Search

Storytelling is a fundamental aspect of human communication, relying heavily on creativity to produce narratives that are novel, appropriate, and surprising. While large language models (LLMs) have recently demonstrated the ability to generate high-quality stories, their creative capabilities remain underexplored. Previous research has either focused on creativity tests requiring short responses or primarily compared model performance in story generation to that of professional writers. However, the question of whether LLMs exhibit creativity in writing short stories on par with the average human remains unanswered. In this work, we conduct a systematic analysis of creativity in short story generation across LLMs and everyday people. Using a five-sentence creative story task, commonly employed in psychology to assess human creativity, we automatically evaluate model- and human-generated stories across several dimensions of creativity, including novelty, surprise, and diversity. Our findings reveal that while LLMs can generate stylistically complex stories, they tend to fall short in terms of creativity when compared to average human writers., Comment: 14 pages

Record linkage is a crucial concept for integrating data from multiple sources, particularly when datasets lack exact identifiers, and it has diverse applications in real-world data analysis. Privacy-Preserving Record Linkage (PPRL) ensures this integration occurs securely, protecting sensitive information from unauthorized access. This is especially important in sectors such as healthcare, where datasets include private identity information (IDAT) governed by strict privacy laws. However, maintaining both privacy and efficiency in large-scale record linkage poses significant challenges. Consequently, researchers must develop advanced methods to protect data privacy while optimizing processing performance. This paper presents a novel and efficient PPRL method based on a secure 3-party computation (MPC) framework. Our approach allows multiple parties to compute linkage results without exposing their private inputs and significantly improves the speed of linkage process compared to existing privacy-preserving solutions. We demonstrated that our method preserves the linkage quality of the state-of-the-art PPRL method while achieving up to 14 times faster performance. For example, linking a record against a database of 10,000 records takes just 8.74 seconds in a realistic network with 700 Mbps bandwidth and 60 ms latency. Even on a slower internet connection with 100 Mbps bandwidth and 60 ms latency, the linkage completes in 28 seconds, highlighting the scalability and efficiency of our solution.

Creativity is the ability to produce novel, useful, and surprising ideas, and has been widely studied as a crucial aspect of human cognition. Machine creativity on the other hand has been a long-standing challenge. With the rise of advanced generative AI, there has been renewed interest and debate regarding AI's creative capabilities. Therefore, it is imperative to revisit the state of creativity in AI and identify key progresses and remaining challenges. In this work, we survey leading works studying the creative capabilities of AI systems, focusing on creative problem-solving, linguistic, artistic, and scientific creativity. Our review suggests that while the latest AI models are largely capable of producing linguistically and artistically creative outputs such as poems, images, and musical pieces, they struggle with tasks that require creative problem-solving, abstract thinking and compositionality and their generations suffer from a lack of diversity, originality, long-range incoherence and hallucinations. We also discuss key questions concerning copyright and authorship issues with generative models. Furthermore, we highlight the need for a comprehensive evaluation of creativity that is process-driven and considers several dimensions of creativity. Finally, we propose future research directions to improve the creativity of AI outputs, drawing inspiration from cognitive science and psychology., Comment: 44 pages

Medical imaging is key in modern medicine. From magnetic resonance imaging (MRI) to microscopic imaging for blood cell detection, diagnostic medical imaging reveals vital insights into patient health. To predict diseases or provide individualized therapies, machine learning techniques like kernel methods have been widely used. Nevertheless, there are multiple challenges for implementing kernel methods. Medical image data often originates from various hospitals and cannot be combined due to privacy concerns, and the high dimensionality of image data presents another significant obstacle. While randomised encoding offers a promising direction, existing methods often struggle with a trade-off between accuracy and efficiency. Addressing the need for efficient privacy-preserving methods on distributed image data, we introduce OKRA (Orthonormal K-fRAmes), a novel randomized encoding-based approach for kernel-based machine learning. This technique, tailored for widely used kernel functions, significantly enhances scalability and speed compared to current state-of-the-art solutions. Through experiments conducted on various clinical image datasets, we evaluated model quality, computational performance, and resource overhead. Additionally, our method outperforms comparable approaches

Large language models (LLMs) have demonstrated significant progress in various natural language generation and understanding tasks. However, their linguistic generalization capabilities remain questionable, raising doubts about whether these models learn language similarly to humans. While humans exhibit compositional generalization and linguistic creativity in language use, the extent to which LLMs replicate these abilities, particularly in morphology, is under-explored. In this work, we systematically investigate the morphological generalization abilities of LLMs through the lens of compositionality. We define morphemes as compositional primitives and design a novel suite of generative and discriminative tasks to assess morphological productivity and systematicity. Focusing on agglutinative languages such as Turkish and Finnish, we evaluate several state-of-the-art instruction-finetuned multilingual models, including GPT-4 and Gemini. Our analysis shows that LLMs struggle with morphological compositional generalization particularly when applied to novel word roots, with performance declining sharply as morphological complexity increases. While models can identify individual morphological combinations better than chance, their performance lacks systematicity, leading to significant accuracy gaps compared to humans., Comment: 33 pages

