Search

We study fuzzy axion dark matter in type IIB string theory, for axions descending from the Ramond-Ramond four-form in compactifications on orientifolds of Calabi-Yau hypersurfaces. Such models can be tested by cosmological measurements if a significant relic abundance of fuzzy dark matter arises, which we argue is most common in models with small numbers of axions. We construct a topologically exhaustive ensemble of more than 350,000 Calabi-Yau compactifications yielding up to seven axions, and in this setting we perform a systematic analysis of misalignment production of fuzzy dark matter. In typical regions of moduli space, the fuzzy axion, the QCD axion, and other axions have comparable decay constants of $f_a \approx 10^{16}$ GeV. We find that overproduction of heavier axions is problematic, except at special loci in moduli space where decay constant hierarchies can occur: without a contrived reheating epoch, it is necessary to fine-tune initial displacements. The resulting dark matter is typically a mix of fuzzy axions and heavier axions, including the QCD axion. Dark photons are typically present as a consequence of the orientifold projection. We examine the signatures of these models by simulating halos with multiple fuzzy axions, and by computing new cosmological constraints on ultralight axions and dark radiation. We also give evidence that cosmic birefringence is possible in this setting. Our findings determine the phenomenological correlates of fuzzy axion dark matter in a corner of the landscape., Comment: 62 pages, 16 figures. Supplementary materials: https://github.com/sheride/fuzzy_axions

Minority languages are vital to preserving cultural heritage, yet they face growing risks of extinction due to limited digital resources and the dominance of artificial intelligence models trained on high-resource languages. This white paper proposes a framework to generate linguistic tools for low-resource languages, focusing on data creation to support the development of language models that can aid in preservation efforts. Sardinian, an endangered language, serves as the case study to demonstrate the framework's effectiveness. By addressing the data scarcity that hinders intelligent applications for such languages, we contribute to promoting linguistic diversity and support ongoing efforts in language standardization and revitalization through modern technologies.

Several methods have been developed to improve the predictions of density functional theory (DFT) in the case of strongly correlated electron systems. Out of these approaches, DFT+$U$, which corresponds to a static treatment of the local interaction, and DFT combined with dynamical mean field theory (DFT+DMFT), which considers local fluctuations, have both proven incredibly valuable in tackling the description of materials with strong local electron-electron interactions. While it is in principle known that the Hartree-Fock (HF) limit of the DFT+DMFT approach should recover DFT+$U$, demonstrating this equivalence in practice is challenging, due to the very different ways in which the two approaches are generally implemented. In this work, we introduce a way to perform DFT+$U$ calculations in Quantum ESPRESSO using Wannier functions as calculated by Wannier90, which allows us to use the same Hubbard projector functions both in DFT+$U$ and in DFT+DMFT. We benchmark these DFT+$U$ calculations against DFT+DMFT calculations where the DMFT impurity problem is solved within the HF approximation. Considering a number of prototypical materials including NiO, MnO, LaMnO$_3$, and LuNiO$_3$, we establish the sameness of the two approaches. Finally, we showcase the versatility of our approach by going beyond the commonly used atomic orbital-like projectors by performing DFT+$U$ calculations for VO$_2$ using a special set of bond-centered Wannier functions.

We study Higgs branches of field theories with 8 supercharges in 5 and 6 dimensions, focusing on theories realised on 5-brane webs in Type IIB with an O$7^+$ plane, or a D6-D8-NS5 brane system in Type IIA in the presence of an O$8^+$ plane. We find magnetic quivers for the Higgs branches of these theories. The main consequence of the presence of the orientifold is that it renders the magnetic quiver to be non-simply-laced. We propose a contribution of the O$7^+$ to the usual stable intersection number of 5-branes from tropical geometry, and show that it is consistent with Fayet-Iliopoulos deformations of magnetic quivers which represent mass deformations of 5d SQFTs. From the magnetic quivers, we compute phase diagrams and highest weight generating functions for the Higgs branches, enabling us to identify the global form of the flavour symmetry for several families of 5d SQFTs; among them Bhardwaj's rank-1 theory. For 6d theories realised on a $-4$ curve, we observe the appearance of an additional $D_4$ slice on top of the phase diagram as one goes to the tensionless limit., Comment: V3: typos corrected, 94.2 pages, many figures

Reinforcement learning (RL) is a promising approach for aligning large language models (LLMs) knowledge with sequential decision-making tasks. However, few studies have thoroughly investigated the impact on LLM agents capabilities of fine-tuning them with RL in a specific environment. In this paper, we propose a novel framework to analyze the sensitivity of LLMs to prompt formulations following RL training in a textual environment. Our findings reveal that the performance of LLMs degrades when faced with prompt formulations different from those used during the RL training phase. Besides, we analyze the source of this sensitivity by examining the model's internal representations and salient tokens. Finally, we propose to use a contrastive loss to mitigate this sensitivity and improve the robustness and generalization capabilities of LLMs.

