Search

We uncover activity-driven crossover from phase separation to a new turbulent state in a two-dimensional system of counter-rotating spinners. We study the statistical properties of this active-rotor turbulence using the active-rotor Cahn-Hilliard-Navier-Stokes model, and show that the vorticity $\omega \propto \phi$, the scalar field that distinguishes regions with different rotating states. We explain this intriguing proportionality theoretically, and we characterize power-law energy and concentration spectra, intermittency, and flow-topology statistics. We suggest biological implications of such turbulence., Comment: 8 pages, 5 figures

Active-fluid turbulence has been found in bacterial suspensions, but not so far in their algal counterparts. We present the first experimental evidence for turbulence in dense algal suspensions of Chlamydomonas reinhardtii. We carry out a detailed analysis of the statistical properties of the flow present in these cell suspensions and show that they are quantitatively distinct from their counterparts in two-dimensional fluid and bacterial turbulence. Both kinetic-energy and density spectra of the fluid flow in algal turbulence show power-law regimes with unique scaling exponents. The fluid velocity probability distribution function (PDF) is strongly non-Gaussian and the length dependence of the PDF of fluid-velocity increments indicates small-scale intermittency. We compare and contrast our results with recent theoretical predictions for active-scalar turbulence and active glasses. Overall, our results highlight that active turbulence can arise, even in absence of orientational instabilities, so it is not limited to bacterial suspensions but it can also be found in many biological systems with free-swimming micro-organisms., Comment: 9 pages (main article), 5 figures (main article), 3 pages (supplementary information), 3 figures (supplementary information)

We show that for a rational homology 3-sphere $Y$ equipped with a self-conjugate spin$^c$-structure $\mathfrak s$, the $\operatorname{Pin}(2)$-equivariant monopole Floer homology of $(Y,\mathfrak s)$, as defined by Lin, is isomorphic to the $\operatorname{Pin}(2)$-equivariant Seiberg-Witten Floer homology of $(Y,\mathfrak s)$ defined by Manolescu., Comment: 34 pages, 1 figure. arXiv admin note: text overlap with arXiv:1404.4561, arXiv:1603.00582 by other authors

Recent applications of kernel methods in machine learning have seen a renewed interest in the Laplacian kernel, due to its stability to the bandwidth hyperparameter in comparison to the Gaussian kernel, as well as its expressivity being equivalent to that of the neural tangent kernel of deep fully connected networks. However, unlike the Gaussian kernel, the Laplacian kernel is not separable. This poses challenges for techniques to approximate it, especially via the random Fourier features (RFF) methodology and its variants. In this work, we provide random features for the Laplacian kernel and its two generalizations: Mat\'{e}rn kernel and the Exponential power kernel. We provide efficiently implementable schemes to sample weight matrices so that random features approximate these kernels. These weight matrices have a weakly coupled heavy-tailed randomness. Via numerical experiments on real datasets we demonstrate the efficacy of these random feature maps.

Advancements in Large Language Models (LLMs) and their increasing use in medical question-answering necessitate rigorous evaluation of their reliability. A critical challenge lies in hallucination, where models generate plausible yet factually incorrect outputs. In the medical domain, this poses serious risks to patient safety and clinical decision-making. To address this, we introduce MedHallu, the first benchmark specifically designed for medical hallucination detection. MedHallu comprises 10,000 high-quality question-answer pairs derived from PubMedQA, with hallucinated answers systematically generated through a controlled pipeline. Our experiments show that state-of-the-art LLMs, including GPT-4o, Llama-3.1, and the medically fine-tuned UltraMedical, struggle with this binary hallucination detection task, with the best model achieving an F1 score as low as 0.625 for detecting "hard" category hallucinations. Using bidirectional entailment clustering, we show that harder-to-detect hallucinations are semantically closer to ground truth. Through experiments, we also show incorporating domain-specific knowledge and introducing a "not sure" category as one of the answer categories improves the precision and F1 scores by up to 38% relative to baselines., Comment: Code and dataset are available at https://medhallu.github.io/

We combine machine-learning (ML) techniques with Monte Carlo (MC) simulations and finite-size scaling (FSS) to study continuous and first-order phase transitions in Ising, Blume-Capel, and Ising-metamagnet spin models. We go beyond earlier studies that had concentrated on obtaining the correlation-length exponent $\nu$. In particular, we show (a) how to combine neural networks (NNs), trained with data from MC simulations of Ising-type spin models on finite lattices, with FSS to obtain both thermal magnetic exponents $y_t = 1/\nu$ and $y_h$, respectively, at both critical and tricritical points, (b) how to obtain the NN counterpart of two-scale-factor universality at an Ising-type critical point, and (c) FSS at a first-order transition. We also obtain the FSS forms for the output of our trained NNs as functions of both the temperature and the magnetic field.