Fine-tuning has become a popular approach to adapting large foundational models to specific tasks. As the size of models and datasets grows, parameter-efficient fine-tuning techniques are increasingly important. One of the most widely used methods is Low-Rank Adaptation (LoRA), with adaptation update expressed as the product of two low-rank matrices. While LoRA was shown to possess strong performance in fine-tuning, it often under-performs when compared to full-parameter fine-tuning (FPFT). Although many variants of LoRA have been extensively studied empirically, their theoretical optimization analysis is heavily under-explored. The starting point of our work is a demonstration that LoRA and its two extensions, Asymmetric LoRA and Chain of LoRA, indeed encounter convergence issues. To address these issues, we propose Randomized Asymmetric Chain of LoRA (RAC-LoRA) -- a general optimization framework that rigorously analyzes the convergence rates of LoRA-based methods. Our approach inherits the empirical benefits of LoRA-style heuristics, but introduces several small but important algorithmic modifications which turn it into a provably convergent method. Our framework serves as a bridge between FPFT and low-rank adaptation. We provide provable guarantees of convergence to the same solution as FPFT, along with the rate of convergence. Additionally, we present a convergence analysis for smooth, non-convex loss functions, covering gradient descent, stochastic gradient descent, and federated learning settings. Our theoretical findings are supported by experimental results., Comment: 36 pages, 4 figures, 2 algorithms

Genome-wide association studies are pivotal in understanding the genetic underpinnings of complex traits and diseases. Collaborative, multi-site GWAS aim to enhance statistical power but face obstacles due to the sensitive nature of genomic data sharing. Current state-of-the-art methods provide a privacy-focused approach utilizing computationally expensive methods such as Secure Multi-Party Computation and Homomorphic Encryption. In this context, we present a novel algorithm PP-GWAS designed to improve upon existing standards in terms of computational efficiency and scalability without sacrificing data privacy. This algorithm employs randomized encoding within a distributed architecture to perform stacked ridge regression on a Linear Mixed Model to ensure rigorous analysis. Experimental evaluation with real world and synthetic data indicates that PP-GWAS can achieve computational speeds twice as fast as similar state-of-the-art algorithms while using lesser computational resources, all while adhering to a robust security model that caters to an all-but-one semi-honest adversary setting. We have assessed its performance using various datasets, emphasizing its potential in facilitating more efficient and private genomic analyses.

The current cybersecurity landscape is increasingly complex, with traditional Static Application Security Testing (SAST) tools struggling to capture complex and emerging vulnerabilities due to their reliance on rule-based matching. Meanwhile, Large Language Models (LLMs) have demonstrated powerful code analysis capabilities, but their static training data and privacy risks limit their effectiveness. To overcome the limitations of both approaches, we propose LSAST, a novel approach that integrates LLMs with SAST scanners to enhance vulnerability detection. LSAST leverages a locally hostable LLM, combined with a state-of-the-art knowledge retrieval system, to provide up-to-date vulnerability insights without compromising data privacy. We set a new benchmark for static vulnerability analysis, offering a robust, privacy-conscious solution that bridges the gap between traditional scanners and advanced AI-driven analysis. Our evaluation demonstrates that incorporating SAST results into LLM analysis significantly improves detection accuracy, identifying vulnerabilities missed by conventional methods., Comment: Under Review of IEEE SaTML 2024

The existence and uniqueness of canonical singular solutions of the J-equation and the deformed Hermitian Yang Mills (dHYM) equation was proved in \cite{DMS24} on compact K\"{a}hler surfaces. In this paper, we study the singularity formation of the dHYM cotangent flow on the one-point blow up of $\mathbb{C}\mathbb{P}^3$ using Calabi ansatz. In particular, we provide an explicit example where the flow develops a singularity along the exceptional divisor. Moreover, the limit satisfies corresponding singular dHYM equation in the sense of \cite{DMS24} and provides some evidence for Conjecture $1.12$ in \cite{DMS24} on this three-dimensional manifold with symmetry., Comment: 20 pages

