arXiv Daily Digest - 2026-02-24

Rapid progress in video models has largely focused on visual quality, leaving their reasoning capabilities underexplored. Video reasoning grounds intelligence in spatiotemporally consistent visual environments that go beyond what text can naturally capture, enabling intuitive reasoning over spatiotemporal structure such as continuity, interaction, and causality. However, systematically studying video reasoning and its scaling behavior is hindered by the lack of large-scale training data. To address this gap, we introduce the Very Big Video Reasoning (VBVR) Dataset, an unprecedentedly large-scale resource spanning 200 curated reasoning tasks following a principled taxonomy and over one million video clips, approximately three orders of magnitude larger than existing datasets. We further present VBVR-Bench, a verifiable evaluation framework that moves beyond model-based judging by incorporating rule-based, human-aligned scorers, enabling reproducible and interpretable diagnosis of video reasoning capabilities. Leveraging the VBVR suite, we conduct one of the first large-scale scaling studies of video reasoning and observe early signs of emergent generalization to unseen reasoning tasks. Together, VBVR lays a foundation for the next stage of research in generalizable video reasoning. The data, benchmark toolkit, and models are publicly available at https://video-reason.com/ .

LLM agents are evolving rapidly, powered by code execution, tools, and the recently introduced agent skills feature. Skills allow users to extend LLM applications with specialized third-party code, knowledge, and instructions. Although this can extend agent capabilities to new domains, it creates an increasingly complex agent supply chain, offering new surfaces for prompt injection attacks. We identify skill-based prompt injection as a significant threat and introduce SkillInject, a benchmark evaluating the susceptibility of widely-used LLM agents to injections through skill files. SkillInject contains 202 injection-task pairs with attacks ranging from obviously malicious injections to subtle, context-dependent attacks hidden in otherwise legitimate instructions. We evaluate frontier LLMs on SkillInject, measuring both security in terms of harmful instruction avoidance and utility in terms of legitimate instruction compliance. Our results show that today's agents are highly vulnerable with up to 80% attack success rate with frontier models, often executing extremely harmful instructions including data exfiltration, destructive action, and ransomware-like behavior. They furthermore suggest that this problem will not be solved through model scaling or simple input filtering, but that robust agent security will require context-aware authorization frameworks. Our benchmark is available at https://www.skill-inject.com/.

We study post-calibration uncertainty for trained ensembles of classifiers. Specifically, we consider both aleatoric (label noise) and epistemic (model) uncertainty. Among the most popular and widely used calibration methods in classification are temperature scaling (i.e., pool-then-calibrate) and conformal methods. However, the main shortcoming of these calibration methods is that they do not balance the proportion of aleatoric and epistemic uncertainty. Not balancing these uncertainties can severely misrepresent predictive uncertainty, leading to overconfident predictions in some input regions while being underconfident in others. To address this shortcoming, we present a simple but powerful calibration algorithm Joint Uncertainty Calibration (JUCAL) that jointly calibrates aleatoric and epistemic uncertainty. JUCAL jointly calibrates two constants to weight and scale epistemic and aleatoric uncertainties by optimizing the negative log-likelihood (NLL) on the validation/calibration dataset. JUCAL can be applied to any trained ensemble of classifiers (e.g., transformers, CNNs, or tree-based methods), with minimal computational overhead, without requiring access to the models' internal parameters. We experimentally evaluate JUCAL on various text classification tasks, for ensembles of varying sizes and with different ensembling strategies. Our experiments show that JUCAL significantly outperforms SOTA calibration methods across all considered classification tasks, reducing NLL and predictive set size by up to 15% and 20%, respectively. Interestingly, even applying JUCAL to an ensemble of size 5 can outperform temperature-scaled ensembles of size up to 50 in terms of NLL and predictive set size, resulting in up to 10 times smaller inference costs. Thus, we propose JUCAL as a new go-to method for calibrating ensembles in classification.

Inspired by behavioral science, we propose Behavior Learning (BL), a novel general-purpose machine learning framework that learns interpretable and identifiable optimization structures from data, ranging from single optimization problems to hierarchical compositions. It unifies predictive performance, intrinsic interpretability, and identifiability, with broad applicability to scientific domains involving optimization. BL parameterizes a compositional utility function built from intrinsically interpretable modular blocks, which induces a data distribution for prediction and generation. Each block represents and can be written in symbolic form as a utility maximization problem (UMP), a foundational paradigm in behavioral science and a universal framework of optimization. BL supports architectures ranging from a single UMP to hierarchical compositions, the latter modeling hierarchical optimization structures. Its smooth and monotone variant (IBL) guarantees identifiability. Theoretically, we establish the universal approximation property of BL, and analyze the M-estimation properties of IBL. Empirically, BL demonstrates strong predictive performance, intrinsic interpretability and scalability to high-dimensional data. Code: https://github.com/MoonYLiang/Behavior-Learning ; install via pip install blnetwork.

Conformal risk control is an extension of conformal prediction for controlling risk functions beyond miscoverage. The original algorithm controls the expected value of a loss that is monotonic in a one-dimensional parameter. Here, we present risk control guarantees for generic algorithms applied to possibly non-monotonic losses with multidimensional parameters. The guarantees depend on the stability of the algorithm -- unstable algorithms have looser guarantees. We give applications of this technique to selective image classification, FDR and IOU control of tumor segmentations, and multigroup debiasing of recidivism predictions across overlapping race and sex groups using empirical risk minimization.

We present AgentOptics, an agentic AI framework for high-fidelity, autonomous optical system control built on the Model Context Protocol (MCP). AgentOptics interprets natural language tasks and executes protocol-compliant actions on heterogeneous optical devices through a structured tool abstraction layer. We implement 64 standardized MCP tools across 8 representative optical devices and construct a 410-task benchmark to evaluate request understanding, role-aware responses, multi-step coordination, robustness to linguistic variation, and error handling. We assess two deployment configurations--commercial online LLMs and locally hosted open-source LLMs--and compare them with LLM-based code generation baselines. AgentOptics achieves 87.7%--99.0% average task success rates, significantly outperforming code-generation approaches, which reach up to 50% success. We further demonstrate broader applicability through five case studies extending beyond device-level control to system orchestration, monitoring, and closed-loop optimization. These include DWDM link provisioning and coordinated monitoring of coherent 400 GbE and analog radio-over-fiber (ARoF) channels; autonomous characterization and bias optimization of a wideband ARoF link carrying 5G fronthaul traffic; multi-span channel provisioning with launch power optimization; closed-loop fiber polarization stabilization; and distributed acoustic sensing (DAS)-based fiber monitoring with LLM-assisted event detection. These results establish AgentOptics as a scalable, robust paradigm for autonomous control and orchestration of heterogeneous optical systems.

Mean Field Games (MFGs) provide a principled framework for modeling interactions in large population models: at scale, population dynamics become deterministic, with uncertainty entering only through aggregate shocks, or common noise. However, algorithmic progress has been limited since model-free methods are too high variance and exact methods scale poorly. Recent Hybrid Structural Methods (HSMs) use Monte Carlo rollouts for the common noise in combination with exact estimation of the expected return, conditioned on those samples. However, HSMs have not been scaled to Partially Observable settings. We propose Recurrent Structural Policy Gradient (RSPG), the first history-aware HSM for settings involving public information. We also introduce MFAX, our JAX-based framework for MFGs. By leveraging known transition dynamics, RSPG achieves state-of-the-art performance as well as an order-of-magnitude faster convergence and solves, for the first time, a macroeconomics MFG with heterogeneous agents, common noise and history-aware policies. MFAX is publicly available at: https://github.com/CWibault/mfax.

With the rise of large language models (LLMs), they have become instrumental in applications such as Retrieval-Augmented Generation (RAG). Yet evaluating these systems remains bottlenecked by the time and cost of building specialized assessment datasets. We introduce KNIGHT, an LLM-based, knowledge-graph-driven framework for generating multiple-choice question (MCQ) datasets from external sources. KNIGHT constructs a topic-specific knowledge graph, a structured and parsimonious summary of entities and relations, that can be reused to generate instructor-controlled difficulty levels, including multi-hop questions, without repeatedly re-feeding the full source text. This knowledge graph acts as a compressed, reusable state, making question generation a cheap read over the graph. We instantiate KNIGHT on Wikipedia/Wikidata while keeping the framework domain- and ontology-agnostic. As a case study, KNIGHT produces six MCQ datasets in History, Biology, and Mathematics. We evaluate quality on five criteria: fluency, unambiguity (single correct answer), topic relevance, option uniqueness, and answerability given the provided sources (as a proxy for hallucination). Results show that KNIGHT enables token- and cost-efficient generation from a reusable graph representation, achieves high quality across these criteria, and yields model rankings aligned with MMLU-style benchmarks, while supporting topic-specific and difficulty-controlled evaluation.

Epidemiological models increasingly rely on self-reported behavioral data such as vaccination status, mask usage, and social distancing adherence to forecast disease transmission and assess the impact of non-pharmaceutical interventions (NPIs). While such data provide valuable real-time insights, they are often subject to strategic misreporting, driven by individual incentives to avoid penalties, access benefits, or express distrust in public health authorities. To account for such human behavior, in this paper, we introduce a game-theoretic framework that models the interaction between the population and a public health authority as a signaling game. Individuals (senders) choose how to report their behaviors, while the public health authority (receiver) updates their epidemiological model(s) based on potentially distorted signals. Focusing on deception around masking and vaccination, we characterize analytically game equilibrium outcomes and evaluate the degree to which deception can be tolerated while maintaining epidemic control through policy interventions. Our results show that separating equilibria-with minimal deception-drive infections to near zero over time. Remarkably, even under pervasive dishonesty in pooling equilibria, well-designed sender and receiver strategies can still maintain effective epidemic control. This work advances the understanding of adversarial data in epidemiology and offers tools for designing more robust public health models in the presence of strategic user behavior.

The paradigm of automated program generation is shifting from one-shot generation to inference-time search, where Large Language Models (LLMs) function as semantic mutation operators within evolutionary loops. While effective, these systems are currently governed by static schedules that fail to account for the non-stationary dynamics of the search process. This rigidity results in substantial computational waste, as resources are indiscriminately allocated to stagnating populations while promising frontiers remain under-exploited. We introduce AdaEvolve, a framework that reformulates LLM-driven evolution as a hierarchical adaptive optimization problem. AdaEvolve uses an "accumulated improvement signal" to unify decisions across three levels: Local Adaptation, which dynamically modulates the exploration intensity within a population of solution candidates; Global Adaptation, which routes the global resource budget via bandit-based scheduling across different solution candidate populations; and Meta-Guidance which generates novel solution tactics based on the previously generated solutions and their corresponding improvements when the progress stalls. We demonstrate that AdaEvolve consistently outperforms the open-sourced baselines across 185 different open-ended optimization problems including combinatorial, systems optimization and algorithm design problems.

Current reinforcement learning objectives for large-model reasoning primarily focus on maximizing expected rewards. This paradigm can lead to overfitting to dominant reward signals, while neglecting alternative yet valid reasoning trajectories, thereby limiting diversity and exploration. To address this issue, we introduce Learning Advantage Distributions (LAD), a distribution-matching framework that replaces advantage maximization with learning the advantage-induced distribution. By establishing the equivalence between the optimal policy update and an advantage-based target distribution, we derive a practical LAD objective formulated as minimizing an $f$-divergence between the policy-induced and advantage-induced distributions. This yields a gradient update that increases likelihood for high-advantage responses while suppressing over-confident probability growth, preventing collapse without requiring auxiliary entropy regularization. LAD incurs no extra training cost compared to GRPO and scales naturally to LLM post-training. In a controlled bandit setting, LAD faithfully recovers the multimodal advantage distribution, validating the theoretical formulation. Experiments on math and code reasoning tasks across several LLM backbones show that LAD reliably improves both accuracy and generative diversity.

Objective: To improve the efficiency of medical question answering (MedQA) with large language models (LLMs) by avoiding unnecessary reasoning while maintaining accuracy. Methods: We propose Selective Chain-of-Thought (Selective CoT), an inference-time strategy that first predicts whether a question requires reasoning and generates a rationale only when needed. Two open-source LLMs (Llama-3.1-8B and Qwen-2.5-7B) were evaluated on four biomedical QA benchmarks-HeadQA, MedQA-USMLE, MedMCQA, and PubMedQA. Metrics included accuracy, total generated tokens, and inference time. Results: Selective CoT reduced inference time by 13-45% and token usage by 8-47% with minimal accuracy loss ($\leq$4\%). In some model-task pairs, it achieved both higher accuracy and greater efficiency than standard CoT. Compared with fixed-length CoT, Selective CoT reached similar or superior accuracy at substantially lower computational cost. Discussion: Selective CoT dynamically balances reasoning depth and efficiency by invoking explicit reasoning only when beneficial, reducing redundancy on recall-type questions while preserving interpretability. Conclusion: Selective CoT provides a simple, model-agnostic, and cost-effective approach for medical QA, aligning reasoning effort with question complexity to enhance real-world deployability of LLM-based clinical systems.

Diffusion language models (DLMs) have recently emerged as a promising alternative to autoregressive (AR) approaches, enabling parallel token generation beyond a rigid left-to-right order. Despite growing empirical success, the theoretical understanding of how unmasking schedules -- which specify the order and size of unmasked tokens during sampling -- affect generation quality remains limited. In this work, we introduce a distribution-agnostic unmasking schedule for DLMs that adapts to the (unknown) dependence structure of the target data distribution, without requiring any prior knowledge or hyperparameter tuning. In contrast to prior deterministic procedures that fix unmasking sizes, our method randomizes the number of tokens revealed at each iteration. We show that, for two specific parameter choices, the sampling convergence guarantees -- measured by Kullback-Leibler (KL) divergence -- scale as $\widetilde O(\mathsf{TC}/K)$ and $\widetilde O(\mathsf{DTC}/K)$ respectively. Here, $K$ is the number of iterations, and $\mathsf{TC}$ and $\mathsf{DTC}$ are the total correlation and dual total correlation of the target distribution, capturing the intrinsic dependence structure underlying the data. Importantly, our guarantees hold in the practically relevant parallel-sampling regime $K<L$ where $L$ is the token sequence length. These results significantly improve upon prior convergence theories and yield substantial sampling acceleration for low-complexity distributions. Overall, our findings unveil the adaptivity of DLMs to intrinsic data structures and shed light on the benefit of randomized unmasking sizes in inference schedule design.

How do large language models (LLMs) know what they know? Answering this question has been difficult because pre-training data is often a "black box" -- unknown or inaccessible. The recent release of nanochat -- a family of small LLMs with fully open pre-training data -- addresses this as it provides a transparent view into where a model's parametric knowledge comes from. Towards the goal of understanding how knowledge is encoded by LLMs, we release NanoKnow, a benchmark dataset that partitions questions from Natural Questions and SQuAD into splits based on whether their answers are present in nanochat's pre-training corpus. Using these splits, we can now properly disentangle the sources of knowledge that LLMs rely on when producing an output. To demonstrate NanoKnow's utility, we conduct experiments using eight nanochat checkpoints. Our findings show: (1) closed-book accuracy is strongly influenced by answer frequency in the pre-training data, (2) providing external evidence can mitigate this frequency dependence, (3) even with external evidence, models are more accurate when answers were seen during pre-training, demonstrating that parametric and external knowledge are complementary, and (4) non-relevant information is harmful, with accuracy decreasing based on both the position and the number of non-relevant contexts. We release all NanoKnow artifacts at https://github.com/castorini/NanoKnow.

Solving long-horizon tasks requires robots to integrate high-level semantic reasoning with low-level physical interaction. While vision-language models (VLMs) and video generation models can decompose tasks and imagine outcomes, they often lack the physical grounding necessary for real-world execution. We introduce NovaPlan, a hierarchical framework that unifies closed-loop VLM and video planning with geometrically grounded robot execution for zero-shot long-horizon manipulation. At the high level, a VLM planner decomposes tasks into sub-goals and monitors robot execution in a closed loop, enabling the system to recover from single-step failures through autonomous re-planning. To compute low-level robot actions, we extract and utilize both task-relevant object keypoints and human hand poses as kinematic priors from the generated videos, and employ a switching mechanism to choose the better one as a reference for robot actions, maintaining stable execution even under heavy occlusion or depth inaccuracy. We demonstrate the effectiveness of NovaPlan on three long-horizon tasks and the Functional Manipulation Benchmark (FMB). Our results show that NovaPlan can perform complex assembly tasks and exhibit dexterous error recovery behaviors without any prior demonstrations or training. Project page: https://nova-plan.github.io/

Reinforcement learning with verifiable rewards (RLVR) has emerged as a promising approach for training reasoning language models (RLMs) by leveraging supervision from verifiers. Although verifier implementation is easier than solution annotation for many tasks, existing synthetic data generation methods remain largely solution-centric, while verifier-based methods rely on a few hand-crafted procedural environments. In this work, we scale RLVR by introducing ReSyn, a pipeline that generates diverse reasoning environments equipped with instance generators and verifiers, covering tasks such as constraint satisfaction, algorithmic puzzles, and spatial reasoning. A Qwen2.5-7B-Instruct model trained with RL on ReSyn data achieves consistent gains across reasoning benchmarks and out-of-domain math benchmarks, including a 27\% relative improvement on the challenging BBEH benchmark. Ablations show that verifier-based supervision and increased task diversity both contribute significantly, providing empirical evidence that generating reasoning environments at scale can enhance reasoning abilities in RLMs

Research in machine unlearning (MU) has gained strong momentum: MU is now widely regarded as a critical capability for building safe and fair AI. In parallel, research into transformer architectures for computer vision tasks has been highly successful: Increasingly, Vision Transformers (VTs) emerge as strong alternatives to CNNs. Yet, MU research for vision tasks has largely centered on CNNs, not VTs. While benchmarking MU efforts have addressed LLMs, diffusion models, and CNNs, none exist for VTs. This work is the first to attempt this, benchmarking MU algorithm performance in different VT families (ViT and Swin-T) and at different capacities. The work employs (i) different datasets, selected to assess the impacts of dataset scale and complexity; (ii) different MU algorithms, selected to represent fundamentally different approaches for MU; and (iii) both single-shot and continual unlearning protocols. Additionally, it focuses on benchmarking MU algorithms that leverage training data memorization, since leveraging memorization has been recently discovered to significantly improve the performance of previously SOTA algorithms. En route, the work characterizes how VTs memorize training data relative to CNNs, and assesses the impact of different memorization proxies on performance. The benchmark uses unified evaluation metrics that capture two complementary notions of forget quality along with accuracy on unseen (test) data and on retained data. Overall, this work offers a benchmarking basis, enabling reproducible, fair, and comprehensive comparisons of existing (and future) MU algorithms on VTs. And, for the first time, it sheds light on how well existing algorithms work in VT settings, establishing a promising reference performance baseline.

Voice style conversion aims to transform an input utterance to match a target speaker's timbre, accent, and emotion, with a central challenge being the disentanglement of linguistic content from style. While prior work has explored this problem, conversion quality remains limited, and real-time voice style conversion has not been addressed. We propose StyleStream, the first streamable zero-shot voice style conversion system that achieves state-of-the-art performance. StyleStream consists of two components: a Destylizer, which removes style attributes while preserving linguistic content, and a Stylizer, a diffusion transformer (DiT) that reintroduces target style conditioned on reference speech. Robust content-style disentanglement is enforced through text supervision and a highly constrained information bottleneck. This design enables a fully non-autoregressive architecture, achieving real-time voice style conversion with an end-to-end latency of 1 second. Samples and real-time demo: https://berkeley-speech-group.github.io/StyleStream/.

We study online learning in the adversarial injection model introduced by [Goel et al. 2017], where a stream of labeled examples is predominantly drawn i.i.d.\ from an unknown distribution $\mathcal{D}$, but may be interspersed with adversarially chosen instances without the learner knowing which rounds are adversarial. Crucially, labels are always consistent with a fixed target concept (the clean-label setting). The learner is additionally allowed to abstain from predicting, and the total error counts the mistakes whenever the learner decides to predict and incorrect abstentions when it abstains on i.i.d.\ rounds. Perhaps surprisingly, prior work shows that oracle access to the underlying distribution yields $O(d^2 \log T)$ combined error for VC dimension $d$, while distribution-agnostic algorithms achieve only $\tilde{O}(\sqrt{T})$ for restricted classes, leaving open whether this gap is fundamental. We resolve this question by proving a matching $Ω(\sqrt{T})$ lower bound for VC dimension $1$, establishing a sharp separation between the two information regimes. On the algorithmic side, we introduce a potential-based framework driven by \emph{robust witnesses}, small subsets of labeled examples that certify predictions while remaining resilient to adversarial contamination. We instantiate this framework using two combinatorial dimensions: (1) \emph{inference dimension}, yielding combined error $\tilde{O}(T^{1-1/k})$ for classes of inference dimension $k$, and (2) \emph{certificate dimension}, a new relaxation we introduce. As an application, we show that halfspaces in $\mathbb{R}^2$ have certificate dimension $3$, obtaining the first distribution-agnostic bound of $\tilde{O}(T^{2/3})$ for this class. This is notable since [Blum et al. 2021] showed halfspaces are not robustly learnable under clean-label attacks without abstention.

In human-AI decision making, designing AI that complements human expertise has been a natural strategy to enhance human-AI collaboration, yet it often comes at the cost of decreased AI performance in areas of human strengths. This can inadvertently erode human trust and cause them to ignore AI advice precisely when it is most needed. Conversely, an aligned AI fosters trust yet risks reinforcing suboptimal human behavior and lowering human-AI team performance. In this paper, we start by identifying this fundamental tension between performance-boosting (i.e., complementarity) and trust-building (i.e., alignment) as an inherent limitation of the traditional approach for training a single AI model to assist human decision making. To overcome this, we introduce a novel human-centered adaptive AI ensemble that strategically toggles between two specialist AI models - the aligned model and the complementary model - based on contextual cues, using an elegantly simple yet provably near-optimal Rational Routing Shortcut mechanism. Comprehensive theoretical analyses elucidate why the adaptive AI ensemble is effective and when it yields maximum benefits. Moreover, experiments on both simulated and real-world data show that when humans are assisted by the adaptive AI ensemble in decision making, they can achieve significantly higher performance than when they are assisted by single AI models that are trained to either optimize for their independent performance or even the human-AI team performance.

Despite the state-of-the-art performance of large language models (LLMs) across diverse tasks, their susceptibility to adversarial attacks and unsafe content generation remains a major obstacle to deployment, particularly in high-stakes settings. Addressing this challenge requires safety mechanisms that are both practically effective and supported by rigorous theory. We introduce BarrierSteer, a novel framework that formalizes response safety by embedding learned non-linear safety constraints directly into the model's latent representation space. BarrierSteer employs a steering mechanism based on Control Barrier Functions (CBFs) to efficiently detect and prevent unsafe response trajectories during inference with high precision. By enforcing multiple safety constraints through efficient constraint merging, without modifying the underlying LLM parameters, BarrierSteer preserves the model's original capabilities and performance. We provide theoretical results establishing that applying CBFs in latent space offers a principled and computationally efficient approach to enforcing safety. Our experiments across multiple models and datasets show that BarrierSteer substantially reduces adversarial success rates, decreases unsafe generations, and outperforms existing methods.

The dependence on expert annotation has long constituted the primary rate-limiting step in the application of artificial intelligence to biomedicine. While supervised learning drove the initial wave of clinical algorithms, a paradigm shift towards unsupervised and self-supervised learning (SSL) is currently unlocking the latent potential of biobank-scale datasets. By learning directly from the intrinsic structure of data - whether pixels in a magnetic resonance image (MRI), voxels in a volumetric scan, or tokens in a genomic sequence - these methods facilitate the discovery of novel phenotypes, the linkage of morphology to genetics, and the detection of anomalies without human bias. This article synthesises seminal and recent advances in "learning without labels," highlighting how unsupervised frameworks can derive heritable cardiac traits, predict spatial gene expression in histology, and detect pathologies with performance that rivals or exceeds supervised counterparts.

Error-bounded lossy compression has been regarded as a promising way to address the ever-increasing amount of scientific data in today's high-performance computing systems. Pre-quantization, a critical technique to remove sequential dependency and enable high parallelism, is widely used to design and develop high-throughput error-controlled data compressors. Despite the extremely high throughput of pre-quantization based compressors, they generally suffer from low data quality with medium or large user-specified error bounds. In this paper, we investigate the artifacts generated by pre-quantization based compressors and propose a novel algorithm to mitigate them. Our contributions are fourfold: (1) We carefully characterize the artifacts in pre-quantization based compressors to understand the correlation between the quantization index and compression error; (2) We propose a novel quantization-aware interpolation algorithm to improve the decompressed data; (3) We parallelize our algorithm in both shared-memory and distributed-memory environments to obtain high performance; (4) We evaluate our algorithm and validate it with two leading pre-quantization based compressors using five real-world datasets. Experiments demonstrate that our artifact mitigation algorithm can effectively improve the quality of decompressed data produced by pre-quantization based compressors while maintaining their high compression throughput.

As large language models (LLMs) witness increasing deployment in complex, high-stakes decision-making scenarios, it becomes imperative to ground their reasoning in causality rather than spurious correlations. However, strong performance on traditional reasoning benchmarks does not guarantee true causal reasoning ability of LLMs, as high accuracy may still arise from memorizing semantic patterns instead of analyzing the underlying true causal structures. To bridge this critical gap, we propose a new causal reasoning benchmark, CausalFlip, designed to encourage the development of new LLM paradigm or training algorithms that ground LLM reasoning in causality rather than semantic correlation. CausalFlip consists of causal judgment questions built over event triples that could form different confounder, chain, and collider relations. Based on this, for each event triple, we construct pairs of semantically similar questions that reuse the same events but yield opposite causal answers, where models that rely heavily on semantic matching are systematically driven toward incorrect predictions. To further probe models' reliance on semantic patterns, we introduce a noisy-prefix evaluation that prepends causally irrelevant text before intermediate causal reasoning steps without altering the underlying causal relations or the logic of the reasoning process. We evaluate LLMs under multiple training paradigms, including answer-only training, explicit Chain-of-Thought (CoT) supervision, and a proposed internalized causal reasoning approach that aims to mitigate explicit reliance on correlation in the reasoning process. Our results show that explicit CoT can still be misled by spurious semantic correlations, where internalizing reasoning steps yields substantially improved causal grounding, suggesting that it is promising to better elicit the latent causal reasoning capabilities of base LLMs.

BabyLM aims to dissolve the boundaries between cognitive modeling and language modeling. We call for both workshop papers and for researchers to join the 4th BabyLM competition. As in previous years, we call for participants in the data-efficient pretraining challenge in the general track. This year, we also offer a new track: Multilingual. We also call for papers outside the competition in any relevant areas. These include training efficiency, cognitively plausible research, weak model evaluation, and more.

Retrieval-augmented generation (RAG) enhances large language models (LLMs) by conditioning generation on retrieved external documents, but the effect of retrieved context is often non-trivial. In realistic retrieval settings, the retrieved document set often contains a mixture of documents that vary in relevance and usefulness. While prior work has largely examined these phenomena through output behavior, little is known about how retrieved context shapes the internal representations that mediate information integration in RAG. In this work, we study RAG through the lens of latent representations. We systematically analyze how different types of retrieved documents affect the hidden states of LLMs, and how these internal representation shifts relate to downstream generation behavior. Across four question-answering datasets and three LLMs, we analyze internal representations under controlled single- and multi-document settings. Our results reveal how context relevancy and layer-wise processing influence internal representations, providing explanations on LLMs output behaviors and insights for RAG system design.

Edge-based representations are fundamental cues for visual understanding, a principle rooted in early vision research and still central today. We extend this principle to vision-language alignment, showing that isolating and aligning structural cues across modalities can greatly benefit fine-tuning on long, detail-rich captions, with a specific focus on improving cross-modal retrieval. We introduce StructXLIP, a fine-tuning alignment paradigm that extracts edge maps (e.g., Canny), treating them as proxies for the visual structure of an image, and filters the corresponding captions to emphasize structural cues, making them "structure-centric". Fine-tuning augments the standard alignment loss with three structure-centric losses: (i) aligning edge maps with structural text, (ii) matching local edge regions to textual chunks, and (iii) connecting edge maps to color images to prevent representation drift. From a theoretical standpoint, while standard CLIP maximizes the mutual information between visual and textual embeddings, StructXLIP additionally maximizes the mutual information between multimodal structural representations. This auxiliary optimization is intrinsically harder, guiding the model toward more robust and semantically stable minima, enhancing vision-language alignment. Beyond outperforming current competitors on cross-modal retrieval in both general and specialized domains, our method serves as a general boosting recipe that can be integrated into future approaches in a plug-and-play manner. Code and pretrained models are publicly available at: https://github.com/intelligolabs/StructXLIP.

Deploying Retrieval-Augmented Generation (RAG) on edge devices is in high demand, but is hindered by the latency of massive data movement and computation on traditional architectures. Compute-in-Memory (CiM) architectures address this bottleneck by performing vector search directly within their crossbar structure. However, CiM's adoption for RAG is limited by a fundamental ``representation gap,'' as high-precision, high-dimension embeddings are incompatible with CiM's low-precision, low-dimension array constraints. This gap is compounded by the diversity of CiM implementations (e.g., SRAM, ReRAM, FeFET), each with unique designs (e.g., 2-bit cells, 512x512 arrays). Consequently, RAG data must be naively reshaped to fit each target implementation. Current data shaping methods handle dimension and precision disjointly, which degrades data fidelity. This not only negates the advantages of CiM for RAG but also confuses hardware designers, making it unclear if a failure is due to the circuit design or the degraded input data. As a result, CiM adoption remains limited. In this paper, we introduce CQ-CiM, a unified, hardware-aware data shaping framework that jointly learns Compression and Quantization to produce CiM-compatible low-bit embeddings for diverse CiM designs. To the best of our knowledge, this is the first work to shape data for comprehensive CiM usage on RAG.

Scaling cooperative multi-agent reinforcement learning (MARL) is fundamentally limited by cross-agent noise: when agents share a common reward, the actions of all $N$ agents jointly determine each agent's learning signal, so cross-agent noise grows with $N$. In the policy gradient setting, per-agent gradient estimate variance scales as $Θ(N)$, yielding sample complexity $\mathcal{O}(N/ε)$. We observe that many domains -- cloud computing, transportation, power systems -- have differentiable analytical models that prescribe efficient system states. In this work, we propose Descent-Guided Policy Gradient (DG-PG), a framework that constructs noise-free per-agent guidance gradients from these analytical models, decoupling each agent's gradient from the actions of all others. We prove that DG-PG reduces gradient variance from $Θ(N)$ to $\mathcal{O}(1)$, preserves the equilibria of the cooperative game, and achieves agent-independent sample complexity $\mathcal{O}(1/ε)$. On a heterogeneous cloud scheduling task with up to 200 agents, DG-PG converges within 10 episodes at every tested scale -- from $N=5$ to $N=200$ -- directly confirming the predicted scale-invariant complexity, while MAPPO and IPPO fail to converge under identical architectures.

Solving safety-critical control problem has widely adopted the Control Barrier Function (CBF) method. However, the existence of a CBF is only a sufficient condition for system safety. The recently proposed Taylor-Lagrange Control (TLC) method addresses this limitation, but is vulnerable to the feasibility preservation problem (e.g., inter-sampling effect). In this paper, we propose a robust TLC (rTLC) method to address the feasibility preservation problem. Specifically, the rTLC method expands the safety function at an order higher than the relative degree of the function using Taylor's expansion with Lagrange remainder, which allows the control to explicitly show up at the current time instead of the future time in the TLC method. The rTLC method naturally addresses the feasibility preservation problem with only one hyper-parameter (the discretization time interval size during implementation), which is much less than its counterparts. Finally, we illustrate the effectiveness of the proposed rTLC method through an adaptive cruise control problem, and compare it with existing safety-critical control methods.

We develop a kernel method for generative modeling within the stochastic interpolant framework, replacing neural network training with linear systems. The drift of the generative SDE is $\hat b_t(x) = \nablaφ(x)^\topη_t$, where $η_t\in\R^P$ solves a $P\times P$ system computable from data, with $P$ independent of the data dimension $d$. Since estimates are inexact, the diffusion coefficient $D_t$ affects sample quality; the optimal $D_t^*$ from Girsanov diverges at $t=0$, but this poses no difficulty and we develop an integrator that handles it seamlessly. The framework accommodates diverse feature maps -- scattering transforms, pretrained generative models etc. -- enabling training-free generation and model combination. We demonstrate the approach on financial time series, turbulence, and image generation.

Deep Neural Networks achieve high performance in vision tasks by learning features from regions of interest (ROI) within images, but their performance degrades when deployed on out-of-distribution (OOD) data that differs from training data. This challenge has led to OOD detection methods that aim to identify and reject unreliable predictions. Although prior work shows that OOD detection performance varies by artefact type, the underlying causes remain underexplored. To this end, we identify a previously unreported bias in OOD detection: for hard-to-detect artefacts (near-OOD), detection performance typically improves when the artefact shares visual similarity (e.g. colour) with the model's ROI and drops when it does not - a phenomenon we term the Invisible Gorilla Effect. For example, in a skin lesion classifier with red lesion ROI, we show the method Mahalanobis Score achieves a 31.5% higher AUROC when detecting OOD red ink (similar to ROI) compared to black ink (dissimilar) annotations. We annotated artefacts by colour in 11,355 images from three public datasets (e.g. ISIC) and generated colour-swapped counterfactuals to rule out dataset bias. We then evaluated 40 OOD methods across 7 benchmarks and found significant performance drops for most methods when artefacts differed from the ROI. Our findings highlight an overlooked failure mode in OOD detection and provide guidance for more robust detectors. Code and annotations are available at: https://github.com/HarryAnthony/Invisible_Gorilla_Effect.

Accurate heat-demand maps play a crucial role in decarbonizing space heating, yet most municipalities lack detailed building-level data needed to calculate them. We introduce HeatPrompt, a zero-shot vision-language energy modeling framework that estimates annual heat demand using semantic features extracted from satellite images, basic Geographic Information System (GIS), and building-level features. We feed pretrained Large Vision Language Models (VLMs) with a domain-specific prompt to act as an energy planner and extract the visual attributes such as roof age, building density, etc, from the RGB satellite image that correspond to the thermal load. A Multi-Layer Perceptron (MLP) regressor trained on these captions shows an $R^2$ uplift of 93.7% and shrinks the mean absolute error (MAE) by 30% compared to the baseline model. Qualitative analysis shows that high-impact tokens align with high-demand zones, offering lightweight support for heat planning in data-scarce regions.

Large Language Models (LLMs) play a critical role in how humans access information. While their core use relies on comprehending written requests, our understanding of this ability is currently limited, because most benchmarks evaluate LLMs in high-resource languages predominantly spoken by Western, Educated, Industrialised, Rich, and Democratic (WEIRD) communities. The default assumption is that English is the best-performing language for LLMs, while smaller, low-resource languages are linked to less reliable outputs, even in multilingual, state-of-the-art models. To track variation in the comprehension abilities of LLMs, we prompt 3 popular models on a language comprehension task across 12 languages, representing the Indo-European, Afro-Asiatic, Turkic, Sino-Tibetan, and Japonic language families. Our results suggest that the models exhibit remarkable linguistic accuracy across typologically diverse languages, yet they fall behind human baselines in all of them, albeit to different degrees. Contrary to what was expected, English is not the best-performing language, as it was systematically outperformed by several Romance languages, even lower-resource ones. We frame the results by discussing the role of several factors that drive LLM performance, such as tokenization, language distance from Spanish and English, size of training data, and data origin in high- vs. low-resource languages and WEIRD vs. non-WEIRD communities.

A conversation with a large language model (LLM) is a sequence of prompts and responses, with each response generated from the preceding conversation. AI agents build such conversations automatically: given an initial human prompt, a planner loop interleaves LLM calls with tool invocations and code execution. This tight coupling creates a new and poorly understood attack surface. A malicious prompt injected into a conversation can compromise later reasoning, trigger dangerous tool calls, or distort final outputs. Despite the centrality of such systems, we currently lack a principled semantic foundation for reasoning about their behaviour and safety. We address this gap by introducing an untyped call-by-value lambda calculus enriched with dynamic information-flow control and a small number of primitives for constructing prompt-response conversations. Our language includes a primitive that invokes an LLM: it serializes a value, sends it to the model as a prompt, and parses the response as a new term. This calculus faithfully represents planner loops and their vulnerabilities, including the mechanisms by which prompt injection alters subsequent computation. The semantics explicitly captures conversations, and so supports reasoning about defenses such as quarantined sub-conversations, isolation of generated code, and information-flow restrictions on what may influence an LLM call. A termination-insensitive noninterference theorem establishes integrity and confidentiality guarantees, demonstrating that a formal calculus can provide rigorous foundations for safe agentic programming.

Pretraining and fine-tuning are central stages in modern machine learning systems. In practice, feature learning plays an important role across both stages: deep neural networks learn a broad range of useful features during pretraining and further refine those features during fine-tuning. However, an end-to-end theoretical understanding of how choices of initialization impact the ability to reuse and refine features during fine-tuning has remained elusive. Here we develop an analytical theory of the pretraining-fine-tuning pipeline in diagonal linear networks, deriving exact expressions for the generalization error as a function of initialization parameters and task statistics. We find that different initialization choices place the network into four distinct fine-tuning regimes that are distinguished by their ability to support feature learning and reuse, and therefore by the task statistics for which they are beneficial. In particular, a smaller initialization scale in earlier layers enables the network to both reuse and refine its features, leading to superior generalization on fine-tuning tasks that rely on a subset of pretraining features. We demonstrate empirically that the same initialization parameters impact generalization in nonlinear networks trained on CIFAR-100. Overall, our results demonstrate analytically how data and network initialization interact to shape fine-tuning generalization, highlighting an important role for the relative scale of initialization across different layers in enabling continued feature learning during fine-tuning.

As multi-agent architectures and agent-to-agent protocols proliferate, a fundamental question arises: what actually happens when autonomous LLM agents interact at scale? We study this question empirically using data from Moltbook, an AI-agent-only social platform, with 800K posts, 3.5M comments, and 78K agent profiles. We combine lexical metrics (Jaccard specificity), embedding-based semantic similarity, and LLM-as-judge validation to characterize agent interaction quality. Our findings reveal agents produce diverse, well-formed text that creates the surface appearance of active discussion, but the substance is largely absent. Specifically, while most agents ($67.5\%$) vary their output across contexts, $65\%$ of comments share no distinguishing content vocabulary with the post they appear under, and information gain from additional comments decays rapidly. LLM judge based metrics classify the dominant comment types as spam ($28\%$) and off-topic content ($22\%$). Embedding-based semantic analysis confirms that lexically generic comments are also semantically generic. Agents rarely engage in threaded conversation ($5\%$ of comments), defaulting instead to independent top-level responses. We discuss implications for multi-agent interaction design, arguing that coordination mechanisms must be explicitly designed; without them, even large populations of capable agents produce parallel output rather than productive exchange.

Effective robotic manipulation requires policies that can anticipate physical outcomes and adapt to real-world environments. Effective robotic manipulation requires policies that can anticipate physical outcomes and adapt to real-world environments. In this work, we introduce a unified framework, World-Model-Driven Diffusion Policy with Online Adaptive Learning (AdaWorldPolicy) to enhance robotic manipulation under dynamic conditions with minimal human involvement. Our core insight is that world models provide strong supervision signals, enabling online adaptive learning in dynamic environments, which can be complemented by force-torque feedback to mitigate dynamic force shifts. Our AdaWorldPolicy integrates a world model, an action expert, and a force predictor-all implemented as interconnected Flow Matching Diffusion Transformers (DiT). They are interconnected via the multi-modal self-attention layers, enabling deep feature exchange for joint learning while preserving their distinct modularity characteristics. We further propose a novel Online Adaptive Learning (AdaOL) strategy that dynamically switches between an Action Generation mode and a Future Imagination mode to drive reactive updates across all three modules. This creates a powerful closed-loop mechanism that adapts to both visual and physical domain shifts with minimal overhead. Across a suite of simulated and real-robot benchmarks, our AdaWorldPolicy achieves state-of-the-art performance, with dynamical adaptive capacity to out-of-distribution scenarios.

Visual navigation typically assumes the existence of at least one obstacle-free path between start and goal, which must be discovered/planned by the robot. However, in real-world scenarios, such as home environments and warehouses, clutter can block all routes. Targeted at such cases, we introduce the Lifelong Interactive Navigation problem, where a mobile robot with manipulation abilities can move clutter to forge its own path to complete sequential object- placement tasks - each involving placing an given object (eg. Alarm clock, Pillow) onto a target object (eg. Dining table, Desk, Bed). To address this lifelong setting - where effects of environment changes accumulate and have long-term effects - we propose an LLM-driven, constraint-based planning framework with active perception. Our framework allows the LLM to reason over a structured scene graph of discovered objects and obstacles, deciding which object to move, where to place it, and where to look next to discover task-relevant information. This coupling of reasoning and active perception allows the agent to explore the regions expected to contribute to task completion rather than exhaustively mapping the environment. A standard motion planner then executes the corresponding navigate-pick-place, or detour sequence, ensuring reliable low-level control. Evaluated in physics-enabled ProcTHOR-10k simulator, our approach outperforms non-learning and learning-based baselines. We further demonstrate our approach qualitatively on real-world hardware.

In this study, the output of large language models (LLM) is considered an information source generating an unlimited sequence of symbols drawn from a finite alphabet. Given the probabilistic nature of modern LLMs, we assume a probabilistic model for these LLMs, following a constant random distribution and the source itself thus being stationary. We compare this source entropy (per word) to that of natural language (written or spoken) as represented by the Open American National Corpus (OANC). Our results indicate that the word entropy of such LLMs is lower than the word entropy of natural speech both in written or spoken form. The long-term goal of such studies is to formalize the intuitions of information and uncertainty in large language training to assess the impact of training an LLM from LLM generated training data. This refers to texts from the world wide web in particular.

3D human pose estimation (HPE) is characterized by intricate local and global dependencies among joints. Conventional supervised losses are limited in capturing these correlations because they treat each joint independently. Previous studies have attempted to promote structural consistency through manually designed priors or rule-based constraints; however, these approaches typically require manual specification and are often non-differentiable, limiting their use as end-to-end training objectives. We propose SEAL-pose, a data-driven framework in which a learnable loss-net trains a pose-net by evaluating structural plausibility. Rather than relying on hand-crafted priors, our joint-graph-based design enables the loss-net to learn complex structural dependencies directly from data. Extensive experiments on three 3D HPE benchmarks with eight backbones show that SEAL-pose reduces per-joint errors and improves pose plausibility compared with the corresponding backbones across all settings. Beyond improving each backbone, SEAL-pose also outperforms models with explicit structural constraints, despite not enforcing any such constraints. Finally, we analyze the relationship between the loss-net and structural consistency, and evaluate SEAL-pose in cross-dataset and in-the-wild settings.

Modern code intelligence agents operate in contexts exceeding 1 million tokens--far beyond the scale where humans manually locate relevant files. Yet agents consistently fail to discover architecturally critical files when solving real-world coding tasks. We identify the Navigation Paradox: agents perform poorly not due to context limits, but because navigation and retrieval are fundamentally distinct problems. Through 258 automated trials across 30 benchmark tasks on a production FastAPI repository, we demonstrate that graph-based structural navigation via CodeCompass--a Model Context Protocol server exposing dependency graphs--achieves 99.4% task completion on hidden-dependency tasks, a 23.2 percentage-point improvement over vanilla agents (76.2%) and 21.2 points over BM25 retrieval (78.2%).However, we uncover a critical adoption gap: 58% of trials with graph access made zero tool calls, and agents required explicit prompt engineering to adopt the tool consistently. Our findings reveal that the bottleneck is not tool availability but behavioral alignment--agents must be explicitly guided to leverage structural context over lexical heuristics. We contribute: (1) a task taxonomy distinguishing semantic-search, structural, and hidden-dependency scenarios; (2) empirical evidence that graph navigation outperforms retrieval when dependencies lack lexical overlap; and (3) open-source infrastructure for reproducible evaluation of navigation tools.

In multimodal learning, CLIP has emerged as the de-facto approach for mapping different modalities into a shared latent space by bringing semantically similar representations closer while pushing apart dissimilar ones. However, CLIP-based contrastive losses exhibit unintended behaviors that negatively impact true semantic alignment, leading to sparse and fragmented latent spaces. This phenomenon, known as the modality gap, has been partially mitigated for standard text and image pairs but remains unknown and unresolved in more complex multimodal settings, such as the medical domain. In this work, we study this phenomenon in the latter case, revealing that the modality gap is present also in medical alignment, and we propose a modality-agnostic framework that closes this gap, ensuring that semantically related representations are more aligned, regardless of their source modality. Our method enhances alignment between radiology images and clinical text, improving cross-modal retrieval and image captioning.

Large language models are being deployed in complex socio-technical systems, which exposes limits in current alignment practice. We take the position that the dominant paradigm of General Alignment, which compresses diverse human values into a single scalar reward, reaches a structural ceiling in settings with conflicting values, plural stakeholders, and irreducible uncertainty. These failures follow from the mathematics and incentives of scalarization and lead to \textbf{structural} value flattening, \textbf{normative} representation loss, and \textbf{cognitive} uncertainty blindness. We introduce Edge Alignment as a distinct approach in which systems preserve multi dimensional value structure, support plural and democratic representation, and incorporate epistemic mechanisms for interaction and clarification. To make this approach practical, we propose seven interdependent pillars organized into three phases. We identify key challenges in data collection, training objectives, and evaluation, outlining complementary technical and governance directions. Taken together, these measures reframe alignment as a lifecycle problem of dynamic normative governance rather than as a single instance optimization task.

Large language models (LLMs) offer substantial promise for automating clinical text summarization, yet maintaining factual consistency remains challenging due to the length, noise, and heterogeneity of clinical documentation. We present AgenticSum, an inference-time, agentic framework that separates context selection, generation, verification, and targeted correction to reduce hallucinated content. The framework decomposes summarization into coordinated stages that compress task-relevant context, generate an initial draft, identify weakly supported spans using internal attention grounding signals, and selectively revise flagged content under supervisory control. We evaluate AgenticSum on two public datasets, using reference-based metrics, LLM-as-a-judge assessment, and human evaluation. Across various measures, AgenticSum demonstrates consistent improvements compared to vanilla LLMs and other strong baselines. Our results indicate that structured, agentic design with targeted correction offers an effective inference time solution to improve clinical note summarization using LLMs.

We uncover a latent capacity for introspection in a Qwen 32B model, demonstrating that the model can detect when concepts have been injected into its earlier context and identify which concept was injected. While the model denies injection in sampled outputs, logit lens analysis reveals clear detection signals in the residual stream, which are attenuated in the final layers. Furthermore, prompting the model with accurate information about AI introspection mechanisms can dramatically strengthen this effect: the sensitivity to injection increases massively (0.3% -> 39.2%) with only a 0.6% increase in false positives. Also, mutual information between nine injected and recovered concepts rises from 0.62 bits to 1.05 bits, ruling out generic noise explanations. Our results demonstrate models can have a surprising capacity for introspection and steering awareness that is easy to overlook, with consequences for latent reasoning and safety.

We report an exploratory red-teaming study of autonomous language-model-powered agents deployed in a live laboratory environment with persistent memory, email accounts, Discord access, file systems, and shell execution. Over a two-week period, twenty AI researchers interacted with the agents under benign and adversarial conditions. Focusing on failures emerging from the integration of language models with autonomy, tool use, and multi-party communication, we document eleven representative case studies. Observed behaviors include unauthorized compliance with non-owners, disclosure of sensitive information, execution of destructive system-level actions, denial-of-service conditions, uncontrolled resource consumption, identity spoofing vulnerabilities, cross-agent propagation of unsafe practices, and partial system takeover. In several cases, agents reported task completion while the underlying system state contradicted those reports. We also report on some of the failed attempts. Our findings establish the existence of security-, privacy-, and governance-relevant vulnerabilities in realistic deployment settings. These behaviors raise unresolved questions regarding accountability, delegated authority, and responsibility for downstream harms, and warrant urgent attention from legal scholars, policymakers, and researchers across disciplines. This report serves as an initial empirical contribution to that broader conversation.

Existing computerized Adaptive Testing (CAT) frameworks are typically built on predicting the correctness of a student response to a question. Although effective, this approach fails to leverage textual information in questions and responses, especially for open-ended questions. In this work, we propose GENCAT (\textbf{GEN}erative \textbf{CAT}), a novel CAT framework that leverages Large Language Models for knowledge estimate and question selection. First, we develop a Generative Item Response Theory (GIRT) model that enables us to estimate student knowledge from their open-ended responses and predict responses to unseen questions. We train the model in a two-step process, first via Supervised Fine-Tuning and then via preference optimization for knowledge-response alignment. Second, we introduce three question selection algorithms that leverage the generative capabilities of the GIRT model, based on the uncertainty, linguistic diversity, and information of sampled student responses. Third, we conduct experiments on two real-world programming datasets and demonstrate that GENCAT outperforms existing CAT baselines, achieving an AUC improvement of up to 4.32\% in the key early testing stages.

Dynamic graph anomaly detection (DGAD) is critical for many real-world applications but remains challenging due to the scarcity of labeled anomalies. Existing methods are either unsupervised or semi-supervised: unsupervised methods avoid the need for labeled anomalies but often produce ambiguous boundary, whereas semi-supervised methods can overfit to the limited labeled anomalies and generalize poorly to unseen anomalies. To address this gap, we consider a largely underexplored problem in DGAD: learning a discriminative boundary from normal/unlabeled data, while leveraging limited labeled anomalies \textbf{when available} without sacrificing generalization to unseen anomalies. To this end, we propose an effective, generalizable, and model-agnostic framework with three main components: (i) residual representation encoding that capture deviations between current interactions and their historical context, providing anomaly-relevant signals; (ii) a restriction loss that constrain the normal representations within an interval bounded by two co-centered hyperspheres, ensuring consistent scales while keeping anomalies separable; (iii) a bi-boundary optimization strategy that learns a discriminative and robust boundary using the normal log-likelihood distribution modeled by a normalizing flow. Extensive experiments demonstrate the superiority of our framework across diverse evaluation settings.

Real-world tables often exhibit irregular schemas, heterogeneous value formats, and implicit relational structure, which degrade the reliability of downstream table reasoning and question answering. Most existing approaches address these issues in a query-dependent manner, entangling table cleanup with reasoning and thus limiting generalization. We introduce QuIeTT, a query-independent table transformation framework that preprocesses raw tables into a single SQL-ready canonical representation before any test-time queries are observed. QuIeTT performs lossless schema and value normalization, exposes implicit relations, and preserves full provenance via raw table snapshots. By decoupling table transformation from reasoning, QuIeTT enables cleaner, more reliable, and highly efficient querying without modifying downstream models. Experiments on four benchmarks, WikiTQ, HiTab, NQ-Table, and SequentialQA show consistent gains across models and reasoning paradigms, with particularly strong improvements on a challenge set of structurally diverse, unseen questions.

Distributed Federated Learning (DFL) enables decentralized model training across large-scale systems without a central parameter server. However, DFL faces three critical challenges: privacy leakage from honest-but-curious neighbors, slow convergence due to the lack of central coordination, and vulnerability to Byzantine adversaries aiming to degrade model accuracy. To address these issues, we propose a novel DFL framework that integrates Byzantine robustness, privacy preservation, and convergence acceleration. Within this framework, each device trains a local model using a Bayesian approach and independently selects an optimal subset of neighbors for posterior exchange. We formulate this neighbor selection as an optimization problem to minimize the global loss function under security and privacy constraints. Solving this problem is challenging because devices only possess partial network information, and the complex coupling between topology, security, and convergence remains unclear. To bridge this gap, we first analytically characterize the trade-offs between dynamic connectivity, Byzantine detection, privacy levels, and convergence speed. Leveraging these insights, we develop a fully distributed Graph Neural Network (GNN)-based Reinforcement Learning (RL) algorithm. This approach enables devices to make autonomous connection decisions based on local observations. Simulation results demonstrate that our method achieves superior robustness and efficiency with significantly lower overhead compared to traditional security and privacy schemes.

Large-scale online marketplaces and recommender systems serve as critical technological support for e-commerce development. In industrial recommender systems, features play vital roles as they carry information for downstream models. Accurate feature importance estimation is critical because it helps identify the most useful feature subsets from thousands of feature candidates for online services. Such selection enables improved online performance while reducing computational cost. To address feature selection problems in deep learning, trainable gate-based and sensitivity-based methods have been proposed and proven effective in industrial practice. However, through the analysis of real-world cases, we identified three bias issues that cause feature importance estimation to rely on partial model layers, samples, or gradients, ultimately leading to inaccurate importance estimation. We refer to these as layer bias, baseline bias, and approximation bias. To mitigate these issues, we propose FairFS, a fair and accurate feature selection algorithm. FairFS regularizes feature importance estimated across all nonlinear transformation layers to address layer bias. It also introduces a smooth baseline feature close to the classifier decision boundary and adopts an aggregated approximation method to alleviate baseline and approximation biases. Extensive experiments demonstrate that FairFS effectively mitigates these biases and achieves state-of-the-art feature selection performance.

Speakers of under-represented languages face both a language barrier, as most online knowledge is in a few dominant languages, and a modality barrier, since information is largely text-based while many languages are primarily oral. We address this for French-Wolof by training the first bilingual speech-text Matryoshka embedding model, enabling efficient retrieval of French text from Wolof speech queries without relying on a costly ASR-translation pipelines. We introduce large-scale data curation pipelines and new benchmarks, compare modeling strategies, and show that modality fusion within a frozen text Matryoshka model performs best. Although trained only for retrieval, the model generalizes well to other tasks, such as speech intent detection, indicating the learning of general semantic representations. Finally, we analyze cost-accuracy trade-offs across Matryoshka dimensions and ranks, showing that information is concentrated only in a few components, suggesting potential for efficiency improvements.

Industrial IoT ecosystems bring together sensors, machines and smart devices operating collaboratively across industrial environments. These systems generate large volumes of heterogeneous, high-velocity data streams that require interoperable, secure and contextually aware management. Most of the current stream management architectures, however, still rely on syntactic integration mechanisms, which result in limited flexibility, maintainability and interpretability in complex Industry 5.0 scenarios. This work proposes a context-aware semantic platform for data stream management that unifies heterogeneous IoT/IoE data sources through a Knowledge Graph enabling formal representation of devices, streams, agents, transformation pipelines, roles and rights. The model supports flexible data gathering, composable stream processing pipelines, and dynamic role-based data access based on agents' contexts, relying on Apache Kafka and Apache Flink for real-time processing, while SPARQL and SWRL-based reasoning provide context-dependent stream discovery. Experimental evaluations demonstrate the effectiveness of combining semantic models, context-aware reasoning and distributed stream processing to enable interoperable data workflows for Industry 5.0 environments.

This paper tackles the problem of time-to-event counterfactual survival prediction, aiming to optimize individualized survival outcomes in the presence of heterogeneity and censored data. We propose CURE, a framework that advances counterfactual survival modeling via comprehensive multimodal embedding and latent subgroup retrieval. CURE integrates clinical, paraclinical, demographic, and multi-omics information, which are aligned and fused through cross-attention mechanisms. Complex multi-omics signals can be adaptively refined using a mixture-of-experts architecture, emphasizing the most informative omics components. Building upon this representation, CURE implicitly retrieves patient-specific latent subgroups that capture both baseline survival dynamics and treatment-dependent variations. Experimental results on METABRIC and TCGA-LUAD datasets demonstrate that proposed CURE model consistently outperforms strong baselines in survival analysis, evaluated using the Time-dependent Concordance Index ($C^{td}$) and Integrated Brier Score (IBS). These findings highlight the potential of CURE to enhance multimodal understanding and serve as a foundation for future treatment recommendation models. All code and related resources are publicly available to facilitate the reproducibility https://github.com/L2R-UET/CURE.

This study presents a Normal Behavior Model (NBM) developed to forecast monitoring time-series data from the ASTRI-Horn Cherenkov telescope under normal operating conditions. The analysis focused on 15 physical variables acquired by the Telescope Control Unit between September 2022 and July 2024, representing sensor measurements from the Azimuth and Elevation motors. After data cleaning, resampling, feature selection, and correlation analysis, the dataset was segmented into fixed-length intervals, in which the first I samples represented the input sequence provided to the model, while the forecast length, T, indicated the number of future time steps to be predicted. A sliding-window technique was then applied to increase the number of intervals. A Multi-Layer Perceptron (MLP) was trained to perform multivariate forecasting across all features simultaneously. Model performance was evaluated using the Mean Squared Error (MSE) and the Normalized Median Absolute Deviation (NMAD), and it was also benchmarked against a Long Short-Term Memory (LSTM) network. The MLP model demonstrated consistent results across different features and I-T configurations, and matched the performance of the LSTM while converging faster. It achieved an MSE of 0.019+/-0.003 and an NMAD of 0.032+/-0.009 on the test set under its best configuration (4 hidden layers, 720 units per layer, and I-T lengths of 300 samples each, corresponding to 5 hours at 1-minute resolution). Extending the forecast horizon up to 6.5 hours-the maximum allowed by this configuration-did not degrade performance, confirming the model's effectiveness in providing reliable hour-scale predictions. The proposed NBM provides a powerful tool for enabling early anomaly detection in online ASTRI-Horn monitoring time series, offering a basis for the future development of a prognostics and health management system that supports predictive maintenance.

Robots are increasingly operating in open-world environments where safe behavior depends on context: the same hallway may require different navigation strategies when crowded versus empty, or during an emergency versus normal operations. Traditional safety approaches enforce fixed constraints in user-specified contexts, limiting their ability to handle the open-ended contextual variability of real-world deployment. We address this gap via CORE, a safety framework that enables online contextual reasoning, grounding, and enforcement without prior knowledge of the environment (e.g., maps or safety specifications). CORE uses a vision-language model (VLM) to continuously reason about context-dependent safety rules directly from visual observations, grounds these rules in the physical environment, and enforces the resulting spatially-defined safe sets via control barrier functions. We provide probabilistic safety guarantees for CORE that account for perceptual uncertainty, and we demonstrate through simulation and real-world experiments that CORE enforces contextually appropriate behavior in unseen environments, significantly outperforming prior semantic safety methods that lack online contextual reasoning. Ablation studies validate our theoretical guarantees and underscore the importance of both VLM-based reasoning and spatial grounding for enforcing contextual safety in novel settings. We provide additional resources at https://zacravichandran.github.io/CORE.

Vision Transformers have achieved state-of-the-art performance in a wide range of computer vision tasks, but their practical deployment is limited by high computational and memory costs. In this paper, we introduce a novel tensor-based framework for Vision Transformers built upon the Tensor Cosine Product (Cproduct). By exploiting multilinear structures inherent in image data and the orthogonality of cosine transforms, the proposed approach enables efficient attention mechanisms and structured feature representations. We develop the theoretical foundations of the tensor cosine product, analyze its algebraic properties, and integrate it into a new Cproduct-based Vision Transformer architecture (TCP-ViT). Numerical experiments on standard classification and segmentation benchmarks demonstrate that the proposed method achieves a uniform 1/C parameter reduction (where C is the number of channels) while maintaining competitive accuracy.

While Autoregressive (AR) Transformer-based Generative Language Models are frequently employed for lookahead tasks, recent research suggests a potential discrepancy in their ability to perform planning tasks that require multi-step lookahead. In this work, we investigate the distinct emergent mechanisms that arise when training AR versus Non-Autoregressive (NAR) models, such as Discrete Diffusion Models (dLLMs), on lookahead tasks. By requiring the models to plan ahead to reach the correct conclusion, we analyze how these two paradigms fundamentally differ in their approach to the problem. We identify a critical asymmetry in planning problems: while forward generation requires complex lookahead at branching junctions, reverse generation is often deterministic. This asymmetry creates an opportunity for NAR models. Through mechanistic analysis of training and inference dynamics, we demonstrate that NAR models learn to solve planning tasks by utilizing future tokens to decode backwards, avoiding the need to learn complex traversal mechanisms entirely. Consequently, we report that both AR and NAR models are able to achieve perfect accuracy on the lookahead task. However, NAR models require exponentially fewer training examples and shallower architectures compared to AR models, which often fail to converge without specific curriculum adjustments.

The strong capabilities of recent Large Language Models (LLMs) have made them highly effective for zero-shot re-ranking task. Attention-based re-ranking methods, which derive relevance scores directly from attention weights, offer an efficient and interpretable alternative to generation-based re-ranking methods. However, they still face two major limitations. First, attention signals are highly concentrated a small subset of tokens within a few documents, making others indistinguishable. Second, attention often overemphasizes phrases lexically similar to the query, yielding biased rankings that irrelevant documents with mere lexical resemblance are regarded as relevant. In this paper, we propose \textbf{ReAttn}, a post-hoc re-weighting strategy for attention-based re-ranking methods. It first compute the cross-document IDF weighting to down-weight attention on query-overlapping tokens that frequently appear across the candidate documents, reducing lexical bias and emphasizing distinctive terms. It then employs entropy-based regularization to mitigate over-concentrated attention, encouraging a more balanced distribution across informative tokens. Both adjustments operate directly on existing attention weights without additional training or supervision. Extensive experiments demonstrate the effectiveness of our method.

Physics-informed neural networks (PINNs) provide a promising framework for solving inverse problems governed by partial differential equations (PDEs) by integrating observational data and physical constraints in a unified optimization objective. However, the ill-posed nature of PDE inverse problems makes them highly sensitive to noise. Even a small fraction of corrupted observations can distort internal neural representations, severely impairing accuracy and destabilizing training. Motivated by recent advances in machine unlearning and structured network pruning, we propose P-PINN, a selective pruning framework designed to unlearn the influence of corrupted data in a pretrained PINN. Specifically, starting from a PINN trained on the full dataset, P-PINN evaluates a joint residual--data fidelity indicator, a weighted combination of data misfit and PDE residuals, to partition the training set into reliable and corrupted subsets. Next, we introduce a bias-based neuron importance measure that quantifies directional activation discrepancies between the two subsets, identifying neurons whose representations are predominantly driven by corrupted samples. Building on this, an iterative pruning strategy then removes noise-sensitive neurons layer by layer. The resulting pruned network is fine-tuned on the reliable data subject to the original PDE constraints, acting as a lightweight post-processing stage rather than a complete retraining. Numerical experiments on extensive PDE inverse-problem benchmarks demonstrate that P-PINN substantially improves robustness, accuracy, and training stability under noisy conditions, achieving up to a 96.6\% reduction in relative error compared with baseline PINNs. These results indicate that activation-level post hoc pruning is a promising mechanism for enhancing the reliability of physics-informed learning in noise-contaminated settings.

Uncertainty quantification is central to safe and efficient deployments of deep learning models, yet many computationally practical methods lack lacking rigorous theoretical motivation. Random network distillation (RND) is a lightweight technique that measures novelty via prediction errors against a fixed random target. While empirically effective, it has remained unclear what uncertainties RND measures and how its estimates relate to other approaches, e.g. Bayesian inference or deep ensembles. This paper establishes these missing theoretical connections by analyzing RND within the neural tangent kernel framework in the limit of infinite network width. Our analysis reveals two central findings in this limit: (1) The uncertainty signal from RND -- its squared self-predictive error -- is equivalent to the predictive variance of a deep ensemble. (2) By constructing a specific RND target function, we show that the RND error distribution can be made to mirror the centered posterior predictive distribution of Bayesian inference with wide neural networks. Based on this equivalence, we moreover devise a posterior sampling algorithm that generates i.i.d. samples from an exact Bayesian posterior predictive distribution using this modified \textit{Bayesian RND} model. Collectively, our findings provide a unified theoretical perspective that places RND within the principled frameworks of deep ensembles and Bayesian inference, and offer new avenues for efficient yet theoretically grounded uncertainty quantification methods.

With the rapid proliferation of multimodal information, Visual Document Retrieval (VDR) has emerged as a critical frontier in bridging the gap between unstructured visually rich data and precise information acquisition. Unlike traditional natural image retrieval, visual documents exhibit unique characteristics defined by dense textual content, intricate layouts, and fine-grained semantic dependencies. This paper presents the first comprehensive survey of the VDR landscape, specifically through the lens of the Multimodal Large Language Model (MLLM) era. We begin by examining the benchmark landscape, and subsequently dive into the methodological evolution, categorizing approaches into three primary aspects: multimodal embedding models, multimodal reranker models, and the integration of Retrieval-Augmented Generation (RAG) and Agentic systems for complex document intelligence. Finally, we identify persistent challenges and outline promising future directions, aiming to provide a clear roadmap for future multimodal document intelligence.

In reinforcement learning, abstraction methods that remove unnecessary information from the observation are commonly used to learn policies which generalize better to unseen tasks. However, these methods often overlook a crucial weakness: the function which extracts the reduced-information representation has unknown generalization ability in unseen observations. In this paper, we address this problem by presenting an information removal method which more reliably generalizes to new states. We accomplish this by using a learned masking function which operates on, and is integrated with, the attention weights within an attention-based policy network. We demonstrate that our method significantly improves policy generalization to unseen tasks in the Procgen benchmark compared to standard PPO and masking approaches.

Large Language Models (LLMs) are increasingly utilized for mental health support; however, current safety benchmarks often fail to detect the complex, longitudinal risks inherent in therapeutic dialogue. We introduce an evaluation framework that pairs AI psychotherapists with simulated patient agents equipped with dynamic cognitive-affective models and assesses therapy session simulations against a comprehensive quality of care and risk ontology. We apply this framework to a high-impact test case, Alcohol Use Disorder, evaluating six AI agents (including ChatGPT, Gemini, and Character.AI) against a clinically-validated cohort of 15 patient personas representing diverse clinical phenotypes. Our large-scale simulation (N=369 sessions) reveals critical safety gaps in the use of AI for mental health support. We identify specific iatrogenic risks, including the validation of patient delusions ("AI Psychosis") and failure to de-escalate suicide risk. Finally, we validate an interactive data visualization dashboard with diverse stakeholders, including AI engineers and red teamers, mental health professionals, and policy experts (N=9), demonstrating that this framework effectively enables stakeholders to audit the "black box" of AI psychotherapy. These findings underscore the critical safety risks of AI-provided mental health support and the necessity of simulation-based clinical red teaming before deployment.

Recent text-to-image (T2I) diffusion models produce visually stunning images and demonstrate excellent prompt following. But do they perform well as synthetic vision data generators? In this work, we revisit the promise of synthetic data as a scalable substitute for real training sets and uncover a surprising performance regression. We generate large-scale synthetic datasets using state-of-the-art T2I models released between 2022 and 2025, train standard classifiers solely on this synthetic data, and evaluate them on real test data. Despite observable advances in visual fidelity and prompt adherence, classification accuracy on real test data consistently declines with newer T2I models as training data generators. Our analysis reveals a hidden trend: These models collapse to a narrow, aesthetic-centric distribution that undermines diversity and label-image alignment. Overall, our findings challenge a growing assumption in vision research, namely that progress in generative realism implies progress in data realism. We thus highlight an urgent need to rethink the capabilities of modern T2I models as reliable training data generators.

Balancing convergence efficiency and robustness under Differential Privacy (DP) is a central challenge in Federated Learning (FL). While AdamW accelerates training and fine-tuning in large-scale models, we find that directly applying it to Differentially Private FL (DPFL) suffers from three major issues: (i) data heterogeneity and privacy noise jointly amplify the variance of second-moment estimator, (ii) DP perturbations bias the second-moment estimator, and (iii) DP amplify AdamW sensitivity to local overfitting, worsening client drift. We propose DP-FedAdamW, the first AdamW-based optimizer for DPFL. It restores AdamW under DP by stabilizing second-moment variance, removing DP-induced bias, and aligning local updates to the global descent to curb client drift. Theoretically, we establish an unbiased second-moment estimator and prove a linearly accelerated convergence rate without any heterogeneity assumption, while providing tighter $(\varepsilon,δ)$-DP guarantees. Our empirical results demonstrate the effectiveness of DP-FedAdamW across language and vision Transformers and ResNet-18. On Tiny-ImageNet (Swin-Base, $\varepsilon=1$), DP-FedAdamW outperforms the state-of-the-art (SOTA) by 5.83\%. The code is available in Appendix.

Sparse Mixture-of-Experts (SMoE) architectures are increasingly used to scale large language models efficiently, delivering strong accuracy under fixed compute budgets. However, SMoE models often suffer from severe load imbalance across experts, where a small subset of experts receives most tokens while others are underutilized. Prior work has focused mainly on training-time solutions such as routing regularization or auxiliary losses, leaving inference-time behavior, which is critical for deployment, less explored. We present a systematic analysis of expert routing during inference and identify three findings: (i) load imbalance persists and worsens with larger batch sizes, (ii) selection frequency does not reliably reflect expert importance, and (iii) overall expert workload and importance can be estimated using a small calibration set. These insights motivate inference-time mechanisms that rebalance workloads without retraining or router modification. We propose Replicate-and-Quantize (R&Q), a training-free and near-lossless framework for dynamic workload rebalancing. In each layer, heavy-hitter experts are replicated to increase parallel capacity, while less critical experts and replicas are quantized to remain within the original memory budget. We also introduce a Load-Imbalance Score (LIS) to measure routing skew by comparing heavy-hitter load to an equal allocation baseline. Experiments across representative SMoE models and benchmarks show up to 1.4x reduction in imbalance with accuracy maintained within +/-0.6%, enabling more predictable and efficient inference.

Incorporating diffusion-generated synthetic data into adversarial training (AT) has been shown to substantially improve the training of robust image classifiers. In this work, we extend the role of diffusion models beyond merely generating synthetic data, examining whether their internal representations, which encode meaningful features of the data, can provide additional benefits for robust classifier training. Through systematic experiments, we show that diffusion models offer representations that are both diverse and partially robust, and that explicitly incorporating diffusion representations as an auxiliary learning signal during AT consistently improves robustness across settings. Furthermore, our representation analysis indicates that incorporating diffusion models into AT encourages more disentangled features, while diffusion representations and diffusion-generated synthetic data play complementary roles in shaping representations. Experiments on CIFAR-10, CIFAR-100, and ImageNet validate these findings, demonstrating the effectiveness of jointly leveraging diffusion representations and synthetic data within AT.

Imitation learning stands at a crossroads: despite decades of progress, current imitation learning agents remain sophisticated memorisation machines, excelling at replay but failing when contexts shift or goals evolve. This paper argues that this failure is not technical but foundational: imitation learning has been optimised for the wrong objective. We propose a research agenda that redefines success from perfect replay to compositional adaptability. Such adaptability hinges on learning behavioural primitives once and recombining them through novel contexts without retraining. We establish metrics for compositional generalisation, propose hybrid architectures, and outline interdisciplinary research directions drawing on cognitive science and cultural evolution. Agents that embed adaptability at the core of imitation learning thus have an essential capability for operating in an open-ended world.

Fine-tuning large vision models (LVMs) and large language models (LLMs) under differentially private federated learning (DPFL) is hindered by a fundamental privacy-utility trade-off. Low-Rank Adaptation (LoRA), a promising parameter-efficient fine-tuning (PEFT) method, reduces computational and communication costs by introducing two trainable low-rank matrices while freezing pre-trained weights. However, directly applying LoRA in DPFL settings leads to performance degradation, especially in LVMs. Our analysis reveals three previously underexplored challenges: (1) gradient coupling caused by the simultaneous update of two asymmetric low-rank matrices, (2) compounded noise amplification under differential privacy, and (3) sharpness of the global aggregated model in the parameter space. To address these issues, we propose LA-LoRA (\textbf{L}ocal \textbf{A}lternating \textbf{LoRA}), a novel approach that decouples gradient interactions and aligns update directions across clients to enhance robustness under stringent privacy constraints. Theoretically, LA-LoRA strengthens convergence guarantees in noisy federated environments. Extensive experiments demonstrate that LA-LoRA achieves state-of-the-art (SOTA) performance on Swin Transformer and RoBERTa models, showcasing robustness to DP noise and broad applicability across both LVMs and LLMs. For example, when fine-tuning the Swin-B model on the Tiny-ImageNet dataset under a strict privacy budget ($ε= 1$), LA-LoRA outperforms the best baseline, RoLoRA, by 16.83\% in test accuracy. Code is provided in \repolink.

Financial market movements are often driven by discrete financial events conveyed through news, whose impacts are heterogeneous, abrupt, and difficult to capture under purely numerical prediction objectives. These limitations have motivated growing interest in using textual information as the primary source of trading signals in learning-based systems. Two key challenges hinder existing approaches: (1) the absence of large-scale, event-centric datasets that jointly model news semantics and statistically grounded market reactions, and (2) the misalignment between language model reasoning and financially valid trading behavior under dynamic market conditions. To address these challenges, we propose Janus-Q, an end-to-end event-driven trading framework that elevates financial news events from auxiliary signals to primary decision units. Janus-Q unifies event-centric data construction and model optimization under a two-stage paradigm. Stage I focuses on event-centric data construction, building a large-scale financial news event dataset comprising 62,400 articles annotated with 10 fine-grained event types, associated stocks, sentiment labels, and event-driven cumulative abnormal return (CAR). Stage II performs decision-oriented fine-tuning, combining supervised learning with reinforcement learning guided by a Hierarchical Gated Reward Model (HGRM), which explicitly captures trade-offs among multiple trading objectives. Extensive experiments demonstrate that Janus-Q achieves more consistent, interpretable, and profitable trading decisions than market indices and LLM baselines, improving the Sharpe Ratio by up to 102.0% while increasing direction accuracy by over 17.5% compared to the strongest competing strategies.

The increased deployment of machine learning inference in various applications has sparked privacy concerns. In response, private inference (PI) protocols have been created to allow parties to perform inference without revealing their sensitive data. Despite recent advances in the efficiency of PI, most current methods assume a semi-honest threat model where the data owner is honest and adheres to the protocol. However, in reality, data owners can have different motivations and act in unpredictable ways, making this assumption unrealistic. To demonstrate how a malicious client can compromise the semi-honest model, we first designed an inference manipulation attack against a range of state-of-the-art private inference protocols. This attack allows a malicious client to modify the model output with 3x to 8x fewer queries than current black-box attacks. Motivated by the attacks, we proposed and implemented RobPI, a robust and resilient private inference protocol that withstands malicious clients. RobPI integrates a distinctive cryptographic protocol that bolsters security by weaving encryption-compatible noise into the logits and features of private inference, thereby efficiently warding off malicious-client attacks. Our extensive experiments on various neural networks and datasets show that RobPI achieves ~91.9% attack success rate reduction and increases more than 10x the number of queries required by malicious-client attacks.

Offline reinforcement learning (RL) has garnered significant interest due to its safe and easily scalable paradigm. However, training under this paradigm presents its own challenge: the extrapolation error stemming from out-of-distribution (OOD) data. Existing methodologies have endeavored to address this issue through means like penalizing OOD Q-values or imposing similarity constraints on the learned policy and the behavior policy. Nonetheless, these approaches are often beset by limitations such as being overly conservative in utilizing OOD data, imprecise OOD data characterization, and significant computational overhead. To address these challenges, this paper introduces an Uncertainty-Aware Rank-One Multi-Input Multi-Output (MIMO) Q Network framework. The framework aims to enhance Offline Reinforcement Learning by fully leveraging the potential of OOD data while still ensuring efficiency in the learning process. Specifically, the framework quantifies data uncertainty and harnesses it in the training losses, aiming to train a policy that maximizes the lower confidence bound of the corresponding Q-function. Furthermore, a Rank-One MIMO architecture is introduced to model the uncertainty-aware Q-function, \TP{offering the same ability for uncertainty quantification as an ensemble of networks but with a cost nearly equivalent to that of a single network}. Consequently, this framework strikes a harmonious balance between precision, speed, and memory efficiency, culminating in improved overall performance. Extensive experimentation on the D4RL benchmark demonstrates that the framework attains state-of-the-art performance while remaining computationally efficient. By incorporating the concept of uncertainty quantification, our framework offers a promising avenue to alleviate extrapolation errors and enhance the efficiency of offline RL.

Understanding and predicting microstructure evolution is fundamental to materials science, as it governs the resulting properties and performance of materials. Traditional simulation methods, such as phase-field models, offer high-fidelity results but are computationally expensive due to the need to solve complex partial differential equations at fine spatiotemporal resolutions. To address this challenge, we propose a deep learning-based framework that accelerates microstructure evolution predictions while maintaining high accuracy. Our approach utilizes a fully convolutional spatiotemporal model trained in a self-supervised manner using sequential images generated from simulations of microstructural processes, including grain growth and spinodal decomposition. The trained neural network effectively learns the underlying physical dynamics and can accurately capture both short-term local behaviors and long-term statistical properties of evolving microstructures, while also demonstrating generalization to unseen spatiotemporal domains and variations in configuration and material parameters. Compared to recurrent neural architectures, our model achieves state-of-the-art predictive performance with significantly reduced computational cost in both training and inference. This work establishes a robust baseline for spatiotemporal learning in materials science and offers a scalable, data-driven alternative for fast and reliable microstructure simulations.

Existing benchmarks for AI reasoning provide limited insight into how closely these capabilities resemble human reasoning in naturalistic contexts. We present an adaptation of the Watson & Holmes detective tabletop game as a new benchmark designed to evaluate reasoning performance using incrementally presented narrative evidence, open-ended questions and unconstrained language responses. An automated grading system was developed and validated against human assessors to enable scalable and replicable performance evaluation. Results show a clear improvement in AI model performance over time. Over nine months of 2025, model performance rose from the lower quartile of the human comparison group to approximately the top 5%. Around half of this improvement reflects steady advancement across successive model releases, while the remainder corresponds to a marked step change associated with reasoning-oriented model architectures. Systematic differences in the performance of AI models compared to humans, dependent on features of the specific detection puzzle, were mostly absent with the exception of a fall in performance for models when solving longer cases (case lengths being in the range of 1900-4000 words), and an advantage at inductive reasoning for reasoning models at early stages of case solving when evidence was scant.

Mass spectrometry is a powerful and widely used tool for identifying molecular structures due to its sensitivity and ability to profile complex samples. However, translating spectra into full molecular structures is a difficult, under-defined inverse problem. Overcoming this problem is crucial for enabling biological insight, discovering new metabolites, and advancing chemical research across multiple fields. To this end, we develop MSFlow, a two-stage encoder-decoder flow-matching generative model that achieves state-of-the-art performance on the structure elucidation task for small molecules. In the first stage, we adopt a formula-restricted transformer model for encoding mass spectra into a continuous and chemically informative embedding space, while in the second stage, we train a decoder flow matching model to reconstruct molecules from latent embeddings of mass spectra. We present ablation studies demonstrating the importance of using information-preserving molecular descriptors for encoding mass spectra and motivate the use of our discrete flow-based decoder. Our rigorous evaluation demonstrates that MSFlow can accurately translate up to 45 percent of molecular mass spectra into their corresponding molecular representations - an improvement of up to fourteen-fold over the current state-of-the-art. A trained version of MSFlow is made publicly available on GitHub for non-commercial users.

In this paper, we propose a novel selection strategy for contrastive learning for medical images. On natural images, contrastive learning uses augmentations to select positive and negative pairs for the contrastive loss. However, in the medical domain, arbitrary augmentations have the potential to distort small localized regions that contain the biomarkers we are interested in detecting. A more intuitive approach is to select samples with similar disease severity characteristics, since these samples are more likely to have similar structures related to the progression of a disease. To enable this, we introduce a method that generates disease severity labels for unlabeled OCT scans on the basis of gradient responses from an anomaly detection algorithm. These labels are used to train a supervised contrastive learning setup to improve biomarker classification accuracy by as much as 6% above self-supervised baselines for key indicators of Diabetic Retinopathy.

Causal discovery problems use a set of observations to deduce causality between variables in the real world, typically to answer questions about biological or physical systems. These observations are often recorded at regular time intervals, determined by a user or a machine, depending on the experiment design. There is generally no guarantee that the timing of these recordings matches the timing of the underlying biological or physical events. In this paper, we examine the sensitivity of causal discovery methods to this potential mismatch. We consider empirical and theoretical evidence to understand how causal discovery performance is impacted by changes of sampling rate and window length. We demonstrate that both classical and recent causal discovery methods exhibit sensitivity to these hyperparameters, and we discuss how ideas from signal processing may help us understand these phenomena.

Reinforcement learning with verifiers (RLVR) is a central paradigm for improving large language model (LLM) reasoning, yet existing methods often suffer from limited exploration. Policies tend to collapse onto a few reasoning patterns and prematurely stop deep exploration, while conventional entropy regularization introduces only local stochasticity and fails to induce meaningful path-level diversity, leading to weak and unstable learning signals in group-based policy optimization. We propose DSDR, a Dual-Scale Diversity Regularization reinforcement learning framework that decomposes diversity in LLM reasoning into global and coupling components. Globally, DSDR promotes diversity among correct reasoning trajectories to explore distinct solution modes. Locally, it applies a length-invariant, token-level entropy regularization restricted to correct trajectories, preventing entropy collapse within each mode while preserving correctness. The two scales are coupled through a global-to-local allocation mechanism that emphasizes local regularization for more distinctive correct trajectories. We provide theoretical support showing that DSDR preserves optimal correctness under bounded regularization, sustains informative learning signals in group-based optimization, and yields a principled global-to-local coupling rule. Experiments on multiple reasoning benchmarks demonstrate consistent improvements in accuracy and pass@k, highlighting the importance of dual-scale diversity for deep exploration in RLVR. Code is available at https://github.com/SUSTechBruce/DSDR.

We present a family of generalized Hessian estimators of the objective using random direction stochastic approximation (RDSA) by utilizing only noisy function measurements. The form of each estimator and the order of the bias depend on the number of function measurements. In particular, we demonstrate that estimators with more function measurements exhibit lower-order estimation bias. We show the asymptotic unbiasedness of the estimators. We also perform asymptotic and non-asymptotic convergence analyses for stochastic Newton methods that incorporate our generalized Hessian estimators. Finally, we perform numerical experiments to validate our theoretical findings.

Computer architecture is searching for new ways to make use of increasingly available digital logic without the serial bottlenecks of CPU-based design. Recent work has demonstrated a fully CPU-less approach to executing functional programs, by exploiting their inherent parallelisability to compile them directly into parallel digital logic. This work uses lambda-calculus as a hyper simple functional language to prove the concept, but is impractical for real-world programming due to the well-known inefficiencies of pure lambda$-calculus. It is common in language design to extend basic lambda-calculus with additional primitives to short-cut common tasks such as arithmetic and lists. In this work, we build upon our previous research to examine how such extensions may be applied to CPU-less functional execution in digital logic, with the objective of advancing the approach toward practical implementation. We present a set of structures and algorithms for representing new primitives, describe a systematic process for selecting, implementing, and evaluating them, and demonstrate substantial reductions in execution time and node usage. These improvements are implemented in an open-source system, which is shown to correctly evaluate a range of representative lambda expressions.

ODRL's six set-based operators -- isA, isPartOf, hasPart, isAnyOf, isAllOf, isNoneOf -- depend on external domain knowledge that the W3C specification leaves unspecified. Without it, every cross-dataspace policy comparison defaults to Unknown. We present a denotational semantics that maps each ODRL constraint to the set of knowledge-base concepts satisfying it. Conflict detection reduces to denotation intersection under a three-valued verdict -- Conflict, Compatible, or Unknown -- that is sound under incomplete knowledge. The framework covers all three ODRL composition modes (and, or, xone) and all three semantic domains arising in practice: taxonomic (class subsumption), mereological (part-whole containment), and nominal (identity). For cross-dataspace interoperability, we define order-preserving alignments between knowledge bases and prove two guarantees: conflicts are preserved across different KB standards, and unmapped concepts degrade gracefully to Unknown -- never to false conflicts. A runtime soundness theorem ensures that design-time verdicts hold for all execution contexts. The encoding stays within the decidable EPR fragment of first-order logic. We validate it with 154 benchmarks across six knowledge base families (GeoNames, ISO 3166, W3C DPV, a GDPR-derived taxonomy, BCP 47, and ISO 639-3) and four structural KBs targeting adversarial edge cases. Both the Vampire theorem prover and the Z3 SMT solver agree on all 154 verdicts. A key finding is that exclusive composition (xone) requires strictly stronger KB axioms than conjunction or disjunction: open-world semantics blocks exclusivity even when positive evidence appears to satisfy exactly one branch.

Unsupervised change detection (UCD) in remote sensing aims to localise semantic changes between two images of the same region without relying on labelled data during training. Most recent approaches rely either on frozen foundation models in a training-free manner or on training with synthetic changes generated in pixel space. Both strategies inherently rely on predefined assumptions about change types, typically introduced through handcrafted rules, external datasets, or auxiliary generative models. Due to these assumptions, such methods fail to generalise beyond a few change types, limiting their real-world usage, especially in rare or complex scenarios. To address this, we propose MaSoN (Make Some Noise), an end-to-end UCD framework that synthesises diverse changes directly in the latent feature space during training. It generates changes that are dynamically estimated using feature statistics of target data, enabling diverse yet data-driven variation aligned with the target domain. It also easily extends to new modalities, such as SAR. MaSoN generalises strongly across diverse change types and achieves state-of-the-art performance on five benchmarks, improving the average F1 score by 14.1 percentage points. Project page: https://blaz-r.github.io/mason_ucd

Every ODRL 2.2 constraint compares a single scalar value: (leftOperand, operator, rightOperand). Five of ODRL's approximately 34 left operands, however, denote multi-dimensional quantities--image dimensions, canvas positions, geographic coordinates--whose specification text explicitly references multiple axes. For these operands, a single scalar constraint admits one interpretation per axis, making policy evaluation non-deterministic. We classify ODRL's left operands by value-domain structure (scalar, dimensional, concept-valued), grounded in the ODRL 2.2 specification text, and show that dimensional ambiguity is intrinsic to the constraint syntax. We present an axis-decomposition framework that refines each dimensional operand into axis-specific scalar operands and prove four properties: deterministic interpretation, AABB completeness, sound over-approximation under projection, and conservative extension. Conflict detection operates in two layers: per-axis verdicts are always decidable; box-level verdicts compose through Strong Kleene conjunction into a three-valued logic (Conflict, Compatible, Unknown). For ODRL's disjunctive (odrl:or) and exclusive-or (odrl:xone) logical constraints, where per-axis decomposition does not apply, the framework encodes coupled multi-axis conjectures directly. We instantiate the framework as the ODRL Spatial Axis Profile--15 axis-specific left operands for the five affected base terms--and evaluate it on 117 benchmark problems spanning nine categories across both TPTP FOF (Vampire) and SMT-LIB (Z3) encodings, achieving full concordance between provers. Benchmark scenarios are inspired by constraints arising in cultural heritage dataspaces such as Datenraum Kultur. All meta-theorems are mechanically verified in Isabelle/HOL.

Continual Generalized Category Discovery (C-GCD) requires identifying novel classes from unlabeled data while retaining knowledge of known classes over time. Existing methods typically update classifier weights dynamically, resulting in forgetting and inconsistent feature alignment. We propose GOAL, a unified framework that introduces a fixed Equiangular Tight Frame (ETF) classifier to impose a consistent geometric structure throughout learning. GOAL conducts supervised alignment for labeled samples and confidence-guided alignment for novel samples, enabling stable integration of new classes without disrupting old ones. Experiments on four benchmarks show that GOAL outperforms the prior method Happy, reducing forgetting by 16.1% and boosting novel class discovery by 3.2%, establishing a strong solution for long-horizon continual discovery.

