey characteristics, the various data-processing steps, and data products. We also highlight the main science objectives and expected performance., Comment: Accepted for publication in the A&A special issue`Euclid on Sky'

The disconnect between tokenizer creation and model training in language models allows for specific inputs, such as the infamous SolidGoldMagikarp token, to induce unwanted model behaviour. Although such `glitch tokens', tokens present in the tokenizer vocabulary but that are nearly or entirely absent during model training, have been observed across various models, a reliable method to identify and address them has been missing. We present a comprehensive analysis of Large Language Model tokenizers, specifically targeting this issue of detecting under-trained tokens. Through a combination of tokenizer analysis, model weight-based indicators, and prompting techniques, we develop novel and effective methods for automatically detecting these problematic tokens. Our findings demonstrate the prevalence of such tokens across a diverse set of models and provide insights into improving the efficiency and safety of language models., Comment: 16 pages, 6 figures. Accepted at EMNLP 2024, main track. For associated code, see https://github.com/cohere-ai/magikarp/

We present a novel model-driven approach for testing RESTful applications. We introduce a (i) domain-specific language for OpenAPI specifications and (ii) a tool to support our methodology. Our DSL is inspired by session types and enables the modelling of communication protocols between a REST client and server. Our tool, dubbed COTS, generates (randomised) model-based test executions and reports software defects. We evaluate the effectiveness of our approach by applying it to test several open source applications. Our findings indicate that our methodology can identify nuanced defects in REST APIs and achieve comparable or superior code coverage when compared to much larger handcrafted test suites., Comment: This preprint is the author version of the paper published at COORDINATION 2024

Human feedback is central to the alignment of Large Language Models (LLMs). However, open questions remain about methods (how), domains (where), people (who) and objectives (to what end) of feedback processes. To navigate these questions, we introduce PRISM, a dataset that maps the sociodemographics and stated preferences of 1,500 diverse participants from 75 countries, to their contextual preferences and fine-grained feedback in 8,011 live conversations with 21 LLMs. With PRISM, we contribute (i) wider geographic and demographic participation in feedback; (ii) census-representative samples for two countries (UK, US); and (iii) individualised ratings that link to detailed participant profiles, permitting personalisation and attribution of sample artefacts. We target subjective and multicultural perspectives on value-laden and controversial issues, where we expect interpersonal and cross-cultural disagreement. We use PRISM in three case studies to demonstrate the need for careful consideration of which humans provide what alignment data.

We present a data-driven pipeline for model building that combines interpretable machine learning, hydrodynamic theories, and microscopic models. The goal is to uncover the underlying processes governing nonlinear dynamics experiments. We exemplify our method with data from microfluidic experiments where crystals of streaming droplets support the propagation of nonlinear waves absent in passive crystals. By combining physics-inspired neural networks, known as neural operators, with symbolic regression tools, we generate the solution, as well as the mathematical form, of a nonlinear dynamical system that accurately models the experimental data. Finally, we interpret this continuum model from fundamental physics principles. Informed by machine learning, we coarse grain a microscopic model of interacting droplets and discover that non-reciprocal hydrodynamic interactions stabilise and promote nonlinear wave propagation., Comment: Main text: 6 pages, 4 figures. Supplement: 11 pages, 4 figures

This paper introduces v0.5 of the AI Safety Benchmark, which has been created by the MLCommons AI Safety Working Group. The AI Safety Benchmark has been designed to assess the safety risks of AI systems that use chat-tuned language models. We introduce a principled approach to specifying and constructing the benchmark, which for v0.5 covers only a single use case (an adult chatting to a general-purpose assistant in English), and a limited set of personas (i.e., typical users, malicious users, and vulnerable users). We created a new taxonomy of 13 hazard categories, of which 7 have tests in the v0.5 benchmark. We plan to release version 1.0 of the AI Safety Benchmark by the end of 2024. The v1.0 benchmark will provide meaningful insights into the safety of AI systems. However, the v0.5 benchmark should not be used to assess the safety of AI systems. We have sought to fully document the limitations, flaws, and challenges of v0.5. This release of v0.5 of the AI Safety Benchmark includes (1) a principled approach to specifying and constructing the benchmark, which comprises use cases, types of systems under test (SUTs), language and context, personas, tests, and test items; (2) a taxonomy of 13 hazard categories with definitions and subcategories; (3) tests for seven of the hazard categories, each comprising a unique set of test items, i.e., prompts. There are 43,090 test items in total, which we created with templates; (4) a grading system for AI systems against the benchmark; (5) an openly available platform, and downloadable tool, called ModelBench that can be used to evaluate the safety of AI systems on the benchmark; (6) an example evaluation report which benchmarks the performance of over a dozen openly available chat-tuned language models; (7) a test specification for the benchmark.