Spin-photon interfaces based on group-IV colour centres in diamond offer a promising platform for quantum networks. A key challenge in the field is realizing precise single-defect positioning and activation, which is crucial for scalable device fabrication. Here we address this problem by demonstrating a two-step fabrication method for tin vacancy (SnV-) centres that uses site-controlled ion implantation followed by local femtosecond laser annealing with in-situ spectral monitoring. The ion implantation is performed with sub-50 nm resolution and a dosage that is controlled from hundreds of ions down to single ions per site, limited by Poissonian statistics. Using this approach, we successfully demonstrate site-selective creation and modification of single SnV- centres. The technique opens a window onto materials tuning at the single defect level, and provides new insight into defect structures and dynamics during the annealing process. While demonstrated for SnV- centres, this versatile approach can be readily generalised to other implanted colour centres in diamond and wide-bandgap materials., Comment: 10 pages, 4 figures, methods, and supplementary information

Quantum sensing based on solid-state spin defects provides a uniquely versatile platform for imaging physical properties at the nanoscale under diverse environmental conditions. Operation of most sensors used to-date is based on projective measurement along a single axis combined with computational extrapolation. Here, we show that the individually addressable carbon-related spin defect in hexagonal boron nitride is a multi-axis spin system for vectorial nanoscale magnetometry. We demonstrate how its low symmetry and strongly spin-selective direct and reverse intersystem crossing dynamics provide sub-$\mu$T/$\sqrt{\text{Hz}}$ magnetic-field sensitivity for both on and off-axis bias magnetic field exceeding 50 mT. Alongside these features, the room-temperature operation and the nanometer-scale proximity enabled by the van der Waals host material further consolidate this system as an exciting quantum sensing platform.

Quantum computing promises to revolutionize machine learning, offering significant efficiency gains in tasks such as clustering and distance estimation. Additionally, it provides enhanced security through fundamental principles like the measurement postulate and the no-cloning theorem, enabling secure protocols such as quantum teleportation and quantum key distribution. While advancements in secure quantum machine learning are notable, the development of secure and distributed quantum analogues of kernel-based machine learning techniques remains underexplored. In this work, we present a novel approach for securely computing common kernels, including polynomial, radial basis function (RBF), and Laplacian kernels, when data is distributed, using quantum feature maps. Our methodology introduces a robust framework that leverages quantum teleportation to ensure secure and distributed kernel learning. The proposed architecture is validated using IBM's Qiskit Aer Simulator on various public datasets., Comment: This paper contains 17 pages, 4 figures, 1 table and 1 appendix. For associated supplementary code, see https://github.com/mdppml/distributed-secure-kernel-based-QML

We introduce Imagen 3, a latent diffusion model that generates high quality images from text prompts. We describe our quality and responsibility evaluations. Imagen 3 is preferred over other state-of-the-art (SOTA) models at the time of evaluation. In addition, we discuss issues around safety and representation, as well as methods we used to minimize the potential harm of our models.

AI assistants are being increasingly used by students enrolled in higher education institutions. While these tools provide opportunities for improved teaching and education, they also pose significant challenges for assessment and learning outcomes. We conceptualize these challenges through the lens of vulnerability, the potential for university assessments and learning outcomes to be impacted by student use of generative AI. We investigate the potential scale of this vulnerability by measuring the degree to which AI assistants can complete assessment questions in standard university-level STEM courses. Specifically, we compile a novel dataset of textual assessment questions from 50 courses at EPFL and evaluate whether two AI assistants, GPT-3.5 and GPT-4 can adequately answer these questions. We use eight prompting strategies to produce responses and find that GPT-4 answers an average of 65.8% of questions correctly, and can even produce the correct answer across at least one prompting strategy for 85.1% of questions. When grouping courses in our dataset by degree program, these systems already pass non-project assessments of large numbers of core courses in various degree programs, posing risks to higher education accreditation that will be amplified as these models improve. Our results call for revising program-level assessment design in higher education in light of advances in generative AI., Comment: 20 pages, 8 figures

We consider a mean-field optimal control problem with general dynamics including common noise and jumps and show that its minimizers are Nash equilibria of an associated mean-field game of controls. These types of games are necessarily potential, and the Nash equilibria obtained as the minimizers of the control problems are naturally related to the McKean-Vlasov equations of Langevin type. We provide several examples to motivate the general theory including the Cucker-Smale Flocking and Kuramoto mean-field games. The invariance property of the value function of these control problems, utilized in our proof, is also proved.

