As artificial intelligence systems become increasingly integrated into daily life, the field of explainability has gained significant attention. This trend is particularly driven by the complexity of modern AI models and their decision-making processes. The advent of foundation models, characterized by their extensive generalization capabilities and emergent uses, has further complicated this landscape. Foundation models occupy an ambiguous position in the explainability domain: their complexity makes them inherently challenging to interpret, yet they are increasingly leveraged as tools to construct explainable models. In this survey, we explore the intersection of foundation models and eXplainable AI (XAI) in the vision domain. We begin by compiling a comprehensive corpus of papers that bridge these fields. Next, we categorize these works based on their architectural characteristics. We then discuss the challenges faced by current research in integrating XAI within foundation models. Furthermore, we review common evaluation methodologies for these combined approaches. Finally, we present key observations and insights from our survey, offering directions for future research in this rapidly evolving field.

We study a limiting absorption principle for the boundary-value problem describing a hybrid plasma resonance, with a regular coefficient in the principal part of the operator that vanishes on a curve inside the domain and changes its sign across this curve. We prove the limiting absorption principle by establishing a priori bounds on the solution in certain weighted Sobolev spaces. Next, we show that the solution can be decomposed into regular and singular parts. A peculiar property of this decomposition enables us to introduce a radiation-like condition in a bounded domain and to state a well-posed problem satisfied by the limiting absorption solution.

Deep Neural Networks (DNNs) have demonstrated remarkable performance across various domains, including computer vision and natural language processing. However, they often struggle to accurately quantify the uncertainty of their predictions, limiting their broader adoption in critical real-world applications. Uncertainty Quantification (UQ) for Deep Learning seeks to address this challenge by providing methods to improve the reliability of uncertainty estimates. Although numerous techniques have been proposed, a unified tool offering a seamless workflow to evaluate and integrate these methods remains lacking. To bridge this gap, we introduce Torch-Uncertainty, a PyTorch and Lightning-based framework designed to streamline DNN training and evaluation with UQ techniques and metrics. In this paper, we outline the foundational principles of our library and present comprehensive experimental results that benchmark a diverse set of UQ methods across classification, segmentation, and regression tasks. Our library is available at this https URL

We introduce SamBa-GQW, a novel quantum algorithm for solving binary combinatorial optimization problems of arbitrary degree with no use of any classical optimizer. The algorithm is based on a continuous-time quantum walk on the solution space represented as a graph. The walker explores the solution space to find its way to vertices that minimize the cost function of the optimization problem. The key novelty of our algorithm is an offline classical sampling protocol that gives information about the spectrum of the problem Hamiltonian. Then, the extracted information is used to guide the walker to high quality solutions via a quantum walk with a time-dependent hopping rate. We investigate the performance of SamBa-GQW on several quadratic problems, namely MaxCut, maximum independent set, portfolio optimization, and higher-order polynomial problems such as LABS, MAX-$k$-SAT and a quartic reformulation of the travelling salesperson problem. We empirically demonstrate that SamBa-GQW finds high quality approximate solutions on problems up to a size of $n=20$ qubits by only sampling $n^2$ states among $2^n$ possible decisions. SamBa-GQW compares very well also to other guided quantum walks and QAOA.

Robots are still limited to controlled conditions, that the robot designer knows with enough details to endow the robot with the appropriate models or behaviors. Learning algorithms add some flexibility with the ability to discover the appropriate behavior given either some demonstrations or a reward to guide its exploration with a reinforcement learning algorithm. Reinforcement learning algorithms rely on the definition of state and action spaces that define reachable behaviors. Their adaptation capability critically depends on the representations of these spaces: small and discrete spaces result in fast learning while large and continuous spaces are challenging and either require a long training period or prevent the robot from converging to an appropriate behavior. Beside the operational cycle of policy execution and the learning cycle, which works at a slower time scale to acquire new policies, we introduce the redescription cycle, a third cycle working at an even slower time scale to generate or adapt the required representations to the robot, its environment and the task. We introduce the challenges raised by this cycle and we present DREAM (Deferred Restructuring of Experience in Autonomous Machines), a developmental cognitive architecture to bootstrap this redescription process stage by stage, build new state representations with appropriate motivations, and transfer the acquired knowledge across domains or tasks or even across robots. We describe results obtained so far with this approach and end up with a discussion of the questions it raises in Neuroscience.

Although deep learning has demonstrated remarkable capabilities in learning from unstructured data, modern tree-based ensemble models remain superior in extracting relevant information and learning from structured datasets. While several efforts have been made to accelerate tree-based models, the inherent characteristics of the models pose significant challenges for conventional accelerators. Recent research leveraging content-addressable memory (CAM) offers a promising solution for accelerating tree-based models, yet existing designs suffer from excessive memory consumption and low utilization. This work addresses these challenges by introducing RETENTION, an end-to-end framework that significantly reduces CAM capacity requirement for tree-based model inference. We propose an iterative pruning algorithm with a novel pruning criterion tailored for bagging-based models (e.g., Random Forest), which minimizes model complexity while ensuring controlled accuracy degradation. Additionally, we present a tree mapping scheme that incorporates two innovative data placement strategies to alleviate the memory redundancy caused by the widespread use of don't care states in CAM. Experimental results show that implementing the tree mapping scheme alone achieves $1.46\times$ to $21.30 \times$ better space efficiency, while the full RETENTION framework yields $4.35\times$ to $207.12\times$ improvement with less than 3% accuracy loss. These results demonstrate that RETENTION is highly effective in reducing CAM capacity requirement, providing a resource-efficient direction for tree-based model acceleration.