The past years have seen Large Language Models (LLMs) strive not only as generative models but also as agents solving textual sequential decision-making tasks. When facing complex environments where their zero-shot abilities are insufficient, recent work showed online Reinforcement Learning (RL) could be used for the LLM agent to discover and learn efficient strategies interactively. However, most prior work sticks to on-policy algorithms, which greatly reduces the scope of methods such agents could use for both exploration and exploitation, such as experience replay and hindsight relabeling. Yet, such methods may be key for LLM learning agents, and in particular when designing autonomous intrinsically motivated agents sampling and pursuing their own goals (i.e. autotelic agents). This paper presents and studies an adaptation of Soft Actor-Critic and hindsight relabeling to LLM agents. Our method not only paves the path towards autotelic LLM agents that learn online but can also outperform on-policy methods in more classic multi-goal RL environments.

Despite their popularity, deep neural networks (DNNs) applied to time series forecasting often fail to beat simpler statistical models. One of the main causes of this suboptimal performance is the data non-stationarity present in many processes. In particular, changes in the mean and variance of the input data can disrupt the predictive capability of a DNN. In this paper, we first show how DNN forecasting models fail in simple non-stationary settings. We then introduce GAS-Norm, a novel methodology for adaptive time series normalization and forecasting based on the combination of a Generalized Autoregressive Score (GAS) model and a Deep Neural Network. The GAS approach encompasses a score-driven family of models that estimate the mean and variance at each new observation, providing updated statistics to normalize the input data of the deep model. The output of the DNN is eventually denormalized using the statistics forecasted by the GAS model, resulting in a hybrid approach that leverages the strengths of both statistical modeling and deep learning. The adaptive normalization improves the performance of the model in non-stationary settings. The proposed approach is model-agnostic and can be applied to any DNN forecasting model. To empirically validate our proposal, we first compare GAS-Norm with other state-of-the-art normalization methods. We then combine it with state-of-the-art DNN forecasting models and test them on real-world datasets from the Monash open-access forecasting repository. Results show that deep forecasting models improve their performance in 21 out of 25 settings when combined with GAS-Norm compared to other normalization methods., Comment: Accepted at CIKM '24

We consider the $k$-min-sum-radii ($k$-MSR) clustering problem with fairness constraints. The $k$-min-sum-radii problem is a mixture of the classical $k$-center and $k$-median problems. We are given a set of points $P$ in a metric space and a number $k$ and aim to partition the points into $k$ clusters, each of the clusters having one designated center. The objective to minimize is the sum of the radii of the $k$ clusters (where in $k$-center we would only consider the maximum radius and in $k$-median we would consider the sum of the individual points' costs). Various notions of fair clustering have been introduced lately, and we follow the definitions due to Chierichetti, Kumar, Lattanzi and Vassilvitskii [NeurIPS 2017] which demand that cluster compositions shall follow the proportions of the input point set with respect to some given sensitive attribute. For the easier case where the sensitive attribute only has two possible values and each is equally frequent in the input, the aim is to compute a clustering where all clusters have a 1:1 ratio with respect to this attribute. We call this the 1:1 case. There has been a surge of FPT-approximation algorithms for the $k$-MSR problem lately, solving the problem both in the unconstrained case and in several constrained problem variants. We add to this research area by designing an FPT $(6+\epsilon)$-approximation that works for $k$-MSR under the mentioned general fairness notion. For the special 1:1 case, we improve our algorithm to achieve a $(3+\epsilon)$-approximation., Comment: Full version of the paper accepted at ISAAC 2024

Accurately detecting 3D objects from monocular images in dynamic roadside scenarios remains a challenging problem due to varying camera perspectives and unpredictable scene conditions. This paper introduces a two-stage training strategy to address these challenges. Our approach initially trains a model on the large-scale synthetic dataset, RoadSense3D, which offers a diverse range of scenarios for robust feature learning. Subsequently, we fine-tune the model on a combination of real-world datasets to enhance its adaptability to practical conditions. Experimental results of the Cube R-CNN model on challenging public benchmarks show a remarkable improvement in detection performance, with a mean average precision rising from 0.26 to 12.76 on the TUM Traffic A9 Highway dataset and from 2.09 to 6.60 on the DAIR-V2X-I dataset when performing transfer learning. Code, data, and qualitative video results are available on the project website: https://roadsense3d.github.io., Comment: 18 pages. Accepted for ECVA European Conference on Computer Vision 2024 (ECCV'24)