Generalization capabilities, or rather a lack thereof, is one of the most important unsolved problems in the field of robot learning, and while several large scale efforts have set out to tackle this problem, unsolved it remains. In this paper, we hypothesize that learning temporal action abstractions using latent variable models (LVMs), which learn to map data to a compressed latent space and back, is a promising direction towards low-level skills that can readily be used for new tasks. Although several works have attempted to show this, they have generally been limited by architectures that do not faithfully capture shareable representations. To address this we present Quantized Skill Transformer (QueST), which learns a larger and more flexible latent encoding that is more capable of modeling the breadth of low-level skills necessary for a variety of tasks. To make use of this extra flexibility, QueST imparts causal inductive bias from the action sequence data into the latent space, leading to more semantically useful and transferable representations. We compare to state-of-the-art imitation learning and LVM baselines and see that QueST's architecture leads to strong performance on several multitask and few-shot learning benchmarks. Further results and videos are available at https://quest-model.github.io/, Comment: Keywords: Behavior Clonning, Action Quantization, Self Supervised Skill Abstraction, Few-shot Imitation Learning

In this paper, we address a recent trend in robotic home appliances to include vision systems on personal devices, capable of personalizing the appliances on the fly. In particular, we formulate and address an important technical task of personal object search, which involves localization and identification of personal items of interest on images captured by robotic appliances, with each item referenced only by a few annotated images. The task is crucial for robotic home appliances and mobile systems, which need to process personal visual scenes or to operate with particular personal objects (e.g., for grasping or navigation). In practice, personal object search presents two main technical challenges. First, a robot vision system needs to be able to distinguish between many fine-grained classes, in the presence of occlusions and clutter. Second, the strict resource requirements for the on-device system restrict the usage of most state-of-the-art methods for few-shot learning and often prevent on-device adaptation. In this work, we propose Swiss DINO: a simple yet effective framework for one-shot personal object search based on the recent DINOv2 transformer model, which was shown to have strong zero-shot generalization properties. Swiss DINO handles challenging on-device personalized scene understanding requirements and does not require any adaptation training. We show significant improvement (up to 55%) in segmentation and recognition accuracy compared to the common lightweight solutions, and significant footprint reduction of backbone inference time (up to 100x) and GPU consumption (up to 10x) compared to the heavy transformer-based solutions., Comment: 8 pages, 2 figures, accepted to IROS2024

Developing a reliable vision system is a fundamental challenge for robotic technologies (e.g., indoor service robots and outdoor autonomous robots) which can ensure reliable navigation even in challenging environments such as adverse weather conditions (e.g., fog, rain), poor lighting conditions (e.g., over/under exposure), or sensor degradation (e.g., blurring, noise), and can guarantee high performance in safety-critical functions. Current solutions proposed to improve model robustness usually rely on generic data augmentation techniques or employ costly test-time adaptation methods. In addition, most approaches focus on addressing a single vision task (typically, image recognition) utilising synthetic data. In this paper, we introduce Per-corruption Adaptation of Normalization statistics (PAN) to enhance the model robustness of vision systems. Our approach entails three key components: (i) a corruption type identification module, (ii) dynamic adjustment of normalization layer statistics based on identified corruption type, and (iii) real-time update of these statistics according to input data. PAN can integrate seamlessly with any convolutional model for enhanced accuracy in several robot vision tasks. In our experiments, PAN obtains robust performance improvement on challenging real-world corrupted image datasets (e.g., OpenLoris, ExDark, ACDC), where most of the current solutions tend to fail. Moreover, PAN outperforms the baseline models by 20-30% on synthetic benchmarks in object recognition tasks.

Guardrails have emerged as an alternative to safety alignment for content moderation of large language models (LLMs). Existing model-based guardrails have not been designed for resource-constrained computational portable devices, such as mobile phones, more and more of which are running LLM-based applications locally. We introduce LoRA-Guard, a parameter-efficient guardrail adaptation method that relies on knowledge sharing between LLMs and guardrail models. LoRA-Guard extracts language features from the LLMs and adapts them for the content moderation task using low-rank adapters, while a dual-path design prevents any performance degradation on the generative task. We show that LoRA-Guard outperforms existing approaches with 100-1000x lower parameter overhead while maintaining accuracy, enabling on-device content moderation.

Recent years have seen object detection robotic systems deployed in several personal devices (e.g., home robots and appliances). This has highlighted a challenge in their design, i.e., they cannot efficiently update their knowledge to distinguish between general classes and user-specific instances (e.g., a dog vs. user's dog). We refer to this challenging task as Instance-level Personalized Object Detection (IPOD). The personalization task requires many samples for model tuning and optimization in a centralized server, raising privacy concerns. An alternative is provided by approaches based on recent large-scale Foundation Models, but their compute costs preclude on-device applications. In our work we tackle both problems at the same time, designing a Few-Shot IPOD strategy called AuXFT. We introduce a conditional coarse-to-fine few-shot learner to refine the coarse predictions made by an efficient object detector, showing that using an off-the-shelf model leads to poor personalization due to neural collapse. Therefore, we introduce a Translator block that generates an auxiliary feature space where features generated by a self-supervised model (e.g., DINOv2) are distilled without impacting the performance of the detector. We validate AuXFT on three publicly available datasets and one in-house benchmark designed for the IPOD task, achieving remarkable gains in all considered scenarios with excellent time-complexity trade-off: AuXFT reaches a performance of 80% its upper bound at just 32% of the inference time, 13% of VRAM and 19% of the model size., Comment: Accepted at IROS 2024, 8 pages, 4 figures, 6 tables

