arXiv:2605.18937v2 Announce Type: replace
Abstract: Patient-managed Personal Health Records (PHRs) promises to empower patients to better understand their health; but information in the record is complex, potentially hindering insights. In this study, we assess the potential of large language models (LLMs, Gemini 3.0 Flash) to provide helpful answers to user health queries, when provided clinical data from PHRs as context. A total of 2,257 user queries were drawn from 3 different distributions to represent patient questions: shorter web search queries, longer questions derived from templates of chatbot conversations, and questions patients asked to their healthcare team (patient calls). Queries were matched with de-identified PHRs (from a pool of 1,945). Gemini responses were generated (1) without PHR context; (2) with a basic summary of demographics, conditions, and medications; (3) with full, extensive clinical notes. For evaluation, we leveraged an existing rating framework (SHARP), and developed a new framework for specific error modes when interpreting PHRs. Evaluation was performed using autoraters for the full set, and with clinician ratings for a subset (n=95), with both sets of raters knowing the full PHR context. We see significant improvements in the helpfulness of answers to all question types with PHR data (p < 0.001, paired t-test). We also observe potential gains in safety, accuracy, relevance and personalization of answers. Our PHR evaluation framework further identifies gaps in LLM understanding of particular aspects of complex PHRs, such as temporal disorientation, and rare but meaningful confabulations. These results suggest potential for PHR data to help people with a wide range of user needs; and provide a framework for monitoring for gaps in LLM answers based on PHR context. This study motivates further work to assess and realize potential benefits to users from understanding their health records.
Science Journals
arXiv:2606.06266v1 Announce Type: new
Abstract: Hate speech detection is inherently subjective: people from different demographic groups perceive the same content very differently. Collecting enough annotations from multiple demographic groups is costly and difficult to scale. Persona-conditioned Large Language Models (models prompted to adopt a specific demographic identity) have been proposed as a way to simulate diverse perspectives at scale. But do they actually reflect how different groups disagree? We evaluate three aspects of human social judgement: (i) whether personas from different groups disagree in human-like ways (inter-group disagreement), (ii) whether they become more sensitive when content targets their own identity (in-group sensitivity), and (iii) whether they can accurately predict how another group would react (vicarious prediction). Our results show that no model consistently captures all three dimensions, and performance is highly model-dependent and does not emerge reliably from minimal identity prompts alone. However, vicarious prompting with Llama 3.1 yields the highest cross-group agreement in most demographic axes and provides the closest overall approximation to human disagreement patterns, indicating that this configuration may provide a more reliable setting for automatic annotation aligned with human judgements.
arXiv:2606.06268v1 Announce Type: new
Abstract: Physics-informed neural networks (PINNs) have been widely applied to solve Navier-Stokes equations by enforcing outputs and gradients of deep models to satisfy target equations. However, conventional PINNs only constrain the boundary terms by means of the $L^2$-norm when addressing the equations with non-homogeneous boundary conditions. This single constraint strategy may cause inaccurate boundary simulation, further resulting in the decline of prediction accuracy. To resolve this critical issue, this paper proposes an improved physics-informed neural network by correcting the error of the boundary value, which is called Tr-PINNs. Based on the results of nonhomogeneous Stokes problem, the algorithm error analysis of Tr-PINNs is established. The efficacy of the Tr-PINNs algorithm is demonstrated via numerical experiments, which further demonstrate that the Tr-PINNs algorithm achieves a remarkable improvement in computational accuracy.
arXiv:2511.21858v2 Announce Type: replace
Abstract: We propose an asympotically optimal choice of shared concentrated real poles of a family of rational approximants of time-dependent exponential functions $\exp(-tz)$ for $z \geq 0$ and $t$ in a positive time interval $T$. Our result extends a classical result by J.-E. Andersson [J. Approx. Theory, 32(2):85--95, 1981] on the asymptotic best rational approximation of $\exp(-z)$ with real poles. Numerical experiments demonstrate the near-optimality of our choice for various time ranges and for both small and large approximation degrees. An application of the uniform-in-time rational approximation using our proposed concentrated real poles to a linear constant-coefficient initial-value problem is also discussed.