The use of AI in healthcare has the potential to improve patient care, optimize clinical workflows, and enhance decision-making. However, bias, data incompleteness, and inaccuracies in training datasets can lead to unfair outcomes and amplify existing disparities. This research investigates the current state of dataset documentation practices, focusing on their ability to address these challenges and support ethical AI development. We identify shortcomings in existing documentation methods, which limit the recognition and mitigation of bias, incompleteness, and other issues in datasets. We propose the 'Healthcare AI Datasheet' to address these gaps, a dataset documentation framework that promotes transparency and ensures alignment with regulatory requirements. Additionally, we demonstrate how it can be expressed in a machine-readable format, facilitating its integration with datasets and enabling automated risk assessments. The findings emphasise the importance of dataset documentation in fostering responsible AI development., Comment: Irish Conference on Artificial Intelligence and Cognitive Science (AICS), December 2024, Ireland

People exhibit unique emotional responses. In the same scenario, the emotional reactions of two individuals can be either similar or vastly different. For instance, consider one person's reaction to an invitation to smoke versus another person's response to a query about their sleep quality. The identification of these individual traits through the observation of common physical parameters opens the door to a wide range of applications, including psychological analysis, criminology, disease prediction, addiction control, and more. While there has been previous research in the fields of psychometrics, inertial sensors, computer vision, and audio analysis, this paper introduces a novel technique for identifying human traits in real time using brainwave data. To achieve this, we begin with an extensive study of brainwave data collected from 80 participants using a portable EEG headset. We also conduct a statistical analysis of the collected data utilizing box plots. Our analysis uncovers several new insights, leading us to a groundbreaking unified approach for identifying diverse human traits by leveraging machine learning techniques on EEG data. Our analysis demonstrates that this proposed solution achieves high accuracy. Moreover, we explore two deep-learning models to compare the performance of our solution. Consequently, we have developed an integrated, real-time trait identification solution using EEG data, based on the insights from our analysis. To validate our approach, we conducted a rigorous user evaluation with an additional 20 participants. The outcomes of this evaluation illustrate both high accuracy and favorable user ratings, emphasizing the robust potential of our proposed method to serve as a versatile solution for human trait identification., Comment: 11th International Conference on Networking, Systems, and Security (NSysS '24)

The AI Act introduces the obligation to conduct a Fundamental Rights Impact Assessment (FRIA), with the possibility to reuse a Data Protection Impact Assessment (DPIA), and requires the EU Commission to create of an automated tool to support the FRIA process. In this article, we provide our novel exploration of the DPIA and FRIA as information processes to enable the creation of automated tools. We first investigate the information involved in DPIA and FRIA, and then use this to align the two to state where a DPIA can be reused in a FRIA. We then present the FRIA as a 5-step process and discuss the role of an automated tool for each step. Our work provides the necessary foundation for creating and managing information for FRIA and supporting it through an automated tool as required by the AI Act., Comment: Presented at CLAIRvoyant (ConventicLE on Artificial Intelligence Regulation) Workshop 2024

The recently published EU Artificial Intelligence Act (AI Act) is a landmark regulation that regulates the use of AI technologies. One of its novel requirements is the obligation to conduct a Fundamental Rights Impact Assessment (FRIA), where organisations in the role of deployers must assess the risks of their AI system regarding health, safety, and fundamental rights. Another novelty in the AI Act is the requirement to create a questionnaire and an automated tool to support organisations in their FRIA obligations. Such automated tools will require a machine-readable form of information involved within the FRIA process, and additionally also require machine-readable documentation to enable further compliance tools to be created. In this article, we present our novel representation of the FRIA as an ontology based on semantic web standards. Our work builds upon the existing state of the art, notably the Data Privacy Vocabulary (DPV), where similar works have been established to create tools for GDPR's Data Protection Impact Assessments (DPIA) and other obligations. Through our ontology, we enable the creation and management of FRIA, and the use of automated tool in its various steps., Comment: Presented at CLAIRvoyant (ConventicLE on Artificial Intelligence Regulation) Workshop 2024