Merging Large Language Models (LLMs) is a cost-effective technique for combining multiple expert LLMs into a single versatile model, retaining the expertise of the original ones. However, current approaches often overlook the importance of safety alignment during merging, leading to highly misaligned models. This work investigates the effects of model merging on alignment. We evaluate several popular model merging techniques, demonstrating that existing methods do not only transfer domain expertise but also propagate misalignment. We propose a simple two-step approach to address this problem: (i) generating synthetic safety and domain-specific data, and (ii) incorporating these generated data into the optimization process of existing data-aware model merging techniques. This allows us to treat alignment as a skill that can be maximized in the resulting merged LLM. Our experiments illustrate the effectiveness of integrating alignment-related data during merging, resulting in models that excel in both domain expertise and alignment., Comment: Under review

Interlayer excitons in transition-metal dichalcogenide heterobilayers combine high binding energy and valley-contrasting physics with long optical lifetime and strong dipolar character. Their permanent electric dipole enables electric-field control of emission energy, lifetime, and location. Device material and geometry impacts the nature of the interlayer excitons via their real- and momentum-space configurations. Here, we show that interlayer excitons in MoS$_{2}$/MoSe$_{2}$ heterobilayers are formed by charge carriers residing at the Brillouin zone edges, with negligible interlayer hybridization. We find that the moir\'e superlattice leads to the reversal of the valley-dependent optical selection rules, yielding a positively valued g-factor and cross-polarized photoluminescence. Time-resolved photoluminescence measurements reveal that the interlayer exciton population retains the optically induced valley polarization throughout its microsecond-long lifetime. The combination of long optical lifetime and valley polarization retention makes MoS$_{2}$/MoSe$_{2}$ heterobilayers a promising platform for studying fundamental bosonic interactions and developing excitonic circuits for optical information processing.

Optically addressable spin-based quantum sensors enable nanoscale measurements of temperature, magnetic field, pH, and other physical properties of a system. Advancing the sensors beyond proof-of-principle demonstrations in living cells and multicellular organisms towards reliable, damage-free quantum sensing poses three distinct technical challenges. First, spin-based quantum sensing requires optical accessibility and microwave delivery. Second, any microelectronics must be biocompatible and designed for imaging living specimens. Third, efficient microwave delivery and temperature control are essential to reduce unwanted heating and to maintain an optimal biological environment. Here, we present the Quantum Biosensing Chip (Q-BiC), which facilitates microfluidic-compatible microwave delivery and includes on-chip temperature control. We demonstrate the use of Q-BiC in conjunction with nanodiamonds containing nitrogen vacancy centers to perform optically detected magnetic resonance in living systems. We quantify the biocompatibility of microwave excitation required for optically detected magnetic resonance both in vitro in HeLa cells and in vivo in the nematode Caenorhabditis elegans for temperature measurements and determine the microwave-exposure range allowed before detrimental effects are observed. In addition, we show that nanoscale quantum thermometry can be performed in immobilised but non-anaesthetised adult nematodes with minimal stress. These results enable the use of spin-based quantum sensors without damaging the biological system under study, facilitating the investigation of the local thermodynamic and viscoelastic properties of intracellular processes.

Handling missing data is crucial in machine learning, but many datasets contain gaps due to errors or non-response. Unlike traditional methods such as listwise deletion, which are simple but inadequate, the literature offers more sophisticated and effective methods, thereby improving sample size and accuracy. However, these methods require accessing the whole dataset, which contradicts the privacy regulations when the data is distributed among multiple sources. Especially in the medical and healthcare domain, such access reveals sensitive information about patients. This study addresses privacy-preserving imputation methods for sensitive data using secure multi-party computation, enabling secure computations without revealing any party's sensitive information. In this study, we realized the mean, median, regression, and kNN imputation methods in a privacy-preserving way. We specifically target the medical and healthcare domains considering the significance of protection of the patient data, showcasing our methods on a diabetes dataset. Experiments on the diabetes dataset validated the correctness of our privacy-preserving imputation methods, yielding the largest error around $3 \times 10^{-3}$, closely matching plaintext methods. We also analyzed the scalability of our methods to varying numbers of samples, showing their applicability to real-world healthcare problems. Our analysis demonstrated that all our methods scale linearly with the number of samples. Except for kNN, the runtime of all our methods indicates that they can be utilized for large datasets., Comment: Submitted to IEEE International Conference on E-health Networking, Application & Services