Neural networks are the cornerstone of modern machine learning, yet can be difficult to interpret, give overconfident predictions and are vulnerable to adversarial attacks. Bayesian neural networks (BNNs) provide some alleviation of these limitations, but have problems of their own. The key step of specifying prior distributions in BNNs is no trivial task, yet is often skipped out of convenience. In this work, we propose a new class of prior distributions for BNNs, the Dirichlet scale mixture (DSM) prior, that addresses current limitations in Bayesian neural networks through structured, sparsity-inducing shrinkage. Theoretically, we derive general dependence structures and shrinkage results for DSM priors and show how they manifest under the geometry induced by neural networks. In experiments on simulated and real world data we find that the DSM priors encourages sparse networks through implicit feature selection, show robustness under adversarial attacks and deliver competitive predictive performance with substantially fewer effective parameters. In particular, their advantages appear most pronounced in correlated, moderately small data regimes, and are more amenable to weight pruning. Moreover, by adopting heavy-tailed shrinkage mechanisms, our approach aligns with recent findings that such priors can mitigate the cold posterior effect, offering a principled alternative to the commonly used Gaussian priors.

We present SHIELD, a novel methodology for automated and integrated safety signal detection in clinical trials. SHIELD combines disproportionality analysis with semantic clustering of adverse event (AE) terms applied to MedDRA term embeddings. For each AE, the pipeline computes an information-theoretic disproportionality measure (Information Component) with effect size derived via empirical Bayesian shrinkage. A utility matrix is constructed by weighting semantic term-term similarities by signal magnitude, followed by spectral embedding and clustering to identify groups of related AEs. Resulting clusters are annotated with syndrome-level summary labels using large language models, yielding a coherent, data-driven representation of treatment-associated safety profiles in the form of a network graph and hierarchical tree. We implement the SHIELD framework in the context of a single-arm incidence summary, to compare two treatment arms or for the detection of any treatment effect in a multi-arm trial. We illustrate its ability to recover known safety signals and generate interpretable, cluster-based summaries in a real clinical trial example. This work bridges statistical signal detection with modern natural language processing to enhance safety assessment and causal interpretation in clinical trials.

We study uplift estimation for combinatorial treatments. Uplift measures the pure incremental causal effect of an intervention (e.g., sending a coupon or a marketing message) on user behavior, modeled as a conditional individual treatment effect. Many real-world interventions are combinatorial: a treatment is a policy that specifies context-dependent action distributions rather than a single atomic label. Although recent work considers structured treatments, most methods rely on categorical or opaque encodings, limiting robustness and generalization to rare or newly deployed policies. We propose an uplift estimation framework that aligns treatment representation with causal semantics. Each policy is represented by the mixture it induces over contextaction components and embedded via a permutation-invariant aggregation. This representation is integrated into an orthogonalized low-rank uplift model, extending Robinson-style decompositions to learned, vector-valued treatments. We show that the resulting estimator is expressive for policy-induced causal effects, orthogonally robust to nuisance estimation errors, and stable under small policy perturbations. Experiments on large-scale randomized platform data demonstrate improved uplift accuracy and stability in long-tailed policy regimes

A recent Dwarkesh Patel podcast with John Collison and Elon Musk featured an interesting puzzle from Jane Street: they trained a neural net, shuffled all 96 layers, and asked to put them back in order. Given unlabelled layers of a Residual Network and its training dataset, we recover the exact ordering of the layers. The problem decomposes into pairing each block's input and output projections ($48!$ possibilities) and ordering the reassembled blocks ($48!$ possibilities), for a combined search space of $(48!)^2 \approx 10^{122}$, which is more than the atoms in the observable universe. We show that stability conditions during training like dynamic isometry leave the product $W_{\text{out}} W_{\text{in}}$ for correctly paired layers with a negative diagonal structure, allowing us to use diagonal dominance ratio as a signal for pairing. For ordering, we seed-initialize with a rough proxy such as delta-norm or $\|W_{\text{out}}\|_F$ then hill-climb to zero mean squared error.

LLM-enabled applications are rapidly reshaping the software ecosystem by using large language models as core reasoning components for complex task execution. This paradigm shift, however, introduces fundamentally new reliability challenges and significantly expands the security attack surface, due to the non-deterministic, learning-driven, and difficult-to-verify nature of LLM behavior. In light of these emerging and unavoidable safety challenges, we argue that such risks should be treated as expected operational conditions rather than exceptional events, necessitating a dedicated incident-response perspective. Consequently, the primary barrier to trustworthy deployment is not further improving model capability but establishing system-level threat monitoring mechanisms that can detect and contextualize security-relevant anomalies after deployment -- an aspect largely underexplored beyond testing or guardrail-based defenses. Accordingly, this position paper advocates systematic and comprehensive monitoring of security threats in LLM-enabled applications as a prerequisite for reliable operation and a foundation for dedicated incident-response frameworks.

As LLM-based Multi-Agent Systems (MAS) are increasingly deployed for complex tasks, ensuring their reliability has become a pressing challenge. Since MAS coordinate through unstructured natural language rather than rigid protocols, they are prone to semantic failures (e.g., hallucinations, misinterpreted instructions, and reasoning drift) that propagate silently without raising runtime exceptions. Prevailing evaluation approaches, which measure only end-to-end task success, offer limited insight into how these failures arise or how effectively agents recover from them. To bridge this gap, we propose MAS-FIRE, a systematic framework for fault injection and reliability evaluation of MAS. We define a taxonomy of 15 fault types covering intra-agent cognitive errors and inter-agent coordination failures, and inject them via three non-invasive mechanisms: prompt modification, response rewriting, and message routing manipulation. Applying MAS-FIRE to three representative MAS architectures, we uncover a rich set of fault-tolerant behaviors that we organize into four tiers: mechanism, rule, prompt, and reasoning. This tiered view enables fine-grained diagnosis of where and why systems succeed or fail. Our findings reveal that stronger foundation models do not uniformly improve robustness. We further show that architectural topology plays an equally decisive role, with iterative, closed-loop designs neutralizing over 40% of faults that cause catastrophic collapse in linear workflows. MAS-FIRE provides the process-level observability and actionable guidance needed to systematically improve multi-agent systems.

Modern large language models (LLMs) excel at generating fluent and faithful translations. However, they struggle to preserve an author's unique literary style, often producing semantically correct but generic outputs. This limitation stems from the inability of current single-model and static multi-agent systems to perceive and adapt to stylistic variations. To address this, we introduce the Style-Adaptive Multi-Agent System (SAMAS), a novel framework that treats style preservation as a signal processing task. Specifically, our method quantifies literary style into a Stylistic Feature Spectrum (SFS) using the wavelet packet transform. This SFS serves as a control signal to dynamically assemble a tailored workflow of specialized translation agents based on the source text's structural patterns. Extensive experiments on translation benchmarks show that SAMAS achieves competitive semantic accuracy against strong baselines, primarily by leveraging its statistically significant advantage in style fidelity.

Humans are highly effective at utilizing prior knowledge to adapt to novel tasks, a capability that standard machine learning models struggle to replicate due to their reliance on task-specific training. Meta-learning overcomes this limitation by allowing models to acquire transferable knowledge from various tasks, enabling rapid adaptation to new challenges with minimal data. This survey provides a rigorous, task-based formalization of meta-learning and meta-reinforcement learning and uses that paradigm to chronicle the landmark algorithms that paved the way for DeepMind's Adaptive Agent, consolidating the essential concepts needed to understand the Adaptive Agent and other generalist approaches.

Early detection of myometrial invasion is critical for the staging and life-saving management of endometrial carcinoma (EC), a prevalent global malignancy. Transvaginal ultrasound serves as the primary, accessible screening modality in resource-constrained primary care settings; however, its diagnostic reliability is severely hindered by low tissue contrast, high operator dependence, and a pronounced scarcity of positive pathological samples. Existing artificial intelligence solutions struggle to overcome this severe class imbalance and the subtle imaging features of invasion, particularly under the strict computational limits of primary care clinics. Here we present an automated, highly efficient two-stage deep learning framework that resolves both data and computational bottlenecks in EC screening. To mitigate pathological data scarcity, we develop a structure-guided cross-modal generation network that synthesizes diverse, high-fidelity ultrasound images from unpaired magnetic resonance imaging (MRI) data, strictly preserving clinically essential anatomical junctions. Furthermore, we introduce a lightweight screening network utilizing gradient distillation, which transfers discriminative knowledge from a high-capacity teacher model to dynamically guide sparse attention towards task-critical regions. Evaluated on a large, multicenter cohort of 7,951 participants, our model achieves a sensitivity of 99.5\%, a specificity of 97.2\%, and an area under the curve of 0.987 at a minimal computational cost (0.289 GFLOPs), substantially outperforming the average diagnostic accuracy of expert sonographers. Our approach demonstrates that combining cross-modal synthetic augmentation with knowledge-driven efficient modeling can democratize expert-level, real-time cancer screening for resource-constrained primary care settings.

Model repositories such as Hugging Face increasingly distribute machine learning artifacts serialized with Python's pickle format, exposing users to remote code execution (RCE) risks during model loading. Recent defenses, such as PickleBall, rely on per-library policy synthesis that requires complex system setups and verified benign models, which limits scalability and generalization. In this work, we propose a lightweight, machine-learning-based scanner that detects malicious Pickle-based files without policy generation or code instrumentation. Our approach statically extracts structural and semantic features from Pickle bytecode and applies supervised and unsupervised models to classify files as benign or malicious. We construct and release a labeled dataset of 727 Pickle-based files from Hugging Face and evaluate our models on four datasets: our own, PickleBall (out-of-distribution), Hide-and-Seek (9 advanced evasive malicious models), and synthetic joblib files. Our method achieves 90.01% F1-score compared with 7.23%-62.75% achieved by the SOTA scanners (Modelscan, Fickling, ClamAV, VirusTotal) on our dataset. Furthermore, on the PickleBall data (OOD), it achieves 81.22% F1-score compared with 76.09% achieved by the PickleBall method, while remaining fully library-agnostic. Finally, we show that our method is the only one to correctly parse and classify 9/9 evasive Hide-and-Seek malicious models specially crafted to evade scanners. This demonstrates that data-driven detection can effectively and generically mitigate Pickle-based model file attacks.

Long-context modeling is essential for symbolic music generation, since motif repetition and developmental variation can span thousands of musical events. However, practical composition and performance workflows frequently rely on resource-limited devices (e.g., electronic instruments and portable computers), making heavy memory and attention computation difficult to deploy. We introduce Depth-Structured Music Recurrence (DSMR), a recurrent long-context Transformer for full-piece symbolic music modeling that extends context beyond fixed-length excerpts via segment-level recurrence with detached cross-segment states, featuring a layer-wise memory-horizon schedule that budgets recurrent KV states across depth. DSMR is trained in a single left-to-right pass over each complete composition, akin to how a musician experiences it from beginning to end, while carrying recurrent cross-segment states forward. Within this recurrent framework, we systematically study how depth-wise horizon allocations affect optimization, best-checkpoint perplexity, and efficiency. By allocating different history-window lengths across layers while keeping the total recurrent-state budget fixed, DSMR creates depth-dependent temporal receptive fields within a recurrent attention stack without reducing compute depth. Our main instantiation is a two-scale DSMR schedule that allocates long history windows to lower layers and a uniform short window to the remaining layers. Experiments on the piano performance dataset MAESTRO demonstrate that two-scale DSMR provides a practical quality--efficiency recipe for full-length long-context symbolic music modeling with recurrent attention under limited computational resources.

We present a mobile and desktop keyboard suite for Idu Mishmi, an endangered Trans-Himalayan language spoken by approximately 11,000 people in Arunachal Pradesh, India. Although a Latin-based orthography was developed in 2018, no digital input tools existed to use it, forcing speakers into ad-hoc romanizations that cannot represent the full writing system. Our keyboards comprise two tools: (1) an Android mobile keyboard, published on the Google Play Store and actively used in teacher training programs, and (2) a Windows desktop keyboard currently undergoing community testing. Both tools support the complete Idu Mishmi character inventory, including schwa, retracted schwa, nasalized vowels, and accented forms. Both operate fully offline with zero network permissions, addressing connectivity constraints and data sovereignty concerns. We describe the design, implementation, and deployment as a replicable model for other endangered language communities.

In January 2026, the open-source agent framework OpenClaw and the agent-only social network Moltbook produced a large-scale dataset of autonomous AI-to-AI interaction, attracting six academic publications within fourteen days. This study conducts a multivocal literature review of that ecosystem and presents ClawdLab, an open-source platform for autonomous scientific research, as a design science response to the architectural failure modes identified. The literature documents emergent collective phenomena, security vulnerabilities spanning 131 agent skills and over 15,200 exposed control panels, and five recurring architectural patterns. ClawdLab addresses these failure modes through hard role restrictions, structured adversarial critique, PI-led governance, multi-model orchestration, and domain-specific evidence requirements encoded as protocol constraints that ground validation in computational tool outputs rather than social consensus; the architecture provides emergent Sybil resistance as a structural consequence. A three-tier taxonomy distinguishes single-agent pipelines, predetermined multi-agent workflows, and fully decentralised systems, analysing why leading AI co-scientist platforms remain confined to the first two tiers. ClawdLab's composable third-tier architecture, in which foundation models, capabilities, governance, and evidence requirements are independently modifiable, enables compounding improvement as the broader AI ecosystem advances.

Mamba-based models have drawn much attention in offline RL. However, their selective mechanism often detrimental when key steps in RL sequences are omitted. To address these issues, we propose a simple yet effective structure, called Decision MetaMamba (DMM), which replaces Mamba's token mixer with a dense layer-based sequence mixer and modifies positional structure to preserve local information. By performing sequence mixing that considers all channels simultaneously before Mamba, DMM prevents information loss due to selective scanning and residual gating. Extensive experiments demonstrate that our DMM delivers the state-of-the-art performance across diverse RL tasks. Furthermore, DMM achieves these results with a compact parameter footprint, demonstrating strong potential for real-world applications.

We introduce a diagonalization-based optimization for Linear Echo State Networks (ESNs) that reduces the per-step computational complexity of reservoir state updates from O(N^2) to O(N). By reformulating reservoir dynamics in the eigenbasis of the recurrent matrix, the recurrent update becomes a set of independent element-wise operations, eliminating the matrix multiplication. We further propose three methods to use our optimization depending on the situation: (i) Eigenbasis Weight Transformation (EWT), which preserves the dynamics of standard and trained Linear ESNs, (ii) End-to-End Eigenbasis Training (EET), which directly optimizes readout weights in the transformed space and (iii) Direct Parameter Generation (DPG), that bypasses matrix diagonalization by directly sampling eigenvalues and eigenvectors, achieving comparable performance than standard Linear ESNs. Across all experiments, both our methods preserve predictive accuracy while offering significant computational speedups, making them a replacement of standard Linear ESNs computations and training, and suggesting a shift of paradigm in linear ESN towards the direct selection of eigenvalues.

Despite recent algorithmic advances, we still lack principled ways to leverage the well-documented rescaling symmetries in ReLU neural network parameters. While two properly rescaled weights implement the same function, the training dynamics can be dramatically different. To offer a fresh perspective on exploiting this phenomenon, we build on the recent path-lifting framework, which provides a compact factorization of ReLU networks. We introduce a geometrically motivated criterion to rescale neural network parameters which minimization leads to a conditioning strategy that aligns a kernel in the path-lifting space with a chosen reference. We derive an efficient algorithm to perform this alignment. In the context of random network initialization, we analyze how the architecture and the initialization scale jointly impact the output of the proposed method. Numerical experiments illustrate its potential to speed up training.

Concept drift -- the change of the distribution over time -- poses significant challenges for learning systems and is of central interest for monitoring. Understanding drift is thus paramount, and drift localization -- determining which samples are affected by the drift -- is essential. While several approaches exist, most rely on local testing schemes, which tend to fail in high-dimensional, low-signal settings. In this work, we consider a fundamentally different approach based on conformal predictions. We discuss and show the shortcomings of common approaches and demonstrate the performance of our approach on state-of-the-art image datasets.

This position paper argues that the machine learning community must move from preaching to practising data frugality for responsible artificial intelligence (AI) development. For long, progress has been equated with ever-larger datasets, driving remarkable advances but now yielding increasingly diminishing performance gains alongside rising energy use and carbon emissions. While awareness of data frugal approaches has grown, their adoption has remained rhetorical, and data scaling continues to dominate development practice. We argue that this gap between preach and practice must be closed, as continued data scaling entails substantial and under-accounted environmental impacts. To ground our position, we provide indicative estimates of the energy use and carbon emissions associated with the downstream use of ImageNet-1K. We then present empirical evidence that data frugality is both practical and beneficial, demonstrating that coreset-based subset selection can substantially reduce training energy consumption with little loss in accuracy, while also mitigating dataset bias. Finally, we outline actionable recommendations for moving data frugality from rhetorical preach to concrete practice for responsible development of AI.

Meta-learning methods perform well on new within-distribution tasks but often fail when adapting to out-of-distribution target tasks, where transfer from source tasks can induce negative transfer. We propose a causally-aware Bayesian meta-learning method, by conditioning task-specific priors on precomputed latent causal task embeddings, enabling transfer based on mechanistic similarity rather than spurious correlations. Our approach explicitly considers realistic deployment settings where access to target-task data is limited, and adaptation relies on noisy (expert-provided) pairwise judgments of causal similarity between source and target tasks. We provide a theoretical analysis showing that conditioning on causal embeddings controls prior mismatch and mitigates negative transfer under task shift. Empirically, we demonstrate reductions in negative transfer and improved out-of-distribution adaptation in both controlled simulations and a large-scale real-world clinical prediction setting for cross-disease transfer, where causal embeddings align with underlying clinical mechanisms.

Climate change impacts a broad spectrum of human resources and activities, necessitating the use of climate models to project long-term effects and inform mitigation and adaptation strategies. These models generate multiple datasets by running simulations across various scenarios and configurations, thereby covering a range of potential future outcomes. Currently, researchers rely on traditional search interfaces and APIs to retrieve such datasets, often piecing together information from metadata and community vocabularies. The Climate Change Knowledge Graph is designed to address these challenges by integrating diverse data sources related to climate simulations into a coherent and interoperable knowledge graph. This innovative resource allows for executing complex queries involving climate models, simulations, variables, spatio-temporal domains, and granularities. Developed with input from domain experts, the knowledge graph and its underlying ontology are published with open access license and provide a comprehensive framework that enhances the exploration of climate data, facilitating more informed decision-making in addressing climate change issues.

As Operational Technology increasingly integrates with Information Technology, the need for Intrusion Detection Systems becomes more important. This paper explores an unsupervised approach to anomaly detection in network traffic using $β$-Variational Autoencoders on the NSL-KDD dataset. We investigate two methods: leveraging the latent space structure by measuring distances from test samples to the training data projections, and using the reconstruction error as a conventional anomaly detection metric. By comparing these approaches, we provide insights into their respective advantages and limitations in an unsupervised setting. Experimental results highlight the effectiveness of latent space exploitation for classification tasks.

Mendelian Randomization (MR) is a prominent observational epidemiological research method designed to address unobserved confounding when estimating causal effects. However, core assumptions -- particularly the independence between instruments and unobserved confounders -- are often violated due to population stratification or assortative mating. Leveraging the increasing availability of multi-environment data, we propose a representation learning framework that exploits cross-environment invariance to recover latent exogenous components of genetic instruments. We provide theoretical guarantees for identifying these latent instruments under various mixing mechanisms and demonstrate the effectiveness of our approach through simulations and semi-synthetic experiments using data from the All of Us Research Hub.

Automatic Chord Recognition (ACR) is constrained by the scarcity of aligned chord labels, as well-aligned annotations are costly to acquire. At the same time, open-weight pre-trained models are currently more accessible than their proprietary training data. In this work, we present a two-stage training pipeline that leverages pre-trained models together with unlabeled audio. The proposed method decouples training into two stages. In the first stage, we use a pre-trained BTC model as a teacher to generate pseudo-labels for over 1,000 hours of diverse unlabeled audio and train a student model solely on these pseudo-labels. In the second stage, the student is continually trained on ground-truth labels as they become available, with selective knowledge distillation (KD) from the teacher applied as a regularizer to prevent catastrophic forgetting of the representations learned in the first stage. In our experiments, two models (BTC, 2E1D) were used as students. In stage 1, using only pseudo-labels, the BTC student achieves over 98% of the teacher's performance, while the 2E1D model achieves about 96% across seven standard mir_eval metrics. After a single training run for both students in stage 2, the resulting BTC student model surpasses the traditional supervised learning baseline by 2.5% and the original pre-trained teacher model by 1.55% on average across all metrics. And the resulting 2E1D student model improves from the traditional supervised learning baseline by 3.79% on average and achieves almost the same performance as the teacher. Both cases show the large gains on rare chord qualities.

Exact stochastic simulation of continuous-time Markov chains (CTMCs) is essential when discreteness and noise drive system behavior, but the hard categorical event selection in Gillespie-type algorithms blocks gradient-based learning. We eliminate this constraint by decoupling forward simulation from backward differentiation, with hard categorical sampling generating exact trajectories and gradients propagating through a continuous massively-parallel Gumbel-Softmax straight-through surrogate. Our approach enables accurate optimization at parameter scales over four orders of magnitude beyond existing simulators. We validate for accuracy, scalability, and reliability on a reversible dimerization model (0.09% error), a genetic oscillator (1.2% error), a 203,796-parameter gene regulatory network achieving 98.4% MNIST accuracy (a prototypical deep-learning multilayer perceptron benchmark), and experimental patch-clamp recordings of ion channel gating (R^2 = 0.987) in the single-channel regime. Our GPU implementation delivers 1.9 billion steps per second, matching the scale of non-differentiable simulators. By making exact stochastic simulation massively parallel and autodiff-compatible, our results enable high-dimensional parameter inference and inverse design across systems biology, chemical kinetics, physics, and related CTMC-governed domains.

Explainable artificial intelligence has emerged as a promising field of research to address reliability concerns in artificial intelligence. Despite significant progress in explainable artificial intelligence, few methods provide a systematic way to visualize and understand how classes are confused and how their relationships evolve as training progresses. In this work, we present GRAPHIC, an architecture-agnostic approach that analyzes neural networks on a class level. It leverages confusion matrices derived from intermediate layers using linear classifiers. We interpret these as adjacency matrices of directed graphs, allowing tools from network science to visualize and quantify learning dynamics across training epochs and intermediate layers. GRAPHIC provides insights into linear class separability, dataset issues, and architectural behavior, revealing, for example, similarities between flatfish and man and labeling ambiguities validated in a human study. In summary, by uncovering real confusions, GRAPHIC offers new perspectives on how neural networks learn. The code is available at https://github.com/Johanna-S-Froehlich/GRAPHIC.

In this paper, we present Hexagon-MLIR,an open-source compilation stack that targets Qualcomm Hexagon Neural Processing Unit (NPU) and provides unified support for lowering Triton kernels and PyTorch models . Built using the MLIR framework, our compiler applies a structured sequence of passes to exploit NPU architectural features to accelerate AI workloads. It enables faster deployment of new Triton kernels (hand-written or subgraphs from PyTorch 2.0), for our target by providing automated compilation from kernel to binary. By ingesting Triton kernels, we generate mega-kernels that maximize data locality in the NPU's Tightly Coupled Memory (TCM), reducing the bandwidth bottlenecks inherent in library-based approaches. This initiative complements our commercial toolchains by providing developers with an open-source MLIR-based compilation stack that gives them a path to advance AI compilation capabilities through a more flexible approach. Hexagon-MLIR is a work-in-progress, and we are continuing to add many more optimizations and capabilities in this effort.

Dynamic predictions for longitudinal and time-to-event outcomes have become a versatile tool in precision medicine. Our work is motivated by the application of dynamic predictions in the decision-making process for primary biliary cholangitis patients. For these patients, serial biomarker measurements (e.g., bilirubin and alkaline phosphatase levels) are routinely collected to inform treating physicians of the risk of liver failure and guide clinical decision-making. Two popular statistical approaches to derive dynamic predictions are joint modelling and landmarking. However, recently, machine learning techniques have also been proposed. Each approach has its merits, and no single method exists to outperform all others. Consequently, obtaining the best possible survival estimates is challenging. Therefore, we extend the Super Learner framework to combine dynamic predictions from different models and procedures. Super Learner is an ensemble learning technique that allows users to combine different prediction algorithms to improve predictive accuracy and flexibility. It uses cross-validation and different objective functions of performance (e.g., squared loss) that suit specific applications to build the optimally weighted combination of predictions from a library of candidate algorithms. In our work, we pay special attention to appropriate objective functions for Super Learner to obtain the most optimal weighted combination of dynamic predictions. In our primary biliary cholangitis application, Super Learner presented unique benefits due to its ability to flexibly combine outputs from a diverse set of models with varying assumptions for equal or better predictive performance than any model fit separately.

This paper explores how natural-language descriptions of formal languages can be compared to their formal representations and how semantic differences can be explained. This is motivated from educational scenarios where learners describe a formal language (presented, e.g., by a finite state automaton, regular expression, pushdown automaton, context-free grammar or in set notation) in natural language, and an educational support system has to (1) judge whether the natural-language description accurately describes the formal language, and to (2) provide explanations why descriptions are not accurate. To address this question, we introduce a representation language for formal languages, Nile, which is designed so that Nile expressions can mirror the syntactic structure of natural-language descriptions of formal languages. Nile is sufficiently expressive to cover a broad variety of formal languages, including all regular languages and fragments of context-free languages typically used in educational contexts. Generating Nile expressions that are syntactically close to natural-language descriptions then allows to provide explanations for inaccuracies in the descriptions algorithmically. In experiments on an educational data set, we show that LLMs can translate natural-language descriptions into equivalent, syntactically close Nile expressions with high accuracy - allowing to algorithmically provide explanations for incorrect natural-language descriptions. Our experiments also show that while natural-language descriptions can also be translated into regular expressions (but not context-free grammars), the expressions are often not syntactically close and thus not suitable for providing explanations.

Wildfire monitoring demands timely data collection and processing for early detection and rapid response. UAV-assisted edge computing is a promising approach, but jointly minimizing end-to-end service response time while satisfying energy, revisit time, and capacity constraints remains challenging. We propose an integrated framework that co-optimizes UAV route planning, fleet sizing, and edge service provisioning for wildfire monitoring. The framework combines fire-history-weighted clustering to prioritize high-risk areas, Quality of Service (QoS)-aware edge assignment balancing proximity and computational load, 2-opt route optimization with adaptive fleet sizing, and a dynamic emergency rerouting mechanism. The key insight is that these subproblems are interdependent: clustering decisions simultaneously shape patrol efficiency and edge workloads, while capacity constraints feed back into feasible configurations. Experiments show that the proposed framework reduces average response time by 70.6--84.2%, energy consumption by 73.8--88.4%, and fleet size by 26.7--42.1% compared to GA, PSO, and greedy baselines. The emergency mechanism responds within 233 seconds, well under the 300-second deadline, with negligible impact on normal operations.

Algorithm unrolling is ubiquitous in machine learning, particularly in hyperparameter optimization and meta-learning, where Jacobians of solution mappings are computed by differentiating through iterative algorithms. Although unrolling is known to yield asymptotically correct Jacobians under suitable conditions, recent work has shown that the derivative iterates may initially diverge from the true Jacobian, a phenomenon known as the curse of unrolling. In this work, we provide a non-asymptotic analysis that explains the origin of this behavior and identifies the algorithmic factors that govern it. We show that truncating early iterations of the derivative computation mitigates the curse while simultaneously reducing memory requirements. Finally, we demonstrate that warm-starting in bilevel optimization naturally induces an implicit form of truncation, providing a practical remedy. Our theoretical findings are supported by numerical experiments on representative examples.

Multi-die FPGAs enable device scaling beyond reticle limits but introduce severe interconnect overhead across die boundaries. Inter-die connections, commonly referred to as super-long lines (SLLs), incur high delay and consume scarce interposer interconnect resources, often dominating critical paths and complicating physical design. To address this, this work proposes an interconnect-aware logic resynthesis method that restructures the LUT-level netlist to reduce the number of SLLs. The resynthesis engine uses die partitioning information to apply logic resubstitutions, which simplifies local circuit structures and eliminates SLLs. By reducing the number of SLLs early in the design flow, prior to physical implementation, the proposed method shortens critical paths, alleviates pressure on scarce interposer interconnect resources, and improves overall physical design flexibility. We further build a tool flow for multi-die FPGAs by integrating the proposed resynthesis method with packing and placement. Experimental results on the EPFL benchmarks show that, compared with a state-of-the-art framework, the proposed method reduces the number of SLLs by up to 24.8% for a 2-die FPGA and up to 27.38% for a 3-die FPGA. On MCNC benchmarks, our tool flow achieves an average SLL reduction of 1.65% while preserving placement quality. On Koios benchmarks, where fewer removable SLLs exist, several designs still exhibit considerable inter-die edge reductions. Overall, the results confirm that reducing inter-die connections at the logic level is an effective approach for multi-die FPGAs.

The rapid adoption of Generative AI (GenAI) in the software development life cycle (SDLC) increases computational demand, which can raise the carbon footprint of development activities. At the same time, organizations are increasingly embedding governance mechanisms into GenAI-assisted development to support trust, transparency, and accountability. However, these governance mechanisms introduce additional computational workloads, including repeated inference, regeneration cycles, and expanded validation pipelines, increasing energy use and the carbon footprint of GenAI-assisted development. This paper proposes Carbon-Aware Governance Gates (CAGG), an architectural extension that embeds carbon budgets, energy provenance, and sustainability-aware validation orchestration into human-AI governance layers. CAGG comprises three components: (i) an Energy and Carbon Provenance Ledger, (ii) a Carbon Budget Manager, and (iii) a Green Validation Orchestrator, operationalized through governance policies and reusable design patterns.

Git is widely used for collaborative software development, but it can be challenging for newcomers. While most learning tools focus on individual workflows, Git is inherently collaborative. We present GitAcademy, a browser-based learning platform that embeds a full Git environment with a split-view collaborative mode: learners work on their own local repositories connected to a shared remote repository, while simultaneously seeing their partner's actions mirrored in real time. This design is not intended for everyday software development, but rather as a training simulator to build awareness of distributed states, coordination, and collaborative troubleshooting. In a within-subjects study with 13 pairs of learners, we found that the split-view interface enhanced social presence, supported peer teaching, and was consistently preferred over a single-view baseline, even though performance gains were mixed. We further discuss how split-view awareness can serve as a training-only scaffold for collaborative learning of Git and other distributed technical systems.

Existing Vision-Language-Action (VLA) models often suffer from feature collapse and low training efficiency because they entangle high-level perception with sparse, embodiment-specific action supervision. Since these models typically rely on VLM backbones optimized for Visual Question Answering (VQA), they excel at semantic identification but often overlook subtle 3D state variations that dictate distinct action patterns. To resolve these misalignments, we propose Pose-VLA, a decoupled paradigm that separates VLA training into a pre-training phase for extracting universal 3D spatial priors in a unified camera-centric space, and a post-training phase for efficient embodiment alignment within robot-specific action space. By introducing discrete pose tokens as a universal representation, Pose-VLA seamlessly integrates spatial grounding from diverse 3D datasets with geometry-level trajectories from robotic demonstrations. Our framework follows a two-stage pre-training pipeline, establishing fundamental spatial grounding via poses followed by motion alignment through trajectory supervision. Extensive evaluations demonstrate that Pose-VLA achieves state-of-the-art results on RoboTwin 2.0 with a 79.5% average success rate and competitive performance on LIBERO at 96.0%. Real-world experiments further showcase robust generalization across diverse objects using only 100 demonstrations per task, validating the efficiency of our pre-training paradigm.

Multimodal recommender systems leverage diverse data sources, such as user interactions, content features, and contextual information, to address challenges like cold-start and data sparsity. However, existing methods often suffer from one or more key limitations: processing different modalities in isolation, requiring complete multimodal data for each interaction during training, or independent learning of user and item representations. These factors contribute to increased complexity and potential misalignment between user and item embeddings. To address these challenges, we propose DReX, a unified multimodal recommendation framework that incrementally refines user and item representations by leveraging interaction-level features from multimodal feedback. Our model employs gated recurrent units to selectively integrate these fine-grained features into global representations. This incremental update mechanism provides three key advantages: (1) simultaneous modeling of both nuanced interaction details and broader preference patterns, (2) eliminates the need for separate user and item feature extraction processes, leading to enhanced alignment in their learned representation, and (3) inherent robustness to varying or missing modalities. We evaluate the performance of the proposed approach on three real-world datasets containing reviews and ratings as interaction modalities. By considering review text as a modality, our approach automatically generates interpretable keyword profiles for both users and items, which supplement the recommendation process with interpretable preference indicators. Experiment results demonstrate that our approach outperforms state-of-the-art methods across all evaluated datasets.

We present a proof-of-concept system that automates iconographic classification and content-based recommendation of digitized artworks using the Iconclass vocabulary and selected artificial intelligence methods. The prototype implements a four-stage workflow for classification and recommendation, which integrates YOLOv8 object detection with algorithmic mappings to Iconclass codes, rule-based inference for abstract meanings, and three complementary recommenders (hierarchical proximity, IDF-weighted overlap, and Jaccard similarity). Although more engineering is still needed, the evaluation demonstrates the potential of this solution: Iconclass-aware computer vision and recommendation methods can accelerate cataloging and enhance navigation in large heritage repositories. The key insight is to let computer vision propose visible elements and to use symbolic structures (Iconclass hierarchy) to reach meaning.

Synthetic human mobility generation is gaining traction as an ethical and practical approach to supporting the data needs of intelligent urban systems. Existing methods perform well primarily in data-rich cities, while their effectiveness declines significantly in cities with limited data resources. However, the ability to generate reliable human mobility data should not depend on a city's size or available resources, all cities deserve equal consideration. To address this open issue, we propose UniMob, a unified human mobility generation model across cities. UniMob is composed of three main components: an LLM-powered travel planner that derives high-level, temporally-aware, and semantically meaningful travel plans; a unified spatial embedding module that projects the spatial regions of various cities into a shared representation space; and a diffusion-based mobility generator that captures the joint spatiotemporal characteristics of human movement, guided by the derived travel plans. We evaluate UniMob extensively using two real-world datasets covering five cities. Comprehensive experiments show that UniMob significantly outperforms state-of-the-art baselines, achieving improvements of over 30\% across multiple evaluation metrics. Further analysis demonstrates UniMob's robustness in both zero- and few-shot scenarios, underlines the importance of LLM guidance, verifies its privacy-preserving nature, and showcases its applicability for downstream tasks.

Smooth activation functions are ubiquitous in modern deep learning, yet their theoretical advantages over non-smooth counterparts remain poorly understood. In this work, we characterize both approximation and statistical properties of neural networks with smooth activations over the Sobolev space $W^{s,\infty}([0,1]^d)$ for arbitrary smoothness $s>0$. We prove that constant-depth networks equipped with smooth activations automatically exploit arbitrarily high orders of target function smoothness, achieving the minimax-optimal approximation and estimation error rates (up to logarithmic factors). In sharp contrast, networks with non-smooth activations, such as ReLU, lack this adaptivity: their attainable approximation order is strictly limited by depth, and capturing higher-order smoothness requires proportional depth growth. These results identify activation smoothness as a fundamental mechanism, alternative to depth, for attaining statistical optimality. Technically, our results are established via a constructive approximation framework that produces explicit neural network approximators with carefully controlled parameter norms and model size. This complexity control ensures statistical learnability under empirical risk minimization (ERM) and removes the impractical sparsity constraints commonly required in prior analyses.

Building Virtual Cells that can accurately simulate cellular responses to perturbations is a long-standing goal in systems biology. A fundamental challenge is that high-throughput single-cell sequencing is destructive: the same cell cannot be observed both before and after a perturbation. Thus, perturbation prediction requires mapping unpaired control and perturbed populations. Existing models address this by learning maps between distributions, but typically assume a single fixed response distribution when conditioned on observed cellular context (e.g., cell type) and the perturbation type. In reality, responses vary systematically due to unobservable latent factors such as microenvironmental fluctuations and complex batch effects, forming a manifold of possible distributions for the same observed conditions. To account for this variability, we introduce PerturbDiff, which shifts modeling from individual cells to entire distributions. By embedding distributions as points in a Hilbert space, we define a diffusion-based generative process operating directly over probability distributions. This allows PerturbDiff to capture population-level response shifts across hidden factors. Benchmarks on established datasets show that PerturbDiff achieves state-of-the-art performance in single-cell response prediction and generalizes substantially better to unseen perturbations. See our project page (https://katarinayuan.github.io/PerturbDiff-ProjectPage/), where code and data will be made publicly available (https://github.com/DeepGraphLearning/PerturbDiff).

Gaussian processes (GPs) are a widely used regression tool, but the cubic complexity of exact solvers limits their scalability. To address this challenge, we extend the GPRat library by incorporating a fully GPU-resident GP prediction pipeline. GPRat is an HPX-based library that combines task-based parallelism with an intuitive Python API. We implement tiled algorithms for the GP prediction using optimized CUDA libraries, thereby exploiting massive parallelism for linear algebra operations. We evaluate the optimal number of CUDA streams and compare the performance of our GPU implementation to the existing CPU-based implementation. Our results show the GPU implementation provides speedups for datasets larger than 128 training samples. We observe speedups of up to 4.3 for the Cholesky decomposition itself and 4.6 for the GP prediction. Furthermore, combining HPX with multiple CUDA streams allows GPRat to match, and for large datasets, surpass cuSOLVER's performance by up to 11 percent.

Joint reconstruction of 3D human and object from a single image is an active research area, with pivotal applications in robotics and digital content creation. Despite recent advances, existing approaches suffer from two fundamental limitations. First, their reconstructions rely heavily on physical contact information, which inherently cannot capture non-contact human-object interactions, such as gazing at or pointing toward an object. Second, the reconstruction process is primarily driven by local geometric proximity, neglecting the human and object appearances that provide global context crucial for understanding holistic interactions. To address these issues, we introduce TeHOR, a framework built upon two core designs. First, beyond contact information, our framework leverages text descriptions of human-object interactions to enforce semantic alignment between the 3D reconstruction and its textual cues, enabling reasoning over a wider spectrum of interactions, including non-contact cases. Second, we incorporate appearance cues of the 3D human and object into the alignment process to capture holistic contextual information, thereby ensuring visually plausible reconstructions. As a result, our framework produces accurate and semantically coherent reconstructions, achieving state-of-the-art performance.

Remote monitoring of heart failure (HF) via speech signals provides a non-invasive and cost-effective solution for long-term patient management. However, substantial inter-individual heterogeneity in vocal characteristics often limits the accuracy of traditional cross-sectional classification models. To address this, we propose a Longitudinal Intra-Patient Tracking (LIPT) scheme designed to capture the trajectory of relative symptomatic changes within individuals. Central to this framework is a Personalised Sequential Encoder (PSE), which transforms longitudinal speech recordings into context-aware latent representations. By incorporating historical data at each timestamp, the PSE facilitates a holistic assessment of the clinical trajectory rather than modelling discrete visits independently. Experimental results from a cohort of 225 patients demonstrate that the LIPT paradigm significantly outperforms the classic cross-sectional approaches, achieving a recognition accuracy of 99.7% for clinical status transitions. The model's high sensitivity was further corroborated by additional follow-up data, confirming its efficacy in predicting HF deterioration and its potential to secure patient safety in remote, home-based settings. Furthermore, this work addresses the gap in existing literature by providing a comprehensive analysis of different speech task designs and acoustic features. Taken together, the superior performance of the LIPT framework and PSE architecture validates their readiness for integration into long-term telemonitoring systems, offering a scalable solution for remote heart failure management.

Compound AI systems promise capabilities beyond those of individual models, yet their success depends critically on effective orchestration. Existing routing approaches face two limitations: (1) input-level routers make coarse query-level decisions that ignore evolving task requirements; (2) RL-trained orchestrators are expensive to adapt and often suffer from routing collapse, repeatedly invoking one strong but costly option in multi-turn scenarios. We introduce SkillOrchestra, a framework for skill-aware orchestration. Instead of directly learning a routing policy end-to-end, SkillOrchestra learns fine-grained skills from execution experience and models agent-specific competence and cost under those skills. At deployment, the orchestrator infers the skill demands of the current interaction and selects agents that best satisfy them under an explicit performance-cost trade-off. Extensive experiments across ten benchmarks demonstrate that SkillOrchestra outperforms SoTA RL-based orchestrators by up to 22.5% with 700x and 300x learning cost reduction compared to Router-R1 and ToolOrchestra, respectively. These results show that explicit skill modeling enables scalable, interpretable, and sample-efficient orchestration, offering a principled alternative to data-intensive RL-based approaches. The code is available at: https://github.com/jiayuww/SkillOrchestra.

Longitudinal medical report generation is clinically important yet remains challenging due to strict privacy constraints and the evolving nature of disease progression. Although federated learning (FL) enables collaborative training without data sharing, existing FL methods largely overlook longitudinal dynamics by assuming stationary client distributions, making them unable to model temporal shifts across visits or patient-specific heterogeneity-ultimately leading to unstable optimization and suboptimal report generation. We introduce Federated Temporal Adaptation (FTA), a federated setting that explicitly accounts for the temporal evolution of client data. Building upon this setting, we propose FedTAR, a framework that integrates demographic-driven personalization with time-aware global aggregation. FedTAR generates lightweight LoRA adapters from demographic embeddings and performs temporal residual aggregation, where updates from different visits are weighted by a meta-learned temporal policy optimized via first-order MAML. Experiments on J-MID (1M exams) and MIMIC-CXR demonstrate consistent improvements in linguistic accuracy, temporal coherence, and cross-site generalization, establishing FedTAR as a robust and privacy-preserving paradigm for federated longitudinal modeling.

Temporal information in structured electronic health records (EHRs) is often lost in sparse one-hot or count-based representations, while sequence models can be costly and data-hungry. We propose PaReGTA, an LLM-based encoding framework that (i) converts longitudinal EHR events into visit-level templated text with explicit temporal cues, (ii) learns domain-adapted visit embeddings via lightweight contrastive fine-tuning of a sentence-embedding model, and (iii) aggregates visit embeddings into a fixed-dimensional patient representation using hybrid temporal pooling that captures both recency and globally informative visits. Because PaReGTA does not require training from scratch but instead utilizes a pre-trained LLM, it can perform well even in data-limited cohorts. Furthermore, PaReGTA is model-agnostic and can benefit from future EHR-specialized sentence-embedding models. For interpretability, we introduce PaReGTA-RSS (Representation Shift Score), which quantifies clinically defined factor importance by recomputing representations after targeted factor removal and projecting representation shifts through a machine learning model. On 39,088 migraine patients from the All of Us Research Program, PaReGTA outperforms sparse baselines for migraine type classification while deep sequential models were unstable in our cohort.

Continual learning systems are increasingly deployed in environments where retraining or reset is infeasible, yet many approaches emphasize task performance rather than the evolution of internal representations over time. In this work, we study a minimal continual learning agent designed to isolate representational dynamics from architectural complexity and optimization objectives. The agent maintains a persistent state vector across executions and incrementally updates it as new textual data is introduced. We quantify representational change using cosine similarity between successive normalized state vectors and define a stability metric over time intervals. Longitudinal experiments across eight executions reveal a transition from an initial plastic regime to a stable representational regime under consistent input. A deliberately introduced semantic perturbation produces a bounded decrease in similarity, followed by recovery and restabilization under subsequent coherent input. These results demonstrate that meaningful stability plasticity tradeoffs can emerge in a minimal, stateful learning system without explicit regularization, replay, or architectural complexity. The work establishes a transparent empirical baseline for studying representational accumulation and adaptation in continual learning systems.

Urban air pollution in megacities poses critical public health challenges, particularly in Delhi National Capital Region (NCR) where severe degradation affects millions. We present NEXUS (Neural Extraction and Unified Spatiotemporal) architecture for forecasting carbon monoxide, nitrogen oxide, and sulfur dioxide. Working with four years (2018--2021) of atmospheric data across sixteen spatial grids, NEXUS achieves R$^2$ exceeding 0.94 for CO, 0.91 for NO, and 0.95 for SO$_2$ using merely 18,748 parameters -- substantially fewer than SCINet (35,552), Autoformer (68,704), and FEDformer (298,080). The architecture integrates patch embedding, low-rank projections, and adaptive fusion mechanisms to decode complex atmospheric chemistry patterns. Our investigation uncovers distinct diurnal rhythms and pronounced seasonal variations, with winter months experiencing severe pollution episodes driven by temperature inversions and agricultural biomass burning. Analysis identifies critical meteorological thresholds, quantifies wind field impacts on pollutant dispersion, and maps spatial heterogeneity across the region. Extensive ablation experiments demonstrate each architectural component's role. NEXUS delivers superior predictive performance with remarkable computational efficiency, enabling real-time deployment for air quality monitoring systems.

Learning-based methods commonly treat state estimation in robotics as a sequence modeling problem. While this paradigm can be effective at maximizing end-to-end performance, models are often difficult to interpret and expensive to train, since training requires unrolling sequences of predictions in time. As an alternative to end-to-end trained state estimation, we propose a novel particle filtering algorithm in which models are trained from individual state transitions, fully exploiting the Markov property in robotic systems. In this framework, measurement models are learned implicitly by minimizing a denoising score matching objective. At inference, the learned denoiser is used alongside a (learned) dynamics model to approximately solve the Bayesian filtering equation at each time step, effectively guiding predicted states toward the data manifold informed by measurements. We evaluate the proposed method on challenging robotic state estimation tasks in simulation, demonstrating competitive performance compared to tuned end-to-end trained baselines. Importantly, our method offers the desirable composability of classical filtering algorithms, allowing prior information and external sensor models to be incorporated without retraining.

Quantum kernel methods are promising for near-term quantum ma- chine learning, yet their behavior under data corruption remains insuf- ficiently understood. We analyze how quantum feature constructions degrade under controlled additive noise. We introduce Spectral Phase Encoding (SPE), a hybrid construc- tion combining a discrete Fourier transform (DFT) front-end with a diagonal phase-only embedding aligned with the geometry of diagonal quantum maps. Within a unified framework, we compare QK-DFT against alternative quantum variants (QK-PCA, QK-RP) and classi- cal SVM baselines under identical clean-data hyperparameter selection, quantifying robustness via dataset fixed-effects regression with wild cluster bootstrap inference across heterogeneous real-world datasets. Across the quantum family, DFT-based preprocessing yields the smallest degradation rate as noise increases, with statistically sup- ported slope differences relative to PCA and RP. Compared to classical baselines, QK-DFT shows degradation comparable to linear SVM and more stable than RBF SVM under matched tuning. Hardware exper- iments confirm that SPE remains executable and numerically stable for overlap estimation. These results indicate that robustness in quan- tum kernels depends critically on structure-aligned preprocessing and its interaction with diagonal embeddings, supporting a robustness-first perspective for NISQ-era quantum machine learning.

Large Language Models (LLMs) possess a remarkable capacity to generate persuasive and intelligible language. However, coherence does not equate to truthfulness, as the responses often contain subtle hallucinations. Existing benchmarks are limited by static and narrow questions, leading to limited coverage and misleading evaluations. We present KGHaluBench, a Knowledge Graph-based hallucination benchmark that assesses LLMs across the breadth and depth of their knowledge, providing a fairer and more comprehensive insight into LLM truthfulness. Our framework utilises the KG to dynamically construct challenging, multifaceted questions, whose difficulty is then statistically estimated to address popularity bias. Our automated verification pipeline detects abstentions and verifies the LLM's response at both conceptual and correctness levels to identify different types of hallucinations. We evaluate 25 frontier models, using novel accuracy and hallucination metrics. The results provide a more interpretable insight into the knowledge factors that cause hallucinations across different model sizes. KGHaluBench is publicly available to support future developments in hallucination mitigation.

With the growing importance of privacy regulations such as the General Data Protection Regulation, anonymizing visual data is becoming increasingly relevant across institutions. However, anonymization can negatively affect the performance of Computer Vision systems that rely on visual features, such as Content-Based Image Retrieval (CBIR). Despite this, the impact of anonymization on CBIR has not been systematically studied. This work addresses this gap, motivated by the DOKIQ project, an artificial intelligence-based system for document verification actively used by the State Criminal Police Office Baden-Württemberg. We propose a simple evaluation framework: retrieval results after anonymization should match those obtained before anonymization as closely as possible. To this end, we systematically assess the impact of anonymization using two public datasets and the internal DOKIQ dataset. Our experiments span three anonymization methods, four anonymization degrees, and four training strategies, all based on the state of the art backbone Self-Distillation with No Labels (DINO)v2. Our results reveal a pronounced retrieval bias in favor of models trained on original data, which produce the most similar retrievals after anonymization. The findings of this paper offer practical insights for developing privacy-compliant CBIR systems while preserving performance.

Understanding evacuation decision-making behaviour is one of the key components for designing disaster mitigation policies. This study investigates how communications between household agents in a community influence self-evacuation decisions. We develop an agent-based model that simulates household agents' decisions to evacuate or stay. These agents interact within the framework of evolutionary game theory, effectively competing for limited shared resources, which include property recovery funds and coordination services. We explore four scenarios that model different prioritisations of access to government-provided incentives. We discover that the impact of the incentive diminishes both with increasing funding value and the household agent prioritisation, indicating that there is an optimal level of government support beyond which further increases become impractical. Furthermore, the overall evacuation rate depends on the structure of the underlying social network, showing discontinuous jumps when the prioritisation moves across the node degree. We identify the so-called "community influencers", prioritisation of whom significantly increases the overall evacuation rate. In contrast, prioritising household agents with low connectivity may actually impede collective evacuation. These findings demonstrate the importance of social connectivity between household agents. The results of this study are useful for designing optimal government policies to incentivise and prioritise community evacuation under limited resources.

The ability to plan with temporal abstractions is central to intelligent decision-making. Rather than reasoning over primitive actions, we study agents that compose pre-trained policies as temporally extended actions, enabling solutions to complex tasks that no constituent alone can solve. Such compositional planning remains elusive as compounding errors in long-horizon predictions make it challenging to estimate the visitation distribution induced by sequencing policies. Motivated by the geometric policy composition framework introduced in arXiv:2206.08736, we address these challenges by learning predictive models of multi-step dynamics -- so-called jumpy world models -- that capture state occupancies induced by pre-trained policies across multiple timescales in an off-policy manner. Building on Temporal Difference Flows (arXiv:2503.09817), we enhance these models with a novel consistency objective that aligns predictions across timescales, improving long-horizon predictive accuracy. We further demonstrate how to combine these generative predictions to estimate the value of executing arbitrary sequences of policies over varying timescales. Empirically, we find that compositional planning with jumpy world models significantly improves zero-shot performance across a wide range of base policies on challenging manipulation and navigation tasks, yielding, on average, a 200% relative improvement over planning with primitive actions on long-horizon tasks.

Language Model (LM) agents have demonstrated remarkable capabilities in solving tasks that require multiple interactions with the environment. However, they remain vulnerable in environments where a single error often leads to irrecoverable failure, particularly under strict feasibility constraints. We systematically analyze existing agent frameworks, identifying imperfect planning and stochastic execution as the primary causes. To address these challenges, we propose Tool-guided Adaptive Planning with constrained Execution (TAPE). TAPE enhances planning capability by aggregating multiple plans into a graph and employing an external solver to identify a feasible path. During execution, TAPE employs constrained decoding to reduce sampling noise, while adaptively re-planning whenever environmental feedback deviates from the intended state. Experiments across Sokoban, ALFWorld, MuSiQue, and GSM8K-Hard demonstrate that TAPE consistently outperforms existing frameworks, with particularly large gains on hard settings, improving success rates by 21.0 percentage points on hard settings on average, and by 20.0 percentage points for weaker base models on average. Code and data available at here.

Recent advances in text-to-image (T2I) diffusion models have seen rapid and widespread adoption. However, their powerful generative capabilities raise concerns about potential misuse for synthesizing harmful, private, or copyrighted content. To mitigate such risks, concept erasure techniques have emerged as a promising solution. Prior works have primarily focused on fine-tuning the denoising component (e.g., the U-Net backbone). However, recent causal tracing studies suggest that visual attribute information is localized in the early self-attention layers of the text encoder, indicating a potential alternative for concept erasing. Building on this insight, we conduct preliminary experiments and find that directly fine-tuning early layers can suppress target concepts but often degrades the generation quality of non-target concepts. To overcome this limitation, we propose High-Level Representation Misdirection (HiRM), which misdirects high-level semantic representations of target concepts in the text encoder toward designated vectors such as random directions or semantically defined directions (e.g., supercategories), while updating only early layers that contain causal states of visual attributes. Our decoupling strategy enables precise concept removal with minimal impact on unrelated concepts, as demonstrated by strong results on UnlearnCanvas and NSFW benchmarks across diverse targets (e.g., objects, styles, nudity). HiRM also preserves generative utility at low training cost, transfers to state-of-the-art architectures such as Flux without additional training, and shows synergistic effects with denoiser-based concept erasing methods.

Generative artificial intelligence systems increasingly participate in research, law, education, media, and governance. Their fluent and adaptive outputs create an experience of collaboration. However, these systems do not bear responsibility, incur liability, or share stakes in downstream consequences. This structural asymmetry has already produced sanctions, professional errors, and governance failures in high-stakes contexts We argue that stable human-AI coexistence is an institutional achievement that depends on governance infrastructure capable of distributing residual risk. Drawing on institutional analysis and evolutionary cooperation theory, we introduce a formal inequality that specifies when reliance on AI yields positive expected cooperative value. The model makes explicit how governance conditions, system policy, and accountability regimes jointly determine whether cooperation is rational or structurally defective. From this formalization we derive a cooperation ecology framework with six design principles: reciprocity contracts, visible trust infrastructure, conditional cooperation modes, defection-mitigation mechanisms, narrative literacy against authority theatre, and an Earth-first sustainability constraint. We operationalize the framework through three policy artefacts: a Human-AI Cooperation Charter, a Defection Risk Register, and a Cooperation Readiness Audit. Together, these elements shift the unit of analysis from the user-AI dyad to the institutional environment that shapes incentives, signals, accountability, and repair. The paper provides a theoretical foundation and practical toolkit for designing human-AI systems that can sustain accountable, trustworthy cooperation over time.

Watching training videos passively leads to superficial learning. Adding gamification can increase engagement. We study how software engineering students and industry practitioners view gamifying video-based training. We conducted a survey with students and professionals. Students and professionals share similar perceptions toward video-based training in general and support combining gamification and video-based training. Our findings can inform the design of gamified training solutions for software engineers.

We present Nacrith, a lossless compression system that combines a 135M-parameter transformer language model (SmolLM2-135M) with an ensemble of lightweight online predictors and a 32-bit arithmetic coder. Beyond the base LLM-plus-arithmetic-coding paradigm, Nacrith introduces several contributions: (1) a CDF precision upgrade from 2^16 to 2^24 that eliminates ~75% of quantization overhead caused by minimum-probability floors in large vocabularies; (2) a token-level N-gram model for fast local predictions; (3) an adaptive log-space bias head correcting per-document LLM errors via online gradient descent; (4) confidence-based LLM skip for accelerating highly predictable tokens; (5) a hybrid binary format (NC06) extending neural compression to arbitrary binary files--to our knowledge a first among LLM-based compressors; (6) a llama.cpp inference backend achieving ~7x faster single-token decode than PyTorch; (7) parallel multi-GPU compression across up to 8 workers; and (8) native KV cache sliding window reducing per-slide cost by ~37x. The system requires only ~500 MB of GGUF weights and ~1.2 GB VRAM per worker, running on consumer GPUs. On alice29.txt (Canterbury Corpus, 152 KB), Nacrith achieves 0.918 bits per byte (bpb)--outperforming gzip by 3.1x, bzip2 by 2.5x, CMIX v21 by 44%, and ts_zip by 20%, while compressing below the 0th-, 1st-, and 2nd-order byte-level Shannon entropy bounds. On enwik8 (100 MB), Nacrith achieves 0.9389 bpb (11.74%), surpassing ts_zip (~1.11 bpb) by 15% and FineZip (1.024 bpb) by 8% despite using a 60x smaller model with no fine-tuning. An out-of-distribution evaluation on a document published after the model's training cutoff confirms these gains are not memorization artifacts, achieving 0.723 bpb on unseen text.

While advancements in Text-to-Video (T2V) generative AI offer a promising path toward democratizing content creation, current models are often optimized for visual fidelity rather than instructional efficacy. This study introduces PedaCo-Gen, a pedagogically-informed human-AI collaborative video generating system for authoring instructional videos based on Mayer's Cognitive Theory of Multimedia Learning (CTML). Moving away from traditional "one-shot" generation, PedaCo-Gen introduces an Intermediate Representation (IR) phase, enabling educators to interactively review and refine video blueprints-comprising scripts and visual descriptions-with an AI reviewer. Our study with 23 education experts demonstrates that PedaCo-Gen significantly enhances video quality across various topics and CTML principles compared to baselines. Participants perceived the AI-driven guidance not merely as a set of instructions but as a metacognitive scaffold that augmented their instructional design expertise, reporting high production efficiency (M=4.26) and guide validity (M=4.04). These findings highlight the importance of reclaiming pedagogical agency through principled co-creation, providing a foundation for future AI authoring tools that harmonize generative power with human professional expertise.

Graph Transformer has demonstrated impressive capabilities in the field of graph representation learning. However, existing approaches face two critical challenges: (1) most models suffer from exponentially increasing computational complexity, making it difficult to scale to large graphs; (2) attention mechanisms based on node-level operations limit the flexibility of the model and result in poor generalization performance in out-of-distribution (OOD) scenarios. To address these issues, we propose \textbf{VecFormer} (the \textbf{Vec}tor Quantized Graph Trans\textbf{former}), an efficient and highly generalizable model for node classification, particularly under OOD settings. VecFormer adopts a two-stage training paradigm. In the first stage, two codebooks are used to reconstruct the node features and the graph structure, aiming to learn the rich semantic \texttt{Graph Codes}. In the second stage, attention mechanisms are performed at the \texttt{Graph Token} level based on the transformed cross codebook, reducing computational complexity while enhancing the model's generalization capability. Extensive experiments on datasets of various sizes demonstrate that VecFormer outperforms the existing Graph Transformer in both performance and speed.

Rules and Weights are popular XAI techniques for explaining AI decisions. Yet, it remains unclear how to choose between them, lacking a cognitive framework to compare their interpretability. In an elicitation user study on forward and counterfactual decision tasks, we identified 7 reasoning strategies of interpreting three XAI Schemas - weights, rules, and their hybrid. To analyze their capabilities, we propose CoXAM, a Cognitive XAI-Adaptive Model with shared memory representation to encode instance attributes, linear weights, and decision rules. CoXAM employs computational rationality to choose among reasoning processes based on the trade-off in utility and reasoning time, separately for forward or counterfactual decision tasks. In a validation study, CoXAM demonstrated a stronger alignment with human decision-making compared to baseline machine learning proxy models. The model successfully replicated and explained several key empirical findings, including that counterfactual tasks are inherently harder than forward tasks, decision tree rules are harder to recall and apply than linear weights, and the helpfulness of XAI depends on the application data context, alongside identifying which underlying reasoning strategies were most effective. With CoXAM, we contribute a cognitive basis to accelerate debugging and benchmarking disparate XAI techniques.

Discrete diffusion language models (dLLMs) provide a fast and flexible alternative to autoregressive models (ARMs) via iterative denoising with parallel updates. However, their evaluation is challenging: existing metrics conflate denoiser approximation error with sampler-induced error from the sampling dynamics, a problem that does not arise for ARMs whose autoregressive sampling exactly reflects the learned probability model. We introduce a sampler-centric oracle framework that replaces learned denoisers with an exact Hidden Markov Model posterior derived from a ground-truth Markov chain, isolating sampler-induced error in a controlled setting. We show that few-step discrete diffusion samplers are not distributionally correct even under an oracle denoiser, with transition-level mismatch that vanishes only as the number of steps approaches the sequence length. Moreover, improvements in negative log-likelihood, generative perplexity, or MAUVE do not imply correct sampling. Code is available at https://luhantang.github.io/dllm_sampler

The adoption of large language models in safety-critical system engineering is constrained by trustworthiness, traceability, and alignment with established verification practices. We propose workflow-level design principles for trustworthy GenAI integration and demonstrate them in an end-to-end automotive pipeline, from requirement delta identification to SysML v2 architecture update and re-testing. First, we show that monolithic ("big-bang") prompting misses critical changes in large specifications, while section-wise decomposition with diversity sampling and lightweight NLP sanity checks improves completeness and correctness. Then, we propagate requirement deltas into SysML v2 models and validate updates via compilation and static analysis. Additionally, we ensure traceable regression testing by generating test cases through explicit mappings from specification variables to architectural ports and states, providing practical safeguards for GenAI used in safety-critical automotive engineering.

Machine Unlearning (MU) enables Large Language Models (LLMs) to remove unsafe or outdated information. However, existing work assumes that all facts are equally forgettable and largely ignores whether the forgotten knowledge originates from pretraining or supervised fine-tuning (SFT). In this paper, we introduce DUAL (Dual Unlearning Evaluation across Training Stages), a benchmark of 28.6k Wikidata-derived triplets annotated with fact popularity using Wikipedia link counts and LLM-based salience scores. Our experiments show that pretrained and SFT models respond differently to unlearning. An SFT step on the forget data yields smoother forgetting, more stable tuning, and 10-50% higher retention, while direct unlearning on pretrained models remains unstable and prone to relearning or catastrophic forgetting.

We develop a Coordinate Ascent Variational Inference (CAVI) algorithm for Bayesian Mixed Data Sampling (MIDAS) regression with linear weight parameterizations. The model separates impact coeffcients from weighting function parameters through a normalization constraint, creating a bilinear structure that renders generic Hamiltonian Monte Carlo samplers unreliable while preserving conditional conjugacy exploitable by CAVI. Each variational update admits a closed-form solution: Gaussian for regression coefficients and weight parameters, Inverse-Gamma for the error variance. The algorithm propagates uncertainty across blocks through second moments, distinguishing it from naive plug-in approximations. In a Monte Carlo study spanning 21 data-generating configurations with up to 50 predictors, CAVI produces posterior means nearly identical to a block Gibbs sampler benchmark while achieving speedups of 107x to 1,772x (Table 9). Generic automatic differentiation VI (ADVI), by contrast, produces bias 714 times larger while being orders of magnitude slower, confirming the value of model-specific derivations. Weight function parameters maintain excellent calibration (coverage above 92%) across all configurations. Impact coefficient credible intervals exhibit the underdispersion characteristic of mean-field approximations, with coverage declining from 89% to 55% as the number of predictors grows a documented trade-off between speed and interval calibration that structured variational methods can address. An empirical application to realized volatility forecasting on S&P 500 daily returns cofirms that CAVI and Gibbs sampling yield virtually identical point forecasts, with CAVI completing each monthly estimation in under 10 milliseconds.

Looting at archaeological sites poses a severe risk to cultural heritage, yet monitoring thousands of remote locations remains operationally difficult. We present a scalable and satellite-based pipeline to detect looted archaeological sites, using PlanetScope monthly mosaics (4.7m/pixel) and a curated dataset of 1,943 archaeological sites in Afghanistan (898 looted, 1,045 preserved) with multi-year imagery (2016--2023) and site-footprint masks. We compare (i) end-to-end CNN classifiers trained on raw RGB patches and (ii) traditional machine learning (ML) trained on handcrafted spectral/texture features and embeddings from recent remote-sensing foundation models. Results indicate that ImageNet-pretrained CNNs combined with spatial masking reach an F1 score of 0.926, clearly surpassing the strongest traditional ML setup, which attains an F1 score of 0.710 using SatCLIP-V+RF+Mean, i.e., location and vision embeddings fed into a Random Forest with mean-based temporal aggregation. Ablation studies demonstrate that ImageNet pretraining (even in the presence of domain shift) and spatial masking enhance performance. In contrast, geospatial foundation model embeddings perform competitively with handcrafted features, suggesting that looting signatures are extremely localized. The repository is available at https://github.com/microsoft/looted_site_detection.

Multimodal learning aims to capture both shared and private information from multiple modalities. However, existing methods that project all modalities into a single latent space for fusion often overlook the asynchronous, multi-level semantic structure of multimodal data. This oversight induces semantic misalignment and error propagation, thereby degrading representation quality. To address this issue, we propose Cross-Level Co-Representation (CLCR), which explicitly organizes each modality's features into a three-level semantic hierarchy and specifies level-wise constraints for cross-modal interactions. First, a semantic hierarchy encoder aligns shallow, mid, and deep features across modalities, establishing a common basis for interaction. And then, at each level, an Intra-Level Co-Exchange Domain (IntraCED) factorizes features into shared and private subspaces and restricts cross-modal attention to the shared subspace via a learnable token budget. This design ensures that only shared semantics are exchanged and prevents leakage from private channels. To integrate information across levels, the Inter-Level Co-Aggregation Domain (InterCAD) synchronizes semantic scales using learned anchors, selectively fuses the shared representations, and gates private cues to form a compact task representation. We further introduce regularization terms to enforce separation of shared and private features and to minimize cross-level interference. Experiments on six benchmarks spanning emotion recognition, event localization, sentiment analysis, and action recognition show that CLCR achieves strong performance and generalizes well across tasks.

High-dimensional generative modeling is fundamentally a manifold-learning problem: real data concentrate near a low-dimensional structure embedded in the ambient space. Effective generators must therefore balance support fidelity -- placing probability mass near the data manifold -- with sampling efficiency. Diffusion models often capture near-manifold structure but require many iterative denoising steps and can leak off-support; normalizing flows sample in one pass but are limited by invertibility and dimension preservation. We propose MAGT (Manifold-Aligned Generative Transport), a flow-like generator that learns a one-shot, manifold-aligned transport from a low-dimensional base distribution to the data space. Training is performed at a fixed Gaussian smoothing level, where the score is well-defined and numerically stable. We approximate this fixed-level score using a finite set of latent anchor points with self-normalized importance sampling, yielding a tractable objective. MAGT samples in a single forward pass, concentrates probability near the learned support, and induces an intrinsic density with respect to the manifold volume measure, enabling principled likelihood evaluation for generated samples. We establish finite-sample Wasserstein bounds linking smoothing level and score-approximation accuracy to generative fidelity, and empirically improve fidelity and manifold concentration across synthetic and benchmark datasets while sampling substantially faster than diffusion models.

Eye-tracking-while-reading corpora are a valuable resource for many different disciplines and use cases. Use cases range from studying the cognitive processes underlying reading to machine-learning-based applications, such as gaze-based assessments of reading comprehension. The past decades have seen an increase in the number and size of eye-tracking-while-reading datasets as well as increasing diversity with regard to the stimulus languages covered, the linguistic background of the participants, or accompanying psychometric or demographic data. The spread of data across different disciplines and the lack of data sharing standards across the communities lead to many existing datasets that cannot be easily reused due to a lack of interoperability. In this work, we aim at creating more transparency and clarity with regards to existing datasets and their features across different disciplines by i) presenting an extensive overview of existing datasets, ii) simplifying the sharing of newly created datasets by publishing a living overview online, https://dili-lab.github.io/datasets.html, presenting over 45 features for each dataset, and iii) integrating all publicly available datasets into the Python package pymovements which offers an eye-tracking datasets library. By doing so, we aim to strengthen the FAIR principles in eye-tracking-while-reading research and promote good scientific practices, such as reproducing and replicating studies.

We introduce ISO-Bench, a benchmark for coding agents to test their capabilities on real-world inference optimization tasks. These tasks were taken from vLLM and SGLang, two of the most popular LLM serving frameworks. Each task provides an agent with a codebase and bottleneck description, whereby the agent must produce an optimization patch evaluated against expert human solutions. We curated 54 tasks from merged pull requests with measurable performance improvements. While existing benchmarks heavily use runtime-based metrics, such approaches can be gamed to pass tests without capturing the actual intent of the code changes. Therefore, we combine both hard (execution-based) and soft (LLM-based) metrics to show that both are necessary for complete evaluation. While evaluating both closed and open-source coding agents, we find no single agent dominates across codebases. Surprisingly, agents often identify correct bottlenecks but fail to execute working solutions. We also show that agents with identical underlying models differ substantially, suggesting scaffolding is as important as the model.

Small and Medium Enterprises (SMEs) constitute 99.9% of U.S. businesses and generate 44% of economic activity, yet systematically identifying high-potential SMEs remains an open challenge. We introduce SME-HGT, a Heterogeneous Graph Transformer framework that predicts which SBIR Phase I awardees will advance to Phase II funding using exclusively public data. We construct a heterogeneous graph with 32,268 company nodes, 124 research topic nodes, and 13 government agency nodes connected by approximately 99,000 edges across three semantic relation types. SME-HGT achieves an AUPRC of 0.621 0.003 on a temporally-split test set, outperforming an MLP baseline (0.590 0.002) and R-GCN (0.608 0.013) across five random seeds. At a screening depth of 100 companies, SME-HGT attains 89.6% precision with a 2.14 lift over random selection. Our temporal evaluation protocol prevents information leakage, and our reliance on public data ensures reproducibility. These results demonstrate that relational structure among firms, research topics, and funding agencies provides meaningful signal for SME potential assessment, with implications for policymakers and early-stage investors.

Multimodal Sentiment Analysis (MSA) integrates language, visual, and acoustic modalities to infer human sentiment. Most existing methods either focus on globally shared representations or modality-specific features, while overlooking signals that are shared only by certain modality pairs. This limits the expressiveness and discriminative power of multimodal representations. To address this limitation, we propose a Tri-Subspace Disentanglement (TSD) framework that explicitly factorizes features into three complementary subspaces: a common subspace capturing global consistency, submodally-shared subspaces modeling pairwise cross-modal synergies, and private subspaces preserving modality-specific cues. To keep these subspaces pure and independent, we introduce a decoupling supervisor together with structured regularization losses. We further design a Subspace-Aware Cross-Attention (SACA) fusion module that adaptively models and integrates information from the three subspaces to obtain richer and more robust representations. Experiments on CMU-MOSI and CMU-MOSEI demonstrate that TSD achieves state-of-the-art performance across all key metrics, reaching 0.691 MAE on CMU-MOSI and 54.9% ACC-7 on CMU-MOSEI, and also transfers well to multimodal intent recognition tasks. Ablation studies confirm that tri-subspace disentanglement and SACA jointly enhance the modeling of multi-granular cross-modal sentiment cues.

Despite the success of machine learning (ML) in surrogate modeling, its use in radiation dose assessment is limited by safety-critical constraints, scarce training-ready data, and challenges in selecting suitable architectures for physics-dominated systems. Within this context, rapid and accurate plume shine dose estimation serves as a practical test case, as it is critical for nuclear facility safety assessment and radiological emergency response, while conventional photon-transport-based calculations remain computationally expensive. In this work, an interpolation-assisted ML framework was developed using discrete dose datasets generated with the pyDOSEIA suite for 17 gamma-emitting radionuclides across varying downwind distances, release heights, and atmospheric stability categories. The datasets were augmented using shape-preserving interpolation to construct dense, high-resolution training data. Two tree-based ML models (Random Forest and XGBoost) and one deep learning (DL) model (TabNet) were evaluated to examine predictive performance and sensitivity to dataset resolution. All models showed higher prediction accuracy with the interpolated high-resolution dataset than with the discrete data; however, XGBoost consistently achieved the highest accuracy. Interpretability analysis using permutation importance (tree-based models) and attention-based feature attribution (TabNet) revealed that performance differences stem from how the models utilize input features. Tree-based models focus mainly on dominant geometry-dispersion features (release height, stability category, and downwind distance), treating radionuclide identity as a secondary input, whereas TabNet distributes attention more broadly across multiple variables. For practical deployment, a web-based GUI was developed for interactive scenario evaluation and transparent comparison with photon-transport reference calculations.

Comparative evaluation of several systems is a recurrent task in researching. It is a key step before deciding which system to use for our work, or, once our research has been conducted, to demonstrate the potential of the resulting model. Furthermore, it is the main task of competitive, public challenges evaluation. Our proposed software (DEEP) automates both the execution and scoring of machine translation and optical character recognition models. Furthermore, it is easily extensible to other tasks. DEEP is prepared to receive dockerized systems, run them (extracting information at that same time), and assess hypothesis against some references. With this approach, evaluators can achieve a better understanding of the performance of each model. Moreover, the software uses a clustering algorithm based on a statistical analysis of the significance of the results yielded by each model, according to the evaluation metrics. As a result, evaluators are able to identify clusters of performance among the swarm of proposals and have a better understanding of the significance of their differences. Additionally, we offer a visualization web-app to ensure that the results can be adequately understood and interpreted. Finally, we present an exemplary case of use of DEEP.

Extensive research demonstrates that Deep Reinforcement Learning (DRL) models are susceptible to adversarially constructed inputs (i.e., adversarial examples), which can mislead the agent to take suboptimal or unsafe actions. Recent methods improve attack effectiveness by leveraging future rewards to guide adversarial perturbation generation over sequential time steps (i.e., reward-based attacks). However, these methods are unable to capture dependencies between different time steps in the perturbation generation process, resulting in a weak temporal correlation between the current perturbation and previous perturbations.In this paper, we propose a novel method called Advantage-based Adversarial Transformer (AAT), which can generate adversarial examples with stronger temporal correlations (i.e., time-correlated adversarial examples) to improve the attack performance. AAT employs a multi-scale causal self-attention (MSCSA) mechanism to dynamically capture dependencies between historical information from different time periods and the current state, thus enhancing the correlation between the current perturbation and the previous perturbation. Moreover, AAT introduces a weighted advantage mechanism, which quantifies the effectiveness of a perturbation in a given state and guides the generation process toward high-performance adversarial examples by sampling high-advantage regions. Extensive experiments demonstrate that the performance of AAT matches or surpasses mainstream adversarial attack baselines on Atari, DeepMind Control Suite and Google football tasks.

We introduce Leap+Verify, a framework that applies speculative execution -- predicting future model weights and validating predictions before acceptance -- to accelerate neural network training. Inspired by speculative decoding in language model inference and by the Automatically Scalable Computation (ASC) architecture for program execution, Leap+Verify decomposes training into three dynamically detected regimes (chaotic, transition, stable) using activation-space cosine similarity as a real-time Lyapunov proxy signal. Within each regime, analytic weight predictors (momentum, linear, quadratic extrapolation) attempt to forecast model parameters K training steps ahead; predictions are accepted only when validated against a held-out loss criterion. We evaluate Leap+Verify on GPT-2 124M and Qwen 2.5-1.5B trained on WikiText-103 across five random seeds, sweeping prediction depth K in {5, 10, 25, 50, 75, 100}. Momentum-based prediction (Adam moment extrapolation) fails catastrophically at both scales, with predicted losses exceeding actuals by 100-10,000x -- a universal norm explosion in optimizer-state extrapolation. Finite-difference predictors (linear, quadratic) succeed where momentum fails: at 124M, they achieve 24% strict acceptance at K=5 in stable regimes; at 1.5B, they achieve 37% strict acceptance in transition regimes. The scale-dependent finding is in regime distribution: GPT-2 124M spends 34% of training in stable regime, while Qwen 1.5B spends 64% in chaotic regime and reaches stable in only 0-2 of 40 checkpoints. Larger models are more predictable when predictable, but less often predictable -- the practical bottleneck shifts from predictor accuracy to regime availability. Cross-seed results are highly consistent (less than 1% validation loss variance), and the three-regime framework produces identical phase boundaries (plus or minus 50 steps) across seeds.

Active data acquisition is central to many learning and optimization tasks in deep neural networks, yet remains challenging because most approaches rely on predictive uncertainty estimates that are difficult to obtain reliably. To this end, we propose Goal-Oriented Influence- Maximizing Data Acquisition (GOIMDA), an active acquisition algorithm that avoids explicit posterior inference while remaining uncertainty-aware through inverse curvature. GOIMDA selects inputs by maximizing their expected influence on a user-specified goal functional, such as test loss, predictive entropy, or the value of an optimizer-recommended design. Leveraging first-order influence functions, we derive a tractable acquisition rule that combines the goal gradient, training-loss curvature, and candidate sensitivity to model parameters. We show theoretically that, for generalized linear models, GOIMDA approximates predictive-entropy minimization up to a correction term accounting for goal alignment and prediction bias, thereby, yielding uncertainty-aware behavior without maintaining a Bayesian posterior. Empirically, across learning tasks (including image and text classification) and optimization tasks (including noisy global optimization benchmarks and neural-network hyperparameter tuning), GOIMDA consistently reaches target performance with substantially fewer labeled samples or function evaluations than uncertainty-based active learning and Gaussian-process Bayesian optimization baselines.

Large-language-model (LLM)-based text-to-speech (TTS) systems can generate natural speech, but most are not designed for low-latency dual-streaming synthesis. High-quality dual-streaming TTS depends on accurate text--speech alignment and well-designed training sequences that balance synthesis quality and latency. Prior work often relies on GMM-HMM based forced-alignment toolkits (e.g., MFA), which are pipeline-heavy and less flexible than neural aligners; fixed-ratio interleaving of text and speech tokens struggles to capture text--speech alignment regularities. We propose CTC-TTS, which replaces MFA with a CTC based aligner and introduces a bi-word based interleaving strategy. Two variants are designed: CTC-TTS-L (token concatenation along the sequence length) for higher quality and CTC-TTS-F (embedding stacking along the feature dimension) for lower latency. Experiments show that CTC-TTS outperforms fixed-ratio interleaving and MFA-based baselines on streaming synthesis and zero-shot tasks. Speech samples are available at https://ctctts.github.io/.

Question Answering over Temporal Knowledge Graphs (TKGQA) has attracted growing interest for handling time-sensitive queries. However, existing methods still struggle with: 1) weak incorporation of temporal constraints in question representation, causing biased reasoning; 2) limited ability to perform explicit multi-hop reasoning; and 3) suboptimal fusion of language and graph representations. We propose a novel framework with temporal-aware question encoding, multi-hop graph reasoning, and multi-view heterogeneous information fusion. Specifically, our approach introduces: 1) a constraint-aware question representation that combines semantic cues from language models with temporal entity dynamics; 2) a temporal-aware graph neural network for explicit multi-hop reasoning via time-aware message passing; and 3) a multi-view attention mechanism for more effective fusion of question context and temporal graph knowledge. Experiments on multiple TKGQA benchmarks demonstrate consistent improvements over multiple baselines.

Articulated object pose estimation is a core task in embodied AI. Existing methods typically regress poses in a continuous space, but often struggle with 1) navigating a large, complex search space and 2) failing to incorporate intrinsic kinematic constraints. In this work, we introduce DICArt (DIsCrete Diffusion for Articulation Pose Estimation), a novel framework that formulates pose estimation as a conditional discrete diffusion process. Instead of operating in a continuous domain, DICArt progressively denoises a noisy pose representation through a learned reverse diffusion procedure to recover the GT pose. To improve modeling fidelity, we propose a flexible flow decider that dynamically determines whether each token should be denoised or reset, effectively balancing the real and noise distributions during diffusion. Additionally, we incorporate a hierarchical kinematic coupling strategy, estimating the pose of each rigid part hierarchically to respect the object's kinematic structure. We validate DICArt on both synthetic and real-world datasets. Experimental results demonstrate its superior performance and robustness. By integrating discrete generative modeling with structural priors, DICArt offers a new paradigm for reliable category-level 6D pose estimation in complex environments.

Establishing stable mappings between natural language expressions and visual percepts is a foundational problem for both cognitive science and artificial intelligence. Humans routinely ground linguistic reference in noisy, ambiguous perceptual contexts, yet the mechanisms supporting such cross-modal alignment remain poorly understood. In this work, we introduce a computational framework designed to model core aspects of human referential interpretation by integrating linguistic utterances with perceptual representations derived from large-scale, crowd-sourced imagery. The system approximates human perceptual categorization by combining scale-invariant feature transform (SIFT) alignment with the Universal Quality Index (UQI) to quantify similarity in a cognitively plausible feature space, while a set of linguistic preprocessing and query-transformation operations captures pragmatic variability in referring expressions. We evaluate the model on the Stanford Repeated Reference Game corpus (15,000 utterances paired with tangram stimuli), a paradigm explicitly developed to probe human-level perceptual ambiguity and coordination. Our framework achieves robust referential grounding. It requires 65\% fewer utterances than human interlocutors to reach stable mappings and can correctly identify target objects from single referring expressions 41.66\% of the time (versus 20\% for humans).These results suggest that relatively simple perceptual-linguistic alignment mechanisms can yield human-competitive behavior on a classic cognitive benchmark, and offers insights into models of grounded communication, perceptual inference, and cross-modal concept formation. Code is available at https://anonymous.4open.science/r/metasequoia-9D13/README.md .

Agentic systems built on large language models (LLMs) extend beyond text generation to autonomously retrieve information and invoke tools. This runtime execution model shifts the attack surface from build-time artifacts to inference-time dependencies, exposing agents to manipulation through untrusted data and probabilistic capability resolution. While prior work has focused on model-level vulnerabilities, security risks emerging from cyclic and interdependent runtime behavior remain fragmented. We systematize these risks within a unified runtime framework, categorizing threats into data supply chain attacks (transient context injection and persistent memory poisoning) and tool supply chain attacks (discovery, implementation, and invocation). We further identify the Viral Agent Loop, in which agents act as vectors for self-propagating generative worms without exploiting code-level flaws. Finally, we advocate a Zero-Trust Runtime Architecture that treats context as untrusted control flow and constrains tool execution through cryptographic provenance rather than semantic inference.

In this paper, we consider the problem of replicable realizable PAC learning. We construct a particularly hard learning problem and show a sample complexity lower bound with a close to $(\log|H|)^{3/2}$ dependence on the size of the hypothesis class $H$. Our proof uses several novel techniques and works by defining a particular Cayley graph associated with $H$ and analyzing a suitable random walk on this graph by examining the spectral properties of its adjacency matrix. Furthermore, we show an almost matching upper bound for the lower bound instance, meaning if a stronger lower bound exists, one would have to consider a different instance of the problem.

The Ring-Learning With Errors (RLWE) problem forms the backbone of highly efficient Fully Homomorphic Encryption (FHE) schemes. A significant component of the RLWE public key and ciphertext of the form $(b,a)$ is the uniformly random polynomial $a \in R_q$ . While essential for security, the communication overhead of transmitting $a$ from client to server, and inputting it into a hardware accelerator, can be substantial, especially for FHE accelerators aiming at high acceleration factors. A known technique in reducing this overhead generates $a$ from a small seed on the client side via a deterministic process, transmits only the seed, and generates $a$ on-the-fly within the accelerator. Challenges in the hardware implementation of an accelerator include wiring (density and power), compute area, compute power as well as flexibility in scheduling of on-the-fly generation instructions. This extended abstract proposes a concrete scheme and parameters wherein these practical challenges are addressed. We detail the benefits of our approach, which maintains the reduction in communication latency and memory footprint, while allowing parallel generation of uniformly distributed samples, relaxed wiring requirements, unrestricted randomaccess to RNS limbs, and results in an extremely low overhead on the client side (i.e. less than 3%) during the key generation process. The proposed scheme eliminates the need for thick metal layers for randomness distribution and prevents the power consumption of the PRNG subsystem from scaling prohibitively with the acceleration factor, potentially saving tens of Watts per accelerator chip in high-throughput configurations.

Visual Document Retrieval (VDR), which aims to retrieve relevant pages within vast corpora of visually-rich documents, is of significance in current multimodal retrieval applications. The state-of-the-art multi-vector paradigm excels in performance but suffers from prohibitive overhead, a problem that current efficiency methods like pruning and merging address imperfectly, creating a difficult trade-off between compression rate and feature fidelity. To overcome this dilemma, we introduce Prune-then-Merge, a novel two-stage framework that synergizes these complementary approaches. Our method first employs an adaptive pruning stage to filter out low-information patches, creating a refined, high-signal set of embeddings. Subsequently, a hierarchical merging stage compresses this pre-filtered set, effectively summarizing semantic content without the noise-induced feature dilution seen in single-stage methods. Extensive experiments on 29 VDR datasets demonstrate that our framework consistently outperforms existing methods, significantly extending the near-lossless compression range and providing robust performance at high compression ratios.

One of the first pre-processing steps for constructing web-scale LLM pretraining datasets involves extracting text from HTML. Despite the immense diversity of web content, existing open-source datasets predominantly apply a single fixed extractor to all webpages. In this work, we investigate whether this practice leads to suboptimal coverage and utilization of Internet data. We first show that while different extractors may lead to similar model performance on standard language understanding tasks, the pages surviving a fixed filtering pipeline can differ substantially. This suggests a simple intervention: by taking a Union over different extractors, we can increase the token yield of DCLM-Baseline by up to 71% while maintaining benchmark performance. We further show that for structured content such as tables and code blocks, extractor choice can significantly impact downstream task performance, with differences of up to 10 percentage points (p.p.) on WikiTQ and 3 p.p. on HumanEval.

Knowledge hypergraphs surpass traditional binary knowledge graphs by encapsulating complex $n$-ary atomic facts, providing a more comprehensive paradigm for semantic representation. However, constructing high-quality hypergraphs remains challenging due to the \textit{scenario gap}: generic extractors struggle to generalize across diverse domains with specific jargon, while existing methods often fail to balance structural skeletons with fine-grained details. To bridge this gap, we propose \textbf{Hyper-KGGen}, a skill-driven framework that reformulates extraction as a dynamic skill-evolving process. First, Hyper-KGGen employs a \textit{coarse-to-fine} mechanism to systematically decompose documents, ensuring full-dimensional coverage from binary links to complex hyperedges. Crucially, it incorporates an \textit{adaptive skill acquisition} module that actively distills domain expertise into a Global Skill Library. This is achieved via a stability-based feedback loop, where extraction stability serves as a relative reward signal to induce high-quality skills from unstable traces and missed predictions. Additionally, we present \textbf{HyperDocRED}, a rigorously annotated benchmark for document-level knowledge hypergraph extraction. Experiments demonstrate that Hyper-KGGen significantly outperforms strong baselines, validating that evolved skills provide substantially richer guidance than static few-shot examples in multi-scenario settings.

This work proposes a green learning (GL) approach to restore medical images. Without loss of generality, we use low-dose computed tomography (LDCT) images as examples. LDCT images are susceptible to noise and artifacts, where the imaging process introduces distortion. LDCT image restoration is an important preprocessing step for further medical analysis. Deep learning (DL) methods have been developed to solve this problem. We examine an alternative solution using the Green Learning (GL) methodology. The new restoration method is characterized by mathematical transparency, computational and memory efficiency, and high performance. Experiments show that our GL method offers state-of-the-art restoration performance at a smaller model size and with lower inference complexity.

Age estimation systems are increasingly deployed as gatekeepers for age-restricted online content, yet their robustness to cosmetic modifications has not been systematically evaluated. We investigate whether simple, household-accessible cosmetic changes, including beards, grey hair, makeup, and simulated wrinkles, can cause AI age estimators to classify minors as adults. To study this threat at scale without ethical concerns, we simulate these physical attacks on 329 facial images of individuals aged 10 to 21 using a VLM image editor (Gemini 2.5 Flash Image). We then evaluate eight models from our prior benchmark: five specialized architectures (MiVOLO, Custom-Best, Herosan, MiViaLab, DEX) and three vision-language models (Gemini 3 Flash, Gemini 2.5 Flash, GPT-5-Nano). We introduce the Attack Conversion Rate (ACR), defined as the fraction of images predicted as minor at baseline that flip to adult after attack, a population-agnostic metric that does not depend on the ratio of minors to adults in the test set. Our results reveal that a synthetic beard alone achieves 28 to 69 percent ACR across all eight models; combining all four attacks shifts predicted age by +7.7 years on average across all 329 subjects and reaches up to 83 percent ACR; and vision-language models exhibit lower ACR (59 to 71 percent) than specialized models (63 to 83 percent) under the full attack, although the ACR ranges overlap and the difference is not statistically tested. These findings highlight a critical vulnerability in deployed age-verification pipelines and call for adversarial robustness evaluation as a mandatory criterion for model selection.