The European Union's Artificial Intelligence Act (AI Act) requires providers and deployers of high-risk AI applications to register their systems into the EU database, wherein the information should be represented and maintained in an easily-navigable and machine-readable manner. Given the uptake of open data and Semantic Web-based approaches for other EU repositories, in particular the use of the Data Catalogue vocabulary Application Profile (DCAT-AP), a similar solution for managing the EU database of high-risk AI systems is needed. This paper introduces AICat - an extension of DCAT for representing catalogues of AI systems that provides consistency, machine-readability, searchability, and interoperability in managing open metadata regarding AI systems. This open approach to cataloguing ensures transparency, traceability, and accountability in AI application markets beyond the immediate needs of high-risk AI compliance in the EU. AICat is available online at https://w3id.org/aicat under the CC-BY-4.0 license., Comment: Presented at 37th International Conference on Legal Knowledge and Information Systems (JURIX) 2024

The EU GDPR is a landmark regulation that introduced several rights for individuals to obtain information and control how their personal data is being processed, as well as receive a copy of it. However, there are gaps in the effective use of rights due to each organisation developing custom methods for rights declaration and management. Simultaneously, there is a technological gap as there is no single consistent standards-based mechanism that can automate the handling of rights for both organisations and individuals. In this article, we present a specification for exercising and managing rights in a machine-interpretable format based on semantic web standards. Our approach uses the comprehensive Data Privacy Vocabulary to create a streamlined workflow for individuals to understand what rights exist, how and where to exercise them, and for organisations to effectively manage them. This work pushes the state of the art in GDPR rights management and is crucial for data reuse and rights management under technologically intensive developments, such as Data Spaces., Comment: Presented at 37th International Conference on Legal Knowledge and Information Systems (JURIX) 2024

The existence of EFX allocations is one of the most significant open questions in fair division. Recent work by Christodolou, Fiat, Koutsoupias, and Sgouritsa ("Fair allocation in graphs", EC 2023) establishes the existence of EFX allocations for graphical valuations, when agents are vertices in a graph, items are edges, and each item has zero value for all agents other than those at its end-points. Thus in this setting, each good has non-zero value for at most two agents, and there is at most one good valued by any pair of agents. This marks one of the few cases when an exact and complete EFX allocation is known to exist for arbitrary agents. In this work, we extend these results to multi-graphs, when each pair of vertices can have more than one edge between them. The existence of EFX allocations in multi-graphs is a natural open question given their existence in simple graphs. We show that EFX allocations exist, and can be computed in polynomial time, for agents with cancellable valuations in the following cases: (i) bipartite multi-graphs, (ii) multi-trees with monotone valuations, and (iii) multi-graphs with girth $(2t-1)$, where $t$ is the chromatic number of the multi-graph. The existence in multi-cycles follows from (i) and (iii).

In the fair division of items among interested agents, envy-freeness is possibly the most favoured and widely studied formalisation of fairness. For indivisible items, envy-free allocations may not exist in trivial cases, and hence research and practice focus on relaxations, particularly envy-freeness up to one item (EF1). A significant reason for the popularity of EF1 allocations is its simple fact of existence. It is known that EF1 allocations exist for two agents with arbitrary valuations; agents with doubly-monotone valuations; agents with Boolean valuations; and identical agents with negative Boolean valuations. We consider two new but natural classes of valuations, and partly extend results on the existence of EF1 allocations to these valuations. Firstly, we consider trilean valuations - an extension of Boolean valuations - when the value of any subset is 0, $a$, or $b$ for any integers $a$ and $b$. Secondly, we define separable single-peaked valuations, when the set of items is partitioned into types. For each type, an agent's value is a single-peaked function of the number of items of the type. The value for a set of items is the sum of values for the different types. We prove EF1 existence for identical trilean valuations for any number of agents, and for separable single-peaked valuations for three agents. For both classes of valuations, we also show that EFX allocations do not exist.

Classical kernel machines have historically faced significant challenges in scaling to large datasets and model sizes--a key ingredient that has driven the success of neural networks. In this paper, we present a new methodology for building kernel machines that can scale efficiently with both data size and model size. Our algorithm introduces delayed projections to Preconditioned Stochastic Gradient Descent (PSGD) allowing the training of much larger models than was previously feasible, pushing the practical limits of kernel-based learning. We validate our algorithm, EigenPro4, across multiple datasets, demonstrating drastic training speed up over the existing methods while maintaining comparable or better classification accuracy., Comment: arXiv admin note: text overlap with arXiv:2302.02605

We show that flocking of microswimmers in a turbulent flow can enhance the efficacy of reinforcement-learning-based path-planning of microswimmers in turbulent flows. In particular, we develop a machine-learning strategy that incorporates Vicsek-model-type flocking in microswimmer assemblies in a statistically homogeneous and isotropic turbulent flow in two dimensions (2D). We build on the adversarial-reinforcement-learning of Ref.~\cite{alageshan2020machine} for non-interacting microswimmers in turbulent flows. Such microswimmers aim to move optimally from an initial position to a target. We demonstrate that our flocking-aided version of the adversarial-reinforcement-learning strategy of Ref.~\cite{alageshan2020machine} can be superior to earlier microswimmer path-planning strategies.

