arXiv subjects

Pascal Bouvry

Publications and source records attributed to Pascal Bouvry.

At least 19 recordsLinked to original sources

arXiv2026

Dynamics-aware bandwidth selection for smoothed charge deposition in 2D electrostatic particle-in-cell

Charge deposition in the particle-in-cell (PIC) method can be viewed as kernel density estimation on a spatial mesh. This allows the smoothing bandwidth to be selected through statistical criteria such as mean integrated squared error or cross-validation, but it also changes the electric field acting on the particles. Therefore, a selected bandwidth can alter the growth or damping of physical modes. We develop a dynamics-aware bandwidth-selection procedure that combines a density-error objective with constraints on specified linear modes. These constraints follow from the plasma equilibrium and the combined transfer of deposition, filtering and field gathering within the model rate-tolerance boundary. We demonstrate the need for equilibrium information through a controlled two-stream experiment. The controlled two-stream scan varies the beam drift while holding the initial spatial density, thermal samples, wavevector and fixed-filter transfer unchanged. Across four resolved equilibria approaching marginal stability, this filter reduces the measured growth rate by $0.7\%$ to $9.1\%$. Equilibrium-dependent interior settings satisfy the prescribed $10\%$ rate tolerance. We then apply the constrained selection rule to 85 Gaussian candidates, that reduces density and grid-field mean-square errors by $7.40\%$ and $3.70\%$ on independently evaluated snapshots. A separate projected-linear calculation at that exact width gives a ratio of mean individual growth rates of $0.9556$, with paired $95\%$ interval $[0.9453,0.9640]$, consistent with the prediction $0.9513$. Analysis of finite-marker loading and ensemble averaging explains how the measured rates relate to the response model. Together, these results demonstrate a filter-design procedure that combines statistical error reduction with subsequent validation of the selected linear dynamics.

physics.plasm-ph↗

arXiv2026

Beyond Compression: Diagnosing How Post-Training Changes Mathematical Reasoning

Post-training is central to mathematical reasoning in modern large language models (LLMs), but endpoint pass@1 alone underidentifies what has changed. Gains may reflect newly reachable solutions, cheaper sampling of latent solutions, surface robustness, or memorisation. We compare three post-training paths under a common diagnostic readout: our sufficiently trained off-policy distillation trajectories, released Qwen3 off-policy-plus-on-policy distillation endpoints, and a released DeepSeek-Math endpoint trained with Group Relative Policy Optimisation (GRPO). Our probe uses cross-surface pass@K over verbatim prompts, paraphrases, numerical isomorphisms, and translations, plus consistency, distribution-shape, and verified supervised-fine-tuning (SFT) membership analyses. We find two regimes. On easier AMC problems, large-K ceilings are near saturation, so post-training mainly compresses sample cost. On harder AIME problems, post-training expands the large-K ceiling over the base model: sufficient off-policy distillation already raises this ceiling, Qwen3 released endpoints raise it further, and DeepSeek-Math GRPO does not dominate sufficient off-policy distillation at large K. English-dominant distillation improves non-English reasoning but preserves language-tier gaps. A controlled-overfit audit finds limited sensitivity in current SFT-membership probes. Compression is one regime of post-training, not a universal explanation.

cs.LG↗

arXiv2026

What Does Post-Training Change in Multilingual Reasoning?

Open-source reasoning models provide unequal access to reasoning capability across languages. When a model can solve a problem but cannot deliver a complete solution in the user's language, language becomes an access barrier rather than merely a source of performance variation. We audit Qwen3 checkpoints on competition-mathematics tasks in eleven languages. Across the ten non-English languages, only 15.4-17.9% of problems receive a correct, terminating solution with visible reasoning in the requested language in any of 16 samples, compared with 92.9% in English. To identify the source of this disparity, we evaluate thirteen endpoints from one model family, spanning released checkpoints, multilingual supervised fine-tuning (SFT) at two scales, controlled SFT ablations, and three reinforcement-learning (RL) reward formulations. We jointly track correctness, language adherence, termination, and delivery efficiency. The dominant bottleneck shifts across post-training stages. Released models often reason in English. Multilingual SFT restores target-language reasoning, but accuracy declines across multilingual, English-only, and single-language SFT runs, showing that this cost is not specific to multilingual mixing; non-English reasoning traces additionally become prone to non-terminating loops. RL restores termination in both arms at no cost in accuracy, but only the arm whose reward includes a language term delivers: rewarding correctness alone returns the model to English. Together, these stages establish a constructive post-training path from English-pivoted capability to multilingual reasoning that is reliably delivered.