arXiv:2605.30507v2 Announce Type: replace
Abstract: This work introduces a self-optimizing virtual processor (VP) for numerical array programs that shifts parallelization from a manual developer task to a cooperative, agent-like runtime mechanism. Instead of relying on centralized task-graph scheduling, static compiler optimization, or explicitly annotated parallel constructs, the VP uses a decentralized network of cooperative execution segments, derived from the stream of numerical instructions and their data dependencies at runtime. Each segment makes only local decisions about when, where, and how to prepare and execute its computation, including task placement, kernel preparation, and data movement. No central scheduler or mapper instance determines the execution globally; instead, scheduling itself is parallelized and distributed over time - asynchronously and strictly dependency driven. The overall execution strategy emerges from concurrently executing local segments, continuously responding to data availability, cost estimates, system state, hardware capabilities, and problem size. While preserving the sequential semantics of the program our VP automatically exploits parallelism across large program regions rather than being limited to individual loop bodies, modules, or explicitly marked parallel sections; developers are not required to design or encode a parallelization strategy. The current VP primarily targets low-latency strong scaling on local heterogeneous hardware, covering workloads from small, latency-sensitive array operations to large data-parallel computations. The current implementation targets the predefined array instruction set of the ILNumerics ONAL domain-specific language, accessible https://github.com/ILNumerics/ILNumerics.ONAL , while the underlying concept is applicable to general array-based numerical programming models such as MATLAB and NumPy.
arXiv:2606.05591v1 Announce Type: new
Abstract: Indirect reciprocity explains how cooperation evolves through social reputations. People observe others, assign reputations, and condition their future actions on these assignments. This process is cognitively demanding, and stereotyping offers a simpler alternative by replacing individual-level reputation with group-level information. Theoretical models commonly implement stereotyping through exogenously given group labels. In real societies, however, group-level impressions may be associated with observable patterns of behavior. Here we propose a framework of stereotyping by strategy standing, in which mutants may condition their actions on the overall reputation level associated with a resident strategy rather than on the recipient's reputation. We show that this form of stereotyping can diversify stable cooperation in indirect reciprocity. As the strength of stereotyping increases, additional cooperative evolutionarily stable norm-strategy (ESS) pairs emerge in substantial numbers. In particular, we identify eight highly cooperative ESS pairs that become stable under very weak stereotyping. These pairs, which we call the counterparts of the leading eight, share the same social norms as the classical leading eight and differ only in how they prescribe behavior between bad individuals. They are unstable without stereotyping because they can be invaded by their corresponding leading strategies, but they become stable once stereotyping exceeds a critical threshold. Our results suggest that group-level impressions based on strategy standing can provide a coarse-grained informational route to stable cooperation and offer a more behaviorally grounded perspective on how stereotyping affects indirect reciprocity.
arXiv:2605.30467v2 Announce Type: replace
Abstract: This study introduces a novel Arctic-focused remote sensing foundation model (RSFM) by combining diversity-aware regional-scale image curation with masked autoencoder (MAE) self-supervised pretraining of a Vision Transformer (ViT) encoder for very-high-spatial-resolution (VHSR) satellite image analysis. Spectral and acquisition-metadata descriptors were used in a scalable affinity-propagation clustering workflow to select approximately 3 million chips from 267 TB of Vantor VHSR imagery This curation strategy was designed to reduce oversampling of visually repetitive or low-information areas while preserving broad scene diversity across the study domain. We pretrained a ViT-Large encoder on the curated corpus using a domain-adapted MAE reconstruction objective, producing Arctic-specific transformer weights for downstream feature mapping. The pretrained encoder was integrated into an existing location-aware detection and segmentation framework and evaluated across four hand-labeled Arctic datasets. Compared to ImageNet-initialized ViT-Large baseline, Arctic MAE pretraining produced consistent improvements in foreground mean F1 scores of 0.87, 0.72, 0.93, and 0.87, for infrastructure, IWP, RTS, and TCNs, with approximately 5-8 percentage increase. The proposed model also outperformed Prithvi-EO-2.0 in all downstream comparisons, with the smallest gain corresponding to at least a 15 percentage improvement mean F1, suggesting that domain-specific self-supervised pretraining on curated Arctic VHSR imagery provides more transferable representations for fine-scale Arctic mapping than a general-purpose Earth observation foundation model. These results demonstrate that optimizing the pretraining data distribution at regional scale, while keeping the architecture and MAE objective fixed, can produce a reusable Arctic-domain encoder for multiple VHSR remote sensing applications.