The use of assistive robots in domestic environments can raise significant ethical concerns, from the risk of individual ethical harm to wider societal ethical impacts including culture flattening and compromise of human dignity. It is therefore essential to ensure that technological development of these robots is informed by robust and inclusive techniques for mitigating ethical concerns. This paper presents EHAZOP, a method for conducting an ethical hazard analysis on an assistive robot. EHAZOP draws upon collaborative, creative and structured processes originating within safety engineering, using these to identify ethical concerns associated with the operation of a given assistive robot. We present the results of a proof of concept study of EHAZOP, demonstrating the potential for this process to identify diverse ethical hazards in these systems., Comment: Presented in ICRA 2024 WORKSHOP ON ROBOT ETHICS Ethical, Legal and User Perspectives in Robotics & Automation WOROBET: https://sites.google.com/view/icra2024-robot-ethics

The crystal structures of ANb$_3$Br$_7$S (A = Rb and Cs) have been refined by single crystal X-ray diffraction, and are found to form highly anisotropic materials based on chains of the triangular Nb$_3$ cluster core. The Nb$_3$ cluster core contains seven valence electrons, six of them being assigned to Nb-Nb bonds within the Nb$_3$ triangle and one unpaired d electron. The presence of this surplus electron gives rise to the formation of correlated electronic states. The connectivity in the structures is represented by one-dimensional [Nb$_3$Br$_7$S]$^-$ chains, containing a sulphur atom capping one face ($\mu_3$) of the triangular niobium cluster, which is believed to induce an important electronic feature. Several types of studies are undertaken to obtain deeper insight into the understanding of this unusual type of material: the crystal structure, morphology and elastic properties are analysed, as well the (photo-) electrical properties and NMR relaxation. Electronic structure (DFT) calculations are performed in order to understand the electronic structure and transport in these compounds, and, based on the experimental and theoretical results, we propose that the electronic interactions along the Nb chains are sufficiently one-dimensional to give rise to Luttinger liquid (rather than Fermi liquid) behaviour of the metallic electrons., Comment: 14 pages, 19 figures

We use a combination of density functional theory (DFT) and dynamical mean-field theory (DMFT) to investigate the potential emergence of a charge-disproportionated insulating phase in SrCrO$_{3}$, whereby the Cr cations disproportionate according to $3\text{Cr}^{4+} \rightarrow 2\text{Cr}^{3+} + \text{Cr}^{6+}$ and arrange in ordered planes perpendicular to the cubic [111] direction. We show that the charge disproportionation couples to a structural distortion where the oxygen octahedra around the nominal Cr$^{6+}$ sites contract, while the octahedra surrounding the Cr$^{3+}$ sites expand and distort. Our results indicate that the charge-disproportionated phase can be stabilized for realistic values of the Hubbard $U$ and Hund's $J$ parameters, but this requires a scaling of the DFT+DMFT double counting correction by at least 35%. The disproportionated state can also be stabilized in DFT+U calculations for a specific magnetic configuration with antiparallel magnetic moments on adjacent Cr$^{3+}$ sites and no magnetic moment on the Cr$^{6+}$ sites. While this phase is higher in energy than other magnetic states that can arise in SrCrO$_{3}$, the presence of an energy minimum suggests that the charge disproportionated state represents a metastable state of SrCrO$_{3}$., Comment: 10 pages, 8 figures

The International Fingerprint Liveness Detection Competition (LivDet) is a biennial event that invites academic and industry participants to prove their advancements in Fingerprint Presentation Attack Detection (PAD). This edition, LivDet2023, proposed two challenges, Liveness Detection in Action and Fingerprint Representation, to evaluate the efficacy of PAD embedded in verification systems and the effectiveness and compactness of feature sets. A third, hidden challenge is the inclusion of two subsets in the training set whose sensor information is unknown, testing participants ability to generalize their models. Only bona fide fingerprint samples were provided to participants, and the competition reports and assesses the performance of their algorithms suffering from this limitation in data availability., Comment: 9 pages, 10 tables, IEEE International Joint Conference on Biometrics (IJCB 2023)

Online continual learning aims to get closer to a live learning experience by learning directly on a stream of data with temporally shifting distribution and by storing a minimum amount of data from that stream. In this empirical evaluation, we evaluate various methods from the literature that tackle online continual learning. More specifically, we focus on the class-incremental setting in the context of image classification, where the learner must learn new classes incrementally from a stream of data. We compare these methods on the Split-CIFAR100 and Split-TinyImagenet benchmarks, and measure their average accuracy, forgetting, stability, and quality of the representations, to evaluate various aspects of the algorithm at the end but also during the whole training period. We find that most methods suffer from stability and underfitting issues. However, the learned representations are comparable to i.i.d. training under the same computational budget. No clear winner emerges from the results and basic experience replay, when properly tuned and implemented, is a very strong baseline. We release our modular and extensible codebase at https://github.com/AlbinSou/ocl_survey based on the avalanche framework to reproduce our results and encourage future research., Comment: ICCV Visual Continual Learning Workshop 2023 accepted paper