Recent advances in machine learning have led to increased interest in reproducing kernel Banach spaces (RKBS) as a more general framework that extends beyond reproducing kernel Hilbert spaces (RKHS). These works have resulted in the formulation of representer theorems under several regularized learning schemes. However, little is known about an optimization method that encompasses these results in this setting. This paper addresses a learning problem on Banach spaces endowed with a reproducing kernel, focusing on efficient optimization within RKBS. To tackle this challenge, we propose an algorithm based on mirror descent (MDA). Our approach involves an iterative method that employs gradient steps in the dual space of the Banach space using the reproducing kernel. We analyze the convergence properties of our algorithm under various assumptions and establish two types of results: first, we identify conditions under which a linear convergence rate is achievable, akin to optimization in the Euclidean setting, and provide a proof of the linear rate; second, we demonstrate a standard convergence rate in a constrained setting. Moreover, to instantiate this algorithm in practice, we introduce a novel family of RKBSs with $p$-norm ($p \neq 2$), characterized by both an explicit dual map and a kernel., Comment: 42 pages, 3 figures

This systematic review explores the theoretical foundations, evolution, applications, and future potential of Kolmogorov-Arnold Networks (KAN), a neural network model inspired by the Kolmogorov-Arnold representation theorem. KANs distinguish themselves from traditional neural networks by using learnable, spline-parameterized functions instead of fixed activation functions, allowing for flexible and interpretable representations of high-dimensional functions. This review details KAN's architectural strengths, including adaptive edge-based activation functions that improve parameter efficiency and scalability in applications such as time series forecasting, computational biomedicine, and graph learning. Key advancements, including Temporal-KAN, FastKAN, and Partial Differential Equation (PDE) KAN, illustrate KAN's growing applicability in dynamic environments, enhancing interpretability, computational efficiency, and adaptability for complex function approximation tasks. Additionally, this paper discusses KAN's integration with other architectures, such as convolutional, recurrent, and transformer-based models, showcasing its versatility in complementing established neural networks for tasks requiring hybrid approaches. Despite its strengths, KAN faces computational challenges in high-dimensional and noisy data settings, motivating ongoing research into optimization strategies, regularization techniques, and hybrid models. This paper highlights KAN's role in modern neural architectures and outlines future directions to improve its computational efficiency, interpretability, and scalability in data-intensive applications.

We explore the relationship between complexity and duality in quantum systems, focusing on how local and non-local operators evolve under time evolution. We find that non-local operators, which are dual to local operators under specific mappings, exhibit behavior that mimics the growth of their local counterparts, particularly when considering state complexity. For the open transverse Ising model this leads to a neat organisation of the operator dynamics on either side of the duality, both consistent with growth expected in a quadratic fermion model like the Kitaev chain. When examing periodic chains, however, the mapping of boundary terms provides access to multiple branches of highly complex operators. These give rise to much larger saturation values of complexity for parity-mixing operators and are in contrast to what one would expect for a quadratic Hamiltonian. Our results shed light on the intricate relationship between non-locality, complexity growth, and duality in quantum systems., Comment: 19+2 pages

Quantization is a powerful tool for accelerating large language model (LLM) inference, but the accuracy-performance trade-offs across different formats remain unclear. In this paper, we conduct the most comprehensive empirical study to date, evaluating FP8, INT8, and INT4 quantization across academic benchmarks and real-world tasks on the entire Llama-3.1 model family. Through over 500,000 evaluations, our investigation yields several key findings: (1) FP8 (W8A8-FP) is effectively lossless across all model scales, (2) well-tuned INT8 (W8A8-INT) achieves surprisingly low (1-3\%) accuracy degradation, and (3) INT4 weight-only (W4A16-INT) is more competitive than expected, rivaling 8-bit quantization. Further, we investigate the optimal quantization format for different deployments by analyzing inference performance through the popular vLLM framework. Our analysis provides clear deployment recommendations: W4A16 is the most cost-efficient for synchronous setups, while W8A8 dominates in asynchronous continuous batching. For mixed workloads, the optimal choice depends on the specific use case. Our findings offer practical, data-driven guidelines for deploying quantized LLMs at scale -- ensuring the best balance between speed, efficiency, and accuracy.