Federated learning (FL) allows clients to collaboratively train a global model without sharing their local data with a server. However, clients' contributions to the server can still leak sensitive information. Differential privacy (DP) addresses such leakage by providing formal privacy guarantees, with mechanisms that add randomness to the clients' contributions. The randomness makes it infeasible to train large transformer-based models, common in modern federated learning systems. In this work, we empirically evaluate the practicality of fine-tuning large scale on-device transformer-based models with differential privacy in a federated learning system. We conduct comprehensive experiments on various system properties for tasks spanning a multitude of domains: speech recognition, computer vision (CV) and natural language understanding (NLU). Our results show that full fine-tuning under differentially private federated learning (DP-FL) generally leads to huge performance degradation which can be alleviated by reducing the dimensionality of contributions through parameter-efficient fine-tuning (PEFT). Our benchmarks of existing DP-PEFT methods show that DP-Low-Rank Adaptation (DP-LoRA) consistently outperforms other methods. An even more promising approach, DyLoRA, which makes the low rank variable, when naively combined with FL would straightforwardly break differential privacy. We therefore propose an adaptation method that can be combined with differential privacy and call it DP-DyLoRA. Finally, we are able to reduce the accuracy degradation and word error rate (WER) increase due to DP to less than 2% and 7% respectively with 1 million clients and a stringent privacy budget of $\epsilon=2$., Comment: 16 pages, 10 figures, 5 tables

Large Language Models (LLMs) are becoming crucial across various fields, emphasizing the urgency for high-quality models in underrepresented languages. This study explores the unique challenges faced by low-resource languages, such as data scarcity, model selection, evaluation, and computational limitations, with a special focus on Turkish. We conduct an in-depth analysis to evaluate the impact of training strategies, model choices, and data availability on the performance of LLMs designed for underrepresented languages. Our approach includes two methodologies: (i) adapting existing LLMs originally pretrained in English to understand Turkish, and (ii) developing a model from the ground up using Turkish pretraining data, both supplemented with supervised fine-tuning on a novel Turkish instruction-tuning dataset aimed at enhancing reasoning capabilities. The relative performance of these methods is evaluated through the creation of a new leaderboard for Turkish LLMs, featuring benchmarks that assess different reasoning and knowledge skills. Furthermore, we conducted experiments on data and model scaling, both during pretraining and fine-tuning, simultaneously emphasizing the capacity for knowledge transfer across languages and addressing the challenges of catastrophic forgetting encountered during fine-tuning on a different language. Our goal is to offer a detailed guide for advancing the LLM framework in low-resource linguistic contexts, thereby making natural language processing (NLP) benefits more globally accessible.

Quantum networks require quantum nodes with coherent optical interfaces and multiple stationary qubits. In terms of optical properties, semiconductor quantum dots are highly compelling, but their adoption as quantum nodes has been impaired by the lack of auxiliary qubits. Here, we demonstrate a functional quantum register in a semiconductor quantum dot leveraging the dense, always-present nuclear spin ensemble. We prepare 13,000 host nuclear spins into a single many-body dark state to operate as the register logic state $|0\rangle$. The logic state $|1\rangle$ is defined as a single nuclear magnon excitation, enabling controlled quantum-state transfer between the electron spin qubit and the nuclear magnonic register. Using 130-ns SWAP gates, we implement a full write-store-retrieve-readout protocol with 68.6(4)% raw overall fidelity and a storage time of 130(16) $\mu$s in the absence of dynamical decoupling. Our work establishes how many-body physics can add step-change functionality to quantum devices, in this case transforming quantum dots into multi-qubit quantum nodes with deterministic registers.

A central spin qubit interacting coherently with an ensemble of proximal spins can be used to engineer entangled collective states or a multi-qubit register. Making full use of this many-body platform requires tuning the interaction between the central spin and its spin register. GaAs quantum dots offer a model realization of the central spin system where an electron qubit interacts with multiple ensembles of $\sim 10^{4}$ nuclear spins. In this work, we demonstrate tuning of the interaction between the electron qubit and the nuclear many-body system in a GaAs quantum dot. The homogeneity of the GaAs system allows us to perform high-precision and isotopically selective nuclear sideband spectroscopy, which reveals the single-nucleus electronic Knight field. Together with time-resolved spectroscopy of the nuclear field, this fully characterizes the electron-nuclear interaction for a priori control. An algorithmic feedback sequence selects the nuclear polarization precisely, which adjusts the electron-nuclear exchange interaction in situ via the electronic g-factor anisotropy. This allows us to tune directly the activation rate of a collective nuclear excitation (magnon) and the coherence time of the electron qubit. Our method is applicable to similar central-spin systems and enables the programmable tuning of coherent interactions in the many-body regime., Comment: 9 pages, 5 figures, and Supplementary Materials