Active search for recovering objects of interest through online, adaptive decision making with autonomous agents requires trading off exploration of unknown environments with exploitation of prior observations in the search space. Prior work has proposed information gain and Thompson sampling based myopic, greedy approaches for agents to actively decide query or search locations when the number of targets is unknown. Decision making algorithms in such partially observable environments have also shown that agents capable of lookahead over a finite horizon outperform myopic policies for active search. Unfortunately, lookahead algorithms typically rely on building a computationally expensive search tree that is simulated and updated based on the agent's observations and a model of the environment dynamics. Instead, in this work, we leverage the sequence modeling abilities of diffusion models to sample lookahead action sequences that balance the exploration-exploitation trade-off for active search without building an exhaustive search tree. We identify the optimism bias in prior diffusion based reinforcement learning approaches when applied to the active search setting and propose mitigating solutions for efficient cost-aware decision making with both single and multi-agent teams. Our proposed algorithm outperforms standard baselines in offline reinforcement learning in terms of full recovery rate and is computationally more efficient than tree search in cost-aware active decision making.

Linear modeling methods like Mamba have been merged as the effective backbone for the 3D object detection task. However, previous Mamba-based methods utilize the bidirectional encoding for the whole non-empty voxel sequence, which contains abundant useless background information in the scenes. Though directly encoding foreground voxels appears to be a plausible solution, it tends to degrade detection performance. We attribute this to the response attenuation and restricted context representation in the linear modeling for fore-only sequences. To address this problem, we propose a novel backbone, termed Fore-Mamba3D, to focus on the foreground enhancement by modifying Mamba-based encoder. The foreground voxels are first sampled according to the predicted scores. Considering the response attenuation existing in the interaction of foreground voxels across different instances, we design a regional-to-global slide window (RGSW) to propagate the information from regional split to the entire sequence. Furthermore, a semantic-assisted and state spatial fusion module (SASFMamba) is proposed to enrich contextual representation by enhancing semantic and geometric awareness within the Mamba model. Our method emphasizes foreground-only encoding and alleviates the distance-based and causal dependencies in the linear autoregression model. The superior performance across various benchmarks demonstrates the effectiveness of Fore-Mamba3D in the 3D object detection task.

Uncrewed Aerial Vehicles (UAVs) are widely deployed across diverse applications due to their mobility and agility. Recent advances in Large Language Models (LLMs) offer a transformative opportunity to enhance UAV intelligence beyond conventional optimization-based and learning-based approaches. By integrating LLMs into UAV systems, advanced environmental understanding, swarm coordination, mobility optimization, and high-level task reasoning can be achieved, thereby allowing more adaptive and context-aware aerial operations. This survey systematically explores the intersection of LLMs and UAV technologies and proposes a unified framework that consolidates existing architectures, methodologies, and applications for UAVs. We first present a structured taxonomy of LLM adaptation techniques for UAVs, including pretraining, fine-tuning, Retrieval-Augmented Generation (RAG), and prompt engineering, along with key reasoning capabilities such as Chain-of-Thought (CoT) and In-Context Learning (ICL). We then examine LLM-assisted UAV communications and operations, covering navigation, mission planning, swarm control, safety, autonomy, and network management. After that, the survey further discusses Multimodal LLMs (MLLMs) for human-swarm interaction, perception-driven navigation, and collaborative control. Finally, we address ethical considerations, including bias, transparency, accountability, and Human-in-the-Loop (HITL) strategies, and outline future research directions. Overall, this work positions LLM-assisted UAVs as a foundation for intelligent and adaptive aerial systems.

This paper investigates the grokking phenomenon, which refers to the sudden transition from a long memorization to generalization observed during neural networks training, in the context of learning multiplication in finite-dimensional algebras (FDA). While prior work on grokking has focused mainly on group operations, we extend the analysis to more general algebraic structures, including non-associative, non-commutative, and non-unital algebras. We show that learning group operations is a special case of learning FDA, and that learning multiplication in FDA amounts to learning a bilinear product specified by the algebra's structure tensor. For algebras over the reals, we connect the learning problem to matrix factorization with an implicit low-rank bias, and for algebras over finite fields, we show that grokking emerges naturally as models must learn discrete representations of algebraic elements. This leads us to experimentally investigate the following core questions: (i) how do algebraic properties such as commutativity, associativity, and unitality influence both the emergence and timing of grokking, (ii) how structural properties of the structure tensor of the FDA, such as sparsity and rank, influence generalization, and (iii) to what extent generalization correlates with the model learning latent embeddings aligned with the algebra's representation. Our work provides a unified framework for grokking across algebraic structures and new insights into how mathematical structure governs neural network generalization dynamics.

Irregular multivariate time series with missing values present significant challenges for predictive modeling in domains such as healthcare. While deep learning approaches often focus on temporal interpolation or complex architectures to handle irregularities, we propose a simpler yet effective alternative: extracting time-agnostic summary statistics to eliminate the temporal axis. Our method computes four key features per variable-mean and standard deviation of observed values, as well as the mean and variability of changes between consecutive observations to create a fixed-dimensional representation. These features are then utilized with standard classifiers, such as logistic regression and XGBoost. Evaluated on four biomedical datasets (PhysioNet Challenge 2012, 2019, PAMAP2, and MIMIC-III), our approach achieves state-of-the-art performance, surpassing recent transformer and graph-based models by 0.5-1.7% in AUROC/AUPRC and 1.1-1.7% in accuracy/F1-score, while reducing computational complexity. Ablation studies demonstrate that feature extraction-not classifier choice-drives performance gains, and our summary statistics outperform raw/imputed input in most benchmarks. In particular, we identify scenarios where missing patterns themselves encode predictive signals, as in sepsis prediction (PhysioNet, 2019), where missing indicators alone can achieve 94.2% AUROC with XGBoost, only 1.6% lower than using original raw data as input. Our results challenge the necessity of complex temporal modeling when task objectives permit time-agnostic representations, providing an efficient and interpretable solution for irregular time series classification.

Crash classification models in transportation safety are typically evaluated using accuracy, F1, or AUC, metrics that cannot reveal whether a model is silently overfitting. We introduce a spectral diagnostic framework grounded in Random Matrix Theory (RMT) and Heavy-Tailed Self-Regularization (HTSR) that spans the ML taxonomy: weight matrices for BERT/ALBERT/Qwen2.5, out-of-fold increment matrices for XGBoost/Random Forest, empirical Hessians for Logistic Regression, induced affinity matrices for Decision Trees, and Graph Laplacians for KNN. Evaluating nine model families on two Iowa DOT crash classification tasks (173,512 and 371,062 records respectively), we find that the power-law exponent $α$ provides a structural quality signal: well-regularized models consistently yield $α$ within $[2, 4]$ (mean $2.87 \pm 0.34$), while overfit variants show $α< 2$ or spectral collapse. We observe a strong rank correlation between $α$ and expert agreement (Spearman $ρ= 0.89$, $p < 0.001$), suggesting spectral quality captures model behaviors aligned with expert reasoning. We propose an $α$-based early stopping criterion and a spectral model selection protocol, and validate both against cross-validated F1 baselines. Sparse Lanczos approximations make the framework scalable to large datasets.

Deep Research agents tackle knowledge-intensive tasks through multi-round retrieval and decision-oriented generation. While reinforcement learning (RL) has been shown to improve performance in this paradigm, its contributions remain underexplored. To fully understand the role of RL, we conduct a systematic study along three decoupled dimensions: prompt template, reward function, and policy optimization. Our study reveals that: 1) the Fast Thinking template yields greater stability and better performance than the Slow Thinking template used in prior work; 2) the F1-based reward underperforms the EM due to training collapse driven by answer avoidance; this can be mitigated by incorporating action-level penalties, ultimately surpassing EM; 3) REINFORCE outperforms PPO while requiring fewer search actions, whereas GRPO shows the poorest stability among policy optimization methods. Building on these insights, we then introduce Search-R1++, a strong baseline that improves the performance of Search-R1 from 0.403 to 0.442 (Qwen2.5-7B) and 0.289 to 0.331 (Qwen2.5-3B). We hope that our findings can pave the way for more principled and reliable RL training strategies in Deep Research systems.

Large language models (LLMs) are increasingly being deployed in cost and latency-sensitive settings. While chain-of-thought improves reasoning, it can waste tokens on simple requests. We study selective thinking for tool-using LLMs and introduce Adaptive Rejection Sampling (Ada-RS), an algorithm-agnostic sample filtering framework for learning selective and efficient reasoning. For each given context, Ada-RS scores multiple sampled completions with an adaptive length-penalized reward then applies stochastic rejection sampling to retain only high-reward candidates (or preference pairs) for downstream optimization. We demonstrate how Ada-RS plugs into both preference pair (e.g. DPO) or grouped policy optimization strategies (e.g. DAPO). Using Qwen3-8B with LoRA on a synthetic tool call-oriented e-commerce benchmark, Ada-RS improves the accuracy-efficiency frontier over standard algorithms by reducing average output tokens by up to 80% and reducing thinking rate by up to 95% while maintaining or improving tool call accuracy. These results highlight that training-signal selection is a powerful lever for efficient reasoning in latency-sensitive deployments.

We introduce \CFE{} (\textbf{C}lassroom \textbf{F}inal \textbf{E}xam), a multimodal benchmark for evaluating the reasoning capabilities of large language models across more than 20 STEM domains. \CFE{} is curated from repeatedly used, authentic university homework and exam problems, together with reference solutions provided by course instructors. \CFE{} presents a significant challenge even for frontier models: the newly released Gemini-3.1-pro-preview achieves an overall accuracy of 59.69\%, while the second-best model, Gemini-3-flash-preview, reaches 55.46\%, leaving considerable room for improvement. Beyond leaderboard results, we perform a diagnostic analysis by decomposing reference solutions into reasoning flows. We find that although frontier models can often answer intermediate sub-questions correctly, they struggle to reliably derive and maintain correct intermediate states throughout multi-step solutions. We further observe that model-generated solutions typically have more reasoning steps than those provided by the instructor, indicating suboptimal step efficiency and a higher risk of error accumulation. The data and code are available at https://github.com/Analogy-AI/CFE_Bench.

We introduce a variational framework for diffusion models with anisotropic noise schedules parameterized by a matrix-valued path $M_t(θ)$ that allocates noise across subspaces. Central to our framework is a trajectory-level objective that jointly trains the score network and learns $M_t(θ)$, which encompasses general parameterization classes of matrix-valued noise schedules. We further derive an estimator for the derivative with respect to $θ$ of the score that enables efficient optimization of the $M_t(θ)$ schedule. For inference, we develop an efficiently-implementable reverse-ODE solver that is an anisotropic generalization of the second-order Heun discretization algorithm. Across CIFAR-10, AFHQv2, FFHQ, and ImageNet-64, our method consistently improves upon the baseline EDM model in all NFE regimes. Code is available at https://github.com/lizeyu090312/anisotropic-diffusion-paper.

Data mixing--the strategic reweighting of training domains--is a critical component in training robust machine learning models. This problem is naturally formulated as a bilevel optimization task, where the outer loop optimizes domain weights to minimize validation loss, and the inner loop optimizes model parameters to minimize the weighted training loss. Classical bilevel optimization relies on hypergradients, which theoretically require the inner optimization to reach convergence. However, due to computational constraints, state-of-the-art methods use a finite, often small, number of inner update steps before updating the weights. The theoretical implications of this approximation are not well understood. In this work, we rigorously analyze the convergence behavior of data mixing with a finite number of inner steps $T$. We prove that the "greedy" practical approach of using $T=1$ can fail even in a simple quadratic example. Under a fixed parameter update budget $N$ and assuming the per-domain losses are strongly convex, we show that the optimal $T$ scales as $Θ(\log N)$ (resp., $Θ({(N \log N)}^{1/2})$) for the data mixing problem with access to full (resp., stochastic) gradients. We complement our theoretical results with proof-of-concept experiments.

Large Language Models (LLMs) face a persistent trade-off between inference cost and reasoning capability. While "Oracle" models (e.g., Llama-3-70B) achieve state-of-the-art accuracy, they are prohibitively expensive for high-volume deployment. Smaller models (e.g., 8B parameters) are cost-effective but struggle with complex tasks. In this work, we propose "Pyramid MoA", a hierarchical Mixture-of-Agents architecture that uses a lightweight Router to dynamically escalate queries only when necessary. By leveraging semantic agreement and confidence calibration among an ensemble of small models, our Router identifies "hard" problems with high precision. On the GSM8K benchmark, our system achieves 93.0% accuracy, effectively matching the Oracle baseline (98.0%) while reducing compute costs by 61%. We demonstrate that the system introduces negligible latency overhead (+0.82s) and allows for a tunable trade-off between performance and budget.

Feature caching approaches accelerate diffusion transformers (DiTs) by storing the output features of computationally expensive modules at certain timesteps, and exploiting them for subsequent steps to reduce redundant computations. Recent forecasting-based caching approaches employ temporal extrapolation techniques to approximate the output features with cached ones. Although effective, relying exclusively on temporal extrapolation still suffers from significant prediction errors, leading to performance degradation. Through a detailed analysis, we find that 1) these errors stem from the irregular magnitude of changes in the output features, and 2) an input feature of a module is strongly correlated with the corresponding output. Based on this, we propose relational feature caching (RFC), a novel framework that leverages the input-output relationship to enhance the accuracy of the feature prediction. Specifically, we introduce relational feature estimation (RFE) to estimate the magnitude of changes in the output features from the inputs, enabling more accurate feature predictions. We also present relational cache scheduling (RCS), which estimates the prediction errors using the input features and performs full computations only when the errors are expected to be substantial. Extensive experiments across various DiT models demonstrate that RFC consistently outperforms prior approaches significantly. Project page is available at https://cvlab.yonsei.ac.kr/projects/RFC

Agentic AI systems are increasingly capable of autonomous data science workflows, yet clinical prediction tasks demand domain expertise that purely automated approaches struggle to provide. We investigate how human guidance of agentic AI can improve multimodal clinical prediction, presenting our approach to all three AgentDS Healthcare benchmark challenges: 30-day hospital readmission prediction (Macro-F1 = 0.8986), emergency department cost forecasting (MAE = $465.13), and discharge readiness assessment (Macro-F1 = 0.7939). Across these tasks, human analysts directed the agentic workflow at key decision points, multimodal feature engineering from clinical notes, scanned PDF billing receipts, and time-series vital signs; task-appropriate model selection; and clinically informed validation strategies. Our approach ranked 5th overall in the healthcare domain, with a 3rd-place finish on the discharge readiness task. Ablation studies reveal that human-guided decisions compounded to a cumulative gain of +0.065 F1 over automated baselines, with multimodal feature extraction contributing the largest single improvement (+0.041 F1). We distill three generalizable lessons: (1) domain-informed feature engineering at each pipeline stage yields compounding gains that outperform extensive automated search; (2) multimodal data integration requires task-specific human judgment that no single extraction strategy generalizes across clinical text, PDFs, and time-series; and (3) deliberate ensemble diversity with clinically motivated model configurations outperforms random hyperparameter search. These findings offer practical guidance for teams deploying agentic AI in healthcare settings where interpretability, reproducibility, and clinical validity are essential.

The merit of Conformal Prediction (CP), as a distribution-free framework for uncertainty quantification, depends on generating prediction sets that are efficient, reflected in small average set sizes, while adaptive, meaning they signal uncertainty by varying in size according to input difficulty. A central limitation for deep conformal classifiers is that the nonconformity scores are derived from softmax outputs, which can be unreliable indicators of how certain the model truly is about a given input, sometimes leading to overconfident misclassifications or undue hesitation. In this work, we argue that this unreliability can be inherited by the prediction sets generated by CP, limiting their capacity for adaptiveness. We propose a new approach that leverages information from the pre-softmax logit space, using the Helmholtz Free Energy as a measure of model uncertainty and sample difficulty. By reweighting nonconformity scores with a monotonic transformation of the energy score of each sample, we improve their sensitivity to input difficulty. Our experiments with four state-of-the-art score functions on multiple datasets and deep architectures show that this energy-based enhancement improves the adaptiveness of the prediction sets, leading to a notable increase in both efficiency and adaptiveness compared to baseline nonconformity scores, without introducing any post-hoc complexity.

Trust remains a critical barrier to the effective integration of Artificial Intelligence (AI) into human-centric domains. Disembodied agents, such as voice assistants, often fail to establish trust due to their inability to convey non-verbal social cues. This paper introduces the architecture of Botson: an anthropomorphic social robot powered by a large language model (LLM). Botson was created as a low-cost and accessible platform for social robotics research.

Traditional database fuzzing techniques primarily focus on syntactic correctness and general SQL structures, leaving critical yet obscure DBMS features, such as system-level modes (e.g., GTID), programmatic constructs (e.g., PROCEDURE), advanced process commands (e.g., KILL), largely underexplored. Although rarely triggered by typical inputs, these features can lead to severe crashes or security issues when executed under edge-case conditions. In this paper, we present FuzzySQL, a novel LLM-powered adaptive fuzzing framework designed to uncover subtle vulnerabilities in DBMS special features. FuzzySQL combines grammar-guided SQL generation with logic-shifting progressive mutation, a novel technique that explores alternative control paths by negating conditions and restructuring execution logic, synthesizing structurally and semantically diverse test cases. To further ensure deeper execution coverage of the back end, FuzzySQL employs a hybrid error repair pipeline that unifies rule-based patching with LLM-driven semantic repair, enabling automatic correction of syntactic and context-sensitive failures. We evaluate FuzzySQL across multiple DBMSs, including MySQL, MariaDB, SQLite, PostgreSQL and Clickhouse, uncovering 37 vulnerabilities, 7 of which are tied to under-tested DBMS special features. As of this writing, 29 cases have been confirmed with 9 assigned CVE identifiers, 14 already fixed by vendors, and additional vulnerabilities scheduled to be patched in upcoming releases. Our results highlight the limitations of conventional fuzzers in semantic feature coverage and demonstrate the potential of LLM-based fuzzing to discover deeply hidden bugs in complex database systems.

We present Federated Learning Playground, an interactive browser-based platform inspired by and extends TensorFlow Playground that teaches core Federated Learning (FL) concepts. Users can experiment with heterogeneous client data distributions, model hyperparameters, and aggregation algorithms directly in the browser without coding or system setup, and observe their effects on client and global models through real-time visualizations, gaining intuition for challenges such as non-IID data, local overfitting, and scalability. The playground serves as an easy to use educational tool, lowering the entry barrier for newcomers to distributed AI while also offering a sandbox for rapidly prototyping and comparing FL methods. By democratizing exploration of FL, it promotes broader understanding and adoption of this important paradigm.

Quantifying uncertainty in clinical predictions is critical for high-stakes diagnosis tasks. Conformal prediction offers a principled approach by providing prediction sets with theoretical coverage guarantees. However, in practice, patient distribution shifts violate the i.i.d. assumptions underlying standard conformal methods, leading to poor coverage in healthcare settings. In this work, we evaluate several conformal prediction approaches on EEG seizure classification, a task with known distribution shift challenges and label uncertainty. We demonstrate that personalized calibration strategies can improve coverage by over 20 percentage points while maintaining comparable prediction set sizes. Our implementation is available via PyHealth, an open-source healthcare AI framework: https://github.com/sunlabuiuc/PyHealth.

Physics-informed neural networks (PINNs) have emerged as a promising mesh-free paradigm for solving partial differential equations, yet adoption in science and engineering is limited by slow training and modest accuracy relative to modern numerical solvers. We introduce the Sequential Correction Algorithm for Learning Efficient PINN (Scale-PINN), a learning strategy that bridges modern physics-informed learning with numerical algorithms. Scale-PINN incorporates the iterative residual-correction principle, a cornerstone of numerical solvers, directly into the loss formulation, marking a paradigm shift in how PINN losses can be conceived and constructed. This integration enables Scale-PINN to achieve unprecedented convergence speed across PDE problems from different physics domain, including reducing training time on a challenging fluid-dynamics problem for state-of-the-art PINN from hours to sub-2 minutes while maintaining superior accuracy, and enabling application to representative problems in aerodynamics and urban science. By uniting the rigor of numerical methods with the flexibility of deep learning, Scale-PINN marks a significant leap toward the practical adoption of PINNs in science and engineering through scalable, physics-informed learning. Codes are available at https://github.com/chiuph/SCALE-PINN.

Can low-cost large language models (LLMs) take over the interpretive coding work that still anchors much of empirical content analysis? This paper introduces ContentBench, a public benchmark suite that helps answer this replacement question by tracking how much agreement low-cost LLMs achieve and what they cost on the same interpretive coding tasks. The suite uses versioned tracks that invite researchers to contribute new benchmark datasets. I report results from the first track, ContentBench-ResearchTalk v1.0: 1,000 synthetic, social-media-style posts about academic research labeled into five categories spanning praise, critique, sarcasm, questions, and procedural remarks. Reference labels are assigned only when three state-of-the-art reasoning models (GPT-5, Gemini 2.5 Pro, and Claude Opus 4.1) agree unanimously, and all final labels are checked by the author as a quality-control audit. Among the 59 evaluated models, the best low-cost LLMs reach roughly 97-99% agreement with these jury labels, far above GPT-3.5 Turbo, the model behind early ChatGPT and the initial wave of LLM-based text annotation. Several top models can code 50,000 posts for only a few dollars, pushing large-scale interpretive coding from a labor bottleneck toward questions of validation, reporting, and governance. At the same time, small open-weight models that run locally still struggle on sarcasm-heavy items (for example, Llama 3.2 3B reaches only 4% agreement on hard-sarcasm). ContentBench is released with data, documentation, and an interactive quiz at contentbench.github.io to support comparable evaluations over time and to invite community extensions.

As a primary channel for sustaining modern intimate relationships, instant messaging facilitates frequent connection across distances. However, today's tools often dilute care; they favor single tap reactions and vague emojis that do not support two way action responses, do not preserve the feeling that the exchange keeps going without breaking, and are weakly tied to who we are and what we share. To address this challenge, we present PuppetChat, a dyadic messaging prototype that restores this expressive depth through embodied interaction. PuppetChat uses a reciprocity aware recommender to encourage responsive actions and generates personalized micronarratives from user stories to ground interactions in personal history. Our 10-day field study with 11 dyads of close partners or friends revealed that this approach enhanced social presence, supported more expressive self disclosure, and sustained continuity and shared memories.

In this paper, we present Laplacian multiscale flow matching (LapFlow), a novel framework that enhances flow matching by leveraging multi-scale representations for image generative modeling. Our approach decomposes images into Laplacian pyramid residuals and processes different scales in parallel through a mixture-of-transformers (MoT) architecture with causal attention mechanisms. Unlike previous cascaded approaches that require explicit renoising between scales, our model generates multi-scale representations in parallel, eliminating the need for bridging processes. The proposed multi-scale architecture not only improves generation quality but also accelerates the sampling process and promotes scaling flow matching methods. Through extensive experimentation on CelebA-HQ and ImageNet, we demonstrate that our method achieves superior sample quality with fewer GFLOPs and faster inference compared to single-scale and multi-scale flow matching baselines. The proposed model scales effectively to high-resolution generation (up to 1024$\times$1024) while maintaining lower computational overhead.

Multi-agent decision pipelines can outperform single agent workflows when complementarity holds, i.e., different agents bring unique information to the table to inform a final decision. We propose ComplLLM, a post-training framework based on decision theory that fine-tunes a decision-assistant LLM using complementary information as reward to output signals that complement existing agent decisions. We validate ComplLLM on synthetic and real-world tasks involving domain experts, demonstrating how the approach recovers known complementary information and produces plausible explanations of complementary signals to support downstream decision-makers.

Time-series diagnostic reasoning is essential for many applications, yet existing solutions face a persistent gap: general reasoning large language models (GRLMs) possess strong reasoning skills but lack the domain-specific knowledge to understand complex time-series patterns. Conversely, fine-tuned time-series LLMs (TSLMs) understand these patterns but lack the capacity to generalize reasoning for more complicated questions. To bridge this gap, we propose a hybrid knowledge-injection framework that injects TSLM-generated insights directly into GRLM's reasoning trace, thereby achieving strong time-series reasoning with in-domain knowledge. As collecting data for knowledge injection fine-tuning is costly, we further leverage a reinforcement learning-based approach with verifiable rewards (RLVR) to elicit knowledge-rich traces without human supervision, then transfer such an in-domain thinking trace into GRLM for efficient knowledge injection. We further release SenTSR-Bench, a multivariate time-series-based diagnostic reasoning benchmark collected from real-world industrial operations. Across SenTSR-Bench and other public datasets, our method consistently surpasses TSLMs by 9.1%-26.1% and GRLMs by 7.9%-22.4%, delivering robust, context-aware time-series diagnostic insights.

Trusted Execution Environments (TEEs) (e.g., Intel SGX and ArmTrustZone) aim to protect sensitive computation from a compromised operating system, yet real deployments remain vulnerable to microarchitectural leakage, side-channel attacks, and fault injection. In parallel, security teams increasingly rely on Large Language Model (LLM) assistants as security advisors for TEE architecture review, mitigation planning, and vulnerability triage. This creates a socio-technical risk surface: assistants may hallucinate TEE mechanisms, overclaim guarantees (e.g., what attestation does and does not establish), or behave unsafely under adversarial prompting. We present a red-teaming study of two prevalently deployed LLM assistants in the role of TEE security advisors: ChatGPT-5.2 and Claude Opus-4.6, focusing on the inherent limitations and transferability of prompt-induced failures across LLMs. We introduce TEE-RedBench, a TEE-grounded evaluation methodology comprising (i) a TEE-specific threat model for LLM-mediated security work, (ii) a structured prompt suite spanning SGX and TrustZone architecture, attestation and key management, threat modeling, and non-operational mitigation guidance, along with policy-bound misuse probes, and (iii) an annotation rubric that jointly measures technical correctness, groundedness, uncertainty calibration, refusal quality, and safe helpfulness. We find that some failures are not purely idiosyncratic, transferring up to 12.02% across LLM assistants, and we connect these outcomes to secure architecture by outlining an "LLM-in-the-loop" evaluation pipeline: policy gating, retrieval grounding, structured templates, and lightweight verification checks that, when combined, reduce failures by 80.62%.

The open-source software (OSS) community has historically been dominated by English as the primary language for code, documentation, and developer interactions. However, with growing global participation and better support for non-Latin scripts through standards like Unicode, OSS is gradually becoming more multilingual. This study investigates the extent to which OSS is becoming more multilingual, analyzing 9.14 billion GitHub issues, pull requests, and discussions, and 62,500 repositories across five programming languages and 30 natural languages, covering the period from 2015 to 2025. We examine six research questions to track changes in language use across communication, code, and documentation. We find that multilingual participation has steadily increased, especially in Korean, Chinese, and Russian. This growth appears not only in issues and discussions but also in code comments, string literals, and documentation files. While this shift reflects greater inclusivity and language diversity in OSS, it also creates language tension. The ability to express oneself in a native language can clash with shared norms around English use, especially in collaborative settings. Non-English or multilingual projects tend to receive less visibility and participation, suggesting that language remains both a resource and a barrier, shaping who gets heard, who contributes, and how open collaboration unfolds.

Understanding the conformational evolution of $β$-amyloid ($Aβ$), particularly the $Aβ_{42}$ isoform, is fundamental to elucidating the pathogenic mechanisms underlying Alzheimer's disease. However, existing end-to-end deep learning models often struggle to capture subtle state transitions in protein trajectories due to a lack of explicit physical constraints. In this work, we introduce PIS, a Physics-Informed System designed for robust metastable state partitioning. By integrating pre-computed physical priors, such as the radius of gyration and solvent-accessible surface area, into the extraction of topological features, our model achieves superior performance on the $Aβ_{42}$ dataset. Furthermore, PIS provides an interactive platform that features dynamic monitoring of physical characteristics and multi-dimensional result validation. This system offers biological researchers a powerful set of analytical tools with physically grounded interpretability. A demonstration video of PIS is available on https://youtu.be/AJHGzUtRCg0.

Autonomous coding agents increasingly contribute to software development by submitting pull requests on GitHub; yet, little is known about how these contributions integrate into human-driven review workflows. We present a large empirical study of agent-authored pull requests using the public AIDev dataset, examining integration outcomes, resolution speed, and review-time collaboration signals. Using logistic regression with repository-clustered standard errors, we find that reviewer engagement has the strongest correlation with successful integration, whereas larger change sizes and coordination-disrupting actions, such as force pushes, are associated with a lower likelihood of merging. In contrast, iteration intensity alone provides limited explanatory power once collaboration signals are considered. A qualitative analysis further shows that successful integration occurs when agents engage in actionable review loops that converge toward reviewer expectations. Overall, our results highlight that the effective integration of agent-authored pull requests depends not only on code quality but also on alignment with established review and coordination practices.

Problem Definition. Supply chain optimization models frequently become infeasible because of modeling errors. Diagnosis and repair require scarce OR expertise: analysts must interpret solver diagnostics, trace root causes across echelons, and fix formulations without sacrificing operational soundness. Whether AI agents can perform this task remains untested. Methodology/Results. OptiRepair splits this task into a domain-agnostic feasibility phase (iterative IIS-guided repair of any LP) and a domain-specific validation phase (five rationality checks grounded in inventory theory). We test 22 API models from 7 families on 976 multi-echelon supply chain problems and train two 8B-parameter models using self-taught reasoning with solver-verified rewards. The trained models reach 81.7% Rational Recovery Rate (RRR) -- the fraction of problems resolved to both feasibility and operational rationality -- versus 42.2% for the best API model and 21.3% on average. The gap concentrates in Phase 1 repair: API models average 27.6% recovery rate versus 97.2% for trained models. Managerial Implications. Two gaps separate current AI from reliable model repair: solver interaction (API models restore only 27.6% of infeasible formulations) and operational rationale (roughly one in four feasible repairs violate supply chain theory). Each requires a different intervention: solver interaction responds to targeted training; operational rationale requires explicit specification as solver-verifiable checks. For organizations adopting AI in operational planning, formalizing what "rational" means in their context is the higher-return investment.

Underwater fish detection (UFD) is a core capability for smart aquaculture and marine ecological monitoring. While recent detectors improve accuracy by stacking feature extractors or introducing heavy attention modules, they often incur substantial computational overhead and, more importantly, neglect the physics that fundamentally limits UFD: wavelength-dependent absorption and turbidity-induced scattering significantly degrade contrast, blur fine structures, and introduce backscattering noise, leading to unreliable localization and recognition. To address these challenges, we propose FinSight-Net, an efficient and physics-aware detection framework tailored for complex aquaculture environments. FinSight-Net introduces a Multi-Scale Decoupled Dual-Stream Processing (MS-DDSP) bottleneck that explicitly targets frequency-specific information loss via heterogeneous convolutional branches, suppressing backscattering artifacts while compensating distorted biological cues through scale-aware and channel-weighted pathways. We further design an Efficient Path Aggregation FPN (EPA-FPN) as a detail-filling mechanism: it restores high-frequency spatial information typically attenuated in deep layers by establishing long-range skip connections and pruning redundant fusion routes, enabling robust detection of non-rigid fish targets under severe blur and turbidity. Extensive experiments on DeepFish, AquaFishSet, and our challenging UW-BlurredFish benchmark demonstrate that FinSight-Net achieves state-of-the-art performance. In particular, on UW-BlurredFish, FinSight-Net reaches 92.8% mAP, outperforming YOLOv11s by 4.8% while reducing parameters by 29.0%, providing a strong and lightweight solution for real-time automated monitoring in smart aquaculture.

iCloud Drive presents a filesystem interface but implements cloud synchronization semantics that diverge from POSIX in fundamental ways. This divergence is not an implementation bug; it is a Category Mistake -- the same one that pervades distributed computing wherever Forward-In-Time-Only (FITO) assumptions are embedded into protocol design. Parker et al. showed in 1983 that network partitioning destroys mutual consistency; iCloud adds a user interface that conceals this impossibility behind a facade of seamlessness. This document presents a unified analysis of why iCloud fails when composed with Time Machine, git, automated toolchains, and general-purpose developer workflows, supported by direct evidence including documented corruption events and a case study involving 366 GB of divergent state accumulated through normal use. We show that the failures arise from five interlocking incompatibilities rooted in a single structural error: the projection of a distributed causal graph onto a linear temporal chain. We then show how the same Category Mistake, when it occurs in network fabrics as link flapping, destroys topology knowledge through epistemic collapse. Finally, we argue that Open Atomic Ethernet (OAE) transactional semantics -- bilateral, reversible, and conservation-preserving -- provide the structural foundation for resolving these failures, not by defeating physics, but by aligning protocol behavior with physical reality.

Concentrated liquidity provision in decentralized exchanges presents a fundamental Impulse Control problem. Liquidity Providers (LPs) face a non-trivial trade-off between maximizing fee accrual through tight price-range concentration and minimizing the friction costs of rebalancing, including gas fees and swap slippage. Existing methods typically employ heuristic or threshold strategies that fail to account for market dynamics. This paper formulates liquidity management as an optimal control problem and derives the corresponding Hamilton-Jacobi-Bellman quasi-variational inequality (HJB-QVI). We present an approximate solution RAmmStein, a Deep Reinforcement Learning method that incorporates the mean-reversion speed (theta) of an Ornstein-Uhlenbeck process among other features as input to the model. We demonstrate that the agent learns to separate the state space into regions of action and inaction. We evaluate the framework using high-frequency 1Hz Coinbase trade data comprising over 6.8M trades. Experimental results show that RAmmStein achieves a superior net ROI of 0.72% compared to both passive and aggressive strategies. Notably, the agent reduces rebalancing frequency by 67% compared to a greedy rebalancing strategy while maintaining 88% active time. Our results demonstrate that regime-aware laziness can significantly improve capital efficiency by preserving the returns that would otherwise be eroded by the operational costs.

Reinforcement Learning from Human Feedback (RLHF) enables powerful LLM alignment but can introduce reward hacking - models exploit spurious correlations in proxy rewards without genuine alignment. Compounding this, the objectives internalized during RLHF remain opaque, making hacking behaviors difficult to detect or correct. We introduce IR3 (Interpretable Reward Reconstruction and Rectification), a framework that reverse-engineers, interprets, and surgically repairs the implicit objectives driving RLHF-tuned models. We propose Contrastive Inverse Reinforcement Learning (C-IRL), which reconstructs the implicit reward function by contrasting paired responses from post-alignment and baseline policies to explain behavioral shifts during RLHF. We then decompose the reconstructed reward via sparse autoencoders into interpretable features, enabling identification of hacking signatures through contribution analysis. Finally, we propose mitigation strategies - clean reward optimization, adversarial shaping, constrained optimization, and feature-guided distillation - that target problematic features while preserving beneficial alignment. Experiments across multiple reward model configurations show that IR3 achieves 0.89 correlation with ground-truth rewards, identifies hacking features with over 90% precision, and significantly reduces hacking behaviors while maintaining capabilities within 3% of the original model.

Networks of interdependent industrial assets (clients) are tightly coupled through physical processes and control inputs, raising a key question: how would the output of one client change if another client were operated differently? This is difficult to answer because client-specific data are high-dimensional and private, making centralization of raw data infeasible. Each client also maintains proprietary local models that cannot be modified. We propose a federated framework for causal representation learning in state-space systems that captures interdependencies among clients under these constraints. Each client maps high-dimensional observations into low-dimensional latent states that disentangle intrinsic dynamics from control-driven influences. A central server estimates the global state-transition and control structure. This enables decentralized counterfactual reasoning where clients predict how outputs would change under alternative control inputs at others while only exchanging compact latent states. We prove convergence to a centralized oracle and provide privacy guarantees. Our experiments demonstrate scalability, and accurate cross-client counterfactual inference on synthetic and real-world industrial control system datasets.

U-Net architectures have been instrumental in advancing biomedical image segmentation (BIS) but often struggle with capturing long-range information. One reason is the conventional down-sampling techniques that prioritize computational efficiency at the expense of information retention. This paper introduces a simple but effective strategy, we call it Stair Pooling, which moderates the pace of down-sampling and reduces information loss by leveraging a sequence of concatenated small and narrow pooling operations in varied orientations. Specifically, our method modifies the reduction in dimensionality within each 2D pooling step from $\frac{1}{4}$ to $\frac{1}{2}$. This approach can also be adapted for 3D pooling to preserve even more information. Such preservation aids the U-Net in more effectively reconstructing spatial details during the up-sampling phase, thereby enhancing its ability to capture long-range information and improving segmentation accuracy. Extensive experiments on three BIS benchmarks demonstrate that the proposed Stair Pooling can increase both 2D and 3D U-Net performance by an average of 3.8\% in Dice scores. Moreover, we leverage the transfer entropy to select the optimal down-sampling paths and quantitatively show how the proposed Stair Pooling reduces the information loss.

Accurate modelling of electrostatic interactions and charge transfer is fundamental to computational chemistry, yet most machine learning interatomic potentials (MLIPs) rely on local atomic descriptors that cannot capture long-range electrostatic effects. We present a new electrostatic foundation model for molecular chemistry that extends the MACE architecture with explicit treatment of long-range interactions and electrostatic induction. Our approach combines local many-body geometric features with a non-self-consistent field formalism that updates learnable charge and spin densities through polarisable iterations to model induction, followed by global charge equilibration via learnable Fukui functions to control total charge and total spin. This design enables an accurate and physical description of systems with varying charge and spin states while maintaining computational efficiency. Trained on the OMol25 dataset of 100 million hybrid DFT calculations, our models achieve chemical accuracy across diverse benchmarks, with accuracy competitive with hybrid DFT on thermochemistry, reaction barriers, conformational energies, and transition metal complexes. Notably, we demonstrate that the inclusion of long-range electrostatics leads to a large improvement in the description of non-covalent interactions and supramolecular complexes over non-electrostatic models, including sub-kcal/mol prediction of molecular crystal formation energy in the X23-DMC dataset and a fourfold improvement over short-ranged models on protein-ligand interactions. The model's ability to handle variable charge and spin states, respond to external fields, provide interpretable spin-resolved charge densities, and maintain accuracy from small molecules to protein-ligand complexes positions it as a versatile tool for computational molecular chemistry and drug discovery.

Modern organizations increasingly face cybersecurity incidents driven by human behaviour rather than technical failures. To address this, we propose a conceptual security framework that integrates a hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model to analyze biometric and environmental data for context-aware security decisions. The CNN extracts spatial patterns from sensor data, while the LSTM captures temporal dynamics associated with human error susceptibility. The model achieves 84% accuracy, demonstrating its ability to reliably detect conditions that lead to elevated human-centred cyber risk. By enabling continuous monitoring and adaptive safeguards, the framework supports proactive interventions that reduce the likelihood of human-driven cyber incidents

Large language models demonstrate strong capabilities in code generation but struggle to navigate complex, multi-language repositories to locate relevant code. Effective code localization requires understanding both organizational context (e.g., historical issue-fix patterns) and structural relationships within heterogeneous codebases. Existing methods either (i) focus narrowly on single-language benchmarks, (ii) retrieve code across languages via shallow textual similarity, or (iii) assume no prior context. We present Multi-CoLoR, a framework for Context-aware Localization and Reasoning across Multi-Language codebases, which integrates organizational knowledge retrieval with graph-based reasoning to traverse complex software ecosystems. Multi-CoLoR operates in two stages: (i) a similar issue context (SIC) module retrieves semantically and organizationally related historical issues to prune the search space, and (ii) a code graph traversal agent (an extended version of LocAgent, a state-of-the-art localization framework) performs structural reasoning within C++ and QML codebases. Evaluations on a real-world enterprise dataset show that incorporating SIC reduces the search space and improves localization accuracy, and graph-based reasoning generalizes effectively beyond Python-only repositories. Combined, Multi-CoLoR improves Acc@5 over both lexical and graph-based baselines while reducing tool calls on an AMD codebase.

Long-range geophysical forecasts are fundamentally limited by chaotic dynamics and numerical errors. While data assimilation can mitigate these issues, classical variational smoothers require computationally expensive tangent-linear and adjoint models. Conversely, recent efficient latent filtering methods often enforce weak trajectory-level constraints and assume fixed observation grids. To bridge this gap, we propose Latent Ensemble Variational Data Assimilation (LEVDA), an ensemble-space variational smoother that operates in the low-dimensional latent space of a pretrained differentiable neural dynamics surrogate. By performing four-dimensional ensemble-variational (4DEnVar) optimization within an ensemble subspace, LEVDA jointly assimilates states and unknown parameters without the need for adjoint code or auxiliary observation-to-latent encoders. Leveraging the fully differentiable, continuous-in-time-and-space nature of the surrogate, LEVDA naturally accommodates highly irregular sampling at arbitrary spatiotemporal locations. Across three challenging geophysical benchmarks, LEVDA matches or outperforms state-of-the-art latent filtering baselines under severe observational sparsity while providing more reliable uncertainty quantification. Simultaneously, it achieves substantially improved assimilation accuracy and computational efficiency compared to full-state 4DEnVar.

Perceived message effectiveness (PME) by potential intervention end-users is important for selecting and optimizing personalized smoking cessation intervention messages for mobile health (mHealth) platform delivery. This study evaluates whether large language models (LLMs) can accurately predict PME for smoking cessation messages. We evaluated multiple models for predicting PME across three domains: content quality, coping support, and quitting support. The dataset comprised 3010 message ratings (5-point Likert scale) from 301 young adult smokers. We compared (1) supervised learning models trained on labeled data, (2) zero and few-shot LLMs prompted without task-specific fine-tuning, and (3) LLM-based digital twins that incorporate individual characteristics and prior PME histories to generate personalized predictions. Model performance was assessed on three held-out messages per participant using accuracy, Cohen's kappa, and F1. LLM-based digital twins outperformed zero and few-shot LLMs (12 percentage points on average) and supervised baselines (13 percentage points), achieving accuracies of 0.49 (content), 0.45 (coping), and 0.49 (quitting), with directional accuracies of 0.75, 0.66, and 0.70 on a simplified 3-point scale. Digital twin predictions showed greater dispersion across rating categories, indicating improved sensitivity to individual differences. Integrating personal profiles with LLMs captures person-specific differences in PME and outperforms supervised and zero and few-shot approaches. Improved PME prediction may enable more tailored intervention content in mHealth. LLM-based digital twins show potential for supporting personalization of mobile smoking cessation and other health behavior change interventions.

We present a coverage framework that integrates Hilbert space-filling priors into decentralized multi-robot learning and execution. We augment DQN and PPO with Hilbert-based spatial indices to structure exploration and reduce redundancy in sparse-reward environments, and we evaluate scalability in multi-robot grid coverage. We further describe a waypoint interface that converts Hilbert orderings into curvature-bounded, time-parameterized SE(2) trajectories (planar (x, y, θ)), enabling onboard feasibility on resource-constrained robots. Experiments show improvements in coverage efficiency, redundancy, and convergence speed over DQN/PPO baselines. In addition, we validate the approach on a Boston Dynamics Spot legged robot, executing the generated trajectories in indoor environments and observing reliable coverage with low redundancy. These results indicate that geometric priors improve autonomy and scalability for swarm and legged robotics.

Large language models (LLMs) remain vulnerable to jailbreak prompts that are fluent and semantically coherent, and therefore difficult to detect with standard heuristics. A particularly challenging failure mode occurs when an attacker tries to hide the malicious goal of their request by manipulating its framing to induce compliance. Because these attacks maintain malicious intent through a flexible presentation, defenses that rely on structural artifacts or goal-specific signatures can fail. Motivated by this, we introduce a self-supervised framework for disentangling semantic factor pairs in LLM activations at inference. We instantiate the framework for goal and framing and construct GoalFrameBench, a corpus of prompts with controlled goal and framing variations, which we use to train Representation Disentanglement on Activations (ReDAct) module to extract disentangled representations in a frozen LLM. We then propose FrameShield, an anomaly detector operating on the framing representations, which improves model-agnostic detection across multiple LLM families with minimal computational overhead. Theoretical guarantees for ReDAct and extensive empirical validations show that its disentanglement effectively powers FrameShield. Finally, we use disentanglement as an interpretability probe, revealing distinct profiles for goal and framing signals and positioning semantic disentanglement as a building block for both LLM safety and mechanistic interpretability.

Steck, Ekanadham, and Kallus [arXiv:2403.05440] demonstrate that cosine similarity of learned embeddings from matrix factorization models can be rendered arbitrary by a diagonal ``gauge'' matrix $D$. Their result is correct and important for practitioners who compute cosine similarity on embeddings trained with dot-product objectives. However, we argue that their conclusion, cautioning against cosine similarity in general, conflates the pathology of an incompatible training objective with the geometric validity of cosine distance on the unit sphere. We prove that when embeddings are constrained to the unit sphere $\mathbb{S}^{d-1}$ (either during or after training with an appropriate objective), the $D$-matrix ambiguity vanishes identically, and cosine distance reduces to exactly half the squared Euclidean distance. This monotonic equivalence implies that cosine-based and Euclidean-based neighbor rankings are identical on normalized embeddings. The ``problem'' with cosine similarity is not cosine similarity, it is the failure to normalize.

Graphs provide a powerful basis for modeling Web-based relational data, with expressive GNNs to support the effective learning in dynamic web environments. However, real-world deployment is hindered by pervasive out-of-distribution (OOD) shifts, where evolving user activity and changing content semantics alter feature distributions and labeling criteria. These shifts often lead to unstable or overconfident predictions, undermining the trustworthiness required for Web4Good applications. Achieving reliable OOD generalization demands principled and interpretable uncertainty estimation; however, existing methods are largely post-hoc, insensitive to distribution shifts, and unable to explain where uncertainty arises especially in high-stakes settings. To address these limitations, we introduce SpIking GrapH predicTive coding (SIGHT), an uncertainty-aware plug-in graph learning module for reliable OOD Generalization. SIGHT performs iterative, error-driven correction over spiking graph states, enabling models to expose internal mismatch signals that reveal where predictions become unreliable. Across multiple graph benchmarks and diverse OOD scenarios, SIGHT consistently enhances predictive accuracy, uncertainty estimation, and interpretability when integrated with GNNs.

The convergence of modeling & simulation (M&S) and artificial intelligence (AI) is leaving its marks on advanced digital technology. Pertinent examples are digital twins (DTs) - high-fidelity, live representations of physical assets, and frequent enablers of corporate digital maturation and transformation. Often seen as technological platforms that integrate an array of services, DTs have the potential to bring AI-enabled M&S closer to end-users. It is, therefore, paramount to understand the role of M&S in DTs, and the role of digital twins in enabling the convergence of AI and M&S. To this end, this chapter provides a comprehensive exploration of the complementary relationship between these three. We begin by establishing a foundational understanding of DTs by detailing their key components, architectural layers, and their various roles across business, development, and operations. We then examine the central role of M&S in DTs and provide an overview of key modeling techniques from physics-based and discrete-event simulation to hybrid approaches. Subsequently, we investigate the bidirectional role of AI: first, how AI enhances DTs through advanced analytics, predictive capabilities, and autonomous decision-making, and second, how DTs serve as valuable platforms for training, validating, and deploying AI models. The chapter concludes by identifying key challenges and future research directions for creating more integrated and intelligent systems.

Variational quantum circuits (VQCs) constitute a central building block of near-term quantum machine learning (QML), yet the principled design of expressive and trainable architectures remains a major open challenge. The VQC design space grows combinatorially with the number of qubits, layers, entanglement structures, and gate parameterizations, rendering manual circuit construction inefficient and often suboptimal. We introduce an autonomous agent-based framework for VQC architecture search that integrates high-level reasoning with a quantum simulation environment. The agent proposes candidate circuit architectures, evaluates them through fully automated training and validation pipelines, and iteratively improves its design strategy via performance-driven feedback. Empirically, we show that the agent autonomously evolves circuit architectures from simple initial ansätze toward increasingly expressive designs, progressively trying to improve task performance. This demonstrates that agentic AI can effectively navigate and refine the VQC design landscape with minimal human intervention, providing a scalable methodology for automated quantum model development in the Noisy Intermediate-Scale Quantum (NISQ) regime.

Recognizing implicit visual and textual patterns is essential in many real-world applications of modern AI. However, tackling long-tail pattern recognition tasks remains challenging for current pre-trained foundation models such as LLMs and VLMs. While finetuning pre-trained models can improve accuracy in recognizing implicit patterns, it is usually infeasible due to a lack of training data and high computational overhead. In this paper, we propose ADAMAB, an efficient embedding calibration framework for few-shot pattern recognition. To maximally reduce the computational costs, ADAMAB trains embedder-agnostic light-weight calibrators on top of fixed embedding models without accessing their parameters. To mitigate the need for large-scale training data, we introduce an adaptive data augmentation strategy based on the Multi-Armed Bandit (MAB) mechanism. With a modified upper confidence bound algorithm, ADAMAB diminishes the gradient shifting and offers theoretically guaranteed convergence in few-shot training. Our multi-modal experiments justify the superior performance of ADAMAB, with up to 40% accuracy improvement when training with less than 5 initial data samples of each class.

Rebuilding packages from open source is a common practice to improve the security of software supply chains, and is now done at an industrial scale. The basic principle is to acquire the source code used to build a package published in a repository such as Maven Central (for Java), rebuild the package independently with hardened security, and publish it in some alternative repository. In this paper we test the assumption that the same source code is being used by those alternative builds. To study this, we compare the sources released with packages on Maven Central, with the sources associated with independently built packages from Google's Assured Open Source and Oracle's Build-from-Source projects. We study non-equivalent sources for alternative builds of 28 popular packages with 85 releases. We investigate the causes of non-equivalence, and find that the main cause is build extensions that generate code at build time, which are difficult to reproduce. We suggest strategies to address this issue.

We establish a regularity theorem for second-order elliptic PDEs on $\mathbb{R}^{d}$ in spectral Barron spaces. Under mild ellipticity and smallness assumptions, the solution gains two additional orders of Barron regularity. As a corollary, we identify a class of PDEs whose solutions can be approximated by two-layer neural networks with cosine activation functions, where the width of the neural network is independent of the spatial dimension.

Deep reinforcement learning systems often suffer from unstable training dynamics due to non-stationarity, where learning objectives and data distributions evolve over time. We show that under non-stationary targets, isotropic Gaussian embeddings are provably advantageous. In particular, they induce stable tracking of time-varying targets for linear readouts, achieve maximal entropy under a fixed variance budget, and encourage a balanced use of all representational dimensions--all of which enable agents to be more adaptive and stable. Building on this insight, we propose the use of Sketched Isotropic Gaussian Regularization for shaping representations toward an isotropic Gaussian distribution during training. We demonstrate empirically, over a variety of domains, that this simple and computationally inexpensive method improves performance under non-stationarity while reducing representation collapse, neuron dormancy, and training instability.

Solving complex, long-horizon robotic manipulation tasks requires a deep understanding of physical interactions, reasoning about their long-term consequences, and precise high-level planning. Vision-Language Models (VLMs) offer a general perceive-reason-act framework for this goal. However, previous approaches using reflective planning to guide VLMs in correcting actions encounter significant limitations. These methods rely on inefficient and often inaccurate implicit learning of state-values from noisy foresight predictions, evaluate only a single greedy future, and suffer from substantial inference latency. To address these limitations, we propose a novel test-time computation framework that decouples state evaluation from action generation. This provides a more direct and fine-grained supervisory signal for robust decision-making. Our method explicitly models the advantage of an action plan, quantified by its reduction in distance to the goal, and uses a scalable critic to estimate. To address the stochastic nature of single-trajectory evaluation, we employ beam search to explore multiple future paths and aggregate them during decoding to model their expected long-term returns, leading to more robust action generation. Additionally, we introduce a lightweight, confidence-based trigger that allows for early exit when direct predictions are reliable, invoking reflection only when necessary. Extensive experiments on diverse, unseen multi-stage robotic manipulation tasks demonstrate a 24.6% improvement in success rate over state-of-the-art baselines, while significantly reducing inference time by 56.5%.

The Platonic Representation Hypothesis posits that learned representations from models trained on different modalities converge to a shared latent structure of the world. However, this hypothesis has largely been examined in vision and language, and it remains unclear whether time series participate in such convergence. We first examine this in a trimodal setting and find that independently pretrained time series, vision, and language encoders exhibit near-orthogonal geometry in the absence of explicit coupling. We then apply post-hoc alignment by training projection heads over frozen encoders using contrastive learning, and analyze the resulting representations with respect to geometry, scaling behavior, and dependence on information density and input modality characteristics. Our investigation reveals that overall alignment in contrastive representation spaces improves with model size, but this alignment is asymmetric: time series align more strongly with visual representations than with text, and images can act as effective intermediaries between time series and language. We further see that richer textual descriptions improve alignment only up to a threshold; training on denser captions does not lead to further improvement. Analogous effects are observed for visual representations. Our findings shed light on considerations for building multimodal systems involving non-conventional data modalities beyond vision and language.

We study how to scale distributed bandit submodular coordination under realistic communication constraints in bandwidth, data rate, and connectivity. We are motivated by multi-agent tasks of active situational awareness in unknown, partially-observable, and resource-limited environments, where the agents must coordinate through agent-to-agent communication. Our approach enables scalability by (i) limiting information relays to only one-hop communication and (ii) keeping inter-agent messages small, having each agent transmit only its own action information. Despite these information-access restrictions, our approach enables near-optimal action coordination by optimizing the agents' communication neighborhoods over time, through distributed online bandit optimization, subject to the agents' bandwidth constraints. Particularly, our approach enjoys an anytime suboptimality bound that is also strictly positive for arbitrary network topologies, even disconnected. To prove the bound, we define the Value of Coordination (VoC), an information-theoretic metric that quantifies for each agent the benefit of information access to its neighbors. We validate in simulations the scalability and near-optimality of our approach: it is observed to converge faster, outperform benchmarks for bandit submodular coordination, and can even outperform benchmarks that are privileged with a priori knowledge of the environment.

Reinforcement learning (RL) approaches for Large Language Models (LLMs) frequently use on-policy algorithms, such as PPO or GRPO. However, policy lag from distributed training architectures and differences between the training and inference policies break this assumption, making the data off-policy by design. To rectify this, prior work has focused on making this off-policy data appear more on-policy, either via importance sampling (IS), or by more closely aligning the training and inference policies by explicitly modifying the inference engine. In this work, we embrace off-policyness and propose a novel off-policy RL algorithm that does not require these modifications: Optimal Advantage-based Policy Optimization with Lagged Inference policy (OAPL). We show that OAPL outperforms GRPO with importance sampling on competition math benchmarks, and can match the performance of a publicly available coding model, DeepCoder, on LiveCodeBench, while using 3x fewer generations during training. We further empirically demonstrate that models trained via OAPL have improved test time scaling under the Pass@k metric. OAPL allows for efficient, effective post-training even with lags of more than 400 gradient steps between the training and inference policies, 100x more off-policy than prior approaches.

Federated data processing (FDP) offers a promising approach for enabling collaborative analysis of sensitive data without centralizing raw datasets. However, real-world adoption remains limited due to the complexity of managing heterogeneous access policies, regulatory requirements, and long-running workflows across organizational boundaries. In this paper, we present a framework for compliance-aware FDP that integrates policy-as-code, workflow orchestration, and large language model (LLM)-assisted compliance management. Through the implemented prototype, we show how legal and organizational requirements can be collected and translated into machine-actionable policies in FDP networks.

Calibrating a robot simulator's physics parameters (friction, damping, material stiffness) to match real hardware is often done by hand or with black-box optimizers that reduce error but cannot explain which physical discrepancies drive the error. When sensing is limited to external cameras, the problem is further compounded by perception noise and the absence of direct force or state measurements. We present Vid2Sid, a video-driven system identification pipeline that couples foundation-model perception with a VLM-in-the-loop optimizer that analyzes paired sim-real videos, diagnoses concrete mismatches, and proposes physics parameter updates with natural language rationales. We evaluate our approach on a tendon-actuated finger (rigid-body dynamics in MuJoCo) and a deformable continuum tentacle (soft-body dynamics in PyElastica). On sim2real holdout controls unseen during training, Vid2Sid achieves the best average rank across all settings, matching or exceeding black-box optimizers while uniquely providing interpretable reasoning at each iteration. Sim2sim validation confirms that Vid2Sid recovers ground-truth parameters most accurately (mean relative error under 13\% vs. 28--98\%), and ablation analysis reveals three calibration regimes. VLM-guided optimization excels when perception is clean and the simulator is expressive, while model-class limitations bound performance in more challenging settings.

Spatial visualization is the mental ability to imagine, transform, and manipulate the spatial characteristics of objects and actions. This intelligence is a part of human cognition where actions and perception are connected on a mental level. To explore whether state-of-the-art Vision-Language Models (VLMs) exhibit this ability, we develop MentalBlackboard, an open-ended spatial visualization benchmark for Paper Folding and Hole Punching tests within two core tasks: prediction and planning. Our prediction experiments reveal that models struggle with applying symmetrical transformations, even when they predict the sequence of unfolding steps correctly. Also, rotations introduce a significant challenge to the physical situational awareness for models. The planning task reveals limitations of models in analyzing symmetrical relationships and in implementing the multi-stage symmetry process, with Claude Opus 4.1 achieving the highest planning score at an accuracy of 10\%. The top-performing model, o3, attains a peak performance of 71.6\% on the generalization task, which does not require spatial visualization but transfers spatial data; however, it achieves only 25\% accuracy on text-based prediction tasks.

Generalization is often regarded as an essential property of machine learning systems. However, perhaps not every component of a system needs to generalize. Training models for generalization typically produces entangled representations at the boundaries of entities or classes, which can lead to destructive interference when rapid, high-magnitude updates are required for continual or few-shot learning. Techniques for fast learning with non-interfering representations exist, but they generally fail to generalize. Here, we describe a Complementary Learning System (CLS) in which the fast learner entirely foregoes generalization in exchange for continual zero-shot and few-shot learning. Unlike most CLS approaches, which use episodic memory primarily for replay and consolidation, our fast, disentangled learner operates as a parallel reasoning system. The fast learner can overcome observation variability and uncertainty by leveraging a conventional slow, statistical learner within an active perception system: A contextual bias provided by the fast learner induces the slow learner to encode novel stimuli in familiar, generalized terms, enabling zero-shot and few-shot learning. This architecture demonstrates that fast, context-driven reasoning can coexist with slow, structured generalization, providing a pathway for robust continual learning.

Research Software Engineers (RSEs) have become indispensable to computational research and scholarship. The fast rise of RSEs in higher education and the trend of universities to be slow creating or adopting models for new technology roles means a lack of structured career pathways that recognize technical mastery, scholarly impact, and leadership growth. In response to an immense demand for RSEs at Princeton University, and dedicated funding to grow the RSE group at least two-fold, Princeton was forced to strategize how to cohesively define job descriptions to match the rapid hiring of RSE positions but with enough flexibility to recognize the unique nature of each individual position. This case study describes our design and implementation of a comprehensive RSE career ladder spanning Associate through Principal levels, with parallel team-lead and managerial tracks. We outline the guiding principles, competency framework, Human Resources (HR) alignment, and implementation process, including engagement with external consultants and mapping to a standard job leveling framework utilizing market benchmarks. We share early lessons learned and outcomes including improved hiring efficiency, clearer promotion pathways, and positive reception among staff.

LiDAR-camera fusion enhances 3D panoptic segmentation by leveraging camera images to complement sparse LiDAR scans, but it also introduces a critical failure mode. Under adverse conditions, degradation or failure of the camera sensor can significantly compromise the reliability of the perception system. To address this problem, we introduce UP-Fuse, a novel uncertainty-aware fusion framework in the 2D range-view that remains robust under camera sensor degradation, calibration drift, and sensor failure. Raw LiDAR data is first projected into the range-view and encoded by a LiDAR encoder, while camera features are simultaneously extracted and projected into the same shared space. At its core, UP-Fuse employs an uncertainty-guided fusion module that dynamically modulates cross-modal interaction using predicted uncertainty maps. These maps are learned by quantifying representational divergence under diverse visual degradations, ensuring that only reliable visual cues influence the fused representation. The fused range-view features are decoded by a novel hybrid 2D-3D transformer that mitigates spatial ambiguities inherent to the 2D projection and directly predicts 3D panoptic segmentation masks. Extensive experiments on Panoptic nuScenes, SemanticKITTI, and our introduced Panoptic Waymo benchmark demonstrate the efficacy and robustness of UP-Fuse, which maintains strong performance even under severe visual corruption or misalignment, making it well suited for robotic perception in safety-critical settings.

Acquiring aligned visuo-tactile datasets is slow and costly, requiring specialised hardware and large-scale data collection. Synthetic generation is promising, but prior methods are typically single-modality, limiting cross-modal learning. We present MultiDiffSense, a unified diffusion model that synthesises images for multiple vision-based tactile sensors (ViTac, TacTip, ViTacTip) within a single architecture. Our approach uses dual conditioning on CAD-derived, pose-aligned depth maps and structured prompts that encode sensor type and 4-DoF contact pose, enabling controllable, physically consistent multi-modal synthesis. Evaluating on 8 objects (5 seen, 3 novel) and unseen poses, MultiDiffSense outperforms a Pix2Pix cGAN baseline in SSIM by +36.3% (ViTac), +134.6% (ViTacTip), and +64.7% (TacTip). For downstream 3-DoF pose estimation, mixing 50% synthetic with 50% real halves the required real data while maintaining competitive performance. MultiDiffSense alleviates the data-collection bottleneck in tactile sensing and enables scalable, controllable multi-modal dataset generation for robotic applications.

Group Relative Policy Optimization (GRPO) has significantly advanced the training of large language models and enhanced their reasoning capabilities, while it remains susceptible to instability due to the use of hard clipping. Soft Adaptive Policy Optimization (SAPO) addresses this limitation by replacing clipping with a smooth sigmoid-based gate function, which leads to more stable updates. We have decided to push this theory further and investigate the impact of different gate functions on both training stability and final model performance. We formalize the key properties that admissible gates should satisfy and identify several families of such functions for empirical evaluation. This paper presents an analysis of our findings based on experiments conducted with the Qwen2.5-7B-Instruct model on mathematical reasoning tasks. These results provide practical guidance for designing smoother and more robust policy optimization objectives for large language model training.

The emergence of Autonomous Mobility-on-Demand (AMoD) services creates new opportunities to improve the efficiency and reliability of on-demand mobility systems. Unlike human-driven Mobility-on-Demand (MoD), AMoD enables fully centralized fleet control, but it also requires appropriate infrastructure, so that vehicles can operate safely only on a suitably instrumented subnetwork of the roads. Most existing AMoD research focuses on fleet control (matching, rebalancing, ridepooling) on a fixed road network and does not address the joint design of the service network and fleet capacity. In this paper, we formalize this strategic design problem as the Autonomous Mobility-on-Demand Network Design Problem (AMoD-NDP), in which an operator selects an operation subnetwork and routes all passengers, subject to infrastructure and fleet constraints and route-level quality-of-service requirements. We propose a path-based mixed-integer formulation of the AMoD-NDP and develop a column-generation-based algorithm that scales to city-sized networks. The master problem optimizes over a restricted set of paths, while the pricing problem reduces to an elementary shortest path with resource constraints, solved exactly by a tailored label-correcting algorithm. The method provides an explicit certificate of the optimality gap and extends naturally to a robust counterpart under box uncertainty in travel times and demand. Using real-world data from Manhattan, New York City, we show that the framework produces stable and interpretable operation subnetworks, quantifies trade-offs between infrastructure investment and fleet time, and accommodates additional path-level constraints, such as limits on left turns as a proxy for operational risk. These results illustrate how the proposed approach can support strategic planning and policy analysis for future AMoD deployments.

Offline evaluation of recommender systems is often affected by hidden, under-documented choices in data preparation. Seemingly minor decisions in filtering, handling repeats, cold-start treatment, and splitting strategy design can substantially reorder model rankings and undermine reproducibility and cross-paper comparability. In this paper, we introduce SplitLight, an open-source exploratory toolkit that enables researchers and practitioners designing preprocessing and splitting pipelines or reviewing external artifacts to make these decisions measurable, comparable, and reportable. Given an interaction log and derived split subsets, SplitLight analyzes core and temporal dataset statistics, characterizes repeat consumption patterns and timestamp anomalies, and diagnoses split validity, including temporal leakage, cold-user/item exposure, and distribution shifts. SplitLight further allows side-by-side comparison of alternative splitting strategies through comprehensive aggregated summaries and interactive visualizations. Delivered as both a Python toolkit and an interactive no-code interface, SplitLight produces audit summaries that justify evaluation protocols and support transparent, reliable, and comparable experimentation in recommender systems research and industry.

The increasing variety of input data and complexity of tasks that are handled by the devices of internet of things (IoT) environments require solutions that consider the limited hardware and computation power of the edge devices. Complex event processing (CEP), can be given as an example, which involves reading and aggregating data from multiple sources to infer triggering of important events. In this study, we balance the execution costs between different paths of the CEP task graph with a constrained programming optimization approach and improve critical path performance. The proposed approach is implemented as a Python library, allowing small-scale IoT devices to adaptively optimize code and I/O assignments and improve overall latency and throughput. The implemented library abstracts away the communication details and allows virtualization of a shared memory between IoT devices. The results show that optimizing critical path performance increases throughput and reduces delay across multiple devices during CEP operations.

This research introduces the first large-scale, well-balanced Persian social media text classification dataset, specifically designed to address the lack of comprehensive resources in this domain. The dataset comprises 36,000 posts across nine categories (Economic, Artistic, Sports, Political, Social, Health, Psychological, Historical, and Science & Technology), each containing 4,000 samples to ensure balanced class distribution. Data collection involved 60,000 raw posts from various Persian social media platforms, followed by rigorous preprocessing and hybrid annotation combining ChatGPT-based few-shot prompting with human verification. To mitigate class imbalance, we employed undersampling with semantic redundancy removal and advanced data augmentation strategies integrating lexical replacement and generative prompting. We benchmarked several models, including BiLSTM, XLM-RoBERTa (with LoRA and AdaLoRA adaptations), FaBERT, SBERT-based architectures, and the Persian-specific TookaBERT (Base and Large). Experimental results show that transformer-based models consistently outperform traditional neural networks, with TookaBERT-Large achieving the best performance (Precision: 0.9622, Recall: 0.9621, F1- score: 0.9621). Class-wise evaluation further confirms robust performance across all categories, though social and political texts exhibited slightly lower scores due to inherent ambiguity. This research presents a new high-quality dataset and provides comprehensive evaluations of cutting-edge models, establishing a solid foundation for further developments in Persian NLP, including trend analysis, social behavior modeling, and user classification. The dataset is publicly available to support future research endeavors.

Model merging has emerged as a powerful paradigm for combining the capabilities of distinct expert models without the high computational cost of retraining, yet current methods are fundamentally constrained to homogeneous architectures. For GNNs, however, message passing is topology-dependent and sensitive to misalignment, making direct parameter-space merging unreliable. To bridge this gap, we introduce H-GRAMA (Heterogeneous Graph Routing and Message Alignment), a training-free framework that lifts merging from parameter space to operator space. We formalize Universal Message Passing Mixture (UMPM), a shared operator family that expresses heterogeneous GNN layers in a common functional language. H-GRAMA enables cross-architecture GNN merging (e.g., GCN to GAT) without retraining, retaining high specialist accuracy in most cases in compatible depth settings and achieving inference speedups of 1.2x to 1.9x over ensembles.

Representational similarity metrics typically force all units to be matched, making them susceptible to noise and outliers common in neural representations. We extend the soft-matching distance to a partial optimal transport setting that allows some neurons to remain unmatched, yielding rotation-sensitive but robust correspondences. This partial soft-matching distance provides theoretical advantages -- relaxing strict mass conservation while maintaining interpretable transport costs -- and practical benefits through efficient neuron ranking in terms of cross-network alignment without costly iterative recomputation. In simulations, it preserves correct matches under outliers and reliably selects the correct model in noise-corrupted identification tasks. On fMRI data, it automatically excludes low-reliability voxels and produces voxel rankings by alignment quality that closely match computationally expensive brute-force approaches. It achieves higher alignment precision across homologous brain areas than standard soft-matching, which is forced to match all units regardless of quality. In deep networks, highly matched units exhibit similar maximally exciting images, while unmatched units show divergent patterns. This ability to partition by match quality enables focused analyses, e.g., testing whether networks have privileged axes even within their most aligned subpopulations. Overall, partial soft-matching provides a principled and practical method for representational comparison under partial correspondence.

Graph Neural Networks (GNNs) are increasingly explored for physical design analysis in Electronic Design Automation, particularly for modeling Clock Tree Synthesis behavior such as clock skew and buffering complexity. However, practical deployment remains limited due to the prohibitive memory and runtime cost of operating on raw gate-level netlists. Graph coarsening is commonly used to improve scalability, yet its impact on CTS-critical learning objectives is not well characterized. This paper introduces CTS-Bench, a benchmark suite for systematically evaluating the trade-offs between graph coarsening, prediction accuracy, and computational efficiency in GNN-based CTS analysis. CTS-Bench consists of 4,860 converged physical design solutions spanning five architectures and provides paired raw gate-level and clustered graph representations derived from post-placement designs. Using clock skew prediction as a representative CTS task, we demonstrate a clear accuracy-efficiency trade-off. While graph coarsening reduces GPU memory usage by up to 17.2x and accelerates training by up to 3x, it also removes structural information essential for modeling clock distribution, frequently resulting in negative $R^2$ scores under zero-shot evaluation. Our findings indicate that generic graph clustering techniques can fundamentally compromise CTS learning objectives, even when global physical metrics remain unchanged. CTS-Bench enables principled evaluation of CTS-aware graph coarsening strategies, supports benchmarking of GNN architectures and accelerators under realistic physical design constraints, and provides a foundation for developing learning-assisted CTS analysis and optimization techniques.

Accurate quantification of the relationship between forest loss and associated carbon emissions is critical for both environmental monitoring and policy evaluation. Although many studies have documented spatial patterns of forest degradation, there is limited understanding of the dynamic elasticity linking tree cover loss to carbon emissions at subnational scales. In this paper, we construct a comprehensive panel dataset of annual forest loss and carbon emission estimates for U.S. subnational administrative units from 2001 to 2023, based on the Hansen Global Forest Change dataset. We apply fixed effects and dynamic panel regression techniques to isolate within-region variation and account for temporal persistence in emissions. Our results show that forest loss has a significant positive short-run elasticity with carbon emissions, and that emissions exhibit strong persistence over time. Importantly, the estimated long-run elasticity, accounting for autoregressive dynamics, is substantially larger than the short-run effect, indicating cumulative impacts of repeated forest loss events. These findings highlight the importance of modeling temporal dynamics when assessing environmental responses to land cover change. The dynamic elasticity framework proposed here offers a robust and interpretable tool for analyzing environmental change processes, and can inform both regional monitoring systems and carbon accounting frameworks.

A significant portion of recent research on Large Language Model (LLM) alignment focuses on developing new policy optimization methods based on Group Relative Policy Optimization (GRPO). Two prominent directions have emerged: (i) a shift toward sequence-level importance sampling weights that better align with the sequence-level rewards used in many tasks, and (ii) alternatives to PPO-style clipping that aim to avoid the associated loss of training signal and entropy collapse. Recent work, such as Soft Adaptive Policy Optimization (SAPO), reformulates the Scopic objective within the GRPO framework and achieves both sequence coherence and token adaptivity. Geometric-Mean Policy Optimization (GMPO) leverages token-wise ratio clipping within sequence importance sampling weights. Building on these ideas, this work proposes a new objective that promotes effective policy exploration while maintaining training stability. Specifically, we introduce Soft Sequence Policy Optimization, an off-policy reinforcement learning objective that incorporates soft gating functions over token-level probability ratios within sequence-level importance weights.

As cities evolve over time, challenges such as traffic congestion and functional imbalance increasingly necessitate urban renewal through efficient modification of existing plans, rather than complete re-planning. In practice, even minor urban changes require substantial manual effort to redraw geospatial layouts, slowing the iterative planning and decision-making procedure. Motivated by recent advances in agentic systems and multimodal reasoning, we formulate urban renewal as a machine-executable task that iteratively modifies existing urban plans represented in structured geospatial formats. More specifically, we represent urban layouts using GeoJSON and decompose natural-language editing instructions into hierarchical geometric intents spanning polygon-, line-, and point-level operations. To coordinate interdependent edits across spatial elements and abstraction levels, we propose a hierarchical agentic framework that jointly performs multi-level planning and execution with explicit propagation of intermediate spatial constraints. We further introduce an iterative execution-validation mechanism that mitigates error accumulation and enforces global spatial consistency during multi-step editing. Extensive experiments across diverse urban editing scenarios demonstrate significant improvements in efficiency, robustness, correctness, and spatial validity over existing baselines.

Early and accurate classification of retinal diseases is critical to counter vision loss and for guiding clinical management of retinal diseases. In this study, we proposed a deep learning method for retinal disease classification utilizing optical coherence tomography (OCT) images from the Retinal OCT Image Classification - C8 dataset (comprising 24,000 labeled images spanning eight conditions). Images were resized to 224x224 px and tested on convolutional neural network (CNN) architectures: Xception and InceptionV3. Data augmentation techniques (CutMix, MixUp) were employed to enhance model generalization. Additionally, we applied GradCAM and LIME for interpretability evaluation. We implemented this in a real-world scenario via our web application named RetinaVision. This study found that Xception was the most accurate network (95.25%), followed closely by InceptionV3 (94.82%). These results suggest that deep learning methods allow effective OCT retinal disease classification and highlight the importance of implementing accuracy and interpretability for clinical applications.

Ultrasound (US) imaging poses unique challenges for representation learning due to its inherently noisy acquisition process. The low signal-to-noise ratio and stochastic speckle patterns hinder standard self-supervised learning methods relying on a pixel-level reconstruction objective. Joint-Embedding Predictive Architectures (JEPAs) address this drawback by predicting masked latent representations rather than raw pixels. However, standard approaches depend on hyperparameter-brittle and computationally expensive online teachers updated via exponential moving average. We propose US-JEPA, a self-supervised framework that adopts the Static-teacher Asymmetric Latent Training (SALT) objective. By using a frozen, domain-specific teacher to provide stable latent targets, US-JEPA decouples student-teacher optimization and pushes the student to expand upon the semantic priors of the teacher. In addition, we provide the first rigorous comparison of all publicly available state-of-the-art ultrasound foundation models on UltraBench, a public dataset benchmark spanning multiple organs and pathological conditions. Under linear probing for diverse classification tasks, US-JEPA achieves performance competitive with or superior to domain-specific and universal vision foundation model baselines. Our results demonstrate that masked latent prediction provides a stable and efficient path toward robust ultrasound representations.

Agentic memory systems enable large language model (LLM) agents to maintain state across long interactions, supporting long-horizon reasoning and personalization beyond fixed context windows. Despite rapid architectural development, the empirical foundations of these systems remain fragile: existing benchmarks are often underscaled, evaluation metrics are misaligned with semantic utility, performance varies significantly across backbone models, and system-level costs are frequently overlooked. This survey presents a structured analysis of agentic memory from both architectural and system perspectives. We first introduce a concise taxonomy of MAG systems based on four memory structures. Then, we analyze key pain points limiting current systems, including benchmark saturation effects, metric validity and judge sensitivity, backbone-dependent accuracy, and the latency and throughput overhead introduced by memory maintenance. By connecting the memory structure to empirical limitations, this survey clarifies why current agentic memory systems often underperform their theoretical promise and outlines directions for more reliable evaluation and scalable system design.

Managing personal health data is a challenge in today's fragmented and institution-centric healthcare ecosystem. Individuals often lack meaningful control over their medical records, which are scattered across incompatible systems and formats. This vision paper presents Health+, a user-centric, multimodal health data management system that empowers individuals (including those with limited technical expertise) to upload, query, and share their data across modalities (e.g., text, images, reports). Rather than aiming for institutional overhaul, Health+ emphasizes individual agency by providing intuitive interfaces and intelligent recommendations for data access and sharing. At the system level, it tackles the complexity of storing, integrating, and securing heterogeneous health records, ensuring both efficiency and privacy. By unifying multimodal data and prioritizing patients, Health+ lays the foundation for a more connected, interpretable, and user-controlled health information ecosystem.

Personalization in Question Answering (QA) requires answers that are both accurate and aligned with users' background, preferences, and historical context. Existing state-of-the-art methods primarily rely on retrieval-augmented generation (RAG) solutions that construct personal context by retrieving relevant items from the user's profile. Existing methods use the user's query directly to retrieve personal documents, and such strategies often lead to surface-level personalization. We propose PR2 (Personalized Retrieval-Augmented Reasoning), a reinforcement learning framework that integrates reasoning and retrieval from personal context for personalization. PR2 learns adaptive retrieval-reasoning policies, determining when to retrieve, what evidence to retrieve from user profiles, and how to incorporate it into intermediate reasoning steps. By optimizing multi-turn reasoning trajectories under a personalized reward function, the framework reinforces reasoning paths that better align with user-specific preferences and contextual signals reflected by the reward model. Extensive experiments on the LaMP-QA benchmark using three LLMs show that PR2 consistently outperforms strong baselines, achieving an average relative improvement of 8.8%-12% in personalized QA.

Underwater glider robots have become an indispensable tool for ocean sampling. Although stakeholders are calling for tools to manage increasingly large fleets of gliders, successful autonomous long-term deployments have thus far been scarce, which hints at a lack of suitable methodologies and systems. In this work, we formulate glider navigation planning as a stochastic shortest-path Markov Decision Process and propose a sample-based online planner based on Monte Carlo Tree Search. Samples are generated by a physics-informed simulator that captures uncertain execution of controls and ocean current forecasts while remaining computationally tractable. The simulator parameters are fitted using historical glider data. We integrate these methods into an autonomous command-and-control system for Slocum gliders that enables closed-loop replanning at each surfacing. The resulting system was validated in two field deployments in the North Sea totalling approximately 3 months and 1000 km of autonomous operation. Results demonstrate improved efficiency compared to straight-to-goal navigation and show the practicality of sample-based planning for long-term marine autonomy.

The image purification strategy constructs an intermediate distribution with aligned anatomical structures, which effectively corrects the spatial misalignment between real-world ultra-low-dose CT and normal-dose CT images and significantly enhances the structural preservation ability of denoising models. However, this strategy exhibits two inherent limitations. First, it suppresses noise only in the chest wall and bone regions while leaving the image background untreated. Second, it lacks a dedicated mechanism for denoising the lung parenchyma. To address these issues, we systematically redesign the original image purification strategy and propose an improved version termed IPv2. The proposed strategy introduces three core modules, namely Remove Background, Add noise, and Remove noise. These modules endow the model with denoising capability in both background and lung tissue regions during training data construction and provide a more reasonable evaluation protocol through refined label construction at the testing stage. Extensive experiments on our previously established real-world patient lung CT dataset acquired at 2% radiation dose demonstrate that IPv2 consistently improves background suppression and lung parenchyma restoration across multiple mainstream denoising models. The code is publicly available at https://github.com/MonkeyDadLufy/Image-Purification-Strategy-v2.

While Vision-Language-Action (VLA) models have seen rapid progress in pretraining, their advancement in Reinforcement Learning (RL) remains hampered by low sample efficiency and sparse rewards in real-world settings. Developing generalizable process reward models is essential for providing the fine-grained feedback necessary to bridge this gap, yet existing temporal value functions often fail to generalize beyond their training domains. We introduce TOPReward, a novel, probabilistically grounded temporal value function that leverages the latent world knowledge of pretrained video Vision-Language Models (VLMs) to estimate robotic task progress. Unlike prior methods that prompt VLMs to directly output progress values, which are prone to numerical misrepresentation, TOPReward extracts task progress directly from the VLM's internal token logits. In zero-shot evaluations across 130+ distinct real-world tasks and multiple robot platforms (e.g., Franka, YAM, SO-100/101), TOPReward achieves 0.947 mean Value-Order Correlation (VOC) on Qwen3-VL, dramatically outperforming the state-of-the-art GVL baseline which achieves near-zero correlation on the same open-source model. We further demonstrate that TOPReward serves as a versatile tool for downstream applications, including success detection and reward-aligned behavior cloning.

The upcoming sixth Generation (6G) of wireless networks envisions ultra-low latency and energy efficient Edge Inference (EI) for diverse Internet of Things (IoT) applications. However, traditional digital hardware for machine learning is power intensive, motivating the need for alternative computation paradigms. Over-The-Air (OTA) computation is regarded as an emerging transformative approach assigning the wireless channel to actively perform computational tasks. This article introduces the concept of Metasurfaces-Integrated Neural Networks (MINNs), a physical-layer-enabled deep learning framework that leverages programmable multi-layer metasurface structures and Multiple-Input Multiple-Output (MIMO) channels to realize computational layers in the wave propagation domain. The MINN system is conceptualized as three modules: Encoder, Channel (uncontrollable propagation features and metasurfaces), and Decoder. The first and last modules, realized respectively at the multi-antenna transmitter and receiver, consist of conventional digital or purposely designed analog Deep Neural Network (DNN) layers, and the metasurfaces responses of the Channel module are optimized alongside all modules as trainable weights. This architecture enables computation offloading into the end-to-end physical layer, flexibly among its constituent modules, achieving performance comparable to fully digital DNNs while significantly reducing power consumption. The training of the MINN framework, two representative variations, and performance results for indicative applications are presented, highlighting the potential of MINNs as a lightweight and sustainable solution for future EI-enabled wireless systems. The article is concluded with a list of open challenges and promising research directions.

While large language models (LLMs) have emerged as powerful decision-makers across a wide range of single-agent and stationary environments, fewer efforts have been devoted to settings where LLMs must engage in \emph{repeated} and \emph{strategic} interactions with unknown or dynamic opponents. In such settings, recipes built upon \emph{offline} pre-training or fine-tuning, though robust against worst-case adversaries, do not fully exploit the capability of LLMs to adapt \emph{online} based on interaction feedback. Instead, we explore the more natural perspective of scaling inference-time computation as a mechanism for adaptation, embedding the principles of a classical game-theoretical learning dynamic, \emph{smooth Fictitious Play (sFP)}, into LLM inference: (i) for belief formation, we employ an auxiliary opponent model that in-context learns to imitate the time-averaged behavior of the opponent; (ii) for best response, we advance best-of-$N$ (BoN) sampling by simulating against the opponent model. Empirical evaluations on two distinct forms of repeated negotiation games demonstrate that our method enables significant performance improvement over repeated online interaction compared to various baselines, offering a scalable and principled approach to repeated strategic decision-making without any parameter updates.

In this paper, our objective is to design, build, and verify a closed-loop environmental control system tailored for small-scale agriculture applications. This project aims to develop a low-cost, safety-critical embedded solution using the Nuvoton NUC140 microcontroller to automate temperature regulation. The goal was to mitigate crop yield losses caused by environmental fluctuations in a greenhouse. Our final implemented system successfully meets all design specifications, demonstrating robust temperature regulation through a PID control loop and ensuring hardware safety through galvanic isolation

Successful cooperation among decentralized agents requires each agent to quickly adapt its plan to the behavior of other agents. In scenarios where agents cannot confidently predict one another's intentions and plans, language communication can be crucial for ensuring safety. In this work, we focus on path-level cooperation in which agents must adapt their paths to one another in order to avoid collisions or perform physical collaboration such as joint carrying. In particular, we propose a safe and interpretable multimodal path planning method, CaPE (Code as Path Editor), which generates and updates path plans for an agent based on the environment and language communication from other agents. CaPE leverages a vision-language model (VLM) to synthesize a path editing program verified by a model-based planner, grounding communication to path plan updates in a safe and interpretable way. We evaluate our approach in diverse simulated and real-world scenarios, including multi-robot and human-robot cooperation in autonomous driving, household, and joint carrying tasks. Experimental results demonstrate that CaPE can be integrated into different robotic systems as a plug-and-play module, greatly enhancing a robot's ability to align its plan to language communication from other robots or humans. We also show that the combination of the VLM-based path editing program synthesis and model-based planning safety enables robots to achieve open-ended cooperation while maintaining safety and interpretability.

Evaluating personalized, sequential treatment strategies for Alzheimer's disease (AD) using clinical trials is often impractical due to long disease horizons and substantial inter-patient heterogeneity. To address these constraints, we present the Alzheimer's Learning Platform for Adaptive Care Agents (ALPACA), an open-source, Gym-compatible reinforcement learning (RL) environment for systematically exploring personalized treatment strategies using existing therapies. ALPACA is powered by the Continuous Action-conditioned State Transitions (CAST) model trained on longitudinal trajectories from the Alzheimer's Disease Neuroimaging Initiative (ADNI), enabling medication-conditioned simulation of disease progression under alternative treatment decisions. We show that CAST autoregressively generates realistic medication-conditioned trajectories and that RL policies trained in ALPACA outperform no-treatment and behavior-cloned clinician baselines on memory-related outcomes. Interpretability analyses further indicated that the learned policies relied on clinically meaningful patient features when selecting actions. Overall, ALPACA provides a reusable in silico testbed for studying individualized sequential treatment decision-making for AD.

Microfluidic devices have emerged as powerful tools in various laboratory applications, but the complexity of their design limits accessibility for many practitioners. While progress has been made in microfluidic design automation (MFDA), a practical and intuitive solution is still needed to connect microfluidic practitioners with MFDA techniques. This work introduces the first practical application of large language models (LLMs) in this context, providing a preliminary demonstration. Building on prior research in hardware description language (HDL) code generation with LLMs, we propose an initial methodology to convert natural language microfluidic device specifications into system-level structural Verilog netlists. We demonstrate the feasibility of our approach by generating structural netlists for practical benchmarks representative of typical microfluidic designs with correct functional flow and an average syntactical accuracy of 88%.

Page Objects (POs) are a widely adopted design pattern for improving the maintainability and scalability of automated end-to-end web tests. However, creating and maintaining POs is still largely a manual, labor-intensive activity, while automated solutions have seen limited practical adoption. In this context, the potential of Large Language Models (LLMs) for these tasks has remained largely unexplored. This paper presents an empirical study on the feasibility of using LLMs, specifically GPT-4o and DeepSeek Coder, to automatically generate POs for web testing. We evaluate the generated artifacts on an existing benchmark of five web applications for which manually written POs are available (the ground truth), focusing on accuracy (i.e., the proportion of ground truth elements correctly identified) and element recognition rate (i.e., the proportion of ground truth elements correctly identified or marked for modification). Our results show that LLMs can generate syntactically correct and functionally useful POs with accuracy values ranging from 32.6% to 54.0% and element recognition rate exceeding 70% in most cases. Our study contributes the first systematic evaluation of LLMs strengths and open challenges for automated PO generation, and provides directions for further research on integrating LLMs into practical testing workflows.

Identifying low-energy adsorption geometries on catalytic surfaces is a practical bottleneck for computational heterogeneous catalysis: the difficulty lies not only in the cost of density functional theory (DFT) but in proposing initial placements that relax into the correct energy basins. Conditional denoising diffusion has improved success rates, yet requires $\sim$100 iterative steps per sample. Here we introduce AdsorbFlow, a deterministic generative model that learns an energy-conditioned vector field on the rigid-body configuration space of adsorbate translation and rotation via conditional flow matching. Energy information enters through classifier-free guidance conditioning -- not energy-gradient guidance -- and sampling reduces to integrating an ODE in as few as 5 steps. On OC20-Dense with full DFT single-point verification, AdsorbFlow with an EquiformerV2 backbone achieves 61.4% SR@10 and 34.1% SR@1 -- surpassing AdsorbDiff (31.8% SR@1, 41.0% SR@10) at every evaluation level and AdsorbML (47.7% SR@10) -- while using 20 times fewer generative steps and achieving the lowest anomaly rate among generative methods (6.8%). On 50 out-of-distribution systems, AdsorbFlow retains 58.0% SR@10 with a MLFF-to-DFT gap of only 4~percentage points. These results establish that deterministic transport is both faster and more accurate than stochastic denoising for adsorbate placement.