arXiv:2605.29054v2 Announce Type: replace
Abstract: Coding agents increasingly act as codebase-scale collaborators that can assist with codebase conversion, but this progress has exposed a critical weakness: agents often over-trust their own local validation routines and declare success on artifacts that satisfy surface checks while violating the semantic contracts users actually care about. This problem is especially acute in codebase conversion, where prior evaluation is largely outcome-driven and therefore unstable: two implementations can match on a shallow outcome, such as a single forward loss, while diverging in gradients, optimizer behavior, or short-horizon training dynamics. We introduce T2J-Bench, a benchmark for codebase conversion that reformulates conversion as transfer under a fixed equivalence contract. A fixed verifier then compares source and converted codebases through three ordered stages: Spec (interface admissibility), Numeric (forward outputs, losses, gradients, and objective-specific tensors), and Behavioral (short training dynamics under fixed seeds). Across 355 blind conversion attempts, the best system reaches only 26.7--28.9% overall pass rate despite Spec pass rates up to 91.1%; a 4.7x token-budget spread yields only a 2.2x pass-rate spread; and all systems overestimate success by 66.6--97.8 points relative to the fixed evaluator. This suggests that failures stem more from contract-misaligned self-validation than from limited budget or backbone strength.
arXiv:2606.05609v1 Announce Type: new
Abstract: As large language models (LLMs) are widely deployed, identifying their vulnerability through jailbreak attacks becomes increasingly critical. Optimization-based attacks like Greedy Coordinate Gradient (GCG) have focused on inserting adversarial tokens to the end of prompts. However, GCG restricts adversarial tokens to a fixed insertion point (typically the prompt suffix), leaving the effect of inserting tokens at other positions unexplored. In this paper, we empirically investigate \emph{slots}, i.e., candidate positions within a prompt where tokens can be inserted. We find that vulnerability to jailbreaking is highly related to the selection of the \emph{slots}. Based on these findings, we introduce the \textit{Vulnerable Slot Score} (VSS) to quantify the positional vulnerability to jailbreaking. We then propose SlotGCG, which evaluates all slots with VSS, selects the most vulnerable slots for insertion, and runs a targeted optimization attack at those slots. Our approach provides a position-search mechanism that is attack-agnostic and can be plugged into any optimization-based attack, adding only 200ms of preprocessing time. Experiments across multiple models demonstrate that SlotGCG significantly outperforms existing methods. Specifically, it achieves 14\% higher Attack Success Rates (ASR) over GCG-based attacks, converges faster, and shows superior robustness against defense methods with 42\% higher ASR than baseline approaches. Our implementation is available at \href{https://github.com/youai058/SlotGCG}{https://github.com/youai058/SlotGCG}
arXiv:2606.05339v1 Announce Type: new
Abstract: MCP (Model Context Protocol) enables LLMs (Large Language Models) to interact with external tools and data sources via a standardized protocol. Its rapid adoption in tool-augmented Artificial Intelligence (AI) workflows has introduced new reliability challenges, such as configuration parameters that are accepted but not enforced at runtime, leading to unintended default behavior, whose runtime fault characteristics remain empirically unexamined. We present the first empirical taxonomy of runtime faults in MCP servers. We manually analyzed 837 MCP-specific runtime fault threads from 473 actively maintained MCP server GitHub repositories and derived a taxonomy using a bottom-up open coding procedure. The taxonomy comprises 11 top-level categories and 27 subcategories (73 leaf fault types), covering recurrent failures across protocol interactions, tool invocations, schema enforcement, state management, model-provider integration, security validation, and timeouts or explicit cancellations of in-progress operations. To assess the taxonomy's external validity, we surveyed 55 MCP server developers. Respondents reported experiencing an average of 20 of the 27 fault subcategories, and no category remained unobserved. These results indicate that the taxonomy reflects widely observed runtime failures in MCP-based systems and shall assist AI software maintenance and evolution in the future.