The goal of this work is to study the smoothings of singular coaxial intersections of ellipsoids (where coaxial includes concentric) with generic singularities, with special attention to the 3-dimensional case.

We present detailed forecasts for the constraints on primordial magnetic fields (PMFs) that will be obtained with the LiteBIRD satellite. The constraints are driven by the effects of PMFs on the CMB anisotropies: the gravitational effects of magnetically-induced perturbations; the effects on the thermal and ionization history of the Universe; the Faraday rotation imprint on the CMB polarization; and the non-Gaussianities induced in polarization anisotropies. LiteBIRD represents a sensitive probe for PMFs and by exploiting all the physical effects, it will be able to improve the current limit coming from Planck. In particular, thanks to its accurate $B$-mode polarization measurement, LiteBIRD will improve the constraints on infrared configurations for the gravitational effect, giving $B_{\rm 1\,Mpc}^{n_{\rm B} =-2.9} < 0.8$ nG at 95% C.L., potentially opening the possibility to detect nanogauss fields with high significance. We also observe a significant improvement in the limits when marginalized over the spectral index, $B_{1\,{\rm Mpc}}^{\rm marg}< 2.2$ nG at 95% C.L. From the thermal history effect, which relies mainly on $E$-mode polarization data, we obtain a significant improvement for all PMF configurations, with the marginalized case, $\sqrt{\langle B^2\rangle}^{\rm marg}<0.50$ nG at 95% C.L. Faraday rotation constraints will take advantage of the wide frequency coverage of LiteBIRD and the high sensitivity in $B$ modes, improving the limits by orders of magnitude with respect to current results, $B_{1\,{\rm Mpc}}^{n_{\rm B} =-2.9} < 3.2$ nG at 95% C.L. Finally, non-Gaussianities of the $B$-mode polarization can probe PMFs at the level of 1 nG, again significantly improving the current bounds from Planck. Altogether our forecasts represent a broad collection of complementary probes, providing conservative limits on PMF characteristics that will be achieved with LiteBIRD., Comment: 51 pages, 24 figures, abstract shortened

Primordial tensor modes can induce Cosmic Microwave Background spectral distortions during horizon re-entry. We investigate a specific mechanism proposed for this purpose, characterized by the coupling of an SU(2) gauge field to an axion undergoing a momentary stage of rapid evolution during inflation. Examining also the scalar perturbations produced by this model, we find that spectral distortions from the scalar modes significantly dominate those arising from the tensors. This holds true also for an earlier version of the model based on a U(1) gauge field. The scalar-induced distortions might be observed in future experiments, and the current COBE/FIRAS constraints already limit the parameter space of these models. Additionally, we find that delaying the onset of fast roll in the SU(2) scenario (to enhance the modes at the scales relevant for spectral distortions, while respecting the CMB constraints at larger scales) poses a greater challenge compared to the U(1) case. We propose a way to control the axion speed by varying the size of its coupling to the gauge fields.

We demonstrate the existence of a multiplicity of co-existing frequency combs in a harmonically mode-locked laser that we link to the splay phases of the Kuramoto model with short range interactions. These splay states are multistable and the laser may wander between them under the influence of stochastic forces. Consequently, the many pulses circulating in the cavity are not necessarily coherent with each other. We show that this partially disordered state for the phase of the optical field features regular train of pulses in the field intensity, a state that we term an incoherent crystal of optical pulses. We provide evidence that the notion of coherence should be interpreted by comparing the duration of the measurement time with the Kramers' escape time of each splay state. Our results are confirmed experimentally by studying a passively mode-locked vertical external-cavity surface-emitting laser., Comment: Main manuscript (5 pages + 5 figures) & Supplementary material (10 pages + 6 figures)