cs.CL↗

arXiv2026

Unlocking the Critic: Reward-Free Policy Optimization for LLM Post-Training

Recent approaches to reinforcement learning (RL) post-training for large language models increasingly remove the critic to reduce training instability and memory overhead. Even where a critic is trained, it is discarded once training ends, although it has learned to predict outcomes. We revisit this trend and show that a pretrained critic's ability to predict future outcomes can make it a valuable asset for efficient long-horizon reasoning. First, we find that instability in critic-based RL for long chain-of-thought reasoning is largely an optimization artifact: keeping policy updates small and low in variance restores stable convergence. Second, a well-pretrained critic estimates the posterior probability of eventual success from later trajectory states and unfinished prefixes. Its predictions provide outcome-derived, dense, per-prefix learning signals that, during policy optimization, require neither completed rollouts, step-level annotations, nor external reward labels. Building on this insight, we introduce Reward-Free Policy Optimization (RFPO), which repurposes a single calibrated, frozen critic as a rollout-level reward, a value baseline for generalized advantage estimation, and a success forecaster for unfinished prefixes. We further show that binarizing the debiased score stops the policy from exploiting the critic's length bias. Binarized, RFPO matches supervised PPO without a single label in the training loop, while cutting compute and memory overhead. This makes RFPO well suited to long-horizon reasoning tasks, where outcomes arrive late and generation dominates cost: because rollouts can be rewarded before they finish, training no longer has to pay for waiting on every trajectory to complete. Our findings challenge the prevailing critic-free paradigm and establish critic-based, reward-free optimization as a scalable and computationally efficient path for LLM post-training.

cs.LG↗

arXiv2026

Solving Without Stopping: On-Policy Distillation at Small Scale

On-policy distillation, where a student learns from a stronger teacher's feedback on its own outputs, is a common way to pass reasoning to smaller models. We analyze what it transfers at small scale, distilling Qwen3-8B into Qwen3 4B, 1.7B and 0.6B students, in thinking mode (reason at length, then end the reasoning and answer) and, for comparison, in non-thinking mode (no separate reasoning phase). Long reasoning needs two abilities, solving a problem and knowing when it is solved, and we find that distillation transfers the first, but in thinking mode not the second. Solving improves at every size, up to two ceilings, which we measure comprehensively across both modes and all student sizes: a student's single attempt never exceeds what it could already reach in many attempts before training, and the smaller the student, the further it stays below the teacher. Stopping is where the modes part. In non-thinking mode every student keeps stopping; in thinking mode students stop ending their reasoning early in training, and the smaller the student, the less of this ability survives: the teacher signals a stop almost only where a student already ends its reasoning, so distillation teaches no new stops; it only keeps the student's existing stops that land on a right answer, and a weak student has few such stops. The smallest students often reach the right value but do not commit to it: they either rarely mark it or mark it and write past it. Together, these results describe how small students behave under on-policy distillation, and a diagnostic that separates answer marking, correctness and stopping.

cs.AI↗

arXiv2026

Reward-Guided Autoregressive Graph Generation for Efficient Multi-Agent Communication Topology Design

LLM-based Multi-Agent Systems (MAS) achieve strong performance on complex reasoning tasks by coordinating multiple agents, but at the cost of substantial token consumption. Recent work on automatic topology design, ARG-Designer, has reframed this problem as autoregressive graph generation. However, its training objective provides no explicit incentive for the model to generate sparse and efficient topologies. We address this limitation by introducing a Reward-Guided Autoregressive Graph Generation (RGA-Designer) inspired by Reinforcement Learning from Human Feedback (RLHF). We train a reward model that jointly captures task correctness and structural compactness, and then fine-tune the pretrained graph generator using the reward model as feedback. Our method preserves task accuracy at the level of ARG-Designer while reducing token consumption by an average of 20.5%.

