Large Language Models (LLMs) are increasingly adopted in high-stakes scenarios, yet their safety mechanisms often remain fragile. Simple jailbreak prompts or even benign fine-tuning can bypass these protocols, underscoring the need to understand where and how they fail. Recent findings suggest that vulnerabilities emerge when alignment is confined to only the initial output tokens. Unfortunately, even with the introduction of deep safety alignment, determining the optimal safety depth remains an unresolved challenge. By leveraging the equivalence between autoregressive language models and Markov chains, this paper offers the first theoretical result on how to identify the ideal depth for safety alignment, and demonstrates how permutation-based data augmentation can tighten these bounds. Crucially, we reveal a fundamental interaction between alignment depth and ensemble width-indicating that broader ensembles can compensate for shallower alignments. These insights provide a theoretical foundation for designing more robust, scalable safety strategies that complement existing alignment approaches, opening new avenues for research into safer, more reliable LLMs.

Our friend and mentor Peter Caines has, together with his colleagues, created new foundations for studying collective dynamics in complex systems. Of particular inspiration to us has been his pioneering work in mean-field games (MFGs) launched two decades ago [10, 24, 25], and the related field of mean-field control. Peter pointed the way to both formulate and solve the problem of collective dynamics arising in a large population of heterogeneous dynamical systems. In this paper we survey some elements of MFGs within the context of controlled coupled oscillators. We begin by introducing a model for a single oscillator: dθ(t) = (ω + u(t)) dt + σ dξ(t), mod 2π (1) where θ(t) [0, 2π) is the phase of the oscillator at time t, ω is the nominal frequency with units of radiansper-second, {ξ(t): t 0} is a standard Wiener process, and u(t) is a control signal whose interpretation depends on the context. Unless otherwise noted, the SDEs are interpreted in their Itô form.

Deep Neural Networks (DNNs) excel at many tasks, often rivaling or surpassing human performance. Yet their internal processes remain elusive, frequently described as "black boxes." While performance can be refined experimentally, achieving a fundamental grasp of their inner workings is still a challenge. Statistical Mechanics has long tackled computational problems, and this thesis applies physics-based insights to understand DNNs via three complementary approaches. First, by averaging over data, we derive an asymptotic bound on generalization that depends solely on the size of the last layer, rather than on the total number of parameters -- revealing how deep architectures process information differently across layers. Second, adopting a data-dependent viewpoint, we explore a finite-width thermodynamic limit beyond the infinite-width regime. This leads to: (i) a closed-form expression for the generalization error in a finite-width one-hidden-layer network (regression task); (ii) an approximate partition function for deeper architectures; and (iii) a link between deep networks in this thermodynamic limit and Student's t-processes. Finally, from a task-explicit perspective, we present a preliminary analysis of how DNNs interact with a controlled dataset, investigating whether they truly internalize its structure -- collapsing to the teacher -- or merely memorize it. By understanding when a network must learn data structure rather than just memorize, it sheds light on fostering meaningful internal representations. In essence, this thesis leverages the synergy between Statistical Physics and Machine Learning to illuminate the inner behavior of DNNs.

Ensuring consistent processing quality is challenging in laser processes due to varying material properties and surface conditions. Although some approaches have shown promise in solving this problem via automation, they often rely on predetermined targets or are limited to simulated environments. To address these shortcomings, we propose a novel real-time reinforcement learning approach for laser process control, implemented on a Field Programmable Gate Array to achieve real-time execution. Our experimental results from laser welding tests on stainless steel samples with a range of surface roughnesses validated the method's ability to adapt autonomously, without relying on reward engineering or prior setup information. Specifically, the algorithm learned the correct power profile for each unique surface characteristic, demonstrating significant improvements over hand-engineered optimal constant power strategies -- up to 23% better performance on rougher surfaces and 7% on mixed surfaces. This approach represents a significant advancement in automating and optimizing laser processes, with potential applications across multiple industries.

The notion of generalization has moved away from the classical one defined in statistical learning theory towards an emphasis on out-of-domain generalization (OODG). Recently, there is a growing focus on inductive generalization, where a progression of difficulty implicitly governs the direction of domain shifts. In inductive generalization, it is often assumed that the training data lie in the easier side, while the testing data lie in the harder side. The challenge is that training data are always finite, but a learner is expected to infer an inductive principle that could be applied in an unbounded manner. This emerging regime has appeared in the literature under different names, such as length/logical/algorithmic extrapolation, but a formal definition is lacking. This work provides such a formalization that centers on the concept of model successors. Then we outline directions to adapt well-established techniques towards the learning of model successors. This work calls for restructuring of the research discussion around inductive generalization from fragmented task-centric communities to a more unified effort, focused on universal properties of learning and computation.