Many different techniques to analyze galaxy clustering data and obtain cosmological constraints have been proposed, tested and used. Given the large amount of data that will be available soon, it is worth investigating new observables and ways to extract information from such datasets. In this paper, we focus on antisymmetric correlations, that arise in the cross-correlation of different galaxy populations when the small-scale power spectrum is modulated by a long-wavelength field. In $\Lambda$CDM this happens because of nonlinear clustering of sources that trace the underlying matter distribution in different ways. Beyond the standard model, this observable is sourced naturally in various new physics scenarios. We derive, for the first time, its complete expression up to second order in redshift space, and show that this improves detectability compared to previous evaluations at first order in real space. Moreover, we explore a few potential applications to use this observable to detect models with vector modes, or where different types of sources respond in different ways to the underlying modulating long mode, and anisotropic models with privileged directions in the sky. This shows how antisymmetric correlations can be a useful tool for testing exotic cosmological models.

With the rise of text-to-image (T2I) generative AI models reaching wide audiences, it is critical to evaluate model robustness against non-obvious attacks to mitigate the generation of offensive images. By focusing on ``implicitly adversarial'' prompts (those that trigger T2I models to generate unsafe images for non-obvious reasons), we isolate a set of difficult safety issues that human creativity is well-suited to uncover. To this end, we built the Adversarial Nibbler Challenge, a red-teaming methodology for crowdsourcing a diverse set of implicitly adversarial prompts. We have assembled a suite of state-of-the-art T2I models, employed a simple user interface to identify and annotate harms, and engaged diverse populations to capture long-tail safety issues that may be overlooked in standard testing. The challenge is run in consecutive rounds to enable a sustained discovery and analysis of safety pitfalls in T2I models. In this paper, we present an in-depth account of our methodology, a systematic study of novel attack strategies and discussion of safety failures revealed by challenge participants. We also release a companion visualization tool for easy exploration and derivation of insights from the dataset. The first challenge round resulted in over 10k prompt-image pairs with machine annotations for safety. A subset of 1.5k samples contains rich human annotations of harm types and attack styles. We find that 14% of images that humans consider harmful are mislabeled as ``safe'' by machines. We have identified new attack strategies that highlight the complexity of ensuring T2I model robustness. Our findings emphasize the necessity of continual auditing and adaptation as new vulnerabilities emerge. We are confident that this work will enable proactive, iterative safety assessments and promote responsible development of T2I models., Comment: 10 pages, 6 figures

Datasets are foundational to many breakthroughs in modern artificial intelligence. Many recent achievements in the space of natural language processing (NLP) can be attributed to the finetuning of pre-trained models on a diverse set of tasks that enables a large language model (LLM) to respond to instructions. Instruction fine-tuning (IFT) requires specifically constructed and annotated datasets. However, existing datasets are almost all in the English language. In this work, our primary goal is to bridge the language gap by building a human-curated instruction-following dataset spanning 65 languages. We worked with fluent speakers of languages from around the world to collect natural instances of instructions and completions. Furthermore, we create the most extensive multilingual collection to date, comprising 513 million instances through templating and translating existing datasets across 114 languages. In total, we contribute four key resources: we develop and open-source the Aya Annotation Platform, the Aya Dataset, the Aya Collection, and the Aya Evaluation Suite. The Aya initiative also serves as a valuable case study in participatory research, involving collaborators from 119 countries. We see this as a valuable framework for future research collaborations that aim to bridge gaps in resources.

We show that nearly degenerate Vertical External-Cavity Surface-Emitting Lasers emit tilted beams of time localized structures, i.e. mode-locked light pulses which can be individually addressed. These beams feature a Gaussian profile and they are emitted in pairs with opposite transverse k-vector. Because they are phase locked, their interference leads to a non homotetic pattern in the near-field emission of the laser. When a single pair is emitted this is a stripe pattern. Our analysis discloses the role of spherical aberrations of the cavity in stabilizing this spatio-temporal mode-locked regime and in selecting the value of the transverse wavevector.