We study the Laplacian of the undirected De Bruijn graph over an alphabet $A$ of order $k$. While the eigenvalues of this Laplacian were found in 1998 by Delorme and Tillich [1], an explicit description of its eigenvectors has remained elusive. In this work, we find these eigenvectors in closed form and show that they yield a natural and canonical basis for the cut- and cycle-spaces of De Bruijn graphs. Remarkably, we find that the cycle basis we construct is a basis for the cycle space of both the undirected and the directed De Bruijn graph. This is done by developing an analogue of the Fourier transform on the De Bruijn graph, which acts to diagonalize the Laplacian. Moreover, we show that the cycle-space of De Bruijn graphs, when considering all possible orders of $k$ simultaneously, contains a rich algebraic structure, that of a graded Hopf algebra.

We show that the Gross-Pitaevskii equation coupled with the wave equation for a wire (GP-W) provides a natural theoretical framework for understanding recent experiments employing a nanowire to detect a single quantum vortex in superfluid $^4 {\rm He}$. We uncover the complete spatiotemporal evolution of such wire-based vortex detection via direct numerical simulations of the GP-W system. Furthermore, by computing the spatiotemporal spectrum, we obtain the vortex-capture-induced change in the oscillation frequency of the wire. We quantify this frequency shift by plotting the wire's oscillation frequency versus time and obtain results that closely match experimental observations. In addition, we provide analytical support for our numerical results by deriving the dispersion relation for the oscillating wire, with and without a trapped vortex. We show that the Magnus force opens a gap in the wire dispersion relation. The size of the gap becomes the characteristic frequency of the wire when a vortex is trapped., Comment: 7 pages, 3 figures, Accepted for publication in Physical Review B as a Letter

Liquid-droplet coalescence and the mergers of liquid lenses are problems of great practical and theoretical interest in fluid dynamics and the statistical mechanics of multi-phase flows. During such mergers, there is an interesting and intricate interplay between the shapes of the interfaces, separating two phases, and the background flow field. In experiments, it is easier to visualize concentration fields than to obtain the flow field. We demonstrate that two-dimensional (2D) encoder-decoder CNNs, 2D U-Nets, and three-dimensional (3D) U-Nets can be used to obtain flow fields from concentration fields here. To train these networks, we use concentration and flow fields, which we obtain from extensive direct numerical simulations (DNSs) of (a) the coalescence of two circular droplets in the two-component 2D Cahn-Hilliard-Navier-Stokes (CHNS) partial differential equations (PDEs), (b) liquid-lens mergers in the three-component 2D CHNS PDEs, and (c) spherical-droplet coalescence in the two-component 3D CHNS PDEs. We then show that, given test images of concentration fields, our trained models accurately predict the flow fields at both high and low Ohnesorge numbers $Oh$ (a dimensionless ratio of viscous stresses to the inertial and surface-tension forces). Using autoencoders and fully connected neural networks, we also investigate the mapping between the concentration and vorticity fields via low-dimensional latent variables for droplet mergers in the 2D CHNS system. We compare the accuracies of flow-field reconstruction based on the two approaches we employ. Finally, we use data from recent experiments on droplet coalescence to show how our method can be used to obtain the flow field from measurements of the concentration field.

Ensuring faithfulness to context in large language models (LLMs) and retrieval-augmented generation (RAG) systems is crucial for reliable deployment in real-world applications, as incorrect or unsupported information can erode user trust. Despite advancements on standard benchmarks, faithfulness hallucination-where models generate responses misaligned with the provided context-remains a significant challenge. In this work, we introduce FaithEval, a novel and comprehensive benchmark tailored to evaluate the faithfulness of LLMs in contextual scenarios across three diverse tasks: unanswerable, inconsistent, and counterfactual contexts. These tasks simulate real-world challenges where retrieval mechanisms may surface incomplete, contradictory, or fabricated information. FaithEval comprises 4.9K high-quality problems in total, validated through a rigorous four-stage context construction and validation framework, employing both LLM-based auto-evaluation and human validation. Our extensive study across a wide range of open-source and proprietary models reveals that even state-of-the-art models often struggle to remain faithful to the given context, and that larger models do not necessarily exhibit improved faithfulness.Project is available at: \url{https://github.com/SalesforceAIResearch/FaithEval}.