cs.MA↗

arXiv2026

A Review on Quantum Circuit Optimization using ZX-Calculus

Quantum computing promises significant speed-ups for certain algorithms but the practical use of current noisy intermediate-scale quantum (NISQ) era computers remains limited by resources constraints (e.g., noise, qubits, gates, and circuit depth). Quantum circuit optimization is a key mitigation strategy. In this context, ZX-calculus has emerged as an alternative framework that allows for semantics-preserving quantum circuit optimization. We review ZX-based optimization of quantum circuits, categorizing them by optimization techniques, target metrics and intended quantum computing architecture. In addition, we outline critical challenges and future research directions, such as multi-objective optimization, scalable algorithms, and enhanced circuit extraction methods. This survey is valuable for researchers in both combinatorial optimization and quantum computing. For researchers in combinatorial optimization, we provide the background to understand a new challenging combinatorial problem: ZX-based quantum circuit optimization. For researchers in quantum computing, we classify and explain existing circuit optimization techniques.

quant-ph↗

arXiv2026

SkyHOST: A Unified Architecture for Cross-Cloud Hybrid Object and Stream Transfer

Cloud and big data workloads are increasingly distributing data across multiple cloud providers and regions for rapid decision-making and analytics. Traditional transfer tools are typically specialized for a single paradigm, either stream replication or bulk transfer. This specialization forces users to deploy and manage separate systems with different configurations for each transfer pattern. This paper presents SkyHOST (Hybrid Object and Stream Transfer), a unified data movement architecture built upon the Skyplane framework to bridge the gap between bulk object transfer and streaming workloads through a single control plane and CLI. SkyHOST manages URI-based routing to automatically select the appropriate transfer mechanism, supporting both structured data for record-level ingestion and chunk-based transfer for large binary objects. We demonstrate, through an environmental monitoring use case and empirical evaluation, that SkyHOST provides operational simplicity by consolidating heterogeneous data movement patterns under a single control plane while achieving competitive throughput for cross-region transfers.

cs.DC↗

arXiv2026

Hyperparameter Optimization of Constraint Programming Solvers

The performance of constraint programming solvers is highly sensitive to the choice of their hyperparameters. Manually finding the best solver configuration is a difficult, time-consuming task that typically requires expert knowledge. In this paper, we introduce probe and solve algorithm, a novel two-phase framework for automated hyperparameter optimization integrated into the CPMpy library. This approach partitions the available time budget into two phases: a probing phase that explores different sets of hyperparameters using configurable hyperparameter optimization methods, followed by a solving phase where the best configuration found is used to tackle the problem within the remaining time. We implement and compare two hyperparameter optimization methods within the probe and solve algorithm: Bayesian optimization and Hamming distance search. We evaluate the algorithm on two different constraint programming solvers, ACE and Choco, across 114 combinatorial problem instances, comparing their performance against the solver's default configurations. Results show that using Bayesian optimization, the algorithm outperforms the solver's default configurations, improving solution quality for ACE in 25.4% of instances and matching the default performance in 57.9%, and for Choco, achieving superior results in 38.6% of instances. It also consistently surpasses Hamming distance search within the same framework, confirming the advantage of model-based exploration over simple local search. Overall, the probe and solve algorithm offers a practical, resource-aware approach for tuning constraint solvers that yields robust improvements across diverse problem types.

cs.AI↗

arXiv2025

Multi-objective methods in Federated Learning: A survey and taxonomy

The Federated Learning paradigm facilitates effective distributed machine learning in settings where training data is decentralized across multiple clients. As the popularity of the strategy grows, increasingly complex real-world problems emerge, many of which require balancing conflicting demands such as fairness, utility, and resource consumption. Recent works have begun to recognise the use of a multi-objective perspective in answer to this challenge. However, this novel approach of combining federated methods with multi-objective optimisation has never been discussed in the broader context of both fields. In this work, we offer a first clear and systematic overview of the different ways the two fields can be integrated. We propose a first taxonomy on the use of multi-objective methods in connection with Federated Learning, providing a targeted survey of the state-of-the-art and proposing unambiguous labels to categorise contributions. Given the developing nature of this field, our taxonomy is designed to provide a solid basis for further research, capturing existing works while anticipating future additions. Finally, we outline open challenges and possible directions for further research.