We study data-driven least squares (LS) problems with semidefinite (SD) constraints and derive finite-sample guarantees on the spectrum of their optimal solutions when these constraints are relaxed. In particular, we provide a high confidence bound allowing one to solve a simpler program in place of the full SDLS problem, while ensuring that the eigenvalues of the resulting solution are $\varepsilon$-close of those enforced by the SD constraints. The developed certificate, which consistently shrinks as the number of data increases, turns out to be easy-to-compute, distribution-free, and only requires independent and identically distributed samples. Moreover, when the SDLS is used to learn an unknown quadratic function, we establish bounds on the error between a gradient descent iterate minimizing the surrogate cost obtained with no SD constraints and the true minimizer.

While automatic monitoring and coaching of exercises are showing encouraging results in non-medical applications, they still have limitations such as errors and limited use contexts. To allow the development and assessment of physical rehabilitation by an intelligent tutoring system, we identify in this article four challenges to address and propose a medical dataset of clinical patients carrying out low back-pain rehabilitation exercises. The dataset includes 3D Kinect skeleton positions and orientations, RGB videos, 2D skeleton data, and medical annotations to assess the correctness, and error classification and localisation of body part and timespan. Along this dataset, we perform a complete research path, from data collection to processing, and finally a small benchmark. We evaluated on the dataset two baseline movement recognition algorithms, pertaining to two different approaches: the probabilistic approach with a Gaussian Mixture Model (GMM), and the deep learning approach with a Long-Short Term Memory (LSTM). This dataset is valuable because it includes rehabilitation relevant motions in a clinical setting with patients in their rehabilitation program, using a cost-effective, portable, and convenient sensor, and because it shows the potential for improvement on these challenges.

Global optimization has gained attraction over the past decades, thanks to the development of both theoretical foundations and efficient numerical routines. Among recent advances, Kernel Sum of Squares (KernelSOS) provides a powerful theoretical framework, combining the expressivity of kernel methods with the guarantees of SOS optimization. In this paper, we take KernelSOS from theory to practice and demonstrate its use on challenging control and robotics problems. We identify and address the practical considerations required to make the method work in applied settings: restarting strategies, systematic calibration of hyperparameters, methods for recovering minimizers, and the combination with fast local solvers. As a proof of concept, the application of KernelSOS to robot localization highlights its competitiveness with existing SOS approaches that rely on heuristics and handcrafted reformulations to render the problem polynomial. Even in the high-dimensional, non-parametric setting of trajectory optimization with simulators treated as black boxes, we demonstrate how KernelSOS can be combined with fast local solvers to uncover higher-quality solutions without compromising overall runtimes.

The rise of Large Language Models (LLMs) has triggered legal and ethical concerns, especially regarding the unauthorized use of copyrighted materials in their training datasets. This has led to lawsuits against tech companies accused of using protected content without permission. Membership Inference Attacks (MIAs) aim to detect whether specific documents were used in a given LLM pretraining, but their effectiveness is undermined by biases such as time-shifts and n-gram overlaps. This paper addresses the evaluation of MIAs on LLMs with partially inferable training sets, under the ex-post hypothesis, which acknowledges inherent distributional biases between members and non-members datasets. We propose and validate algorithms to create ``non-biased'' and ``non-classifiable'' datasets for fairer MIA assessment. Experiments using the Gutenberg dataset on OpenLamma and Pythia show that neutralizing known biases alone is insufficient. Our methods produce non-biased ex-post datasets with AUC-ROC scores comparable to those previously obtained on genuinely random datasets, validating our approach. Globally, MIAs yield results close to random, with only one being effective on both random and our datasets, but its performance decreases when bias is removed.

Beam-driven plasma-wakefield acceleration (PWFA) has emerged as a transformative technology with the potential to revolutionize the field of particle acceleration, especially toward compact accelerators for high-energy and high-power applications. Charged particle beams are used to excite density waves in plasma with accelerating fields reaching up to 100 GV/m, thousands of times stronger than the fields provided by radio-frequency cavities. Plasma-wakefield-accelerator research has matured over the span of four decades from basic concepts and proof-of-principle experiments to a rich and rapidly progressing sub-field with dedicated experimental facilities and state-of-the-art simulation codes. We review the physics, including theory of linear and nonlinear plasma wakefields as well as beam dynamics of both the wakefield driver and trailing bunches accelerating in the plasma wake, and address challenges associated with energy efficiency and preservation of beam quality. Advanced topics such as positron acceleration, self-modulation, internal injection, long-term plasma evolution and multistage acceleration are discussed. Simulation codes and major experiments are surveyed, spanning the use of electron, positron and proton bunches as wakefield drivers. Finally, we look ahead to future particle colliders and light sources based on plasma technology.