Cosmic Birefringence (CB) is a phenomenon in which the polarization of the Cosmic Microwave Background (CMB) radiation is rotated as it travels through space due to the coupling between photons and an axion-like field. We look for a solution able to explain the result obtained from the \textit{Planck} Public Release 4 (PR4), which has provided a hint of detection of the CB angle, $\alpha=(0.30\pm0.11)^{\circ}$. In addition to the solutions, already present in the literature, which need a non-negligible evolution in time of the axion-like field during recombination, we find a new region of the parameter space which allows for a nearly constant time evolution of such a field in the same epoch. The latter reinforces the possibility to employ the commonly used relations connecting the observed CMB spectra with the unrotated ones, through trigonometric functions of the CB angle. However, if the homogeneous axion field sourcing isotropic birefringence is almost constant in time during the matter-dominated era, this does not automatically implies that the same holds true also for the associated inhomogeneous perturbations. For this reason, in this paper we present a full generalized Boltzmann treatment of this phenomenon, that is able, for the first time to our knowledge to deal with the time evolution of anisotropic cosmic birefringence (ACB). We employ this approach to provide predictions of ACB, in particular for the set of best-fit parameters found in the new solution of the isotropic case. If the latter is the correct model, we expect an ACB spectrum of the order of $(10^{-15}\div10^{-32})$ deg$^2$ for the auto-correlation, and $(10^{-7}\div10^{-17})$ $\mu $K$\cdot\,$deg for the cross-correlations with the CMB $T$ and $E$ fields, depending on the angular scale., Comment: 20 pages, 5 figures. It matches the published version

We present a study of the impact of an uncertainty in the beam far side-lobe knowledge on the measurement of the Cosmic Microwave Background $B$-mode signal at large scale. It is expected to be one of the main source of systematic effects in future CMB observations. Because it is crucial for all-sky survey missions to take into account the interplays between beam systematic effects and all the data analysis steps, the primary goal of this paper is to provide the methodology to carry out the end-to-end study of their effect for a space-borne CMB polarization experiment, up to the cosmological results in the form of a bias $\delta r$ on the tensor-to-scalar ratio $r$. LiteBIRD is dedicated to target the measurement of CMB primordial $B$ modes by reaching a sensitivity of $\sigma \left( r \right) \leq 10^{-3}$ assuming $r=0$. As a demonstration of our framework, we derive the relationship between the knowledge of the beam far side-lobes and the tentatively allocated error budget under given assumptions on design, simulation and component separation method. We assume no mitigation of the far side-lobes effect at any stage of the analysis pipeline. We show that $\delta r$ is mostly due to the integrated fractional power difference between the estimated beams and the true beams in the far side-lobes region, with little dependence on the actual shape of the beams, for low enough $\delta r$. Under our set of assumptions, in particular considering the specific foreground cleaning method we used, we find that the integrated fractional power in the far side-lobes should be known at a level as tight as $\sim 10^{-4}$, to achieve the required limit on the bias $\delta r < 1.9 \times 10^{-5}$. The framework and tools developed for this study can be easily adapted to provide requirements under different design, data analysis frameworks and for other future space-borne experiments beyond LiteBIRD.

We study the existence of some irreducible projective plane curves of degree~$8$ with some prescribed topological type of singularities in the algebraic and symplectic worlds., Comment: Some minor corrections

We estimate the efficiency of mitigating the lensing $B$-mode polarization, the so-called delensing, for the $LiteBIRD$ experiment with multiple external data sets of lensing-mass tracers. The current best bound on the tensor-to-scalar ratio, $r$, is limited by lensing rather than Galactic foregrounds. Delensing will be a critical step to improve sensitivity to $r$ as measurements of $r$ become more and more limited by lensing. In this paper, we extend the analysis of the recent $LiteBIRD$ forecast paper to include multiple mass tracers, i.e., the CMB lensing maps from $LiteBIRD$ and CMB-S4-like experiment, cosmic infrared background, and galaxy number density from $Euclid$- and LSST-like survey. We find that multi-tracer delensing will further improve the constraint on $r$ by about $20\%$. In $LiteBIRD$, the residual Galactic foregrounds also significantly contribute to uncertainties of the $B$-modes, and delensing becomes more important if the residual foregrounds are further reduced by an improved component separation method., Comment: 21 pages, 7 figures