cs.LG↗

arXiv2025

Trustworthy Efficient Communication for Distributed Learning using LQ-SGD Algorithm

We propose LQ-SGD (Low-Rank Quantized Stochastic Gradient Descent), an efficient communication gradient compression algorithm designed for distributed training. LQ-SGD further develops on the basis of PowerSGD by incorporating the low-rank approximation and log-quantization techniques, which drastically reduce the communication overhead, while still ensuring the convergence speed of training and model accuracy. In addition, LQ-SGD and other compression-based methods show stronger resistance to gradient inversion than traditional SGD, providing a more robust and efficient optimization path for distributed learning systems.

cs.LG↗

arXiv2025

Exhaustive Search for Quantum Circuit Optimization using ZX Calculus

Quantum computers allow a near-exponential speed-up for specific applications when compared to classical computers. Despite recent advances in the hardware of quantum computers, their practical usage is still severely limited due to a restricted number of available physical qubits and quantum gates, short coherence time, and high error rates. This paper lays the foundation towards a metric independent approach to quantum circuit optimization based on exhaustive search algorithms. This work uses depth-first search and iterative deepening depth-first search. We rely on ZX calculus to represent and optimize quantum circuits through the minimization of a given metric (e.g. the T-gate and edge count). ZX calculus formally guarantees that the semantics of the original circuit is preserved. As ZX calculus is a non-terminating rewriting system, we utilise a novel set of pruning rules to ensure termination while still obtaining high-quality solutions. We provide the first formalization of quantum circuit optimization using ZX calculus and exhaustive search. We extensively benchmark our approach on 100 standard quantum circuits. Finally, our implementation is integrated in the well-known libraries PyZX and Qiskit as a compiler pass to ensure applicability of our results.

quant-ph↗

arXiv2025

Trustworthiness of Stochastic Gradient Descent in Distributed Learning

Distributed learning (DL) uses multiple nodes to accelerate training, enabling efficient optimization of large-scale models. Stochastic Gradient Descent (SGD), a key optimization algorithm, plays a central role in this process. However, communication bottlenecks often limit scalability and efficiency, leading to increasing adoption of compressed SGD techniques to alleviate these challenges. Despite addressing communication overheads, compressed SGD introduces trustworthiness concerns, as gradient exchanges among nodes are vulnerable to attacks like gradient inversion (GradInv) and membership inference attacks (MIA). The trustworthiness of compressed SGD remains unexplored, leaving important questions about its reliability unanswered. In this paper, we provide a trustworthiness evaluation of compressed versus uncompressed SGD. Specifically, we conducted empirical studies using GradInv attacks, revealing that compressed SGD demonstrates significantly higher resistance to privacy leakage compared to uncompressed SGD. In addition, our findings suggest that MIA may not be a reliable metric for assessing privacy risks in distributed learning.

cs.LG↗

arXiv2025

Lightweight Trustworthy Distributed Clustering

Ensuring data trustworthiness within individual edge nodes while facilitating collaborative data processing poses a critical challenge in edge computing systems (ECS), particularly in resource-constrained scenarios such as autonomous systems sensor networks, industrial IoT, and smart cities. This paper presents a lightweight, fully distributed k-means clustering algorithm specifically adapted for edge environments, leveraging a distributed averaging approach with additive secret sharing, a secure multiparty computation technique, during the cluster center update phase to ensure the accuracy and trustworthiness of data across nodes.