Image captioning models have been able to generate grammatically correct and human understandable sentences. However most of the captions convey limited information as the model used is trained on datasets that do not caption all possible objects existing in everyday life. Due to this lack of prior information most of the captions are biased to only a few objects present in the scene, hence limiting their usage in daily life. In this paper, we attempt to show the biased nature of the currently existing image captioning models and present a new image captioning dataset, Egoshots, consisting of 978 real life images with no captions. We further exploit the state of the art pre-trained image captioning and object recognition networks to annotate our images and show the limitations of existing works. Furthermore, in order to evaluate the quality of the generated captions, we propose a new image captioning metric, object based Semantic Fidelity (SF). Existing image captioning metrics can evaluate a caption only in the presence of their corresponding annotations; however, SF allows evaluating captions generated for images without annotations, making it highly useful for real life generated captions.

Microgrids are integrated systems that gather and operate energy production units to satisfy consumers demands. This paper details different mathematical methods to design the Energy Management System (EMS) of domestic microgrids. We consider different stocks coupled together - a battery, a domestic hot water tank - and decentralized energy production with solar panel. The main challenge of the EMS is to ensure, at least cost, that supply matches demand for all time, while considering the inherent uncertainties of such systems. We benchmark two optimization algorithms to manage the EMS, and compare them with a heuristic. The Model Predictive Control (MPC) is a well known algorithm which models the future uncertainties with a deterministic forecast. By contrast, Stochastic Dual Dynamic Programming (SDDP) models the future uncertainties as probability distributions to compute optimal policies. We present a fair comparison of these two algorithms to control microgrid. A comprehensive numerical study shows that i) optimization algorithms achieve significant gains compared to the heuristic, ii) SDDP outperforms MPC by a few percents, with a reasonable computational overhead.

This work presents a new approach to the guidance and control of marine craft via HEOL, i.e., a new way of combining flatness-based and model-free controllers. Its goal is to develop a general regulator for Unmanned Surface Vehicles (USV). To do so, the well-known USV maneuvering model is simplified into a nominal Hovercraft model which is flat. A flatness-based controller is derived for the simplified USV model and the loop is closed via an intelligent proportional-derivative (iPD) regulator. We thus associate the well-documented natural robustness of flatness-based control and adaptivity of iPDs. The controller is applied in simulation to two surface vessels, one meeting the simplifying hypotheses, the other one being a generic USV of the literature. It is shown to stabilize both systems even in the presence of unmodeled environmental disturbances.

The latest Deep Learning (DL) models for detection and classification have achieved an unprecedented performance over classical machine learning algorithms. However, DL models are black-box methods hard to debug, interpret, and certify. DL alone cannot provide explanations that can be validated by a non technical audience. In contrast, symbolic AI systems that convert concepts into rules or symbols -- such as knowledge graphs -- are easier to explain. However, they present lower generalisation and scaling capabilities. A very important challenge is to fuse DL representations with expert knowledge. One way to address this challenge, as well as the performance-explainability trade-off is by leveraging the best of both streams without obviating domain expert knowledge. We tackle such problem by considering the symbolic knowledge is expressed in form of a domain expert knowledge graph. We present the eXplainable Neural-symbolic learning (X-NeSyL) methodology, designed to learn both symbolic and deep representations, together with an explainability metric to assess the level of alignment of machine and human expert explanations. The ultimate objective is to fuse DL representations with expert domain knowledge during the learning process to serve as a sound basis for explainability. X-NeSyL methodology involves the concrete use of two notions of explanation at inference and training time respectively: 1) EXPLANet: Expert-aligned eXplainable Part-based cLAssifier NETwork Architecture, a compositional CNN that makes use of symbolic representations, and 2) SHAP-Backprop, an explainable AI-informed training procedure that guides the DL process to align with such symbolic representations in form of knowledge graphs. We showcase X-NeSyL methodology using MonuMAI dataset for monument facade image classification, and demonstrate that our approach improves explainability and performance.

In this work we study how the convergence rate of GMRES is influenced by the properties of linear systems arising from Helmholtz problems near resonances or quasi-resonances. We extend an existing convergence bound to demonstrate that the approximation of small eigenvalues by harmonic Ritz values plays a key role in convergence behavior. Next, we analyze the impact of deflation using carefully selected vectors and combine this with a Complex Shifted Laplacian preconditioner. Finally, we apply these tools to two numerical examples near (quasi-)resonant frequencies, using them to explain how the convergence rate evolves.

We establish improved convergence rates for curved boundary element methods applied to the three-dimensional (3D) Laplace and Helmholtz equations with smooth geometry and data. Our analysis relies on a precise analysis of the consistency errors introduced by the perturbed bilinear and sesquilinear forms. We illustrate our results with numerical experiments in 3D based on basis functions and curved triangular elements up to order four.