Language models have recently been shown capable of performing regression tasks wherein numeric predictions are represented as decoded strings. In this work, we provide theoretical grounds for this capability and furthermore investigate the utility of causal auto-regressive sequence models when they are applied to any feature representation. We find that, despite being trained in the usual way - for next-token prediction via cross-entropy loss - decoding-based regression is as performant as traditional approaches for tabular regression tasks, while being flexible enough to capture arbitrary distributions, such as in the task of density estimation.

In the span of four decades, quantum computation has evolved from an intellectual curiosity to a potentially realizable technology. Today, small-scale demonstrations have become possible for quantum algorithmic primitives on hundreds of physical qubits and proof-of-principle error-correction on a single logical qubit. Nevertheless, despite significant progress and excitement, the path toward a full-stack scalable technology is largely unknown. There are significant outstanding quantum hardware, fabrication, software architecture, and algorithmic challenges that are either unresolved or overlooked. These issues could seriously undermine the arrival of utility-scale quantum computers for the foreseeable future. Here, we provide a comprehensive review of these scaling challenges. We show how the road to scaling could be paved by adopting existing semiconductor technology to build much higher-quality qubits, employing system engineering approaches, and performing distributed quantum computation within heterogeneous high-performance computing infrastructures. These opportunities for research and development could unlock certain promising applications, in particular, efficient quantum simulation/learning of quantum data generated by natural or engineered quantum systems. To estimate the true cost of such promises, we provide a detailed resource and sensitivity analysis for classically hard quantum chemistry calculations on surface-code error-corrected quantum computers given current, target, and desired hardware specifications based on superconducting qubits, accounting for a realistic distribution of errors. Furthermore, we argue that, to tackle industry-scale classical optimization and machine learning problems in a cost-effective manner, heterogeneous quantum-probabilistic computing with custom-designed accelerators should be considered as a complementary path toward scalability.

In reinforcement learning for partially observable environments, many successful algorithms were developed within the asymmetric learning paradigm. This paradigm leverages additional state information available at training time for faster learning. Although the proposed learning objectives are usually theoretically sound, these methods still lack a theoretical justification for their potential benefits. We propose such a justification for asymmetric actor-critic algorithms with linear function approximators by adapting a finite-time convergence analysis to this setting. The resulting finite-time bound reveals that the asymmetric critic eliminates an error term arising from aliasing in the agent state.

Scientific Machine Learning (SciML) is a recently emerged research field which combines physics-based and data-driven models for the numerical approximation of differential problems. Physics-based models rely on the physical understanding of the problem at hand, subsequent mathematical formulation, and numerical approximation. Data-driven models instead aim to extract relations between input and output data without arguing any causality principle underlining the available data distribution. In recent years, data-driven models have been rapidly developed and popularized. Such a diffusion has been triggered by a huge availability of data (the so-called big data), an increasingly cheap computing power, and the development of powerful machine learning algorithms. SciML leverages the physical awareness of physics-based models and, at the same time, the efficiency of data-driven algorithms. With SciML, we can inject physics and mathematical knowledge into machine learning algorithms. Yet, we can rely on data-driven algorithms' capability to discover complex and non-linear patterns from data and improve the descriptive capacity of physics-based models. After recalling the mathematical foundations of digital modelling and machine learning algorithms, and presenting the most popular machine learning architectures, we discuss the great potential of a broad variety of SciML strategies in solving complex problems governed by partial differential equations. Finally, we illustrate the successful application of SciML to the simulation of the human cardiac function, a field of significant socio-economic importance that poses numerous challenges on both the mathematical and computational fronts. The corresponding mathematical model is a complex system of non-linear ordinary and partial differential equations describing the electromechanics, valve dynamics, blood circulation, perfusion in the coronary tree, and torso potential. Despite the robustness and accuracy of physics-based models, certain aspects, such as unveiling constitutive laws for cardiac cells and myocardial material properties, as well as devising efficient reduced order models to dominate the extraordinary computational complexity, have been successfully tackled by leveraging data-driven models.

Autoregressive and recurrent networks have achieved remarkable progress across various fields, from weather forecasting to molecular generation and Large Language Models. Despite their strong predictive capabilities, these models lack a rigorous framework for addressing uncertainty, which is key in scientific applications such as PDE solving, molecular generation and Machine Learning Force Fields. To address this shortcoming we present BARNN: a variational Bayesian Autoregressive and Recurrent Neural Network. BARNNs aim to provide a principled way to turn any autoregressive or recurrent model into its Bayesian version. BARNN is based on the variational dropout method, allowing to apply it to large recurrent neural networks as well. We also introduce a temporal version of the "Variational Mixtures of Posteriors" prior (tVAMP-prior) to make Bayesian inference efficient and well-calibrated. Extensive experiments on PDE modelling and molecular generation demonstrate that BARNN not only achieves comparable or superior accuracy compared to existing methods, but also excels in uncertainty quantification and modelling long-range dependencies.