Nowadays, users demand for increased personalization of vision systems to localize and identify personal instances of objects (e.g., my dog rather than dog) from a few-shot dataset only. Despite outstanding results of deep networks on classical label-abundant benchmarks (e.g., those of the latest YOLOv8 model for standard object detection), they struggle to maintain within-class variability to represent different instances rather than object categories only. We construct an Object-conditioned Bag of Instances (OBoI) based on multi-order statistics of extracted features, where generic object detection models are extended to search and identify personal instances from the OBoI's metric space, without need for backpropagation. By relying on multi-order statistics, OBoI achieves consistent superior accuracy in distinguishing different instances. In the results, we achieve 77.1% personal object recognition accuracy in case of 18 personal instances, showing about 12% relative gain over the state of the art., Comment: ICASSP 2024. Copyright 2024 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other

We investigate first-order notions of correlated equilibria; distributions of actions for smooth, potentially non-concave games such that players do not incur any regret against small modifications to their strategies along a set of continuous vector fields. We define two such notions, based on local deviations and on stationarity of the distribution, and identify the notion of coarseness as the setting where the associated vector fields are in fact gradient fields. For coarse equilibria, we prove that online (projected) gradient decent has a universal approximation property for both variants of equilibrium. In the non-coarse setting, we instead reduce the problem of finding an equilibrium to fixed-point computation via the usual framework of $\Phi$-regret minimisation, and identify tractable instances. Finally, we study the primal-dual framework to our notion of first-order equilibria. For coarse equilibria defined by a family of functions, we find that a dual bound on the worst-case expectation of a performance metric takes the form of a generalised Lyapunov function for the dynamics of the game. Specifically, usual primal-dual price of anarchy analysis for coarse correlated equilibria as well as the smoothness framework of Roughgarden are both equivalent to a problem of general Lyapunov function estimation. For non-coarse equilibria, we instead observe a vector field fit problem for the gradient dynamics of the game. These follow from containment results in normal form games; the usual notion of a (coarse) correlated equilibria is equivalent to our first-order local notions of (coarse) correlated equilibria with respect to an appropriately chosen set of vector fields., Comment: 50 pages

Ensuring data privacy is a significant challenge for machine learning applications, not only during model training but also during evaluation. Federated learning has gained significant research interest in recent years as a result. Current research on federated learning primarily focuses on preserving privacy during the training phase. However, model evaluation has not been adequately addressed, despite the potential for significant privacy leaks during this phase as well. In this paper, we demonstrate that the state-of-the-art AUC computation method for federated learning systems, which utilizes differential privacy, still leaks sensitive information about the test data while also requiring a trusted central entity to perform the computations. More importantly, we show that the performance of this method becomes completely unusable as the data size decreases. In this context, we propose an efficient, accurate, robust, and more secure evaluation algorithm capable of computing the AUC in horizontal federated learning systems. Our approach not only enhances security compared to the current state-of-the-art but also surpasses the state-of-the-art AUC computation method in both approximation performance and computational robustness, as demonstrated by experimental results. To illustrate, our approach can efficiently calculate the AUC of a federated learning system involving 100 parties, achieving 99.93% accuracy in just 0.68 seconds, regardless of data size, while providing complete data privacy.

Improving model robustness in case of corrupted images is among the key challenges to enable robust vision systems on smart devices, such as robotic agents. Particularly, robust test-time performance is imperative for most of the applications. This paper presents a novel approach to improve robustness of any classification model, especially on severely corrupted images. Our method (FROST) employs high-frequency features to detect input image corruption type, and select layer-wise feature normalization statistics. FROST provides the state-of-the-art results for different models and datasets, outperforming competitors on ImageNet-C by up to 37.1% relative gain, improving baseline of 40.9% mCE on severe corruptions., Comment: ICASSP 2024. Copyright 2024 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other

Two-dimensional (2D) Janus Transition Metal Dichalcogenides (TMDs) have attracted much interest due to their exciting quantum properties arising from their unique two-faced structure, broken-mirror symmetry, and consequent colossal polarisation field within the monolayer. While efforts have been made to achieve high-quality Janus monolayers, the existing methods rely on highly energetic processes that introduce unwanted grain-boundary and point defects with still unexplored effects on the material's structural and excitonic properties Through High-resolution scanning transmission electron microscopy (HRSTEM), density functional theory (DFT), and optical spectroscopy measurements; this work introduces the most encountered and energetically stable point defects. It establishes their impact on the material's optical properties. HRSTEM studies show that the most energetically stable point defects are single (Vs and Vse) and double chalcogen vacancy (Vs-Vse), interstitial defects (Mi), and metal impurities (MW) and establish their structural characteristics. DFT further establishes their formation energies and related localized bands within the forbidden band. Cryogenic excitonic studies on h-BN-encapsulated Janus monolayers offer a clear correlation between these structural defects and observed emission features, which closely align with the results of the theory. The overall results introduce the defect genome of Janus TMDs as an essential guideline for assessing their structural quality and device properties.