arXiv:2606.05346v1 Announce Type: new
Abstract: Human language comprehension unfolds sequentially: each word is processed in the context of those that came before, and the interpretation builds incrementally over time. Surprisal, the negative log probability of a word given its context, has been the dominant predictor of incremental processing cost. But surprisal reduces rich sequential representations to a single scalar at each word, discarding information about the direction in which the interpretation has been evolving. Dynamical-systems approaches suggest that the trajectory of the evolving interpretive state, not just its position at each moment,should shape processing, and language itself may have local momentum, since speakers plan utterances a few words at a time. We introduce trajectory extrapolation error: at each word, we fit a linear trajectory to the preceding hidden states of a transformer language model and measure deviation from the extrapolated path. On the Natural Stories corpus, this measure is nearly orthogonal to surprisal (r = .044) and independently predicts self-paced reading times. The effect is especially pronounced in garden-path sentences, strengthens with model scale (GPT-2 Small to Large), and replicates across architectures with different positional encoding schemes (GPT-2 vs. Pythia/RoPE). A displacement control shows the effect is not reducible to representational change magnitude: displacement and extrapolation error predict in opposite directions. These findings reveal two dissociable components of processing cost: word-level prediction error (surprisal) and sensitivity to the local momentum of the unfolding interpretation (trajectory extrapolation error).
arXiv:2606.06338v1 Announce Type: new
Abstract: Video question answering (VideoQA) aims to answer questions about given videos. While existing approaches excel on factoid VideoQA, they struggle with deep video understanding (DVU), which requires the comprehension of complex storylines. This challenge arises from the inherent long-range video content, multi-faceted question types, and instance-level story elements, all of which constrain the scale and diversity of manually constructed DVU datasets.These difficulties constrain the scale and diversity of manually-constructed DVU dataset. To address these, we previously introduced StoryMind to automatically construct DVU datasets with balanced fine-grained topics. Though it can generate high-quality question-answer pairs (QAs) for TV series, it suffers significant performance degradation when handling longer and more complex movies. In this paper, we further design StoryMindv2, an enhanced multi-agent collaboration framework to generate high-quality DVU datasets for both TV series and movies. By integrating a novel supervisor-guided generation mechanism and a refined multi-reviewer voting strategy, the framework is utilized to construct StoryVideoQA, the largest DVU dataset to date, featuring over 363K QAs on 393.2 hours diverse story videos including TV series (avg. 1,635 seconds) and movies (avg. 7,878 seconds). Comprehensive evaluations of 20 state-of-the-art VideoQA methods on this large-scale benchmark reveal that they cannot fully maintain long-range character associations or construct a coherent understanding of complex storylines. To bridge this gap, we propose PlotTree, a novel video understanding agent, re-organizing long-range video content into a hierarchical plot structure, enabling efficient storyline reasoning on StoryVideoQA. Project page: https://github.com/nercms-mmap/StoryVideoQA/
arXiv:2606.05661v1 Announce Type: new
Abstract: Continual learning, the ability of AI systems to improve through sequential experience, has attracted substantial interest, but no high-quality benchmark exists to evaluate it. We introduce Continual Learning Bench (CL-Bench), the first difficult, expert-validated benchmark designed to measure whether LLM-based systems genuinely improve with experience. CL-Bench spans six diverse domains (software engineering, signal processing, disease outbreak forecasting, database querying, strategic game-playing, and demand forecasting), each validated by domain experts and designed so that tasks share a learnable latent structure (codebase layout, disease outbreak dynamics, opponent strategies) that a stateful system can discover online but a stateless one cannot. We evaluate frontier models across several agent architectures, from naive in-context learning (ICL) to dedicated memory systems, introducing a gain metric to isolate learning from prior capabilities. We find that these systems leave headroom for improved continual learning: agents frequently overfit to immediate observations or fail to reuse knowledge across instances, and dedicated memory systems do not fix this -- in fact, naive ICL outperforms systems dedicated to memory management. CL-Bench is the first benchmark to evaluate continual learning across diverse real-world domains with expert-validated tasks and isolate online learning from underlying model capability, showing a need for better continual learning systems.