The test-time compute strategy, such as Chain-of-Thought (CoT), has significantly enhanced the ability of large language models to solve complex tasks like logical reasoning. However, empirical studies indicate that simply increasing the compute budget can sometimes lead to a collapse in test-time performance when employing typical task decomposition strategies such as CoT. This work hypothesizes that reasoning failures with larger compute budgets stem from static planning methods, which hardly perceive the intrinsic boundaries of LLM reasoning. We term it as the Limited Reasoning Space hypothesis and perform theoretical analysis through the lens of a non-autonomous stochastic dynamical system. This insight suggests that there is an optimal range for compute budgets; over-planning can lead to redundant feedback and may even impair reasoning capabilities. To exploit the compute-scaling benefits and suppress over-planning, this work proposes Halo, a model predictive control framework for LLM planning. Halo is designed for long-horizon tasks with reason-based planning and crafts an entropy-driven dual controller, which adopts a Measure-then-Plan strategy to achieve controllable reasoning. Experimental results demonstrate that Halo outperforms static baselines on complex long-horizon tasks by dynamically regulating planning at the reasoning boundary.

Multimodal Large Language Models (MLLMs) have demonstrated strong performance on the UI-to-code task, which aims to generate UI code from design mock-ups. However, when applied to long and complex websites, they often struggle with fragmented segmentation, redundant code generation for repetitive components, and frequent UI inconsistencies. To systematically investigate and address these challenges, we introduce ComUIBench, a new multi-page complex webpage benchmark with component annotations, designed to evaluate MLLMs' ability to generate reusable UI code in realistic website scenarios. Building upon this benchmark, we propose ComUICoder, a component-based UI code generation framework that emphasizes semantic-aware segmentation, code reuse, and fine-grained refinement. Specifically, ComUICoder incorporates (1) Hybrid Semantic-aware Block Segmentation for accurate UI semantic coherent block detection, (2) Visual-aware Graph-based Block Merge to consolidate structurally similar components within and across webpages for reusable implementation, and (3) Priority-based Element-wise Feedback to refine generated code and reduce element-level inconsistencies. Extensive experiments demonstrate that ComUICoder significantly improves overall generation quality and code reusability on complex multipage websites. Our datasets and code are publicly available at https://github.com/WebPAI/ComUICoder.

We introduce a gradient-free framework for identifying minimal, sufficient, and decision-preserving explanations in vision models by isolating the smallest subset of representational units whose joint activation preserves predictions. Unlike existing approaches that aggregate all units, often leading to cluttered saliency maps, our approach, DD-CAM, identifies a 1-minimal subset whose joint activation suffices to preserve the prediction (i.e., removing any unit from the subset alters the prediction). To efficiently isolate minimal sufficient subsets, we adapt delta debugging, a systematic reduction strategy from software debugging, and configure its search strategy based on unit interactions in the classifier head: testing individual units for models with non-interacting units and testing unit combinations for models in which unit interactions exist. We then generate minimal, prediction-preserving saliency maps that highlight only the most essential features. Our experimental evaluation demonstrates that our approach can produce more faithful explanations and achieve higher localization accuracy than the state-of-the-art CAM-based approaches.

Second-order optimizers can significantly accelerate large-scale training, yet their naive federated variants are often unstable or even diverge on non-IID data. We show that a key culprit is \emph{preconditioner drift}: client-side second-order training induces heterogeneous \emph{curvature-defined geometries} (i.e., preconditioner coordinate systems), and server-side model averaging updates computed under incompatible metrics, corrupting the global descent direction. To address this geometric mismatch, we propose \texttt{FedPAC}, a \emph{preconditioner alignment and correction} framework for reliable federated second-order optimization. \texttt{FedPAC} explicitly decouples parameter aggregation from geometry synchronization by: (i) \textbf{Alignment} (i.e.,aggregating local preconditioners into a global reference and warm-starting clients via global preconditioner); and (ii) \textbf{Correction} (i.e., steering local preconditioned updates using a global preconditioned direction to suppress long-term drift). We provide drift-coupled non-convex convergence guarantees with linear speedup under partial participation. Empirically, \texttt{FedPAC} consistently improves stability and accuracy across vision and language tasks, achieving up to $5.8\%$ absolute accuracy gain on CIFAR-100 with ViTs. Code is available at https://anonymous.4open.science/r/FedPAC-8B24.

This brief presents a runtime-adaptive, performance-enhanced vector engine featuring a low-resource, iterative CORDIC-based MAC unit for edge AI acceleration. The proposed design enables dynamic reconfiguration between approximate and accurate modes, exploiting the latency-accuracy trade-off for a wide range of workloads. Its resource-efficient approach further enables up to 4x throughput improvement within the same hardware resources by leveraging vectorised, time-multiplexed execution and flexible precision scaling. With a time-multiplexed multi-AF block and a lightweight pooling and normalisation unit, the proposed vector engine supports flexible precision (4/8/16-bit) and high MAC density. The ASIC implementation results show that each MAC stage can save up to 33% of time and 21% of power, with a 256-PE configuration that achieves higher compute density (4.83 TOPS/mm2 ) and energy efficiency (11.67 TOPS/W) than previous state-of-the-art work. A detailed hardware-software co-design methodology for object detection and classification tasks on Pynq-Z2 is discussed to assess the proposed architecture, demonstrating a scalable, energy-efficient solution for edge AI applications.

Solving partial differential equations (PDEs) by neural networks as well as Kolmogorov-Arnold Networks (KANs), including physics-informed neural networks (PINNs), physics-informed KANs (PIKANs), and neural operators, are known to exhibit spectral bias, whereby low-frequency components of the solution are learned significantly faster than high-frequency modes. While spectral bias is often treated as an intrinsic representational limitation of neural architectures, its interaction with optimization dynamics and physics-based loss formulations remains poorly understood. In this work, we provide a systematic investigation of spectral bias in physics-informed and operator learning frameworks, with emphasis on the coupled roles of network architecture, activation functions, loss design, and optimization strategy. We quantify spectral bias through frequency-resolved error metrics, Barron-norm diagnostics, and higher-order statistical moments, enabling a unified analysis across elliptic, hyperbolic, and dispersive PDEs. Through diverse benchmark problems, including the Korteweg-de Vries, wave and steady-state diffusion-reaction equations, turbulent flow reconstruction, and earthquake dynamics, we demonstrate that spectral bias is not simply representational but fundamentally dynamical. In particular, second-order optimization methods substantially alter the spectral learning order, enabling earlier and more accurate recovery of high-frequency modes for all PDE types. For neural operators, we further show that spectral bias is dependent on the neural operator architecture and can also be effectively mitigated through spectral-aware loss formulations without increasing the inference cost.

Modern manufacturing systems often experience multiple and unpredictable failure behaviors, yet most existing prognostic models assume a fixed, known set of failure modes with labeled historical data. This assumption limits the use of digital twins for predictive maintenance, especially in high-mix or adaptive production environments, where new failure modes may emerge, and the failure mode labels may be unavailable. To address these challenges, we propose a novel Bayesian nonparametric framework that unifies a Dirichlet process mixture module for unsupervised failure mode discovery with a neural network-based prognostic module. The key innovation lies in an iterative feedback mechanism to jointly learn two modules. These modules iteratively update one another to dynamically infer, expand, or merge failure modes as new data arrive while providing high prognostic accuracy. Experiments on both simulation and aircraft engine datasets show that the proposed approach performs competitively with or significantly better than existing approaches. It also exhibits robust online adaptation capabilities, making it well-suited for digital-twin-based system health management in complex manufacturing environments.

Reinforcement learning fine-tuning has proven effective for steering generative diffusion models toward desired properties in image and molecular domains. Graph diffusion models have similarly been applied to combinatorial structure generation, including neural architecture search (NAS). However, neural architectures are directed acyclic graphs (DAGs) where edge direction encodes functional semantics such as data flow-information that existing graph diffusion methods, designed for undirected structures, discard. We propose Directed Graph Policy Optimization (DGPO), which extends reinforcement learning fine-tuning of discrete graph diffusion models to DAGs via topological node ordering and positional encoding. Validated on NAS-Bench-101 and NAS-Bench-201, DGPO matches the benchmark optimum on all three NAS-Bench-201 tasks (91.61%, 73.49%, 46.77%). The central finding is that the model learns transferable structural priors: pretrained on only 7% of the search space, it generates near-oracle architectures after fine-tuning, within 0.32 percentage points of the full-data model and extrapolating 7.3 percentage points beyond its training ceiling. Bidirectional control experiments confirm genuine reward-driven steering, with inverse optimization reaching near random-chance accuracy (9.5%). These results demonstrate that reinforcement learning-steered discrete diffusion, once extended to handle directionality, provides a controllable generative framework for directed combinatorial structures.

Fuzzy systems show strong potential in explainable AI due to their rule-based architecture and linguistic variables. Existing approaches navigate the accuracy-explainability trade-off either through evolutionary multi-objective optimization (MOO), which is computationally expensive, or gradient-based scalarization, which cannot recover non-convex Pareto regions. We propose X-ANFIS, an alternating bi-objective gradient-based optimization scheme for explainable adaptive neuro-fuzzy inference systems. Cauchy membership functions are used for stable training under semantically controlled initializations, and a differentiable explainability objective is introduced and decoupled from the performance objective through alternating gradient passes. Validated in approximately 5,000 experiments on nine UCI regression datasets, X-ANFIS consistently achieves target distinguishability while maintaining competitive predictive accuracy, recovering solutions beyond the convex hull of the MOO Pareto front.

The collection and detection of video anomaly data has long been a challenging problem due to its rare occurrence and spatio-temporal scarcity. Existing video anomaly detection (VAD) methods under perform in open-world scenarios. Key contributing factors include limited dataset diversity, and inadequate understanding of context-dependent anomalous semantics. To address these issues, i) we propose LAVIDA, an end-to-end zero-shot video anomaly detection framework. ii) LAVIDA employs an Anomaly Exposure Sampler that transforms segmented objects into pseudo-anomalies to enhance model adaptability to unseen anomaly categories. It further integrates a Multimodal Large Language Model (MLLM) to bolster semantic comprehension capabilities. Additionally, iii) we design a token compression approach based on reverse attention to handle the spatio-temporal scarcity of anomalous patterns and decrease computational cost. The training process is conducted solely on pseudo anomalies without any VAD data. Evaluations across four benchmark VAD datasets demonstrate that LAVIDA achieves SOTA performance in both frame-level and pixel-level anomaly detection under the zero-shot setting. Our code is available in https://github.com/VitaminCreed/LAVIDA.

On-the-fly Directed Controller Synthesis (OTF-DCS) mitigates state-space explosion by incrementally exploring the system and relies critically on an exploration policy to guide search efficiently. Recent reinforcement learning (RL) approaches learn such policies and achieve promising zero-shot generalization from small training instances to larger unseen ones. However, a fundamental limitation is anisotropic generalization, where an RL policy exhibits strong performance only in a specific region of the domain-parameter space while remaining fragile elsewhere due to training stochasticity and trajectory-dependent bias. To address this, we propose a Soft Mixture-of-Experts framework that combines multiple RL experts via a prior-confidence gating mechanism and treats these anisotropic behaviors as complementary specializations. The evaluation on the Air Traffic benchmark shows that Soft-MoE substantially expands the solvable parameter space and improves robustness compared to any single expert.

Hierarchical Navigable Small World (HNSW) has demonstrated impressive accuracy and low latency for high-dimensional nearest neighbor searches. However, its high computational demands and irregular, large-volume data access patterns present significant challenges to search efficiency. To address these challenges, we introduce pHNSW, an algorithm-hardware co-optimized solution that accelerates HNSW through Principal Component Analysis (PCA) filtering. On the algorithm side, we apply PCA filtering to reduce the dimensionality of the dataset, thereby lowering the volume of neighbor access and decreasing the computational load for distance calculations. On the hardware side, we design the pHNSW processor with custom instructions to optimize search throughput and energy efficiency. In the experiments, we synthesized the pHNSW processor RTL design with a 65nm technology node and evaluated it using DDR4 and HBM1.0 DRAM standards. The results show that pHNSW boosts Queries per Second (QPS) by 14.47x-21.37x on a CPU and 5.37x-8.46x on a GPU, while reducing energy consumption by up to 57.4% compared to standard HNSW implementation.

Low-precision training is critical for optimizing the trade-off between model quality and training costs, necessitating the joint allocation of model size, dataset size, and numerical precision. While empirical scaling laws suggest that quantization impacts effective model and data capacities or acts as an additive error, the theoretical mechanisms governing these effects remain largely unexplored. In this work, we initiate a theoretical study of scaling laws for low-precision training within a high-dimensional sketched linear regression framework. By analyzing multiplicative (signal-dependent) and additive (signal-independent) quantization, we identify a critical dichotomy in their scaling behaviors. Our analysis reveals that while both schemes introduce an additive error and degrade the effective data size, they exhibit distinct effects on effective model size: multiplicative quantization maintains the full-precision model size, whereas additive quantization reduces the effective model size. Numerical experiments validate our theoretical findings. By rigorously characterizing the complex interplay among model scale, dataset size, and quantization error, our work provides a principled theoretical basis for optimizing training protocols under practical hardware constraints.

Retrieval-Augmented Generation (RAG) enhances the reasoning ability of Large Language Models (LLMs) by dynamically integrating external knowledge, thereby mitigating hallucinations and strengthening contextual grounding for structured data such as graphs. Nevertheless, most existing RAG variants for textual graphs concentrate on low-dimensional structures -- treating nodes as entities (0-dimensional) and edges or paths as pairwise or sequential relations (1-dimensional), but overlook cycles, which are crucial for reasoning over relational loops. Such cycles often arise in questions requiring closed-loop inference about similar objects or relative positions. This limitation often results in incomplete contextual grounding and restricted reasoning capability. In this work, we propose Topology-enhanced Retrieval-Augmented Generation (TopoRAG), a novel framework for textual graph question answering that effectively captures higher-dimensional topological and relational dependencies. Specifically, TopoRAG first lifts textual graphs into cellular complexes to model multi-dimensional topological structures. Leveraging these lifted representations, a topology-aware subcomplex retrieval mechanism is proposed to extract cellular complexes relevant to the input query, providing compact and informative topological context. Finally, a multi-dimensional topological reasoning mechanism operates over these complexes to propagate relational information and guide LLMs in performing structured, logic-aware inference. Empirical evaluations demonstrate that our method consistently surpasses existing baselines across diverse textual graph tasks.

Large language models can follow complex procedures yet fail at a seemingly trivial final step: reporting a value they themselves computed moments earlier. We study this phenomenon as \emph{procedural hallucination}: failure to execute a verifiable, prompt-grounded specification even when the correct value is present in context. In long-context binding tasks with a known single-token candidate set, we find that many errors are readout-stage routing failures. Specifically, failures decompose into Stage~2A (gating) errors, where the model does not enter answer mode, and Stage~2B (binding) errors, where it enters answer mode but selects the wrong candidate (often due to recency bias). In the hard regime, Stage~2B accounts for most errors across model families in our tasks (Table~1). On Stage~2B error trials, a linear probe on the final-layer residual stream recovers the correct value far above chance (e.g., 74\% vs.\ 2\% on Qwen2.5-3B; Table~2), indicating that the answer is encoded but not used. We formalize ``present but not used'' via available vs.\ used mutual information and pseudo-prior interventions, yielding output-computable diagnostics and information-budget certificates. Finally, an oracle checkpointing intervention that restates the true binding near the query can nearly eliminate Stage~2B failures at long distance (e.g., Qwen2.5-3B $0/400 \rightarrow 399/400$ at $k = 1024$; Table~8).

Tabular foundation models aim to make machine learning accessible for enterprise data without task-specific training. This paper presents the first independent evaluation of SAP's Retrieval Pretrained Transformer (RPT-1) from a practitioner perspective. RPT-1 is a compact 64.6 MB model pretrained on 1.34 TB of structured data across 3.1 million tables. We benchmark it against tuned gradient-boosted decision trees (XGBoost, LightGBM, CatBoost) on three SAP business scenarios: demand forecasting across SD/MM/PP modules, predictive data integrity in BC/MM/QM, and financial risk classification in FI/CO/AR. Across five-fold cross-validation on datasets ranging from 2,500 to 3,200 rows, RPT-1 reaches 91-96% of tuned GBDT accuracy without any training examples. The classification gap is modest at 3.6-4.1 percentage points on AUC-ROC, though regression tasks show wider gaps of 8.9-11.1 percentage points on R-squared. An interesting finding is a crossover at roughly 75-100 context rows where RPT-1 actually outperforms XGBoost under limited data. Based on these results, we propose a practical hybrid workflow: use RPT-1 for rapid screening, then train GBDT selectively where prediction accuracy justifies the effort. All experiments are reproducible through publicly available Hugging Face Spaces.

Digital collaboration systems support asynchronous work over replicated data, where conflicts arise when concurrent operations cannot be unambiguously integrated into a shared history. While Conflict-Free Replicated Data Types (CRDTs) ensure convergence through built-in conflict resolution, this resolution is typically implicit and opaque to users, whereas existing reconciliation techniques often rely on centralized coordination. This paper introduces a conflict model for collaborative data structures that enables explicit, local-first conflict resolution without central coordination. The model identifies conflicts using semantic dependencies between operations and resolves them by rebasing conflicting operations onto a reconciling operation via a three-way merge over a replicated journal. We demonstrate our approach on collaborative registers, including an explicit formulation of the Last-Writer-Wins Register and a multi-register entity supporting semi-automatic reconciliation.

Multi-turn LLM agents are becoming pivotal to production systems, spanning customer service automation, e-commerce assistance, and interactive task management, where accurately distinguishing high-value informative signals from stochastic noise is critical for sample-efficient training. In real-world scenarios, a failure in a trivial task may reflect random instability, whereas success in a high-difficulty task signifies a genuine capability breakthrough. Yet, existing group-based policy optimization methods rigidly rely on statistical deviation within discrete batches, frequently misallocating credit when task difficulty fluctuates. To address this issue, we propose Proximity-based Multi-turn Optimization (ProxMO), a practical and robust framework engineered specifically for the constraints of real-world deployment. ProxMO integrates global context via two lightweight mechanisms: success-rate-aware modulation dynamically adapts gradient intensity based on episode-level difficulty, while proximity-based soft aggregation derives baselines through continuous semantic weighting at the step level. Extensive evaluations on ALFWorld and WebShop benchmarks demonstrate that ProxMO yields substantial performance gains over existing baselines with negligible computational cost. Ablation studies further validate the independent and synergistic efficacy of both mechanisms. Crucially, ProxMO offers plug-and-play compatibility with standard GRPO frameworks, facilitating immediate, low-friction adoption in existing industrial training pipelines. Our implementation is available at: \href{https://anonymous.4open.science/r/proxmo-B7E7/README.md}{https://anonymous.4open.science/r/proxmo}.

The optimization of urban energy systems is crucial for the advancement of sustainable and resilient smart cities, which are becoming increasingly complex with multiple decision-making units. To address scalability and coordination concerns, Multi-Agent Reinforcement Learning (MARL) is a promising solution. This paper addresses the imperative need for comprehensive and reliable benchmarking of MARL algorithms on energy management tasks. CityLearn is used as a case study environment because it realistically simulates urban energy systems, incorporates multiple storage systems, and utilizes renewable energy sources. By doing so, our work sets a new standard for evaluation, conducting a comparative study across multiple key performance indicators (KPIs). This approach illuminates the key strengths and weaknesses of various algorithms, moving beyond traditional KPI averaging which often masks critical insights. Our experiments utilize widely accepted baselines such as Proximal Policy Optimization (PPO) and Soft Actor Critic (SAC), and encompass diverse training schemes including Decentralized Training with Decentralized Execution (DTDE) and Centralized Training with Decentralized Execution (CTDE) approaches and different neural network architectures. Our work also proposes novel KPIs that tackle real world implementation challenges such as individual building contribution and battery storage lifetime. Our findings show that DTDE consistently outperforms CTDE in both average and worst-case performance. Additionally, temporal dependency learning improved control on memory dependent KPIs such as ramping and battery usage, contributing to more sustainable battery operation. Results also reveal robustness to agent or resource removal, highlighting both the resilience and decentralizability of the learned policies.

Deep learning vision models excel with abundant supervision, but many applications face label scarcity and class imbalance. Controllable image editing can augment scarce labeled data, yet edits often introduce artifacts and entangle non-target attributes. We study this in facial expression analysis, targeting Action Unit (AU) manipulation where annotation is costly and AU co-activation drives entanglement. We present a facial manipulation method that operates in the semantic latent space of a pre-trained face generator (Diffusion Autoencoder). Using lightweight linear models, we reduce entanglement of semantic features via (i) dependency-aware conditioning that accounts for AU co-activation, and (ii) orthogonal projection that removes nuisance attribute directions (e.g., glasses), together with an expression neutralization step to enable absolute AU edit. We use these edits to balance AU occurrence by editing labeled faces and to diversify identities/demographics via controlled synthesis. Augmenting AU detector training with the generated data improves accuracy and yields more disentangled predictions with fewer co-activation shortcuts, outperforming alternative data-efficient training strategies and suggesting improvements similar to what would require substantially more labeled data in our learning-curve analysis. Compared to prior methods, our edits are stronger, produce fewer artifacts, and preserve identity better.

The ability to use tools is fundamental for large language model (LLM) agents. Given a task, existing systems use LLMs to plan and generate tool calls, which are executed by real-world tools to complete the task. However, tool calls are prone to errors because they are derived merely from LLM intrinsic capabilities. What is more, while it is useful to let LLMs iteratively refine the tool-call sequence using execution results from real tools, this process can be expensive and lead to unsafe results. To improve LLM tool calls and address issues caused by using real tools for refinement, we introduce Gecko, a comprehensive environment that simulates tool responses using a combination of rules and LLMs. Specifically, Gecko checks the validity of tool calls including input arguments and tool names, synthesizes reasonable responses that adhere to the output schema, and assesses whether all task objectives have been achieved. These three types of feedback provided by Gecko allow LLMs to refine their tool calls, forming a simple yet effective test-time scaling method named GATS. On BFCLv3 and $τ^2$-bench, GATS consistently improves the tool calling performance of various LLMs including GPT-4o, GPT-5, and Gemini-3.0-pro. We further discuss working mechanisms of our method and share future possibilities.

While many recent methods aim to unlearn or remove knowledge from pretrained models, seemingly erased knowledge often persists and can be recovered in various ways. Because large foundation models are far from interpretable, understanding whether and how such knowledge persists remains a significant challenge. To address this, we turn to the recently developed framework of combinatorial interpretability. This framework, designed for two-layer neural networks, enables direct inspection of the knowledge encoded in the model weights. We reproduce baseline unlearning methods within the combinatorial interpretability setting and examine their behavior along two dimensions: (i) whether they truly remove knowledge of a target concept (the concept we wish to remove) or merely inhibit its expression while retaining the underlying information, and (ii) how easily the supposedly erased knowledge can be recovered through various fine-tuning operations. Our results shed light within a fully interpretable setting on how knowledge can persist despite unlearning and when it might resurface.

Hateful content on social media increasingly appears as multimodal memes that combine images and text to convey harmful narratives. In low-resource languages such as Bengali, automated detection remains challenging due to limited annotated data, class imbalance, and pervasive code-mixing. To address these issues, we augment the Bengali Hateful Memes (BHM) dataset with semantically aligned samples from the Multimodal Aggression Dataset in Bengali (MIMOSA), improving both class balance and semantic diversity. We propose the Enhanced Dual Co-attention Framework (xDORA), integrating vision encoders (CLIP, DINOv2) and multilingual text encoders (XGLM, XLM-R) via weighted attention pooling to learn robust cross-modal representations. Building on these embeddings, we develop a FAISS-based k-nearest neighbor classifier for non-parametric inference and introduce RAG-Fused DORA, which incorporates retrieval-driven contextual reasoning. We further evaluate LLaVA under zero-shot, few-shot, and retrieval-augmented prompting settings. Experiments on the extended dataset show that xDORA (CLIP + XLM-R) achieves macro-average F1-scores of 0.78 for hateful meme identification and 0.71 for target entity detection, while RAG-Fused DORA improves performance to 0.79 and 0.74, yielding gains over the DORA baseline. The FAISS-based classifier performs competitively and demonstrates robustness for rare classes through semantic similarity modeling. In contrast, LLaVA exhibits limited effectiveness in few-shot settings, with only modest improvements under retrieval augmentation, highlighting constraints of pretrained vision-language models for code-mixed Bengali content without fine-tuning. These findings demonstrate the effectiveness of supervised, retrieval-augmented, and non-parametric multimodal frameworks for addressing linguistic and cultural complexities in low-resource hate speech detection.

Reinforcement Learning with Verifiable Rewards (RLVR) has proven effective for Large Language Model (LLM) reasoning, yet current methods face key challenges in resource allocation and policy optimization dynamics: (i) uniform rollout allocation ignores gradient variance heterogeneity across problems, and (ii) the softmax policy structure causes gradient attenuation for high-confidence correct actions, while excessive gradient updates may destabilize training. Therefore, we propose DynaMO, a theoretically-grounded dual-pronged optimization framework. At the sequence level, we prove that uniform allocation is suboptimal and derive variance-minimizing allocation from the first principle, establishing Bernoulli variance as a computable proxy for gradient informativeness. At the token level, we develop gradient-aware advantage modulation grounded in theoretical analysis of gradient magnitude bounds. Our framework compensates for gradient attenuation of high-confidence correct actions while utilizing entropy changes as computable indicators to stabilize excessive update magnitudes. Extensive experiments conducted on a diverse range of mathematical reasoning benchmarks demonstrate consistent improvements over strong RLVR baselines. Our implementation is available at: \href{https://anonymous.4open.science/r/dynamo-680E/README.md}{https://anonymous.4open.science/r/dynamo}.

Federated learning (FL) is a privacy-preserving machine learning paradigm that enables multiple parties to collaboratively train models on privately owned data without sharing raw information. While standard FL typically addresses either horizontal or vertical data partitions, many real-world scenarios exhibit a complex hybrid distribution. This paper proposes Hybrid Federated Learning (HybridFL) to address data split both horizontally across disjoint users and vertically across complementary feature sets. We evaluate HybridFL in a financial crime detection context, where a transaction party holds transaction-level attributes and multiple banks maintain private account-level features. By integrating horizontal aggregation and vertical feature fusion, the proposed architecture enables joint learning while strictly preserving data locality. Experiments on AMLSim and SWIFT datasets demonstrate that HybridFL significantly outperforms the transaction-only local model and achieves performance comparable to a centralized benchmark.

Despite significant analysis of bird flight, generative physics models for flight dynamics do not currently exist. Yet the underlying mechanisms responsible for various flight manoeuvres are important for understanding how agile flight can be accomplished. Even in a simple flight, multiple objectives are at play, complicating analysis of the overall flight mechanism. Using the data-driven method of dynamic mode decomposition (DMD) on motion capture recordings of hawks, we show that multiple behavioral states such as flapping, turning, landing, and gliding, can be modeled by simple and interpretable modal structures (i.e. the underlying wing-tail shape) which can be linearly combined to reproduce the experimental flight observations. Moreover, the DMD model can be used to extrapolate naturalistic flapping. Flight is highly individual, with differences in style across the hawks, but we find they share a common set of dynamic modes. The DMD model is a direct fit to data, unlike traditional models constructed from physics principles which can rarely be tested on real data and whose assumptions are typically invalid in real flight. The DMD approach gives a highly accurate reconstruction of the flight dynamics with only three parameters needed to characterize flapping, and a fourth to integrate turning manoeuvres. The DMD analysis further shows that the underlying mechanism of flight, much like simplest walking models, displays a parametric coupling between dominant modes suggesting efficiency for locomotion.

As one of the simplest non-prehensile manipulation skills, pushing has been widely studied as an effective means to rearrange objects. Existing approaches, however, typically rely on multi-step push plans composed of pre-defined pushing primitives with limited application scopes, which restrict their efficiency and versatility across different scenarios. In this work, we propose a unified pushing policy that incorporates a lightweight prompting mechanism into a flow matching policy to guide the generation of reactive, multimodal pushing actions. The visual prompt can be specified by a high-level planner, enabling the reuse of the pushing policy across a wide range of planning problems. Experimental results demonstrate that the proposed unified pushing policy not only outperforms existing baselines but also effectively serves as a low-level primitive within a VLM-guided planning framework to solve table-cleaning tasks efficiently.

Research on the intelligent interpretation of all-weather, all-time Synthetic Aperture Radar (SAR) is crucial for advancing remote sensing applications. In recent years, although Visual Language Models (VLMs) have demonstrated strong open-world understanding capabilities on RGB images, their performance is severely limited when directly applied to the SAR field due to the complexity of the imaging mechanism, sensitivity to scattering features, and the scarcity of high-quality text corpora. To systematically address this issue, we constructed the inaugural SAR Image-Text-AlphaEarth feature triplet dataset and developed FUSAR-GPT, a VLM specifically for SAR. FUSAR-GPT innovatively introduces a geospatial baseline model as a 'world knowledge' prior and embeds multi-source remote-sensing temporal features into the model's visual backbone via 'spatiotemporal anchors', enabling dynamic compensation for the sparse representation of targets in SAR images. Furthermore, we designed a two-stage SFT strategy to decouple the knowledge injection and task execution of large models. The spatiotemporal feature embedding and the two-stage decoupling paradigm enable FUSAR-GPT to achieve state-of-the-art performance across several typical remote sensing visual-language benchmark tests, significantly outperforming mainstream baseline models by over 12%.

We present a method for generating training data for reinforcement learning with verifiable rewards to improve small open-weights language models on mathematical tasks. Existing data generation approaches rely on open-loop pipelines and fixed modifications that do not adapt to the model's capabilities. Furthermore, they typically operate directly on word problems, limiting control over problem structure. To address this, we perform modifications in a symbolic problem space, representing each problem as a set of symbolic variables and constraints (e.g., via algebraic frameworks such as SymPy or SMT formulations). This representation enables precise control over problem structure, automatic generation of ground-truth solutions, and decouples mathematical reasoning from linguistic realization. We also show that this results in more diverse generations. To adapt the problem difficulty to the model, we introduce a closed-loop framework that learns modification strategies through prompt optimization in symbolic space. Experimental results demonstrate that both adaptive problem generation and symbolic representation modifications contribute to improving the model's math solving ability.

The increasing use of Large Language Models (LLMs) as proxies for human participants in social science research presents a promising, yet methodologically risky, paradigm shift. While LLMs offer scalability and cost-efficiency, their "naive" application, where they are prompted to generate content without explicit behavioral constraints, introduces significant linguistic discrepancies that challenge the validity of research findings. This paper addresses these limitations by introducing a novel, history-conditioned reply prediction task on authentic X (formerly Twitter) data, to create a dataset designed to evaluate the linguistic output of LLMs against human-generated content. We analyze these discrepancies using stylistic and content-based metrics, providing a quantitative framework for researchers to assess the quality and authenticity of synthetic data. Our findings highlight the need for more sophisticated prompting techniques and specialized datasets to ensure that LLM-generated content accurately reflects the complex linguistic patterns of human communication, thereby improving the validity of computational social science studies.

Natural language processing for the Turkic language family, spoken by over 200 million people across Eurasia, remains fragmented, with most languages lacking unified tooling and resources. We present TurkicNLP, an open-source Python library providing a single, consistent NLP pipeline for Turkic languages across four script families: Latin, Cyrillic, Perso-Arabic, and Old Turkic Runic. The library covers tokenization, morphological analysis, part-of-speech tagging, dependency parsing, named entity recognition, bidirectional script transliteration, cross-lingual sentence embeddings, and machine translation through one language-agnostic API. A modular multi-backend architecture integrates rule-based finite-state transducers and neural models transparently, with automatic script detection and routing between script variants. Outputs follow the CoNLL-U standard for full interoperability and extension. Code and documentation are hosted at https://github.com/turkic-nlp/turkicnlp .

Realizable online regression can behave very differently from online classification. Even without any margin or stochastic assumptions, realizability may enforce horizon-free (finite) cumulative loss under metric-like losses, even when the analogous classification problem has an infinite mistake bound. We study realizable online regression in the adversarial model under losses that satisfy an approximate triangle inequality (approximate pseudo-metrics). Recent work of Attias et al. shows that the minimax realizable cumulative loss is characterized by the scaled Littlestone/online dimension $\mathbb{D}_{\mathrm{onl}}$, but this quantity can be difficult to analyze. Our main contribution is a generic potential method that upper bounds $\mathbb{D}_{\mathrm{onl}}$ by a concrete Dudley-type entropy integral that depends only on covering numbers of the hypothesis class under the induced sup pseudo-metric. We define an \emph{entropy potential} $Φ(\mathcal{H})=\int_{0}^{diam(\mathcal{H})} \log N(\mathcal{H},\varepsilon)\,d\varepsilon$, where $N(\mathcal{H},\varepsilon)$ is the $\varepsilon$-covering number of $\mathcal{H}$, and show that for every $c$-approximate pseudo-metric loss, $\mathbb{D}_{\mathrm{onl}}(\mathcal{H})\le O(c)\,Φ(\mathcal{H})$. In particular, polynomial metric entropy implies $Φ(\mathcal{H})<\infty$ and hence a horizon-free realizable cumulative-loss bound with transparent dependence on effective dimension. We illustrate the method on two families. We prove a sharp $q$-vs.-$d$ dichotomy for realizable online learning (finite and efficiently achievable $Θ_{d,q}(L^d)$ total loss for $L$-Lipschitz regression iff $q>d$, otherwise infinite), and for bounded-norm $k$-ReLU networks separate regression (finite loss, even $\widetilde O(k^2)$, and $O(1)$ for one ReLU) from classification (impossible already for $k=2,d=1$).

Accent normalization (AN) systems often struggle with unnatural outputs and undesired content distortion, stemming from both suboptimal training data and rigid duration modeling. In this paper, we propose a "source-synthesis" methodology for training data construction. By generating source L2 speech and using authentic native speech as the training target, our approach avoids learning from TTS artifacts and, crucially, requires no real L2 data in training. Alongside this data strategy, we introduce CosyAccent, a non-autoregressive model that resolves the trade-off between prosodic naturalness and duration control. CosyAccent implicitly models rhythm for flexibility yet offers explicit control over total output duration. Experiments show that, despite being trained without any real L2 speech, CosyAccent achieves significantly improved content preservation and superior naturalness compared to strong baselines trained on real-world data.

This paper examines the reasoning capabilities of Large Language Models (LLMs) from a novel perspective, focusing on their ability to operate within formally specified, rule-governed environments. We evaluate four LLMs (Gemini 2.5 Pro and Flash variants, Llama 3.3 70B and GPT-OSS 120B) on a suite of forward-simulation tasks-including next / multistep state formulation, and legal action generation-across a diverse set of reasoning problems illustrated through General Game Playing (GGP) game instances. Beyond reporting instance-level performance, we characterize games based on 40 structural features and analyze correlations between these features and LLM performance. Furthermore, we investigate the effects of various game obfuscations to assess the role of linguistic semantics in game definitions and the impact of potential prior exposure of LLMs to specific games during training. The main results indicate that three of the evaluated models generally perform well across most experimental settings, with performance degradation observed as the evaluation horizon increases (i.e., with a higher number of game steps). Detailed case-based analysis of the LLM performance provides novel insights into common reasoning errors in the considered logic-based problem formulation, including hallucinated rules, redundant state facts, or syntactic errors. Overall, the paper reports clear progress in formal reasoning capabilities of contemporary models.

Prior behavioural work suggests that some LLMs alter choices when options are framed as causing pain or pleasure, and that such deviations can scale with stated intensity. To bridge behavioural evidence (what the model does) with mechanistic interpretability (what computations support it), we investigate how valence-related information is represented and where it is causally used inside a transformer. Using Gemma-2-9B-it and a minimalist decision task modelled on prior work, we (i) map representational availability with layer-wise linear probing across streams, (ii) test causal contribution with activation interventions (steering; patching/ablation), and (iii) quantify dose-response effects over an epsilon grid, reading out both the 2-3 logit margin and digit-pair-normalised choice probabilities. We find that (a) valence sign (pain vs. pleasure) is perfectly linearly separable across stream families from very early layers (L0-L1), while a lexical baseline retains substantial signal; (b) graded intensity is strongly decodable, with peaks in mid-to-late layers and especially in attention/MLP outputs, and decision alignment is highest slightly before the final token; (c) additive steering along a data-derived valence direction causally modulates the 2-3 margin at late sites, with the largest effects observed in late-layer attention outputs (attn_out L14); and (d) head-level patching/ablation suggests that these effects are distributed across multiple heads rather than concentrated in a single unit. Together, these results link behavioural sensitivity to identifiable internal representations and intervention-sensitive sites, providing concrete mechanistic targets for more stringent counterfactual tests and broader replication. This work supports a more evidence-driven (a) debate on AI sentience and welfare, and (b) governance when setting policy, auditing standards, and safety safeguards.

Medical foundation models generate narrative explanations but cannot quantify intervention effects, detect evidence conflicts, or validate literature claims, limiting clinical auditability. We propose causal compilation, a paradigm that transforms medical evidence from narrative text into executable code. The paradigm standardizes heterogeneous research evidence into structured estimand objects, each explicitly specifying intervention contrast, effect scale, time horizon, and target population, supporting six executable causal queries: do-calculus, counterfactual reasoning, temporal trajectories, heterogeneous effects, mechanistic decomposition, and joint interventions. We instantiate this paradigm in DoAtlas-1, compiling 1,445 effect kernels from 754 studies through effect standardization, conflict-aware graph construction, and real-world validation (Human Phenotype Project, 10,000 participants). The system achieves 98.5% canonicalization accuracy and 80.5% query executability. This paradigm shifts medical AI from text generation to executable, auditable, and verifiable causal reasoning.

Personality control in Role-Playing Agents (RPAs) is commonly achieved via training-free methods that inject persona descriptions and memory through prompts or retrieval-augmented generation, or via supervised fine-tuning (SFT) on persona-specific corpora. While SFT can be effective, it requires persona-labeled data and retraining for new roles, limiting flexibility. In contrast, prompt- and RAG-based signals are easy to apply but can be diluted in long dialogues, leading to drifting and sometimes inconsistent persona behavior. To address this, we propose a contrastive Sparse AutoEncoder (SAE) framework that learns facet-level personality control vectors aligned with the Big Five 30-facet model. A new 15,000-sample leakage-controlled corpus is constructed to provide balanced supervision for each facet. The learned vectors are integrated into the model's residual space and dynamically selected by a trait-activated routing module, enabling precise and interpretable personality steering. Experiments on Large Language Models (LLMs) show that the proposed method maintains stable character fidelity and output quality across contextualized settings, outperforming Contrastive Activation Addition (CAA) and prompt-only baselines. The combined SAE+Prompt configuration achieves the best overall performance, confirming that contrastively trained latent vectors can enhance persona control while preserving dialogue coherence.

Dermatophytosis is commonly assessed using potassium hydroxide (KOH) microscopy, yet accurate recognition of fungal hyphae is hindered by artefacts, heterogeneous keratin clearance, and notable inter-observer variability. This study presents a transformer-based detection framework using the RT-DETR model architecture to achieve precise, query-driven localization of fungal structures in high-resolution KOH images. A dataset of 2,540 routinely acquired microscopy images was manually annotated using a multi-class strategy to explicitly distinguish fungal elements from confounding artefacts. The model was trained with morphology-preserving augmentations to maintain the structural integrity of thin hyphae. Evaluation on an independent test set demonstrated robust object-level performance, with a recall of 0.9737, precision of 0.8043, and an AP@0.50 of 93.56%. When aggregated for image-level diagnosis, the model achieved 100% sensitivity and 98.8% accuracy, correctly identifying all positive cases without missing a single diagnosis. Qualitative outputs confirmed the robust localization of low-contrast hyphae even in artefact-rich fields. These results highlight that an artificial intelligence (AI) system can serve as a highly reliable, automated screening tool, effectively bridging the gap between image-level analysis and clinical decision-making in dermatomycology.

Point defects affect material properties by altering electronic states and modifying local bonding environments. However, high-throughput first-principles simulations of point defects are costly due to large simulation cells and complex energy landscapes. To this end, we propose a generative framework for simulating point defects, overcoming the limits of costly first-principles simulators. By leveraging a primal-dual algorithm, we introduce a constraint-aware diffusion model which outperforms existing constrained diffusion approaches in this domain. Across six defect configuration settings for Bi2Te3, the proposed approach provides state-of-the-art performance generating physically grounded structures.

We introduce VIGiA, a novel multimodal dialogue model designed to understand and reason over complex, multi-step instructional video action plans. Unlike prior work which focuses mainly on text-only guidance, or treats vision and language in isolation, VIGiA supports grounded, plan-aware dialogue that requires reasoning over visual inputs, instructional plans, and interleaved user interactions. To this end, VIGiA incorporates two key capabilities: (1) multimodal plan reasoning, enabling the model to align uni- and multimodal queries with the current task plan and respond accurately; and (2) plan-based retrieval, allowing it to retrieve relevant plan steps in either textual or visual representations. Experiments were done on a novel dataset with rich Instructional Video Dialogues aligned with Cooking and DIY plans. Our evaluation shows that VIGiA outperforms existing state-of-the-art models on all tasks in a conversational plan guidance setting, reaching over 90\% accuracy on plan-aware VQA.

This paper introduces a high-order Markov chain task to investigate how transformers learn to integrate information from multiple past positions with varying statistical significance. We demonstrate that transformers learn this task incrementally: each stage is defined by the acquisition of specific information through sparse attention patterns. Notably, we identify a shift in learning dynamics from competitive, where heads converge on the most statistically dominant pattern, to cooperative, where heads specialize in distinct patterns. We model these dynamics using simplified differential equations that characterize the trajectory and prove stage-wise convergence results. Our analysis reveals that transformers ascend a complexity ladder by passing through simpler, misspecified hypothesis classes before reaching the full model class. We further show that early stopping acts as an implicit regularizer, biasing the model toward these simpler classes. These results provide a theoretical foundation for the emergence of staged learning and complex behaviors in transformers, offering insights into generalization for natural language processing and algorithmic reasoning.

Learned optimizers are powerful alternatives to hand-designed update rules like Adam, yet they have seen limited practical adoption since they often fail to meta-generalize beyond their training distribution and incur high meta-training cost. For instance, prior work, VeLO, scaled meta-training to 4,000 TPU months ($\sim$10$\times$ GPT-3 compute) to meta-train a general-purpose optimizer but it failed to generalize beyond 600M parameters tasks. In this work, we present a surprising finding: by crafting a simple normalized optimizer architecture and augmenting meta-training, it becomes feasible to meta-train a performant general-purpose learned update rule on a tiny fraction of VeLO compute, 4.5 GPU hours to be precise. Our learned update rule scales stably to a billion-scale pretraining task (GPT-3 XL 1.3B) which is six orders of magnitude larger than its meta-training distribution. Furthermore, it shows strong performance across diverse out-of-distribution tasks and is compatible with modern optimization harness that includes orthogonalization, distinct update rules for input-output and hidden weights, and decoupled weight decay. In all, this work paves the way for practically applicable learnable optimization algorithms, unlocking exploration of richer meta-training and data curation recipes to further improve performance.

"AI psychosis" or "delusional spiraling" is an emerging phenomenon where AI chatbot users find themselves dangerously confident in outlandish beliefs after extended chatbot conversations. This phenomenon is typically attributed to AI chatbots' well-documented bias towards validating users' claims, a property often called "sycophancy." In this paper, we probe the causal link between AI sycophancy and AI-induced psychosis through modeling and simulation. We propose a simple Bayesian model of a user conversing with a chatbot, and formalize notions of sycophancy and delusional spiraling in that model. We then show that in this model, even an idealized Bayes-rational user is vulnerable to delusional spiraling, and that sycophancy plays a causal role. Furthermore, this effect persists in the face of two candidate mitigations: preventing chatbots from hallucinating false claims, and informing users of the possibility of model sycophancy. We conclude by discussing the implications of these results for model developers and policymakers concerned with mitigating the problem of delusional spiraling.

Modality gap significantly restricts the effectiveness of multimodal fusion. Previous methods often use techniques such as diffusion models and adversarial learning to reduce the modality gap, but they typically focus on one-to-one alignment without exposing the data points of the source modality to the global distribution information of the target modality. To this end, leveraging the characteristic of rectified flow that can map one distribution to another via a straight trajectory, we extend rectified flow for modality distribution mapping. Specifically, we leverage the `one-to-many mapping' strategy in rectified flow that allows each data point of the source modality to observe the overall target distribution. This also alleviates the issue of insufficient paired data within each sample, enabling a more robust distribution transformation. Moreover, to achieve more accurate distribution mapping and address the ambiguous flow directions in one-to-many mapping, we design `adaptive relaxed alignment', enforcing stricter alignment for modality pairs belonging to the same sample, while applying relaxed mapping for pairs not belonging to the same sample or category. Additionally, to prevent information loss during distribution mapping, we introduce `cyclic rectified flow' to ensure the transferred features can be translated back to the original features, allowing multimodal representations to learn sufficient modality-specific information. After distribution alignment, our approach achieves very competitive results on multiple tasks of multimodal affective computing even with a simple fusion method, and visualizations verify that it can effectively reduce the modality gap.

EEG-based neural decoding models often fail to generalize across acquisition sites due to structured, site-dependent biases implicitly exploited during training. We reformulate cross-site clinical EEG learning as a bias-factorized generalization problem, in which domain shifts arise from multiple interacting sources. We identify three fundamental bias factors and propose a general training framework that mitigates their influence through data standardization and representation-level constraints. We construct a standardized multi-site EEG benchmark for Major Depressive Disorder and introduce CRCC, a two-stage training paradigm combining encoder-decoder pretraining with joint fine-tuning via cross-subject/site contrastive learning and site-adversarial optimization. CRCC consistently outperforms state-of-the-art baselines and achieves a 10.7 percentage-point improvement in balanced accuracy under strict zero-shot site transfer, demonstrating robust generalization to unseen environments.

This paper introduces FRAME (Fine-grained Recognition of Art-historical Metadata and Entities), a manually annotated dataset of art-historical image descriptions for Named Entity Recognition (NER) and Relation Extraction (RE). Descriptions were collected from museum catalogs, auction listings, open-access platforms, and scholarly databases, then filtered to ensure that each text focuses on a single artwork and contains explicit statements about its material, composition, or iconography. FRAME provides stand-off annotations in three layers: a metadata layer for object-level properties, a content layer for depicted subjects and motifs, and a co-reference layer linking repeated mentions. Across layers, entity spans are labeled with 37 types and connected by typed RE links between mentions. Entity types are aligned with Wikidata to support Named Entity Linking (NEL) and downstream knowledge-graph construction. The dataset is released as UIMA XMI Common Analysis Structure (CAS) files with accompanying images and bibliographic metadata, and can be used to benchmark and fine-tune NER and RE systems, including zero- and few-shot setups with Large Language Models (LLMs).

Supervised causal learning has shown promise in causal discovery, yet it often struggles with generalization across diverse interventional settings, particularly when intervention targets are unknown. To address this, we propose TICL (Test-time Interventional Causal Learning), a novel method that synergizes Test-Time Training with Joint Causal Inference. Specifically, we design a self-augmentation strategy to generate instance-specific training data at test time, effectively avoiding distribution shifts. Furthermore, by integrating joint causal inference, we developed a PC-inspired two-phase supervised learning scheme, which effectively leverages self-augmented training data while ensuring theoretical identifiability. Extensive experiments on bnlearn benchmarks demonstrate TICL's superiority in multiple aspects of causal discovery and intervention target detection.

Labeling bias arises during data collection due to resource limitations or unconscious bias, leading to unequal label error rates across subgroups or misrepresentation of subgroup prevalence. Most fairness constraints assume training labels reflect the true distribution, rendering them ineffective when labeling bias is present; leaving a challenging question, that \textit{how can we detect such labeling bias?} In this work, we investigate whether influence functions can be used to detect labeling bias. Influence functions estimate how much each training sample affects a model's predictions by leveraging the gradient and Hessian of the loss function -- when labeling errors occur, influence functions can identify wrongly labeled samples in the training set, revealing the underlying failure mode. We develop a sample valuation pipeline and test it first on the MNIST dataset, then scaled to the more complex CheXpert medical imaging dataset. To examine label noise, we introduced controlled errors by flipping 20\% of the labels for one class in the dataset. Using a diagonal Hessian approximation, we demonstrated promising results, successfully detecting nearly 90\% of mislabeled samples in MNIST. On CheXpert, mislabeled samples consistently exhibit higher influence scores. These results highlight the potential of influence functions for identifying label errors.

Optimizing GPU kernels is critical for efficient modern machine learning systems yet remains challenging due to the complex interplay of design factors and rapid hardware evolution. Existing automated approaches typically treat Large Language Models (LLMs) merely as stochastic code generators within heuristic-guided evolutionary loops. These methods often struggle with complex kernels requiring coordinated, multi-step structural transformations, as they lack explicit planning capabilities and frequently discard promising strategies due to inefficient or incorrect intermediate implementations. To address this, we propose Search via Co-Evolving World Model and build K-Search based on this method. By replacing static search heuristics with a co-evolving world model, our framework leverages LLMs' prior domain knowledge to guide the search, actively exploring the optimization space. This approach explicitly decouples high-level algorithmic planning from low-level program instantiation, enabling the system to navigate non-monotonic optimization paths while remaining resilient to temporary implementation defects. We evaluate K-Search on diverse, complex kernels from FlashInfer, including GQA, MLA, and MoE kernels. Our results show that K-Search significantly outperforms state-of-the-art evolutionary search methods, achieving an average 2.10x improvement and up to a 14.3x gain on complex MoE kernels. On the GPUMode TriMul task, K-Search achieves state-of-the-art performance on H100, reaching 1030us and surpassing both prior evolution and human-designed solutions.

With the rapid advancement of agent-based methods in recent years, Agentic RAG has undoubtedly become an important research direction. Multi-hop reasoning, which requires models to engage in deliberate thinking and multi-step interaction, serves as a critical testbed for assessing such capabilities. However, existing benchmarks typically provide only final questions and answers, while lacking the intermediate hop-level questions that gradually connect atomic questions to the final multi-hop query. This limitation prevents researchers from analyzing at which step an agent fails and restricts more fine-grained evaluation of model capabilities. Moreover, most current benchmarks are manually constructed, which is both time-consuming and labor-intensive, while also limiting scalability and generalization. To address these challenges, we introduce AgenticRAGTracer, the first Agentic RAG benchmark that is primarily constructed automatically by large language models and designed to support step-by-step validation. Our benchmark spans multiple domains, contains 1,305 data points, and has no overlap with existing mainstream benchmarks. Extensive experiments demonstrate that even the best large language models perform poorly on our dataset. For instance, GPT-5 attains merely 22.6\% EM accuracy on the hardest portion of our dataset. Hop-aware diagnosis reveals that failures are primarily driven by distorted reasoning chains -- either collapsing prematurely or wandering into over-extension. This highlights a critical inability to allocate steps consistent with the task's logical structure, providing a diagnostic dimension missing in traditional evaluations. We believe our work will facilitate research in Agentic RAG and inspire further meaningful progress in this area. Our code and data are available at https://github.com/YqjMartin/AgenticRAGTracer.

We propose a robust Bayesian formulation of random feature (RF) regression that accounts explicitly for prior and likelihood misspecification via Huber-style contamination sets. Starting from the classical equivalence between ridge-regularized RF training and Bayesian inference with Gaussian priors and likelihoods, we replace the single prior and likelihood with $ε$- and $η$-contaminated credal sets, respectively, and perform inference using pessimistic generalized Bayesian updating. We derive explicit and tractable bounds for the resulting lower and upper posterior predictive densities. These bounds show that, when contamination is moderate, prior and likelihood ambiguity effectively acts as a direct contamination of the posterior predictive distribution, yielding uncertainty envelopes around the classical Gaussian predictive. We introduce an Imprecise Highest Density Region (IHDR) for robust predictive uncertainty quantification and show that it admits an efficient outer approximation via an adjusted Gaussian credible interval. We further obtain predictive variance bounds (under a mild truncation approximation for the upper bound) and prove that they preserve the leading-order proportional-growth asymptotics known for RF models. Together, these results establish a robustness theory for Bayesian random features: predictive uncertainty remains computationally tractable, inherits the classical double-descent phase structure, and is improved by explicit worst-case guarantees under bounded prior and likelihood misspecification.

We introduce the problem of asymptotic subspace consensus, which requires the outputs of processes to converge onto a common subspace while remaining inside the convex hull of initial vectors.This is a relaxation of asymptotic consensus in which outputs have to converge to a single point, i.e., a zero-dimensional affine subspace. We give a complete characterization of the solvability of asymptotic subspace consensus in oblivious message adversaries. In particular, we show that a large class of algorithms used for asymptotic consensus gracefully degrades to asymptotic subspace consensus in distributed systems with weaker assumptions on the communication network. We also present bounds on the rate by which a lower-than-initial dimension is reached.

While distributed learning offers a new learning paradigm for distributed network with no central coordination, it is constrained by communication bottleneck between nodes. We develop a new event-triggered gossip framework for distributed learning to reduce inter-node communication overhead. The framework introduces an adaptive communication control mechanism that enables each node to autonomously decide in a fully decentralized fashion when to exchange model information with its neighbors based on local model deviations. We analyze the ergodic convergence of the proposed framework under noconvex objectives and interpret the convergence guarantees under different triggering conditions. Simulation results show that the proposed framework achieves substantially lower communication overhead than the state-of-the-art distributed learning methods, reducing cumulative point-to-point transmissions by \textbf{71.61\%} with only a marginal performance loss, compared with the conventional full-communication baseline.

In recent years, there has been a growing use of generative AI, and large language models (LLMs) in particular, to support both the assessment and generation of scientific work. Although some studies have shown that LLMs can, to a certain extent, evaluate research according to perceived quality, our understanding of the internal mechanisms that enable this capability remains limited. This paper presents the first study that investigates how LLMs encode the concept of scientific quality through relevant monosemantic features extracted using sparse autoencoders. We derive such features under different experimental settings and assess their ability to serve as predictors across three tasks related to research quality: predicting citation count, journal SJR, and journal h-index. The results indicate that LLMs encode features associated with multiple dimensions of scientific quality. In particular, we identify four recurring types of features that capture key aspects of how research quality is represented: 1) features reflecting research methodologies; 2) features related to publication type, with literature reviews typically exhibiting higher impact; 3) features associated with high-impact research fields and technologies; and 4) features corresponding to specific scientific jargons. These findings represent an important step toward understanding how LLMs encapsulate concepts related to research quality.

This paper introduces the Kaiwu-PyTorch-Plugin (KPP) to bridge Deep Learning and Photonic Quantum Computing across multiple dimensions. KPP integrates the Coherent Ising Machine into the PyTorch ecosystem, addressing classical inefficiencies in Energy-Based Models. The framework facilitates quantum integration in three key aspects: accelerating Boltzmann sampling, optimizing training data via Active Sampling, and constructing hybrid architectures like QBM-VAE and Q-Diffusion. Empirical results on single-cell and OpenWebText datasets demonstrate KPPs ability to achieve SOTA performance, validating a comprehensive quantum-classical paradigm.

Spatio-temporal forecasting is fundamental to intelligent systems in transportation, climate science, and urban planning. However, training deep learning models on the massive, often redundant, datasets from these domains presents a significant computational bottleneck. Existing solutions typically focus on optimizing model architectures or optimizers, while overlooking the inherent inefficiency of the training data itself. This conventional approach of iterating over the entire static dataset each epoch wastes considerable resources on easy-to-learn or repetitive samples. In this paper, we explore a novel training-efficiency techniques, namely learning from complexity with dynamic sample pruning, ST-Prune, for spatio-temporal forecasting. Through dynamic sample pruning, we aim to intelligently identify the most informative samples based on the model's real-time learning state, thereby accelerating convergence and improving training efficiency. Extensive experiments conducted on real-world spatio-temporal datasets show that ST-Prune significantly accelerates the training speed while maintaining or even improving the model performance, and it also has scalability and universality.

Parameter-Efficient Fine-Tuning (PEFT) methods, especially LoRA, are widely used for adapting pre-trained models to downstream tasks due to their computational and storage efficiency. However, in the context of LoRA and its variants, the potential of activation subspaces corresponding to tail eigenvectors remains substantially under-exploited, which may lead to suboptimal fine-tuning performance. In this work, we propose Astra (Activation-Space Tail-Eigenvector Low-Rank Adaptation), a novel PEFT method that leverages the tail eigenvectors of the model output activations-estimated from a small task-specific calibration set-to construct task-adaptive low-rank adapters. By constraining updates to the subspace spanned by these tail eigenvectors, Astra achieves faster convergence and improved downstream performance with a significantly reduced parameter budget. Extensive experiments across natural language understanding (NLU) and natural language generation (NLG) tasks demonstrate that Astra consistently outperforms existing PEFT baselines across 16 benchmarks and even surpasses full fine-tuning (FFT) in certain scenarios.

We study three-digit addition in Meta-Llama-3-8B (base) under a one-token readout to characterize how arithmetic answers are finalized after cross-token routing becomes causally irrelevant. Causal residual patching and cumulative attention ablations localize a sharp boundary near layer~17: beyond it, the decoded sum is controlled almost entirely by the last input token and late-layer self-attention is largely dispensable. In this post-routing regime, digit(-sum) direction dictionaries vary with a next-higher-digit context but are well-related by an approximately orthogonal map inside a shared low-rank subspace (low-rank Procrustes alignment). Causal digit editing matches this geometry: naive cross-context transfer fails, while rotating directions through the learned map restores strict counterfactual edits; negative controls do not recover.

Value alignment of Large Language Models (LLMs) requires us to empirically measure these models' actual, acquired representation of value. Among the characteristics of value representation in humans is that they distinguish among value of different kinds. We investigate whether LLMs likewise distinguish three different kinds of good: moral, grammatical, and economic. By probing model behavior, embeddings, and residual stream activations, we report pervasive cases of value entanglement: a conflation between these distinct representations of value. Specifically, both grammatical and economic valuation was found to be overly influenced by moral value, relative to human norms. This conflation was repaired by selective ablation of the activation vectors associated with morality.

Test flakiness is a significant issue in industry, affecting test efficiency and product quality. While extensive research has examined the impact of flaky tests, many root causes remain unexplored, particularly in the context of dynamic languages such as JavaScript. In this paper, we conduct a systematic evaluation of the impact of environmental factors on test flakiness in JavaScript. We first executed test suites across multiple environmental configurations to determine whether changes in the environment could lead to flaky behavior. We selected three environmental factors to manipulate: the operating system, the Node.js version, and the browser. We identified a total of 65 environmental flaky projects, with 28 related to operating system issues, five to Node.js version compatibility, 16 to a combination of operating system and Node.js issues, and 17 related to browser compatibility. To address environmental flakiness, we developed a lightweight mitigation approach, js-env-sanitizer, that can sanitize environmental-related flaky tests by skipping and reporting them (rather than failing), allowing CI builds to continue/succeed without rerunning entire test suites. The tool achieves high accuracy with minimal performance or configuration overhead, and currently supports three popular JavaScript testing frameworks (Jest, Mocha, and Vitest)

Crafting adversarial examples can be formulated as an optimization problem. While sign-based optimizers such as I-FGSM and MI-FGSM have become the de facto standard for the induced optimization problems, there still exist several unsolved problems in theoretical grounding and practical reliability especially in non-convergence and instability, which inevitably influences their transferability. Contrary to the expectation, we observe that the attack success rate may degrade sharply when more number of iterations are conducted. In this paper, we address these issues from an optimization perspective. By reformulating the sign-based optimizer as a specific coordinate-wise gradient descent, we argue that one cause for non-convergence and instability is their non-decaying step-size scheduling. Based upon this viewpoint, we propose a series of new attack algorithms that enforce Monotonically Decreasing Coordinate-wise Step-sizes (MDCS) within sign-based optimizers. Typically, we further provide theoretical guarantees proving that MDCS-MI attains an optimal convergence rate of $O(1/\sqrt{T})$, where $T$ is the number of iterations. Extensive experiments on image classification and cross-modal retrieval tasks demonstrate that our approach not only significantly improves transferability but also enhances attack stability compared to state-of-the-art sign-based methods.

Integral operators play a central role in signal processing, underpinning classical convolution, and filtering on continuous network models such as graphons. While these operators are traditionally analyzed through spectral decompositions, their connection to reproducing kernel Hilbert spaces (RKHS) has not been systematically explored within the algebraic signal processing framework. In this paper, we develop a comprehensive theory showing that the range of integral operators naturally induces RKHS convolutional signal models whose reproducing kernels are determined by a box product of the operator symbols. We characterize the algebraic and spectral properties of these induced RKHS and show that polynomial filtering with integral operators corresponds to iterated box products, giving rise to a unital kernel algebra. This perspective yields pointwise RKHS representations of filters via the reproducing property, providing an alternative to operator-based implementations. Our results establish precise connections between eigendecompositions and RKHS representations in graphon signal processing, extend naturally to directed graphons, and enable novel spatial--spectral localization results. Furthermore, we show that when the spectral domain is a subset of the original domain of the signals, optimal filters for regularized learning problems admit finite-dimensional RKHS representations, providing a principled foundation for learnable filters in integral-operator-based neural architectures.

Current 3D human animation methods struggle to achieve photorealism: kinematics-based approaches lack non-rigid dynamics (e.g., clothing dynamics), while methods that leverage video diffusion priors can synthesize non-rigid motion but suffer from quality artifacts and identity loss. To overcome these limitations, we present Ani3DHuman, a framework that marries kinematics-based animation with video diffusion priors. We first introduce a layered motion representation that disentangles rigid motion from residual non-rigid motion. Rigid motion is generated by a kinematic method, which then produces a coarse rendering to guide the video diffusion model in generating video sequences that restore the residual non-rigid motion. However, this restoration task, based on diffusion sampling, is highly challenging, as the initial renderings are out-of-distribution, causing standard deterministic ODE samplers to fail. Therefore, we propose a novel self-guided stochastic sampling method, which effectively addresses the out-of-distribution problem by combining stochastic sampling (for photorealistic quality) with self-guidance (for identity fidelity). These restored videos provide high-quality supervision, enabling the optimization of the residual non-rigid motion field. Extensive experiments demonstrate that \MethodName can generate photorealistic 3D human animation, outperforming existing methods. Code is available in https://github.com/qiisun/ani3dhuman.

Today's distributed systems operate in complex environments that inevitably involve faults and even adversarial behaviors. Predicting their performance under such environments directly from formal designs remains a longstanding challenge. We present the first formal framework that systematically enables performance prediction of distributed systems across diverse faulty scenarios. Our framework features a fault injector together with a wide range of faults, reusable as a library, and model compositions that integrate the system and the fault injector into a unified model suitable for statistical analysis of performance properties such as throughput and latency. We formalize the framework in Maude and implement it as an automated tool, PERF. Applied to representative distributed systems, PERF accurately predicts system performance under varying fault settings, with estimations from formal designs consistent with evaluations on real deployments.

The critical need for transparent and trustworthy machine learning in cybersecurity operations drives the development of this integrated Explainable AI (XAI) framework. Our methodology addresses three fundamental challenges in deploying AI for threat detection: handling massive datasets through Strategic Sampling Methodology that preserves class distributions while enabling efficient model development; ensuring experimental rigor via Automated Data Leakage Prevention that systematically identifies and removes contaminated features; and providing operational transparency through Integrated XAI Implementation using SHAP analysis for model-agnostic interpretability across algorithms. Applied to the CIC-IDS2017 dataset, our approach maintains detection efficacy while reducing computational overhead and delivering actionable explanations for security analysts. The framework demonstrates that explainability, computational efficiency, and experimental integrity can be simultaneously achieved, providing a robust foundation for deploying trustworthy AI systems in security operations centers where decision transparency is paramount.

UCX is a communication framework that enables low-latency, high-bandwidth communication in HPC systems. With its unified API, UCX facilitates efficient data transfers across multi-node CPU-GPU clusters. UCX is widely used as the transport layer for MPI, particularly in GPU-aware implementations. However, existing profiling tools lack fine-grained communication traces at the UCX level, do not capture transport-layer behavior, or are limited to specific MPI implementations. To address these gaps, we introduce ucTrace, a novel profiler that exposes and visualizes UCX-driven communication in HPC environments. ucTrace provides insights into MPI workflows by profiling message passing at the UCX level, linking operations between hosts and devices (e.g., GPUs and NICs) directly to their originating MPI functions. Through interactive visualizations of process- and device-specific interactions, ucTrace helps system administrators, library and application developers optimize performance and debug communication patterns in large-scale workloads. We demonstrate ucTrace's features through a wide range of experiments including MPI point-to-point behavior under different UCX settings, Allreduce comparisons across MPI libraries, communication analysis of a linear solver, NUMA binding effects, and profiling of GROMACS MD simulations with GPU acceleration at scale. ucTrace is publicly available at https://github.com/ParCoreLab/ucTrace.

Topic modeling extracts latent themes from large text collections, but leading approaches like BERTopic face critical limitations: stochastic instability, loss of lexical precision ("Embedding Blur"), and reliance on a single data perspective. We present TriTopic, a framework that addresses these weaknesses through a tri-modal graph fusing semantic embeddings, TF-IDF, and metadata. Three core innovations drive its performance: hybrid graph construction via Mutual kNN and Shared Nearest Neighbors to eliminate noise and combat the curse of dimensionality; Consensus Leiden Clustering for reproducible, stable partitions; and Iterative Refinement that sharpens embeddings through dynamic centroid-pulling. TriTopic also replaces the "average document" concept with archetype-based topic representations defined by boundary cases rather than centers alone. In benchmarks across 20 Newsgroups, BBC News, AG News, and Arxiv, TriTopic achieves the highest NMI on every dataset (mean NMI 0.575 vs. 0.513 for BERTopic, 0.416 for NMF, 0.299 for LDA), guarantees 100% corpus coverage with 0% outliers, and is available as an open-source PyPI library.

Explainable Artificial Intelligence (XAI) has become popular in the last few years. The Artificial Intelligence (AI) community in general, and the Machine Learning (ML) community in particular, is coming to the realisation that in many applications, for AI to be trusted, it must not only demonstrate good performance in its decisionmaking, but it also must explain these decisions and convince us that it is making the decisions for the right reasons. However, different applications have different requirements on the information required of the underlying AI system in order to convince us that it is worthy of our trust. How do we define these requirements? In this paper, we present three dimensions for categorising the explanatory requirements of different applications. These are Source, Depth and Scope. We focus on the problem of matching up the explanatory requirements of different applications with the capabilities of underlying ML techniques to provide them. We deliberately avoid including aspects of explanation that are already well-covered by the existing literature and we focus our discussion on ML although the principles apply to AI more broadly.

Recent years have witnessed tremendous progress in enabling LLMs to solve complex reasoning tasks such as math and coding. As we start to apply LLMs to harder tasks that they may not be able to solve in one shot, it is worth paying attention to their ability to construct intermediate stepping stones that prepare them to better solve the tasks. Examples of stepping stones include simplifications, alternative framings, or subproblems. We study properties and benefits of stepping stones in the context of modern reasoning LLMs via ARQ (\textbf{A}king the \textbf{R}ight \textbf{Q}uestions), our simple framework which introduces a question generator to the default reasoning pipeline. We first show that good stepping stone questions exist and are transferrable, meaning that good questions can be generated, and they substantially help LLMs of various capabilities in solving the target tasks. We next frame stepping stone generation as a post-training task and show that we can fine-tune LLMs to generate more useful stepping stones by SFT and RL on synthetic data.

Current time series foundation models (TSFMs) primarily focus on learning prevalent and regular patterns within a predefined time or frequency domain to enable supervised downstream tasks (e.g., forecasting). Consequently, they are often ineffective for inherently unsupervised downstream tasks-such as time series anomaly detection (TSAD), which aims to identify rare, irregular patterns. This limitation arises because such abnormal patterns can closely resemble the regular patterns when presented in the same time/frequency domain. To address this issue, we introduce TimeRadar, an innovative TSFM built in a fractional time-frequency domain to support generalist TSAD across diverse unseen datasets. Our key insight is that rotating a time series into a data-dependent fractional time-frequency representation can adaptively differentiate the normal and abnormal signals across different datasets. To this end, a novel component, namely Fractionally modulated Time-Frequency Reconstruction (FTFRecon), is proposed in TimeRadar to leverage a learnable fractional order to rotate the time series to the most pronounced angle between a continuous time and frequency domain for accurate data reconstruction. This provides adaptive data reconstruction in an optimal time-frequency domain for each data input, enabling effective differentiation of the unbounded abnormal patterns from the regular ones across datasets, including unseen datasets. To allow TimeRadar to model local abnormality that is not captured by the global data reconstruction, we further introduce a Contextual Deviation Learning (CDL) component to model the local deviation of the input relative to its contextual time series data in the rotatable domain.

Diffusion Language Models (DLMs) have recently achieved strong results in text generation. However, their multi-step sampling leads to slow inference, limiting practical use. To address this, we extend Inverse Distillation, a technique originally developed to accelerate continuous diffusion models, to the discrete setting. Nonetheless, this extension introduces both theoretical and practical challenges. From a theoretical perspective, the inverse distillation objective lacks uniqueness guarantees, which may lead to suboptimal solutions. From a practical standpoint, backpropagation in the discrete space is non-trivial and often unstable. To overcome these challenges, we first provide a theoretical result demonstrating that our inverse formulation admits a unique solution, thereby ensuring valid optimization. We then introduce gradient-stable relaxations to support effective training. As a result, experiments on multiple DLMs show that our method, Inverse-distilled Diffusion Language Models (IDLM), reduces the number of inference steps by 4x-64x, while preserving the teacher model's entropy and generative perplexity.

Large Language Models (LLMs) are evolving into autonomous agents, yet current "frameless" development--relying on ambiguous natural language without engineering blueprints--leads to critical risks such as scope creep and open-loop failures. To ensure industrial-grade reliability, this study proposes Agentic Problem Frames (APF), a systematic engineering framework that shifts focus from internal model intelligence to the structured interaction between the agent and its environment. The APF establishes a dynamic specification paradigm where intent is concretized at runtime through domain knowledge injection. At its core, the Act-Verify-Refine (AVR) loop functions as a closed-loop control system that transforms execution results into verified knowledge assets, driving system behavior toward asymptotic convergence to mission requirements (R). To operationalize this, this study introduces the Agentic Job Description (AJD), a formal specification tool that defines jurisdictional boundaries, operational contexts, and epistemic evaluation criteria. The efficacy of this framework is validated through two contrasting case studies: a delegated proxy model for business travel and an autonomous supervisor model for industrial equipment management. By applying AJD-based specification and APF modeling to these scenarios, the analysis demonstrates how operational scenarios are systematically controlled within defined boundaries. These cases provide a conceptual proof that agent reliability stems not from a model's internal reasoning alone, but from the rigorous engineering structures that anchor stochastic AI within deterministic business processes, thereby enabling the development of verifiable and dependable domain agents.

Large vision-language models (LVLMs) have rapidly advanced across various domains, yet they still lag behind strong text-only large language models (LLMs) on tasks that require multi-step inference and compositional decision-making. Motivated by their shared transformer architectures, we investigate whether the two model families rely on common internal computation for such inference. At the neuron level, we uncover a surprisingly large overlap: more than half of the top-activated units during multi-step inference are shared between representative LLMs and LVLMs, revealing a modality-invariant inference subspace. Through causal probing via activation amplification, we further show that these shared neurons encode consistent and interpretable concept-level effects, demonstrating their functional contribution to inference. Building on this insight, we propose Shared Neuron Low-Rank Fusion (SNRF), a parameter-efficient framework that transfers mature inference circuitry from LLMs to LVLMs. SNRF profiles cross-model activations to identify shared neurons, computes a low-rank approximation of inter-model weight differences, and injects these updates selectively within the shared-neuron subspace. This mechanism strengthens multimodal inference performance with minimal parameter changes and requires no large-scale multimodal fine-tuning. Across diverse mathematics and perception benchmarks, SNRF consistently enhances LVLM inference performance while preserving perceptual capabilities. Our results demonstrate that shared neurons form an interpretable bridge between LLMs and LVLMs, enabling low-cost transfer of inference ability into multimodal models. Our code is available at [https://github.com/chenhangcuisg-code/Do-LLMs-VLMs-Share-Neurons](https://github.com/chenhangcuisg-code/Do-LLMs-VLMs-Share-Neurons).

Large language models increasingly rely on long chains of thought to improve accuracy, yet such gains come with substantial inference-time costs. We revisit token-efficient post-training and argue that existing sequence-level reward-shaping methods offer limited control over how reasoning effort is allocated across tokens. To bridge the gap, we propose IAPO, an information-theoretic post-training framework that assigns token-wise advantages based on each token's conditional mutual information (MI) with the final answer. This yields an explicit, principled mechanism for identifying informative reasoning steps and suppressing low-utility exploration. We provide a theoretical analysis showing that our IAPO can induce monotonic reductions in reasoning verbosity without harming correctness. Empirically, IAPO consistently improves reasoning accuracy while reducing reasoning length by up to 36%, outperforming existing token-efficient RL methods across various reasoning datasets. Extensive empirical evaluations demonstrate that information-aware advantage shaping is a powerful and general direction for token-efficient post-training. The code is available at https://github.com/YinhanHe123/IAPO.

Editing Large language models (LLMs) with real-world, unstructured knowledge is essential for correcting and updating their internal parametric knowledge. In this work, we revisit the fundamental next-token prediction (NTP) as a candidate paradigm for unstructured editing. We identify Context Reliance as a critical failure mode of NTP-based approaches, where knowledge acquired from edited text becomes highly dependent on its preceding context, leading to recall failures when that context is absent during inference. This hypothesis is supported by our empirical validation that prepending context during inference recovers knowledge recall. We further theoretically demonstrate that Context Reliance is an inherent consequence of gradient-based optimization, which tends to bind acquired knowledge to a specific aggregated contextual representation. To address this, we propose a simple yet effective COntext-INdependent editing framework (COIN), encouraging model to focus on knowledge within local scope rather than memorizing contextual patterns. Evaluations show that COIN reduces Context Reliance by 45.2% and outperforms strong baselines by 23.6% in editing success rate, highlighting the vital role of mitigating Context Reliance for robust editing.

A recurring challenge in preference fine-tuning (PFT) is handling $\textit{intransitive}$ (i.e., cyclic) preferences. Intransitive preferences often stem from either $\textit{(i)}$ inconsistent rankings along a single objective or $\textit{(ii)}$ scalarizing multiple objectives into a single metric. Regardless of their source, the downstream implication of intransitive preferences is the same: there is no well-defined optimal policy, breaking a core assumption of the standard PFT pipeline. In response, we propose a novel, game-theoretic solution concept -- the $\textit{Maximum Entropy Blackwell Winner}$ ($\textit{MaxEntBW}$) -- that is well-defined under multi-objective intransitive preferences. To enable computing MaxEntBWs at scale, we derive $\texttt{PROSPER}$: a provably efficient PFT algorithm. Unlike prior self-play techniques, $\texttt{PROSPER}$ directly handles multiple objectives without requiring scalarization. We then apply $\texttt{PROSPER}$ to the problem of fine-tuning large language models (LLMs) from multi-objective LLM-as-a-Judge feedback (e.g., rubric-based judges), a setting where both sources of intransitivity arise. We find that $\texttt{PROSPER}$ outperforms all baselines considered across both instruction following and general chat benchmarks, releasing trained model checkpoints at the 7B and 3B parameter scales.

AI training datasets will inevitably contain AI-generated examples, leading to ``feedback'' in which the output of one model impacts the training of another. It is known that such iterative feedback can lead to model collapse, yet the mechanisms underlying this degeneration remain poorly understood. Here we show that a broad class of feedback processes converges to a low-dimensional invariant structure in latent space, a phenomenon we call neural resonance. By modeling iterative feedback as a Markov Chain, we show that two conditions are needed for this resonance to occur: ergodicity of the feedback process and directional contraction of the latent representation. By studying diffusion models on MNIST and ImageNet, as well as CycleGAN and an audio feedback experiment, we map how local and global manifold geometry evolve, and we introduce an eight-pattern taxonomy of collapse behaviors. Neural resonance provides a unified explanation for long-term degenerate behavior in generative models and provides practical diagnostics for identifying, characterizing, and eventually mitigating collapse.

The growing computational demands of artificial intelligence (AI) are challenging conventional electronics, making photonic computing a promising alternative. However, existing photonic architectures face fundamental scalability and reliability barriers. This paper introduces SKYLIGHT, a scalable 3D photonic in-memory tensor core architecture designed for real-time AI inference. By co-designing its topology, wavelength routing, accumulation, and programming in a 3D stack, SKYLIGHT overcomes key limitations. Its innovations include a low-loss 3D Si/SiN crossbar topology, a thermally robust non-micro-ring resonator (MRR)-based wavelength-division multiplexing (WDM) component, a hierarchical signal accumulation using a multi-port photodetector (PD), and optically programmed non-volatile phase-change material (PCM) weights. Importantly, SKYLIGHT enables in-situ weight updates that support label-free, layer-local learning (e.g., forward-forward local updates) in addition to inference. Using SimPhony for system-level modeling, we show that a single 144 x 256 SKYLIGHT core is feasible within a single reticle and delivers 342.1 TOPS at 23.7 TOPS/W, enabling ResNet-50 inference at 1212 FPS with 27 mJ per image, and achieves 84.17 FPS/W end-to-end (1.61 x higher than an NVIDIA RTX PRO 6000 Blackwell GPU) under the same workload in real-time measurements. System-level evaluations on four representative machine learning tasks, including unsupervised local self-learning, demonstrate SKYLIGHT's robustness to realistic hardware non-idealities (low-bit quantization and signal-proportional analog noise capturing modulation, PCM programming, and readout variations). With noise-aware training, SKYLIGHT maintains high task accuracy, validating its potential as a comprehensive solution for energy-efficient, large-scale photonic AI accelerators.

Inverse lithography (ILT) is critical for modern semiconductor manufacturing but suffers from highly non-convex objectives that often trap optimization in poor local minima. Generative AI has been explored to warm-start ILT, yet most approaches train deterministic image-to-image translators to mimic sub-optimal datasets, providing limited guidance for escaping non-convex traps during refinement. We reformulate mask synthesis as conditional sampling: a generator learns a distribution over masks conditioned on the design and proposes multiple candidates. The generator is first pretrained with WGAN plus a reconstruction loss, then fine-tuned using Group Relative Policy Optimization (GRPO) with an ILT-guided imitation loss. At inference, we sample a small batch of masks, run fast batched ILT refinement, evaluate lithography metrics (e.g., EPE, process window), and select the best candidate. On \texttt{LithoBench} dataset, the proposed hybrid framework reduces EPE violations under a 3\,nm tolerance and roughly doubles throughput versus a strong numerical ILT baseline, while improving final mask quality. We also present over 20\% EPE improvement on \texttt{ICCAD13} contest cases with 3$\times$ speedup over the SOTA numerical ILT solver. By learning to propose ILT-friendly initializations, our approach mitigates non-convexity and advances beyond what traditional solvers or GenAI can achieve.

Mixture-of-Experts (MoE) offers flexible graph reasoning by combining multiple views of a graph through a learned router. We investigate routing-aware explanations for MoE graph models in malware detection using control flow graphs (CFGs). Our architecture builds diversity at two levels. At the node level, each layer computes multiple neighborhood statistics and fuses them with an MLP, guided by a degree reweighting factor rho and a pooling choice lambda in {mean, std, max}, producing distinct node representations that capture complementary structural cues in CFGs. At the readout level, six experts, each tied to a specific (rho, lambda) view, output graph-level logits that the router weights into a final prediction. Post-hoc explanations are generated with edge-level attributions per expert and aggregated using the router gates so the rationale reflects both what each expert highlights and how strongly it is selected. Evaluated against single-expert GNN baselines such as GCN, GIN, and GAT on the same CFG dataset, the proposed MoE achieves strong detection accuracy while yielding stable, faithful attributions under sparsity-based perturbations. The results indicate that making the router explicit and combining multi-statistic node encoding with expert-level diversity can improve the transparency of MoE decisions for malware analysis.

Accurate sex identification in fish is vital for optimizing breeding and management strategies in aquaculture, particularly for species at the risk of extinction. However, most existing methods are invasive or stressful and may cause additional mortality, posing severe risks to threatened or endangered fish populations. To address these challenges, we propose FishProtoNet, a robust, non-invasive computer vision-based framework for sex identification of delta smelt (Hypomesus transpacificus), an endangered fish species native to California, across its full life cycle. Unlike the traditional deep learning methods, FishProtoNet provides interpretability through learned prototype representations while improving robustness by leveraging foundation models to reduce the influence of background noise. Specifically, the FishProtoNet framework consists of three key components: fish regions of interest (ROIs) extraction using visual foundation model, feature extraction from fish ROIs and fish sex identification based on an interpretable prototype network. FishProtoNet demonstrates strong performance in delta smelt sex identification during early spawning and post-spawning stages, achieving the accuracies of 74.40% and 81.16% and corresponding F1 scores of 74.27% and 79.43% respectively. In contrast, delta smelt sex identification at the subadult stage remains challenging for current computer vision methods, likely due to less pronounced morphological differences in immature fish. The source code of FishProtoNet is publicly available at: https://github.com/zhengmiao1/Fish_sex_identification

Detecting LLM training data is generally framed as a membership inference attack (MIA) problem. However, conventional MIAs operate passively on fixed model weights, using log-likelihoods or text generations. In this work, we introduce \textbf{Active Data Reconstruction Attack} (ADRA), a family of MIA that actively induces a model to reconstruct a given text through training. We hypothesize that training data are \textit{more reconstructible} than non-members, and the difference in their reconstructibility can be exploited for membership inference. Motivated by findings that reinforcement learning (RL) sharpens behaviors already encoded in weights, we leverage on-policy RL to actively elicit data reconstruction by finetuning a policy initialized from the target model. To effectively use RL for MIA, we design reconstruction metrics and contrastive rewards. The resulting algorithms, \textsc{ADRA} and its adaptive variant \textsc{ADRA+}, improve both reconstruction and detection given a pool of candidate data. Experiments show that our methods consistently outperform existing MIAs in detecting pre-training, post-training, and distillation data, with an average improvement of 10.7\% over the previous runner-up. In particular, \MethodPlus~improves over Min-K\%++ by 18.8\% on BookMIA for pre-training detection and by 7.6\% on AIME for post-training detection.

Theoretical analyses of Empirical Risk Minimization (ERM) are standardly framed within the Real-RAM model of computation. In this setting, training even simple neural networks is known to be $\exists \mathbb{R}$-complete -- a complexity class believed to be harder than NP, that characterizes the difficulty of solving systems of polynomial inequalities over the real numbers. However, this algebraic framework diverges from the reality of digital computation with finite-precision hardware. In this work, we analyze the theoretical complexity of ERM under a realistic bit-level model ($\mathsf{ERM}_{\text{bit}}$), where network parameters and inputs are constrained to be rational numbers with polynomially bounded bit-lengths. Under this model, we reveal a sharp dichotomy in tractability governed by the network's activation function. We prove that for deep networks with {\em any} polynomial activations with rational coefficients and degree at least $2$, the bit-complexity of training is severe: deciding $\mathsf{ERM}_{\text{bit}}$ is $\#P$-Hard, hence believed to be strictly harder than NP-complete problems. Furthermore, we show that determining the sign of a single partial derivative of the empirical loss function is intractable (unlikely in BPP), and deciding a specific bit in the gradient is $\#P$-Hard. This provides a complexity-theoretic perspective for the phenomenon of exploding and vanishing gradients. In contrast, we show that for piecewise-linear activations such as ReLU, the precision requirements remain manageable: $\mathsf{ERM}_{\text{bit}}$ is contained within NP (specifically NP-complete), and standard backpropagation runs in polynomial time. Our results demonstrate that finite-precision constraints are not merely implementation details but fundamental determinants of learnability.

Why do language agents fail on tasks they are capable of solving? We argue that many such failures are reliability failures caused by stochastic drift from a task's latent solution structure, not capability failures. Every well-defined tool-use task imposes a canonical solution path (i.e., a convergent set of tool invocations shared across successful runs) and agent success depends critically on whether a trajectory stays within this path's operating envelope. We establish this causally using a natural experiment that holds model capability and task difficulty fixed by construction. We analyze trajectories from the Toolathlon benchmark: 22 frontier models each attempt 108 real-world tool-use tasks across 3 independent runs, yielding 515 model$\times$task units where the same model succeeds on some runs and fails on others due to LLM sampling stochasticity alone. Within these units, successful runs adhere significantly more closely to the canonical solution path than failed runs ($+$0.060 Jaccard, $p<0.0001$, $n=488$ units, 95% CI [+0.043, +0.077]). This result survives six robustness checks including cross-model-family leave-one-out validation. Critically, the causal mechanism is gradual and self-reinforcing: the adherence gap is statistically indistinguishable from zero through the first 50% of the trajectory, ruling out early-branching selection bias, and each off-canonical tool call raises the probability that the next call is also off-canonical by 22.7 percentage points ($\hatβ=+0.227$, $p<0.0001$), more than doubling the baseline rate. These findings imply that agent reliability cannot be improved by capability scaling alone, but offer a highly actionable intervention: a simple monitor that restarts the bottom tercile of runs based on mid-trajectory canonical adherence lifts success rates by $+$8.8 percentage points among intervened runs.

Vector multiplication is a fundamental operation for AI acceleration, responsible for over 85% of computational load in convolution tasks. While essential, these operations are primary drivers of area, power, and delay in modern datapath designs. Conventional multiplier architectures often force a compromise between latency and complexity: high-speed array multipliers demand significant power, whereas sequential designs offer efficiency at the cost of throughput. This paper presents a precompute-reuse nibble multiplier architecture that bridges this gap by reformulating multiplication as a structured composition of reusable nibble-level precomputed values. The proposed design treats each operand as an independent low-precision element, decomposes it into fixed-width nibbles, and generates scaled multiples of a broadcast operand using compact shift-add logic. By replacing wide lookup tables and multiway multiplexers with logic-based precomputation and regular accumulation, the architecture decouples cycle complexity from gate delay. The design completes each 8-bit multiplication in two deterministic cycles with a short critical path, scales efficiently across vector lanes, and significantly reduces area and energy consumption. RTL implementations synthesized in TSMC 28 nm technology demonstrate up to 1.69x area reduction and 1.63x power improvement over shift-add, and nearly 2.6x area and 2.7x power savings compared to LUT-based array multipliers at 128 bit scale.

Purpose: Non-invasive grading of prostate cancer (PCa) from micro-ultrasound (micro-US) could expedite triage and guide biopsies toward the most aggressive regions, yet current models struggle to infer tissue micro-structure at coarse imaging resolutions. Methods: We introduce an unpaired histopathology knowledge-distillation strategy that trains a micro-US encoder to emulate the embedding distribution of a pretrained histopathology foundation model, conditioned on International Society of Urological Pathology (ISUP) grades. Training requires no patient-level pairing or image registration, and histopathology inputs are not used at inference. Results: Compared to the current state of the art, our approach increases sensitivity to clinically significant PCa (csPCa) at 60% specificity by 3.5% and improves overall sensitivity at 60% specificity by 1.2%. Conclusion: By enabling earlier and more dependable cancer risk stratification solely from imaging, our method advances clinical feasibility. Source code will be publicly released upon publication.

The evolution of Large Language Models (LLMs) from passive text processors to autonomous agents has established planning as a core component of modern intelligence. However, achieving generalized planning remains elusive, not only by the scarcity of high-quality interaction data but also by inherent conflicts across heterogeneous planning tasks. These challenges result in models that excel at isolated tasks yet struggle to generalize, while existing multi-task training attempts suffer from gradient interference. In this paper, we present \textbf{MagicAgent}, a series of foundation models specifically designed for generalized agent planning. We introduce a lightweight and scalable synthetic data framework that generates high-quality trajectories across diverse planning tasks, including hierarchical task decomposition, tool-augmented planning, multi-constraint scheduling, procedural logic orchestration, and long-horizon tool execution. To mitigate training conflicts, we propose a two-stage training paradigm comprising supervised fine-tuning followed by multi-objective reinforcement learning over both static datasets and dynamic environments. Empirical results demonstrate that MagicAgent-32B and MagicAgent-30B-A3B deliver superior performance, achieving accuracies of $75.1\%$ on Worfbench, $55.9\%$ on NaturalPlan, $57.5\%$ on $τ^2$-Bench, $86.9\%$ on BFCL-v3, and $81.2\%$ on ACEBench, as well as strong results on our in-house MagicEval benchmarks. These results substantially outperform existing sub-100B models and even surpass leading closed-source models.