Retrieval Augmented Generation (RAG), a paradigm that integrates external contextual information with large language models (LLMs) to enhance factual accuracy and relevance, has emerged as a pivotal area in generative AI. The LLMs used in RAG applications are required to faithfully and completely comprehend the provided context and users' questions, avoid hallucination, handle unanswerable, counterfactual or otherwise low-quality and irrelevant contexts, perform complex multi-hop reasoning and produce reliable citations. In this paper, we introduce SFR-RAG, a small LLM that is instruction-tuned with an emphasis on context-grounded generation and hallucination minimization. We also present ContextualBench, a new evaluation framework compiling multiple popular and diverse RAG benchmarks, such as HotpotQA and TriviaQA, with consistent RAG settings to ensure reproducibility and consistency in model assessments. Experimental results demonstrate that our SFR-RAG-9B model outperforms leading baselines such as Command-R+ (104B) and GPT-4o, achieving state-of-the-art results in 3 out of 7 benchmarks in ContextualBench with significantly fewer parameters. The model is also shown to be resilient to alteration in the contextual information and behave appropriately when relevant context is removed. Additionally, the SFR-RAG model maintains competitive performance in general instruction-following tasks and function-calling capabilities., Comment: Technical report

We consider Carrollian conformal field theories in two dimensions and construct the boundary Carrollian conformal algebra (BCCA), opening up innumerable possibilities for further studies, given the growing relevance of Carrollian symmetries. We prove that the BCCA emerges by contracting a single copy of the Virasoro algebra. As an application, we construct, for the first time, open null strings and show that, for Dirichlet boundary conditions, we recover the BCCA as the algebra of constraints. We finally reconstruct our string results by taking the null limit of tensile open strings., Comment: 7pp, 1fig

Voice cloning is a prominent feature in personalized speech interfaces. A neural vocal cloning system can mimic someone's voice using just a few audio samples. Both speaker encoding and speaker adaptation are topics of research in the field of voice cloning. Speaker adaptation relies on fine-tuning a multi-speaker generative model, which involves training a separate model to infer a new speaker embedding used for speaker encoding. Both methods can achieve excellent performance, even with a small number of cloning audios, in terms of the speech's naturalness and similarity to the original speaker. Speaker encoding approaches are more appropriate for low-resource deployment since they require significantly less memory and have a faster cloning time than speaker adaption, which can offer slightly greater naturalness and similarity. The main goal is to create a vocal cloning system that produces audio output with a Nepali accent or that sounds like Nepali. For the further advancement of TTS, the idea of transfer learning was effectively used to address several issues that were encountered in the development of this system, including the poor audio quality and the lack of available data., Comment: 6 pages, 10 figures

Catastrophic interference, the loss of previously learned information when learning new information, remains a major challenge in machine learning. Since living organisms do not seem to suffer from this problem, researchers have taken inspiration from biology to improve memory retention in artificial intelligence systems. However, previous attempts to use bio-inspired mechanisms have typically resulted in systems that rely on task boundary information during training and/or explicit task identification during inference, information that is not available in real-world scenarios. Here, we show that neuro-inspired mechanisms such as synaptic consolidation and metaplasticity can mitigate catastrophic interference in a spiking neural network, using only synapse-local information, with no need for task awareness, and with a fixed memory size that does not need to be increased when training on new tasks. Our model, TACOS, combines neuromodulation with complex synaptic dynamics to enable new learning while protecting previous information. We evaluate TACOS on sequential image recognition tasks and demonstrate its effectiveness in reducing catastrophic interference. Our results show that TACOS outperforms existing regularization techniques in domain-incremental learning scenarios. We also report the results of an ablation study to elucidate the contribution of each neuro-inspired mechanism separately.

The implementation of Artificial Intelligence (AI) in the healthcare industry has garnered considerable attention, attributable to its prospective enhancement of clinical outcomes, expansion of access to superior healthcare, cost reduction, and elevation of patient satisfaction. Nevertheless, the primary hurdle that persists is related to the quality of accessible multi-modal healthcare data in conjunction with the evolution of AI methodologies. This study delves into the adoption of large language models to address specific challenges, specifically, the standardization of healthcare data. We advocate the use of these models to identify and map clinical data schemas to established data standard attributes, such as the Fast Healthcare Interoperability Resources. Our results illustrate that employing large language models significantly diminishes the necessity for manual data curation and elevates the efficacy of the data standardization process. Consequently, the proposed methodology has the propensity to expedite the integration of AI in healthcare, ameliorate the quality of patient care, whilst minimizing the time and financial resources necessary for the preparation of data for AI., Comment: 11 pages, 2 figures, 4 tables

Recent results suggest that highly active, chaotic, non-equilibrium states of living fluids might share much in common with high Reynolds number, inertial turbulence. We now show, by using a hydrodynamical model, the onset of intermittency and the consequent multiscaling of Eulerian and Lagrangian structure functions as a function of the bacterial activity. Our results bridge the worlds of low and high Reynolds number flows as well as open up intriguing possibilities of what makes flows intermittent.