arXiv:2606.06440v1 Announce Type: new
Abstract: Data-driven causal relationship identification is pertinent to advancing understanding of complex systems both within and beyond science. Bayesian networks offer a probabilistic method for modelling generic causal relationships via directed acyclic graphs (DAGs). However, typical techniques for constructing Bayesian networks rely on optimization, which can be ill-suited for learning causal relationships because the underlying data may admit multiple chains of causation. More data-faithful representations of causal relationships would provide frameworks for constructing multiple causal maps that are consistent with the variability that is inherent in underlying data. Here, we show that entropy-based inference generates atlases of plausible causal relationships that are consistent with underlying data. On simulated noisy data of 2- and 20-node linear structural equation models, we sample a maximum-entropy ensemble of graphs that allow us to quantify the inherent structural ambiguity in underlying causal relationships. Our method shows that "optimized" DAGs can contain causal artifacts are not consistent across equivalently accurate topologies.
arXiv:2606.05648v1 Announce Type: new
Abstract: The k-nearest neighbor query (kNNQ) is a core component of modern location-based services (LBS) and has been widely adopted in popular features such as ``people nearby''. However, its potential privacy risks have long been overlooked. In this work, we present the first two attacks against kNNQ, namely the geometric intersection location inference attack (GI-LIA) and the zero-order optimization location inference attack (ZO-LIA), revealing the inherent location privacy risks posed by kNNQ. To mitigate these privacy risks, we further propose DPRS, a differential privacy framework for kNNQ protection. The core idea of DPRS is to incorporate a rejection sampling mechanism within a constrained perturbation interval, thereby mitigating the distance distortion caused by excessive noise injection. In addition, we design a private interval construction algorithm to construct the perturbation interval, enabling the rejection sampling mechanism to achieve a more favorable trade-off between privacy protection and query utility in kNNQ. Extensive experiments on real-world spatial datasets demonstrate that DPRS outperforms existing methods in both privacy protection and query utility. Our code is available at https://github.com/reanatom/DPRS.
arXiv:2606.05649v1 Announce Type: cross
Abstract: Scenario generation is a critical component in stochastic programming (SP), as it directly influences the quality of decision-making under uncertainty. Existing approaches predominantly rely on either sampling-based techniques or supervised learning using neural networks. Sampling-based techniques often struggle to capture complex dependencies and rare but plausible events, while supervised learning requires fixed input-output pairs for training and is limited in its ability to generate a wide variety of realistic scenarios that are not restricted by predefined patterns or rules. To address these limitations, we introduce Diff2SP, a diffusion-based generative framework that incorporates downstream optimization objectives directly into scenario generation. Unlike conventional methods that treat scenario generation and decision-making as separate steps, Diff2SP embeds stochastic optimization into the training process, enabling the generation of scenarios that are both statistically coherent and decision-aware. To formally justify this optimization-aware design, we establish a regret bounds that link distributional accuracy to decision quality, and establish sample complexity guarantees showing faster convergence than traditional generative models such as GANs. Empirical results on both synthetic and power-system datasets validate these theoretical insights, demonstrating that Diff2SP consistently improves both statistical fidelity and downstream optimization outcomes.