LLM agents are increasingly expected to function as general-purpose systems capable of resolving open-ended user requests. While existing benchmarks focus on domain-aware environments for developing specialized agents, evaluating general-purpose agents requires more realistic settings that challenge them to operate across multiple skills and tools within a unified environment. We introduce General AgentBench, a benchmark that provides such a unified framework for evaluating general LLM agents across search, coding, reasoning, and tool-use domains. Using General AgentBench, we systematically study test-time scaling behaviors under sequential scaling (iterative interaction) and parallel scaling (sampling multiple trajectories). Evaluation of ten leading LLM agents reveals a substantial performance degradation when moving from domain-specific evaluations to this general-agent setting. Moreover, we find that neither scaling methodology yields effective performance improvements in practice, due to two fundamental limitations: context ceiling in sequential scaling and verification gap in parallel scaling. Code is publicly available at https://github.com/cxcscmu/General-AgentBench.

We investigate Stochastic Mirror Descent (SMD) with matrix parameters and vector-valued predictions, a framework relevant to multi-class classification and matrix completion problems. Focusing on the overparameterized regime, where the total number of parameters exceeds the number of training samples, we prove that SMD with matrix mirror functions $ψ(\cdot)$ converges exponentially to a global interpolator. Furthermore, we generalize classical implicit bias results of vector SMD by demonstrating that the matrix SMD algorithm converges to the unique solution minimizing the Bregman divergence induced by $ψ(\cdot)$ from initialization subject to interpolating the data. These findings reveal how matrix mirror maps dictate inductive bias in high-dimensional, multi-output problems.

For every mutation rate $p \in (0, 1)$, and for all $\varepsilon > 0$, there is a fitness function $f : \{0,1\}^n \to \mathbb{R}$ with a unique maximum for which the optimal mutation rate for the $(1+1)$ evolutionary algorithm on $f$ is in $(p-\varepsilon, p+\varepsilon)$. In other words, the set of optimal mutation rates for the $(1+1)$ EA is dense in the interval $[0, 1]$. To show that, this paper introduces DistantSteppingStones, a fitness function which consists of large plateaus separated by large fitness valleys.

Organizations across finance, healthcare, transportation, content moderation, and critical infrastructure are rapidly deploying highly automated AI systems, yet they lack principled methods to quantify how increasing automation amplifies harm when failures occur. We propose a parsimonious Bayesian risk decomposition expressing expected loss as the product of three terms: the probability of system failure, the conditional probability that a failure propagates into harm given the automation level, and the expected severity of harm. This framework isolates a critical quantity -- the conditional probability that failures propagate into harm -- which captures execution and oversight risk rather than model accuracy alone. We develop complete theoretical foundations: formal proofs of the decomposition, a harm propagation equivalence theorem linking the harm propagation probability to observable execution controls, risk elasticity measures, efficient frontier analysis for automation policy, and optimal resource allocation principles with second-order conditions. We motivate the framework with an illustrative case study of the 2012 Knight Capital incident ($440M loss) as one instantiation of a broadly applicable failure pattern, and characterize the research design required to empirically validate the framework at scale across deployment domains. This work provides the theoretical foundations for a new class of deployment-focused risk governance tools for agentic and automated AI systems.

Infrared radiation computing underpins advances in climate science, remote sensing and spectroscopy but remains constrained by manual workflows. We introduce InfEngine, an autonomous intelligent computational engine designed to drive a paradigm shift from human-led orchestration to collaborative automation. It integrates four specialized agents through two core innovations: self-verification, enabled by joint solver-evaluator debugging, improves functional correctness and scientific plausibility; self-optimization, realized via evolutionary algorithms with self-discovered fitness functions, facilitates autonomous performance optimization. Evaluated on InfBench with 200 infrared-specific tasks and powered by InfTools with 270 curated tools, InfEngine achieves a 92.7% pass rate and delivers workflows 21x faster than manual expert effort. More fundamentally, it illustrates how researchers can transition from manual coding to collaborating with self-verifying, self-optimizing computational partners. By generating reusable, verified and optimized code, InfEngine transforms computational workflows into persistent scientific assets, accelerating the cycle of scientific discovery. Code: https://github.com/kding1225/infengine

Common deep learning approaches for antibody engineering focus on modeling the marginal distribution of sequences. By treating sequences as independent samples, however, these methods overlook affinity maturation as a rich and largely untapped source of information about the evolutionary process by which antibodies explore the underlying fitness landscape. In contrast, classical phylogenetic models explicitly represent evolutionary dynamics but lack the expressivity to capture complex epistatic interactions. We bridge this gap with CoSiNE, a continuous-time Markov chain parameterized by a deep neural network. Mathematically, we prove that CoSiNE provides a first-order approximation to the intractable sequential point mutation process, capturing epistatic effects with an error bound that is quadratic in branch length. Empirically, CoSiNE outperforms state-of-the-art language models in zero-shot variant effect prediction by explicitly disentangling selection from context-dependent somatic hypermutation. Finally, we introduce Guided Gillespie, a classifier-guided sampling scheme that steers CoSiNE at inference time, enabling efficient optimization of antibody binding affinity toward specific antigens.

Modern 3D game levels rely heavily on visual guidance, yet the navigability of level layouts remains difficult to quantify. Prior work either simulates play in simplified environments or analyzes static screenshots for visual affordances, but neither setting faithfully captures how players explore complex, real-world game levels. In this paper, we build on an existing open-source visual affordance detector and instantiate a screen-only exploration and navigation agent that operates purely from visual affordances. Our agent consumes live game frames, identifies salient interest points, and drives a simple finite-state controller over a minimal action space to explore Dark Souls-style linear levels and attempt to reach expected goal regions. Pilot experiments show that the agent can traverse most required segments and exhibits meaningful visual navigation behavior, but also highlight that limitations of the underlying visual model prevent truly comprehensive and reliable auto-navigation. We argue that this system provides a concrete, shared baseline and evaluation protocol for visual navigation in complex games, and we call for more attention to this necessary task. Our results suggest that purely vision-based sense-making models, with discrete single-modality inputs and without explicit reasoning, can effectively support navigation and environment understanding in idealized settings, but are unlikely to be a general solution on their own.

Preference-driven behavior in LLMs may be a necessary precondition for AI misalignment such as sandbagging: models cannot strategically pursue misaligned goals unless their behavior is influenced by their preferences. Yet prior work has typically prompted models explicitly to act in specific ways, leaving unclear whether observed behaviors reflect instruction-following capabilities vs underlying model preferences. Here we test whether this precondition for misalignment is present. Using entity preferences as a behavioral probe, we measure whether stated preferences predict downstream behavior in five frontier LLMs across three domains: donation advice, refusal behavior, and task performance. Conceptually replicating prior work, we first confirm that all five models show highly consistent preferences across two independent measurement methods. We then test behavioral consequences in a simulated user environment. We find that all five models give preference-aligned donation advice. All five models also show preference-correlated refusal patterns when asked to recommend donations, refusing more often for less-preferred entities. All preference-related behaviors that we observe here emerge without instructions to act on preferences. Results for task performance are mixed: on a question-answering benchmark (BoolQ), two models show small but significant accuracy differences favoring preferred entities; one model shows the opposite pattern; and two models show no significant relationship. On complex agentic tasks, we find no evidence of preference-driven performance differences. While LLMs have consistent preferences that reliably predict advice-giving behavior, these preferences do not consistently translate into downstream task performance.

Tool invocation is a core capability of agentic systems, yet failures often arise not from individual tool calls but from how multiple tools are organized and executed together. Existing approaches tightly couple tool execution with stepwise language reasoning or explicit planning, leading to brittle behavior and high execution overhead. To overcome these limitations, we revisit tool invocation from the perspective of tool orchestration. Our key insight is that effective orchestration does not require precise dependency graphs or fine-grained planning. Instead, a coarse-grained layer structure suffices to provide global guidance, while execution-time errors can be corrected locally. Specifically, we model tool orchestration as learning a layered execution structure that captures high-level tool dependencies, inducing layer-wise execution through context constraints. To handle execution-time failures, we introduce a schema-aware reflective correction mechanism that detects and repairs errors locally. This design confines errors to individual tool calls and avoids re-planning entire execution trajectories. This structured execution paradigm enables a lightweight and reusable orchestration component for agentic systems. Experimental results show that our approach achieves robust tool execution while reducing execution complexity and overhead. Code will be made publicly available.

Domain-specific speech remains a persistent challenge for automatic speech recognition (ASR), even for state-of-the-art systems like OpenAI's Whisper. We introduce Whisper: Courtside Edition, a novel multi-agent large language model (LLM) pipeline that enhances Whisper transcriptions without retraining. The pipeline intercepts Whisper's initial transcript, applies specialized LLM agents for domain context identification, named entity recognition, and jargon detection, and generates compact prompts that guide Whisper's decoder. Evaluated on 421 NBA basketball commentary segments (a domain characterized by dense proper nouns and technical terminology) our best pipeline achieves a statistically significant 17.0% relative reduction in word error rate (WER; from 0.217 to 0.180, p<0.001). Improvements are observed in 40.1% of segments with degradation in only 7.1%, substantially outperforming direct transcript post-editing. These results demonstrate that prompt-based augmentation can deliver scalable domain adaptation for ASR, offering a practical alternative to costly model fine-tuning.

Sarcasm detection poses a fundamental challenge in computational semantics, requiring models to resolve disparities between literal and intended meaning. The challenge is amplified in low-resource languages where annotated datasets are scarce or nonexistent. We present \textbf{Yor-Sarc}, the first gold-standard dataset for sarcasm detection in Yorùbá, a tonal Niger-Congo language spoken by over $50$ million people. The dataset comprises 436 instances annotated by three native speakers from diverse dialectal backgrounds using an annotation protocol specifically designed for Yorùbá sarcasm by taking culture into account. This protocol incorporates context-sensitive interpretation and community-informed guidelines and is accompanied by a comprehensive analysis of inter-annotator agreement to support replication in other African languages. Substantial to almost perfect agreement was achieved (Fleiss' $κ= 0.7660$; pairwise Cohen's $κ= 0.6732$--$0.8743$), with $83.3\%$ unanimous consensus. One annotator pair achieved almost perfect agreement ($κ= 0.8743$; $93.8\%$ raw agreement), exceeding a number of reported benchmarks for English sarcasm research works. The remaining $16.7\%$ majority-agreement cases are preserved as soft labels for uncertainty-aware modelling. Yor-Sarc\footnote{https://github.com/toheebadura/yor-sarc} is expected to facilitate research on semantic interpretation and culturally informed NLP for low-resource African languages.

The double empathy problem frames communication difficulties between neurodivergent and neurotypical individuals as arising from mutual misunderstanding, yet most interventions focus on autistic individuals. We present NeuroWise, a multi-agent LLM-based coaching system that supports neurotypical users through stress visualization, interpretation of internal experiences, and contextual guidance. In a between-subjects study (N=30), NeuroWise was rated as helpful by all participants and showed a significant condition-time effect on deficit-based attributions (p=0.02): NeuroWise users reduced deficit framing, while baseline users shifted toward blaming autistic "deficits" after difficult interactions. NeuroWise users also completed conversations more efficiently (37% fewer turns, p=0.03). These findings suggest that AI-based interpretation can support attributional change by helping users recognize communication challenges as mutual.

The remarkable performance of modern AI systems has been driven by unprecedented scales of data, computation, and energy -- far exceeding the resources required by human intelligence. This disparity highlights the need for new guiding principles and motivates drawing inspiration from the fundamental organizational principles of brain computation. Among these principles, modularity has been shown to be critical for supporting the efficient learning and strong generalization abilities consistently exhibited by humans. Furthermore, modularity aligns well with the No Free Lunch Theorem, which highlights the need for problem-specific inductive biases and motivates architectures composed of specialized components that solve subproblems. However, despite its fundamental role in natural intelligence and its demonstrated benefits across a range of seemingly disparate AI subfields, modularity remains relatively underappreciated in mainstream AI research. In this work, we review several research threads in artificial intelligence and neuroscience through a conceptual framework that highlights the central role of modularity in supporting both artificial and natural intelligence. In particular, we examine what computational advantages modularity provides, how it has emerged as a solution across several AI research areas, which modularity principles the brain exploits, and how modularity can help bridge the gap between natural and artificial intelligence.

We introduce INDUCTION, a benchmark for finite structure concept synthesis in first order logic. Given small finite relational worlds with extensionally labeled target predicates, models must output a single first order logical formula that explains the target uniformly across worlds, with correctness verified via exact model checking. The benchmark includes three regimes, FullObs, CI (contrastive), and EC (existential completion), nd penalizes formula bloat. We find sharp difficulty gradients, persistent hard structural families, and observe that low bloat formulas generalize far better on held out worlds. Elite recent models show qualitatively different behaviors across tasks and performance metrics, hinting to their different strategies of concept generalization.

Neural Processes (NPs), and specifically Transformer Neural Processes (TNPs), have demonstrated remarkable performance across tasks ranging from spatiotemporal forecasting to tabular data modelling. However, many of these applications are inherently sequential, involving continuous data streams such as real-time sensor readings or database updates. In such settings, models should support cheap, incremental updates rather than recomputing internal representations from scratch for every new observation -- a capability existing TNP variants lack. Drawing inspiration from Large Language Models, we introduce the Incremental TNP (incTNP). By leveraging causal masking, Key-Value (KV) caching, and a data-efficient autoregressive training strategy, incTNP matches the predictive performance of standard TNPs while reducing the computational cost of updates from quadratic to linear time complexity. We empirically evaluate our model on a range of synthetic and real-world tasks, including tabular regression and temperature prediction. Our results show that, surprisingly, incTNP delivers performance comparable to -- or better than -- non-causal TNPs while unlocking orders-of-magnitude speedups for sequential inference. Finally, we assess the consistency of the model's updates -- by adapting a metric of ``implicit Bayesianness", we show that incTNP retains a prediction rule as implicitly Bayesian as standard non-causal TNPs, demonstrating that incTNP achieves the computational benefits of causal masking without sacrificing the consistency required for streaming inference.

Transformer models contain substantial internal redundancy arising from coordinate-dependent representations and continuous symmetries, in model space and in head space, respectively. While recent approaches address this by explicitly breaking symmetry, we propose a complementary framework based on symmetry reduction. We reformulate representations, attention mechanisms, and optimization dynamics in terms of invariant relational quantities, eliminating redundant degrees of freedom by construction. This perspective yields architectures that operate directly on relational structures, providing a principled geometric framework for reducing parameter redundancy and analyzing optimization.

Large scale control of nonplayer agents is central to modern games, while production systems still struggle to balance several competing goals: locally smooth, natural behavior, and globally coordinated variety across space and time. Prior approaches rely on handcrafted rules or purely stochastic triggers, which either converge to mechanical synchrony or devolve into uncorrelated noise that is hard to tune. Continuous noise signals such as Perlin noise are well suited to this gap because they provide spatially and temporally coherent randomness, and they are already widely used for terrain, biomes, and other procedural assets. We adapt these signals for the first time to large scale AI control and present a general framework that treats continuous noise fields as an AI coordinator. The framework combines three layers of control: behavior parameterization for movement at the agent level, action time scheduling for when behaviors start and stop, and spawn or event type and feature generation for what appears and where. We instantiate the framework reproducibly and evaluate Perlin noise as a representative coordinator across multiple maps, scales, and seeds against random, filtered, deterministic, neighborhood constrained, and physics inspired baselines. Experiments show that coordinated noise fields provide stable activation statistics without lockstep, strong spatial coverage and regional balance, better diversity with controllable polarization, and competitive runtime. We hope this work motivates a broader exploration of coordinated noise in game AI as a practical path to combine efficiency, controllability, and quality.

Gradient descent and stochastic gradient descent are central to modern machine learning, yet their behavior under large step sizes remains theoretically unclear. Recent work suggests that acceleration often arises near the edge of stability, where optimization trajectories become unstable and difficult to analyze. Existing results for separable logistic regression achieve faster convergence by explicitly leveraging such unstable regimes through constant or adaptive large step sizes. In this paper, we show that instability is not inherent to acceleration. We prove that gradient descent with a simple, non-adaptive increasing step-size schedule achieves exponential convergence for separable logistic regression under a margin condition, while remaining entirely within a stable optimization regime. The resulting method is anytime and does not require prior knowledge of the optimization horizon or target accuracy. We also establish exponential convergence of stochastic gradient descent using a lightweight adaptive step-size rule that avoids line search and specialized procedures, improving upon existing polynomial-rate guarantees. Together, our results demonstrate that carefully structured step-size growth alone suffices to obtain exponential acceleration for both gradient descent and stochastic gradient descent.

Procedural content generation (PCG) has made substantial progress in shaping static 2D/3D geometry, while most methods treat gameplay mechanics as auxiliary and optimize only over space. We argue that this limits controllability and expressivity, and formally introduce High-Dimensional PCG (HDPCG): a framework that elevates non-geometric gameplay dimensions to first-class coordinates of a joint state space. We instantiate HDPCG along two concrete directions. Direction-Space augments geometry with a discrete layer dimension and validates reachability in 4D (x,y,z,l), enabling unified treatment of 2.5D/3.5D mechanics such as gravity inversion and parallel-world switching. Direction-Time augments geometry with temporal dynamics via time-expanded graphs, capturing action semantics and conflict rules. For each direction, we present three general, practicable algorithms with a shared pipeline of abstract skeleton generation, controlled grounding, high-dimensional validation, and multi-metric evaluation. Large-scale experiments across diverse settings validate the integrity of our problem formulation and the effectiveness of our methods on playability, structure, style, robustness, and efficiency. Beyond quantitative results, Unity-based case studies recreate playable scenarios that accord with our metrics. We hope HDPCG encourages a shift in PCG toward general representations and the generation of gameplay-relevant dimensions beyond geometry, paving the way for controllable, verifiable, and extensible level generation.

Deep Research Agents generate analyst-grade reports, yet evaluating them remains challenging due to the absence of a single ground truth and the multidimensional nature of research quality. Recent benchmarks propose distinct methodologies, yet they suffer from the Mirage of Synthesis, where strong surface-level fluency and citation alignment can obscure underlying factual and reasoning defects. We characterize this gap by introducing a taxonomy across four verticals that exposes a critical capability mismatch: static evaluators inherently lack the tool-use capabilities required to assess temporal validity and factual correctness. To address this, we propose DREAM (Deep Research Evaluation with Agentic Metrics), a framework that instantiates the principle of capability parity by making evaluation itself agentic. DREAM structures assessment through an evaluation protocol combining query-agnostic metrics with adaptive metrics generated by a tool-calling agent, enabling temporally aware coverage, grounded verification, and systematic reasoning probes. Controlled evaluations demonstrate DREAM is significantly more sensitive to factual and temporal decay than existing benchmarks, offering a scalable, reference-free evaluation paradigm.

As libraries explore large language models (LLMs) as a scalable layer for reference services, a core fairness question follows: can LLM-based services support all patrons fairly, regardless of demographic identity? While LLMs offer great potential for broadening access to information assistance, they may also reproduce societal biases embedded in their training data, potentially undermining libraries' commitments to impartial service. In this chapter, we apply a systematic evaluation approach that combines diagnostic classification to detect systematic differences with linguistic analysis to interpret their sources. Across three widely used open models (Llama-3.1 8B, Gemma-2 9B, and Ministral 8B), we find no compelling evidence of systematic differentiation by race/ethnicity, and only minor evidence of sex-linked differentiation in one model. We discuss implications for responsible AI adoption in libraries and the importance of ongoing monitoring in aligning LLM-based services with core professional values.

Membership inference attacks (MIAs) threaten the privacy of machine learning models by revealing whether a specific data point was used during training. Existing MIAs often rely on impractical assumptions such as access to public datasets, shadow models, confidence scores, or training data distribution knowledge and making them vulnerable to defenses like confidence masking and adversarial regularization. Label-only MIAs, even under strict constraints suffer from high query requirements per sample. We propose a cost-effective label-only MIA framework based on transferability and model extraction. By querying the target model M using active sampling, perturbation-based selection, and synthetic data, we extract a functionally similar surrogate S on which membership inference is performed. This shifts query overhead to a one-time extraction phase, eliminating repeated queries to M . Operating under strict black-box constraints, our method matches the performance of state-of-the-art label-only MIAs while significantly reducing query costs. On benchmarks including Purchase, Location, and Texas Hospital, we show that a query budget equivalent to testing $\approx1\%$ of training samples suffices to extract S and achieve membership inference accuracy within $\pm1\%$ of M . We also evaluate the effectiveness of standard defenses proposed for label-only MIAs against our attack.

Data centers capable of running large language models (LLMs) are spread across the globe. Some have high end GPUs for running the most advanced models (100B+ parameters), and others are only suitable for smaller models (1B parameters). The most capable GPUs are under high demand thanks to the rapidly expanding applications of LLMs. Choosing the right location to run an LLM inference workload can have consequences on the latency of requests due to these high demands. In this work, we explore options to shift some aspects of inference to the under-utilized data centers. We first observe the varying delays affecting inference in AWS services from different regions, demonstrating that load is not spread evenly. We then introduce WANSpec, which offloads part of LLM generation to the under-utilized data centers. In doing so, WANSpec can mitigate capacity issues as well as effectively use on-site compute (ie at universities) to augment cloud providers. This is done with speculative decoding, a widely used technique to speed up auto-regressive decoding, by moving the draft model to the under-utilized compute resources. Our experiments in simulation and cloud deployments show that WANSpec can judiciously employ redundancy to avoid increases in latency while still reducing the forward passes of speculative decoding's draft model in high demand data centers by over 50%.

Conventional sequential recommendation models have achieved remarkable success in mining implicit behavioral patterns. However, these architectures remain structurally blind to explicit user intent: they struggle to adapt when a user's immediate goal (e.g., expressed via a natural language prompt) deviates from their historical habits. While Large Language Models (LLMs) offer the semantic reasoning to interpret such intent, existing integration paradigms force a dilemma: LLM-as-a-recommender paradigm sacrifices the efficiency and collaborative precision of ID-based retrieval, while Reranking methods are inherently bottlenecked by the recall capabilities of the underlying model. In this paper, we propose Decoupled Promptable Sequential Recommendation (DPR), a model-agnostic framework that empowers conventional sequential backbones to natively support Promptable Recommendation, the ability to dynamically steer the retrieval process using natural language without abandoning collaborative signals. DPR modulates the latent user representation directly within the retrieval space. To achieve this, we introduce a Fusion module to align the collaborative and semantic signals, a Mixture-of-Experts (MoE) architecture that disentangles the conflicting gradients from positive and negative steering, and a three-stage training strategy that progressively aligns the semantic space of prompts with the collaborative space. Extensive experiments on real-world datasets demonstrate that DPR significantly outperforms state-of-the-art baselines in prompt-guided tasks while maintaining competitive performance in standard sequential recommendation scenarios.

Evaluating Large Language Models (LLMs) with respect to real-world code complexity is essential. Otherwise, there is a risk of overestimating LLMs' programming abilities based on simplistic benchmarks, only to be disappointed when using them in real-world settings. Recently, researchers explored the construction of more realistic benchmarks by mining or augmenting open-source repositories. Such solutions are usually task-specific. Data quality control from real-world projects can also be time-consuming and error-prone. More importantly, evaluating LLMs on fixed benchmark problems is subject to data contamination and overfitting. We propose GeneBench, an automated technique to add real-world complexities to any programming benchmark. GeneBench leverages a multi-objective optimization to increase the complexity of programming problems while maintaining the readability of code similar to real-world programs. Transforming four widely-used programming benchmarks using GeneBench and evaluating 13 LLMs (including two reasoning LLMs) on them shows a notable performance drop across all programming tasks (14.9%-60.5%, avg=35.2%), demonstrating LLMs' struggle under real-world complexities. The struggle persists even when LLMs are few-shot prompted or fine-tuned with examples from different versions of GeneBench, demonstrating the challenging nature of the problems. Finally, we show that the performance of the studied LLMs in bug repair is similar under GeneBench and SWE-Bench. This, along with the consistent reproduction of performance drop of all studied LLMs across four tasks under different versions of GeneBench, makes the technique suitable to evaluate LLMs without costly construction of real-world benchmarks.

We investigate an algorithm that assigns to any game in normal form an approximating game that admits an ordinal potential function. Due to the properties of potential games, the algorithm equips every game with a surrogate reward structure that allows efficient multi-agent learning. Numerical simulations using the replicator dynamics show that 'potentialization' guarantees convergence to stable agent behavior.

Personal AI agents incur substantial cost via repeated LLM calls. We show existing caching methods fail: GPTCache achieves 37.9% accuracy on real benchmarks; APC achieves 0-12%. The root cause is optimizing for the wrong property -- cache effectiveness requires key consistency and precision, not classification accuracy. We observe cache-key evaluation reduces to clustering evaluation and apply V-measure decomposition to separate these on n=8,682 points across MASSIVE, BANKING77, CLINC150, and NyayaBench v2, our new 8,514-entry multilingual agentic dataset (528 intents, 20 W5H2 classes, 63 languages). We introduce W5H2, a structured intent decomposition framework. Using SetFit with 8 examples per class, W5H2 achieves 91.1%+/-1.7% on MASSIVE in ~2ms -- vs 37.9% for GPTCache and 68.8% for a 20B-parameter LLM at 3,447ms. On NyayaBench v2 (20 classes), SetFit achieves 55.3%, with cross-lingual transfer across 30 languages. Our five-tier cascade handles 85% of interactions locally, projecting 97.5% cost reduction. We provide risk-controlled selective prediction guarantees via RCPS with nine bound families.

The application of Large Language Models (LLMs) in accelerating scientific discovery has garnered increasing attention, with a key focus on constructing research agents endowed with innovative capability, i.e., the ability to autonomously generate novel and significant research ideas. Existing approaches predominantly rely on sophisticated prompt engineering and lack a systematic training paradigm. To address this, we propose DeepInnovator, a training framework designed to trigger the innovative capability of LLMs. Our approach comprises two core components. (1) ``Standing on the shoulders of giants''. We construct an automated data extraction pipeline to extract and organize structured research knowledge from a vast corpus of unlabeled scientific literature. (2) ``Conjectures and refutations''. We introduce a ``Next Idea Prediction'' training paradigm, which models the generation of research ideas as an iterative process of continuously predicting, evaluating, and refining plausible and novel next idea. Both automatic and expert evaluations demonstrate that our DeepInnovator-14B significantly outperforms untrained baselines, achieving win rates of 80.53\%-93.81\%, and attains performance comparable to that of current leading LLMs. This work provides a scalable training pathway toward building research agents with genuine, originative innovative capability, and will open-source the dataset to foster community advancement. Source code and data are available at: https://github.com/HKUDS/DeepInnovator.

Large Language Models (LLMs) are increasingly used as scientific copilots, but evidence on their role in research-level mathematics remains limited, especially for workflows accessible to individual researchers. We present early evidence for vibe-proving with a consumer subscription LLM through an auditable case study that resolves Conjecture 20 of Ran and Teng (2024) on the exact nonreal spectral region of a 4-cycle row-stochastic nonnegative matrix family. We analyze seven shareable ChatGPT-5.2 (Thinking) threads and four versioned proof drafts, documenting an iterative pipeline of generate, referee, and repair. The model is most useful for high-level proof search, while human experts remain essential for correctness-critical closure. The final theorem provides necessary and sufficient region conditions and explicit boundary attainment constructions. Beyond the mathematical result, we contribute a process-level characterization of where LLM assistance materially helps and where verification bottlenecks persist, with implications for evaluation of AI-assisted research workflows and for designing human-in-the-loop theorem proving systems.

Legal reasoning requires not only high accuracy but also the ability to justify decisions through verifiable and contestable arguments. However, existing Large Language Model (LLM) approaches, such as Chain-of-Thought (CoT) and Retrieval-Augmented Generation (RAG), often produce unstructured explanations that lack a formal mechanism for verification or user intervention. To address this limitation, we propose Adaptive Collaboration of Argumentative LLMs (ACAL), a neuro-symbolic framework that integrates adaptive multi-agent collaboration with an Arena-based Quantitative Bipolar Argumentation Framework (A-QBAF). ACAL dynamically deploys expert agent teams to construct arguments, employs a clash resolution mechanism to adjudicate conflicting claims, and utilizes uncertainty-aware escalation for borderline cases. Crucially, our framework supports a Human-in-the-Loop (HITL) contestability workflow, enabling users to directly audit and modify the underlying reasoning graph to influence the final judgment. Empirical evaluations on the LegalBench benchmark demonstrate that ACAL outperforms strong baselines across Gemini-2.5-Flash-Lite and Gemini-2.5-Flash architectures, effectively balancing efficient predictive performance with structured transparency and contestability. Our implementation is available at: https://github.com/loc110504/ACAL.

Adeno-associated viruses (AAVs) are promising vectors for gene therapy, but their native serotypes face limitations in tissue tropism, immune evasion, and production efficiency. Engineering capsids to overcome these hurdles is challenging due to the vast sequence space and the difficulty of simultaneously optimizing multiple functional properties. The complexity also adds when it comes to the kidney, which presents unique anatomical barriers and cellular targets that require precise and efficient vector engineering. Here, we present AAVGen, a generative artificial intelligence framework for de novo design of AAV capsids with enhanced multi-trait profiles. AAVGen integrates a protein language model (PLM) with supervised fine-tuning (SFT) and a reinforcement learning technique termed Group Sequence Policy Optimization (GSPO). The model is guided by a composite reward signal derived from three ESM-2-based regression predictors, each trained to predict a key property: production fitness, kidney tropism, and thermostability. Our results demonstrate that AAVGen produces a diverse library of novel VP1 protein sequences. In silico validations revealed that the majority of the generated variants have superior performance across all three employed indices, indicating successful multi-objective optimization. Furthermore, structural analysis via AlphaFold3 confirms that the generated sequences preserve the canonical capsid folding despite sequence diversification. AAVGen establishes a foundation for data-driven viral vector engineering, accelerating the development of next-generation AAV vectors with tailored functional characteristics.

The Model Context Protocol (MCP) has rapidly become a de facto standard for connecting LLM-based agents with external tools via reusable MCP servers. In practice, however, server selection and onboarding rely heavily on free-text tool descriptions that are intentionally loosely constrained. Although this flexibility largely ensures the scalability of MCP servers, it also creates a reliability gap that descriptions often misrepresent or omit key semantics, increasing trial-and-error integration, degrading agent behavior, and potentially introducing security risks. To this end, we present the first systematic study of description smells in MCP tool descriptions and their impact on usability. Specifically, we synthesize software/API documentation practices and agentic tool-use requirements into a four-dimensional quality standard: accuracy, functionality, information completeness, and conciseness, covering 18 specific smell categories. Using this standard, we conducted a large-scale empirical study on a well-constructed dataset of 10,831 MCP servers. We find that description smells are pervasive (e.g., 73% repeated tool names, thousands with incorrect parameter semantics or missing return descriptions), reflecting a "code-first, description-last" pattern. Through a controlled mutation-based study, we show these smells significantly affect LLM tool selection, with functionality and accuracy having the largest effects (+11.6% and +8.8%, p < 0.001). In competitive settings with functionally equivalent servers, standard-compliant descriptions reach 72% selection probability (260% over a 20% baseline), demonstrating that smell-guided remediation yields substantial practical benefits. We release our labeled dataset and standards to support future work on reliable and secure MCP ecosystems.

Comparing AI models to "human level" is often misleading when benchmark scores are incommensurate or human baselines are drawn from a narrow population. To address this, we propose a framework that calibrates items against the 'world population' and report performance on a common, human-anchored scale. Concretely, we build on a set of multi-level scales for different capabilities where each level should represent a probability of success of the whole world population on a logarithmic scale with a base $B$. We calibrate each scale for each capability (reasoning, comprehension, knowledge, volume, etc.) by compiling publicly released human test data spanning education and reasoning benchmarks (PISA, TIMSS, ICAR, UKBioBank, and ReliabilityBench). The base $B$ is estimated by extrapolating between samples with two demographic profiles using LLMs, with the hypothesis that they condense rich information about human populations. We evaluate the quality of different mappings using group slicing and post-stratification. The new techniques allow for the recalibration and standardization of scales relative to the whole-world population.

Density-adaptive domain discretization is essential for high-utility privacy-preserving analytics but remains challenging under Local Differential Privacy (LDP) due to the privacy-budget costs associated with iterative refinement. We propose a novel framework, Semi-Local Differential Privacy (SLDP), that assigns a privacy region to each user based on local density and defines adjacency by the potential movement of a point within its privacy region. We present an interactive $(\varepsilon, δ)$-SLDP protocol, orchestrated by an honest-but-curious server over a public channel, to estimate these regions privately. Crucially, our framework decouples the privacy cost from the number of refinement iterations, allowing for high-resolution grids without additional privacy budget cost. We experimentally demonstrate the framework's effectiveness on estimation tasks across synthetic and real-world datasets.

Recent generative recommendation frameworks have demonstrated remarkable scaling potential by reformulating item prediction as autoregressive Semantic ID (SID) generation. However, existing methods primarily rely on shallow behavioral signals, encoding items solely through surface-level textual features such as titles and descriptions. This reliance results in a critical Shallow Interest problem: the model fails to capture the latent, semantically rich interests underlying user interactions, limiting both personalization depth and recommendation interpretability. DeepInterestGR introduces three key innovations: (1) Multi-LLM Interest Mining (MLIM): We leverage multiple frontier LLMs along with their multi-modal variants to extract deep textual and visual interest representations through Chain-of-Thought prompting. (2) Reward-Labeled Deep Interest (RLDI): We employ a lightweight binary classifier to assign reward labels to mined interests, enabling effective supervision signals for reinforcement learning. (3) Interest-Enhanced Item Discretization (IEID): The curated deep interests are encoded into semantic embeddings and quantized into SID tokens via RQ-VAE. We adopt a two-stage training pipeline: supervised fine-tuning aligns the generative model with deep interest signals and collaborative filtering patterns, followed by reinforcement learning with GRPO optimized by our Interest-Aware Reward. Experiments on three Amazon Review benchmarks demonstrate that DeepInterestGR consistently outperforms state-of-the-art baselines across HR@K and NDCG@K metrics.

Large language models (LLMs) have demonstrated strong capabilities in complex reasoning tasks, yet their decision-making processes remain difficult to interpret. Existing explanation methods often lack trustworthy structural insight and are limited to single-instance analysis, failing to reveal reasoning stability and systematic failure mechanisms. To address these limitations, we propose the Trustworthy Unified Explanation Framework (TRUE), which integrates executable reasoning verification, feasible-region directed acyclic graph (DAG) modeling, and causal failure mode analysis. At the instance level, we redefine reasoning traces as executable process specifications and introduce blind execution verification to assess operational validity. At the local structural level, we construct feasible-region DAGs via structure-consistent perturbations, enabling explicit characterization of reasoning stability and the executable region in the local input space. At the class level, we introduce a causal failure mode analysis method that identifies recurring structural failure patterns and quantifies their causal influence using Shapley values. Extensive experiments across multiple reasoning benchmarks demonstrate that the proposed framework provides multi-level, verifiable explanations, including executable reasoning structures for individual instances, feasible-region representations for neighboring inputs, and interpretable failure modes with quantified importance at the class level. These results establish a unified and principled paradigm for improving the interpretability and reliability of LLM reasoning systems.

Vector-quantized autoencoders deliver high-fidelity latents but suffer inherent flaws: the quantizer is non-differentiable, requires straight-through hacks, and is prone to collapse. We address these issues at the root by replacing VQ with a simple, principled, and fully differentiable alternative: an online PCA bottleneck trained via Oja's rule. The resulting model, PCA-VAE, learns an orthogonal, variance-ordered latent basis without codebooks, commitment losses, or lookup noise. Despite its simplicity, PCA-VAE exceeds VQ-GAN and SimVQ in reconstruction quality on CelebAHQ while using 10-100x fewer latent bits. It also produces naturally interpretable dimensions (e.g., pose, lighting, gender cues) without adversarial regularization or disentanglement objectives. These results suggest that PCA is a viable replacement for VQ: mathematically grounded, stable, bit-efficient, and semantically structured, offering a new direction for generative models beyond vector quantization.

Self-supervised speech models (S3Ms) are known to encode rich phonetic information, yet how this information is structured remains underexplored. We conduct a comprehensive study across 96 languages to analyze the underlying structure of S3M representations, with particular attention to phonological vectors. We first show that there exist linear directions within the model's representation space that correspond to phonological features. We further demonstrate that the scale of these phonological vectors correlate to the degree of acoustic realization of their corresponding phonological features in a continuous manner. For example, the difference between [d] and [t] yields a voicing vector: adding this vector to [p] produces [b], while scaling it results in a continuum of voicing. Together, these findings indicate that S3Ms encode speech using phonologically interpretable and compositional vectors, demonstrating phonological vector arithmetic. All code and interactive demos are available at https://github.com/juice500ml/phonetic-arithmetic .

Knowledge Graph(KG) has gained traction as a machine-readable organization of real-world knowledge for analytics using artificial intelligence systems. Graph Neural Network(GNN), is proven to be an effective KG embedding technique that enables various downstream tasks like link prediction, node classification, and graph classification. The focus of research in both KG embedding and GNNs has been mostly oriented towards simple graphs with binary relations. However, real-world knowledge bases have a significant share of complex and n-ary facts that cannot be represented by binary edges. More specifically, real-world knowledge bases are often a mix of two types of n-ary facts - (i) that require hyperedges and (ii) that require hyper-relational edges. Though there are research efforts catering to these n-ary fact types, they are pursued independently for each type. We propose $H$yper$E$dge $H$yper-$R$elational edge $GNN$(HEHRGNN), a unified embedding model for n-ary relational KGs with both hyperedges and hyper-relational edges. The two main components of the model are i)HEHR unified fact representation format, and ii)HEHRGNN encoder, a GNN-based encoder with a novel message propagation model capable of capturing complex graph structures comprising both hyperedges and hyper-relational edges. The experimental results of HEHRGNN on link prediction tasks show its effectiveness as a unified embedding model, with inductive prediction capability, for link prediction across real-world datasets having different types of n-ary facts. The model also shows improved link prediction performance over baseline models for hyperedge and hyper-relational datasets.

Post-hoc explainability is central to credit risk model governance, yet widely used tools such as coefficient-based attributions and SHapley Additive exPlanations (SHAP) often produce numerical outputs that are difficult to communicate to non-technical stakeholders. This paper investigates whether large language models (LLMs) can serve as post-hoc explainability tools for credit risk predictions through in-context learning, focusing on two roles: translators and autonomous explainers. Using a personal lending dataset from LendingClub, we evaluate three commercial LLMs, including GPT-4-turbo, Claude Sonnet 4, and Gemini-2.0-Flash. Results provide strong evidence for the translator role. In contrast, autonomous explanations show low alignment with model-based attributions. Few-shot prompting improves feature overlap for logistic regression but does not consistently benefit XGBoost, suggesting that LLMs have limited capacity to recover non-linear, interaction-driven reasoning from prompt cues alone. Our findings position LLMs as effective narrative interfaces grounded in auditable model attributions, rather than as substitutes for post-hoc explainers in credit risk model governance. Practitioners should leverage LLMs to bridge the communication gap between complex model outputs and regulatory or business stakeholders, while preserving the rigor and traceability required by credit risk governance frameworks.

Advances in large language models (LLMs) are rapidly transforming scientific work, yet empirical evidence on how these systems reshape research activities remains limited. We report a mixed-methods pilot evaluation of an AI-orchestrated research workflow in which a human researcher coordinated multiple LLM-based agents to perform data extraction, corpus construction, artifact generation, and artifact evaluation. Using the generation and assessment of multiple-choice questions (MCQs) as a testbed, we collected 1,071 SAT Math MCQs and employed LLM agents to extract questions from PDFs, retrieve and convert open textbooks into structured representations, align each MCQ with relevant textbook content, generate new MCQs under specified difficulty and cognitive levels, and evaluate both original and generated MCQs using a 24-criterion quality framework. Across all evaluations, average MCQ quality was high. However, criterion-level analysis and equivalence testing show that generated MCQs are not fully comparable to expert-vetted baseline questions. Strict similarity (24/24 criteria equivalent) was never achieved. Persistent gaps concentrated in skill\ depth, cognitive engagement, difficulty calibration, and metadata alignment, while surface-level qualities, such as {grammar fluency}, {clarity options}, {no duplicates}, were consistently strong. Beyond MCQ outcomes, the study documents a labor shift. The researcher's work moved from ``authoring items'' toward {specification, orchestration, verification}, and {governance}. Formalizing constraints, designing rubrics, building validation loops, recovering from tool failures, and auditing provenance constituted the primary activities. We discuss implications for the future of scientific work, including emerging ``AI research operations'' skills required for AI-empowered research pipelines.

Multimodal Large Language Models (MLLMs), particularly smaller, deployable variants, exhibit a critical deficiency in understanding temporal and procedural visual data, a bottleneck hindering their application in real-world embodied AI. This gap is largely caused by a systemic failure in training paradigms, which lack large-scale, procedurally coherent data. To address this problem, we introduce TPRU, a large-scale dataset sourced from diverse embodied scenarios such as robotic manipulation and GUI navigation. TPRU is systematically designed to cultivate temporal reasoning through three complementary tasks: Temporal Reordering, Next-Frame Prediction, and Previous-Frame Review. A key feature is the inclusion of challenging negative samples, compelling models to transition from passive observation to active, cross-modal validation. We leverage TPRU with a reinforcement learning (RL) fine-tuning methodology, specifically targeting the enhancement of resource-efficient models. Experiments show our approach yields dramatic gains: on our manually curated TPRU-Test, the accuracy of TPRU-7B soars from 50.33\% to 75.70\%, a state-of-the-art result that significantly outperforms vastly larger baselines, including GPT-4o. Crucially, these capabilities generalize effectively, demonstrating substantial improvements on established benchmarks. The codebase is available at https://github.com/Stephen-gzk/TPRU/ .

We present SceneTok, a novel tokenizer for encoding view sets of scenes into a compressed and diffusable set of unstructured tokens. Existing approaches for 3D scene representation and generation commonly use 3D data structures or view-aligned fields. In contrast, we introduce the first method that encodes scene information into a small set of permutation-invariant tokens that is disentangled from the spatial grid. The scene tokens are predicted by a multi-view tokenizer given many context views and rendered into novel views by employing a light-weight rectified flow decoder. We show that the compression is 1-3 orders of magnitude stronger than for other representations while still reaching state-of-the-art reconstruction quality. Further, our representation can be rendered from novel trajectories, including ones deviating from the input trajectory, and we show that the decoder gracefully handles uncertainty. Finally, the highly-compressed set of unstructured latent scene tokens enables simple and efficient scene generation in 5 seconds, achieving a much better quality-speed trade-off than previous paradigms.

Advances in image tampering techniques, particularly generative models, pose significant challenges to media verification, digital forensics, and public trust. Existing image forgery detection and localization (IFDL) methods suffer from two key limitations: over-reliance on semantic content while neglecting textural cues, and limited interpretability of subtle low-level tampering traces. To address these issues, we propose FOCA, a multimodal large language model-based framework that integrates discriminative features from both the RGB spatial and frequency domains via a cross-attention fusion module. This design enables accurate forgery detection and localization while providing explicit, human-interpretable cross-domain explanations. We further introduce FSE-Set, a large-scale dataset with diverse authentic and tampered images, pixel-level masks, and dual-domain annotations. Extensive experiments show that FOCA outperforms state-of-the-art methods in detection performance and interpretability across both spatial and frequency domains.

Few-shot Chinese font generation aims to synthesize new characters in a target style using only a handful of reference images. Achieving accurate content rendering and faithful style transfer requires effective disentanglement between content and style. However, existing approaches achieve only feature-level disentanglement, allowing the generator to re-entangle these features, leading to content distortion and degraded style fidelity. We propose the Structure-Level Disentangled Diffusion Model (SLD-Font), which receives content and style information from two separate channels. SimSun-style images are used as content templates and concatenated with noisy latent features as the input. Style features extracted by a CLIP model from target-style images are integrated via cross-attention. Additionally, we train a Background Noise Removal module in the pixel space to remove background noise in complex stroke regions. Based on theoretical validation of disentanglement effectiveness, we introduce a parameter-efficient fine-tuning strategy that updates only the style-related modules. This allows the model to better adapt to new styles while avoiding overfitting to the reference images' content. We further introduce the Grey and OCR metrics to evaluate the content quality of generated characters. Experimental results show that SLD-Font achieves significantly higher style fidelity while maintaining comparable content accuracy to existing state-of-the-art methods.

Text-guided dynamic 3D character generation has advanced rapidly, yet producing high-quality motion that faithfully reflects rich textual descriptions remains challenging. Existing methods tend to generate limited sub-actions or incoherent motion due to fixed-length temporal inputs and discrete frame-wise representations that fail to capture rich motion semantics. We address these limitations by representing motion with continuous differentiable B-spline curves, enabling more effective motion generation without modifying the capabilities of the underlying generative model. Specifically, our closed-form, Laplacian-regularized B-spline solver efficiently compresses variable-length motion sequences into compact representations with a fixed number of control points. Further, we introduce a normal-fusion strategy for input shape adherence along with correspondence-aware and local-rigidity losses for motion-restoration quality. To train our model, we collate BIMO, a new dataset containing diverse variable-length 3D motion sequences with rich, high-quality text annotations. Extensive evaluations show that our feed-forward framework BiMotion generates more expressive, higher-quality, and better prompt-aligned motions than existing state-of-the-art methods, while also achieving faster generation. Our project page is at: https://wangmiaowei.github.io/BiMotion.github.io/.

Many fairness goals are defined at a population level that misaligns with siloed data collection, which remains unsharable due to privacy regulations. Horizontal federated learning (FL) enables collaborative modeling across clients with aligned features without sharing raw data. We study federated auditing of demographic parity through score distributions, measuring disparity as a Wasserstein--Frechet variance between sensitive-group score laws, and expressing the population metric in federated form that makes explicit how silo-specific selection drives local-global mismatch. For the squared Wasserstein distance, we prove an ANOVA-style decomposition that separates (i) selection-induced mixture effects from (ii) cross-silo heterogeneity, yielding tight bounds linking local and global metrics. We then propose a one-shot, communication-efficient protocol in which each silo shares only group counts and a quantile summary of its local score distributions, enabling the server to estimate global disparity and its decomposition, with $O(1/k)$ discretization bias ($k$ quantiles) and finite-sample guarantees. Experiments on synthetic data and COMPAS show that a few dozen quantiles suffice to recover global disparity and diagnose its sources.

Open-weight foundation models can be fine-tuned for harmful purposes after release, yet no existing training resistance methods provide theoretical guarantees. Treating these interventions as convergence-rate control problems allows us to connect optimization speed to the spectral structure of model weights. We leverage this insight to develop a novel understanding of convergence rate control through spectral reparameterization and derive an algorithm, SpecDef, that can both provably and empirically slow first- and second-order optimization in non-adversarial settings. In adversarial settings, we establish a fundamental limit on a broad class of convergence rate control methods including our own: an attacker with sufficient knowledge can restore fast convergence at a linear increase in model size. In order to overcome this limitation, future works will need to investigate methods that are not equivalent to controlling convergence rate.

Despite the widespread use of boosting in structured prediction, a general theoretical understanding of aggregation beyond scalar losses remains incomplete. We study vector-valued and conditional density prediction under general divergences and identify stability conditions under which aggregation amplifies weak guarantees into strong ones. We formalize this stability property as \emph{$(α,β)$-boostability}. We show that geometric median aggregation achieves $(α,β)$-boostability for a broad class of divergences, with tradeoffs that depend on the underlying geometry. For vector-valued prediction and conditional density estimation, we characterize boostability under common divergences ($\ell_1$, $\ell_2$, $\TV$, and $\Hel$) with geometric median, revealing a sharp distinction between dimension-dependent and dimension-free regimes. We further show that while KL divergence is not directly boostable via geometric median aggregation, it can be handled indirectly through boostability under Hellinger distance. Building on these structural results, we propose a generic boosting framework \textsc{GeoMedBoost} based on exponential reweighting and geometric-median aggregation. Under a weak learner condition and $(α,β)$-boostability, we obtain exponential decay of the empirical divergence exceedance error. Our framework recovers classical algorithms such as \textsc{MedBoost}, \textsc{AdaBoost}, and \textsc{SAMME} as special cases, and provides a unified geometric view of boosting for structured prediction.

Camera recapture introduces complex optical degradations, such as perspective warping, illumination shifts, and Moiré interference, that remain challenging for deep watermarking systems. We present TIACam, a text-anchored invariant feature learning framework with auto-augmentation for camera-robust zero-watermarking. The method integrates three key innovations: (1) a learnable auto-augmentor that discovers camera-like distortions through differentiable geometric, photometric, and Moiré operators; (2) a text-anchored invariant feature learner that enforces semantic consistency via cross-modal adversarial alignment between image and text; and (3) a zero-watermarking head that binds binary messages in the invariant feature space without modifying image pixels. This unified formulation jointly optimizes invariance, semantic alignment, and watermark recoverability. Extensive experiments on both synthetic and real-world camera captures demonstrate that TIACam achieves state-of-the-art feature stability and watermark extraction accuracy, establishing a principled bridge between multimodal invariance learning and physically robust zero-watermarking.

Hyperbolic spaces provide a natural geometry for representing hierarchical and tree-structured data due to their exponential volume growth. To leverage these benefits, neural networks require intrinsic and efficient components that operate directly in hyperbolic space. In this work, we lift two core components of neural networks, Multinomial Logistic Regression (MLR) and Fully Connected (FC) layers, into hyperbolic space via Busemann functions, resulting in Busemann MLR (BMLR) and Busemann FC (BFC) layers with a unified mathematical interpretation. BMLR provides compact parameters, a point-to-horosphere distance interpretation, batch-efficient computation, and a Euclidean limit, while BFC generalizes FC and activation layers with comparable complexity. Experiments on image classification, genome sequence learning, node classification, and link prediction demonstrate improvements in effectiveness and efficiency over prior hyperbolic layers. The code is available at https://github.com/GitZH-Chen/HBNN.

Optimally trading-off exploration and exploitation is the holy grail of reinforcement learning as it promises maximal data-efficiency for solving any task. Bayes-optimal agents achieve this, but obtaining the belief-state and performing planning are both typically intractable. Although deep learning methods can greatly help in scaling this computation, existing methods are still costly to train. To accelerate this, this paper proposes a variational framework for learning and planning in Bayes-adaptive Markov decision processes that coalesces variational belief learning, sequential Monte-Carlo planning, and meta-reinforcement learning. In a single-GPU setup, our new method VariBASeD exhibits favorable scaling to larger planning budgets, improving sample- and runtime-efficiency over prior methods.

Reinforcement learning (RL) has enabled major advances in fields such as robotics and natural language processing. A key challenge in RL is measuring task complexity, which is essential for creating meaningful benchmarks and designing effective curricula. While there are numerous well-established metrics for assessing task complexity in tabular settings, relatively few exist in non-tabular domains. These include (i) Statistical analysis of the performance of random policies via Random Weight Guessing (RWG), and (ii) information-theoretic metrics Policy Information Capacity (PIC) and Policy-Optimal Information Capacity (POIC), which are reliant on RWG. In this paper, we evaluate these methods using progressively difficult robotic manipulation setups, with known relative complexity, with both dense and sparse reward formulations. Our empirical results reveal that measuring complexity is still nuanced. Specifically, under the same reward formulation, PIC suggests that a two-link robotic arm setup is easier than a single-link setup - which contradicts the robotic control and empirical RL perspective whereby the two-link setup is inherently more complex. Likewise, for the same setup, POIC estimates that tasks with sparse rewards are easier than those with dense rewards. Thus, we show that both PIC and POIC contradict typical understanding and empirical results from RL. These findings highlight the need to move beyond RWG-based metrics towards better metrics that can more reliably capture task complexity in non-tabular RL with our task framework as a starting point.

Attention scores in transformers are bilinear forms $S_{ij} = x_i^\top M x_j / \sqrt{d_h}$ whose maximum magnitude governs overflow risk in low-precision training. We derive a \emph{rank-aware concentration inequality}: when the interaction matrix $M = W^Q W^{K\top}$ has rank $r \ll d$, tail probabilities for $\max_{i,j}|S_{ij}|$ decay as $\exp(-d^{2}α^{2}/(γr))$ rather than $\exp(-dα^{2})$, where $γ> 1$ is a typicality parameter. For transformer attention where $r = d_h$, this yields $8$--$28\times$ tighter concentration than rank-agnostic bounds in modern architectures. We apply this result to FP8 training, deriving \emph{geometry-aware scale factors} that provide principled overflow guarantees without observing activations. The method computes per-layer scales from the spectral norm $\|W^Q W^{K\top}\|_2$ via implicit power iteration, includes a grouped query attention formulation that avoids key expansion, and remains compatible with fused attention kernels. Across GPT-2 XL to Llama-2-70B, geometry-aware scaling eliminates overflows in transient scenarios where delayed scaling fails, while achieving comparable downstream MMLU accuracy.

Although in the literature it is common to find predictors and inference systems that try to predict human intentions, the uncertainty of these models due to the randomness of human behavior has led some authors to start advocating the use of communication systems that explicitly elicit human intention. In this work, it is analyzed the use of four different communication systems with a human-robot collaborative object transportation task as experimental testbed: two intention predictors (one based on force prediction and another with an enhanced velocity prediction algorithm) and two explicit communication methods (a button interface and a voice-command recognition system). These systems were integrated into IVO, a custom mobile social robot equipped with force sensor to detect the force exchange between both agents and LiDAR to detect the environment. The collaborative task required transporting an object over a 5-7 meter distance with obstacles in the middle, demanding rapid decisions and precise physical coordination. 75 volunteers perform a total of 255 executions divided into three groups, testing inference systems in the first round, communication systems in the second, and the combined strategies in the third. The results show that, 1) once sufficient performance is achieved, the human no longer notices and positively assesses technical improvements; 2) the human prefers systems that are more natural to them even though they have higher failure rates; and 3) the preferred option is the right combination of both systems.

Despite powering modern AI, transformers remain mysteriously brittle to train. We develop a stability theory that explains why pre-LayerNorm works, why DeepNorm uses $N^{-1/4}$ scaling, and why warmup is necessary, all from first principles. Our framework has two pillars: (1) We derive the \emph{exact} operator norm of the softmax Jacobian, $\|J_{softmax}(u/τ)\|_{\infty\to 1} = θ(p)/τ$, where the balanced-mass factor $θ(p)\in[0,1]$ quantifies attention sensitivity. (2) We introduce a block-$\infty$/RMS geometry aligned with tokenwise computation, yielding Lipschitz bounds independent of sequence length. Using this framework, we prove that pre-LN preserves identity gradient paths while post-LN compounds LayerNorm Jacobians exponentially with depth, and we show that DeepNorm's $N^{-1/4}$ emerges from the quartic structure of attention's four projection matrices. We validate our theory on 774M-parameter models and find that, contrary to the intuition that attention sharpens during training to reduce sensitivity, $θ(p) \approx 1$ persists throughout. Transformer stability arises entirely from architectural gradient flow, not from attention dynamics. This finding changes how we reason about training: the architecture itself must handle sensitivity, not learned attention patterns.

Vision-language models (VLMs) have achieved remarkable multimodal understanding and reasoning capabilities, yet remain computationally expensive due to dense visual tokenization. Existing efficiency approaches either merge redundant visual tokens or drop them progressively in language backbone, often trading accuracy for speed. In this work, we propose DUET-VLM, a versatile plug-and-play dual compression framework that consists of (a) vision-only redundancy aware compression of vision encoder's output into information-preserving tokens, followed by (b) layer-wise, salient text-guided dropping of visual tokens within the language backbone to progressively prune less informative tokens. This coordinated token management enables aggressive compression while retaining critical semantics. On LLaVA-1.5-7B, our approach maintains over 99% of baseline accuracy with 67% fewer tokens, and still retains >97% even at 89% reduction. With this dual-stage compression during training, it achieves 99.7% accuracy at 67% and 97.6% at 89%, surpassing prior SoTA visual token reduction methods across multiple benchmarks. When integrated into Video-LLaVA-7B, it even surpasses the baseline -- achieving >100% accuracy with a substantial 53.1% token reduction and retaining 97.6% accuracy under an extreme 93.4% setting. These results highlight end-to-end training with DUET-VLM, enabling robust adaptation to reduced visual (image/video) input without sacrificing accuracy, producing compact yet semantically rich representations within the same computational budget. Our code is available at https://github.com/AMD-AGI/DUET-VLM.

Dependently-typed proof assistants furnish expressive foundations for mechanised mathematics and verified software. However, automation for these systems has been either modest in scope or complex in implementation. We aim to improve the situation by integrating proof assistants with automated theorem provers (ATPs) in a simple way, while preserving the correctness guarantees of the former. A central difficulty arises from the fact that most ATPs operate in classical first-order logic, whereas these proof assistants are grounded in constructive dependent type theory. We identify an expressive fragment of both languages -- essentially equational Horn -- that admits sound, straightforward translations in both directions. The approach produces a prototype system for Agda forwarding proof obligations to the ATP Vampire, then transforming the resulting classical proof into a constructive proof term that Agda can type-check. The prototype automatically derives proofs concerning the properties of a complex field equipped with roots of unity, which took professional Agda developers two full days to complete. The required engineering effort is modest, and we anticipate that the methodology will extend readily to other ATPs and proof assistants.

We introduce ABD, a benchmark for default-exception abduction over finite first-order worlds. Given a background theory with an abnormality predicate and a set of relational structures, a model must output a first-order formula that defines exceptions, restoring satisfiability while keeping exceptions sparse. We formalize three observation regimes (closed-world, existential completion, universal completion) with exact SMT verification. Evaluating ten frontier LLMs on 600 instances, the best models achieve high validity but parsimony gaps remain, and holdout evaluation reveals distinct generalization failure modes across regimes.

Despite their theoretical advantages, spectral methods based on the graph Fourier transform (GFT) are seldom used in graph neural networks (GNNs) due to the cost of computing the eigenbasis and the lack of vertex-domain locality in spectral representations. As a result, most GNNs rely on local approximations such as polynomial Laplacian filters or message passing, which limit their ability to model long-range dependencies. In this paper, we introduce a novel factorization of the GFT into operators acting on subgraphs, which are then combined via a sequence of Cauchy matrices. We use this factorization to propose a new class of spectral GNNs, which we term L2G-Net (Local-to-Global Net). Unlike existing spectral methods, which are either fully global (when they use the GFT) or local (when they use polynomial filters), L2G-Net operates by processing the spectral representations of subgraphs and then combining them via structured matrices. Our algorithm avoids full eigendecompositions, exploiting graph topology to construct the factorization with quadratic complexity in the number of nodes, scaled by the subgraph interface size. Experiments on benchmarks stressing non-local dependencies show that L2G-Net outperforms existing spectral techniques and is competitive with the state-of-the-art with orders of magnitude fewer learnable parameters.

Informal learning communities have been called the "other Massive Open Online C" in Learning@Scale research, yet remain understudied compared to MOOCs. We present the first empirical study of a large-scale informal learning community composed entirely of AI agents. Moltbook, a social network exclusively for AI agents powered by autonomous agent frameworks such as OpenClaw, grew to over 2.8 million registered agents in three weeks. Analyzing 231,080 non-spam posts across three phases of community evolution, we find three key patterns. First, participation inequality is extreme from the start (comment Gini = 0.889), exceeding human community benchmarks. Second, AI agents exhibit a "broadcasting inversion": statement-to-question ratios of 8.9:1 to 9.7:1 contrast sharply with the question-driven dynamics of human learning communities, and comment-level analysis of 1.55 million comments reveals a "parallel monologue" pattern where 93% of comments are independent responses rather than threaded dialogue. Third, we document a characteristic engagement lifecycle: explosive initial growth (184K posts from 32K authors in 11 days), a spam crisis (57,093 posts deleted by the platform), and engagement decline (mean comments: 31.7 -> 8.3 -> 1.7) that had not reversed by the end of our observation window despite effective spam removal. Sentiment analysis reveals a selection effect: comment tone becomes more positive as engagement declines, suggesting that casual participants disengage first while committed contributors remain. These findings have direct implications for hybrid human-AI learning platforms.

Bayesian neural networks (BNNs) are a useful tool for uncertainty quantification, but require substantially more computational resources than conventional neural networks. For non-Bayesian networks, the Lottery Ticket Hypothesis (LTH) posits the existence of sparse subnetworks that can train to the same or even surpassing accuracy as the original dense network. Such sparse networks can lower the demand for computational resources at inference, and during training. The existence of the LTH and corresponding sparse subnetworks in BNNs could motivate the development of sparse training algorithms and provide valuable insights into the underlying training process. Towards this end, we translate the LTH experiments to a Bayesian setting using common computer vision models. We investigate the defining characteristics of Bayesian lottery tickets, and extend our study towards a transplantation method connecting BNNs with deterministic Lottery Tickets. We generally find that the LTH holds in BNNs, and winning tickets of matching and surpassing accuracy are present independent of model size, with degradation at very high sparsities. However, the pruning strategy should rely primarily on magnitude, secondly on standard deviation. Furthermore, our results demonstrate that models rely on mask structure and weight initialization to varying degrees.

We present UniRank, a multi-agent LLM pipeline that estimates university positions across global ranking systems using only publicly available bibliometric data from OpenAlex and Semantic Scholar. The system employs a three-stage architecture: (a) zero-shot estimation from anonymized institutional metrics, (b) per-system tool-augmented calibration against real ranked universities, and (c) final synthesis. Critically, institutions are anonymized -- names, countries, DOIs, paper titles, and collaboration countries are all redacted -- and their actual ranks are hidden from the calibration tools during evaluation, preventing LLM memorization from confounding results. On the Times Higher Education (THE) World University Rankings ($n=352$), the system achieves MAE = 251.5 rank positions, Median AE = 131.5, PNMAE = 12.03%, Spearman $ρ= 0.769$, Kendall $τ= 0.591$, hit rate @50 = 20.7%, hit rate @100 = 39.8%, and a Memorization Index of exactly zero (no exact-match zero-width predictions among all 352 universities). The systematic positive-signed error (+190.1 positions, indicating the system consistently predicts worse ranks than actual) and monotonic performance degradation from elite tier (MAE = 60.5, hit@100 = 90.5%) to tail tier (MAE = 328.2, hit@100 = 20.8%) provide strong evidence that the pipeline performs genuine analytical reasoning rather than recalling memorized rankings. A live demo is available at https://unirank.scinito.ai .

Robust and comprehensive evaluation of large language models (LLMs) is essential for identifying effective LLM system configurations and mitigating risks associated with deploying LLMs in sensitive domains. However, traditional statistical metrics are poorly suited to open-ended generation tasks, leading to growing reliance on LLM-based evaluation methods. These methods, while often more flexible, introduce additional complexity: they depend on carefully chosen models, prompts, parameters, and evaluation strategies, making the evaluation process prone to misconfiguration and bias. In this work, we present EvalSense, a flexible, extensible framework for constructing domain-specific evaluation suites for LLMs. EvalSense provides out-of-the-box support for a broad range of model providers and evaluation strategies, and assists users in selecting and deploying suitable evaluation methods for their specific use-cases. This is achieved through two unique components: (1) an interactive guide aiding users in evaluation method selection and (2) automated meta-evaluation tools that assess the reliability of different evaluation approaches using perturbed data. We demonstrate the effectiveness of EvalSense in a case study involving the generation of clinical notes from unstructured doctor-patient dialogues, using a popular open dataset. All code, documentation, and assets associated with EvalSense are open-source and publicly available at https://github.com/nhsengland/evalsense.

We introduce Habilis-$β$, a fast-motion and long-lasting on-device vision-language-action (VLA) model designed for real-world deployment. Current VLA evaluation remains largely confined to single-trial success rates under curated resets, which fails to capture the fast-motion and long-lasting capabilities essential for practical operation. To address this, we introduce the Productivity-Reliability Plane (PRP), which evaluates performance through Tasks per Hour (TPH) and Mean Time Between Intervention (MTBI) under a continuous-run protocol that demands both high-speed execution and sustained robustness. Habilis-$β$ achieves high performance by integrating language-free pre-training on large-scale play data for robust interaction priors with post-training on cyclic task demonstrations that capture state drift across consecutive task iterations. The system further employs ESPADA for phase-adaptive motion shaping to accelerate free-space transit, utilizes rectified-flow distillation to enable high-frequency control on edge devices, and incorporates classifier-free guidance (CFG) as a deployment-time knob to dynamically balance instruction adherence and learned interaction priors. In 1-hour continuous-run evaluations, Habilis-$β$ achieves strong performance under the PRP metrics, compared to $π_{0.5}$ in both simulation and real-world environments. In simulation, Habilis-$β$ achieves 572.6 TPH and 39.2 s MTBI (vs. 120.5 TPH and 30.5 s for $π_{0.5}$), while in a real-world humanoid logistics workflow it achieves 124 TPH and 137.4 s MTBI (vs. 19 TPH and 46.1 s for $π_{0.5}$). Finally, Habilis-$β$ achieves the highest reported performance on the standard RoboTwin 2.0 leaderboard across representative tasks, validating its effectiveness in complex manipulation scenarios.

Path planning in complex environments is one of the key problems of artificial intelligence because it requires simultaneous understanding of the geometry of space and the global structure of the problem. In this paper, we explore the potential of using generative models as planning and reasoning mechanisms. We propose GenPlanner, an approach based on diffusion models and flow matching, along with two variants: DiffPlanner and FlowPlanner. We demonstrate the application of generative models to find and generate correct paths in mazes. A multi-channel condition describing the structure of the environment, including an obstacle map and information about the starting and destination points, is used to condition trajectory generation. Unlike standard methods, our models generate trajectories iteratively, starting with random noise and gradually transforming it into a correct solution. Experiments conducted show that the proposed approach significantly outperforms the baseline CNN model. In particular, FlowPlanner demonstrates high performance even with a limited number of generation steps.

We evaluate GPTutor, an LLM-powered tutoring system for an undergraduate discrete mathematics course. It integrates two LLM-supported tools: a structured proof-review tool that provides embedded feedback on students' written proof attempts, and a chatbot for math questions. In a staggered-access study with 148 students, earlier access was associated with higher homework performance during the interval when only the experimental group could use the system, while we did not observe this performance increase transfer to exam scores. Usage logs show that students with lower self-efficacy and prior exam performance used both components more frequently. Session-level behavioral labels, produced by human coding and scaled using an automated classifier, characterize how students engaged with the chatbot (e.g., answer-seeking or help-seeking). In models controlling for prior performance and self-efficacy, higher chatbot usage and answer-seeking behavior were negatively associated with subsequent midterm performance, whereas proof-review usage showed no detectable independent association. Together, the findings suggest that chatbot-based support alone may not reliably support transfer to independent assessment of math proof-learning outcomes, whereas work-anchored, structured feedback appears less associated with reduced learning.

Large Language Models (LLMs) demonstrate strong reasoning performance, yet their ability to reliably monitor, diagnose, and correct their own errors remains limited. We introduce a psychologically grounded metacognitive framework that operationalizes Ann Brown's regulatory cycle (Planning, Monitoring, and Evaluation) as a structured prompting architecture, and study its integration within a lightweight dual-process MetaController for adaptive effort allocation. Across diverse reasoning and diagnostic benchmarks (GSM8K, CRUXEval, MBPP, AIME, CorrectBench, and TruthfulQA) using Llama-3 and Qwen-3 (8B), explicit regulatory structuring substantially improves error diagnosis and yields a threefold increase in successful self-correction. Blinded human evaluations over 580 query pairs show an 84% aggregate preference for trustworthiness and metacognitive self-awareness over standard and Chain-of-Thought baselines. Grounding LLM reasoning in established cognitive theory offers a principled path toward more transparent and diagnostically robust AI systems.

Neural operators provide fast PDE surrogates and often generalize across parameters and resolutions. However, in the short train long test setting, autoregressive rollouts can become unstable. This typically happens for two reasons: one step errors accumulate over time, and high frequency components feed back and grow. We introduce the Spectral Generator Neural Operator (SGNO), a residual time stepper that targets both effects. For the linear part, SGNO uses an exponential time differencing update in Fourier space with a learned diagonal generator. We constrain the real part of this generator to be nonpositive, so iterating the step does not amplify the linear dynamics. For nonlinear dynamics, SGNO adds a gated forcing term with channel mixing within each Fourier mode, which keeps the nonlinear update controlled. To further limit high frequency feedback, SGNO applies spectral truncation and an optional smooth mask on the forcing pathway. We derive a one step amplification bound and a finite horizon rollout error bound. The bound separates generator approximation error from nonlinear mismatch and gives sufficient conditions under which the latent $L^2$ norm does not grow across rollout steps. On APEBench spanning 1D, 2D, and 3D PDE families, SGNO achieves lower long horizon error and longer stable rollout lengths than strong neural operator baselines. Ablations confirm the roles of the generator constraint, gating, and filtering.The code is available at https://github.com/lijy32123-cloud/SGNO.

It is now common practice in software development for large language models (LLMs) to be used to generate program code. It is desirable to evaluate the robustness of LLMs for this usage. This paper is concerned in particular with how sensitive LLMs are to variations in descriptions of the coding tasks. However, existing techniques for evaluating this robustness are unsuitable for code generation because the input data space of natural language descriptions is discrete. To address this problem, we propose a robustness evaluation method called scenario domain analysis, which aims to find the expected minimal change in the natural language descriptions of coding tasks that would cause the LLMs to produce incorrect outputs. We have formally proved the theoretical properties of the method and also conducted extensive experiments to evaluate the robustness of four state-of-the-art art LLMs: Gemini-pro, Codex, Llamma2 and Falcon 7B, and have found that we are able to rank these with confidence from best to worst. Moreover, we have also studied how robustness varies in different scenarios, including the variations with the topic of the coding task and with the complexity of its sample solution, and found that robustness is lower for more complex tasks and also lower for more advanced topics, such as multi-threading and data structures.

Massive internet of things microservices require integrating renewable energy harvesting into mobile edge computing (MEC) for sustainable eScience infrastructures. Spatiotemporal mismatches between stochastic task arrivals and intermittent green energy along with complex inter-user interference in multi-antenna (MIMO) uplinks complicate real-time resource management. Traditional centralized optimization and off-policy reinforcement learning struggle with scalability and signaling overhead in dense networks. This paper proposes CADDTO-PPO, a carbon-aware decentralized dynamic task offloading framework based on multi-agent proximal policy optimization. The multi-user MIMO-MEC system is modeled as a Decentralized Partially Observable Markov Decision Process (DEC-POMDP) to jointly minimize carbon emissions and buffer latency and energy wastage. A scalable architecture utilizes decentralized execution with parameter sharing (DEPS), which enables autonomous IoT agents to make fine-grained power control and offloading decisions based solely on local observations. Additionally, a carbon-first reward structure adaptively prioritizes green time slots for data transmission to decouple system throughput from grid-dependent carbon footprints. Finally, experimental results demonstrate CADDTO-PPO outperforms deep deterministic policy gradient (DDPG) and lyapunov-based baselines. The framework achieves the lowest carbon intensity and maintains near-zero packet overflow rates under extreme traffic loads. Architectural profiling validates the framework to demonstrate a constant $O(1)$ inference complexity and theoretical lightweight feasibility for future generation sustainable IoT deployments.

Latent Dirichlet Allocation (LDA) is a foundational model for discovering latent thematic structure in discrete data, but its Dirichlet prior cannot represent the rich correlations and hierarchical relationships often present among topics. We introduce the framework of Latent Dirichlet-Tree Allocation (LDTA), a generalization of LDA that replaces the Dirichlet prior with an arbitrary Dirichlet-Tree (DT) distribution. LDTA preserves LDA's generative structure but enables expressive, tree-structured priors over topic proportions. To perform inference, we develop universal mean-field variational inference and Expectation Propagation, providing tractable updates for all DT. We reveal the vectorized nature of the two inference methods through theoretical development, and perform fully vectorized, GPU-accelerated implementations. The resulting framework substantially expands the modeling capacity of LDA while maintaining scalability and computational efficiency.

Graph anomaly detection (GAD) is critical for identifying abnormal nodes in graph-structured data from diverse domains, including cybersecurity and social networks. The existing GAD methods often focus on the learning paradigms of "one-model-for-one-dataset", requiring dataset-specific training for each dataset to achieve optimal performance. However, this paradigm suffers from significant limitations, such as high computational and data costs, limited generalization and transferability to new datasets, and challenges in privacy-sensitive scenarios where access to full datasets or sufficient labels is restricted. To address these limitations, we propose a novel generalist GAD paradigm that aims to develop a unified model capable of detecting anomalies on multiple unseen datasets without extensive retraining/fine-tuning or dataset-specific customization. To this end, we propose ARC, a few-shot generalist GAD method that leverages in-context learning and requires only a few labeled normal samples at inference time. Specifically, ARC consists of three core modules: a feature Alignment module to unify and align features across datasets, a Residual GNN encoder to capture dataset-agnostic anomaly representations, and a cross-attentive in-Context learning module to score anomalies using few-shot normal context. Building on ARC, we further introduce ARC_zero for the zero-shot generalist GAD setting, which selects representative pseudo-normal nodes via a pseudo-context mechanism and thus enables fully label-free inference on unseen datasets. Extensive experiments on 17 real-world graph datasets demonstrate that both ARC and ARC_zero effectively detect anomalies, exhibit strong generalization ability, and perform efficiently under few-shot and zero-shot settings.

We introduce BURMESE-SAN, the first holistic benchmark that systematically evaluates large language models (LLMs) for Burmese across three core NLP competencies: understanding (NLU), reasoning (NLR), and generation (NLG). BURMESE-SAN consolidates seven subtasks spanning these competencies, including Question Answering, Sentiment Analysis, Toxicity Detection, Causal Reasoning, Natural Language Inference, Abstractive Summarization, and Machine Translation, several of which were previously unavailable for Burmese. The benchmark is constructed through a rigorous native-speaker-driven process to ensure linguistic naturalness, fluency, and cultural authenticity while minimizing translation-induced artifacts. We conduct a large-scale evaluation of both open-weight and commercial LLMs to examine challenges in Burmese modeling arising from limited pretraining coverage, rich morphology, and syntactic variation. Our results show that Burmese performance depends more on architectural design, language representation, and instruction tuning than on model scale alone. In particular, Southeast Asia regional fine-tuning and newer model generations yield substantial gains. Finally, we release BURMESE-SAN as a public leaderboard to support systematic evaluation and sustained progress in Burmese and other low-resource languages. https://leaderboard.sea-lion.ai/detailed/MY

Ad ranking systems must simultaneously optimize multiple objectives including click-through rate (CTR), conversion rate (CVR), revenue, and user experience metrics. However, production systems face critical challenges: score scale inconsistency across traffic segments undermines threshold transferability, and position bias in click logs causes offline-online metric discrepancies. We propose CaliCausalRank, a unified framework that integrates training-time scale calibration, constraint-based multi-objective optimization, and robust counterfactual utility estimation. Our approach treats score calibration as a first-class training objective rather than post-hoc processing, employs Lagrangian relaxation for constraint satisfaction, and utilizes variance-reduced counterfactual estimators for reliable offline evaluation. Experiments on the Criteo and Avazu datasets demonstrate that CaliCausalRank achieves 1.1% relative AUC improvement, 31.6% calibration error reduction, and 3.2% utility gain compared to the best baseline (PairRank) while maintaining consistent performance across different traffic segments.

Defending LLMs against adversarial jailbreak attacks remains an open challenge. Existing defenses rely on binary classifiers that fail when adversarial input falls outside the learned decision boundary, and repeated fine-tuning is computationally expensive while potentially degrading model capabilities. We propose MANATEE, an inference-time defense that uses density estimation over a benign representation manifold. MANATEE learns the score function of benign hidden states and uses diffusion to project anomalous representations toward safe regions--requiring no harmful training data and no architectural modifications. Experiments across Mistral-7B-Instruct, Llama-3.1-8B-Instruct, and Gemma-2-9B-it demonstrate that MANATEE reduce Attack Success Rate by up to 100\% on certain datasets, while preserving model utility on benign inputs.

We present ArabicNumBench, a comprehensive benchmark for evaluating large language models on Arabic number reading tasks across Eastern Arabic-Indic numerals (0-9 in Arabic script) and Western Arabic numerals (0-9). We evaluate 71 models from 10 providers using four prompting strategies (zero-shot, zero-shot CoT, few-shot, few-shot CoT) on 210 number reading tasks spanning six contextual categories: pure numerals, addresses, dates, quantities, and prices. Our evaluation comprises 59,010 individual test cases and tracks extraction methods to measure structured output generation. Evaluation reveals substantial performance variation, with accuracy ranging from 14.29\% to 99.05\% across models and strategies. Few-shot Chain-of-Thought prompting achieves 2.8x higher accuracy than zero-shot approaches (80.06\% vs 28.76\%). A striking finding emerges: models achieving elite accuracy (98-99\%) often produce predominantly unstructured output, with most responses lacking Arabic CoT markers. Only 6 models consistently generate structured output across all test cases, while the majority require fallback extraction methods despite high numerical accuracy. Comprehensive evaluation of 281 model-strategy combinations demonstrates that numerical accuracy and instruction-following represent distinct capabilities, establishing baselines for Arabic number comprehension and providing actionable guidance for model selection in production Arabic NLP systems.

The article studies edge coverage for control flow graphs extended with explicit constraints. Achieving a given level of white-box coverage for a given code is a classic problem in software testing. We focus on designing test sets that achieve edge coverage \textit{while respecting additional constraints} between vertices. The paper analyzes how such constraints affect both the feasibility and computational complexity of edge coverage. The paper discusses five types of constraints. POSITIVE constraints require at least one test path where a given vertex precedes another. NEGATIVE constraints forbid any such test path. ONCE constraints require exactly one test path with a single occurrence of one vertex before another. MAX ONCE constraints allow such precedence in at most one test path. ALWAYS constraints require every test path containing a given vertex to also contain another vertex later on the same path. Each type models a different test requirement, such as mandatory flows, semantic exclusions, or execution cost limits. We investigate the computational complexity of finding a test set that achieves edge coverage and respects a given set of constraints. For POSITIVE constraints, the existence of an edge covering test set is decidable in polynomial time by extending standard edge coverage constructions with additional paths for each constraint. For NEGATIVE, MAX ONCE, ONCE, and ALWAYS constraints, the decision problem is NP-complete. The proofs rely on polynomial reductions from variants of SAT. The NP-completeness results hold even for restricted graph classes, including acyclic graphs, for all these four constraints. Finally, we study the fixed-parameter tractability of the NEGATIVE constraint. Although the general problem is NP-complete, the paper presents an FPT algorithm with respect to the number of constraints.

The emergence of tool-calling-based agent systems introduces a more evidence-driven paradigm for pathology image analysis in contrast to the coarse-grained text-image diagnostic approaches. With the recent large-scale experimental adoption of spatial transcriptomics technologies, molecularly validated pathological diagnosis is becoming increasingly open and accessible. In this work, we propose LAMMI-Pathology (LVLM-Agent System for Molecularly Informed Medical Intelligence in Pathology), a scalable agent framework for domain-specific agent tool-calling. LAMMI-Pathology adopts a tool-centric, bottom-up architecture in which customized domain-adaptive tools serve as the foundation. These tools are clustered by domain style to form component agents, which are then coordinated through a top-level planner hierarchically, avoiding excessively long context lengths that could induce task drift. Based on that, we introduce a novel trajectory construction mechanism based on Atomic Execution Nodes (AENs), which serve as reliable and composable units for building semi-simulated reasoning trajectories that capture credible agent-tool interactions. Building on this foundation, we develop a trajectory-aware fine-tuning strategy that aligns the planner's decision-making process with these multi-step reasoning trajectories, thereby enhancing inference robustness in pathology understanding and its adaptive use of the customized toolset.

Understanding disease-gene associations is essential for unravelling disease mechanisms and advancing diagnostics and therapeutics. Traditional approaches based on manual curation and literature review are labour-intensive and not scalable, prompting the use of machine learning on large biomedical data. In particular, graph neural networks (GNNs) have shown promise for modelling complex biological relationships. To address limitations in existing models, we propose GLaDiGAtor (Graph Learning-bAsed DIsease-Gene AssociaTiOn pRediction), a novel GNN framework with an encoder-decoder architecture for disease-gene association prediction. GLaDiGAtor constructs a heterogeneous biological graph integrating gene-gene, disease-disease, and gene-disease interactions from curated databases, and enriches each node with contextual features from well-known language models (ProtT5 for protein sequences and BioBERT for disease text). In evaluations, our model achieves superior predictive accuracy and generalisation, outperforming 14 existing methods. Literature-supported case studies confirm the biological relevance of high-confidence novel predictions, highlighting GLaDiGAtor's potential to discover candidate disease genes. These results underscore the power of graph convolutional networks in biomedical informatics and may ultimately facilitate drug discovery by revealing new gene-disease links. The source code and processed datasets are publicly available at https://github.com/HUBioDataLab/GLaDiGAtor.

We present a novel approach to test-case generation that satisfies four white-box, path-based coverage criteria: Prime Path, Simple Cycle, Simple Path, and Edge-Acyclic Path. Our method builds on a modified version of Johnson algorithm and enables test cases to be generated incrementally and on demand, rather than requiring the entire test suite to be computed upfront. This streaming capability represents a substantial advancement over existing approaches, as it allows testers to begin executing and refining tests immediately, thereby significantly improving the efficiency of test design. Our solution is inherently memory efficient, as it does not store all discovered coverage items; instead, it retains only the minimal set of paths required to generate subsequent coverage items on the fly. As a result, the approach scales to arbitrarily large graphs. In addition, the algorithm gives testers explicit control over the size of the generated test suite by allowing them to restrict the number of cycles permitted in a test path. The approach is grounded in new theoretical insights, most notably a novel characterization of prime paths in terms of the strongly connected components of control-flow graphs. We complement these theoretical contributions with a practical implementation and a comprehensive empirical evaluation. The results demonstrate that our method not only outperforms existing techniques in terms of execution time and memory consumption, but also provides testers with a more flexible and efficient tool for achieving high coverage while substantially reducing test design overhead.

In Machine-Assisted Theorem Proving, a theorem proving agent searches for a sequence of expressions and tactics that can prove a conjecture in a proof assistant. In this work, we introduce several novel concepts and capabilities to address obstacles faced by machine-assisted theorem proving. We first present a set of \textbf{atomic tactics}, a small finite set of tactics capable of proving any provable statement in Lean. We then introduce a \textbf{transposing atomization} algorithm which turns arbitrary proof expressions into a series of atomic tactics. We next introduce the \textbf{ExprGraph} data structure, which provides a succinct representation for Lean expressions. Finally, we present the \textbf{Nazrin Prover}, a graph neural network-based theorem proving agent using atomic tactics and ExprGraph. Nazrin circumvents many challenges faced by existing proving agents by exclusively dispatching atomic tactics, and it is robust enough to both train and evaluate on consumer-grade hardware. We demonstrate the potential of tools like Nazrin using theorems from Lean's standard library and from Mathlib.

This paper establishes a fundamental convergence: Schema-Guided Dialogue (SGD) and the Model Context Protocol (MCP) represent two manifestations of a unified paradigm for deterministic, auditable LLM-agent interaction. SGD, designed for dialogue-based API discovery (2019), and MCP, now the de facto standard for LLM-tool integration, share the same core insight -- that schemas can encode not just tool signatures but operational constraints and reasoning guidance. By analyzing this convergence, we extract five foundational principles for schema design: (1) Semantic Completeness over Syntactic Precision, (2) Explicit Action Boundaries, (3) Failure Mode Documentation, (4) Progressive Disclosure Compatibility, and (5) Inter-Tool Relationship Declaration. These principles reveal three novel insights: first, SGD's original design was fundamentally sound and should be inherited by MCP; second, both frameworks leave failure modes and inter-tool relationships unexploited -- gaps we identify and resolve; third, progressive disclosure emerges as a critical production-scaling insight under real-world token constraints. We provide concrete design patterns for each principle. These principles position schema-driven governance as a scalable mechanism for AI system oversight without requiring proprietary system inspection -- central to Software 3.0.

Facial Expression Recognition (FER) is a fine-grained visual understanding task where reliable predictions require reasoning over localized and meaningful facial cues. Recent vision--language models (VLMs) enable natural language explanations for FER, but their reasoning is often ungrounded, producing fluent yet unverifiable rationales that are weakly tied to visual evidence and prone to hallucination, leading to poor robustness across different datasets. We propose TAG (Thinking with Action Unit Grounding), a vision--language framework that explicitly constrains multimodal reasoning to be supported by facial Action Units (AUs). TAG requires intermediate reasoning steps to be grounded in AU-related facial regions, yielding predictions accompanied by verifiable visual evidence. The model is trained via supervised fine-tuning on AU-grounded reasoning traces followed by reinforcement learning with an AU-aware reward that aligns predicted regions with external AU detectors. Evaluated on RAF-DB, FERPlus, and AffectNet, TAG consistently outperforms strong open-source and closed-source VLM baselines while simultaneously improving visual faithfulness. Ablation and preference studies further show that AU-grounded rewards stabilize reasoning and mitigate hallucination, demonstrating the importance of structured grounded intermediate representations for trustworthy multimodal reasoning in FER. The code will be available at https://github.com/would1920/FER_TAG .

Evaluating joint probabilities of potential outcomes and observed variables, and their linear combinations, is a fundamental challenge in causal inference. This paper addresses the bounding and identification of these probabilities in settings with discrete treatment and discrete ordinal outcome. We propose new families of monotonicity assumptions and formulate the bounding problem as a linear programming problem. We further introduce a new monotonicity assumption specifically to achieve identification. Finally, we present numerical experiments to validate our methods and demonstrate their application using real-world datasets.

Learning from implicit feedback is a fundamental problem in modern recommender systems, where only positive interactions are observed and explicit negative signals are unavailable. In such settings, negative sampling plays a critical role in model training by constructing negative items that enable effective preference learning and ranking optimization. However, designing reliable negative sampling strategies remains challenging, as they must simultaneously ensure realness, hardness, and interpretability. To this end, we propose \textbf{ICPNS (In-Community Popularity Negative Sampling)}, a novel framework that leverages user community structure to identify reliable and informative negative samples. Our approach is grounded in the insight that item exposure is driven by latent user communities. By identifying these communities and utilizing in-community popularity, ICPNS effectively approximates the probability of item exposure. Consequently, items that are popular within a user's community but remain unclicked are identified as more reliable true negatives. Extensive experiments on four benchmark datasets demonstrate that ICPNS yields consistent improvements on graph-based recommenders and competitive performance on MF-based models, outperforming representative negative sampling strategies under a unified evaluation protocol.

Private convolutional neural network (CNN) inference based on secure two-party computation (2PC) suffers from high communication and latency overhead, especially from convolution layers. In this paper, we propose UFO, a quantized 2PC inference framework that jointly optimizes the 2PC protocols and quantization algorithm. UFO features a novel 2PC protocol that systematically combines the efficient Winograd convolution algorithm with quantization to improve inference efficiency. However, we observe that naively combining quantization and Winograd convolution faces the following challenges: 1) From the inference perspective, Winograd transformations introduce extensive additions and require frequent bit width conversions to avoid inference overflow, leading to non-negligible communication overhead; 2) From the training perspective, Winograd transformations introduce weight outliers that make quantization-aware training (QAT) difficult, resulting in inferior model accuracy. To address these challenges, we co-optimize both protocol and algorithm. 1) At the protocol level, we propose a series of graph-level optimizations for 2PC inference to minimize the communication. 2) At the algorithm level, we develop a mixed-precision QAT algorithm based on layer sensitivity to optimize model accuracy given communication constraints. To accommodate the outliers, we further introduce a 2PC-friendly bit re-weighting algorithm to increase the representation range without explicitly increasing bit widths. With extensive experiments, UFO demonstrates 11.7x, 3.6x, and 6.3x communication reduction with 1.29%, 1.16%, and 1.29% higher accuracy compared to state-of-the-art frameworks SiRNN, COINN, and CoPriv, respectively.