cs.DC↗

arXiv2025

FedPref: Federated Learning Across Heterogeneous Multi-objective Preferences

Federated Learning (FL) is a distributed machine learning strategy, developed for settings where training data is owned by distributed devices and cannot be shared. FL circumvents this constraint by carrying out model training in distribution. The parameters of these local models are shared intermittently among participants and aggregated to enhance model accuracy. This strategy has been rapidly adopted by the industry in efforts to overcome privacy and resource constraints in model training. However, the application of FL to real-world settings brings additional challenges associated with heterogeneity between participants. Research into mitigating these difficulties in FL has largely focused on only two types of heterogeneity: the unbalanced distribution of training data, and differences in client resources. Yet more types of heterogeneity are becoming relevant as the capability of FL expands to cover more complex problems, from the tuning of LLMs to enabling machine learning on edge devices. In this work, we discuss a novel type of heterogeneity that is likely to become increasingly relevant in future applications: this is preference heterogeneity, emerging when clients learn under multiple objectives, with different importance assigned to each objective on different clients. In this work, we discuss the implications of this type of heterogeneity and propose FedPref, a first algorithm designed to facilitate personalised FL in this setting. We demonstrate the effectiveness of the algorithm across different problems, preference distributions and model architectures. In addition, we introduce a new analytical point of view, based on multi-objective metrics, for evaluating the performance of FL algorithms in this setting beyond the traditional client-focused metrics. We perform a second experimental analysis based in this view, and show that FedPref outperforms compared algorithms.

cs.LG↗

arXiv2025

A Unified Hyperparameter Optimization Pipeline for Transformer-Based Time Series Forecasting Models

Transformer-based models for time series forecasting (TSF) have attracted significant attention in recent years due to their effectiveness and versatility. However, these models often require extensive hyperparameter optimization (HPO) to achieve the best possible performance, and a unified pipeline for HPO in transformer-based TSF remains lacking. In this paper, we present one such pipeline and conduct extensive experiments on several state-of-the-art (SOTA) transformer-based TSF models. These experiments are conducted on standard benchmark datasets to evaluate and compare the performance of different models, generating practical insights and examples. Our pipeline is generalizable beyond transformer-based architectures and can be applied to other SOTA models, such as Mamba and TimeMixer, as demonstrated in our experiments. The goal of this work is to provide valuable guidance to both industry practitioners and academic researchers in efficiently identifying optimal hyperparameters suited to their specific domain applications. The code and complete experimental results are available on GitHub.

cs.LG↗

arXiv2024

Dataset | Mindset = Explainable AI | Interpretable AI

We often use "explainable" Artificial Intelligence (XAI)" and "interpretable AI (IAI)" interchangeably when we apply various XAI tools for a given dataset to explain the reasons that underpin machine learning (ML) outputs. However, these notions can sometimes be confusing because interpretation often has a subjective connotation, while explanations lean towards objective facts. We argue that XAI is a subset of IAI. The concept of IAI is beyond the sphere of a dataset. It includes the domain of a mindset. At the core of this ambiguity is the duality of reasons, in which we can reason either outwards or inwards. When directed outwards, we want the reasons to make sense through the laws of nature. When turned inwards, we want the reasons to be happy, guided by the laws of the heart. While XAI and IAI share reason as the common notion for the goal of transparency, clarity, fairness, reliability, and accountability in the context of ethical AI and trustworthy AI (TAI), their differences lie in that XAI emphasizes the post-hoc analysis of a dataset, and IAI requires a priori mindset of abstraction. This hypothesis can be proved by empirical experiments based on an open dataset and harnessed by High-Performance Computing (HPC). The demarcation of XAI and IAI is indispensable because it would be impossible to determine regulatory policies for many AI applications, especially in healthcare, human resources, banking, and finance. We aim to clarify these notions and lay the foundation of XAI, IAI, EAI, and TAI for many practitioners and policymakers in future AI applications and research.

cs.AI↗

arXiv2024

Heterogeneity: An Open Challenge for Federated On-board Machine Learning

The design of satellite missions is currently undergoing a paradigm shift from the historical approach of individualised monolithic satellites towards distributed mission configurations, consisting of multiple small satellites. With a rapidly growing number of such satellites now deployed in orbit, each collecting large amounts of data, interest in on-board orbital edge computing is rising. Federated Learning is a promising distributed computing approach in this context, allowing multiple satellites to collaborate efficiently in training on-board machine learning models. Though recent works on the use of Federated Learning in orbital edge computing have focused largely on homogeneous satellite constellations, Federated Learning could also be employed to allow heterogeneous satellites to form ad-hoc collaborations, e.g. in the case of communications satellites operated by different providers. Such an application presents additional challenges to the Federated Learning paradigm, arising largely from the heterogeneity of such a system. In this position paper, we offer a systematic review of these challenges in the context of the cross-provider use case, giving a brief overview of the state-of-the-art for each, and providing an entry point for deeper exploration of each issue.

cs.DC↗