Given a set of co-located mobile robots in an unknown anonymous graph, the robots must relocate themselves in distinct graph nodes to solve the dispersion problem. In this paper, we consider the dispersion problem for silent robots \cite{gorain2024collaborative}, i.e., no direct, explicit communication between any two robots placed in the nodes of an oriented $n$ node ring network. The robots operate in synchronous rounds. The dispersion problem for silent mobile robots has been studied in arbitrary graphs where the robots start from a single source. In this paper, we focus on the dispersion problem for silent mobile robots where robots can start from multiple sources. The robots have unique labels from a range $[0,\;L]$ for some positive integer $L$. Any two co-located robots do not have the information about the label of the other robot. The robots have weak multiplicity detection capability, which means they can determine if it is alone on a node. The robots are assumed to be able to identify an increase or decrease in the number of robots present on a node in a particular round. However, the robots can not get the exact number of increase or decrease in the number of robots. We have proposed a deterministic distributed algorithm that solves the dispersion of $k$ robots in an oriented ring in $O(\log L+k)$ synchronous rounds with $O(\log L)$ bits of memory for each robot. A lower bound $\Omega(\log L+k)$ on time for the dispersion of $k$ robots on a ring network is presented to establish the optimality of the proposed algorithm.

Kernel ridge regression (KRR) is a popular class of machine learning models that has become an important tool for understanding deep learning. Much of the focus has been on studying the proportional asymptotic regime, $n \asymp d$, where $n$ is the number of training samples and $d$ is the dimension of the dataset. In this regime, under certain conditions on the data distribution, the kernel random matrix involved in KRR exhibits behavior akin to that of a linear kernel. In this work, we extend the study of kernel regression to the quadratic asymptotic regime, where $n \asymp d^2$. In this regime, we demonstrate that a broad class of inner-product kernels exhibit behavior similar to a quadratic kernel. Specifically, we establish an operator norm approximation bound for the difference between the original kernel random matrix and a quadratic kernel random matrix with additional correction terms compared to the Taylor expansion of the kernel functions. The approximation works for general data distributions under a Gaussian-moment-matching assumption with a covariance structure. This new approximation is utilized to obtain a limiting spectral distribution of the original kernel matrix and characterize the precise asymptotic training and generalization errors for KRR in the quadratic regime when $n/d^2$ converges to a non-zero constant. The generalization errors are obtained for both deterministic and random teacher models. Our proof techniques combine moment methods, Wick's formula, orthogonal polynomials, and resolvent analysis of random matrices with correlated entries., Comment: 75 pages

Neural networks trained to solve modular arithmetic tasks exhibit grokking, a phenomenon where the test accuracy starts improving long after the model achieves 100% training accuracy in the training process. It is often taken as an example of "emergence", where model ability manifests sharply through a phase transition. In this work, we show that the phenomenon of grokking is not specific to neural networks nor to gradient descent-based optimization. Specifically, we show that this phenomenon occurs when learning modular arithmetic with Recursive Feature Machines (RFM), an iterative algorithm that uses the Average Gradient Outer Product (AGOP) to enable task-specific feature learning with general machine learning models. When used in conjunction with kernel machines, iterating RFM results in a fast transition from random, near zero, test accuracy to perfect test accuracy. This transition cannot be predicted from the training loss, which is identically zero, nor from the test loss, which remains constant in initial iterations. Instead, as we show, the transition is completely determined by feature learning: RFM gradually learns block-circulant features to solve modular arithmetic. Paralleling the results for RFM, we show that neural networks that solve modular arithmetic also learn block-circulant features. Furthermore, we present theoretical evidence that RFM uses such block-circulant features to implement the Fourier Multiplication Algorithm, which prior work posited as the generalizing solution neural networks learn on these tasks. Our results demonstrate that emergence can result purely from learning task-relevant features and is not specific to neural architectures nor gradient descent-based optimization methods. Furthermore, our work provides more evidence for AGOP as a key mechanism for feature learning in neural networks.

We demonstrate the existence of interface-induced turbulence, an emergent nonequilibrium statistically steady state (NESS) with spatiotemporal chaos, which is induced by interfacial fluctuations in low-Reynolds-number binary-fluid mixtures. We uncover the properties of this NESS via direct numerical simulations (DNSs) of cellular flows in the Cahn-Hilliard-Navier-Stokes (CHNS) equations for binary fluids. We show that, in this NESS, the shell-averaged energy spectrum $E(k)$ is spread over several decades in the wavenumber $k$ and it exhibits a power-law region, indicative of turbulence but without a conventional inertial cascade. To characterize the statistical properties of this turbulence, we compute, in addition to $E(k)$, the time series $e(t)$ of the kinetic energy and its power spectrum, scale-by-scale energy transfer as a function of $k$, and the energy dissipation resulting from interfacial stresses. Furthermore, we analyze the mixing properties of this low-Reynolds-number turbulence via the mean-square displacement (MSD) of Lagrangian tracer particles, for which we demonstrate diffusive behavior at long times, a hallmark of strong mixing in turbulent flows., Comment: 9 pages, 6 figures