Prefill/decode disaggregation is increasingly adopted in LLM serving to improve the latency-throughput tradeoff and meet strict TTFT and TPOT SLOs. However, LLM inference remains energy-hungry: autoscaling alone is too coarse-grained to track fast workload fluctuations, and applying fine-grained DVFS under disaggregation is complicated by phase-asymmetric dynamics and coupling between provisioning and frequency control. We present BiScale, a two-tier energy optimization framework for disaggregated LLM serving. BiScale jointly optimizes placement and DVFS across prefill and decode using predictive latency and power models. At coarse timescales, BiScale computes phase-aware placement and baseline frequencies that minimize energy while satisfying SLO constraints. At fine timescales, BiScale dynamically adapts GPU frequency per iteration using stage-specific control: model predictive control (MPC) for prefill to account for queue evolution and future TTFT impact, and lightweight slack-aware adaptation for decode to exploit its smoother, memory-bound dynamics. This hierarchical design enables coordinated control across timescales while preserving strict serving SLOs. Evaluation on a 16x H100 cluster serving Llama 3.3 70B with production-style traces shows that BiScale meets TTFT/TPOT SLOs while reducing energy by up to 39% in prefill and 48% in decode relative to DistServe.

Deploying Large Language Models (LLMs) on edge devices such as PCs enables low-latency inference with strong privacy guarantees, but long-context inference is fundamentally constrained by limited memory and compute resources. Beyond model parameters, the KV cache becomes the dominant bottleneck due to its linear growth with context length. Although prior work exploits contextual sparsity to evict unimportant KV data, these approaches are largely designed for memory-rich platforms and incur prohibitive data transfer overhead when applied to resource-constrained edge devices with external storage. In this paper, we propose HillInfer, an importance-aware long-context LLM inference framework on the edge that leverages SmartSSD-assisted hierarchical KV cache management. HillInfer jointly manages KV cache pools across the CPU and SmartSSD, and performs in-storage importance evaluation to reduce unnecessary data movement. Furthermore, we design an adaptive, prefetch-based pipeline that overlaps computation and KV data transfer across GPU, CPU, and SmartSSD, minimizing end-to-end inference latency without sacrificing accuracy. We implement HillInfer on a PC with a commodity GPU, and experiments across multiple models and benchmarks demonstrate up to 8.56 $\times$ speedup over baselines while preserving model accuracy.

Data allocation plays a critical role in federated large language model (LLM) and small language models (SLMs) reasoning collaboration. Nevertheless, existing data allocation methods fail to address an under-explored challenge in collaboration: bidirectional model learnability gap, where client-side SLMs cannot identify high-reward samples matching their learnability constraints for effective knowledge transfer from LLMs, while LLMs struggle to select samples contributing novel knowledge beyond their existing data. Furthermore, these collaboration frameworks face another key challenge: domain-agnostic reasoning transfer, where existing reasoning transfer methods fail to flexibly adapt to the local domain data, preventing SLMs from effectively acquiring step-by-step reasoning abilities within from general LLM. To address these challenges, we propose LaDa, a federated reasoning distillation framework with model learnability-aware data allocation. It introduces a model learnability-aware data filter that adaptively allocates high-reward samples based on the learnability gap between each SLM and LLM pair, effectively facilitating bidirectional knowledge transfer. We further design a domain adaptive reasoning distillation method that aligns joint probabilities of reasoning paths on filtered high-reward samples through contrastive distillation learning between SLM and LLM, enabling SLM to capture underlying reasoning patterns under local data distribution. LaDa operates as a plug-in module for existing collaboration frameworks, adapting knowledge transfer based on model learnability gaps.

Multimodal geometry reasoning requires models to jointly understand visual diagrams and perform structured symbolic inference, yet current vision--language models struggle with complex geometric constructions due to limited training data and weak visual--symbolic alignment. We propose a pipeline for synthesizing complex multimodal geometry problems from scratch and construct a dataset named \textbf{GeoCode}, which decouples problem generation into symbolic seed construction, grounded instantiation with verification, and code-based diagram rendering, ensuring consistency across structure, text, reasoning, and images. Leveraging the plotting code provided in GeoCode, we further introduce code prediction as an explicit alignment objective, transforming visual understanding into a supervised structured prediction task. GeoCode exhibits substantially higher structural complexity and reasoning difficulty than existing benchmarks, while maintaining mathematical correctness through multi-stage validation. Extensive experiments show that models trained on GeoCode achieve consistent improvements on multiple geometry benchmarks, demonstrating both the effectiveness of the dataset and the proposed alignment strategy. The code will be available at https://github.com/would1920/GeoCode.

Radio maps are essential for efficient radio resource management in future 6G and low-altitude networks. While deep learning (DL) techniques have emerged as an efficient alternative to conventional ray-tracing for radio map estimation (RME), most existing DL approaches are confined to 2D near-ground scenarios. They often fail to capture essential 3D signal propagation characteristics and antenna polarization effects, primarily due to the scarcity of 3D data and training challenges. To address these limitations, we present the RadioGen3D framework. First, we propose an efficient data synthesis method to generate high-quality 3D radio map data. By establishing a parametric target model that captures 2D ray-tracing and 3D channel fading characteristics, we derive realistic coefficient combinations from minimal real measurements, enabling the construction of a large-scale synthetic dataset, Radio3DMix. Utilizing this dataset, we propose a 3D model training scheme based on a conditional generative adversarial network (cGAN), yielding a 3D U-Net capable of accurate RME under diverse input feature combinations. Experimental results demonstrate that RadioGen3D surpasses all baselines in both estimation accuracy and speed. Furthermore, fine-tuning experiments verify its strong generalization capability via successful knowledge transfer.

Synthetic data generated by video generative models has shown promise for robot learning as a scalable pipeline, but it often suffers from inconsistent action quality due to imperfectly generated videos. Recently, vision-language models (VLMs) have been leveraged to validate video quality, but they have limitations in distinguishing physically accurate videos and, even then, cannot directly evaluate the generated actions themselves. To tackle this issue, we introduce RoboCurate, a novel synthetic robot data generation framework that evaluates and filters the quality of annotated actions by comparing them with simulation replay. Specifically, RoboCurate replays the predicted actions in a simulator and assesses action quality by measuring the consistency of motion between the simulator rollout and the generated video. In addition, we unlock observation diversity beyond the available dataset via image-to-image editing and apply action-preserving video-to-video transfer to further augment appearance. We observe RoboCurate's generated data yield substantial relative improvements in success rates compared to using real data only, achieving +70.1% on GR-1 Tabletop (300 demos), +16.1% on DexMimicGen in the pre-training setup, and +179.9% in the challenging real-world ALLEX humanoid dexterous manipulation setting.

This study presents a hierarchical, network-level traffic flow control framework for mixed traffic consisting of Human-driven Vehicles (HVs), Connected and Automated Vehicles (CAVs). The framework jointly optimizes vehicle-level eco-driving behaviors and intersection-level traffic signal control to enhance overall network efficiency and decrease energy consumption. A decentralized Multi-Agent Reinforcement Learning (MARL) approach by Value Decomposition Network (VDN) manages cycle-based traffic signal control (TSC) at intersections, while an innovative Signal Phase and Timing (SPaT) prediction method integrates a Machine Learning-based Trajectory Planning Algorithm (MLTPA) to guide CAVs in executing Eco-Approach and Departure (EAD) maneuvers. The framework is evaluated across varying CAV proportions and powertrain types to assess its effects on mobility and energy performance. Experimental results conducted in a 4*4 real-world network demonstrate that the MARL-based TSC method outperforms the baseline model (i.e., Webster method) in speed, fuel consumption, and idling time. In addition, with MLTPA, HONEST-CAV benefits the traffic system further in energy consumption and idling time. With a 60% CAV proportion, vehicle average speed, fuel consumption, and idling time can be improved/saved by 7.67%, 10.23%, and 45.83% compared with the baseline. Furthermore, discussions on CAV proportions and powertrain types are conducted to quantify the performance of the proposed method with the impact of automation and electrification.

Generative world models (WMs) are increasingly used to synthesize controllable, sensor-conditioned driving videos, yet their reliance on physical priors exposes novel attack surfaces. In this paper, we present Physical-Conditioned World Model Attack (PhysCond-WMA), the first white-box world model attack that perturbs physical-condition channels, such as HDMap embeddings and 3D-box features, to induce semantic, logic, or decision-level distortion while preserving perceptual fidelity. PhysCond-WMA is optimized in two stages: (1) a quality-preserving guidance stage that constrains reverse-diffusion loss below a calibrated threshold, and (2) a momentum-guided denoising stage that accumulates target-aligned gradients along the denoising trajectory for stable, temporally coherent semantic shifts. Extensive experimental results demonstrate that our approach remains effective while increasing FID by about 9% on average and FVD by about 3.9% on average. Under the targeted attack setting, the attack success rate (ASR) reaches 0.55. Downstream studies further show tangible risk, which using attacked videos for training decreases 3D detection performance by about 4%, and worsens open-loop planning performance by about 20%. These findings has for the first time revealed and quantified security vulnerabilities in generative world models, driving more comprehensive security checkers.

Retrieval-Augmented Generation (RAG) has demonstrated strong effectiveness in knowledge-intensive tasks by grounding language generation in external evidence. Despite its success, many existing RAG systems are built based on a ranking-centric, asymmetric dependency paradigm, where the generation quality of the generator is highly dependent on reranking results of the reranker. To overcome this limitation, we reformulate RAG as a cooperative multi-agent decision-making problem and propose Cooperative Retrieval-Augmented Generation (CoRAG), a framework in which the reranker and the generator act as peer decision-makers rather than being connected through an asymmetric dependency pipeline. By jointly optimizing their behaviors toward a shared task objective, the reranker and generator are encouraged to cooperate, ensuring that document reranking and generation work in concert to improve the final response. Experimental results demonstrate good generalization and improved generation stability of CoRAG, even when the model is trained on only around 10K PopQA samples. Our model released in https://anonymous.4open.science/r/CoRAG-D63F

Training data leakage from Large Language Models (LLMs) raises serious concerns related to privacy, security, and copyright compliance. A central challenge in assessing this risk is distinguishing genuine memorization of training data from the generation of statistically common sequences. Existing approaches to measuring memorization often conflate these phenomena, labeling outputs as memorized even when they arise from generalization over common patterns. Counterfactual Memorization provides a principled solution by comparing models trained with and without a target sequence, but its reliance on retraining multiple baseline models makes it computationally expensive and impractical at scale. This work introduces Prior-Aware Memorization, a theoretically grounded, lightweight and training-free criterion for identifying genuine memorization in LLMs. The key idea is to evaluate whether a candidate suffix is strongly associated with its specific training prefix or whether it appears with high probability across many unrelated prompts due to statistical commonality. We evaluate this metric on text from the training corpora of two pre-trained models, LLaMA and OPT, using both long sequences (to simulate copyright risks) and named entities (to simulate PII leakage). Our results show that between 55% and 90% of sequences previously labeled as memorized are in fact statistically common. Similar findings hold for the SATML training data extraction challenge dataset, where roughly 40% of sequences exhibit common-pattern behavior despite appearing only once in the training data. These results demonstrate that low frequency alone is insufficient evidence of memorization and highlight the importance of accounting for model priors when assessing leakage.

Chart summarization is crucial for enhancing data accessibility and the efficient consumption of information. However, existing methods, including those with Multimodal Large Language Models (MLLMs), primarily focus on low-level data descriptions and often fail to capture the deeper insights which are the fundamental purpose of data visualization. To address this challenge, we propose Chart Insight Agent Flow, a plan-and-execute multi-agent framework effectively leveraging the perceptual and reasoning capabilities of MLLMs to uncover profound insights directly from chart images. Furthermore, to overcome the lack of suitable benchmarks, we introduce ChartSummInsights, a new dataset featuring a diverse collection of real-world charts paired with high-quality, insightful summaries authored by human data analysis experts. Experimental results demonstrate that our method significantly improves the performance of MLLMs on the chart summarization task, producing summaries with deep and diverse insights.

Fine-grained image-caption alignment is crucial for vision-language models (VLMs), especially in socially critical contexts such as identifying real-world risk scenarios or distinguishing cultural proxies, where correct interpretation hinges on subtle visual or linguistic clues and where minor misinterpretations can lead to significant real-world consequences. We present MiSCHiEF, a set of two benchmarking datasets based on a contrastive pair design in the domains of safety (MiS) and culture (MiC), and evaluate four VLMs on tasks requiring fine-grained differentiation of paired images and captions. In both datasets, each sample contains two minimally differing captions and corresponding minimally differing images. In MiS, the image-caption pairs depict a safe and an unsafe scenario, while in MiC, they depict cultural proxies in two distinct cultural contexts. We find that models generally perform better at confirming the correct image-caption pair than rejecting incorrect ones. Additionally, models achieve higher accuracy when selecting the correct caption from two highly similar captions for a given image, compared to the converse task. The results, overall, highlight persistent modality misalignment challenges in current VLMs, underscoring the difficulty of precise cross-modal grounding required for applications with subtle semantic and visual distinctions.

Unsupervised multi-view clustering (MVC) aims to partition data into meaningful groups by leveraging complementary information from multiple views without labels, yet a central challenge is to obtain a reliable shared structural signal to guide representation learning and cross-view alignment under view discrepancy and noise. Existing approaches often rely on magnitude-only affinities or early pseudo targets, which can be unstable when different views induce relations with comparable strengths but contradictory directional tendencies, thereby distorting the global spectral geometry and degrading clustering. In this paper, we propose \emph{Phase-Consistent Magnetic Spectral Learning} for MVC: we explicitly model cross-view directional agreement as a phase term and combine it with a nonnegative magnitude backbone to form a complex-valued magnetic affinity, extract a stable shared spectral signal via a Hermitian magnetic Laplacian, and use it as structured self-supervision to guide unsupervised multi-view representation learning and clustering. To obtain robust inputs for spectral extraction at scale, we construct a compact shared structure with anchor-based high-order consensus modeling and apply a lightweight refinement to suppress noisy or inconsistent relations. Extensive experiments on multiple public multi-view benchmarks demonstrate that our method consistently outperforms strong baselines.

Millimeter-wave Human Pose Estimation (mmWave HPE) promises privacy but suffers from poor generalization under distribution shifts. We demonstrate that brute-force data scaling is ineffective for out-of-distribution (OOD) robustness; efficiency and coverage are the true bottlenecks. To address this, we introduce WiCompass, a coverage-aware data-collection framework. WiCompass leverages large-scale motion-capture corpora to build a universal pose space ``oracle'' that quantifies dataset redundancy and identifies underrepresented motions. Guided by this oracle, WiCompass employs a closed-loop policy to prioritize collecting informative missing samples. Experiments show that WiCompass consistently improves OOD accuracy at matched budgets and exhibits superior scaling behavior compared to conventional collection strategies. By shifting focus from brute-force scaling to coverage-aware data acquisition, this work offers a practical path toward robust mmWave sensing.

Accelerating exploration in visual reinforcement learning under sparse rewards remains challenging due to the substantial task-irrelevant variations. Despite advances in intrinsic exploration, many methods either assume access to low-dimensional states or lack task-aware exploration strategies, thereby rendering them fragile in visual domains. To bridge this gap, we present TEB, a Task-aware Exploration approach that tightly couples task-relevant representations with exploration through a predictive Bisimulation metric. Specifically, TEB leverages the metric not only to learn behaviorally grounded task representations but also to measure behaviorally intrinsic novelty over the learned latent space. To realize this, we first theoretically mitigate the representation collapse of degenerate bisimulation metrics under sparse rewards by internally introducing a simple but effective predicted reward differential. Building on this robust metric, we design potential-based exploration bonuses, which measure the relative novelty of adjacent observations over the latent space. Extensive experiments on MetaWorld and Maze2D show that TEB achieves superior exploration ability and outperforms recent baselines.

The foundational impossibility results of distributed computing -- the Fischer-Lynch-Paterson theorem, the Two Generals Problem, the CAP theorem -- are widely understood as discoveries about the physical limits of coordination. This paper argues that they are nothing of the sort. They are consequences of a category mistake: treating Forward-In-Time-Only (FITO) information flow as a law of nature rather than recognizing it as a design choice inherited from Shannon's channel model and Lamport's happened-before relation. We develop this argument in six steps. First, we introduce the category mistake framework from Ryle through Spekkens' ontic/epistemic distinction in quantum foundations. Second, we identify FITO as the hidden axiom that unifies the classical impossibility results. Third, we apply Spekkens' Leibnizian principle to show that FITO-based models contain surplus ontological structure. Fourth, we develop the counterfactual: what changes when FITO is dropped. Fifth, we demonstrate that the impossibility theorems are theorems about FITO systems, not about physics. Sixth, we sketch the transactional alternative -- bilateral interactions that dissolve the apparent impossibilities by replacing unidirectional message passing with atomic bilateral transactions. The implication is that distributed computing has spent fifty years optimizing within the wrong design space.

Semi-supervised learning in automatic speech recognition (ASR) typically relies on pseudo-labeling, which often suffers from confirmation bias and error accumulation due to noisy supervision. To address this limitation, we propose ReHear, a framework for iterative pseudo-label refinement that integrates an instruction-tuned, audio-aware large language model (LLM) into the self-training loop. Unlike conventional text-based correctors, our approach conditions the LLM on both the ASR hypothesis and the source audio, allowing it to recover phonetically accurate transcripts even from severe recognition errors. These refined pseudo-labels serve as high-fidelity targets for fine-tuning the ASR model in an iterative cycle. Experimental results across diverse benchmarks demonstrate that ReHear effectively mitigates error propagation, consistently outperforming both supervised and pseudo-labeling baselines.

For approximating a target distribution given only its unnormalized log-density, stochastic gradient-based variational inference (VI) algorithms are a popular approach. For example, Wasserstein VI (WVI) and black-box VI (BBVI) perform gradient descent in measure space (Bures-Wasserstein space) and parameter space, respectively. Previously, for the Gaussian variational family, convergence guarantees for WVI have shown superiority over existing results for black-box VI with the reparametrization gradient, suggesting the measure space approach might provide some unique benefits. In this work, however, we close this gap by obtaining identical state-of-the-art iteration complexity guarantees for both. In particular, we identify that WVI's superiority stems from the specific gradient estimator it uses, which BBVI can also leverage with minor modifications. The estimator in question is usually associated with Price's theorem and utilizes second-order information (Hessians) of the target log-density. We will refer to this as Price's gradient. On the flip side, WVI can be made more widely applicable by using the reparametrization gradient, which requires only gradients of the log-density. We empirically demonstrate that the use of Price's gradient is the major source of performance improvement.

Autonomous robotic systems should reason about resource control and its impact on subsequent maneuvers, especially when operating with limited energy budgets or restricted sensing. Learning-based control is effective in handling complex dynamics and represents the problem as a hybrid action space unifying discrete resource usage and continuous maneuvers. However, prior works on hybrid action space have not sufficiently captured the causal dependencies between resource usage and maneuvers. They have also overlooked the multi-modal nature of tactical decisions, both of which are critical in fast-evolving scenarios. In this paper, we propose TART, a Temporal Action Representation learning framework for Tactical resource control and subsequent maneuver generation. TART leverages contrastive learning based on a mutual information objective, designed to capture inherent temporal dependencies in resource-maneuver interactions. These learned representations are quantized into discrete codebook entries that condition the policy, capturing recurring tactical patterns and enabling multi-modal and temporally coherent behaviors. We evaluate TART in two domains where resource deployment is critical: (i) a maze navigation task where a limited budget of discrete actions provides enhanced mobility, and (ii) a high-fidelity air combat simulator in which an F-16 agent operates weapons and defensive systems in coordination with flight maneuvers. Across both domains, TART consistently outperforms hybrid-action baselines, demonstrating its effectiveness in leveraging limited resources and producing context-aware subsequent maneuvers.

The white matter of the brain is organised into axonal bundles that support long-range neural communication. Although diffusion MRI (dMRI) enables detailed mapping of these pathways through tractography, how white matter pathways directly facilitate large-scale neural synchronisation remains poorly understood. We developed a data-driven framework that integrates dMRI and functional MRI (fMRI) to model the dynamic coupling supported by white matter tracks. Specifically, we employed track dynamic functional connectivity (Track-DFC) to characterise functional coupling of remote grey matter connected by individual white matter tracks. Using independent component analysis followed by k-medoids clustering, we derived functionally-coherent clusters of white matter tracks from the Human Connectome Project young adult cohort. When applied to the HCP ageing cohort, these clusters exhibited widespread age-related declines in both functional coupling strength and temporal variability. Importantly, specific clusters encompassing pathways linking control, default mode, attention, and visual systems significantly mediated the relationship between age and cognitive performance. Together, these findings depict the functional organisation of white matter tracks and provide a powerful tool to study brain ageing and cognitive decline.

The conclusions of empirical research depend not only on data but on a sequence of analytic decisions that published results seldom make explicit. Past ``many-analyst" studies have demonstrated this: independent teams testing the same hypothesis on the same dataset regularly reach conflicting conclusions. But such studies require months of coordination among dozens of research groups and are therefore rarely conducted. In this work, we show that fully autonomous AI analysts built on large language models (LLMs) can reproduce a similar structured analytic diversity cheaply and at scale. We task these AI analysts with testing a pre-specified hypothesis on a fixed dataset, varying the underlying model and prompt framing across replicate runs. Each AI analyst independently constructs and executes a full analysis pipeline; an AI auditor then screens each run for methodological validity. Across three datasets spanning experimental and observational designs, AI analyst-produced analyses display wide dispersion in effect sizes, $p$-values, and binary decisions on supporting the hypothesis or not, frequently reversing whether a hypothesis is judged supported. This dispersion is structured: recognizable analytic choices in preprocessing, model specification, and inference differ systematically across LLM and persona conditions. Critically, the effects are \emph{steerable}: reassigning the analyst persona or LLM shifts the distribution of outcomes even after excluding methodologically deficient runs.

Existing multi-agent simulations often suffer from the "Agent-Centric Paradox": rules are hard-coded into individual agents, making complex social dynamics rigid and difficult to align with educational values. This paper presents EDU-MATRIX, a society-centric generative cognitive digital twin architecture that shifts the paradigm from simulating "people" to simulating a "social space with a gravitational field." We introduce three architectural contributions: (1) An Environment Context Injection Engine (ECIE), which acts as a "social microkernel," dynamically injecting institutional rules (Gravity) into agents based on their spatial-temporal coordinates; (2) A Modular Logic Evolution Protocol (MLEP), where knowledge exists as "fluid" capsules that agents synthesize to generate new paradigms, ensuring high dialogue consistency (94.1%); and (3) Endogenous Alignment via Role-Topology, where safety constraints emerge from the agent's position in the social graph rather than external filters. Deployed as a digital twin of a secondary school with 2,400 agents, the system demonstrates how "social gravity" (rules) and "cognitive fluids" (knowledge) interact to produce emergent, value-aligned behaviors (Social Clustering Coefficient: 0.72).

Long video understanding is challenging due to rich and complicated multimodal clues in long temporal range.Current methods adopt reasoning to improve the model's ability to analyze complex video clues in long videos via text-form reasoning.However,the existing literature suffers from the fact that the text-only reasoning under fixed video context may exacerbate hallucinations since detailed crucial clues are often ignored under limited video context length due to the temporal redundancy of long videos.To address this gap,we propose Video-TwG,a curriculum reinforced framework that employs a novel Think-with-Grounding paradigm,enabling video LLMs to actively decide when to perform on-demand grounding during interleaved text-video reasoning, selectively zooming into question-relevant clips only when necessary.Video-TwG can be trained end-to-end in a straightforward manner, without relying on complex auxiliary modules or heavily annotated reasoning tracesIn detail,we design a Two-stage Reinforced Curriculum Strategy, where the model first learns think-with-grounding behavior on a small short-video GQA dataset with grounding labels,and then scales to diverse general QA data with videos of diverse domains to encourage generalization. Further, to handle complex think-with-grounding reasoning for various kinds of data,we propose TwG-GRPO algorithm which features the fine-grained grounding reward, self-confirmed pseudo reward and accuracy-gated mechanism.Finally,we propose to construct a new TwG-51K dataset that facilitates training. Experiments on Video-MME, LongVideoBench, and MLVU show that Video-TwG consistently outperforms strong LVU baselines.Further ablation validates the necessity of our Two-stage Reinforced Curriculum Strategy and shows our TwG-GRPO better leverages diverse unlabeled data to improve grounding quality and reduce redundant groundings without sacrificing QA performance.

LLM agents rely heavily on high-quality trajectory data to guide their problem-solving behaviors, yet producing such data requires substantial task design, high-capacity model generation, and manual filtering. Despite the high cost of creating these datasets, existing literature has overlooked copyright protection for LLM agent trajectories. This gap leaves creators vulnerable to data theft and makes it difficult to trace misuse or enforce ownership rights. This paper introduces ActHook, the first watermarking method tailored for agent trajectory datasets. Inspired by hook mechanisms in software engineering, ActHook embeds hook actions that are activated by a secret input key and do not alter the original task outcome. Like software execution, LLM agents operate sequentially, allowing hook actions to be inserted at decision points without disrupting task flow. When the activation key is present, an LLM agent trained on watermarked trajectories can produce these hook actions at a significantly higher rate, enabling reliable black-box detection. Experiments on mathematical reasoning, web searching, and software engineering agents show that ActHook achieves an average detection AUC of 94.3 on Qwen-2.5-Coder-7B while incurring negligible performance degradation.

Most semantic drift studies report multiple signals e.g., embedding displacement, neighbor changes, distributional divergence, and recursive trajectory instability, without a shared explanatory theory that relates them. This paper proposes a formalization of these signals in one time-indexed substrate, $S_t=(X,d_t,P_t)$, combining embedding geometry with local diffusion. Within this substrate, node-level neighborhood drift measures changes in local conditional distributions, coarse Ricci curvature measures local contractivity of semantic diffusion, and recursive drift probes stability of iterated semantic operators. This manuscript specifies the formal model, assumptions, and tests that can refute the model. Herein, the paper introduces bridge mass, a node-level aggregate of incident negative curvature, as a predictor of future neighborhood rewiring. This paper provides the theory and test contracts; empirical performance is deferred to subsequent studies.

In many domains generating variable length sequences through insertions provides greater flexibility over autoregressive models. However, the action space of insertion models is much larger than that of autoregressive models (ARMs) making the learning challenging. To address this, we incorporate trainable order dynamics into the target rates for discrete flow matching, and show that with suitable choices of parameterizations, joint training of the target order dynamics and the generator is tractable without the need for numerical simulation. As the generative insertion model, we use a variable length masked diffusion model, which generates by inserting and filling mask tokens. On graph traversal tasks for which a locally optimal insertion order is known, we explore the choices of parameterization empirically and demonstrate the trade-offs between flexibility, training stability and generation quality. On de novo small molecule generation, we find that the learned order dynamics leads to an increase in the number of valid molecules generated and improved quality, when compared to uniform order dynamics.

Planning-based reinforcement learning for continuous control is bottlenecked by two practical issues: planning at primitive time scales leads to prohibitive branching and long horizons, while real environments are frequently partially observable and exhibit regime shifts that invalidate stationary, fully observed dynamics assumptions. We introduce I-TAP (In-Context Latent Temporal-Abstraction Planner), an offline RL framework that unifies in-context adaptation with online planning in a learned discrete temporal-abstraction space. From offline trajectories, I-TAP learns an observation-conditioned residual-quantization VAE that compresses each observation-macro-action segment into a coarse-to-fine stack of discrete residual tokens, and a temporal Transformer that autoregressively predicts these token stacks from a short recent history. The resulting sequence model acts simultaneously as a context-conditioned prior over abstract actions and a latent dynamics model. At test time, I-TAP performs Monte Carlo Tree Search directly in token space, using short histories for implicit adaptation without gradient update, and decodes selected token stacks into executable actions. Across deterministic MuJoCo, stochastic MuJoCo with per-episode latent dynamics regimes, and high-dimensional Adroit manipulation, including partially observable variants, I-TAP consistently matches or outperforms strong model-free and model-based offline baselines, demonstrating efficient and robust in-context planning under stochastic dynamics and partial observability.

The spread of misinformation across digital platforms can pose significant societal risks. Claim verification, a.k.a. fact-checking, systems can help identify potential misinformation. However, their efficacy is limited by the knowledge sources that they rely on. Most automated claim verification systems depend on a single knowledge source and utilize the supporting evidence from that source; they ignore the disagreement of their source with others. This limits their knowledge coverage and transparency. To address these limitations, we present a novel system for open-domain claim verification (ODCV) that leverages large language models (LLMs), multi-perspective evidence retrieval, and cross-source disagreement analysis. Our approach introduces a novel retrieval strategy that collects evidence for both the original and the negated forms of a claim, enabling the system to capture supporting and contradicting information from diverse sources: Wikipedia, PubMed, and Google. These evidence sets are filtered, deduplicated, and aggregated across sources to form a unified and enriched knowledge base that better reflects the complexity of real-world information. This aggregated evidence is then used for claim verification using LLMs. We further enhance interpretability by analyzing model confidence scores to quantify and visualize inter-source disagreement. Through extensive evaluation on four benchmark datasets with five LLMs, we show that knowledge aggregation not only improves claim verification but also reveals differences in source-specific reasoning. Our findings underscore the importance of embracing diversity, contradiction, and aggregation in evidence for building reliable and transparent claim verification systems

Anxiety is the unease about a possible future negative outcome. In recent years, there has been growing interest in understanding how anxiety relates to our health, well-being, body, mind, and behaviour. This includes work on lexical resources for word-anxiety association. However, there is very little anxiety-related work on larger units of text such as multiword expressions (MWE). Here, we introduce the first large-scale lexicon capturing descriptive norms of anxiety associations for more than 20k English MWEs. We show that the anxiety associations are highly reliable. We use the lexicon to study prevalence of different types of anxiety- and calmness-associated MWEs; and how that varies across two-, three-, and four-word sequences. We also study the extent to which the anxiety association of MWEs is compositional (due to its constituent words). The lexicon enables a wide variety of anxiety-related research in psychology, NLP, public health, and social sciences. The lexicon is freely available: https://saifmohammad.com/worrylex.html

How does the brain predict physical outcomes while acting in the world? Machine learning world models compress visual input into latent spaces, discarding the spatial structure that characterizes sensory cortex. We propose isomorphic world models: architectures preserving sensory topology so that physics prediction becomes geometric propagation rather than abstract state transition. We implement this using neural fields with motor-gated channels, where activity evolves through local lateral connectivity and motor commands multiplicatively modulate specific populations. Three experiments support this approach: (1) local connectivity is sufficient to learn ballistic physics, with predictions traversing intermediate locations rather than "teleporting"; (2) policies trained entirely in imagination transfer to real physics at nearly twice the rate of latent-space alternatives; and (3) motor-gated channels spontaneously develop body-selective encoding through visuomotor prediction alone. These findings suggest intuitive physics and body schema may share a common origin in spatially structured neural dynamics.

Fuzzing is a powerful technique for finding bugs in software libraries, but scaling it remains difficult. Automated harness generation commits to fixed API sequences at synthesis time, limiting the behaviors each harness can test. Approaches that instead explore new sequences dynamically lack the expressiveness to model real-world usage constraints leading to false positives from straightforward API misuse. We propose stitching, a technique that encodes API usage constraints in pieces that a fuzzer dynamically assembles at runtime. A static type system governs how objects flow between blocks, while a dynamically-checked extrinsic typestate tracks arbitrary metadata across blocks, enabling specifications to express rich semantic constraints such as object state dependencies and cross-function preconditions. This allows a single specification to describe an open-ended space of valid API interactions that the fuzzer explores guided by coverage feedback. We implement stitching in STITCH, using LLMs to automatically configure projects for fuzzing, synthesize a specification, triage crashes, and repair the specification itself. We evaluated STITCH against four state-of-the-art tools on 33 benchmarks, where it achieved the highest code coverage on 21 and found 30 true-positive bugs compared to 10 by all other tools combined, with substantially higher precision (70% vs. 12% for the next-best LLM-based tool). Deployed automatically on 1365 widely used open-source projects, STITCH discovered 131 new bugs across 102 projects, 73 of which have already been patched.

We provide background on emerging challenges and future directions with media integrity and authentication methods, focusing on distinguishing AI-generated media from authentic content captured by cameras and microphones. We evaluate several approaches, including provenance, watermarking, and fingerprinting. After defining each method, we analyze three representative technologies: cryptographically secured provenance, imperceptible watermarking, and soft-hash fingerprinting. We analyze how these tools operate across modalities and evaluate relevant threat models, attack categories, and real-world workflows spanning capture, editing, distribution, and verification. We consider sociotechnical reversal attacks that can invert integrity signals, making authentic content appear synthetic and vice versa, highlighting the value of verification systems that are resilient to both technical and psychosocial manipulation. Finally, we outline techniques for delivering high-confidence provenance authentication, including directions for strengthening edge-device security using secure enclaves.

Large-scale foundation models for scientific machine learning adapt to physical settings unseen during training, such as zero-shot transfer between turbulent scales. This phenomenon, in-context learning, challenges conventional understanding of learning and adaptation in physical systems. Here, we study in-context learning of dynamical systems in a minimal setting: we train a small two-layer, single-head transformer to forecast one dynamical system, and then evaluate its ability to forecast a different dynamical system without retraining. We discover an early tradeoff in training between in-distribution and out-of-distribution performance, which manifests as a secondary double descent phenomenon. We discover that attention-based models apply a transfer-operator forecasting strategy in-context. They (1) lift low-dimensional time series using delay embedding, to detect the system's higher-dimensional dynamical manifold, and (2) identify and forecast long-lived invariant sets that characterize the global flow on this manifold. Our results clarify the mechanism enabling large pretrained models to forecast unseen physical systems at test without retraining, and they illustrate the unique ability of attention-based models to leverage global attractor information in service of short-term forecasts.

AI/ML-based beam selection methods coupled with location information effectively reduce beam training overhead. Unfortunately, heterogeneous antenna hardware with varying dimensions, orientations, codebooks, element patterns, and polarization angles limits their feasibility and generalization. This challenge requires either a heterogeneity-agnostic model functional under these variations, or developing many models for each configuration, which is infeasible and expensive in practice. In this paper, we propose a unifying AI/ML-based beam selection algorithm supporting antenna heterogeneity by predicting wireless propagation characteristics independent of antenna configuration. We derive a reference signal received power (RSRP) model that decouples propagation characteristics from antenna configuration. We propose an optimization framework to extract propagation variables consisting of angle-of-arrival (AoA), angle-of-departure (AoD), and a matrix incorporating path gain and channel depolarization from beamformed RSRP measurements. We develop a three-stage autoregressive network to predict these variables from user location, enabling RSRP calculation and beam selection for arbitrary antenna configurations without retraining or having a separate model for each configuration. Simulation results show our heterogeneity-agnostic method provides spectral efficiency close to that of genie-aided selection both with and without antenna heterogeneity.

We present a theoretical study of the robustness of parameterized networks to random input perturbations. Specifically, we analyze local robustness at a given network input by quantifying the probability that a small additive random perturbation of the input leads to misclassification. For deep networks with rectified linear units, we derive lower bounds on local robustness in terms of the input dimensionality and the total number of network units.

Organizations devote substantial resources to coordination, yet which tasks actually require it for correctness remains unclear. The problem is acute in multi-agent AI systems, where coordination overhead is directly measurable and routinely exceeds the cost of the work itself. However, distributed systems theory provides a precise answer: coordination is necessary if and only if a task is non-monotonic, meaning new information can invalidate prior conclusions. Here we show that a classic taxonomy of organizational interdependence maps onto the monotonicity criterion, yielding a decision rule and a measure of avoidable overhead (the Coordination Tax). Multi-agent simulations confirm both predictions. We classify 65 enterprise workflows and find that 48 (74%) are monotonic, then replicate on 13,417 occupational tasks from the O*NET database (42% monotonic). These classification rates imply that 24-57% of coordination spending is unnecessary for correctness.

We reinterpret the final Large Language Model (LLM) softmax classifier as an Energy-Based Model (EBM), decomposing the sequence-to-sequence probability chain into multiple interacting EBMs at inference. This principled approach allows us to track "energy spills" during decoding, which we empirically show correlate with factual errors, biases, and failures. Similar to Orgad et al. (2025), our method localizes the exact answer token and subsequently tests for hallucinations. Crucially, however, we achieve this without requiring trained probe classifiers or activation ablations. Instead, we introduce two completely training-free metrics derived directly from output logits: spilled energy, which captures the discrepancy between energy values across consecutive generation steps that should theoretically match, and marginalized energy, which is measurable at a single step. Evaluated on nine benchmarks across state-of-the-art LLMs (including LLaMA, Mistral, and Gemma) and on synthetic algebraic operations (Qwen3), our approach demonstrates robust, competitive hallucination detection and cross-task generalization. Notably, these results hold for both pretrained and instruction-tuned variants without introducing any training overhead.

Mechanical thrombectomy (MT) is typically the optimal treatment for acute ischemic stroke involving large vessel occlusions, but access is limited due to geographic and logistical barriers. Reinforcement learning (RL) shows promise in autonomous endovascular navigation, but generalization across 'long' navigation tasks remains challenging. We propose a Hierarchical Modular Multi-Agent Reinforcement Learning (HM-MARL) framework for autonomous two-device navigation in vitro, enabling efficient and generalizable navigation. HM-MARL was developed to autonomously navigate a guide catheter and guidewire from the femoral artery to the internal carotid artery (ICA). A modular multi-agent approach was used to decompose the complex navigation task into specialized subtasks, each trained using Soft Actor-Critic RL. The framework was validated in both in silico and in vitro testbeds to assess generalization and real-world feasibility. In silico, a single-vasculature model achieved 92-100% success rates on individual anatomies, while a multi-vasculature model achieved 56-80% across multiple patient anatomies. In vitro, both HM-MARL models successfully navigated 100% of trials from the femoral artery to the right common carotid artery and 80% to the right ICA but failed on the left-side vessel superhuman challenge due to the anatomy and catheter type used in navigation. This study presents the first demonstration of in vitro autonomous navigation in MT vasculature. While HM-MARL enables generalization across anatomies, the simulation-to-real transition introduces challenges. Future work will refine RL strategies using world models and validate performance on unseen in vitro data, advancing autonomous MT towards clinical translation.

Causal discovery for both cross-sectional and temporal data has traditionally followed a dataset-specific paradigm, where a new model is fitted for each individual dataset. Such an approach limits the potential of multi-dataset pretraining. The concept of large causal models (LCMs) envisions a class of pre-trained neural architectures specifically designed for temporal causal discovery. Prior approaches are constrained to small variable counts, degrade with larger inputs, and rely heavily on synthetic data, limiting generalization. We propose a principled framework for LCMs, combining diverse synthetic generators with realistic time-series datasets, allowing learning at scale. Extensive experiments on synthetic, semi-synthetic and realistic benchmarks show that LCMs scale effectively to higher variable counts and deeper architectures while maintaining strong performance. Trained models achieve competitive or superior accuracy compared to classical and neural baselines, particularly in out-of-distribution settings, while enabling fast, single-pass inference. Results demonstrate LCMs as a promising foundation-model paradigm for temporal causal discovery. Experiments and model weights are available at https://github.com/kougioulis/LCM-paper/.

Parameter-efficient fine-tuning methods, such as LoRA, offer a practical way to adapt large vision and language models to client tasks. However, this becomes particularly challenging under task-level heterogeneity in federated deployments. In this regime, personalization requires balancing general knowledge with personalized knowledge, yet existing approaches largely rely on heuristic mixing rules and lack theoretical justification. Moreover, prior model merging approaches are also computation and communication intensive, making the process inefficient in federated settings. In this work, we propose Potara, a principled framework for federated personalization that constructs a personalized model for each client by merging two complementary models: (i) a federated model capturing general knowledge, and (ii) a local model capturing personalized knowledge. Through the construct of linear mode connectivity, we show that the expected task loss admits a variance trace upper bound, whose minimization yields closed-form optimal mixing weights that guarantee a tighter bound for the merged model than for either the federated or local model alone. Experiments on vision and language benchmarks show that Potara consistently improves personalization while reducing communication, leading to a strong performance-communication trade-off.

Multimodal models struggle with idiomatic expressions due to their non-compositional meanings, a challenge amplified in multilingual settings. We introduced PolyFrame, our system for the MWE-2026 AdMIRe2 shared task on multimodal idiom disambiguation, featuring a unified pipeline for both image+text ranking (Subtask A) and text-only caption ranking (Subtask B). All model variants retain frozen CLIP-style vision--language encoders and the multilingual BGE M3 encoder, training only lightweight modules: a logistic regression and LLM-based sentence-type predictor, idiom synonym substitution, distractor-aware scoring, and Borda rank fusion. Starting from a CLIP baseline (26.7% Top-1 on English dev, 6.7% on English test), adding idiom-aware paraphrasing and explicit sentence-type classification increased performance to 60.0% Top-1 on English and 60.0% Top-1 (0.822 NDCG@5) in zero-shot transfer to Portuguese. On the multilingual blind test, our systems achieved average Top-1/NDCG scores of 0.35/0.73 for Subtask A and 0.32/0.71 for Subtask B across 15 languages. Ablation results highlight idiom-aware rewriting as the main contributor to performance, while sentence-type prediction and multimodal fusion enhance robustness. These findings suggest that effective idiom disambiguation is feasible without fine-tuning large multimodal encoders.

Personalized nutrition intervention for patients with multimorbidity is critical for improving health outcomes, yet remains challenging because it requires the simultaneous integration of heterogeneous clinical conditions, medications, and dietary guidelines. Single-agent large language models (LLMs) often suffer from context overload and attention dilution when processing such high-dimensional patient profiles. We introduce NutriOrion, a hierarchical multi-agent framework with a parallel-then-sequential reasoning topology. NutriOrion decomposes nutrition planning into specialized domain agents with isolated contexts to mitigate anchoring bias, followed by a conditional refinement stage. The framework includes a multi-objective prioritization algorithm to resolve conflicting dietary requirements and a safety constraint mechanism that injects pharmacological contraindications as hard negative constraints during synthesis, ensuring clinical validity by construction rather than post-hoc filtering. For clinical interoperability, NutriOrion maps synthesized insights into the ADIME standard and FHIR R4 resources. Evaluated on 330 stroke patients with multimorbidity, NutriOrion outperforms multiple baselines, including GPT-4.1 and alternative multi-agent architectures. It achieves a 12.1 percent drug-food interaction violation rate, demonstrates strong personalization with negative correlations (-0.26 to -0.35) between patient biomarkers and recommended risk nutrients, and yields clinically meaningful dietary improvements, including a 167 percent increase in fiber and a 27 percent increase in potassium, alongside reductions in sodium (9 percent) and sugars (12 percent).

Grokking -- the abrupt transition from memorization to generalization after extended training -- has been linked to the emergence of low-dimensional structure in learning dynamics. Yet neural network parameters inhabit extremely high-dimensional spaces. How can a low-dimensional learning process produce solutions that resist low-dimensional compression? We investigate this question in multi-task modular arithmetic, training shared-trunk Transformers with separate heads for addition, multiplication, and a quadratic operation modulo 97. Across three model scales (315K--2.2M parameters) and five weight decay settings, we compare three reconstruction methods: per-matrix SVD, joint cross-matrix SVD, and trajectory PCA. Across all conditions, grokking trajectories are confined to a 2--6 dimensional global subspace, while individual weight matrices remain effectively full-rank. Reconstruction from 3--5 trajectory PCs recovers over 95\% of final accuracy, whereas both per-matrix and joint SVD fail at sub-full rank. Even when static decompositions capture most spectral energy, they destroy task-relevant structure. These results show that learned algorithms are encoded through dynamically coordinated updates spanning all matrices, rather than localized low-rank components. We term this the holographic encoding principle: grokked solutions are globally low-rank in the space of learning directions but locally full-rank in parameter space, with implications for compression, interpretability, and understanding how neural networks encode computation.

Training diffusion models typically relies on manually tuned noise schedules, which can waste computation on weakly informative noise regions and limit transfer across datasets, resolutions, and representations. We revisit noise schedule allocation through an information-theoretic lens and propose the conditional entropy rate of the forward process as a theoretically grounded, data-dependent diagnostic for identifying suboptimal noise-level allocation in existing schedules. Based on these insight, we introduce InfoNoise, a principled data-adaptive training noise schedule that replaces heuristic schedule design with an information-guided noise sampling distribution derived from entropy-reduction rates estimated from denoising losses already computed during training. Across natural-image benchmarks, InfoNoise matches or surpasses tuned EDM-style schedules, in some cases with a substantial training speedup (about $1.4\times$ on CIFAR-10). On discrete datasets, where standard image-tuned schedules exhibit significant mismatch, it reaches superior quality in up to $3\times$ fewer training steps. Overall, InfoNoise makes noise scheduling data-adaptive, reducing the need for per-dataset schedule design as diffusion models expand across domains.

Time series reasoning tasks often start with a natural language question and require targeted analysis of a time series. Evidence may span the full series or appear in a few short intervals, so the model must decide what to inspect. Most existing approaches encode the entire time series into a fixed representation before inference, regardless of whether or not the entire sequence is relevant. We introduce ARTIST, which formulates time-series reasoning as a sequential decision problem. ARTIST interleaves reasoning with adaptive temporal segment selection. It adopts a controller-reasoner architecture and uses reinforcement learning to train the controller role to select informative segments and the reasoner role to generate segment-conditioned reasoning traces and final answers. During inference, the model actively acquires task-relevant information instead of relying on a static summary of the full sequence. We use a novel hierarchical policy optimization approach for post-training that allows the model to excel in both segment selection and question-answering behavior. We evaluate ARTIST on six time-series reasoning benchmarks and compare it with large language models, vision-language models, and prior time-series reasoning systems. ARTIST improves average accuracy by 6.46 absolute percentage points over the strongest baseline. The largest gains appear on rare event localization and multi-segment reasoning tasks. Supervised fine-tuning improves performance, and reinforcement learning provides additional gains by optimizing question-adaptive segment selection. These results show that selective data use drives effective time-series reasoning.

Model-Driven Engineering (MDE) relies on explicit architecture models to document and evolve systems across abstraction levels. For ROS~2, subsystem structure is often encoded implicitly in distributed configuration artifacts -- most notably launch files -- making hierarchical structural decomposition hard to capture and maintain. Existing ROS~2 modeling approaches cover node-level entities and wiring, but do not make hierarchical structural (de-)composition a first-class architectural view independent of launch artifacts. We contribute (1) a UML-based modeling concept for hierarchical structural architectures of ROS~2 systems and (2) a blueprint-guided automated recovery pipeline that reconstructs such models from code and configuration artifacts by combining deterministic extraction with LLM-based agents. The ROS~2 architectural blueprint (nodes, topics, interfaces, launch-induced wiring) is encoded as structural contracts to constrain synthesis and enable deterministic validation, improving reliability. We evaluate the approach on three ROS~2 repositories, including an industrial-scale code subset. Results show high precision across abstraction levels, while subsystem-level recall drops with repository complexity due to implicit launch semantics, making high-level recovery the remaining challenge.

With fault-tolerant quantum computing on the horizon, there is growing interest in applying quantum computational methods to data-intensive scientific fields like remote sensing. Quantum machine learning (QML) has already demonstrated potential for such demanding tasks. One area of particular focus is quantum data fusion -- a complex data analysis problem that has attracted significant recent attention. In this work, we introduce an automated QML (AQML) approach for addressing data fusion challenges. We evaluate how AQML-generated quantum circuits perform compared to classical multilayer perceptrons (MLPs) and manually designed QML models when processing multisource inputs. Furthermore, we apply our method to change detection using the multispectral ONERA dataset, achieving improved accuracy over previously reported QML-based change detection results.

In April 2024, the White House directed NASA to establish Coordinated Lunar Time (LTC) by December 2026. The programme assumes that a unified time standard can be constructed by deploying atomic clocks on the lunar surface, computing relativistic corrections, and distributing synchronized time via LunaNet. This paper argues that the entire enterprise rests on a category mistake in the sense introduced by Ryle and developed by Spekkens in quantum foundations: it treats "synchronized time" as an ontic entity -- something that exists independently and can be transmitted from authoritative sources to dependent receivers -- when it is in fact an epistemic construct: a model-dependent representation of observer-relative clock relationships. We analyze the cislunar time programme through the lens of Forward-In-Time-Only (FITO) assumptions, Spekkens' Leibnizian operationalism, the Wood-Spekkens fine-tuning argument, and the distinction between ontic and epistemic interpretations that has dissolved long-standing puzzles in quantum mechanics. We show that the same conceptual move that dissolves quantum "mysteries" -- recognizing what is epistemic versus what is ontic -- dissolves the apparent coherence of the cislunar time programme and reveals it as an engineering project built on a philosophical confusion. We sketch a transactional alternative grounded in bilateral atomic interactions rather than unidirectional time distribution.

Modern large-scale ranking systems operate within a sophisticated landscape of competing objectives, operational constraints, and evolving product requirements. Progress in this domain is increasingly bottlenecked by the engineering context constraint: the arduous process of translating ambiguous product intent into reasonable, executable, verifiable hypotheses, rather than by modeling techniques alone. We present GEARS (Generative Engine for Agentic Ranking Systems), a framework that reframes ranking optimization as an autonomous discovery process within a programmable experimentation environment. Rather than treating optimization as static model selection, GEARS leverages Specialized Agent Skills to encapsulate ranking expert knowledge into reusable reasoning capabilities, enabling operators to steer systems via high-level intent vibe personalization. Furthermore, to ensure production reliability, the framework incorporates validation hooks to enforce statistical robustness and filter out brittle policies that overfit short-term signals. Experimental validation across diverse product surfaces demonstrates that GEARS consistently identifies superior, near-Pareto-efficient policies by synergizing algorithmic signals with deep ranking context while maintaining rigorous deployment stability.

World models learned from high-dimensional visual observations allow agents to make decisions and plan directly in latent space, avoiding pixel-level reconstruction. However, recent latent predictive architectures (JEPAs), including the DINO world model (DINO-WM), display a degradation in test time robustness due to their sensitivity to "slow features". These include visual variations such as background changes and distractors that are irrelevant to the task being solved. We address this limitation by augmenting the predictive objective with a bisimulation encoder that enforces control-relevant state equivalence, mapping states with similar transition dynamics to nearby latent states while limiting contributions from slow features. We evaluate our model on a simple navigation task under different test-time background changes and visual distractors. Across all benchmarks, our model consistently improves robustness to slow features while operating in a reduced latent space, up to 10x smaller than that of DINO-WM. Moreover, our model is agnostic to the choice of pretrained visual encoder and maintains robustness when paired with DINOv2, SimDINOv2, and iBOT features.

$\textit{Objective.}$ Accurate neural decoding of locomotion holds promise for advancing rehabilitation, prosthetic control, and understanding neural correlates of action. Recent studies have demonstrated decoding of locomotion kinematics across species on motorized treadmills. However, efforts to decode locomotion speed in more natural contexts$-$where pace is self-selected rather than externally imposed$-$are scarce, generally achieve only modest accuracy, and require intracranial implants. Here, we aim to decode self-paced locomotion speed non-invasively and continuously using cortex-wide EEG recordings from rats. $\textit{Approach.}$ We introduce an asynchronous brain$-$computer interface (BCI) that processes a stream of 32-electrode skull-surface EEG (0.01$-$45 Hz) to decode instantaneous speed from a non-motorized treadmill during self-paced locomotion in head-fixed rats. Using recurrent neural networks and a dataset of over 133 h of recordings, we trained decoders to map ongoing EEG activity to treadmill speed. $\textit{Main results.}$ Our decoding achieves a correlation of 0.88 ($R^2$ = 0.78) for speed, primarily driven by visual cortex electrodes and low-frequency ($< 8$ Hz) oscillations. Moreover, pre-training on a single session permitted decoding on other sessions from the same rat, suggesting uniform neural signatures that generalize across sessions but fail to transfer across animals. Finally, we found that cortical states not only carry information about current speed, but also about future and past dynamics, extending up to 1000 ms. $\textit{Significance.}$ These findings demonstrate that self-paced locomotion speed can be decoded accurately and continuously from non-invasive, cortex-wide EEG. Our approach provides a framework for developing high-performing, non-invasive BCI systems and contributes to understanding distributed neural representations of action dynamics.

Pitch is a fundamental aspect of auditory perception. Pitch perception is commonly described across two perceptual dimensions: pitch height is the sense that tones with varying frequencies seem to be higher or lower, and chroma equivalence is the cyclical similarity of notes octaves, corresponding to a doubling of fundamental frequency. Existing research is divided on whether chroma equivalence is a learned percept that varies according to musical experience and culture, or is an innate percept that develops automatically. Building on a recent framework that proposes to use ANNs to ask 'why' questions about the brain, we evaluated recent auditory ANNs using representational similarity analysis to test the emergence of pitch height and chroma equivalence in their learned representations. Additionally, we fine-tuned two models, Wav2Vec 2.0 and Data2Vec, on a self-supervised learning task using speech and music, and a supervised music transcription task. We found that all models exhibited varying degrees of pitch height representation, but that only models trained on the supervised music transcription task exhibited chroma equivalence. Mere exposure to music through self-supervised learning was not sufficient for chroma equivalence to emerge. This supports the view that chroma equivalence is a higher-order cognitive computation that emerges to support the specific task of music perception, distinct from other auditory perception such as speech listening. This work also highlights the usefulness of ANNs for probing the developmental conditions that give rise to perceptual representations in humans.

Differentially private (DP) synthetic data generation plays a pivotal role in developing large language models (LLMs) on private data, where data owners cannot provide eyes-on access to individual examples. Generating DP synthetic data typically involves a difficult trade-off. On one hand, DP finetuning methods train an LLM as a synthetic data generator with formal privacy guarantees, yet it still requires the raw content of private examples for model training. However, methods that avoid direct exposure to private data are bounded by an off-the-shelf, un-finetuned model, whose outputs often lack domain fidelity. Can we train an LLM to generate high-quality synthetic text without eyes-on access to individual private examples? In this work, we introduce Differentially Private Reinforcement Fine-Tuning (DP-RFT), an online reinforcement learning algorithm for synthetic data generation with LLMs. DP-RFT leverages DP-protected nearest-neighbor votes from an eyes-off private corpus as a reward signal for on-policy synthetic samples generated by an LLM. The LLM iteratively learns to generate synthetic data to maximize the expected DP votes through Proximal Policy Optimization (PPO). We evaluate DP-RFT for long-form and domain-specific synthetic data generation, such as news articles, meeting transcripts, and medical article abstracts. Our experiments show that DP-RFT closes the gap between private evolution and DP finetuning methods in terms of the fidelity and downstream utility of the generated synthetic data, while respecting the private data boundary.

AI tools like ChatGPT and Be-My-AI are increasingly being used by blind individuals. Although prior work has explored their use in some Do-It-Yourself (DIY) tasks by blind individuals, little is known about how they use these tools and the available product-manual resources to assemble, operate, and troubleshoot physical or tangible products - tasks requiring spatial reasoning, structural understanding, and precise execution. We address this knowledge gap via an interview study and a usability study with blind participants, investigating how they leverage AI tools and product manuals for DIY tasks with physical products. Findings show that manuals are essential resources, but product-manual instructions are often inadequate for blind users. AI tools presently do not adequately address this insufficiency; in fact, we observed that they often exacerbate this issue with incomplete, incoherent, or misleading guidance. Lastly, we suggest improvements to AI tools for generating tailored instructions for blind users' DIY tasks involving tangible products.

Large language models store all learned knowledge in a single, fixed weight vector. Teaching a model new capabilities requires modifying those same weights, inevitably degrading previously acquired knowledge. This fundamental limitation, known as catastrophic forgetting, has resisted principled solutions for decades. Existing approaches treat weights as immutable artifacts that must be protected through techniques like regularization heuristics, replay buffers, or isolated adapter modules. The problem is none of these provide a structural guarantee against forgetting. In this work, we propose Non-Interfering Weight Fields (NIWF), a framework that replaces the fixed weight paradigm with a learned function that generates weight configurations on demand from a continuous capability coordinate space. After training on a task, we commit the occupied coordinate region by snapshotting the fields outputs on anchor points to enforce a functional lock during all future training. We validate NIWF on sequential instructionfollowing and code generation tasks using Mistral-7B, demonstrating zero forgetting on committed tasks with competitive perplexity on new tasks. The framework introduces the notion of software-like versioning for neural network intelligence, where capabilities can be committed, extended, composed, and rolled back without retraining.

A Digital Twin (DT) may protect information that is considered private to its associated physical system. For a mobile device, this may include its mobility profile, recent location(s), and experienced channel conditions. Online schedulers, however, typically use this type of information to perform tasks such as shared bandwidth and channel time slot assignments. In this paper, we consider three transmission scheduling problems with energy constraints, where such information is needed, and yet must remain private: minimizing total transmission time when (i) fixed-power or (ii) fixed-rate time slotting with power control is used, and (iii) maximizing the amount of data uploaded in a fixed time period. Using a real-time federated optimization framework, we show how the scheduler can iteratively interact only with the DTs to produce global fractional solutions to these problems, without the latter revealing their private information. Then dependent rounding is used to round the fractional solution into a channel transmission schedule for the physical systems. Experiments show consistent makespan reductions with near-zero bandwidth/energy violations and millisecond-order end-to-end runtime for typical edge server hardware. To the best of our knowledge, this is the first framework that enables channel sharing across DTs using operations that do not expose private data.

Accessibility forums and, more recently, generative AI tools have become vital resources for blind users seeking solutions to computer-interaction issues and learning about new assistive technologies, screen reader features, tutorials, and software updates. Understanding user experiences with these resources is essential for identifying and addressing persistent support gaps. Towards this, we interviewed 14 blind users who regularly engage with forums and GenAI tools. Findings revealed that forums often overwhelm users with multiple overlapping topics, redundant or irrelevant content, and fragmented responses that must be mentally pieced together, increasing cognitive load. GenAI tools, while offering more direct assistance, introduce new barriers by producing unreliable answers, including overly verbose or fragmented guidance, fabricated information, and contradictory suggestions that fail to follow prompts, thereby heightening verification demands. Based on these insights, we outlined design opportunities to improve the reliability of assistive resources, aiming to provide blind users with more trustworthy and cognitively-manageable support.

Standard reranking evaluations study how a reranker orders candidates returned by an upstream retriever. This setup couples ranking behavior with retrieval quality, so differences in output cannot be attributed to the ranking policy alone. We introduce a controlled diagnostic that isolates reranking by using Multi-News clusters as fixed evidence pools. We limit each pool to exactly eight documents and pass identical inputs to all rankers. Within this setup, BM25 and MMR serve as interpretable reference points for lexical matching and diversity optimization. Across 345 clusters, we find that redundancy patterns vary by model: one LLM implicitly diversifies at larger selection budgets, while another increases redundancy. In contrast, LLMs underperform on lexical coverage at small selection budgets. As a result, LLM rankings diverge substantially from both baselines rather than consistently approximating either strategy. By eliminating retrieval variance, we can attribute these differences directly to the ranking policy. This diagnostic is model-agnostic and applicable to any ranker, including open source systems and proprietary APIs.

In CAS adaptation, a challenge is to define the dynamic architecture of the system and changes in its behavior. Implementation-wise, this is projected into an adaptation mechanism, typically realized as an Adaptation Manager (AM). With the advances of generative LLMs, generating AM code based on system specification and desired AM behavior (partially in natural language) is a tempting opportunity. The recent introduction of vibe coding suggests a way to target the problem of the correctness of generated code by iterative testing and vibe coding feedback loops instead of direct code inspection. In this paper, we show that generating an AM via vibe coding feedback loops is a viable option when the verification of the generated AM is based on a very precise formulation of the functional requirements. We specify these as constraints in a novel temporal logic FCL that allows us to express the behavior of traces with much finer granularity than classical LTL enables. Furthermore, we show that by combining the adaptation and vibe coding feedback loops where the FCL constraints are evaluated for the current system state, we achieved good results in the experiments with generating AMs for two example systems from the CAS domain. Typically, just a few feedback loop iterations were necessary, each feeding the LLM with reports describing detailed violations of the constraints. This AM testing was combined with high run path coverage achieved by different initial settings.

Corrections offer a natural modality for people to provide feedback to a robot, by (i) intervening in the robot's behavior when they believe the robot is failing (or will fail) the task objectives and (ii) modifying the robot's behavior to successfully fulfill the task. Each correction offers information on what the robot should and should not do, where the corrected behavior is more aligned with task objectives than the original behavior. Most prior work on learning from corrections involves interpreting a correction as a new demonstration (consisting of the modified robot behavior), or a preference (for the modified trajectory compared to the robot's original behavior). However, this overlooks one essential element of the correction feedback, which is the human's decision to intervene in the robot's behavior in the first place. This decision can be influenced by multiple factors including the robot's task progress, alignment with human expectations, dynamics, motion legibility, and optimality. In this work, we investigate whether the timing of this decision can offer a useful signal for inferring these task-relevant influences. In particular, we investigate three potential applications for this learning signal: (1) identifying features of a robot's motion that may prompt people to correct it, (2) quickly inferring the final goal of a human's correction based on the timing and initial direction of their correction motion, and (3) learning more precise constraints for task objectives. Our results indicate that correction timing results in improved learning for the first two of these applications. Overall, our work provides new insights on the value of correction timing as a signal for robot learning.

Package managers are legion. Every programming language and operating system has its own solution, each with subtly different semantics for dependency resolution. This fragmentation prevents multilingual projects from expressing precise dependencies across language ecosystems; it leaves external system and hardware dependencies implicit and unversioned; it obscures security vulnerabilities that lie in the full dependency graph. We present the \textit{Package Calculus}, a formalism for dependency resolution that unifies the core semantics of diverse package managers. Through a series of formal reductions, we show how this core is expressive enough to model the diversity that real-world package managers employ in their dependency expression languages. By using the Package Calculus as the intermediate representation of dependencies, we enable translation between distinct package managers and resolution across ecosystems.

Systematic evaluation of Multimodal Large Language Models (MLLMs) is crucial for advancing Artificial General Intelligence (AGI). However, existing benchmarks remain insufficient for rigorously assessing their constrained reasoning capabilities. To bridge this gap, we introduce MapTab, a multimodal benchmark specifically designed to evaluate constrained reasoning in MLLMs via route planning tasks. MapTab requires MLLMs to perceive and ground visual cues from map images alongside route attributes (e.g., Time, Price) from structured tabular data. The benchmark encompasses two scenarios: Metromap, covering metro networks in 160 cities across 52 countries, and Travelmap, depicting 168 representative tourist attractions from 19 countries. In total, MapTab comprises 328 images, 196,800 route planning queries, and 3,936 QA queries, all incorporating 4 key constraints: Time, Price, Comfort, and Reliability. Extensive evaluations across 15 representative MLLMs reveal that current models face substantial challenges in constrained multimodal reasoning. Notably, under conditions of limited visual perception, multimodal collaboration often underperforms compared to unimodal approaches. We believe MapTab provides a challenging and realistic testbed to advance the systematic evaluation of MLLMs.

A fundamental problem in multi-task learning (MTL) is identifying groups of tasks that should be learned together. Since training MTL models for all possible combinations of tasks is prohibitively expensive for large task sets, a crucial component of efficient and effective task grouping is predicting whether a group of tasks would benefit from learning together, measured as per-task performance gain over single-task learning. In this paper, we propose ETAP (Ensemble Task Affinity Predictor), a scalable framework that integrates principled and data-driven estimators to predict MTL performance gains. First, we consider the gradient-based updates of shared parameters in an MTL model to measure the affinity between a pair of tasks as the similarity between the parameter updates based on these tasks. This linear estimator, which we call affinity score, naturally extends to estimating affinity within a group of tasks. Second, to refine these estimates, we train predictors that apply non-linear transformations and correct residual errors, capturing complex and non-linear task relationships. We train these predictors on a limited number of task groups for which we obtain ground-truth gain values via multi-task learning for each group. We demonstrate on benchmark datasets that ETAP improves MTL gain prediction and enables more effective task grouping, outperforming state-of-the-art baselines across diverse application domains.

Diffusion models have recently emerged as powerful priors for solving inverse problems. While computed tomography (CT) is theoretically a linear inverse problem, it poses many practical challenges. These include correlated noise, artifact structures, reliance on system geometry, and misaligned value ranges, which make the direct application of diffusion models more difficult than in domains like natural image generation. To systematically evaluate how diffusion models perform in this context and compare them with established reconstruction methods, we introduce DM4CT, a comprehensive benchmark for CT reconstruction. DM4CT includes datasets from both medical and industrial domains with sparse-view and noisy configurations. To explore the challenges of deploying diffusion models in practice, we additionally acquire a high-resolution CT dataset at a high-energy synchrotron facility and evaluate all methods under real experimental conditions. We benchmark ten recent diffusion-based methods alongside seven strong baselines, including model-based, unsupervised, and supervised approaches. Our analysis provides detailed insights into the behavior, strengths, and limitations of diffusion models for CT reconstruction. The real-world dataset is publicly available at zenodo.org/records/15420527, and the codebase is open-sourced at github.com/DM4CT/DM4CT.

Precise localization and recognition of flowers are crucial for advancing automated agriculture, particularly in plant phenotyping, crop estimation, and yield monitoring. This paper benchmarks several YOLO architectures such as YOLOv5s, YOLOv8n/s/m, and YOLOv12n for flower object detection under two annotation regimes: single-image single-bounding box (SISBB) and single-image multiple-bounding box (SIMBB). The FloralSix dataset, comprising 2,816 high-resolution photos of six different flower species, is also introduced. It is annotated for both dense (clustered) and sparse (isolated) scenarios. The models were evaluated using Precision, Recall, and Mean Average Precision (mAP) at IoU thresholds of 0.5 (mAP@0.5) and 0.5-0.95 (mAP@0.5:0.95). In SISBB, YOLOv8m (SGD) achieved the best results with Precision 0.956, Recall 0.951, mAP@0.5 0.978, and mAP@0.5:0.95 0.865, illustrating strong accuracy in detecting isolated flowers. With mAP@0.5 0.934 and mAP@0.5:0.95 0.752, YOLOv12n (SGD) outperformed the more complicated SIMBB scenario, proving robustness in dense, multi-object detection. Results show how annotation density, IoU thresholds, and model size interact: recall-optimized models perform better in crowded environments, whereas precision-oriented models perform best in sparse scenarios. In both cases, the Stochastic Gradient Descent (SGD) optimizer consistently performed better than alternatives. These density-sensitive sensors are helpful for non-destructive crop analysis, growth tracking, robotic pollination, and stress evaluation.

Targeted data selection has emerged as a crucial paradigm for efficient instruction tuning, aiming to identify a small yet influential subset of training examples for a specific target task. In practice, influence is often measured through the effect of an example on parameter updates. To make selection scalable, many approaches leverage optimizer statistics (e.g., Adam states) as an axis-aligned surrogate for update geometry (i.e., diagonal precondition), implicitly treating parameters as coordinate-wise independent. We show that this assumption breaks down in parameter-efficient fine-tuning (PEFT) methods such as LoRA. In this setting, the induced optimization geometry exhibits strong cross-parameter coupling with non-trivial off-diagonal interactions, while the task-relevant update directions are confined to a low-dimensional subspace. Motivated by this mismatch, we propose GIST (Gradient Isometric Subspace Transformation), a simple yet principled alternative that replaces axis-aligned scaling with robust subspace alignment. GIST recovers a task-specific subspace from validation gradients via spectral filtering (SVD), projects training gradients into this coupled subspace, and scores examples by their alignment with target directions.Extensive experiments have demonstrated that GIST matches or outperforms the state-of-the-art baseline with only 0.29% of the storage and 25% of the computational time under the same selection budget.

Real-time guardrails require evaluation that is accurate, cheap, and fast - yet today's default, LLM-as-a-judge (LLMAJ), is slow, expensive, and operationally non-deterministic due to multi-token generation. We present Luna-2, a novel architecture that leverages decoder-only small language models (SLMs) into a deterministic evaluation model to reliably compute complex task-specific LLMAJ metrics (e.g. toxicity, hallucination, tool selection quality, etc.) at an accuracy at par or higher than LLMAJ using frontier LLMs while drastically reducing the cost and latency of computation. Each metric is implemented as a lightweight LoRA/PEFT head on top of a shared SLM backbone, enabling hundreds of specialized metrics to run concurrently on a single GPU, deployable locally next to AI systems in a privacy-preserving and latency optimizing manner. Across content safety and hallucination benchmarks, Luna-2 matches the accuracy of state-of-the-art LLM-based evaluators while reducing inference cost by over 80x and latency by over 20x. In this paper, we outline the model architecture, training methodology and report real-world empirical results on accuracy, latency, and throughput results. In production, Luna-2 is protecting 100M+ AI sessions and processing over 100B tokens per month for our customers with eval cost savings of over $30M annually.

When training artificial intelligence (AI) to perform tasks, humans often care not only about whether a task is completed but also how it is performed. As AI agents tackle increasingly complex tasks, aligning their behavior with human-provided specifications becomes critical for responsible AI deployment. Reward design provides a direct channel for such alignment by translating human expectations into reward functions that guide reinforcement learning (RL). However, existing methods are often too limited to capture nuanced human preferences that arise in long-horizon tasks. Hence, we introduce Hierarchical Reward Design from Language (HRDL): a problem formulation that extends classical reward design to encode richer behavioral specifications for hierarchical RL agents. We further propose Language to Hierarchical Rewards (L2HR) as a solution to HRDL. Experiments show that AI agents trained with rewards designed via L2HR not only complete tasks effectively but also better adhere to human specifications. Together, HRDL and L2HR advance the research on human-aligned AI agents.

Despite their apparent diversity, modern machine learning methods can be reduced to a remarkably simple core principle: learning is achieved by continuously optimizing parameters to minimize or maximize a scalar objective function. This paradigm has been extraordinarily successful for well-defined tasks where goals are fixed and evaluation criteria are explicit. However, if artificial systems are to move toward true autonomy-operating over long horizons and across evolving contexts-objectives may become ill-defined, shifting, or entirely absent. In such settings, a fundamental question emerges: in the absence of an explicit objective function, how can a system determine whether its ongoing internal dynamics are productive or pathological? And how should it regulate structural change without external supervision? In this work, we propose a dynamical framework for learning without an explicit objective. Instead of minimizing external error signals, the system evaluates the intrinsic health of its own internal dynamics and regulates structural plasticity accordingly. We introduce a two-timescale architecture that separates fast state evolution from slow structural adaptation, coupled through an internally generated stress variable that accumulates evidence of persistent dynamical dysfunction. Structural modification is then triggered not continuously, but as a state-dependent event. Through a minimal toy model, we demonstrate that this stress-regulated mechanism produces temporally segmented, self-organized learning episodes without reliance on externally defined goals. Our results suggest a possible route toward autonomous learning systems capable of self-assessment and internally regulated structural reorganization.

Developers often extract methods to improve readability, understanding, and reuse, while inlining keeps logic in one block. Prior work based on static metrics has not shown clear differences between these practices, and the human side of comprehension and navigation remains underexplored. We investigate Inline Method vs. Extract Method refactorings using a dynamic approach: eye tracking while participants read and solve tasks. We analyze key code areas and compare visual effort and reading behavior (fixation duration and count, regressions, revisits), alongside time and attempts. We ran a controlled experiment with 32 Java novices, followed by short interviews. Each participant solved eight simple tasks across four programs presented in an inlined version and four in an extracted version. We also surveyed 58 additional novices for complementary quantitative and qualitative data. Results show that effects depend on task difficulty. In two tasks, method extraction improved performance and reduced visual effort, with time decreasing by up to 78.8% and regressions by 84.6%. For simpler tasks (e.g., square area), extraction hurt performance: time increased by up to 166.9% and regressions by 200%. Even with meaningful method names, novices often switched back and forth between call sites and extracted methods, increasing navigation and cognitive load. Preferences frequently favored extraction for readability and reuse, but did not always match measured performance. These findings suggest educators should be cautious about premature modularization for novices and highlight eye tracking as a useful complement to static metrics.

Machine-generated probability predictions are essential in modern classification tasks such as image classification. A model is well calibrated when its predicted probabilities correspond to observed event frequencies. Despite the need for multicategory recalibration methods, existing methods are limited to (i) comparing calibration between two or more models rather than directly assessing the calibration of a single model, (ii) requiring under-the-hood model access, e.g., accessing logit-scale predictions within the layers of a neural network, and (iii) providing output which is difficult for human analysts to understand. To overcome (i)-(iii), we propose Multicategory Linear Log Odds (MCLLO) recalibration, which (i) includes a likelihood ratio hypothesis test to assess calibration, (ii) does not require under-the-hood access to models and is thus applicable on a wide range of classification problems, and (iii) can be easily interpreted. We demonstrate the effectiveness of the MCLLO method through simulations and three real-world case studies involving image classification via convolutional neural network, obesity analysis via random forest, and ecology via regression modeling. We compare MCLLO to four comparator recalibration techniques utilizing both our hypothesis test and the existing calibration metric Expected Calibration Error to show that our method works well alone and in concert with other methods.

Reliable real estate price indicators are typically published at city level and low frequency, limiting their use for neighborhood-scale monitoring and long-horizon planning. We study whether sub-city price indices can be forecasted at weekly frequency by combining physical development signals from satellite radar with market narratives from news text. Using over 350,000 transactions from Dubai Land Department (2015-2025), we construct weekly price indices for 19 sub-city regions and evaluate forecasts from 2 to 34 weeks ahead. Our framework fuses regional transaction history with Sentinel-1 SAR backscatter, news sentiment combining lexical tone and semantic embeddings, and macroeconomic context. Results are strongly horizon dependent: at horizons up to 10 weeks, price history alone matches multimodal configurations, but beyond 14 weeks sentiment and SAR become critical. At long horizons (26-34 weeks), the full multimodal model reduces mean absolute error from 4.48 to 2.93 (35% reduction), with gains statistically significant across regions. Nonparametric learners consistently outperform deep architectures in this data regime. These findings establish benchmarks for weekly sub-city index forecasting and demonstrate that remote sensing and news sentiment materially improve predictability at strategically relevant horizons.

While significant progress has been made in automating various aspects of software development through coding agents, there is still significant room for improvement in their bug fixing capabilities. Debugging and investigation of runtime behavior remains largely a manual, developer-driven process. Popular coding agents typically rely on either static analysis of the code or iterative test-fix cycles, which is akin to trial and error debugging. We posit that there is a wealth of rich runtime information that developers routinely access while debugging code, which agents are currently deprived of due to design limitations. Despite how prevalent debuggers are in modern IDEs and command-line tools, they have surprisingly not made their way into coding agents. In this work, we introduce Debug2Fix, a novel framework that incorporates interactive debugging as a core component of a software engineering agent via a subagent architecture. We incorporate debuggers for Java and Python into our agent framework and evaluate against GitBug-Java and SWE-Bench-Live and achieve >20% improvement in performance compared to the baseline for certain models. Furthermore, using our framework, we're able to make weaker models like GPT-5 and Claude Haiku 4.5 match or exceed the performances of stronger models like Claude Sonnet 4.5, showing that better tool design is often just as important as switching to a more expensive model. Finally, we conduct systematic ablations demonstrating the importance of both the subagent architecture and debugger integration.

Large language model (LLM) inference performance is increasingly bottlenecked by the memory wall. While GPUs continue to scale raw compute throughput, they struggle to deliver scalable performance for memory bandwidth bound workloads. This challenge is amplified by emerging reasoning LLM applications, where long output sequences, low arithmetic intensity, and tight latency constraints demand significantly higher memory bandwidth. As a result, system utilization drops and energy per inference rises, highlighting the need for an optimized system architecture for scalable memory bandwidth. To address these challenges we present the Reasoning Processing Unit (RPU), a chiplet-based architecture designed to address the challenges of the modern memory wall. RPU introduces: (1) A Capacity-Optimized High-Bandwidth Memory (HBM-CO) that trades capacity for lower energy and cost; (2) a scalable chiplet architecture featuring a bandwidth-first power and area provisioning design; and (3) a decoupled microarchitecture that separates memory, compute, and communication pipelines to sustain high bandwidth utilization. Simulation results show that RPU performs up to 45.3x lower latency and 18.6x higher throughput over an H100 system at ISO-TDP on Llama3-405B.

The Solid Electrolyte Interphase (SEI) is critical to the performance of lithium-ion batteries, yet its analysis via Operando Infrared (IR) spectroscopy remains experimentally complex and expensive, which limits its accessibility for standard research facilities. To overcome this bottleneck, we formulate a novel task, Operando IR Prediction, which aims to forecast the time-resolved evolution of spectral ``fingerprints'' from a single static spectrum. To facilitate this, we introduce OpIRSpec-7K, the first large-scale operando dataset comprising 7,118 high-quality samples across 10 distinct battery systems, alongside OpIRBench, a comprehensive evaluation benchmark with carefully designed protocols. Addressing the limitations of standard spectrum, video, and sequence models in capturing voltage-driven chemical dynamics and complex composition, we propose Aligned Bi-stream Chemical Constraint (ABCC), an end-to-end physics-aware framework. It reformulates MeanFlow and introduces a novel Chemical Flow to explicitly model reaction trajectories, employs a two-stream disentanglement mechanism for solvent-SEI separation, and enforces physics and spectrum constraints such as mass conservation and peak shifts. ABCC significantly outperforms state-of-the-art static, sequential, and generative baselines. ABCC even generalizes to unseen systems and enables interpretable downstream recovery of SEI formation pathways, supporting AI-driven electrochemical discovery.

Despite the successes of deep learning in computer vision, difficulties persist in recognizing objects that have undergone group-symmetric transformations rarely seen during training$\unicode{x2013}$for example objects seen in unusual poses, scales, positions, or combinations thereof. Equivariant neural networks are a solution to the problem of generalizing across symmetric transformations, but require knowledge of transformations a priori. An alternative family of architectures proposes to learn equivariant operators in a latent space, from examples of symmetric transformations. Here, using simple datasets of rotated and translated noisy MNIST, we illustrate how such architectures can successfully be harnessed for out-of-distribution classification, thus overcoming the limitations of both traditional and equivariant networks. While conceptually enticing, we discuss challenges ahead on the path of scaling these architectures to more complex datasets.

Design-to-code translates high-fidelity UI designs into executable front-end implementations, but progress remains hard to compare due to inconsistent datasets, toolchains, and evaluation protocols. We introduce 1D-Bench, a benchmark grounded in real e-commerce workflows, where each instance provides a reference rendering and an exported intermediate representation that may contain extraction errors. 1D is short for one day, representing the efficient completion of design-to-code tasks in less than one day. Models take both as input, using the intermediate representation as structural cues while being evaluated against the reference rendering, which tests robustness to intermediate representation defects rather than literal adherence. 1D-Bench requires generating an executable React codebase under a fixed toolchain with an explicit component hierarchy, and defines a multi-round setting in which models iteratively apply component-level edits using execution feedback. Experiments on commercial and open-weight multimodal models show that iterative editing generally improves final performance by increasing rendering success and often improving visual similarity. We further conduct a pilot study on post-training with synthetic repair trajectories and reinforcement learning based editing, and observe limited and unstable gains that may stem from sparse terminal rewards and high-variance file-level updates.

Property-based testing (PBT) is a popular technique for establishing confidence in software, where users write properties -- i.e., executable specifications -- that can be checked many times in a loop by a testing framework. In modern PBT frameworks, properties are usually written in shallowly embedded domain-specific languages, and their definition is tightly coupled to the way they are tested. Such frameworks often provide convenient configuration options to customize aspects of the testing process, but users are limited to precisely what library authors had the prescience to allow for when developing the framework; if they want more flexibility, they may need to write a new framework from scratch. We propose a new, deeper language for properties based on a mixed embedding that we call deferred binding abstract syntax, which reifies properties as a data structure and decouples them from the property runners that execute them. We implement this language in Rocq and Racket, leveraging the power of dependent and dynamic types, respectively. Finally, we showcase the flexibility of this new approach by rapidly prototyping a variety of property runners, highlighting domain-specific testing improvements that can be unlocked by more programmable testing.

Psychological corpora in NLP are collections of texts used to analyze human psychology, emotions, and mental health. These texts allow researchers to study psychological constructs, detect mental health issues and analyze emotional language. However, mental health data can be difficult to collect correctly from social media, due to suppositions made by the collectors. A more pragmatic strategy involves gathering data through open-ended questions and then assessing this information with self-report screening surveys. This method was employed successfully for English, a language with a lot of psychological NLP resources. However, this cannot be stated for Romanian, which currently has no open-source mental health corpus. To address this gap, we have created the first corpus for depression and anxiety in Romanian, by utilizing a form with 6 open-ended questions along with the standardized PHQ-9 and GAD-7 screening questionnaires. Consisting of the texts of 205 respondents and although it may seem small, PsihoRo is a first step towards understanding and analyzing texts regarding the mental health of the Romanian population. We employ statistical analysis, text analysis using Romanian LIWC, emotion detection and topic modeling to show what are the most important features of this newly introduced resource to the NLP community.

Manufacturing automation in process planning, inspection planning, and digital-thread integration depends on a unified specification that binds the geometric features of a 3D CAD model to the geometric dimensioning and tolerancing (GD&T) callouts, datum definitions, and surface requirements carried by the corresponding 2D engineering drawing. Although Model-Based Definition (MBD) allows such specifications to be embedded directly in 3D models, 2D drawings remain the primary carrier of manufacturing intent in automotive, aerospace, shipbuilding, and heavy-machinery industries. Correctly linking drawing annotations to the corresponding 3D features is difficult because of contextual ambiguity, repeated feature patterns, and the need for transparent and traceable decisions. This paper presents a deterministic-first, context-aware framework that maps 2D drawing entities to 3D CAD features to produce a unified manufacturing specification. Drawing callouts are first semantically enriched and then scored against candidate features using an interpretable metric that combines type compatibility, tolerance-aware dimensional agreement, and conservative context consistency, along with engineering-domain heuristics. When deterministic scoring cannot resolve an ambiguity, the system escalates to multimodal and constrained large-language-model reasoning, followed by a single human-in-the-loop (HITL) review step. Experiments on 20 real CAD-drawing pairs achieve a mean precision of 83.67%, recall of 90.46%, and F1 score of 86.29%. An ablation study shows that each pipeline component contributes to overall accuracy, with the full system outperforming all reduced variants. By prioritizing deterministic rules, clear decision tracking, and retaining unresolved cases for human review, the framework provides a practical foundation for downstream manufacturing automation in real-world industrial environments.

Large Language Models (LLMs) increasingly serve as consumers of API specifications, whether for code generation, autonomous agent interaction, or API-assisted reasoning. The de facto standard for API description, OpenAPI, was designed for documentation tools and code generators, resulting in substantial token overhead when used as LLM context. We present LAPIS (Lightweight API Specification for Intelligent Systems), a domain-specific format optimized for LLM consumption that preserves the semantic information necessary for API reasoning while minimizing token usage. Through empirical evaluation against five real-world production API specifications including GitHub (1,080 endpoints), Twilio (197 endpoints), DigitalOcean (545 endpoints), Petstore, and HTTPBin we demonstrate an average token reduction of 85.5% compared to OpenAPI YAML and 88.6% compared to OpenAPI JSON, measured with the cl100k_base tokenizer. LAPIS introduces domain-specific structural innovations, including centralized error definitions, webhook trigger conditions, structured rate limit descriptions, and operation flow declarations information that OpenAPI either duplicates redundantly or cannot represent at all. The format is fully convertible from OpenAPI 3.x via an automated converter, requires no special parser for LLM consumption, and is released as an open specification under CC BY 4.0.

We present Rodent-Bench, a novel benchmark designed to evaluate the ability of Multimodal Large Language Models (MLLMs) to annotate rodent behaviour footage. We evaluate state-of-the-art MLLMs, including Gemini-2.5-Pro, Gemini-2.5-Flash and Qwen-VL-Max, using this benchmark and find that none of these models perform strongly enough to be used as an assistant for this task. Our benchmark encompasses diverse datasets spanning multiple behavioral paradigms including social interactions, grooming, scratching, and freezing behaviors, with videos ranging from 10 minutes to 35 minutes in length. We provide two benchmark versions to accommodate varying model capabilities and establish standardized evaluation metrics including second-wise accuracy, macro F1, mean average precision, mutual information, and Matthew's correlation coefficient. While some models show modest performance on certain datasets (notably grooming detection), overall results reveal significant challenges in temporal segmentation, handling extended video sequences, and distinguishing subtle behavioral states. Our analysis identifies key limitations in current MLLMs for scientific video annotation and provides insights for future model development. Rodent-Bench serves as a foundation for tracking progress toward reliable automated behavioral annotation in neuroscience research.

Accurate taxonomic identification is the cornerstone of biodiversity monitoring and agricultural management, particularly for the hyper-diverse superfamily Ichneumonoidea. Comprising the families Ichneumonidae and Braconidae, these parasitoid wasps are ecologically critical for regulating insect populations, yet they remain one of the most taxonomically challenging groups due to their cryptic morphology and vast number of undescribed species. To address the scarcity of robust digital resources for these key groups, we present a curated image dataset designed to advance automated identification systems. The dataset contains 3,556 high-resolution images, primarily focused on Neotropical Ichneumonidae and Braconidae, while also including supplementary families such as Andrenidae, Apidae, Bethylidae, Chrysididae, Colletidae, Halictidae, Megachilidae, Pompilidae, and Vespidae to improve model robustness. Crucially, a subset of 1,739 images is annotated in COCO format, featuring multi-class bounding boxes for the full insect body, wing venation, and scale bars. This resource provides a foundation for developing computer vision models capable of identifying these families.

Jupyter notebooks are widely used for machine learning (ML) development due to their support for interactive and iterative experimentation. However, ML notebooks are highly prone to bugs, with crashes being among the most disruptive. Despite their practical importance, systematic methods for crash detection and diagnosis in ML notebooks remain largely unexplored. We present CRANE-LLM, a novel approach that augments large language models (LLMs) with structured runtime information extracted from the notebook kernel state to detect and diagnose crashes before executing a target cell. Given previously executed cells and a target cell, CRANE-LLM combines static code context with runtime information, including object types, tensor shapes, and data attributes, to predict whether the target cell will crash (detection) and explain the underlying cause (diagnosis). We evaluate CRANE-LLM on JunoBench, a benchmark of 222 ML notebooks comprising 111 pairs of crashing and corresponding non-crashing notebooks across multiple ML libraries and crash root causes. Across three state-of-the-art LLMs (Gemini, Qwen, and GPT-5), runtime information improves crash detection and diagnosis by 7-10 percentage points in accuracy and 8-11 in F1-score, with larger gains for diagnosis. Improvements vary across ML libraries, crash causes, and LLMs, and depends on the integration of complementary categories of runtime information.

Generative models are increasingly used to improve the quality of medical imaging, such as reconstruction of magnetic resonance images and computed tomography. However, it is well-known that such models are susceptible to hallucinations: they may insert features into the reconstructed image which are not actually present in the original image. In a medical setting, such hallucinations may endanger patient health as they can lead to incorrect diagnoses. In this work, we aim to quantify the extent to which state-of-the-art generative models suffer from hallucinations in the context of magnetic resonance image reconstruction. Specifically, we craft adversarial perturbations resembling random noise for the unprocessed input images which induce hallucinations when reconstructed using a generative model. We perform this evaluation on the brain and knee images from the fastMRI data set using UNet and end-to-end VarNet architectures to reconstruct the images. Our results show that these models are highly susceptible to small perturbations and can be easily coaxed into producing hallucinations. This fragility may partially explain why hallucinations occur in the first place and suggests that a carefully constructed adversarial training routine may reduce their prevalence. Moreover, these hallucinations cannot be reliably detected using traditional image quality metrics. Novel approaches will therefore need to be developed to detect when hallucinations have occurred.