Analog In-Memory Computing (AIMC) is a promising approach to reduce the latency and energy consumption of Deep Neural Network (DNN) inference and training. However, the noisy and non-linear device characteristics, and the non-ideal peripheral circuitry in AIMC chips, require adapting DNNs to be deployed on such hardware to achieve equivalent accuracy to digital computing. In this tutorial, we provide a deep dive into how such adaptations can be achieved and evaluated using the recently released IBM Analog Hardware Acceleration Kit (AIHWKit), freely available at https://github.com/IBM/aihwkit. The AIHWKit is a Python library that simulates inference and training of DNNs using AIMC. We present an in-depth description of the AIHWKit design, functionality, and best practices to properly perform inference and training. We also present an overview of the Analog AI Cloud Composer, a platform that provides the benefits of using the AIHWKit simulation in a fully managed cloud setting along with physical AIMC hardware access, freely available at https://aihw-composer.draco.res.ibm.com. Finally, we show examples on how users can expand and customize AIHWKit for their own needs. This tutorial is accompanied by comprehensive Jupyter Notebook code examples that can be run using AIHWKit, which can be downloaded from https://github.com/IBM/aihwkit/tree/master/notebooks/tutorial.

In the current digitalization era, capturing and effectively representing knowledge is crucial in most real-world scenarios. In this context, knowledge graphs represent a potent tool for retrieving and organizing a vast amount of information in a properly interconnected and interpretable structure. However, their generation is still challenging and often requires considerable human effort and domain expertise, hampering the scalability and flexibility across different application fields. This paper proposes an innovative knowledge graph generation approach that leverages the potential of the latest generative large language models, such as GPT-3.5, that can address all the main critical issues in knowledge graph building. The approach is conveyed in a pipeline that comprises novel iterative zero-shot and external knowledge-agnostic strategies in the main stages of the generation process. Our unique manifold approach may encompass significant benefits to the scientific community. In particular, the main contribution can be summarized by: (i) an innovative strategy for iteratively prompting large language models to extract relevant components of the final graph; (ii) a zero-shot strategy for each prompt, meaning that there is no need for providing examples for "guiding" the prompt result; (iii) a scalable solution, as the adoption of LLMs avoids the need for any external resources or human expertise. To assess the effectiveness of our proposed model, we performed experiments on a dataset that covered a specific domain. We claim that our proposal is a suitable solution for scalable and versatile knowledge graph construction and may be applied to different and novel contexts.

Neural networks are very effective when trained on large datasets for a large number of iterations. However, when they are trained on non-stationary streams of data and in an online fashion, their performance is reduced (1) by the online setup, which limits the availability of data, (2) due to catastrophic forgetting because of the non-stationary nature of the data. Furthermore, several recent works (Caccia et al., 2022; Lange et al., 2023) arXiv:2205.13452 showed that replay methods used in continual learning suffer from the stability gap, encountered when evaluating the model continually (rather than only on task boundaries). In this article, we study the effect of model ensembling as a way to improve performance and stability in online continual learning. We notice that naively ensembling models coming from a variety of training tasks increases the performance in online continual learning considerably. Starting from this observation, and drawing inspirations from semi-supervised learning ensembling methods, we use a lightweight temporal ensemble that computes the exponential moving average of the weights (EMA) at test time, and show that it can drastically increase the performance and stability when used in combination with several methods from the literature., Comment: CoLLAs 2023 accepted paper

In this paper, we discuss the possibility of achieving tunable topologically protected edge modes through the application of uniaxial deformation in an auxetic metamaterial. The proposed structure consists of a thin slab with oriented cuts in a hexagonal lattice, where topologically protected band gaps are opened by introducing a controlled variation in the cut lengths. Numerical simulations demonstrate the existence of topologically protected and scatter-free wave propagation in the structure at the interface between two sub-domains with modified cells, in distinct frequency ranges. This only occurs in the presence of auxeticity. In addition, exploiting geometrical nonlinearity, the application of a uniaxial strain can be used to close the topological band gaps or to modify their frequency range, i.e., to weaken the localization effects or to shift the frequency at which they occur. The spatial and temporal variation of the applied strain field can thus be used for the dynamic tuning of metamaterial topological waveguiding properties, with applications in mechanical devices for logic operations and computations., Comment: 37 pages, 12 figures