We present an efficient method for adapting a monolingual Large Language Model (LLM) to another language, addressing challenges of catastrophic forgetting and tokenizer limitations. We focus this study on adapting Llama 2 to Arabic. Our two-stage approach begins with expanding the vocabulary and training only the embeddings matrix, followed by full model continual pre-training on a bilingual corpus. By continually pre-training on a mix of Arabic and English corpora, the model retains its proficiency in English while acquiring capabilities in Arabic. Our approach results in significant improvements in Arabic and slight enhancements in English, demonstrating cost-effective cross-lingual transfer. We perform ablations on embedding initialization techniques, data mix ratios, and learning rates and release a detailed training recipe. To demonstrate generalizability of this approach we also adapted Llama 3 8B to Arabic and Llama 2 13B to Hindi.

Large language models (LLMs) are increasingly being used to synthesize and reason about source code. However, the static nature of these models' knowledge does not reflect the fact that libraries and API functions they invoke are continuously evolving, with functionality being added or changing. While numerous benchmarks evaluate how LLMs can generate code, no prior work has studied how an LLMs' knowledge about code API functions can be updated. To fill this gap, we present CodeUpdateArena, a benchmark for knowledge editing in the code domain. An instance in our benchmark consists of a synthetic API function update paired with a program synthesis example that uses the updated functionality; our goal is to update an LLM to be able to solve this program synthesis example without providing documentation of the update at inference time. Compared to knowledge editing for facts encoded in text, success here is more challenging: a code LLM must correctly reason about the semantics of the modified function rather than just reproduce its syntax. Our dataset is constructed by first prompting GPT-4 to generate atomic and executable function updates. Then, for each update, we generate program synthesis examples whose code solutions are prone to use the update. Our benchmark covers updates of various types to 54 functions from seven diverse Python packages, with a total of 670 program synthesis examples. Our experiments show that prepending documentation of the update to open-source code LLMs (i.e., DeepSeek, CodeLlama) does not allow them to incorporate changes for problem solving, and existing knowledge editing techniques also have substantial room for improvement. We hope our benchmark will inspire new methods for knowledge updating in code LLMs., Comment: Under Review

Quantum error mitigation schemes (QEM) have greatly enhanced the performance of quantum computers, mostly by reducing errors caused by interactions with the environment. Nevertheless, the presence of coherence errors, typically arising from miscalibration and inter-qubit crosstalk, is a significant challenge to the scalability of quantum computing. Such errors are often addressed using a refined Pauli twirling scheme called Randomized Compiling (RC) that converts the coherent errors into incoherent errors that can then be mitigated by conventional QEM. Unfortunately for multi-qubit gates, RC is restricted to Clifford gates such as CNOT and CPHASE. However, it has been demonstrated experimentally that a direct implementation of multi-qubit non-Clifford gates, i.e. without using multi-qubit Clifford gates, has reduced the depth of the circuit by a factor of four and more. Recently, a framework called pseudo-twirling (PST) for treating coherent error in multi-qubit non-Clifford gates has been introduced and experimentally demonstrated. We show analytically that a higher order correction to the existing PST theory yields an over-rotation coherent error generated by the PST protocol itself. This PST effect has no analogue in RC. Although the small induced over-rotation can amount to a significant coherent error in deep circuits, we explain why it does not degrade the performance of the gate. Interestingly, we find that a simplified twirling scheme that was introduced and exploited experimentally by Kim et al. also displays an induced over-rotation. We study the conditions under which the two twirling schemes display the same over-rotation behavior., Comment: v2 includes an analysis of half-twirling

This paper introduces a collaborative, human-centred taxonomy of AI, algorithmic and automation harms. We argue that existing taxonomies, while valuable, can be narrow, unclear, typically cater to practitioners and government, and often overlook the needs of the wider public. Drawing on existing taxonomies and a large repository of documented incidents, we propose a taxonomy that is clear and understandable to a broad set of audiences, as well as being flexible, extensible, and interoperable. Through iterative refinement with topic experts and crowdsourced annotation testing, we propose a taxonomy that can serve as a powerful tool for civil society organisations, educators, policymakers, product teams and the general public. By fostering a greater understanding of the real-world harms of AI and related technologies, we aim to increase understanding, empower NGOs and individuals to identify and report violations, inform policy discussions, and encourage responsible technology development and deployment., Comment: 21 pages, 2 figures; typos corrected, reference missing fields fixed, printer-friendly version of a diagram added