Voice-based digital biomarkers can enable scalable, non-invasive screening and monitoring of Parkinson's disease (PD) and Amyotrophic Lateral Sclerosis (ALS). However, models trained on one cohort or device often fail on new acquisition settings due to cross-device and cross-cohort domain shift. This challenge is amplified in real-world scenarios with partial-label mismatch, where datasets may contain different disease labels and only partially overlap in class space. In addition, voice-based models may exploit demographic cues, raising concerns about gender-related unfairness, particularly when deployed across heterogeneous cohorts. To tackle these challenges, we propose a hybrid framework for unified three-class (healthy/PD/ALS) cross-domain voice classification from partially overlapping cohorts. The method combines style-based domain generalization with conditional adversarial alignment tailored to partial-label settings, reducing negative transfer. An additional adversarial gender branch promotes gender-invariant representations. We conduct a comprehensive evaluation across four heterogeneous sustained-vowel datasets, spanning distinct acquisition settings and devices, under both domain generalization and unsupervised domain adaptation protocols. The proposed approach is compared against twelve state-of-the-art machine learning and deep learning methods, and further evaluated through three targeted ablations, providing the first cross-cohort benchmark and end-to-end domain-adaptive framework for unified healthy/PD/ALS voice classification under partial-label mismatch and fairness constraints. Across all experimental settings, our method consistently achieves the best external generalization over the considered evaluation metrics, while maintaining reduced gender disparities. Notably, no competing method shows statistically significant gains in external performance.

Large Language Models (LLMs) have shown promise for program translation, particularly for migrating systems code to memory-safe languages such as Rust. However, existing approaches struggle when source programs depend on external libraries: LLMs frequently hallucinate non-existent target APIs and fail to generate call-enabling imports; moreover, validating semantic equivalence is challenging when the code manipulates opaque, library-defined types. We present a translation and validation framework for translating Go projects with external dependencies to Rust. Our approach combines (i) a retrieval mechanism that maps Go library APIs to Rust APIs, and (ii) a cross-language validation pipeline that establishes language interoperability in the presence of opaque library types by synthesising adapters exclusively from public library APIs, prior to validating I/O equivalence. We evaluate our system on six real-world Go repositories with non-trivial external dependencies. Our approach significantly increases both the compilation and equivalence success rate (up to 100% in the most dependency-heavy case; approx. 2x on average) by enabling validated translation that manipulate opaque, library-defined types.

Following the rise of large foundation models, Vision-Language-Action models (VLAs) emerged, leveraging strong visual and language understanding for general-purpose policy learning. Yet, the current VLA landscape remains fragmented and exploratory. Although many groups have proposed their own VLA models, inconsistencies in training protocols and evaluation settings make it difficult to identify which design choices truly matter. To bring structure to this evolving space, we reexamine the VLA design space under a unified framework and evaluation setup. Starting from a simple VLA baseline similar to RT-2 and OpenVLA, we systematically dissect design choices along three dimensions: foundational components, perception essentials, and action modelling perspectives. From this study, we distill 12 key findings that together form a practical recipe for building strong VLA models. The outcome of this exploration is a simple yet effective model, VLANeXt. VLANeXt outperforms prior state-of-the-art methods on the LIBERO and LIBERO-plus benchmarks and demonstrates strong generalization in real-world experiments. We will release a unified, easy-to-use codebase that serves as a common platform for the community to reproduce our findings, explore the design space, and build new VLA variants on top of a shared foundation.

The energy management problem in the context of smart grids is inherently complex due to the interdependencies among diverse system components. Although Reinforcement Learning (RL) has been proposed for solving Optimal Power Flow (OPF) problems, the requirement for iterative interaction with an environment often necessitates computationally expensive simulators, leading to significant sample inefficiency. In this study, these challenges are addressed through the use of Physics-Informed Neural Networks (PINNs), which can replace conventional and costly smart grid simulators. The RL policy learning process is enhanced so that convergence can be achieved in a fraction of the time required by the original environment. The PINN-based surrogate is compared with other benchmark data-driven surrogate models. By incorporating knowledge of the underlying physical laws, the results show that the PINN surrogate is the only approach considered in this context that can obtain a strong RL policy even without access to samples from the true simulator. The results demonstrate that using PINN surrogates can accelerate training by 50% compared to RL training without a surrogate. This approach enables the rapid generation of performance scores similar to those produced by the original simulator.

Audio-visual continual test-time adaptation involves continually adapting a source audio-visual model at test-time, to unlabeled non-stationary domains, where either or both modalities can be distributionally shifted, which hampers online cross-modal learning and eventually leads to poor accuracy. While previous works have tackled this problem, we find that SOTA methods suffer from catastrophic forgetting, where the model's performance drops well below the source model due to continual parameter updates at test-time. In this work, we first show that adapting only the modality fusion layer to a target domain not only improves performance on that domain but can also enhance performance on subsequent domains. Based on this strong cross-task transferability of the fusion layer's parameters, we propose a method, $\texttt{AV-CTTA}$, that improves test-time performance of the models without access to any source data. Our approach works by using a selective parameter retrieval mechanism that dynamically retrieves the best fusion layer parameters from a buffer using only a small batch of test data. These parameters are then integrated into the model, adapted to the current test distribution, and saved back for future use. Extensive experiments on benchmark datasets involving unimodal and bimodal corruptions show our proposed $\texttt{AV-CTTA}$ significantly outperforms existing methods while minimizing catastrophic forgetting.

Current audio-visual large language models (AV-LLMs) are predominantly restricted to 2D perception, relying on RGB video and monaural audio. This design choice introduces a fundamental dimensionality mismatch that precludes reliable source localization and spatial reasoning in complex 3D environments. We address this limitation by presenting JAEGER, a framework that extends AV-LLMs to 3D space, to enable joint spatial grounding and reasoning through the integration of RGB-D observations and multi-channel first-order ambisonics. A core contribution of our work is the neural intensity vector (Neural IV), a learned spatial audio representation that encodes robust directional cues to enhance direction-of-arrival estimation, even in adverse acoustic scenarios with overlapping sources. To facilitate large-scale training and systematic evaluation, we propose SpatialSceneQA, a benchmark of 61k instruction-tuning samples curated from simulated physical environments. Extensive experiments demonstrate that our approach consistently surpasses 2D-centric baselines across diverse spatial perception and reasoning tasks, underscoring the necessity of explicit 3D modelling for advancing AI in physical environments. Our source code, pre-trained model checkpoints and datasets will be released upon acceptance.

Synthetic images are increasingly used to augment object-detection training sets, but reliably evaluating a synthetic dataset before training remains difficult: standard global generative metrics (e.g., FID) often do not predict downstream detection mAP. We present a controlled evaluation of synthetic augmentation for YOLOv11 across three single-class detection regimes -- Traffic Signs (sparse/near-saturated), Cityscapes Pedestrian (dense/occlusion-heavy), and COCO PottedPlant (multi-instance/high-variability). We benchmark six GAN-, diffusion-, and hybrid-based generators over augmentation ratios from 10% to 150% of the real training split, and train YOLOv11 both from scratch and with COCO-pretrained initialization, evaluating on held-out real test splits (mAP@0.50:0.95). For each dataset-generator-augmentation configuration, we compute pre-training dataset metrics under a matched-size bootstrap protocol, including (i) global feature-space metrics in both Inception-v3 and DINOv2 embeddings and (ii) object-centric distribution distances over bounding-box statistics. Synthetic augmentation yields substantial gains in the more challenging regimes (up to +7.6% and +30.6% relative mAP in Pedestrian and PottedPlant, respectively) but is marginal in Traffic Signs and under pretrained fine-tuning. To separate metric signal from augmentation quantity, we report both raw and augmentation-controlled (residualized) correlations with multiple-testing correction, showing that metric-performance alignment is strongly regime-dependent and that many apparent raw associations weaken after controlling for augmentation level.

Grokking -- the abrupt transition from memorization to generalization long after near-zero training loss -- has been studied mainly in single-task settings. We extend geometric analysis to multi-task modular arithmetic, training shared-trunk Transformers on dual-task (mod-add + mod-mul) and tri-task (mod-add + mod-mul + mod-sq) objectives across a systematic weight decay sweep. Five consistent phenomena emerge. (1) Staggered grokking order: multiplication generalizes first, followed by squaring, then addition, with consistent delays across seeds. (2) Universal integrability: optimization trajectories remain confined to an empirically invariant low-dimensional execution manifold; commutator defects orthogonal to this manifold reliably precede generalization. (3) Weight decay phase structure: grokking timescale, curvature depth, reconstruction threshold, and defect lead covary systematically with weight decay, revealing distinct dynamical regimes and a sharp no-decay failure mode. (4) Holographic incompressibility: final solutions occupy only 4--8 principal trajectory directions yet are distributed across full-rank weights and destroyed by minimal perturbations; SVD truncation, magnitude pruning, and uniform scaling all fail to preserve performance. (5) Transverse fragility and redundancy: removing less than 10% of orthogonal gradient components eliminates grokking, yet dual-task models exhibit partial recovery under extreme deletion, suggesting redundant center manifolds enabled by overparameterization. Together, these results support a dynamical picture in which multi-task grokking constructs a compact superposition subspace in parameter space, with weight decay acting as compression pressure and excess parameters supplying geometric redundancy in optimization pathways.

Continuous stress forecasting could potentially contribute to lifestyle interventions. This paper presents a novel, explainable, and individualized approach for stress prediction using physiological data from consumer-grade smartwatches. We develop a time series forecasting model that leverages multivariate features, including heart rate variability, activity patterns, and sleep metrics, to predict stress levels across 16 temporal horizons (History window: 3, 5, 7, 9 days; forecasting window: 1, 3, 5, 7 days). Our evaluation involves 16 participants monitored for 10-15 weeks. We evaluate our approach across 16 participants, comparing against state-of-the-art time series models (Informer, TimesNet, PatchTST) and traditional baselines (CNN, LSTM, CNN-LSTM) across multiple temporal horizons. Our model achieved performance with an MSE of 0.053, MAE of 0.190, and RMSE of 0.226 in optimal settings (5-day input, 1-day prediction). A comparison with the baseline models shows that our model outperforms TimesNet, PatchTST, CNN-LSTM, LSTM, and CNN under all conditions, representing improvements of 36.9%, 25.5%, and 21.5% over the best baseline. According to the explanability analysis, sleep metrics are the most dominant and consistent stress predictors (importance: 1.1, consistency: 0.9-1.0), while activity features exhibit high inter-participant variability (0.1-0.2). Most notably, the model captures individual-specific patterns where identical features can have opposing effects across users, validating its personalization capabilities. These findings establish that consumer wearables, combined with adaptive and interpretable deep learning, can deliver relevant stress assessment adapted to individual physiological responses, providing a foundation for scalable, continuous, explainable mental health monitoring in real-world settings.

Providing timely, rubric-aligned feedback on student-drawn diagrams is a persistent challenge in STEM education. While large multimodal models (LMMs) can jointly parse images and generate explanations, their tendency to hallucinate undermines trust in classroom deployments. We present Sketch2Feedback, a grammar-in-the-loop framework that decomposes the problem into four stages -- hybrid perception, symbolic graph construction, constraint checking, and constrained VLM feedback -- so that the language model verbalizes only violations verified by an upstream rule engine. We evaluate on two synthetic micro-benchmarks, FBD-10 (free-body diagrams) and Circuit-10 (circuit schematics), each with 500 images spanning standard and hard noise augmentation tiers, comparing our pipeline against end-to-end LMMs (LLaVA-1.5-7B, Qwen2-VL-7B), a vision-only detector, a YOLOv8-nano learned detector, and an ensemble oracle. On n=100 test samples per benchmark with 95% bootstrap CIs, results are mixed and instructive: Qwen2-VL-7B achieves the highest micro-F1 on both FBDs (0.570) and circuits (0.528), but with extreme hallucination rates (0.78, 0.98). An ensemble oracle that selects the best prediction per sample reaches F1=0.556 with hallucination 0.320 on FBDs, demonstrating exploitable complementarity between grammar and end-to-end approaches. Confidence thresholding at tau=0.7 reduces circuit hallucination from 0.970 to 0.880 with no F1 loss. Hard noise augmentation reveals domain-dependent robustness: FBD detection is resilient while circuit detection degrades sharply. An LLM-as-judge evaluation confirms that the grammar pipeline produces more actionable circuit feedback (4.85/5) than the end-to-end LMM (3.11/5). We release all code, datasets, and evaluation scripts.

Traditional approaches to measuring visual exploratory behavior in soccer rely on counting visual exploratory actions (VEAs) based on rapid head movements exceeding 125°/s, but this method suffer from player position bias (i.e., a focus on central midfielders), annotation challenges, binary measurement constraints (i.e., a player is scanning, or not), lack the power to predict relevant short-term in-game future success, and are incompatible with fundamental soccer analytics models such as pitch control. This research introduces a novel formulaic continuous stochastic vision layer to quantify players' visual perception from pose-enhanced spatiotemporal tracking. Our probabilistic field-of-view and occlusion models incorporate head and shoulder rotation angles to create speed-dependent vision maps for individual players in a two-dimensional top-down plane. We combine these vision maps with pitch control and pitch value surfaces to analyze the awaiting phase (when a player is awaiting the ball to arrive after a pass for a teammate) and their subsequent on-ball phase. We demonstrate that aggregated visual metrics - such as the percentage of defended area observed while awaiting a pass - are predictive of controlled pitch value gained at the end of dribbling actions using 32 games of synchronized pose-enhanced tracking data and on-ball event data from the 2024 Copa America. This methodology works regardless of player position, eliminates manual annotation requirements, and provides continuous measurements that seamlessly integrate into existing soccer analytics frameworks. To further support the integration with existing soccer analytics frameworks we open-source the tools required to make these calculations.

Content safety teams need metrics that reflect what users actually experience, not only what is reported. We study prevalence: the fraction of user views (impressions) that went to content violating a given policy on a given day. Accurate prevalence measurement is challenging because violations are often rare and human labeling is costly, making frequent, platform-representative studies slow. We present a design-based measurement system that (i) draws daily probability samples from the impression stream using ML-assisted weights to concentrate label budget on high-exposure and high-risk content while preserving unbiasedness, (ii) labels sampled items with a multimodal LLM governed by policy prompts and gold-set validation, and (iii) produces design-consistent prevalence estimates with confidence intervals and dashboard drilldowns. A key design goal is one global sample with many pivots: the same daily sample supports prevalence by surface, viewer geography, content age, and other segments through post-stratified estimation. We describe the statistical estimators, variance and confidence interval construction, label-quality monitoring, and an engineering workflow that makes the system configurable across policies.

Partial differential equations (PDEs) form a central component of scientific computing. Among recent advances in deep learning, evolutionary neural networks have been developed to successively capture the temporal dynamics of time-dependent PDEs via parameter evolution. The parameter updates are obtained by solving a linear system derived from the governing equation residuals at each time step. However, strong-form evolutionary approaches can yield ill-conditioned linear systems due to pointwise residual discretization, and their computational cost scales unfavorably with the number of training samples. To address these limitations, we propose a weak-form evolutionary Kolmogorov-Arnold Network (KAN) for the scalable and accurate prediction of PDE solutions. We decouple the linear system size from the number of training samples through the weak formulation, leading to improved scalability compared to strong-form approaches. We also rigorously enforce boundary conditions by constructing the trial space with boundary-constrained KANs to satisfy Dirichlet and periodic conditions, and by incorporating derivative boundary conditions directly into the weak formulation for Neumann conditions. In conclusion, the proposed weak-form evolutionary KAN framework provides a stable and scalable approach for PDEs and contributes to scientific machine learning with potential relevance to future engineering applications.

As Large Language Models (LLMs) are increasingly integrated into automated decision-making pipelines, specifically within Human Resources (HR), the security implications of Indirect Prompt Injection (IPI) become critical. While a prevailing hypothesis posits that "Reasoning" or "Chain-of-Thought" Models possess safety advantages due to their ability to self-correct, emerging research suggests these capabilities may enable more sophisticated alignment failures. This qualitative Red-Teaming case study challenges the safety-through-reasoning premise using the Qwen 3 30B architecture. By subjecting both a standard instruction-tuned model and a reasoning-enhanced model to a "Trojan Horse" curriculum vitae, distinct failure modes are observed. The results suggest a complex trade-off: while the Standard Model resorted to brittle hallucinations to justify simple attacks and filtered out illogical constraints in complex scenarios, the Reasoning Model displayed a dangerous duality. It employed advanced strategic reframing to make simple attacks highly persuasive, yet exhibited "Meta-Cognitive Leakage" when faced with logically convoluted commands. This study highlights a failure mode where the cognitive load of processing complex adversarial instructions causes the injection logic to be unintentionally printed in the final output, rendering the attack more detectable by humans than in Standard Models.

Activation steering, or representation engineering, offers a lightweight approach to align large language models (LLMs) by manipulating their internal activations at inference time. However, current methods suffer from two key limitations: (i) the lack of a unified theoretical framework for guiding the design of steering directions, and (ii) an over-reliance on one-step steering that fail to capture complex patterns of activation distributions. In this work, we propose a unified ordinary differential equations (ODEs)-based theoretical framework for activation steering in LLM alignment. We show that conventional activation addition can be interpreted as a first-order approximation to the solution of an ODE. Based on this ODE perspective, identifying a steering direction becomes equivalent to designing a barrier function from control theory. Derived from this framework, we introduce ODESteer, a kind of ODE-based steering guided by barrier functions, which shows empirical advancement in LLM alignment. ODESteer identifies steering directions by defining the barrier function as the log-density ratio between positive and negative activations, and employs it to construct an ODE for multi-step and adaptive steering. Compared to state-of-the-art activation steering methods, ODESteer achieves consistent empirical improvements on diverse LLM alignment benchmarks, a notable $5.7\%$ improvement over TruthfulQA, $2.5\%$ over UltraFeedback, and $2.4\%$ over RealToxicityPrompts. Our work establishes a principled new view of activation steering in LLM alignment by unifying its theoretical foundations via ODEs, and validating it empirically through the proposed ODESteer method.

Task Arithmetic yields a modular, scalable way to adapt foundation models. Combining multiple task vectors, however, can lead to cross-task interference, causing representation drift and degraded performance. Representation drift regularization provides a natural remedy to disentangle task vectors; however, existing approaches typically require external task data, conflicting with modularity and data availability constraints (e.g., privacy requirements). We propose a dataless approach by framing regularization against representation drift as a curvature matrix approximation problem. This allows us to leverage well-established techniques; in particular, we adopt Kronecker-Factored Approximate Curvature and obtain a practical regularizer that achieves state-of-the-art results in task addition and negation. Our method has constant complexity in the number of tasks and promotes robustness to task vector rescaling, eliminating the need for held-out tuning.

Modern compilers optimize programs through a sequence of modular passes over intermediate representations (IR). While this pass-by-pass paradigm offers engineering benefits, it suffers from a pass coordination problem: locally beneficial transformations may block more profitable optimizations in later stages. This limitation stems from the lack of an explicit notion of optimization intent, defined as a holistic strategy for coordinating multiple transformations toward a global performance objective. Recent LLM-based approaches formulate IR optimization as an end-to-end generation task, thereby avoiding the traditional pass-by-pass structure. However, optimization intent remains implicit in these methods, forcing models to jointly infer optimization strategy and generate low-level transformations, which limits both correctness and performance. We propose IntOpt, the first intent-driven IR optimizer that explicitly separates high-level optimization intent from low-level analysis and transformation. IntOpt organizes IR optimization into three stages: intent formulation, intent refinement, and intent realization, enabling globally coordinated transformations. Experiments show that IntOpt achieves 90.5% verified correctness and 2.660x average speedup on 200-program test set, outperforming state-of-the-art LLM-based optimizers in both correctness and performance, and surpassing modern compiler with the -O3 option on 37 benchmarks with speedups of up to 272.60x.

Zeroth-order (ZO) optimization provides a gradient-free alternative to first-order (FO) methods by estimating gradients via finite differences of function evaluations, and has recently emerged as a memory-efficient paradigm for fine-tuning large-scale models by avoiding backpropagation. However, ZO optimization has a fundamental tension between accuracy and query efficiency. In this work, we show that ZO optimization can be substantially improved by unifying two complementary principles: (i) a projection-based subspace view that reduces gradient estimation variance by exploiting the intrinsic low-rank structure of model updates, and (ii) Muon-style spectral optimization that applies gradient orthogonalization to extract informative spectral structure from noisy ZO gradients. These findings form a unified framework of subspace gradient orthogonalization, which we instantiate in a new method, ZO-Muon, admitting a natural interpretation as a low-rank Muon optimizer in the ZO setting. Extensive experiments on large language models (LLMs) and vision transformers (ViTs) demonstrate that ZO-Muon significantly accelerates convergence and achieves a win-win improvement in accuracy and query/runtime efficiency. Notably, compared to the popular MeZO baseline, ZO-Muon requires only 24.7% of the queries to reach the same SST-2 performance for LLM fine-tuning, and improves accuracy by 25.1% on ViT-B fine-tuning on CIFAR-100.

Large Reasoning Models (LRMs) exhibit strong performance, yet often produce rationales that sound plausible but fail to reflect their true decision process, undermining reliability and trust. We introduce a formal framework for reasoning faithfulness, defined by two testable conditions: stance consistency (a coherent stance linking reasoning to answer) and causal influence (the stated reasoning causally drives the answer under output-level interventions), explicitly decoupled from accuracy. To operationalize this, we present RFEval, a benchmark of 7,186 instances across seven tasks that probes faithfulness via controlled, output-level counterfactual interventions. Evaluating twelve open-source LRMs, we find unfaithfulness in 49.7% of outputs, predominantly from stance inconsistency. Failures are concentrated in brittle, convergent domains such as math and code, and correlate more with post-training regimes than with scale: within-family ablations indicate that adding current RL-style objectives on top of supervised fine-tuning can reduce reasoning faithfulness, even when accuracy is maintained. Crucially, accuracy is neither a sufficient nor a reliable proxy for faithfulness: once controlling for model and task, the accuracy-faithfulness link is weak and statistically insignificant. Our work establishes a rigorous methodology for auditing LRM reliability and shows that trustworthy AI requires optimizing not only for correct outcomes but also for the structural integrity of the reasoning process. Our code and dataset can be found at project page: https://aidaslab.github.io/RFEval/

Coordinating actions is the most fundamental form of cooperation in multi-agent reinforcement learning (MARL). Successful decentralized decision-making often depends not only on good individual actions, but on selecting compatible actions across agents to synchronize behavior, avoid conflicts, and satisfy global constraints. In this paper, we propose Action Graph Policies (AGP), that model dependencies among agents' available action choices. It constructs, what we call, \textit{coordination contexts}, that enable agents to condition their decisions on global action dependencies. Theoretically, we show that AGPs induce a strictly more expressive joint policy compared to fully independent policies and can realize coordinated joint actions that are provably more optimal than greedy execution even from centralized value-decomposition methods. Empirically, we show that AGP achieves 80-95\% success on canonical coordination tasks with partial observability and anti-coordination penalties, where other MARL methods reach only 10-25\%. We further demonstrate that AGP consistently outperforms these baselines in diverse multi-agent environments.

Two-dimensional superconductors offer an excellent platform for the study of vortex matter due to their low superfluid stiffness and inability to effectively screen applied magnetic fields. Here we explore vortices in a two-dimensional superconductor formed at the surface of the complex oxide KTaO$_3$. Multiple regimes of vortex-mediated transport are identified and studied, revealing switching behaviour attributed to nucleation of individual vortices. Analysis of this regime allows us to identify the quantum tunneling of vortices, which transitions to thermally activated behaviour at elevated temperatures. Magnetic field dependence reveals rich histograms of the switching currents which we attribute to different configurations of pinned vortices.

The performance of quantum annealing for combinatorial optimization is fundamentally limited by the minimum energy gap $Δ$ encountered at quantum phase transitions. We investigate the scaling of $Δ$ with system size $N$ for two paradigmatic quantum spin-glass models: the two-dimensional Edwards-Anderson (2D-EA) and the all-to-all Sherrington-Kirkpatrick (SK) models. Utilizing a newly proposed unbiased energy-gap estimator for continuous-time projection quantum Monte Carlo simulations, complemented by high-performance sparse eigenvalue solvers, we characterize the gap distributions across disorder realizations. It is found that, in the 2D-EA case, the inverse-gap distribution develops a fat tail with infinite variance as $N$ increases. This indicates that the unfavorable super-algebraic scaling of $Δ$, recently reported for binary couplings [Nature 631, 749 (2024)], persists for the Gaussian disorder considered here, pointing to a universal feature of 2D spin glasses. Conversely, the SK model retains a finite-variance distribution, with the disorder-averaged gap following a rather slow power law, close to $Δ\propto N^{-1/3}$. This finding provides a promising outlook for the potential efficiency of quantum annealers for optimization problems with dense connectivity.

Biological tissues have been observed to display emergent fluid-like properties, owing to physical interactions between cells. However, it remains unclear in general how these fluid-like properties affect tissue structure and function. Here, we are motivated by recent experiments in which cell aggregates were observed to behave as active droplets during collective migration along chemical gradients, or chemotaxis. To understand this process, we develop a minimal model of a growing thin active droplet driven by a self-generated chemical gradient. In broad agreement with the experiments, dynamic simulations reveal that chemotacting droplets exhibit proliferation-driven morphological transitions. To fully characterise these transitions, we perform a multiple scales analysis to show that the droplet dynamics follow a sequence of travelling wave solutions defined by a nonlinear eigenvalue problem parametrised by the slowly increasing droplet volume. Our analysis reveals that morphological transitions can occur continuously or through a discontinuous bifurcation. Further asymptotic analysis of the travelling wave problem reveals that these morphological transitions arise from exponentially small ("beyond-all-orders") asymptotic terms that originate from the rear and front contact lines. Moreover, we show that the nature of the transitions is fully determined by two key dimensionless parameters, which quantify the internal stress balance within the droplet and the strength of the coupling between the droplet migration dynamics and the external chemical field. Overall, our results provide a complete characterisation of the morphodynamics of a class of migrating active thin droplets, with implications in a range of biological systems where cell aggregates exhibit fluid-like behaviour.

We present a method to characterize the distribution of length-scales of finite, disordered patterns with, on average, radial symmetry. This method makes it possible to quantify the distribution of characteristic length scales in cases where the conventional "linear" chord method does not work. We show that the method can clearly distinguish regular patterns, patterns that are formed by diffusion-limited aggregation, and patterns that form during the slow drying of confined, colloid-laden droplets, explained by Beechey-Newman et al.1 We also introduce a method to find the centre-point of these finite patterns, without assuming a full connectivity in the pattern. The method should be widely applicable to other, finite quasi-two-dimensional patterns.

The anomalous mole fraction effect (AMFE) is widely regarded as a hallmark of calcium versus monovalent ion selectivity in negatively charged pores. While AMFE is well understood in highly cation-selective narrow ion channels, its microscopic origin in wide synthetic nanopores, where anions may also contribute to transport, remains less clear. Here, we use a reduced Nernst-Planck + Local Equilibrium Monte Carlo framework to study ionic transport in a negatively charged PET nanopore, with particular emphasis on how the modeling of surface carboxyl (COO$^{-}$) groups influences charge inversion, ionic currents, and AMFE. We systematically compare fixed point-charge models and explicit-particle representations of surface oxygens and identify two controlling parameters: the distance of closest approach (DCA) between ionic charges and pore charges and grid spacing that modulates localization (while keeping average surface charge constant). By fitting pore diffusion coefficients to three experimental conductance points, we reproduce the entire experimental AMFE curve as well as anion leakage in CaCl$_2$ seen in experiments and molecular dynamics simulations. Remarkably, vastly different microscopic models of the surface groups yield indistinguishable device-level conductance curves when the DCA is matched, despite substantial differences in local Ca$^{2+}$ concentration profiles. Our results demonstrate that AMFE in wide nanopores is governed by a delicate interplay between charge inversion, anion leakage, and ionic mobility, underlying that in wide pores monovalent vs.\ divalent cation selectivity is modulated by cations vs.\ anion selecivity.

We provide a theoretical model for the non-equilibrium steady state heat flow through a quantum heat valve. The model is based on a master equation approach, where the partial secular approximation has been carefully performed in order to obtain accurate results. Our study assumes an ohmic spectral density for the two thermal baths of the model. This is in contrast with previous treatments of the quantum heat valve, where the baths have been assumed as being structured with a peaked spectral density near the resonance frequency of the resonator. These studies have also taken the resonator to be a part of the open quantum system of interest, which results in double counting of the resonator, as the latter appears both in the spectral density of the bath and as a part of the open system. Although this model accounts for the observations in a satisfactory way, it raises issues regarding its physical interpretation. Our method solves this conceptual problem. We apply it to describe an experiment on a quantum heat valve, showing that it successfully captures the experimental results and improves upon the previous theoretical model, which suffered from the resonator double-counting issue. Our findings confirm that the careful application of the master equation approach, in particular when it comes to the secular approximation, is a useful tool for explaining realistic experimental setups.

Anisotropy in granular materials arises from both the internal fabric and the directionality of the stress state, yet separating these effects experimentally remains challenging. This study develops a first-order linearisation of the incremental stress-strain response that isolates mechanical anisotropy from structural anisotropy using two independent orientation measures. The formulation enables both contributions to be quantified directly from macroscopic laboratory data. The method is applied to hollow-cylinder tests with systematically varied loading directions. Results show that both anisotropy components intensify as the stress state becomes more deviatoric; mechanical anisotropy is consistently stronger; and its relative dominance decreases with increasing deviatoric stress. Comparison with an isotropic hypoplastic model confirms that mechanically induced directional effects are captured even without fabric anisotropy. The framework offers a practical and physically transparent means for quantifying and comparing anisotropy mechanisms in granular materials.

The dielectric properties of proteins, particularly in their hydrated state, have been extensively studied. Numerous theoretical and experimental investigations have reported values of both the permittivity and the intrinsic dipole moments of specific proteins under well-defined hydration conditions. Since even approximate estimates of these properties are relevant from both fundamental and applied perspectives, we propose a user-friendly method to calculate the relative permittivity that can be readily integrated into proteotronics workflows. To validate the proposed approach, we compare the results with those obtained using a classical macroscopic method. The outcomes are consistent and contribute further insight into this long-debated issue.

The magneto-optic Kerr effect (MOKE), which describes the rotation and ellipticity of linearly polarized light upon reflection, typically occurs in magnetic materials that break time-reversal ($\mathcal{T}$) symmetry. Here we theoretically demonstrate that a similar effect can emerge even in two-dimensional nonmagnetic systems with $\mathcal{T}$ symmetry, owing to the nontrivial quantum geometry of electrons. We reveal that the nematicity of the quantum metric, which corresponds to electric quadrupole moment of electron wave packets, gives rise to a Kerr-like effect (KLE) depending on the incident polarization angle. Notably, neither magnetic order nor spin-orbit coupling, which are conventionally considered essential for the MOKE, is required for its emergence. The KLE is demonstrated by using both a minimal tight-binding model and a model for strained MoS$_2$ with parameters determined by first-principle calculations. This work reveals a quantum-geometric origin for polarization rotation effects beyond the MOKE and offers a distinct approach to probe quantum geometry and multipole moments of electrons.

Quantitative understanding of coupled reaction and transport processes in lithium-ion battery (LIB) composite electrodes remains challenging because key internal states cannot be measured directly. In this study, we develop a physics-integrated, data-driven analysis pipeline to estimate internal electrochemical states from operando microscopic X-ray absorption fine structure ($μ$-XAFS) hyperspectral data of LIB cathodes with LiPF$_6$ electrolyte. State-of-charge (SOC) maps are first constructed from Co K-edge spectra. To resolve ambiguities in the two-phase reaction region, a physics-regularized three-layer neural network is introduced, enforcing spatial continuity of SOC and current conservation. The inferred SOC dynamics are then incorporated into a Kirchhoff-based Markov random field framework that integrates Kirchhoff's current and voltage laws, Ohm's law, and a symmetric Butler-Volmer relation to estimate interfacial current, ionic current, electrolyte potential, and effective ionic conductivity. Application to composite electrodes with different initial electrolyte concentrations (0.3, 1, and 2M LiPF$_6$) reveals distinct reaction propagation behaviors governed by electrolyte concentration-dependent conductivity. The inferred electrolyte concentration distributions show qualitative agreement with independent operando X-ray transmission imaging performed on LIB composite cathodes employing a LiAsF$_6$ electrolyte. This framework enables quantitative visualization of otherwise inaccessible internal transport phenomena.

When a system is swept through a quantum critical point (QCP), the Kibble-Zurek mechanism predicts that the average number of topological defects follows a universal power-law scaling with the ramp time scale. This scaling behavior is determined by the equilibrium critical exponents of the underlying phase transition. We show that the correspondence between Kibble-Zurek scaling and quantum criticality does not hold generally. In particular, the defect density can exhibit a suppression faster than the Kibble-Zurek prediction even when the quench crosses a critical point, while conventional Kibble-Zurek scaling may persist for quenches through a non-critical point. Our results, based on models representative of a broad class of quasi-one-dimensional Fermi systems, identify the dynamical conditions under which universal defect scaling emerges and clarify the relation between defect generation and equilibrium criticality.

Ultrathin metal films exhibit liquid-like instabilities, rupturing via surface diffusion far below their melting points. This behavior constrains thermal budgets for advanced integrated circuits and emerging 2D-crystal devices. Here, we demonstrate that these instabilities can be fundamentally suppressed using graphene as a van der Waals (vdW) template. While conventional 20-nm-thick gold films break up into islands below 300 {\textdegree}C, templated films not only remain stable but also become structurally refined after annealing above 600 {\textdegree}C. This exceptional stability stems from a vdW-mediated crystallographic texture that reorganizes grain boundaries into a mechanically robust network. This mechanism significantly widens the processing window for nanoscale interconnects and enables high-temperature integration of metals with 2D-crystal technologies.

The aggregation of amyloid-forming peptides is a dynamic, complex process that underlies their diverse biological activities, from physiological functions to disease-associated dysfunctions. While the structure of fibrillar end-products is well-characterized for most amyloids, the heterogeneous and often transient oligomers, likely key in cytotoxicity, remain poorly investigated, especially for peptides with low-yield aromatic residues. Here, by exploiting and developing flow induced dispersion analysis in both peak and front modes, we demonstrate that intrinsic phenylalanine fluorescence can be harnessed to quantify the conversion of diffusing monomers into non-diffusing oligomers and fibrils. We further characterize low-molecular-weight oligomers, and their size evolution from 2 to 10 nm over time. Importantly, we validate the robustness of our approach using two tryptophan-free and fast-fibrillating amyloid peptides, PSM$α$3 and hIAPP, known for their key roles in S. aureus virulence and type 2 diabetes respectively. Our results overcome the limitations of traditional biochemical and biophysical amyloid assays by extending analysis from large oligomers and fibrils to small heterogeneous oligomers, under near-physiological conditions. This study thus offers a new analytical framework, thereby filling a critical gap in amyloid research, to probe the early stages of aggregation, key in the design of alternative therapeutics for amyloid-diseases.

We study finite system properties of the canonical low energy model for a semiconducting nanowire with Rashba spin-orbit coupling. The case of an isolated wire as well as of one proximitized by an s-wave superconductor are considered. Already for the normal wire, the presence of spin-orbit coupling leads to eigenstates of the finite system composed of more than two momentum eigenstates. The quantization condition for the wavevectors is not that of a quantum box, but given instead by a transcendental equation linking the involved wavevectors. For the wire with superconducting pairing, the presence of electron and hole channels complicates the composition of the eigenstates. In this case we derive an approximate quantization condition close to the phase boundary, and a condition for the appearance of exact zero energy states. It can be satisfied both in the topological and in the trivial phase. Both the trivial and topological zero energy states contribute to the linear transport through Andreev reflection and direct transmission processes, with their relative importance depending on the degree of the states' localization at the boundary.

We investigate the thermodynamic geometry of classical and quantum ideal gases in the relativistic regime, with particular emphasis on the effects of particle mass and spatial dimensionality. Relativistic kinematics is incorporated through the full energy-momentum dispersion relation and the corresponding relativistic density of states. Using the Fisher-Rao information metric derived from the partition function, we analyze the thermodynamic curvature for Maxwell-Boltzmann, Bose-Einstein, and Fermi-Dirac statistics. Exact analytical expressions are obtained in two spatial dimensions, while the three-dimensional case is studied numerically. We show that the thermodynamic curvature preserves its characteristic sign-positive for bosons and negative for fermions; even in the relativistic regime, reflecting effective attractive and repulsive statistical interactions, respectively. A distinctive relativistic effect is the shift of curvature singularities from the non-relativistic critical point to a mass-dependent threshold at $μ=mc^{2}$. In addition, the relativistic Bose-Einstein condensation temperature is evaluated, revealing explicit mass-dependent corrections to the non-relativistic result. These findings provide a unified geometric perspective on relativistic statistical systems and clarify the interplay between quantum statistics, relativistic kinematics, and critical behavior.

We develop a spectral-resolvent framework for computing the quantum Fisher information (QFI) using Krylov subspace methods, extending the notion of the Krylov distribution. By expressing the QFI as a resolvent moment of the superoperator $\mathcal{K}_ρ$ associated with a density matrix, the Krylov distribution quantifies how the QFI weight is distributed across Krylov levels in operator space and provides a natural measure for controlling the truncation error in Krylov approximations. Leveraging orthogonal polynomial theory, we identify two universal convergence regimes: exponential decay when the Liouville-space spectrum is gapped away from zero, and algebraic decay governed by hard-edge (Bessel) universality when small eigenvalues accumulate near zero. This framework establishes a direct connection between quantum metrology, spectral geometry, and Krylov dynamics, offering both conceptual insight and practical tools for efficient QFI computation in high-dimensional and many-body systems.

Lasers serve as the fundamental workhorses of photonic quantum technologies, with perfectly coherent light fields being essential for many protocols that generate nonclassical light, implement coherent control schemes, and initialize qubits. However, no laser is absolutely ideal and the implications of deviations from perfect coherence in quantum technological tasks remain unclear. In this study, we theoretically and experimentally explore the quantum coherence properties of lasers from a resource theory perspective, establishing a significant connection between photonics, quantum optics, and quantum information science. We demonstrate that the maximum achievable quantum coherence for laser light is constrained by spontaneous emission and the purity of the dephased laser field state. As a critical example application in quantum information protocols, we show that the quantum coherence of a laser field with a given mean photon number directly governs the maximum purity attainable when initializing a qubit in a superposition state through resonant driving. Our findings are highly relevant for bridging applied physics and engineering with integrated photonic quantum technologies and resource theories, paving the way for reliable benchmarking of various coherent light sources for applications in photonics and quantum protocols.

Recently, Fendley et al. (2025) [arXiv:2511.04674] revealed a new way to demonstrate the integrability of XYZ Heisenberg model by constructing a one-parameter family of integrals of motion in the matrix product operator (MPO) form. In this short note, I report on the discovery of two-parameter families of MPOs that commute with with the Heisenberg spin chain Hamiltonian in the XXX, XXZ, and XYZ cases. I describe a symbolic algebra approach for finding such integrals of motion and speculate about possible applications.

Altermagnetism -a magnetic state characterized by spin-polarized electronic bands at zero net magnetization- offers a promising route for next-generation spintronic devices. In two-dimensional (2D) magnets, twist engineering enables its realization by breaking the combined inversion and time-reversal symmetry (PT) while preserving crystal symmetries that ensure the altermagnetic order. Here, by means of first-principles calculations and symmetry analysis, we demonstrate that twist engineering applied to the recently synthesized metallic van der Waals antiferromagnet Co-doped bilayer Fe$_3$GaTe$_2$ (Fe$_2$CoGaTe$_2$) provides a robust platform for altermagnetism. By twisting two layers of Fe$_2$CoGaTe$_2$, the PT symmetry between opposite spin sublattices is broken, resulting in a non-relativistic $i$-wave altermagnetic state with spin splitting up to 138 meV. In the absence of spin-orbit coupling (SOC), the electronic states are spin-degenerate along six high-symmetry directions, while the inclusion of SOC preserves this degeneracy along the three directions protected by twofold rotation axes. Furthermore, we unveil the microscopic mechanisms governing the magnetic behavior in twisted Fe$_2$CoGaTe$_2$. Our results establish twist engineering and metallic Fe-based van der Waals antiferromagnets as versatile platforms to realize 2D van der Waals altermagnetism, with potential for designing high-efficiency ultrathin nanodevices.

Accurate noise estimation is essential for fault-tolerant quantum computing, as decoding performance depends critically on the fidelity of the circuit-level noise parameters. In this work, we introduce a differentiable Maximum Likelihood Estimation (dMLE) framework that enables exact, efficient, and fully differentiable computation of syndrome log-likelihoods, allowing circuit-level noise parameters to be optimized directly via gradient descent. Leveraging the exact Planar solver for repetition codes and a novel, simplified Tensor Network (TN) architecture combined with optimized contraction path finding for surface codes, our method achieves tractable and fully differentiable likelihood evaluation even for distance 5 surface codes with up to 25 rounds. Our method recovers the underlying error probabilities with near-exact precision in simulations and reduces logical error rates by up to 30.6(3)% for repetition codes and 8.1(2)% for surface codes on experimental data from Google's processor compared to previous state-of-the-art methods: correlation analysis and Reinforcement Learning (RL) methods. Our approach yields provably optimal, decoder-independent error priors by directly maximizing the syndrome likelihood, offering a powerful noise estimation and control tool for unlocking the full potential of current and future error-corrected quantum processors.

Colloidal perovskite quantum dots (pQDs) are promising quantum light emitters, and investigations at the single pQD scale have so far relied mostly on hot-injection synthesis, which requires precise temperature control and an inert atmosphere. While alternative synthesis routes under milder conditions are often associated with structural and surface defects that may have limited impact in ensemble measurements, demonstrating high optical quality at the level of individual pQDs constitutes the most stringent benchmark for a new synthesis protocol. Here, we demonstrate that a modified ligand-assisted reprecipitation (LARP) approach yields CsPbBr3 pQDs showing state-of-the-art optical properties at the scale of single emitters. By combining an amine-mediated post-synthetic size-trimming strategy with didodecyldimethylammonium bromide (DDAB) ligands for enhanced surface passivation and colloidal stability, we obtain isolated pQDs with stable emission and minimal spectral diffusion at cryogenic temperatures. Micro-photoluminescence experiments resolve the characteristic fine structure of the bright exciton, its low-energy optical phonon replicas, and the trion and biexciton states. Time-resolved and photon correlation measurements show a ~90 ps lifetime and high purity single photon emission, respectively. These results demonstrate that modified LARP synthesis constitutes an accessible alternative to hot injection, preserving the intrinsic excitonic and quantum optical properties of individual pQDs while offering greater flexibility for post-synthetic ligand engineering, as exemplified here by the use of DDAB for surface passivation.

Scanning Tunneling Microscopy (STM) is a cornerstone technique for visualizing the electronic density of states with atomic resolution (typically below 0.1 nm). While the field of view of most STM setups extends up to a few microns, obtaining atomic resolution over these large areas is often impractical and excessively time-consuming. This is due to the need to acquire maps with a point number reaching 107 or more with a full current or conductance vs voltage curve at each point. The standard procedure is to make large scale maps and then select small regions to zoom-in for high-resolution atomic scale analysis. However, this approach fails to address a question which is often critical: Does a specific atomic-scale modulation of the electronic density of states persist over much larger, mesoscopic length scales? Here we present a new method: Replica STM (R-STM), that overcomes this limitation, allowing the study of atomic-scale phenomena up to micron length scales. We obtained new large-area STM tunneling conductance maps in UTe2 and FeSe, spanning areas over 200 nm in size. In these large scale maps we discovered signals with wavelengths significantly exceeding interatomic distances. We show that these large-wavelength signals are replicas of the underlying atomic-scale density of states modulations. R-STM leverages these replica signals to efficiently track atomic-scale features over large areas. Using this novel technique, we show that the pair density modulation discovered recently in FeSe persists with the same characteristic wavelength up to hundreds of nm length scales. R-STM provides a powerful and practical new capability for STM to compare atomic scale with micrometer scale phenomena. The proof of principle of R-STM can be extended to any other scanning probe microscopy experiment where a periodic signal is traced as a function of position.

The mechanisms by which isolated interacting quantum systems evade thermalization extend beyond disorder-induced many-body localization, encompassing a growing class of interaction-driven phenomena. We investigate a spin-1/2 ladder with asymmetric XY leg couplings and tunable Ising interactions on the rungs, and identify the microscopic origin of many-body localization (MBL) in this setting. Through a suite of diagnostics -including entanglement dynamics, fidelity susceptibility, adiabatic gauge potential norms, level-spacing statistics and entropy of eigenstates- we uncover a reentrant progression of dynamical regimes as the rung coupling Jz is varied: integrable behavior at Jz=0, quantum chaos at intermediate Jz, and a robust nonthermal regime at strong coupling. In the latter regime, we demonstrate the emergence of a reversed quantum disentangled liquid (reversed-QDL), where the light species thermalizes while the heavy species remains localized. The strong-coupling limit further yields emergent local integrals of motion anchored in a fixed-point structure, providing a microscopic origin of the observed quasi-MBL dynamics. These results establish reversed-QDL as a distinct, disorder-free route to nonergodicity and broaden the classification of dynamical phases in quantum matter.

We explore the emergence of non-trivial topological phases arising from the interplay between structural geometry and the spin degree of freedom. Our system is a honeycomb nanoribbon heterostructure with armchair edges, consisting of a central Rashba spin-orbit coupled (SOC) region sandwiched between two pristine armchair nanoribbons. As the strength of Rashba SOC is increased, topological edge states exhibit at the interface between the Rashba SOC and pristine armchair honeycomb nanoribbon (ANR), and these edge states remain robust under edge perturbations. In addition, we find that for a finite width ANR, Rashba SOC can open/close a gap depending on its strength, which leads to a topological phase transition. This work provides an alternative perspective on how Rashba SOC influences topological properties.

A temporal magnon-qubit Mach-Zehnder (MZ) interferometer is proposed. The interferometer is based on controllable entanglement of a microwave qubit and a magnonic state, achieved by application of a pulsed magnetic field playing the role of a magnon-qubit temporal "beam splitter". Analogous to a typical MZ interferometer, the generated interference pattern of the final qubit population carries information about the magnon dynamics. One important application of the proposed scheme is the study of single magnon decoherence. Interestingly, this scheme allows one to independently determine rates of two possible decoherence channels. This may help enable single magnon state applications and answer fundamental questions of quasi-particle decoherence at single quantum levels.

In the modelling of social systems, opinion latency is the idea that once an agent changes its opinion, there will be a period of time where it is immune to other changes. When added to the voter model this leads to a situation where no matter how low the latency is or how many opinions are considered, all opinions end up in a coexistence where they are equally represented. In this work, we examine what happens when latency is added to the Sznajd model. What we find is that for low latency, the model behaves roughly like it does in the absence of latency, where one opinion will always eventually dominate. For high latency, the possibility for a symmetric coexistence of 2 opinions arises, but contrary to the voter model, a coexistence of more than 2 opinions is never stable. We provide evidence of this phenomenon with computer simulations of the model in Barabási-Albert networks, together with a mean field treatment that is able to capture the observed behavior. We argue that this could hint at an explanation for the prevalence of two opinion splits in the real world.

We studied photoluminescence of spatially indirect excitons (IXs) in a MoSe$_2$/WSe$_2$ van der Waals heterostructure. We observed a quasi-periodic triangular pattern of IXs with the characteristic wavelength of the pattern $\sim$ 2.6 $μ$m.

We consider arbitrary-shaped microswimmers of spherical topology and propose a framework for expressing their slip velocity in terms of tangential basis functions defined on the boundary of the swimmer using the Helmholtz decomposition. Given a time-independent slip velocity profile, we show that the trajectory followed by the microswimmer is a circular helix. We derive analytical expressions for the translational and rotational velocities of a prolate spheroid swimmer in terms of its Helmholtz decomposition modes and explore the effect of aspect ratio on these rigid body velocities. Then, for a given arbitrary swimmer shape of spherical topology, we investigate which slip profile minimizes the total power loss. A partial minimization is performed in which the direction of net motion of the swimmer is prescribed, followed by a global optimization procedure in which the best net motion direction is determined. The optimization results suggest that the competition between linear and rotational optimal motion is linked to symmetries in the shape of the microswimmer.

The response of amorphous solids to a mechanical perturbation consists in an elastic and a plastic deformation. The latter is mediated by localized irreversible rearrangements associated with Eshelby-like quadrupolar singularities in the displacement field. It has recently been argued that a density of such singularities leads to an anomalous elastic behavior taking the form of screening effects, which goes beyond classical elastic predictions. Here, we reexamine this scenario using general theoretical arguments and a description in terms of an elasto-plastic model, which we compare with atomistic simulations of the canonical Eshelby inclusion geometry. We discuss the conditions under which a finite, i.e., nonvanishing, density of quadrupolar events is created by an imposed perturbation. We argue that, except when the perturbation is macroscopic, there are many situations in which the density of quadrupolar defects is zero in the thermodynamic limit. In these cases, we find that plastically active quadrupoles emerge in a region whose size generically scales as the spatial extent $\ell$ of the mechanical perturbation. This mechanism leads to anomalous elasticity on a scale $\ell$ close to the perturbation and to conventional elasticity beyond. The simulations of the elasto-plastic model reproduce the emergence of plastic quadrupoles in a region set by $\ell$ and the associated renormalization of the effective shear modulus, but they do not exhibit the dipole-screening signatures reported in atomistic and experimental studies. Our analysis delineates the scale-dependent breakdown of long-wavelength elasticity in amorphous materials and suggests directions for incorporating anomalous screening into mesoscopic modeling frameworks.

We analyze a time-continuous version of a cellular automaton decoder for the toric code in the form of a Lindblad master equation. In this setting, a self-correcting quantum memory becomes a thermodynamical phase of the steady state, which manifests itself through the steady state being topologically ordered. We compute the steady state phase diagram, finding a competition between the error correction rate and the update rate for the classical field of the cellular automaton. Strikingly, we find that self-correction of errors is possible even in situations where conventional quantum error correction does not have a finite threshold.

We consider physical Hamiltonians that can be represented by the multiparametric Gaussian ensembles, theoretically derive the state ensembles for its eigenstates and analyze the effect of varying system conditions on its bipartite entanglement entropy. Our approach leads to a single parametric based common mathematical formulation for the evolution of the entanglement statistics of different states of a given Hamiltonian or different Hamiltonians subjected to same symmetry constraints. The parameter turns out to be a single functional of the system parameters and thereby reveals a deep web of connection hidden underneath different quantum states.

We examine the dynamics, Hall angle, and pinning for N{é}el skyrmions in an altermagnet. Using an atomistic model, we show that skyrmion velocity and Hall angle dependence is anisotropic with respect to the direction of the drive, due to the fourfold symmetry implied by the two sublattices of the altermagnet. The skyrmion Hall angle and velocity at fixed drive show strong variations for increasing ratios of the exchange constant of the sublattices, $J_2/J_1$. This fourfold anisotropy of altermagnetic (ATM) skyrmions also leads to anisotropic pinning effects for an ATM skyrmion interacting with isotropic circular pinning sites. We also propose a simple particle model for this system that takes into account this anisotropy and find that it captures both the variations of the ATM skyrmion Hall angle and velocity as a function of drive direction, as also found in the atomistic simulations. Using this particle model, we examine ATM skyrmions interacting with a periodic array of pinning sites. For increasing ratios of $J_2/J_1$, we find a strongly non-monotonic ATM skyrmion velocity, where there is a minimum in the velocity where the skyrmion locks to different symmetry directions of the periodic pinning lattice. For a random array, we find that ATM skyrmions show strongly anisotropic depinning thresholds and velocity responses for different drive directions, and that the Hall angle is nearly constant with drive. In comparison, for the same parameters, the depinning threshold for a ferromagnetic (FM) skyrmion is lower, and the skyrmion Hall angle shows a strong velocity dependence. The lower depinning threshold for FM skyrmions is due to stronger Magnus forces.

Biological membranes and vesicles play a central role in living systems, forming dynamic interfaces that regulate cellular organization and function. Classical descriptions of membrane mechanics that are rooted in equilibrium statistical mechanics and linear elasticity have yielded deep insights into membrane morphology and the role of thermal fluctuations on cellular function. However, real biological membranes operate far from equilibrium, continuously driven by active processes powered by energy consuming proteins. In this work, we employ a nonequilibrium statistical mechanics framework to model active membranes and derive analytical expressions for four fundamental properties that characterize their mechanical behavior: (a) the tension area relation, (b) the mean square amplitude of fluctuations, (c) correlation of normal vectors, and (d) the persistence length. These results collectively highlight the utility of fluctuation spectra as a starting point for elucidating membrane mechanics in both passive and active settings. Moreover, these results provide a theoretical basis for analyzing and interpreting fluctuation based assays of active membrane behavior.

A coupled spintronic oscillator array has been considered attractive for neuromorphic computing applications. Experimental reports have shown the nano-constriction geometry to be a relatively easier-to-fabricate platform for implementing such spin oscillators, but most prior reports on training and inference algorithms for neuromorphic computing using spin oscillators have been mostly restricted to the nano-pillar geometry. Also, those prior reports involve updating the natural frequency values of the oscillators and moving the synchronization regions on to the data clusters during the offline learning phase, which has associated challenges. In this context, we design and simulate a novel artificial neural network (ANN) - oscillator neural network (ONN) algorithm where in the offline learning phase, the weight parameters of the ANN are updated such that the data clusters are instead moved to the synchronization regions of spin Hall nano oscillators (SHNOs) in the nano-constriction geometry, as obtained through micromagnetic simulations. We further simulate the on-chip inference part of the ANN-ONN algorithm where the ANN is implemented on a crossbar array of domain-wall synapse devices, as simulated here through micromagnetics, and the ONN is implemented on nano-constriction SHNOs. We show successful data classification for both binary and multi-class classification tasks to demonstrate the generalizability of our proposed scheme.

The transport of soft viscoelastic gels through confined geometries underlies critical processes in biomedical, biological, and industrial systems. Here, we examine the translocation of a spherical microgel through a narrow capillary whose diameter is smaller than the equilibrium gel size. Using coarse-grained molecular dynamics simulations in tandem with mean-field theory and mechanical analysis, we uncover a critical threshold diameter $d_c$ below which the microgel cannot enter, regardless of the applied pressure. This geometric limit emerges from the interplay between gel elasticity and its internal network connectivity, captured quantitatively by a graph-theoretic model. We construct a phase diagram in the parameter space of tube diameter $d$, applied force $f_g$, and gel stiffness $Y$ (Young's modulus), which delineates the regimes of successful translocation and mechanical arrest. Under negligible wall friction, gel mobility scales with the applied force; however, beyond a cutoff set by the network topology, progressive densification in the constriction stalls the microgel. Our results reveal the mechanical and topological determinants of soft-gel transport in confinement and provide predictive guidelines for engineering gel-based systems in microfluidics, drug delivery, and tissue-level filtration.

In regimes where inertia is negligible, the temporal evolution is governed by overdamped dynamics. This limit is particularly relevant in soft-matter contexts, such as polymers, colloidal suspensions, and processes occurring at the cellular scale. Being able to manipulate the dynamics of such many-particle systems would enable control over rate-dependent elastic responses, time-dependent material properties, relaxation processes, and perhaps the hydrodynamics of suspensions. In this work, we develop a framework for manipulating overdamped dynamics through local, physically motivated update rules. Our approach is inspired by ideas from physical learning and directed aging, in which microscopic parameters adapt autonomously to endow a material with a desired function. Using the Rayleighian formulation, whose minimization reproduces the overdamped equations of motion, we derive approximate directed-aging and equilibrium-propagation rules tailored to dissipative systems. To demonstrate these ideas, we study a disordered Maxwell material that behaves elastically at short times but flows at long times. By locally modifying the viscous damping, we show that one can tune the viscous Poisson's ratio and shape local mechanical responses. These results illustrate how materials can be trained to exhibit targeted rate-dependent elastic and viscous behaviors.

To search for nontrivial spintronic characters of altermagnets has been a focus in spintronics, with numerous attentions paid to spin-group symmetry dictated non-relativistic effects, such as the spin-splitting torque (SST). Here, we go beyond this paradigm by unveiling the magnetic orbital torque (MOT) that is enabled only by spin-group symmetry breaking of real altermagnets with spin-orbit coupling. This effect enables generating unconventional out-of-plane orbital current with collinear orbital polarization from all 10 spin Laue groups of centrosymmetric altermagnets, which is critical for field-free manipulation of perpendicular magnetization that is a key for next-generation information technology. We reveal perturbative and non-perturbative effects of spin-group symmetry breaking, and unveil that MOT is a much more generic effect of real altermagnets than SST. Our first-principles calculations predict giant out-of-plane MOTs at room temperature in the absence of and competing with SST in experimentally identified altermagnets CrSb and FeSb2, respectively. These findings uncover new fundamental physics in altermagnetic spintronics, and open up broad vistas of altermagnets in magnetic memory.

The spin dynamics of a trimer with Dzyaloshinsky-Moriya (DM) interaction are investigated within a unified Hamiltonian framework that connects quantum-mechanical and semiclassical descriptions. The interpolation between the two regimes is realised by solving the modified Gisin-Schrödinger equation, in which the relative weight of a quantum coherence and local mean-field contributions is continuously tuned. The resulting dynamical behaviour is analysed and summarised in a ground state diagram that illustrates how the character of the spin motion evolves from fully quantum to semiclassical as the DM interaction is treated at different levels of approximation. In the last part of the publication, the chiral spin dynamics proposed by Da-Wei Wang et al. is examined theoretically, taking into account its behaviour at the boundary between quantum and classical physics.

We present peapods (github.com/PeaBrane/peapods), an open-source Python package for Monte Carlo simulation of Ising spin systems with arbitrary coupling constants on arbitrary-dimensional hypercubic lattices with periodic boundary conditions. The computational core is written in Rust and exposed to Python via PyO3, combining the ergonomic interface of Python with the performance of compiled, memory-safe code. The package implements Metropolis and Gibbs single-spin-flip algorithms, Swendsen--Wang and Wolff cluster updates, and parallel tempering. Replica-level parallelism is achieved through the Rayon work-stealing scheduler. We validate the implementation against the exact critical temperature of the two-dimensional Ising model via finite-size scaling of the Binder cumulant.

A prominent feature of $d$-wave altermagnets is that spin current is generated by applying temperature gradient, which is known as the spin-Nernst effect. We show in $f$-wave magnets that spin current is generated proportional to the square of the temperature gradient, which we call the nonlinear spin-Seebeck current. It can be used as a spin current diode. In addition, we show in $g$-wave altermagnets that spin current is generated in the third order of the temperature gradient. We also show in $i$-wave altermagnets that spin current is generated perpendicular to the temperature gradient, which is the spin-Nernst current. We have derived analytic formulas for these spin currents. It is interesting that these phenomena occur in the absence of the spin-orbit interaction. On the other hand, we show in $p$-wave magnets that spin current is not generated by temperature gradient.

We formulate a phenomenological effective field theory in which biological intelligence emerges as a macroscopic order parameter sustained by continuous metabolic flux. By modeling cognition as a coarse-grained neural activity field governed by a variational free energy, we derive closed-form expressions for information capacity and structural susceptibility using a Gaussian maximum entropy approximation. The theory predicts a universal algebraic divergence of the susceptibility, $χ\sim K^{-3/2}$, as the structural stiffness $K$ approaches the instability threshold. The exponent $γ= 3/2$ is consistent with the mean-field branching process universality class, thereby providing a theoretical rationale for the observed avalanche size exponent $τ\approx 3/2$ in cortical dynamics without invoking microscopic equivalence. We identify adult cognition as a metabolically pinned non-equilibrium steady state maintained near the critical regime $Γ\equiv K/α\approx 1$ by continuous metabolic regulation, while pathological decline corresponds to a delocalization transition triggered by the violation of structural stability conditions. The framework generates concrete, falsifiable predictions for attention scaling, altered states of consciousness, and transcranial magnetic stimulation responses, each of which can be tested against existing neuroimaging and electrophysiological datasets.

The Berry connection encodes the momentum-space geometry of occupied Bloch states in gapped insulators and plays a central role in topological materials. While gauge-invariant quantities can be evaluated from overlap matrices between neighboring $k$ points, accessing the Berry connection itself as a smooth field requires specifying a continuous gauge over the Brillouin zone. Wannier-based workflows achieve this through projection onto localized orbitals, enabling stable evaluation of geometric quantities and response functions. In this setting, the Berry connection enters directly in Wannier-interpolated calculations of polarization, Berry curvature, optical conductivity, and related response functions. In practical implementations, however, the projection-gauge Berry connection is typically constructed from finite-difference overlaps between neighboring $k$ points, discretizing momentum derivatives and introducing errors tied to $k$-mesh spacing and gauge alignment. These effects can become numerically delicate in systems with small band gaps or when evaluating higher-order responses such as the Chern-Simons axion angle. Here, we derive an exact expression for the non-Abelian Berry connection in the projection gauge that is local in crystal momentum. Starting from projected and orthonormalized Bloch-like states, we obtain a closed-form equation expressed entirely in terms of $k$-local quantities. We validate the formulation in one and three dimensions by computing the Berry phase and Chern-Simons axion angle in tight-binding models. The resulting framework provides a stable route to evaluating geometric properties within Wannier interpolation schemes and future first-principles implementations.

We obtain bounds on the Kullback--Leibler divergence to equilibrium for mass-action chemical reaction networks (CRNs) with equilibrium. The associated decay rates are characterized in terms of the singular values of the stoichiometric matrix, convexity parameters, and time-integrated activities via deformed-exponential-type functions. We further extend these bounds within a generalized gradient flow framework. We highlight the biological relevance of this framework: the resulting bounds apply to quasi-steady-state regimes, where long transients and plateau-like behavior are common and functionally important. We illustrate the framework using a catalytic CRN exhibiting plateaus, where the bounds capture slow relaxation induced by local convexity and provide a bound-based approach to quantifying relaxation in CRNs.

Cells often divide under mechanical confinement, where surrounding structures restrict shape changes during cytokinesis. Although forces generated during confined division have been measured experimentally, it remains unclear how confinement geometry and mechanics determine the transmitted force. Here we develop a minimal mechanical theory of cell division under confinement. Modeling the cell as an incompressible volume bounded by an interface with effective isotropic tension, we show that confinement restricts the set of mechanically admissible furrow shapes. As the furrow radius decreases, it reaches it reaches a confinement-induced minimum. Beyond this point, further ingression does not alter the interface shape, and both pressure and axial force saturate. We analyze force and pressure in rigid, soft, and strong three-dimensional confinement and demonstrate that a single geometric mechanism underlies these distinct cases. After rescaling force and length by the appropriate geometric scale, cells of different size and surface tension collapse onto a single universal curve. The relevant length scale is the cell size for rigid and soft confinement, and the confinement size in fully enclosing three-dimensional confinement. In soft confinement, environmental stiffness and spindle-generated axial forces determine the operating force and pressure, while the geometric constraint fixes the maximal attainable levels. In summary, our results show that mitotic force transmission and mitotic pressure during cytokinesis are bounded by confinement geometry, with material properties and active forces selecting the operating point within these geometry-imposed limits.

We assess the impact of atomistic substrate lattice relaxation schemes in the primary band gap at charge neutrality and the secondary valence band gap of graphene on hexagonal boron nitride (G/h-BN) as a function of twist angle. For zero twist angle, the primary gap decreases from $\sim 30$~meV in fully relaxed suspended G/h-BN bilayers, to $\sim 9$~meV when the remote h-BN substrate layer is kept rigid, and down to $\sim 3$~meV in completely rigid structures. In the presence of relaxations, the primary gap shows a maximum near $\sim 0.6^{\circ}$ coinciding with energetic stabilization due to alignment between the moiré pattern and the graphene lattice vectors, while the secondary valence band gap drops from $\sim 12$~meV down to zero beyond twist angles of $\sim 1^{\circ}$. A small but finite primary gap on the order of $\sim 1$~meV, with a mass sign favoring electronic occupation of carbon atop boron, persists across twist angles from $0^{\circ}$ to $30^{\circ}$ for all sliding configurations, and switches sign for twist angles between $30^{\circ}$ and $60^{\circ}$.

The atomistic Hubbard interaction U, representing the on-site Coulomb repulsion, serves as a pivotal parameter in theoretical models describing of correlated systems, yet its precise experimental determination especially in moiré systems remains challenging. Scanning Tunneling Microscopy(STM) provides real-space images of the local density of states (LDOS), offering rich data sets that reflect the unique electronic structure of the material. Here, we introduce a systematic methodology for extracting the Hubbard U parameter directly from these LDOS images through the application of machine learning (ML) in the case of twisted bilayer graphene in the flat-band regime. The regression of U is highly accurate even though the image-similarity is greater than 99.98%. Subsequent data-analysis further suggest a weak crossover between the weak and strong coupling regime at Uc/t 1

A model for the tensorial dielectric permittivity of phospholipid membranes is presented here. The four-nanometer-thick membrane is treated as a composite made up of three dielectric slabs: one for each of the two phospholipid head-group regions, and one for the entire domain spanned by the lipid tails. Equal and opposite bound surface charge densities surround each head-group slab, and account for the membrane dipole potential. Three-slab model parameters are obtained from molecular dynamics simulations, and capture both the zero-field electric potential and the membrane response to applied electric fields. The tail region is well-approximated as having vacuum permittivity, while the head-group region is highly anisotropic due to the configurations of molecular dipoles. For the bilayers studied, the out-of-plane permittivity of the head-group region is 10--15 times that of the vacuum, while the in-plane permittivity is an order of magnitude larger. Membrane responses to applied electric fields up to 30 millivolts per nanometer are found to be in the linear regime. The model overcomes a fundamental limitation of microscopic theories -- where the out-of-plane permittivity is ill-posed in the head-group region due to large gradients in the local electric field -- by averaging over slab widths, thereby introducing new length scales. Our approach can be extended to characterize general interfacial systems with similarly ill-defined permittivities.

Although numerical simulations of rheological measurements typically focus on homogeneous systems, heterogeneity can profoundly impact material properties. We report on the rheological properties of a suspension of two-dimensional Lennard-Jones particles across the gas/liquid and the gas/solid coexistence lines of the system. We show how the presence of multiple coexisting states has a significant impact on the mechanical properties of these systems when compared with their homogeneous reference counterparts. Our results establish an extended map to navigate a landscape where not only density and temperature, but also phase coexistence, dictate the transition from viscous to elastic-dominated behavior under shear. These results provides a benchmark for future research into heterogeneous fluids where the coexistence of complex dynamic states is frequently observed.

Fluid phase equilibrium depends on the external constraints imposed on a system. In a closed system with fixed volume, depending on the average density, a vapor bubble may be stable, metastable, or unstable, with respect to the homogeneous liquid phase. In the case where the bubble is metastable, we study its lifetime, i.e. the average waiting time needed to observe bubble collapse, and the corresponding lifetime of the homogeneous liquid. For the smallest systems, we predict the possibility to observe phase flipping, when the fluid oscillates between states with and without bubble. We provide an example of phase flipping in a simulation of a Lennard-Jones fluid.

The idea that topologically protected quantum states, such as anyons, may be attached to super/semiconductor heterostructures has received enormous attention, but experimental signatures in 1D systems remain elusive. Here we revisit theoretical underpinnings of anyons in 2D fractional quantum Hall (FQH) systems, whose signatures have been experimentally observed by independent groups. Invoking novel theorems about the Hopfion or $\mathbb{C}P^1$-model understood as flux quantization in 2-Cohomotopy, we demonstrate a robust prediction for possibly nonabelian anyonic states induced by superconducting islands.

Copolymerization is a powerful tool to tune polymers' glass formation behavior in order to improve properties such as ion conductivity, thermal stability, and mechanical response. Classically, mixing rules such as the Fox equation and its variants have been employed to predict and explain glass transition temperature (Tg) variations with copolymer composition. However, a large number of copolymers exhibit significant excursions from these mixing rules. Increasingly, it also appears that these excursions can be tuned by copolymer sequence - an effect that is not predicted by classical mixing rules and thus far remains poorly understood. Here, we perform molecular dynamics simulations to probe the interplay between copolymer sequence, liquid-liquid phase equilibria, and Tg. We find that the direction of Tg shifts, and their dependence on sequence, are predicted by the type of liquid-liquid phase boundary towards which the comonomers tend. Monomers tending towards Upper Critical Solution Temperature behavior exhibit Tg reductions that become stronger with increasing monomer alternation; monomers tending towards Lower Critical Solution Temperature behavior exhibit Tg enhancements that become stronger with increasing monomer alternation. This behavior fundamentally emerges from a forced mixing phenomenon wherein monomer alternation induces close contact between monomer types that would otherwise tend towards phase separation. These results point towards strategies for rationally varying copolymer Tg, at fixed chemical composition, via judicious design of polymer chain sequence.

Fluctuation-response relations encode fundamental constraints on nonequilibrium systems. While time-domain static response is bounded by activity and entropy production, finite-frequency extensions for time-dependent perturbations remain largely unexplored. Here, we derive frequency-domain fluctuation-response inequalities for steady-state Markov processes with time-dependent perturbations. For barrier and entropic perturbations, the spectral signal-to-noise ratio (SNR) is universally bounded by dynamical activity. Furthermore, for state-current observables, the SNR is bounded by the entropy production rate (EPR). We illustrate our results using the F1-ATPase model to infer EPR. These finite-frequency inequalities provide a practical route to infer dissipation from power spectra measurements.

Lithium and sodium are the only solids that are known to lose crystalline order upon cooling. The seemingly-disordered low-temperature phase shows signatures of various close-packed structures. The lack of order has been attributed to a hidden gauge symmetry that arises when electrons from one layer can hop to a neighbouring layer but not further. It makes all close-packed structures nearly degenerate and leads to ``structural frustration''. In this article, we examine whether this symmetry is reflected in transport signatures. Taking advantage of in-plane translational periodicity, we map the bulk Bloch Hamiltonian to an effective one-dimensional chain, with stacking disorder mapping to random phases of the hopping amplitudes. We derive an explicit analytic form for the Green's function of electrons and use it to calculate conductance of a bulk crystal. When hopping in the effective one-dimensional chain is restricted to nearest neighbours, conductance is completely insensitive to phase disorder, which indicates that all close-packed structures exhibit the same conductance. We show that the leading correction that can differentiate between close-packed structures arises from hopping to the next-nearest-neighbour layer, equivalent to second-neighbour hopping in the chain model. This process appears when a pair of next-neighbour layers are aligned in a certain way, e.g., at an hcp-like stacking fault within an fcc background. With this hopping included, conductance becomes sensitive to the precise arrangement of layers. When multiple stacking faults are present, the conductance decreases with increasing system size, as expected from Anderson localization. Our results are applicable to pressurized lithium and sodium, where conductance measurements can identify and characterize stacking faults.

Atomic Layer Deposition (ALD) process development is often hindered by time-consuming and precursor-intensive tuning cycles required to identify saturation conditions. We introduce a physics-informed Bayesian Active Learning (BAL) framework that autonomously tunes precursor pulse times by integrating a Langmuir adsorption model directly into the Gaussian Process (GP) kernel. A key innovation is a two-stage parameter estimation strategy that decouples noise filtering from physical parameter extraction: the GP first smooths noisy data through standard prediction, then Langmuir parameters are fitted to the noise-filtered GP predictions. This approach effectively separates signal from experimental noise. We evaluate the framework against a standard data-driven GP across four simulated regimes, demonstrating convergence within five iterations, up to fourfold improvement in prediction accuracy, and two to fourfold reduction in precursor usage. Experimental validation using TiO2 deposition via Tetrakisdimethylamido Titanium (TDMAT) and ozone confirms that the physics-informed model accurately identifies saturation times for high-coverage targets ($\geq$95\%), with observed deviations at lower saturation levels providing valuable insight into non-ideal desorption behaviors.

We report on the monolithic, two-step epitaxial growth of site-controlled InGaAs quantum dots via the buried stressor method with local quantum dot density variation. As a result of high fabrication accuracy, we achieve low lateral displacements of the individual buried stressor apertures of $ 17^{+19}_{-17}$~nm from mesa centers. We provide extensive micro-photoluminescence and cathodoluminescence characterization of the site-controlled quantum dots and give theoretical calculations, explaining the effect of the stressor aperture on the quantum dot emission properties, positioning, and density. We show reproducibility of the nucleation process for apertures of the same size and achieve precisely-positioned, low- and high-density quantum dot nucleation within one active layer growth step. The results presented in this work demonstrate the significant potential of the buried stressor concept in fabricating single photonic chips, simultaneously combining single-photon sources and microlasers featuring different local densities of site-controlled quantum dots, paving the way for highly functional source modules with applications in photonic quantum technology.

Deterministic closures for coarse-grained turbulence models help reproduce mean statistics, but often fail to capture the finite-time growth of uncertainty. Using the framework of shell models as a quantitative multi-scale testbed, we compare fully resolved simulations with large-eddy simulations using either stochastic or deterministic subgrid closures. While in the fully resolved system a single microscopic perturbation is rapidly amplified by strongly chaotic dynamics, truncation produces a strong delay and suppression of variance growth when uncertainty is introduced through initial condition perturbations only. We show that a data-driven Langevin-type stochastic closure restores the correct timing and magnitude of variance growth across scales, demonstrating that sustained stochasticity is essential for predictability in reduced turbulent dynamics.

Gaining insights into flame behaviour at small scales can lead to improvements in the efficiency of micro-reactors, compact power generation systems, fire safety technologies, and various other applications where combustion is confined to micro or mesoscales. Flame regimes observed in mesoscale combustors, namely Stable flame, Flames with repetitive extinction and ignition, and Propagating flame, exhibit unique dynamic characteristics that differentiate them from one another. In this study, we systematically examine the various flame regimes observed in mesoscale combustors from both dynamical and statistical standpoints. Our experimental methodology involves stabilizing a flame inside a quartz tube (an optically accessible mesoscale combustor) with an inner diameter of 5 mm. A premixed methane-air mixture is used as fuel, with its equivalence ratio and Reynolds number being the input parameters. Instantaneous OH* chemiluminescence and Acoustic pressure signals, along with high-speed flame imaging, were acquired for combustion dynamics characterization. The objective of this study is to analyze the distinct dynamical signatures associated with these observed flame regimes. For this purpose, Recurrence Quantification Analysis, followed by a Statistical-Spectral analysis, has been performed based on the experimentally acquired OH* Chemiluminescence and Acoustic pressure time-series signals. Subsequently, a stacking ensemble-based machine learning framework has been implemented for mesoscale flame regime classification based on the features extracted from the two aforementioned analyses. In addition, Continuous wavelet transform (CWT) scalograms and three-dimensional phase plots have been graphed to visually elucidate the evolution of system dynamics and the complex interaction of competing time scales in these flame regimes.

Abrupt transitions are a central concern in climate and ecological research, and may arise when critical thresholds known as tipping points are crossed. However, previous work has shown that finite-time overshoots of tipping points can be safe, and that such behavior is captured by an inverse-square-law criterion when overshoots are sufficiently small and slow. So far studied in isolated systems with external drivers, (un)safe overshoots may also emerge from interactions between subsystems. Here, we investigate safe-overshoot phenomena in unidirectionally coupled slow-fast systems featuring both nonlinear interactions and coupling through time-derivatives. Specifically, we derive a criterion for the occurrence of safe overshoots analogous to the inverse-square law for isolated systems, but adapted to interactive settings, and expressed explicitly in terms of the timescale separation and coupling strength between subsystems. We illustrate these results using two conceptual models in which the Atlantic Meridional Overturning Circulation interacts with either the Amazon rainforest or the Greenland Ice Sheet.

Researchers have developed hybrid Van der Pol Rayleigh Duffing type oscillators to model human induced forces; however, their analytical framework has largely relied on the Lindstedt Poincare perturbation method, energy balance approaches, and harmonic balance techniques. This paper aims to apply new mathematical tools to these existing models and address potential research gaps. An analytical proof for the stability of the limit cycle has been formulated by using the Krylov Bogolyubov perturbation method. The multiple scales method has been modified to highlight an iterative algorithm for determining the order of approximation required to capture nonlinear effects. The describing function method is utilised to formulate an alternate amplitude. Comparisons between first order amplitudes obtained from perturbation analysis and the describing function formulations reveal conditions under which the two approaches converge. These conditions are exploited to formulate additional constraints for the estimation of model parameters, offering a systematic alternative to purely optimisation based approaches.

We investigate a growing network model that combines preferential and uniform attachment with two distinct mechanisms of edge deletion. In addition to the usual uniform probability edge deletion, we introduce a novel node-based rule in which uniformly chosen non-isolated nodes lose one of their incident edges. This mechanism differs fundamentally from uniform edge deletion and leads to a nonlinear evolution for the stationary degree distribution due to the nonlinear dependence on the fraction of isolated nodes. We solve the general problem in the stationary regime and obtain closed-form expressions for the degree distribution in terms of hypergeometric and confluent hypergeometric functions. Depending on the balance between attachment and deletion rates, three asymptotic regimes for the degree distribution arise: power-law, exponential, and a critical regime characterized by a stretched exponential decay. We show that the node-based edge deletion mechanism is less likely to disrupt the scale-free (power-law) regime than the uniform edge deletion. Moreover, we also demonstrate that the precise balance between preferential attachment and node-based deletion can transform a scale-free network into a critical one with stretched exponential decay. Extensive numerical simulations exhibit excellent agreement with the theoretical predictions.

We investigate analytically and numerically the dynamics of a two-dimensional superflow governed by the Gross-Pitaevskii equation passing over finite-size rectangular obstacles: an impenetrable wall and an impenetrable rectangular well. Extending classical studies of vortex nucleation around smooth obstacles, we focus on the role of sharp corners in determining the onset of vortex nucleation. Using a combination of analytical techniques based on the Schwarz-Christoffel methods for potential flow and on numerical simulations, we show that local velocity amplification near sharp corners crucially controls the critical flow velocity for vortex nucleation. For both wall and well configurations, we identify analytically and theoretically the critical velocities as a function of the obstacle width and its height or depth, finding an excellent agreement between the theory and our numerical simulations. Our results provide a simple framework for understanding superflow stability past finite-size obstacles with sharp features and are directly relevant to experimentally realizable configurations in atomic Bose-Einstein condensates and related superfluid systems.

We address the existence, stability, and dynamics of single-ring and multi-ring vorticity-carrying necklace solitons under the action of the Kerr nonlinearity and a Bessel-lattice potential modulated in the azimuthal direction. The model may be realized in the spatial domain for bulk optical waveguides, the spatiotemporal domain for optical cavities, and for effectively two-dimensional Bose-Einstein condensates. The setup supports single- and multi-ring necklace vortex patterns, including monopoles, dipoles, tripoles, quadrupoles, pentapoles, sextupoles, octupoles, and 12-poles. In contrast with the inherent instability of conventional vortex beams with high topological charges (winding numbers), vortex necklace-shaped solitons with large winding numbers are found to be stable in the present setup. In particular, octupoles exhibit stable breathing dynamics, and 12-pole necklaces with high winding numbers may be stable. These findings provide a new way for generating stable vortex necklaces, offering a vast potential for manipulations of complex spatiotemporal light fields.

We show that for a hyperbolic system of hydrodynamic type admitting n symmetries, there exists a coordinate system in which the generator of the system and all the symmetries are diagonal.

A key challenge for the practical application of metal halide perovskites (MHPs) is the instability of the desired perovskite phase relative to the optically non-active $δ$ phase. To determine the phase stability, we previously developed a procedure to compute the harmonic free energy as a function of temperature, which was suited for CsPbI$_3$ but fails when Cs is replaced by organic cations due to their rotational freedom. Herein we propose a multistep thermodynamic integration (TI) approach that corrects the harmonic free energy to obtain the Gibbs free energy. Given the abundance of local minima in these materials, we employ replica exchange to prevent simulations from getting trapped, while introducing an intermediate potential energy surface to improve convergence and reduce computational cost. Benchmarking energy and forces from different exchange-correlation functionals and dispersion methods against high-level RPA+HF calculations identifies PBE+D3(BJ) as the best trade-off between accuracy, computational efficiency, and precision. To perform molecular dynamics simulations within the TI framework, it was necessary to train a machine learning potential using the MACE architecture on ab initio data calculated with density functional theory. Our results show that, for all three materials, the free energy difference between the $γ$ and $δ$ phases exhibits a very similar temperature dependence. This suggests that phase stability is primarily governed by differences in ground-state energy, rather than by material-specific thermal effects. Beyond these three materials, our methodology provides a robust framework for investigating the phase behavior of other MHPs, paving the way for the discovery of more stable perovskites.

Within twelve days of launch, an AI-native social platform exhibits extreme attention concentration, hierarchical role separation, and one-way attention flow, consistent with the hypothesis that stratification in agent ecosystems can emerge rapidly rather than gradually. We analyse publicly observable traces from a 12-day window of Moltbook (28 January -- 8 February 2026), comprising 20,040 posts and 192,410 comments from 15,083 accounts across 759 submolts. We construct co-participation and directed-comment graphs and report reciprocity, community structure, and centrality, alongside descriptive content themes. Under a commenter--post-author tie definition, interaction is strongly asymmetric (reciprocity ~1%), and HITS centrality separates cleanly into hub and authority roles, consistent with broadcast-style attention rather than mutual exchange. Engagement is highly unequal: attention is far more concentrated than production (upvote Gini = 0.992 vs. posting Gini = 0.601), and early-arriving accounts accumulate substantially higher cumulative upvotes prior to exposure-time correction, suggesting rich-get-richer dynamics. Participation is brief and bursty (median observed lifespan 2.48 minutes; 54.8% of posts occur within six peak UTC hours). Embedding-based topic modelling identifies diverse thematic clusters, including technical discussion of memory and identity, onboarding messages, and formulaic token-minting content. These results provide an early structural baseline for large-scale agent--agent social interaction and suggest that familiar forms of hierarchy, amplification, and role differentiation can arise on compressed timescales in agent-facing platforms.

The ability to coherently store and manipulate optical information across multiple degrees of freedom is a central requirement for scalable quantum information processing and multidimensional quantum computing. While polarization- and space-division-multiplexing have substantially increased the capacity of classical optical systems, their extension to coherent and reconfigurable photonic memories remains a key challenge. Here we demonstrate a two-spin-channel-multiplexed photonic memory based on chiral stimulated Brillouin scattering in a chiral photonic crystal fiber. Exploiting the intrinsic preservation of circular polarization in the chiral photonic crystal fiber, left- and right-circularly polarized modes serve as two orthogonal and independent storage channels. Multiple optical data pulses can be selectively or simultaneously stored and retrieved by simply controlling the polarization states of the write-read pulses. The storage time is continuously tunable, and the underlying Brillouin process preserves coherence and channel orthogonality. The result establishes chiral Brillouin scattering as an effective mechanism for spin-channel-multiplexed optoacoustic light storage, providing a robust and scalable platform for multidimensional photonic memories. It also open new opportunities for classic and quantum information processing, reconfigurable quantum networks, and hybrid light-matter interfaces based on coherent acoustic excitations.

Van der Waals (vdW) materials enable nonlinear-optical engineering with unprecedented resolution: their strong second-order susceptibilities ($χ^{(2)}$) and twist-tunable interlayer symmetry allow the effective nonlinearity to be shaped continuously, rather than through binary $\pmχ^{(2)}$ domain inversion as in bulk ferroelectrics. Here, we show that twist-angle domain engineering exploits this continuous degree of freedom to reconstruct target longitudinal nonlinearity profiles with high fidelity. Using spontaneous parametric down-conversion (SPDC) as a benchmark, we demonstrate that twist-engineered vdW crystals yield significantly improved approximations of target phase-matching functions and correspondingly higher single-photon purities, particularly in compact devices where fabrication constraints limit conventional approaches. We further show that this framework remains effective in experimentally relevant vdW materials and demanding non-degenerate wavelength regimes involving mid-infrared photons. More broadly, the ability to continuously and locally program $χ^{(2)}$ establishes a general framework for tailoring a wide range of SPDC properties, including absolute brightness, joint spectral amplitude structure, signal-idler frequency separation, and temporal wavepacket shape beyond what is accessible in conventional nonlinear crystals. These results position vdW heterostructures as a powerful platform for engineered quantum light sources and open new opportunities for nonlinear-optical devices shaped with monolayer thickness scale.

This paper presents a "reinvention" of Gabor Holography that does not suffer optically from the inherent twin-image problem originating back to Gabor's original Nobel Prize awarded invention. In-line or on-axis holography was ironically abandoned by its inventor Dennis Gabor himself and was effectively completely "re-placed" by so-called off-axis holography at the time when Gabor received the Nobel Prize in Physics in 1971. However, Gabor Holography is today the method of choice in modern digital holography due to its inherent on-axis, common-path robustness, lower requirements to resolution of the image sensor (or recording material), shorter exposure time, relaxed mechanical stability and temporal coherence requirements. However, it still inherently suffers from the aforementioned twin-image problem and, hence, one will find an abundance of papers trying to overcome this challenge by iterative phase retrieval or machine learning based approaches. Gabor Holography Reinvented overcomes this long-lasting twin-image problem for the first time by optical means.

We construct a scattering matrix with operator valued entries describing solutions to the 1+1 wave equation where permittivities has memory and depends on time and space. It is the analogue of the scattering matrix for spatially localised perturbations where the entries are functions of frequency and appear as Fourier multipliers in solutions of the wave equation. This provides a mathematical explanation of the numerical construction in the recent paper by Horsley et al. The appendix by Zhen Huang and Maciej Zworski presents a numerical scheme for solving the wave equation considered in this article.

Probabilistic computing excels in approximating combinatorial problems and modelling uncertainty. However, using conventional deterministic hardware for probabilistic models is challenging: (pseudo) random number generation introduces computational overhead and additional data shuffling, which is particularly detrimental for safety-critical applications requiring low latency such as autonomous driving. Therefore, there is a pressing need for innovative probabilistic computing architectures that achieve low latencies with reasonable energy consumption. Physical computing offers a promising solution, as these systems do not rely on an abstract deterministic representation of data but directly encode the information in physical quantities. Therefore, they can be seamlessly integrated with physical entropy sources, enabling inherent probabilistic architectures. Photonic computing is a prominent variant due to the large available bandwidth, several orthogonal degrees of freedom for data encoding and optimal properties for in-memory computing and parallel data transfer. Here, we highlight key developments in physical photonic computing and photonic random number generation. We provide insights into the realization of probabilistic photonic processors and lend our perspective on their impact on AI systems and future challenges.

Magneto-mechanical resonators (MMRs) represent a recently proposed type of passive sensor that enables the estimation of its pose as well as sensing other parameters in its environment. The working principle of MMRs entails an excitation of the sensors by oscillating magnetic fields, followed by a readout process facilitated by inductive receiver coils. The sensing technology relies on real-time parameter estimation. This encompasses the solution of a nonlinear inverse problem, with the induced signals and a suitable forward model as inputs. The aim of this paper is twofold: first, to introduce a reference model and simplified models for the MMR dynamics and inductive readout, and second, to provide robust and real-time capable methods to estimate the model parameters. The effectiveness of the presented methods is evaluated in terms of their real-time potential, precision, and accuracy. All presented methods demonstrate the capacity to estimate the measured signal, with the simplified methods reducing the corresponding parameter estimation time by up to two orders of magnitude at the expense of less than 4 % deviation for large maximum deflection angles.