arXiv:2606.05421v1 Announce Type: new
Abstract: When a text is translated, does the translation retain the complexity of the original? We introduce ComplexityMT, a new challenge for assessing how text complexity and machine translation interact with and influence each other, using the Common European Framework of Reference for Languages (CEFR) levels as the measure of text complexity. Across six languages, including Arabic, Dutch, English, French, Hindi, and Russian, we evaluate three open-weight models, one closed model, and a commercial machine translation system on two tasks: i) correlation of CEFR with translation difficulty, and ii) shifts in CEFR levels of the source texts. Our experiments show that higher CEFR levels make texts more difficult to translate, and that machine translation shifts the CEFR level of the target text compared to the original source, for most languages. These findings provide new insights for researchers and practitioners working on multilingual pedagogical content generation and machine translation difficulty estimation.
arXiv:2606.05705v1 Announce Type: cross
Abstract: Turning a promising economic idea into a credible empirical finding is, in practice, an expensive undertaking: it demands a great deal of specialised computation, and the results are seldom released in a form that others can check or build upon. Econstellar is our response. It is an open, publicly served research engine that runs publication-grade financial econometrics from an ordinary web browser and explains what the results mean, so that a reader does not merely read a finding but can re-run it, vary its inputs, and trace exactly how it was produced. Three choices give the system its character. The heavy computation is placed on the processor that suits it, rather than forced onto hardware ill-matched to the task, which is much of the reason analysis of this kind is so rarely served to the public. An artificial-intelligence assistant selects and interprets the analyses but never originates a number, so every quantity it reports is a real computation the reader can reproduce. And the engine a visitor exercises is the same code that produced the figures in our published research. We expose seventeen econometric methods, each reported with a verified live value and reproducible at the public endpoint, computed under a single discipline: prices are treated as non-stationary and all methods are applied to returns. The system also regenerates, on demand, the headline result of an accompanying study of financial contagion, from the package that generated it. The platform is the working core of an active research programme spanning three software releases and three preprints, and it is available now, free and open-source, at a live public address. Our aim is a simple one: to shorten the distance between a research claim and the moment another person can independently verify it.
arXiv:2603.04891v2 Announce Type: replace
Abstract: Context: Open Source Software (OSS) is a crucial component of over 90\% of digital infrastructure underpinning industry and public digital services, facilitating collaborative software development and dissemination. Its significance in the European public sector has been emphasised through various Ministerial Declarations, highlighting its potential to accelerate digitalisation, transform businesses, and foster a digitally skilled population. Research Aim: This study aims to explore how the adoption, development, and collaboration on OSS can be enabled through organisational support functions or centres of competency, also known as Open Source Programme Offices (OSPOs) within Public Sector Organisations (PSOs) in the European Union, Norway, Liechtenstein, and Iceland. Methodology: A qualitative research approach was adopted, involving an interview survey of 18 OSPO representatives across 16 cases of public-sector OSPOs. These cases were cross-analysed and categorised into six OSPO archetypes. The findings were validated and enriched through two follow-up focus groups that included earlier interviewees and additional experts. Results: The study identified six distinct OSPO archetypes, providing insights into their organisational structures, responsibilities, and contributions to OSS adoption. The archetypes, along with policy recommendations, offer guidance on how PSOs can design their own OSPOs, taking into account their specific context, resources, and policy goals. Conclusions: The findings enhance the understanding of OSPOs as strategic endeavours aimed at promoting OSS adoption. The study offers practical guidance for PSOs and policymakers on leveraging OSS to achieve strategic objectives, foster digital sovereignty, drive economic growth, and improve the interoperability and quality of digital services.