The recently discovered Neural collapse (NC) phenomenon states that the last-layer weights of Deep Neural Networks (DNN), converge to the so-called Equiangular Tight Frame (ETF) simplex, at the terminal phase of their training. This ETF geometry is equivalent to vanishing within-class variability of the last layer activations. Inspired by NC properties, we explore in this paper the transferability of DNN models trained with their last layer weight fixed according to ETF. This enforces class separation by eliminating class covariance information, effectively providing implicit regularization. We show that DNN models trained with such a fixed classifier significantly improve transfer performance, particularly on out-of-domain datasets. On a broad range of fine-grained image classification datasets, our approach outperforms i) baseline methods that do not perform any covariance regularization (up to 22%), as well as ii) methods that explicitly whiten covariance of activations throughout training (up to 19%). Our findings suggest that DNNs trained with fixed ETF classifiers offer a powerful mechanism for improving transfer learning across domains., Comment: ICASSP 2024. Copyright 2024 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other

Continual Learning (CL) aims to learn a sequence of problems (i.e., tasks and domains) by transferring knowledge acquired on previous problems, whilst avoiding forgetting of past ones. Different from previous approaches which focused on CL for one NLP task or domain in a specific use-case, in this paper, we address a more general CL setting to learn from a sequence of problems in a unique framework. Our method, HOP, permits to hop across tasks and domains by addressing the CL problem along three directions: (i) we employ a set of adapters to generalize a large pre-trained model to unseen problems, (ii) we compute high-order moments over the distribution of embedded representations to distinguish independent and correlated statistics across different tasks and domains, (iii) we process this enriched information with auxiliary heads specialized for each end problem. Extensive experimental campaign on 4 NLP applications, 5 benchmarks and 2 CL setups demonstrates the effectiveness of our HOP., Comment: AAAI 2024. Main + supplmentary

In multimedia understanding tasks, corrupted samples pose a critical challenge, because when fed to machine learning models they lead to performance degradation. In the past, three groups of approaches have been proposed to handle noisy data: i) enhancer and denoiser modules to improve the quality of the noisy data, ii) data augmentation approaches, and iii) domain adaptation strategies. All the aforementioned approaches come with drawbacks that limit their applicability; the first has high computational costs and requires pairs of clean-corrupted data for training, while the others only allow deployment of the same task/network they were trained on (\ie, when upstream and downstream task/network are the same). In this paper, we propose SyMPIE to solve these shortcomings. To this end, we design a small, modular, and efficient (just 2GFLOPs to process a Full HD image) system to enhance input data for robust downstream multimedia understanding with minimal computational cost. Our SyMPIE is pre-trained on an upstream task/network that should not match the downstream ones and does not need paired clean-corrupted samples. Our key insight is that most input corruptions found in real-world tasks can be modeled through global operations on color channels of images or spatial filters with small kernels. We validate our approach on multiple datasets and tasks, such as image classification (on ImageNetC, ImageNetC-Bar, VizWiz, and a newly proposed mixed corruption benchmark named ImageNetC-mixed) and semantic segmentation (on Cityscapes, ACDC, and DarkZurich) with consistent improvements of about 5\% relative accuracy gain across the board. The code of our approach and the new ImageNetC-mixed benchmark will be made available upon publication., Comment: Accepted at ACM MMSys'24. 10 pages, 7 figures, 8 tables

Inferring contextually-relevant and diverse commonsense to understand narratives remains challenging for knowledge models. In this work, we develop a series of knowledge models, DiffuCOMET, that leverage diffusion to learn to reconstruct the implicit semantic connections between narrative contexts and relevant commonsense knowledge. Across multiple diffusion steps, our method progressively refines a representation of commonsense facts that is anchored to a narrative, producing contextually-relevant and diverse commonsense inferences for an input context. To evaluate DiffuCOMET, we introduce new metrics for commonsense inference that more closely measure knowledge diversity and contextual relevance. Our results on two different benchmarks, ComFact and WebNLG+, show that knowledge generated by DiffuCOMET achieves a better trade-off between commonsense diversity, contextual relevance and alignment to known gold references, compared to baseline knowledge models.