arXiv:2606.05862v1 Announce Type: cross
Abstract: Two dimensional site interaction models of water and alcohols are mixed in various proportions and studied by Monte Carlo simulations, with the purpose to clarify problems related to simulation of real micro-heterogeneous systems. Three alcohols are considered, methanol, pentanol and octanol. The main finding is that, while real alcohols demix with water from butanol onward, their 2D analogs are always fully miscible, while developing increasingly pronounced micro-segregation as the alcohol tail length increases. This is not a consequence of the intrinsically higher fluctuations in 2D, but rather a reorganization of these fluctuations under the charge ordering mechanism. The second finding is that water drives the micro-segregation through strong self-aggregation, but this is not enough to achieve full phase separation because of the water-alcohol contact at the outer rim of the water domains. In this work we examine how this local heterogeneity develops with increasing alcohol alkyl tails, monitored with the study of pair correlation functions, structure factors and Kirkwood-Buff integrals. The absence of clear local self-averaging of the latter provides an illustration of the tension between energy driven maintaining of local structures and entropy driven global homogeneity. In that, the 2D modelisation of real hydrogen bonding mixtures allows to better capture and reveal the physics behind the chemistry of these liquids.
arXiv:2606.05870v1 Announce Type: cross
Abstract: Neurodegenerative disorders such as Alzheimer's disease exhibit highly organized patterns of regional brain vulnerability, yet the biological mechanisms underlying this spatial selectivity remain incompletely understood. Existing imaging-transcriptomic studies have largely relied on correlation-based analyses between gene expression and neuroimaging phenotypes, limiting their ability to model how molecular organization gives rise to neurodegeneration. Here, we introduce a cross-scale spatially-aware generative framework for modeling transcriptomic programs underlying cortical neurodegeneration. Regional transcriptomic profiles were derived from the Allen Human Brain Atlas using 910 landmark genes across 68 cortical regions. Neurodegenerative vulnerability maps were constructed from ADNI FreeSurfer cortical thickness measurements by computing regional cortical thinning differences between cognitively normal controls (NC = 926) and Alzheimer's disease subjects (AD = 426). A variational generative architecture was used to learn latent biological programs linking regional gene-expression organization to cortical degeneration while incorporating graph-based spatial smoothness regularization to preserve cortical organization. The proposed framework achieved strong prediction of regional neurodegenerative vulnerability, yielding an explained variance of 0.8604 and a significant spatial correlation between predicted and observed cortical degeneration profiles (r = 0.9439, p < 0.001). The learned latent representations revealed structured transcriptomic organization associated with distributed disease susceptibility. These findings demonstrate that biologically constrained generative modeling can bridge microscale molecular organization with macroscale neurodegeneration, providing a foundation for spatially-aware generative neurobiology and computational neuroscience.
arXiv:2606.05247v1 Announce Type: new
Abstract: Enforcing nonlinear inequality constraints in neural networks remains challenging, especially when the output is subject to many coupled constraints. Existing hard constraint methods often impose structural restrictions on the constraint set or introduce substantial computational overhead for large-scale nonlinear problems. Here, we propose DiffSlack, a differentiable projection layer for nonlinear inequality-constrained neural prediction. DiffSlack reformulates inequalities as equalities with learnable slack variables, which are predicted as part of the augmented network output and provide a data-driven warm start for damped Gauss-Newton projection. The projection layer maps raw predictions onto the augmented feasible manifold while preserving end-to-end differentiability. A two-stage curriculum further stabilizes training and improves constraint satisfaction. We evaluate DiffSlack on vehicle path planning with 200 nonlinear inequality constraints from collision avoidance, curvature limits, and waypoint spacing. Compared with existing learning-based baselines, DiffSlack achieves a higher planning success rate and stronger geometric constraint satisfaction under a comparable inference budget. Ablation studies further show that the hard projection layer reduces sensitivity to supervision quality. Closed-loop tracking in CARLA and real-world vehicle experiments confirms the executability of the generated trajectories. These results demonstrate that DiffSlack provides a practical and scalable approach to embedding hard inequality constraints into neural networks for engineering applications.
arXiv:2606.05499v1 Announce Type: new
Abstract: Mutation testing is widely used to evaluate test-suite effectiveness, yet IEC 61131-3 Structured Text (ST) programs still lack a publicly available benchmark that supports reproducible mutation-based research. This gap is especially important because ST is extensively used in Programmable Logic Controllers (PLCs) that operate in real-time, safety-critical industrial environments, where software faults may cause equipment damage, production loss, or unsafe system behavior. To address this need, we present STMutants, a curated mutation testing dataset for industrial automation software. STMutants contains 110 generated first-order mutants derived from 11 ST programs collected from the OSCAT basic library and industrially relevant sources, of which 108 are retained after observability and equivalence screening. The dataset covers seven mutation operator categories adapted from classical taxonomies for the PLC domain, including value, relational, arithmetic, logical, negation, operation insertion/omission, and initialization faults. Each mutant is constructed through a four-phase methodology: fault-type profiling and operator selection, syntactic transformation, compilability verification, and manual equivalence screening with strong inter-rater agreement (kappa = 0.87). To demonstrate the usefulness of the dataset, we evaluate three large language models (LLMs) in a two-phase setting: test-suite generation followed by mutation kill/survive prediction. Across 108 retained mutants, the models achieve mutation detection accuracies of 86.1%, 94.4%, and 86.1%, respectively, with statistical analysis confirming significant performance differences. By providing the first publicly available mutation benchmark for ST programs, STMutants enables reproducible research on automated test generation, mutation analysis, fault localization, and AI-assisted quality assurance for PLC software.
arXiv:2606.05335v1 Announce Type: new
Abstract: Theoretical studies of machine learning models commonly consider different limiting regimes in which the learning dynamics of gradient descent becomes theoretically tractable. It is, however, desirable to have a systematically obtained picture of all qualitatively different extreme learning regimes for a particular type of models. In this paper we propose such a picture for large weight-tied linear autoencoders characterized by input and latent dimensions, initialization magnitude, and training set size. This model is nonlinear in the weights and its gradient flow does not have a general theoretical solution. We show that at the level of the formal loss-expansion hierarchy, its extreme regimes are naturally associated with faces of a triangular prism. In particular, there are five basic extreme regimes associated with the 2-faces of the prism: (1) large-data, (2) small-data, (3) mean-field, (4) narrow-latent, and (5) free. For regimes (1,2,3,4), we derive explicit expressions for both train and population limiting loss evolutions under gradient flow, obtaining very good agreement with experimental results.
arXiv:2606.05445v1 Announce Type: new
Abstract: We dream of AI agents that can read arbitrary designs and construct real-world objects from reusable building blocks. As a first step toward this vision, we study whether multimodal large language models (MLLMs) possess the visual grounding and spatial reasoning capabilities required for brick assembly. We formulate brick assembly as a sequential decision-making problem, where each step involves two subtasks: brick selection, identifying the target brick from candidate components, and brick pose estimation, predicting where and how the selected brick should be placed. To support this study, we introduce BC-Bench (Brick Construction Benchmark), the first benchmark for evaluating MLLMs on assembly with diverse bricks. Experiments show that current state-of-the-art MLLMs remain far from reliable builders, struggling with fine-grained brick selection and failing at precise pose estimation. To bridge this gap, we propose Brick-Composer, a learning framework that equips MLLMs with assembly skills through three complementary signals: Human Design Sparks, which provide affordance-rich construction demonstrations; World Feedback, which grounds predicted actions in visual and physical consequences; and Synthetic Experience, which scales learning beyond existing object designs. Brick-Composer improves brick selection accuracy by over three times, substantially reduces pose estimation errors, and raises strict step-level assembly success from less than 1% to around 15%. After training, a Qwen-3-8B can correctly compose up to 42% of the steps for a complete object, suggesting that MLLMs can acquire assembly capabilities through targeted, physically grounded learning.