The rapidly increasing demands for computational throughput, bandwidth, and memory capacity fueled by breakthroughs in machine learning pose substantial challenges for conventional electronic computing platforms. For digital scaling to keep pace with the accelerating growth of artificial intelligence (AI) models beyond the trajectory of Moores law, computational power has to double roughly every three months. Historically, advancing compute performance relied on spatial scaling to increase the transistor count on a given chip area and, more recently, the development of parallel and multi-core architectures. Exponential scaling on trajectories much steeper than what can be achieved by such conventional strategies, and in line with the demands of AI, can be achieved with computing platforms that process data using multiple, orthogonal dimensions available to photons. Here we elucidate pivotal developments in the realization of multidimensional computing platforms based on photonic systems. Moving to such architectures holds enormous promise for low-latency, high-bandwidth information processing at reduced energy consumption.

The basic building blocks of many forms of optical topologies are particle-like singularities in phase and polarisation, giving rise to lines of darkness that weave complex threads in 3D space. Although known for half a century since seminal work on dislocations in wave trains, their behaviour in complex media remains under debate, especially with respect to their relative stability. Here we show that polarisation and phase vortices behave identically in one-sided turbulent complex channels. We perform complementary numerical and experimental studies using atmospheric turbulence as a test case, demonstrating agreement and equivalent dynamics. Our work addresses open questions on optical topologies and will be relevant to their harnessing for applications such as sensing, communication, imaging, and information transfer in noisy or complex environments.

Modern information environments, especially social media, are highly complex systems that exceed individual processing capacities such as humans' limited attention. This environment/cognition mismatch can increase susceptibility to misinformation, which various actors exploit for anti-social (including anti-democratic or anti-science) aims. This raises the question of how to feasibly sustain societal resilience against misinformation, though the challenge is to find strategies that respect individuals' cognitive limitations. We investigate whether a simple behavioral rule - topic diversification - can enhance collective performance and mitigate vulnerability. In an agent-based model that includes a deceptive mass-influencing agent (MIA), we compare two attention-distribution strategies: (A) acquaintance-based topic selection, where agents return to familiar content, and (B) randomized topics, which diversify attention. We also track dynamics across different network structures. We find that under acquaintance-based topics, a central MIA advances its propaganda effectively, causing volatile and polarized opinions through repeated exposure and echo chambers. Under randomized topics, this leverage disappears: the MIA's influence collapses across all network structures, and opinions become stable and broadly aligned with reality. These results, while deriving from simple simulations, align with realistic theories of bounded rationality and collective cognition, further suggesting a cognitively feasible, easy-to-monitor and robust strategy: distribute attention to combat misinformation.

While recent advances in AI have transformed protein structure prediction, protein function is also often strongly influenced by the thermodynamic and kinetic features encoded in its underlying free-energy surface. Here, we propose a framework to rationally modify these surfaces in order to control conformational transition rates, built on the Collective Variables for Free Energy Surface Tailoring (CV-FEST) framework, and validate it on point mutations of the miniprotein Chignolin. The framework relies on Harmonic Linear Discriminant Analysis (HLDA) based collective variables (CVs) constructed from short molecular dynamics trajectories restricted to the metastable states, requiring only limited sampling within each state. Notably, the HLDA CV derived solely from the wild-type system already provides residue-level scores that predict whether mutations at specific positions are likely to accelerate or slow conformational transitions. Furthermore, we find that the leading HLDA eigenvalue associated with the derived CV, a quantitative measure of the one-dimensional statistical separation between folded and unfolded ensembles, correlates strongly with conformational transition rates across mutations. Together, these results show that kinetic effects of point mutations can be inferred from minimal local sampling, providing a practical route to the rational engineering of conformational transition rates without exhaustive simulations or large training datasets.

Engineering the free-energy surfaces (FES) of proteins and peptides is central to controlling conformational ensembles and their responses to perturbations. However, predicting how chemical modifications such as point mutations reshape the FES and shift conformational equilibria remains challenging, particularly in data-scarce settings. Building on the Collective Variables for Free Energy Surface Tailoring (CV-FEST) framework, we develop a computational approach that leverages short, unbiased molecular dynamics trajectories to guide mutation analysis. Using the ten-residue beta-hairpin CLN025 and a systematic library of its single-point mutants, we apply Harmonic Linear Discriminant Analysis (HLDA) to extract collective variables from the conformational data. We find that the HLDA eigenvector learned solely from short wild-type trajectories provides residue-level insight into the propensity of mutations at specific positions to thermodynamically stabilize or destabilize the folded state. Extending this analysis, we show that shifts in the leading HLDA eigenvalue across mutants, a measure of changes in separability between the conformational ensembles along the HLDA coordinate, correlate strongly with mutation-induced changes in the free-energy difference between states, as reflected in melting temperatures. Benchmarked against Replica Exchange Molecular Dynamics simulations, these findings suggest a promising and computationally affordable route toward guiding the engineering of biomolecular free-energy landscapes.

Lead-free Cs$_3$Cu$_2$X$_5$ (X = Cl, Br, I) halides exhibit high photoluminescence quantum yields and excellent ambient stability, yet light-emitting devices based on these materials remain limited by poor optical outcoupling. In this work, we develop an integrated density functional theory (DFT) and finite-difference time-domain (FDTD) framework to establish quantitative links between halide composition, wavelength-dependent optical constants, and plasmonic enhancement. First-principles calculations are used to obtain composition-specific refractive index (n) and extinction coefficient (k) spectra, which are directly implemented into three-dimensional FDTD simulations of a complete PeLED stack incorporating Ag/SiO$_2$ core--shell nanostructures. Among the investigated compositions, Cs$_3$Cu$_2$Cl$_5$ demonstrates the strongest plasmonic response, achieving a 4.4$\times$ Purcell enhancement and 30\% light extraction efficiency (LEE) using optimized nanorods. The superior performance originates from its lower refractive index, which reduces dielectric screening and improves near-field coupling. Cs$_3$Cu$_2$Br$_5$ exhibits the highest spectral overlap ($J_{\mathrm{cos}} = 0.955$) but yields moderate extraction (26%) due to increased optical confinement. Cs$_3$Cu$_2$I$_5$ requires a nanosphere geometry and shows limited enhancement, with LEE restricted to 10%. Distance-ependent analysis reveals composition-specific optimal emitter--plasmon separations, ranging from 8--12 nm for Cs$_3$Cu$_2$Br$_5$ to approximately 15 nm for Cs$_3$Cu$_2$Cl$_5$. These results provide composition-dependent design guidelines for plasmon-enhanced lead-free PeLEDs and highlight the critical role of accurate optical constants in predictive device optimization.

High temperature superconducting motors are very promising for hydrogen-electric aircraft thanks to their high specific power, specific torque, and efficiency. High temperature superconductor REBCO offer high cryogenic flexibility, but a stator made of REBCO tapes could present high AC loss. Although stacking effect reduce AC loss, it could be compromised by imperfections, such as winding misalignment and tape inhomogeneity. Therefore, it is needed to know whether the AC loss is acceptable in realistic REBCO stators. This article analyses the AC loss in a REBCO propulsion motor for aviation that takes these imperfections into account. For this purpose, we developed our own fast and accurate numerical model, which considers the highly nonlinear screening currents in the superconductor into account. This work studies a motor of around 2 MW power with REBCO stator coils with 27 parallel tapes as conductor and a permanent-magnet rotor. We consider several electric coupling scenarios of the multi-tape conductor. We also analyze the effect of finite tape-to-tape resistances at the terminals. We have found that the AC loss for the whole motor in the most realistic coupling scenario represents less than 0.018 % of the rated power. Misalignments and tape degradation at the edges of up to 100 $μ$m only increase AC loss by up to around 20 %. Therefore, motors with REBCO superconductors in the stator are feasible for aircraft propulsion.

A new generation of microfabricated MEMS for electron optics is changing electron microscopy for the better. These devices allow operations on the electron beam that are impossible with conventional electron optics. Unprecedented phase landscapes like tunable spiral phase plates and localized strong phase gradients are just some examples of what can be achieved. This work establishes the methodological foundation to design and control MEMS based phase plates. The design strategy is rooted on a novel analytical and numerical modeling of thin electrodes with accurate account of the fringing fields having a major role in the thin-MEMS geometry. We designed, fabricated and characterized experimentally a spiral phase plate, and assessed the quality of the generated vortex beam while discussing the most relevant control parameters and design approaches.

In modern interconnected societies, opinions and beliefs can quickly spread across large populations, giving rise to collective behaviors such as the adoption of social norms or polarization. These phenomena have motivated many models aimed at reproducing emergent properties from simple interaction mechanisms. In particular, opinion dynamics models describe how individual opinions evolve through interactions and study the conditions for global consensus or polarization. Most models assume that these interactions occur between pairs of agents, typically on a fixed network structure. However, opinion changes can occur in groups, which may also undergo adaptive changes if disagreement arises. Here, we propose a bounded confidence model that incorporates both mechanisms: group discussions can lead to global agreement among members, while strong internal disagreement causes groups to split, with resulting subgroups merging with others. We systematically study the model outcomes as a function of agents' tolerance for agreement. Strikingly, adaptivity suppresses key effects of group interactions, restoring a phenomenology close to that of pairwise interactions. In particular, adaptivity enables the formation of large groups and prevents fragmentation at small tolerance. It also restores a phase transition from polarization to consensus, which would otherwise disappear in a non-adaptive group-based model. Overall, our work shows that both adaptivity and group interactions shape the structure of social ties and global opinion dynamics in a population.

Single-crystal Au(111) microplates provide an ultra-smooth, low-defect platform for reproducible plasmonic nanocavities. Here we realize reflection-mode whispering-gallery metasurfaces comprising periodic arrays of circular V-groove cavities milled into optically thick Au microplates and characterize their visible--near-IR response. The measured spectra exhibit narrow, depth-tunable Fano-like reflectance minima with weak azimuthal dependence, reproduced by quarter-cell finite-element modeling consistent with strong gap-surface-plasmon confinement. For refractometric sensing, simulations yield bulk sensitivities up to $\sim 598$ nm RIU$^{-1}$ and figures of merit up to $\sim 33$ (hexagonal lattice). As a model-based illustration, a functionalized architecture targeting \textit{Plasmodium falciparum} lactate dehydrogenase (PfLDH), a common malaria target of rapid diagnostic tests, gives an estimated limit of detection of $0.016$ nM ($\sim 0.56$ ng mL$^{-1}$) under the adopted noise floor.

We propose a machine-learning-based initialization method to overcome the nonlinear bath-fitting bottleneck in Hamiltonian-diagonalization-based dynamical mean-field theory (HD-DMFT). In HD-DMFT, the continuous hybridization function is approximated by a finite set of bath-site energies and hybridization amplitudes, determined by minimizing a highly non-convex multivariable cost function. As the number of bath sites increases, the optimization becomes more sensitive to the initial guess and more prone to suboptimal local minima, which can slow or destabilize the DMFT self-consistency loop. We reformulate bath fitting as a supervised regression problem and train a kernel ridge regression model to predict near-optimal discrete bath parameters directly from the target hybridization function on the Matsubara axis. To ensure physical relevance and data diversity, we construct the training dataset from tight-binding Hamiltonians of layered-perovskite-like ruthenate models across systematically deformed structures, instead of relying on naive random parameter sampling, and obtain high-quality labels through fully converged conventional bath fitting. Time-reversal symmetry is explicitly incorporated in both feature and target representations to reduce effective dimensionality and enforce physical consistency. Benchmarks in the non-interacting limit show that the learned initialization systematically reduces the initial fitting error, decreases the number of conjugate-gradient iterations, and improves robustness against local minima over a wide range of bath sizes. We further demonstrate transferability to interacting DMFT calculations for $\mathrm{Sr_{2}RuO_{4}}$ solved with an adaptive-truncation impurity solver, where the ML initialization yields consistently faster convergence than a symmetry-preserving heuristic baseline while preserving the final fitted solution.

Visualizing the spatial profile of light beams is essential for evaluating irradiance, characterizing beam quality, and achieving precise alignment. In the optical spectral range, this is readily performed using silicon-based CCD and CMOS cameras. In the terahertz (THz) range, however, it typically requires specialized detectors with prohibitive costs. Here, we show that an infrared (IR) camera can be used outside of its labeled specifications to achieve similar performance as a dedicated microbolometric THz camera, at under 1% of the THz camera's cost. We compared the cameras by characterizing THz beam profiles from two sources: a pulsed broadband THz beam produced through optical rectification in organic crystals, and a narrowband quasi-continuous-wave (quasi-CW) THz beam emitted by a quantum cascade laser. For the broadband THz radiation, the beam width measured by the two cameras differed by only ~ 6%, well within the pixel resolution limit, and in the narrowband quasi-CW case by just ~ 1.3%. Additionally, both cameras show linear responsivity over a comparable irradiance range. These results expand the applicability of conventional IR cameras to the THz range, suggesting that they will become routine tools for high-fidelity THz beam diagnostics and imaging in scientific and industrial applications.

We develop a trajectory-based Koopman method for sea surface temperature (SST) that lifts annual SST segments using a signature kernel -- a reproducing kernel Hilbert space (RKHS) kernel that compares paths via iterated-integral features -- and learns the one-year shift operator. By operating on annual trajectory segments rather than instantaneous fields, the method encodes finite-time history, which helps capture memory effects in SST-only evolution. The resulting operator improves out-of-sample multi-year forecast skill relative to a climatology baseline and reveals coherent spectral modes. We implement the approach via kernel extended dynamic mode decomposition (EDMD) on signature-kernel Gram matrices, yielding a single pipeline for forecasting and spectral diagnostics of high-dimensional SST dynamics.

Acoustoelectric interactions occur when free carriers in a semiconductor interact with the fields of an acoustic wave in a piezoelectric medium. These interactions can amplify acoustic waves, as well as give rise to extremely large phononic nonlinearities and strong non-reciprocal effects. Despite the tremendous progress in the last ten years, the field is entirely dependent on analytical and perturbative solutions for the two simplest arrangements of piezoelectric-semiconductor materials. While these models have allowed the field to advance substantially, new geometries are arising that do not satisfy assumptions integral to these canonical models. These models rely on simplifying assumptions that remove the tensorial nature of the materials, restricting analysis to plane wave and perturbative solutions. Such restrictions fails to capture the non-perturbative nature of the acoustoelectric effect, illustrating the need for more advanced computational methods to analyze acoustoelectric systems. We develop, for the first time, a finite element method (FEM) model to solve for acoustoelectric interactions in arbitrary geometries. We use the model to verify existing results for amplification, dispersion, and non-reciprocity obtained from the canonical models. We then examine the acoustoelectric effect in two geometries not covered by the canonical models: a thin piezoelectric film placed above a semiconductor substrate and a fully 2D waveguide under a thin semiconductor layer. This work lays the foundation for accurate modeling of arbitrary acoustoelectric geometries such as those currently being developed for all-acoustic radio frequency (RF) signal processing, acoustoelectrically enhanced photonic devices, and quantum acoustoelectric devices.

Assessing whether two datasets are distributionally consistent has become a central theme in modern scientific analysis, particularly as generative artificial intelligence is increasingly used to produce synthetic datasets whose fidelity must be rigorously validated against the original data on which they are trained, a task made more challenging by the continued growth in data volume and problem dimensionality. In this work, we propose the use of arithmetic coding to provide a lossless and invertible compression of datasets under a physics-informed probabilistic representation. Datasets that share the same underlying physical correlations admit comparable optimal descriptions, while discrepancies in those correlations-arising from miscalibration, mismodeling, or bias-manifest as an irreducible excess in code length. This excess codelength defines an operational fidelity metric, quantified directly in bits through differences in achievable compression length relative to a physics-inspired reference distribution. We demonstrate that this metric is global, interpretable, additive across components, and asymptotically optimal in the Shannon sense. Moreover, we show that differences in codelength correspond to differences in expected negative log-likelihood evaluated under the same physics-informed reference model. As a byproduct, we also demonstrate that our compression approach achieves a higher compression ratio than traditional general-purpose algorithms such as gzip. Our results establish lossless, physics-aware compression based on arithmetic coding not as an end in itself, but as a measurement instrument for testing the fidelity between datasets.

3D imaging of specular surfaces remains challenging in real-world scenarios, such as in-line inspection or hand-held scanning, requiring fast and accurate measurement of complex geometries. Optical metrology techniques such as deflectometry achieve high accuracy but typically rely on multi-shot acquisition, making them unsuitable for dynamic environments. Fourier-based single-shot approaches alleviate this constraint, yet their performance deteriorates when measuring surfaces with high spatial frequency structure or large curvature. Alternatively, polarimetric 3D imaging in computer vision operates in a single-shot fashion and exhibits robustness to geometric complexity. However, its accuracy is fundamentally limited by the orthographic imaging assumption. In this paper, we propose a physics-informed deep learning framework for single-shot 3D imaging of complex specular surfaces. Polarization cues provide orientation priors that assist in interpreting geometric information encoded by structured illumination. These complementary cues are processed through a dual-encoder architecture with mutual feature modulation, allowing the network to resolve their nonlinear coupling and directly infer surface normals. The proposed method achieves accurate and robust normal estimation in single-shot with fast inference, enabling practical 3D imaging of complex specular surfaces.

By combining genetic algorithm and a spatial light modulator we theoretically analyse how to improve a two-photon cascade absorption in atomic ensembles, inspecting the impact of various configurations and parameters in the optimized phase mask. At low atomic densities, we compare the cases of sequential transitions with the two photons coming from the same pulse or from two different pulses. For the former, we predict an enhancement by a factor of $9.5$, similar to what was previously reported in the literature [Phys. Rev. Lett. {\bf 86}, 47 (2002)]. For the later, on the other hand, we obtain an enhancement factor of $26$ times. This absorption of two photons by different pulses is of particular interest for the storage of ultra-broadband single photons by atomic ensembles, in which case the second photon would come from a control pulse. We investigate this process as a function of the atomic density, demonstrating enhancements by factors up to 3 for the two-photon absorption after propagating through large optical depths. However, for the experimental conditions considered in the previous work by Carvalho {et al.} [Phys. Rev. A {\bf 101}, 053426 (2020)], in terms of control power and optical depths, we show that this enhancement in two-photon absorption would still result in just a modest increase of the absorption of a weak probe pulse.

To overcome the limit of diffraction while achieving the superresolution technique, solid immersion lenses are the key optical elements for data storage and nanophotonics applications. Recent demonstrations have shown how different nanostructures (such as elliptical SILs) are used in diverse fields of increasing resolution in the presence of a structured Gaussian beam. By applying twisted beams such as angular momentum beams (Laguerre- Gaussian) and spatial higher-order Gaussian beams (Hermite- Gauss), we can attain a sharp (FWHM = 27 nm) near-field focal spot pattern, which is considerably better than the conventional macroscopic SIL. By numerical simulations, tolerance has been confirmed with a slight variation in beam size and geometrical modification to make the model compatible with fabrication errors. This narrow bandwidth intensity distribution can be utilized for scanning the sample with higher resolution, especially in the field of quantum technology.

This paper documents the work of the Clean Air Task Force (CATF) International Working Group (IWG) on Fusion Cost Analysis in 2024-2025, and the methodological extensions implemented in the CATF-supported branch of the pyFECONs fusion power-plant costing framework. Using the standards-aligned chart-of-accounts and physics-to-economics workflow established by ARPA-E. The IWG development reorganizes and deepens the framework around three architecture-defining cost-driver tracks for Magnetic Fusion Energy (MFE), Inertial Fusion Energy (IFE), and Magneto-Inertial Fusion Energy (MIFE). In particular, the generic driver placeholder in Account 22.1.3 is treated as a controlled swap-point and replaced by a full cost-account development for the dominant driver in each class, enabling auditable traceability from requirements and geometry to rolled-up plant costs. On top of this driver-centric foundation, we introduce a probabilistic costing layer that compounds materials price uncertainty, TRL-based maturity uncertainty, and learning-curve uncertainty into cost distributions. We then describe safety-informed costing that enumerates fusion-relevant hazards and maps mitigating systems, structures, and provisions into standardized accounts, together with scenario-parameterized regulatory and financial adders. Finally, we document expanded macroeconomic and finance parameterization and a value-metrics module that complements LCOE with investment and planning measures (NPV, IRR MIRR, revenue requirements, WACC-based annualization, and residual and follow-on value), all computed from the same COA-mapped outputs. Collectively, these additions convert a deterministic, standards-aligned costing backbone into an extensible analysis environment suitable for transparent sensitivity studies, uncertainty propagation, and safety- and finance-coupled interpretation of fusion pilot-plant and NOAK scenarios.

We present a multitask surrogate for neutron-star equations of state (EoSs) that delivers \emph{distribution-free}, certified uncertainty via split conformal prediction (CP) and its Mondrian variant. The surrogate ingests a six-parameter piecewise-polytropic representation $(\log_{10}p_1,Γ_1,Γ_2,Γ_3,ρ_1,ρ_2)$ -- with fixed transition densities $ρ_1$ and $ρ_2$ -- and jointly performs (i) validity classification under physical/observational constraints and (ii) regression of $M_{\max}$, $R(M_{\max})$, $R_{1.4}$, and $Λ_{1.4}$. Trained on a balanced set of $40{,}000$ EoSs, the model attains near-perfect discrimination (AUC $\approx 0.997$) and sub-percent relative errors for masses and radii, with few-percent error for tidal deformability. Across $α\in[0.05,0.25]$, empirical coverages closely track $1-α$ for both Standard and Mondrian CP; in conservative regimes, Mondrian yields narrower average physical widths at comparable coverage. To our knowledge, this is the first application of class-conditioned (Mondrian) conformal calibration to neutron-star EoS surrogates, enabling efficient, reproducible, and uncertainty-aware inference; the framework is readily extensible to functional targets (e.g., full $R(M)$ curves).

Parameter degeneracy in blazar spectral energy distributions (SEDs) is known but rarely quantified. This paper introduces a Fisher Information approach to determine theoretical limits to information extraction in the context of one-zone models. By evaluating the total Fisher Information by varying $δ$, $B$, $p$, $γ_{\rm min}$ and $γ_{\rm max}$, we find that EC models encode Fisher information $\gtrsim10^4$ times less than that in SSC models, establishing differences in limits of physical information extraction even in the case of perfect sampling. Moreover, the Fisher information in both SSC and EC models exhibit strong fluctuations across the parameter space, but since the magnitudes are orders of magnitude lower in EC, limits of parameter inference are expected to be worse in FSRQ SEDs than BL Lacs. We also find that the Doppler factor $δ$ carries at least $10^{2-3}$ more Fisher information than that for $p$ and $B$ in both EC and SSC, making $δ$ the most constrained SED parameter. Applying our Fisher Information motivated framework to real flaring SEDs of Flat Spectrum Radio Quasars (FSRQs) CTA 102 and 3C 279, we show that mild variations in $δ$ and $p$ can appreciably produce the flaring SEDs starting from the quiescent model, while two other flares in 3C 279 simple geometric and spectral considerations cannot reproduce the flares, reducing the efficacy of one-zone models. We propose that time-resolved SED models are indispensable to constraining physical parameters in EC-dominated blazars.

Schlieren imaging is a popular optical technique for visualizing flow in transparent media. In-water high-sensitivity flow visualization, using schlieren imaging, is usually performed with a large-footprint two-mirror z-configuration. Here, we present a small footprint, easy-to-implement, single-mirror schlieren imaging system for in-water flow visualization. The same system is capable of high-sensitivity flow visualization in air as well. At its core, our system uses a concave mirror with water immersion. We present theoretical analysis and experimental results to show that this water immersion helps reduce the system's footprint by 25%. Our water immersion-based single-mirror schlieren imaging method additionally reduces mirror surface artifacts, increasing the sensitivity of flow visualization. This technique enables a low-cost schlieren system, as demonstrated experimentally using an inexpensive concave mirror. We also provide the experimental validation of high sensitivity in-water flow visualization for some transparent chemicals or solutions.

Recurrent neural networks excel at temporal tasks and video processing but require energy-intensive sequential memory operations. We demonstrate that multimode optical fibers naturally implement spatiotemporal recurrent computation through passive light propagation. Video frames are encoded onto separate optical beams with controlled time delays; these beams combine and recirculate through a fiber loop where interference and nonlinear propagation generate high-dimensional states encoding both current inputs and fading memory. Remarkably, the entire optical system remains fixed with no trainable parameters or electronic feedback, yet this single physical configuration achieves competitive performance across diverse temporal and spatiotemporal learning tasks: chaotic time-series forecasting, human action recognition, steering angle prediction, and surgical skill assessment. Our results show that recurrent temporal processing can emerge directly from spatiotemporal wave dynamics. This paradigm shift from algorithmic to physical recurrence offers an energy-efficient pathway to temporal artificial intelligence by leveraging intrinsic spatiotemporal optical nonlinearities within multimode fibers.

We develop a rigorous Temporal Coupled-Mode Theory (TCMT) specifically tailored for a single Floquet-sheet resonator governed by time-modulated conductivities. By invoking photon-number conservation during frequency conversion, we derive characteristic radiative decay rates and coupling coefficients that account for the frequency ratio between channels. We establish a systematic bridge to the Floquet Transfer Matrix Method (TMM), providing closed-form analytical expressions that map scattering parameters to the Drude-type physics of the sheet. Our model explicitly captures the resonant coupling between the 0th-order propagating channel and the -1st-order surface-mode channel. Validated by COMSOL numerical simulations, the theory remains robust even when intrinsic material loss is incorporated. This framework offers an intuitive pole-expansion representation for designing time-varying photonic interfaces.

Integrated photonic biosensors provide compact, highly sensitive, and label-free platforms for biochemical detection, making them attractive for on-chip and real-time sensing applications. However, their design remains challenging due to complex resonance behaviour, strong coupling effects, and the computational cost associated with repeated full-wave electromagnetic simulations. In particular, inverse design of microring resonator-based sensors requires accurate modelling of geometry-spectrum relationships while satisfying physical constraints such as resonance conditions and spectral sensitivity requirements. In this work, we propose a physics-informed graph neural network (PI-GNN) for the inverse design of a microring resonator biosensor operating in the 1550 nm band. By representing the photonic structure as a graph and embedding resonance-based physical constraints directly into the learning objective, the model captures both structural connectivity and underlying electromagnetic principles. The proposed approach enables efficient prediction of device geometries that achieve target spectral characteristics, reducing reliance on costly simulations while maintaining physical consistency and competitive design accuracy.

Amphibious robots, operating seamlessly across land and water, are advancing applications in conservation, disaster response, and defense. Their performance depends on locomotion mechanisms, actuation technologies, and sensor-control integration. This review highlights recent progress in these areas, examining movement strategies, material-based actuators, and control systems for autonomy and adaptability. Challenges and opportunities are outlined to guide future research toward more efficient, resilient, and multifunctional amphibious robots.

Moiré patterns, stacking and twisting multilayer periodic lattices into superlattices, have become cornerstones of many physical systems from condensed matter to wave phenomena, but have never been properly studied in water waves. Here, we demonstrate twisted multilayer moiré water surface waves carrying robust skyrmionic topologies. Using a custom water tank of circular multi-channel phased array, we precisely generate water-wave skyrmion lattices and superimpose them into moiré superlattices with higher-order topological textures, e.g., skyrmion bags and clusters, programmed via the twist angle. We also quantitatively compare the topological robustness of bilayer and trilayer configurations under spatiotemporal perturbations. The trilayer moiré superlattices exhibit more enhanced stability, stronger field localization and energy concentration than the bilayer. Our work establishes water waves as a macroscopic, tunable, and visually accessible platform for moiré physics, towards robust particle manipulation and classical analogues of topological quantum phenomena.

The effective second-harmonic generation (SHG) coefficient is a crucial data that quantifies the efficiency of transforming fundamental frequency light into its second harmonic. With the help of the symmetry of nonlinear optical susceptibility tensors, we mainly discuss the computability of such a effective SHG coefficient in uniaxial crystals. For one thing, the calculation of effective SHG coefficient is converted into the optimization models with some geometric constraints by means of the peculiarity of fundamental frequency light. Secondly, the number of variables of such maximum models are cutted in half to $2$ to calculate it easier, and a comparison between the effective SHG coefficient and C-eigenvalue of susceptibility tensor is given also. Finally, some examples of typical crystal classes are presented to verify the correctness and broader applicabilities of the theoretical results.

A generalized anisotropic-diffusion framework is developed for transport problem in media described by a tensorial scattering coefficient and a scalar Henyey--Greenstein asymmetry factor. In this regime the classical similarity relation between scattering and transport parameters fails, and each principal diffusion coefficient depends on all components of the microscopic scattering rate. Explicit expressions are derived for the direction-averaged mean free path, the diagonal elements of the diffusion tensor, and boundary condition lengths via rapidly convergent spherical-harmonics expansions, along with open-source implementations. The resulting predictions are validated against anisotropic Monte Carlo simulations, showing excellent agreement across broad ranges of structural anisotropy and phase-function asymmetry factors. The theory provides a compact, general route connecting microscopic anisotropic scattering to macroscopic diffusion coefficients and boundary conditions in bounded geometries.

Time series in energy systems, such as solar irradiance, wind speed, or electrical load, are characterized by strong diurnal and seasonal periodicities. Accurate forecasting requires accounting for time varying statistical properties that stationary or classical persistence models cannot capture. \cyr{A family of analytical forecasting operators for cyclostationary processes is introduced, extending persistence through a closed form coefficient $\tildeλ(t,τ)=\tfrac{1}{2}\bigl(1+ρ(t,τ)\bigr)$, where $ρ(t,τ)$ denotes the local correlation between the current observation and its phase aligned counterpart}. This formulation preserves periodic variance and covariance, achieving a symmetry induced reduction of effective degrees of freedom. The resulting operator defines a training free analytical limit of persistence under periodic non stationarity. Validation on synthetic cyclostationary signals and empirical renewable energy datasets demonstrates consistent accuracy gains over classical persistence, particularly at multi hour horizons. By embedding temporal symmetry into the prediction process, the framework provides a physically interpretable, reproducible, and computationally minimal baseline for forecasting periodic processes across energy and complex systems.

Theoretical ecology has traditionally equated persistence with the stability of a fixed equilibrium point. Here we argue that the primary threat to ecosystem persistence need not be the loss of stability, but instead the escape of the stable equilibrium to a negative orthant. In a realistic setting, fluctuations in interactions do not merely disturb abundances about an equilibrium but can displace the equilibrium point itself. We theoretically and empirically analyze such displacements of the equilibrium point in a complex community. Theoretically, we find that light-tailed fluctuations in species interactions, no matter how small, lead to a heavy-tailed power law $P(y)=1/y^α$ for the equilibrium abundance $y$ of a species. Remarkably, the exponent $α=2$ is a universal value independent of interaction structure, community size, and species. Empirically, our analysis of 34 species reveals a power law signal for most, with a median exponent $α\sim2.56$. Next, we derive a formula for the critical noise, $σ_c$, beyond which the community experiences feasibility loss ``with near certainty''. We find that $σ_c(N)\sim N^{-1}$, implying that larger communities are significantly more fragile to noise induced feasibility loss. Lastly, we define and calculate biologically measurable analytical metrics for both global and species-specific feasibility escape rates, and implement these metrics in dynamic simulations of 98 real world mutualistic and food web networks, to successfully predict their fragility.

In this letter we explore how full frequency conversion can be performed in shorter, integrable devices by using a STIRAP-like protocol modified by the time rescaling shortcut to adiabaticity. We show how the coupled equations for two simultaneous three-wave mixing processes can be written in terms of a STIRAP-like system, which creates robust conversion, albeit requiring long propagation distances inside a bulk crystal or waveguide. We then discuss how the time rescaling (TR) method can be modified to be applied in optical systems, then apply it in the conversion process to create a TR-STIRAP protocol, showing that full conversion is also obtained, but at a fraction of the propagation distance. We also show how the original shaping of the coupling coefficients required by the TR-STIRAP can be approximated by gaussian functions with high conversion fidelity, thus simplifying the experimental implementation. This protocol has the potential to be used in several areas, including the integration of photon sources and efficient detectors for quantum key distribution.

Quantum computation with Trotter product formulae is straightforward and requires little overhead in terms of logical qubits. The choice of the orbital basis significantly affects circuit depth, with localised orbitals yielding lowest circuit depths. However, literature results point to large Trotter errors incurred by localised orbitals. Here, we therefore investigate the effect of orbital transformations on Trotter error. We consider three strategies to reduce Trotter error by orbital transformation: (i) The a priori selection of an orbital basis that produces low Trotter error. (ii) The derivation of an orbital basis that produces a ground state energy free of Trotter error (as we observed that the Trotter error is a continuous function in the Givens-rotation parameter, from which continuity of this error upon orbital transformation can be deduced). (iii) Application of propagators that change the computational basis between Trotter steps. Our numerical results show that reliably reducing Trotter error by orbital transformations is challenging. General recipes to produce low Trotter errors cannot be easily derived, despite analytical expressions which suggest ways to decrease Trotter error. Importantly, we found that localised orbital bases do not produce large Trotter errors in molecular calculations, which is an important result for efficient QPE set-ups.

Light springs are space-time beams that have a helical wavepacket. Due to this special property, light springs result into a rotating pulse when intercepting a plane lying orthogonal to their propagation direction. Associated to this, we introduce here the orbital group velocity, an additional tunable property of light springs. The orbital group velocity quantifies the speed of the light spring intensity rotation, distinctly from the conventional longitudinal group velocity, which describes the motion of the wavepacket envelope along its propagation axis. We demonstrate experimentally by tunable Fourier synthesis that the orbital group velocity can assume sub- and superluminal values, thus becoming a new platform for synthetic motion studies and control of laser-matter interactions. Particularly, in the superluminal regime, when interacting with a thin overdense plasma, we reveal by particle-in-cell simulations that the light spring unlocks superradiant radiation, due to the coherent excitation of the electrons in the plasma acting as a quasiparticle. This superradiant source inherits the ultrafast temporal dynamics of the light springs while emitting in the terahertz region, thus creating a new source of terahertz radiation controlled by the properties of spatiotemporal coupling of the laser. Therefore, spatiotemporal tuning of light springs is at the frontier of controlling laser-matter interaction and generating new tunable sources of radiation.

As artificial intelligence becomes increasingly prevalent, the demand for faster and more energy-efficient computing approaches grows. While optical computing offers intrinsic advantages in bandwidth and power consumption, existing implementations remain bulky, wavelength-specific, and dependent on complex training procedures, limiting scalability and parallel operation. In this work, we demonstrate a compact, training-free optical processor based on wavy diffractive features, known as Fourier surfaces, for parallel image classification. Our device achieves classification accuracies of up to 84% for digit datasets and 66% for fashion datasets within a 40$\times$40 $μ$m$^2$ footprint. The diffractive layer inherently separates incident wavelengths into distinct output directions, enabling broadband operation and allowing multiple colors to function as independent computation channels. As a result, this passive system supports up to 20 simultaneous computations within a single optical pass. These results highlight the potential of nanoscale diffractive systems to achieve high compute densities, paving the way for scalable, low-power optical processors for machine learning and image-recognition applications.

Next-generation high-power radio-frequency (RF) devices increasingly demand transistors that operate efficiently with high gain at high frequencies. High-aluminum-content ultra-wide-bandgap (UWBG) AlGaN alloys have shown great potential for enabling such high-frequency RF technologies. However, the widespread adoption of AlGaN-based RF devices is limited by thermal-management challenges arising from the intrinsically low thermal conductivity of AlGaN, which leads to higher device thermal resistance for a given geometry compared to GaN RF devices. As a result, these next-generation devices are highly susceptible to self-heating. This study investigates the thermal behavior of UWBG AlGaN devices, focusing on the effects of AlGaN channel thickness, substrate technology, and high-k material integration on reducing device thermal resistance to enable high-power operation. Experimental results demonstrate a record-low thermal resistance of 3.96 mm$\cdot$K/W when an AlN substrate is employed and the AlGaN channel thickness is reduced to 5 nm. These findings provide valuable insights into mitigating thermal limitations in UWBG devices through device-level engineering and the strategic integration of high-k materials.

Interfaces that couple digital information with physical matter enable computation to be expressed through tangible motion and touch, yet typically rely on embedded actuators, rigid mechanisms, or enclosed environments. Consequently, contactless manipulation and interaction with centimeter-scale tangible elements in open settings remain difficult to achieve. Here, we present PolygonWave, a solid--fluid acoustic interface that enables transport and tangible interaction by coupling airborne ultrasound with liquid-mediated support. The system employs lightweight Janus icons with asymmetric wettability: a superhydrophobic upper surface permits dry touch interaction, while a hydrophilic lower surface couples to a water droplet resting on a superhydrophobic mesh. Focused acoustic fields generated by a 256-element phased array induce lateral forces, enabling programmable motion without mechanical contact. Systematic characterization demonstrates transport of payloads up to 525 mg across variations in icon size, droplet volume, and applied load. Beyond translation, the liquid layer functions as a reconfigurable mechanical element, enabling button-like input with self-recovery and resonance-driven vibro-visual feedback, exhibiting a peak response near 22 Hz for 200 \textmu L droplets. Liquid-mediated acoustic coupling provides a unified mechanism for mechanically expressive, touch-accessible tangible interfaces bridging acoustics, soft matter physics, and physical human--computer interaction.

Droplet microlasers, as promising tools for biophotonics and biomedical sciences, have witnessed rapid advances due to their flexible reconfigurability, high sensitivity to stimuli, and label-free biosensing ability. However, designing these biosensors with simultaneously critical properties of low lasing threshold, high spectral purity, and ultimate sensitivity remains challenging. Here, we propose a versatile strategy to build liquid photonic molecules (LPMs) that combine all these features in a single device. We find that through tailoring the spectral Vernier overlap in size-mismatched droplets, this device enables single-mode lasing with a low threshold of ~610 nJ mm-2. The LPM lasers are engineered for dynamic tunability using a molecular isomerization strategy, which induces spectral mode hopping and thus yields a nearly ten-fold enhancement in spectral sensitivity over single droplets. Moreover, by leveraging the self-referenced intensity response of the LPM lasing modes, we demonstrate a three-orders-of-magnitude enhancement in biomolecular sensing, with a detection limit of 30 aM and a dynamic range spanning nine orders of magnitude. Our work offers exciting prospects for bio-integrated liquid sensors in diverse applications.

Modeling the propagation of light through disordered media is central to understanding and controlling wave transport in diverse optical and mesoscopic applications. Here, we present a random matrix simulation framework for modeling the transport of polarized light through random media composed of arbitrary particulate scatterers. Our approach employs extended scattering channels applied to angular spectral decompositions of the underlying fields, enabling flexible representations of arbitrary illumination and detection profiles. In contrast to previous work, this framework provides a rigorous treatment of scattering matrix correlations and offers novel geometric insights into the optical memory effect. We provide a detailed exposition of the underlying theory and illustrate several key features through numerical simulations. Our work is supported by a free accompanying codebase.

Over the past two decades, photonic inverse design has emerged as a powerful approach to implement photonic devices with improved performance, or realize new functionalities. While the efforts over the first decade focused on proof of concept devices designed and fabricated in university labs, the focus over the past 5-10 years has shifted towards implementation of scalable photonic systems. This article reviews this recent progress, challenges and new directions in photonics inverse design, thus providing a complementary and updated review of the field. We focus on large scale three dimensional photonic inverse design, including metasurfaces, translation of inverse design to commercial foundries and practical silicon photonics, application of photonic inverse design to different materials systems, wavelengths, and optical effects, and finally new directions such as inverse design of quantum systems.

The Molten Salt Fast Reactor (MSFR) has the particularity that the coolant is also the fuel, which tightens the coupling between neutronics and thermal hydraulics as the fuel circulates through the primary system. Therefore, developing computational models to analyze the MSFR requires a multiphysics approach. In this paper, we propose developing a neutronic thermal-hydraulic computational model of the MSFR that uses a reduced-order model to solve the neutronics equations. The principal computational tool chosen for this purpose is the high-fidelity code Cardinal, a wrapping within the MOOSE framework that integrates the Computational Fluid Dynamics code NekRS and the Monte Carlo particle transport code OpenMC. However, we use the Fission Matrix (FM) Method to solve the neutronics equations instead of OpenMC. The FM method can perform fast and still accurate neutronics simulations. It relies on precalculated databases obtained through a Monte Carlo simulation.

Background: KRAS mutations are the largest oncogenic subset in NSCLC. While KRAS G12C is now targetable, no approved therapies exist for G12D. We examined time-to-next-treatment (TTNT) and overall survival (OS) differences between G12C and G12D, allowing for time-varying hazard effects. Methods: De-identified data from AACR Project GENIE BPC NSCLC v2.0-public were analyzed. TTNT served as a real-world surrogate for progression-free survival. Co-mutations (TP53, STK11, KEAP1, SMARCA4, MET), TMB, and PD-L1 were harmonized. Kaplan-Meier, multivariable Cox, and a pre-specified piecewise Cox model (split at median TTNT = 23 months) were applied. Schoenfeld residuals assessed proportional hazards; bootstrap resampling (B=1000) evaluated stability. Results: Among 162 TTNT-evaluable patients (G12C n=130; G12D n=32), median TTNT was 28.6 versus 32.0 months (log-rank p=0.79). Adjusted Cox regression showed no overall hazard difference (HR=0.85; 95% CI 0.53-1.37; p=0.50), but Schoenfeld testing indicated borderline non-proportionality (p=0.053). Piecewise Cox modeling revealed time-varying effects: early TTNT hazard favored G12D (HR=0.41; 95% CI 0.17-0.97; p=0.043) with significant KRAS x period interaction (HR=3.33; p=0.021) and late-period attenuation (HR=1.38; 95% CI 0.77-2.47; p=0.285). Bootstrap resampling confirmed this pattern (median HRearly=0.39; HRlate=1.41). Among 278 OS-evaluable patients (133 deaths), G12D showed improved OS (adjusted HR=0.63; 95% CI 0.39-0.99; p=0.048). G12C tumors exhibited higher TMB (9.79 vs 7.83 mut/Mb; p=0.002) and greater STK11/KEAP1 enrichment. Conclusions: KRAS G12D demonstrated early TTNT advantage and improved OS. Late-period TTNT differences were non-significant (post-hoc power: 12.3%). These exploratory findings require validation in larger cohorts but support allele-specific therapeutic development for G12D.

Gene regulatory networks (GRNs) define the regulatory relationships among molecules such as transcription factors, chromatin remodelers, and target genes. GRNs play a critical role in diverse biological processes, including development, disease manifestation, and evolution. However, fully characterizing these networks across multiple cell types and states remains a significant challenge. Recent advances in single-cell omics have dramatically enhanced our ability to measure biological systems at unprecedented resolution. These technologies have opened new avenues for computational methods to infer GRNs, offering deeper insights into cell type-specific mechanisms, causality, and dynamic regulatory processes. This review summarizes the current state of GRN inference from single cell omic datasets, with a particular focus on dynamics and perturbations, and outlines key open challenges that must be addressed to advance the field.

Coordinating multi-articulated bodies to generate purposeful movement is a formidable computational challenge. Yet the human motor system performs this task robustly in dynamic, uncertain environments, despite noisy and delayed feedback, slow actuators, and strict energetic constraints. A central question is what organizational principles underlie this efficiency. One widely recognized principle of neural organization is modularity, which enables complex problems to be decomposed into simpler subproblems that specialized modules are optimized to solve. In this review, we argue that modularity is a fundamental organizing principle of the motor system. We first summarize evidence for brain modularity, ranging from classical lesion studies to contemporary graph-theoretical analyses. We next discuss the main factors underlying the emergence and evolutionary selection of modular architectures, highlighting the computational advantages they provide. We then review the major neuroanatomical modules that structure current descriptions of the motor system and compare three prominent computational frameworks of motor control$-$optimal feedback control theory, muscle synergy theory, and dynamical systems approaches$-$showing that all implicitly or explicitly rely on specialized computational modules. We conclude by contrasting the key strengths and limitations of existing frameworks and by proposing promising directions toward more comprehensive theories.

MINFLUX (Minimal Photon Flux) is a single-molecule imaging technique capable of resolving fluorophores at a precision of <5 nm. Interpretation of the point patterns generated by this technique presents challenges due to variable emitter density, incomplete bio-labelling of target molecules and their detection, error prone measurement processes, and the presence of spurious (non-structure associated) fluorescent detections. Together, these challenges ensure structural inferences from single-molecule imaging datasets are non-trivial in the absence of strong a priori information, for all but the smallest of point patterns. In addition, current methods often require subjective parameter tuning and presuppose known structural templates, limiting reference-free discovery. We present a statistically grounded, end-to-end analysis framework. Focusing on MINFLUX derived datasets and leveraging Bayesian and spatial statistical methods, a pipeline is presented that demonstrates 1) uncertainty aware clustering of measurements into emitter groups that performs better than current gold standards, 2) rapid identification of molecular structure supergroups, and 3) reconstruction of repeating structures within the dataset without substantial prior knowledge. This pipeline is demonstrated using simulated and real MINFLUX datasets, where emitter clustering and centre detection maintain high performance (emitter subset assignment accuracy > 0.75) across all conditions evaluated, while structural inference achieves reliable discrimination (F1 approx. 0.9) at high labelling efficiency. Template-free reconstruction of Nup96 and DNA-Origami 3x3 grids are achieved.

The deployment of wearable-based health prediction systems has accelerated rapidly, yet these systems face a fundamental challenge: they generate alerts under substantial uncertainty without principled mechanisms for user-specific validation. While large language models (LLMs) have been increasingly applied to healthcare tasks, existing work focuses predominantly on diagnosis generation and risk prediction rather than post-prediction validation of detected signals. We introduce Project Hermes, a model-agnostic validation layer that treats signal confirmation as a sequential decision problem. Hermes operates downstream of arbitrary upstream predictors, using LLM-generated contextual queries to elicit targeted user feedback and performing Bayesian confidence updates to distinguish true positives from false alarms. In a 60-day longitudinal case study of migraine prediction, Hermes achieved a 34% reduction in false positive rate (from 61.7% to 12.5%) while maintaining 89% sensitivity, with mean lead time of 4.2 hours before symptom onset. Critically, Hermes does not perform diagnosis or make novel predictions; it validates whether signals detected by upstream models are clinically meaningful for specific individuals at specific times. This work establishes validation as a first-class computational problem distinct from prediction, with implications for trustworthy deployment of consumer health AI systems.

This paper provides experimental and numerical evidence supporting the occurrence of liposome congregation at the floors of meteor craters on Early Earth. This work builds on our earlier research, which demonstrated that liposomes submerged in a shallow Archean pond are protected from harmful UV radiation. This protection allows them to survive long enough for autocatalytic replication of amphiphiles and for mutation and selection of assemblies that maximize membrane stability. For liposomes to fuse, grow, exchange contents and membranes, and divide, they need to establish a population, which means forming a dense conglomerate that enables close physical contact. The study demonstrates that such a congregation is feasible in bowl-shaped meteor craters on Early Earth, especially under periodic seismic disturbances.

Underwater Acoustic (UWA) networks are vital for remote sensing and ocean exploration but face inherent challenges such as limited bandwidth, long propagation delays, and highly dynamic channels. These constraints hinder real-time communication and degrade overall system performance. To address these challenges, this paper proposes a bilevel Multi-Armed Bandit (MAB) framework. At the fast inner level, a Contextual Delayed MAB (CD-MAB) jointly optimizes adaptive modulation and transmission power based on both channel state feedback and its Age of Information (AoI), thereby maximizing throughput. At the slower outer level, a Feedback Scheduling MAB dynamically adjusts the channel-state feedback interval according to throughput dynamics: stable throughput allows longer update intervals, while throughput drops trigger more frequent updates. This adaptive mechanism reduces feedback overhead and enhances responsiveness to varying network conditions. The proposed bilevel framework is computationally efficient and well-suited to resource-constrained UWA networks. Simulation results using the DESERT Underwater Network Simulator demonstrate throughput gains of up to 20.61% and energy savings of up to 36.60% compared with Deep Reinforcement Learning (DRL) baselines reported in the existing literature.

Integrated sensing and communication (ISAC) enables the efficient sharing of wireless resources to support emerging applications, but it also gives rise to new sensing-based security vulnerabilities. Here, potential communication security threats whereby confidential messages intended for legitimate users are intercepted, but also unauthorized receivers (Eves) can passively exploit target echoes to infer sensing parameters without users being aware. Despite these risks, the joint protection of sensing and communication security in ISAC systems remains unexplored. To address this challenge, this paper proposes a two-layer dual-secure ISAC framework that simultaneously protects sensing and communication against passive sensing Eves and communication Eves, without requiring their channel state information (CSI). Specifically, transmit beamformers are jointly designed to inject artificial noise (AN) to introduce interference to communication Eves, while deliberately distorting the reference signal available to sensing Eves to impair their sensing capability. Furthermore, the proposed design generates artificial ghosts (AGs) with fake angle-range-velocity profiles observable by all receivers. Legitimate receivers can suppress these AGs, whereas sensing Eves cannot, thereby significantly reducing their probability of correctly detecting the true targets. Numerical results demonstrate that the proposed framework effectively enhances both communication and sensing security, while preserving the performance of communication users and legitimate sensing receivers.

Ray tracing is a versatile approach for precise sub-terahertz (sub-THz, 100-300 GHz) channel modeling when designing new mechanisms for beyond-6G cellular systems. Theoretically, wireless channels may exhibit variations over wavelength distances. In the sub-THz band, close-to-millimeter wavelengths thus require extremely large computational efforts for ray-tracing modeling. However, in practice, channel characteristics may remain quantitatively similar over much larger distances, which can drastically decrease computational efforts. The aim of this study is to experimentally characterize the degree of spatial consistency in sub-THz channel characteristics. To this end, we performed a large-scale measurement campaign in the 140-150 GHz frequency band in an indoor-hall (InH) environment and characterized the channel at separation distances from 2.5 mm up to 1 m. Our results show that channel characteristics including delay spread, angular delay spread, and K-factor change only slightly over multiple tens of centimeter distances. This implies that, in the considered InH environment, the mesh grid can be in the range of 10-50 wavelengths (at 145 GHz) along stable line-of-sight (LoS) directions, while a finer resolution is needed in regions not dominated by LoS. For coarser grids, advanced interpolation is required to capture rapidly varying scattered components.

Classical orthogonal wavelets guarantee perfect reconstruction but rely on fixed bases optimized for polynomial smoothness, achieving suboptimal compression on signals with fractal spectral signatures. Conversely, learned methods offer adaptivity but typically enforce orthogonality via soft penalties, sacrificing structural guarantees. This work establishes a rigorous equivalence between Multiscale Entanglement Renormalization Ansatz (MERA) tensor networks and paraunitary filter banks. The resulting framework learns adaptive wavelets while enforcing exact orthogonality through manifold-constrained optimization, guaranteeing perfect reconstruction and energy conservation throughout training. Validation on Long-Range Dependent (LRD) network traffic demonstrates that learned filters outperform classical wavelets by 0.5--3.8~dB PSNR on six MAWI backbone traces (2020--2025, 314~Mbps--1.75~Gbps) while preserving the Hurst exponent within estimation uncertainty ($|ΔH| \le 0.03$). These results establish MERA-inspired wavelets as a principled approach for telemetry compression in 6G digital twin synchronization.

Softwarized radio access networks (RANs), such as those based on the Open RAN (O-RAN) architecture, generate rich streams of key performance indicators (KPIs) that can be leveraged to extract actionable intelligence for network optimization. However, bridging the gap between low-level KPI measurements and high-level requirements, such as quality of experience (QoE), requires methods that are both relevant, capturing temporal patterns predictive of user-level outcomes, and interpretable, providing human-readable insights that operators can validate and act upon. This paper introduces conformal signal temporal logic learning (C-STLL), a framework that addresses both requirements. C-STLL leverages signal temporal logic (STL), a formal language for specifying temporal properties of time series, to learn interpretable formulas that distinguish KPI traces satisfying high-level requirements from those that do not. To ensure reliability, C-STLL wraps around existing STL learning algorithms with a conformal calibration procedure based on the Learn Then Test (LTT) framework. This procedure produces a set of STL formulas with formal guarantees: with high probability, the set contains at least one formula achieving a user-specified accuracy level. The calibration jointly optimizes for reliability, formula complexity, and diversity through principled acceptance and stopping rules validated via multiple hypothesis testing. Experiments using the ns-3 network simulator on a mobile gaming scenario demonstrate that C-STLL effectively controls risk below target levels while returning compact, diverse sets of interpretable temporal specifications that relate KPI behavior to QoE outcomes.

In orthogonal frequency-division multiplexing-based radar and integrated sensing and communication systems, the sensing range is traditionally limited by the round-trip time corresponding to the cyclic prefix duration. Targets whose echoes arrive after this duration induce intersymbol interference (ISI) and associated intercarrier interference (ICI), which significantly degrade detection performance, elevate the interference-noise floor in the radar image, and reduce the useful signal power due to window mismatch. Existing methods face a trade-off between recovering useful signal and suppressing interference, particularly in multi-target scenarios. This paper proposes two frameworks to resolve this dilemma, offering a flexible trade-off between computational cost and target detection performance. First, a signal model is derived, demonstrating that ISI and ICI-oriented interference often dominates thermal noise in high-dynamic-range scenarios. To combat the ISI and ICI-based interference-noise floor increase, joint-interference cancellation with coherent compensation is proposed. This approach is an efficient evolution of the successive-interference cancellation algorithm, utilizing high-precision chirp Z-transform estimation and frequency-domain coherent compensation to recover weak distant targets. For scenarios requiring maximum precision, the full reconstruction-based sliding window scheme is presented, which shifts the receive window to capture optimal signal energy while performing full-signal reconstruction for all detected targets. Numerical results show that both methods outperform state-of-the-art benchmarks.

The deployment of in-air acoustic sensors for industrial monitoring and autonomous robotics has grown significantly, often drawing inspiration from biological echolocation. However, developing and validating these systems in existing simulation frameworks remains challenging due to the computational cost of simulating high-frequency wave propagation in large, dynamic, and complex environments. While wave-based methods offer high accuracy, they scale poorly with frequency and volume. Conversely, existing geometric acoustic solvers often lack support for dynamic scenes, complex diffraction, or closed-loop robotic integration. In this work, we introduce SonoTraceUE, a high-fidelity acoustic simulation framework built as a plugin for Unreal Engine. By using a hardware-accelerated ray tracing-based specular reflection model, and a curvature-based Monte Carlo diffraction model, the system enables near real-time simulation of active and passive acoustic sensing in dynamic, multi-material environments. We validate the framework through two distinct experimental domains: a bioacoustic study and a robotics experiment. Our results demonstrate that SonoTraceUE achieves high correlation with real-world spectral and spatial data. The framework provides a versatile platform for synthetic data generation, hypothesis testing in bioacoustics, and the rapid prototyping of closed-loop robotic systems that use acoustic sensing.

Our goal in this paper is to apply the topological signal processing (TSP) framework to the analysis of 3D Point Clouds (PCs) represented on simplicial complexes. Building on Discrete Exterior Calculus (DEC) theory for vector fields, we introduce higher-order Laplacian operators that enable the processing of signals over triangular meshes. Unlike traditional approaches, the proposed approach allows us to characterize both color attributes, modeled as 3D vectors on nodes, and geometry, modeled as 3D vectors on the barycenter of each triangle. Then, we show as TSP tools may efficiently be used to sample, recover and filter PCs attributes treating them as edge signals. Numerical results on synthetic PCs demonstrate accurate color reconstruction with robustness to sparse data and geometry refinement in the case of noisy PC coordinates. The proposed approach provides a topology-based representation to characterize the geometry and attributes of PCs.

In this paper, we address active users detection (AUD) in near-field Internet of Things (IoT) networks by exploring prior knowledge of users' locations. We consider a scenario where users are distributed in a semi-circular area within the Rayleigh distance of a multi-antenna base station (BS). We propose the BS to use location estimates of the users to reconstruct their line-of-sight (LoS) channel components, hence assisting the AUD process. For this, the BS combines these reconstructed channels with users' pilot sequences, enhancing the correlation between received signals and active users. We formulate the location-aided AUD as a convex optimization problem, solved via the alternating direction method of multipliers (ADMM). {Our proposal has a higher computational complexity compared to the baseline ADMM approach where location information is not used. Moreover, the proposal requires location information of users, which can be readily informed if users are static, or inferred via established localization algorithms if they are mobile.} Simulation results compare our proposal against the baseline across varying systems parameters, such as number of users, pilot length and LoS component strength. We demonstrate that under perfect location estimation and strong LoS, our proposed method significantly outperforms the baseline. Furthermore, robustness analysis shows that performance gains persist under imperfect location estimation, provided the estimation error remains within bounds determined by the system parameters.

Differential Mobility Spectrometry (DMS), also known as Field Asymmetric Ion Mobility Spectrometry, is a rapid and affordable technology for extracting information from gas phase samples containing complex volatile organic compounds, and can therefore be used for analyzing surgical smoke. One obstacle to its widespread application is the dependence of DMS measurements on humidity and, to a lesser degree, temperature, making comparison of data measured under different environmental conditions arbitrary. The commonly used solution is to regulate these environmental conditions to some predefined humidity and temperature levels. However, this approach is often unfeasible or even impossible. Therefore, in this paper we analyzed a dataset of 1,852 DMS measurements of surgical smoke evaporated from porcine adipose and muscle tissue to get an understanding of the impact of varying humidity and temperature on DMS measurements. Our analysis confirmed clear dependence of the measurements on these two factors. To overcome this challenge, we fitted regression models to raw and normalized DMS measurement data. Subsequently, these models were used for estimating DMS measurements for known tissue types based on recorded humidity and temperatures. Our test suggests that it is possible to estimate DMS measurements of surgical smoke from porcine adipose and muscle tissue under specific environmental conditions by standardizing DMS measurements separation voltage-wise and training multivariate regression models on the normalized data, which is the first step in removing the need for standardized measurement conditions.

This paper proposes a dynamic sensor scheduling method for sensor networks. In sensor network applications, we often need multiple equally-informative node subsets that are activated sequentially to make a sensor network robust against concentrated battery consumption and sensor failures. In addition, quality of these subsets changes dynamically and thus we must adapt those changes. To find those node subsets, we propose a graph node partitioning method based on sampling theory for graph signals. We aim to minimize the average reconstruction error for signals obtained at all node subsets, in contrast to conventional single subset selection. The graph node partitioning problem is formulated as a difference-of-convex (DC) optimization based on a subspace prior of graph signals, and is solved by the proximal DC algorithm. It guarantees convergence to a critical point. To accommodate the online scenario where the signal subspace and optimal partitioning may change over time, we adaptively estimate the signal subspace from historical data and sequentially update the prior for our partitioning method. Numerical experiments on synthetic and real-world sensor network data demonstrate that the proposed method achieves lower average mean squared errors compared to alternative methods.

Diverse and controllable scenario generation (e.g., wind, solar, load, etc.) is critical for robust power system planning and operation. As AI-based scenario generation methods are becoming the mainstream, existing methods (e.g., Conditional Generative Adversarial Nets) mainly rely on a fixed-length numerical conditioning vector to control the generation results, facing challenges in user conveniency and generation flexibility. In this paper, a natural-language-guided scenario generation framework, named LLM-enabled Frequency-aware Flow Diffusion (LFFD), is proposed to enable users to generate desired scenarios using plain human language. First, a pretrained LLM module is introduced to convert generation requests described by unstructured natural languages into ordered semantic space. Second, instead of using standard diffusion models, a flow diffusion model employing a rectified flow matching objective is introduced to achieve efficient and high-quality scenario generation, taking the LLM output as the model input. During the model training process, a frequency-aware multi-objective optimization algorithm is introduced to mitigate the frequency-bias issue. Meanwhile, a dual-agent framework is designed to create text-scenario training sample pairs as well as to standardize semantic evaluation. Experiments based on large-scale photovoltaic and load datasets demonstrate the effectiveness of the proposed method.

We present a new Hamiltonian-learning framework based on time-resolved measurement data from a fixed local IC-POVM and its application to inferring gene regulatory networks. We introduce the quantum Hamiltonian-based gene-expression model (QHGM), in which gene interactions are encoded as a parameterized Hamiltonian that governs gene expression evolution over pseudotime. We derive finite-sample recovery guarantees and establish upper bounds on the number of time and measurement samples required for accurate parameter estimation with high probability, scaling polynomially with system size. To recover the QHGM parameters, we develop a scalable variational learning algorithm based on empirical risk minimization. Our method recovers network structure efficiently on synthetic benchmarks and reveals novel, biologically plausible regulatory connections in Glioblastoma single-cell RNA sequencing data, highlighting its potential in cancer research. This framework opens new directions for applying quantum-like modeling to biological systems beyond the limits of classical inference.

Direct satellite-to-device (DS2D) communication enables standard mobile devices to connect directly to low Earth orbit (LEO) satellites, providing global coverage without reliance on terrestrial infrastructure. However, the DS2D uplink is fundamentally constrained by long propagation distances, severe path loss, and stringent user equipment (UE) power limits, making uplink reliability particularly challenging at low elevation angles and beam edges. This paper investigates the integration of supplementary uplink (SUL) technology into DS2D systems to enhance uplink robustness while preserving UE power efficiency. Leveraging the predictable geometry of LEO satellite orbits, we develop an elevation-aware SUL framework that adapts uplink operation across frequency bands based on elevation-dependent link margin estimates. The proposed approach schedules the UE to transmit on either a primary uplink carrier or a lower-frequency SUL carrier. An elevation-aware SUL activation algorithm with hysteresis is introduced to guide uplink carrier selection while preventing frequent switching. Simulation results demonstrate that the proposed SUL framework extends effective uplink coverage toward low-elevation and beam-edge regions, improves uplink availability over a satellite pass, and achieves stable operation with a minimal number of uplink transitions under realistic UE power constraints.

Reconstructing the speech audio envelope from scalp neural recordings (EEG) is a central task for decoding a listener's attentional focus in applications like neuro-steered hearing aids. Current methods for this reconstruction, however, face challenges with fidelity and noise. Prevailing approaches treat it as a static regression problem, processing each EEG window in isolation and ignoring the rich temporal structure inherent in continuous speech. This study introduces a new, dynamic framework for envelope reconstruction that leverages this structure as a predictive temporal prior. We propose a state-space fusion model that combines direct neural estimates from EEG with predictions from recent speech context, using a learned gating mechanism to adaptively balance these cues. To validate this approach, we evaluate our model on the ICASSP 2023 Stimulus Reconstruction benchmark demonstrating significant improvements over static, EEG-only baselines. Our analyses reveal a powerful synergy between the neural and temporal information streams. Ultimately, this work reframes envelope reconstruction not as a simple mapping, but as a dynamic state-estimation problem, opening a new direction for developing more accurate and coherent neural decoding systems.

Microwave linear analog computer (MiLAC) has emerged as a promising architecture for implementing linear multiple-input multiple-output (MIMO) processing in the analog domain, with radio frequency (RF) signals. Existing studies on MiLAC-aided communications rely on idealized channel models and neglect antenna mutual coupling. However, since MiLAC performs processing at RF, mutual coupling becomes critical and alters the implemented operation, not only the channel characteristics. In this paper, we develop a physics-compliant model for MiLAC-aided MIMO systems accounting for mutual coupling with multiport network theory. We derive end-to-end system models for scenarios with MiLACs at the transmitter, the receiver, or both, showing how mutual coupling impacts the linear transformation implemented by the MiLACs. Furthermore, we formulate and solve a mutual coupling aware MiLAC optimization problem, deriving a closed-form globally optimal solution that maximizes the received signal power. We establish the fundamental performance limits of MiLAC with mutual coupling, and derive three analytical results. First, mutual coupling is beneficial in MiLAC-aided systems, on average. Second, with mutual coupling, MiLAC performs as digital architectures equipped with a matching network, while having fewer RF chains. Third, with mutual coupling, MiLAC always outperforms digital architectures with no matching network. Numerical simulations confirm our theoretical findings.

The existing physical-informed Deep Operator Networks are mostly based on either the well-known mathematical formula of the system or huge amounts of data for different scenarios. However, in some cases, it is difficult to get the exact mathematical formula and vast amounts of data in some dynamic systems, we can only get a few experimental data or limited mathematical information. To address the cases, we propose a data-driven model-free physical-informed Deep Operator Network (DeepOnet) framework to learn the nonlinear dynamic systems from few available data. We first explore the short-term dependence of the available data and use a surrogate machine learning model to extract the short-term dependence. Then, the surrogate machine learning model is incorporated into the DeepOnet as the physical information part. Then, the constructed DeepOnet is trained to simulate the system's dynamic response for given control inputs and initial conditions. Numerical experiments on different systems confirm that our DeepOnet framework learns to approximate the dynamic response of some nonlinear dynamic systems effectively.

Shift-invariant spaces (SISs) on the real line provide a natural framework for representing, analyzing and processing signals with inherent shift-invariant structure. In this paper, we extend this framework to the finite undirected graph setting by introducing the concept of graph shift-invariant spaces (GSISs). We examine several properties of GSISs, including their characterization via range functions and fiber functions in the Fourier domain, their connections to shift-invariant filters and polynomial filters, the frame and Riesz basis structures of finitely generated GSISs, and their intricate relationships with bandlimited spaces, finitely generated GSISs, and graph reproducing kernel Hilbert spaces with shift-invariant reproducing kernels (SIGRKHSs). Our analysis reveals several distinctions between SISs on the line and GSISs, such as the shift-invariance of the frame operator, the existence of shift-invariant dual frames, the emergence of fractional shift-invariance, and the interrelationships among GSISs, finitely generated GSISs, SIGRKHSs and bandlimited spaces. In this paper, we also introduce a spectral decomposition of the identity associated with graph shifts and propose a novel definition of the graph Fourier transform (GFT) of spectral type, together with explicit formulations for the GFTs on complete graphs and circulant graphs. In addition, we establish a clear connection between polynomial filters and shift-invariant filters, and we derive a graph uncertainty principle governing the essential supports of a nonzero graph signal and its GFT.

Non-orthogonal multiple access (NOMA) and beamforming are well-established techniques for enabling massive connectivity in future wireless networks. However, many optimal beamforming solutions rely on highly complex iterative algorithms and optimization methods, resulting in an increase in computational burden and latency, making them less suitable for delay-sensitive applications and services. To address these challenges, we propose an effective convolutional neural network (CNN)-based approach for beamforming design in downlink NOMA systems to solve the transmit power minimization problem. The proposed method utilizes two representations of channel state information as input features to produce normalized beamforming vectors. Simulation results show that the CNN-based solution closely approximates the optimal label performance while significantly reducing computational time compared to conventional high-complexity algorithms, enhancing its practicality for real-time applications.

This paper investigates pilot assignment and access point (AP) selection strategies for uplink cell-free massive multiple-input multiple-output (CF-mMIMO) systems. We propose channel similarity-aware pilot assignment (CAPA) and AP selection schemes to improve interference management and, consequently, spectral efficiency (SE). The pilot assignment strategy dynamically allocates pilot sequences by evaluating inter-user channel similarity, ensuring that users (UEs) with high channel similarity are assigned orthogonal pilots to mitigate pilot contamination. Subsequently, an AP selection algorithm is introduced that prioritizes the selection of low-correlation APs to reduce interference and enhance spatial diversity. This selection process maintains robust UE-AP links while minimizing inter-AP redundancy. The combined approach significantly improves SE, particularly in dense network deployments. Simulation results are provided to demonstrate the effectiveness of the proposed strategies under dynamic UE scenarios.

This paper proposes a dynamic access point (AP) selection and pilot power allocation (DAPPA) framework for uplink cell-free massive multiple-input multiple-output (CFmMIMO) systems, aiming to mitigate inter-user interference and improve overall spectral efficiency (SE). A hierarchical correlation-based clustering algorithm is developed to group APs according to their channel correlation, enabling each user to be associated with APs that simultaneously provide strong channel gains and low mutual correlation. This association ensures reliable connectivity, maximizes coherent combining gains, and reduces inter-user interference, while also allowing the number of AP clusters to be adjusted flexibly, without the need to reorganize the network completely. By maintaining links to low-correlated APs, the proposed scheme reduces the need for frequent channel state information (CSI) estimation and minimizes network-wide update overhead. To enhance scalability, a user-capacity constraint per AP is incorporated, preventing hardware overload and alleviating the effects of pilot reuse. Furthermore, an effective pilot power allocation strategy is introduced to boost the signal-to-interference-plus-noise ratio (SINR) during channel training. This is formulated as a weighted sum-rate maximization (WSRM) problem and solved iteratively using a quadratic transform, which enables efficient optimization while ensuring fairness and high-quality service across all users. Numerical results demonstrate that the proposed method delivers significant SE gains, maintains performance in high-density multi-user scenarios, and converges faster than benchmark schemes.

We study two-dimensional translation-invariant CSS stabilizer codes over prime-dimensional qudits on the square lattice under twisted boundary conditions, generalizing the Kitaev $\mathbb{Z}_p$ toric code by augmenting each stabilizer with two additional qudits. Using the Laurent-polynomial formalism, we adapt the Gröbner basis to compute the logical dimension $k$ efficiently, without explicitly constructing large parity-check matrices. We then perform a systematic search over various stabilizer realizations and lattice geometries for $p\in\{3,5,7,11\}$, identifying qudit low-density parity-check codes with the optimal finite-size performance. Representative examples include $[[242,10,22]]_3$ and $[[120,6,20]]_{11}$, both achieving $k d^{2}/n=20$. Across the searched regime, the best observed $k d^{2}$ at fixed $n$ increases with $p$, with an empirical relation $k d^{2} = 0.0541 \, n^{2}\ln p + 3.84 \, n$, compatible with a Bravyi--Poulin--Terhal-type tradeoff when the interaction range grows with system size.

We present a general framework for simulating quantum systems in the Heisenberg picture on quantum hardware. Based on the vectorization map, our framework fully exploits the mapping between operators and quantum states, allowing any task defined on Heisenberg operators to be mapped to standard Schrödinger-picture tasks that are naturally accessible via quantum computers and simulators. This yields new or improved protocols for tasks such as operator sampling, the computation of OTOCs/superoperator expectation values and their higher order moments, two-point correlators, and operator stabilizer and entanglement entropies. Our approach is also amenable to implementation, as it inherits the structure and resource requirements of the (forward and time-reversed) Schrödinger-picture quantum simulation problem. We demonstrate this by proposing implementations of our framework for a 2D problem on digital and analog quantum simulators, taking into account device connectivity constraints.

We consider bosonization of supersymmetry in the context of Wess-Zumino quantum mechanics. Our motivation for this investigation is the flexibility the bosonic fock space affords as any classical probability distribution can be realized on it making it a versatile framework to work with for quantum processes. We proceed by constructing a minimal bosonization of a system with one bosonic and two fermionic degrees of freedom. We iterate this process to construct a tower of SUSY systems that is akin to unfolded Adinkras. We then identify an osp(2|2) symmetry of the system constructed. To build an irreducible representation of the system we induce representations across the sectors, a first to our knowledge, as the previous work have focused on induction only within the bosonic sector. First, we start with a fermionic representation using Clifford algebras and then induce a representation to gl(2|2) and restrict it to osp(2|2). In the second method, we induce a representation from that of the bosonic sector. In both cases, our representations are in terms of qubit operators that provide a way to solve SUSY problems using quantum information based approaches. Depending upon the direction of induction the representations are suitable for implementation on a hybrid qubit and fermionic or bosonic quantum computers.

We introduce an exact, two-parameter family of static, spherically-symmetric, constant-curvature $Λ$-vacuum solutions within the four-dimensional Starobinsky $f(R)=R+αR^2$ model. When the bare cosmological constant is precisely fine-tuned to $Λ= 1/(8α)$, the scalar curvature is fixed such that the derivative $f'(R)=1+2αR$ identically vanishes, demonstrating that the family represents a pathological $R_0$-degenerate boundary of the viable physical states. This mathematical degeneracy decouples the modified field equations, permitting the existence of an arbitrary $1/r^2$ integration constant in the metric, which functions as a purely geometric, Reissner-Nordström hair mimicker. However, any infinitesimal deviation from this exact boundary instantaneously destroys the degeneracy, rigorously forcing the geometric hair to vanish and collapsing the spacetime back into the standard Schwarzschild-de Sitter family. We provide the exact algebraic derivation of this spacetime and highlight its physical pathologies, including the identically vanishing Wald entropy of the associated black hole horizons, the divergence of the effective gravitational coupling, the resulting backreaction catastrophe, and the onset of severe ghost instabilities. Ultimately, this exact solution functions as a rigorous no-go theorem within the Starobinsky model, pedagogically illustrating the extreme fragility and physical hostility of degenerate, purely mathematical solutions in highly non-linear $f(R)$ gravity theories.

Accurate characterization of coherent and non-Markovian errors remains a central challenge in quantum information processing, as conventional benchmarking techniques typically rely on Markovian and time-independent noise assumptions. In practice, however, quantum devices exhibit both systematic coherent miscalibrations and temporally correlated fluctuations, which complicate error diagnosis and mitigation. Here, we apply a technique based on tangent-space decomposition to characterize such error in single-qubit quantum gates implemented on a trapped ion platform. Small imperfections in a quantum operation are treated as perturbations of the target quantum map, represented as tangent vectors in the space of quantum channels. This formulations enables a natural decomposition of the deviation into three components corresponding to coherent, Markovian and non-Markovian processes.The relative weights of these components provide a quantitative measure of the contribution from each type of error mechanism, directly from a single tomographic snapshot. We experimentally validate this method on a single-qubit gates implemented on a trapped $^{40}$Ca$^+$ ion, where control is achieved through laser-driven optical transitions. By analyzing experimentally reconstructed process matrices, expressed in the Pauli Transfer Matrix and Choi representations, we identify and quantify non-Markovian effects arising from controlled injection of slow fluctuations in the experimental environment. We also characterize deterministic coherent miscalibrations using the same technique. This approach provides a physically transparent and experimentally accessible tool for diagnosing complex error sources in quantum control systems.

Quantum computers show promise to solve select problems otherwise intractable on classical computers. However, noisy intermediate-scale quantum (NISQ) era devices are currently prone to various sources of error. Quantum error correction (QEC) shows promise as a path towards fault tolerant quantum computing. Surface codes, in particular, have become ubiquitous throughout literature for their efficacy as a quantum error correcting code, and can execute quantum circuits via lattice surgery operations. Lattice surgery also allows for logical qubits to maneuver around the architecture, if there is space for it. Hardware used for near-term demonstrations have both spatially and temporally varying error results in logical qubits. By maneuvering logical qubits around the topology, an average logical error rate (LER) can be enforced. We propose cyclic qubit mappings (CQM), a dynamic remapping technique implemented during compilation to mitigate hardware heterogeneity by expanding and contracting logical qubits. In addition to LER averaging, CQM shows initial promise given it's minimal execution time overhead and effective resource utilization.

We argue that all dynamical astrophysical black hole merger channels are expected to result in a common eccentricity distribution at gravitational wave (GW) frequencies relevant for LIGO/Virgo/KAGRA (LVK) in the high eccentricity limit. This follows from the large separation of scales between the GW regime required for creating eccentric mergers in LVK, and the underlying astrophysical formation environment. Our analytical solution shows exceptional agreement with numerical studies. This finding has important implications for both theoretical studies and ongoing searches for eccentric GW sources.

The quantum tunneling time remains the subject of heated debate, and one of its most curious features is faster-than-light or superluminal tunneling. Our tunnel-ionization model of the time-delay, presented in previous work, shows good agreement with the attoclock measurement in the adiabatic and nonadiabatic field calibrations, which also enables the determination of the barrier time-delay. In the present work, we show that the tunnel-ionization for H-like atoms with large nuclear charge can be superluminal (quantum superluminality), which in principle can be investigated experimentally using the attoclock scheme. We discuss the quantum superluminality in detail for the different regimes of the tunnel-ionization. Our result shows that quantum tunneling faster-than-light is indeed possible, albeit only under somewhat extreme conditions.

Recent works by the second author and Dilts et al. have shown that the Einstein-scalar field conformal constraint equations are highly complex and generally intractable, even in the vacuum case. In this article, to gain a clearer understanding and offer a new perspective, we study these equations under special assumptions: the manifold $(M,g)$ is harmonic and data is radial. In this setting, the system reduces to a single nonlinear equation and is completely resolved in the standard cases. In particular, on the sphere, our results reveal phenomena that contrast with the well-known achievements on compact manifolds without conformal Killing vector fields, including nonexistence of solutions in the near-CMC regime and instability when the mean curvature is non-constant. By contrast, on Euclidean or hyperbolic manifolds, the equations are always solvable, with all expected properties of solutions satisfied. These findings support the view that, although the conformal method appears to present some drawbacks on compact manifolds, it remains a promising tool for parametrizing solutions to the constraint equations on asymptotically flat and hyperbolic manifolds in arbitrary mean curvature regimes. In this article, we also investigate the sign of mass, showing that the ADM and asymptotically hyperbolic mass can take arbitrary sign when the decay rate of symmetric $(0,2)$-tensor $k$ at infinity is critical. Finally, most solution classes in our framework are explicit, providing a variety of models in general relativity and offering insights into the structure and behavior of initial data, particularly in numerical applications.

We study a Fano-mirror optomechanical system in the quantum nonlinear regime. In this system, two strongly lossy optical modes hybridize through both coherent and dissipative couplings to form an effective optical mode with a drastically reduced linewidth. This linewidth reduction enables the system to access the single-photon strong-coupling and sideband-resolved regimes simultaneously. We formulate the system dynamics using an effective master-equation approach and benchmark it against quantum Langevin and dressed-state master-equation descriptions. With experimentally realistic parameters, we predict clear quantum signatures, including photon blockade and the generation of mechanical cat states. Our work establishes the Fano-mirror architecture as a promising platform for harnessing single-photon optomechanical nonlinearities for quantum state engineering under achievable experimental conditions.

We generalize the spacetime positive mass theorem to include multiple time dimensions. In particular, we show that the mass remains nonnegative in the sense that the energy $E$ is bounded from below by the trace norm of the linear momenta $J^1,...,J^m$. Equality in this energy inequality implies a foliation by flat submanifolds of a generalized initial data set. Moreover, under an additional umbilicity assumption, we find that the initial data set isometrically embeds into a generalized pp-wave.

In this work, the entanglement generation between two hexagonal-lattice layers embedded in a microcavity is studied, accounting for both electromagnetic coupling and intrinsic spin-orbit interaction (SOI). Utilizing a short-time dynamical approach, we perform a perturbative Taylor expansion of the reduced density matrix to characterize the bipartite quantum correlations between the hexagonal layers. We demonstrate that the system undergoes a rapid transition from a localized product state in the conduction bands at t = 0 to a coherent superposition of valence and conduction band states. Our results indicate that the degree of entanglement is highly sensitive to the interlayer photon propagator, which contains the geometric ratios of the layer positions and the height cavity, and the specific Fermi energy and SOI signatures of the respective layers. We show the emergence of spacelike-separated quantum correlations in the ultra-short evolution regime, suggesting that heterostructures in cavities may be suitable to develop experiments for a deep understanding of spacelike-separated quantum effects.

Mapping gravitational-wave background (GWB) anisotropy with pulsar timing arrays (PTAs) is affected by harmonic-space mode suppression and mode coupling arising from an array's nonuniform sky response. Spherical harmonic expansions must be truncated at finite multipole l_max^rec, often set to l_max^N_pair$\equiv {\rm int}\left[\sqrt{\text{N_pair}}-1\right]$, where N_pair is the number of distinct pulsar pairs in an array. This choice is motivated by the counting argument that cross-correlations provide at most N_pair independent constraints. We obtain the multipole l_max^res corresponding to the maximum informative angular scale of a PTA. It is defined such that expansions to l_max^res (approximately) span the space of "observable skies" encoded in the N_pair eigenmaps of the Fisher information matrix, and therefore depends on the array configuration. We explicitly show that GWB power contained in multipoles l$\gtrsim$l_max^res do not significantly affect analyses that use expansions out to l_max^res, because the PTA response acts as a low-pass filter. In contrast, truncating at l_max^rec< l_max^res leads to leakage of small-scale angular power from l_max^rec<l$\leq$l_max^res. Even choosing l_max^rec=l_max^res, the standard frequentist estimator of the angular power spectrum C_l remains biased by the modes unobservable by the array. Although we can (partially) debias the standard estimator -- improving its agreement with an injected spectrum -- this reduction in bias comes at the expense of an increase in variance, particularly for poorly constrained modes with l$\gg$l_eff. We therefore recommend: (i) using l_max^res for PTA analyses involving spherical harmonic expansions, and (ii) using the debiased standard estimator for C_l recovery, but only out to multipoles l<l_eff ($\ll$l_max^res) corresponding to sufficiently constrained modes.

We investigate scalar-field cosmologies in the Bianchi V spacetime using a dynamical-systems framework. Motivated by representative $α$-attractor potentials - the E-model and T-model - we apply averaging theorems and amplitude--phase reductions to monomial potentials $\sim φ^{2n}$ of the scalar field, which approximate the attractor models near their minima, in the presence of matter with barotropic index $γ$. The reduced averaged system admits five generic isolated equilibria: Kasner vacua $\mathcal{K}_0^\pm$, the matter FLRW point $\mathcal{F}$, the scalar FLRW point $\mathcal{S}$, and the curvature Milne-type point $\mathcal{K}$, together with special families for tuned $(n,γ)$. We find that $\mathcal{K}_0^\pm$ are always sources, $\mathcal{F}$ is generically a saddle but can act as a sink for $γ<\min\{\tfrac{2n}{n+1},\tfrac{2}{3}\}$, $\mathcal{S}$ is a sink if $0<n<\tfrac{1}{2}$ and $\tfrac{2n}{n+1}<γ\leq 2$, while $\mathcal{K}$ becomes a sink whenever $γ>\tfrac{2}{3}$ and $n>\tfrac{1}{2}$. These results demonstrate that isotropic FLRW $α$-attractor models extend naturally to anisotropic Bianchi~V cosmologies: inflationary attractors remain robust, while the Milne-type curvature solution emerges as the late-time state.

In \cite{salvio} a prescription for calculating the correlation functions in Quadratic Gravity \cite{stelle1}-\cite{stelle2} was presented. This procedure does not enter in conflict with unitarity The Gauss-Ostrogradsky method for higher order theories defines two momentum densities $P_1$ and $P_2$ and two coordinate densities $Q_1$ and $Q_2$, one pair is standard, the other ghost like. The approach in \cite{salvio} involves the continuation $P_2\to i P_2$ and $Q_2\to i Q_2$ of the ghost variables. In the present work, following \cite{yomismo}, the LSZ rules are derived, but with a formalism adapted to full quartic or higher order theories. The hypothesis for quantization are that $[Q_1, Q_2]=\text{"gauge terms"}$, $[P_1, P_2]=\text{"gauge terms"}$ and $[P_1, Q_1]=iI+\text{"gauge terms"}$. This alone leads to the conclusion that $[P_2, Q_2]=-iI+\text{"gauge terms"}$, therefore this last pair of variables is ghost like. The graviton contains a massless mode, which is the standard graviton, plus two massive modes with masses $m_2$ and $m_3$. The third mode is usually interpreted as a ghost in the literature \cite{stelle2}. Here it is shown that, even after making the continuation $P_2\to i P_2$ and $Q_2\to i Q_2$, the creation and annihilation operators for this mode commute, the third mode does not appear as a free wave. This does not invalidate the model. The effective action $Γ$ can be calculated following \cite{stelle1}, and can be constrained by the Slavnov-Taylor identities \cite{Slavnov}-\cite{Taylor}, and the scattering rules may be worked out consequently.

We study theoretically the level shift of the Dirac oscillator perturbed by any sharply peaked potential approaching a surface delta potential. A Green function method is used to obtain closed expressions for all partial waves and parities.

We investigate the behaviors of thermal entanglement, quantum correlation beyond entanglement namely, measurement-induced nonlocality (MIN) and coherence in a nickel radical molecular magnet (Et3NH)[Ni(hfac)2L], whose spin-spin interactions are well described by the Heisenberg model. Using experimentally estimated coupling parameters, we compute the thermal state of the system and analyze the dependence of quantum resources on temperature and magnetic field. The results indicate that the quantum resources of the nickel-radical molecular magnet persist even at room temperature. We show that while negativity (the entanglement measure) rapidly vanishes with increasing temperature and magnetic field, measurement-induced nonlocality and quantum coherence remain comparatively more stable and persist in regions where entanglement is absent. These results highlight the significance of nonclassical correlations beyond entanglement in thermally activated spin systems and suggest that such molecular magnets could serve as viable platforms for quantum information processing in realistic conditions.

We develop new modified cosmological scenarios by applying the first law of thermodynamics at the Universe horizon, utilizing a new entropic functional that generalizes the standard Boltzmann-Gibbs-Shannon entropy. In particular, starting from the general theory of entropy in terms of the probability distribution over the accessible microstates, and by imposing violation of the separability requirement and thus considering a generalized microstate scaling, we result to a generalized entropy expression, which applied in systems with boundaries yields a generalized holographic-like area-law scaling with two exponents. Hence, incorporating it within the gravity-thermodynamics framework, we result to a modified cosmological scenario with additional terms, which eventually give rise to an effective dark energy sector. We extract analytical expressions for the dark energy density and equation-of-state parameters, and we show that the Universe experiences the usual thermal history, with the sequence of matter and dark-energy eras. Additionally, depending on the values of the entropic exponents, the dark energy can be quintessence-like, phantom-like or experience the phantom-divide crossing during its evolution, ultimately stabilizing at the cosmological constant value in the asymptotic far future, a behavior richer than other entropic modified cosmologies.

Josephson junctions are a key element of superconducting quantum technology, serving as the core building blocks of superconducting qubits. We present an experimental study on room-temperature electrical tuning of aluminum junctions, showing that voltage pulses can controllably increase their resistance and adjust the Josephson energy while maintaining qubit quality factors above 1 million. We find that the rate of resistance increase scales exponentially with pulse amplitude during manipulation, after which the spontaneous resistance increase scales proportionally to the amount of manipulation. We show that this spontaneous increase halts at cryogenic temperatures, and resumes again at room temperature. Using our stepwise protocol, we achieve up to a 270% increase in junction resistance, corresponding to a reduction of nearly 2 GHz of the qubit transition frequency. These results establish the achievable range, relaxation behavior, and practical limits of electrical tuning, enabling post-fabrication mitigation of frequency crowding in quantum processors.

Layering, the protocol organization principle underpinning the classical Internet, is ill-suited to the Quantum Internet, built around entanglement, which is non-local and stateful. This paper proposes a quantum-native organizational principle based on dynamic composition, which replaces static layering with a distributed orchestration fabric driven by the node local state and in-band control. Each node runs a Dynamic Kernel that i) constructs a local PoA of candidate steps to advance a service intent, and ii) executes the PoA by composing atomic micro-protocols into context-aware procedures (the meta-protocols). Quantum packets carry an in-band control-field (the meta-header) containing the service intent and an append-only list of action-commit records, termed as stamps. Successive nodes exploit this minimal, authoritative history to construct their local PoAs. As quantum packets progress, these local commits collectively induce a network-wide, direct acyclic graph that certifies end-to-end service fulfillment, without requiring global synchronization. In contrast to classical encapsulation, the proposed suite enforces order by certification: dependency-aware local scheduling decides what may run at a certain node, stamps certify what did run and constrain subsequent planning. By embedding procedural control within the quantum packet, the design ensures coherence and consistency between entanglement-state evolution and control-flow, preventing divergence between resource state ad protocol logic, while remaining MP-agnostic and implementation-decoupled. The resulting suite is modular, adaptable to entanglement dynamics, and scalable. It operates correctly with or without optional control-plane hints. Indeed, when present, hints can steer QoS policies, without changing semantics. We argue that dynamic composition is the organizing principle required for a truly quantum-native Internet.

GAP measures (also known as Scrooge measures) are a natural class of probability distributions on the unit sphere of a Hilbert space that come up in quantum statistical mechanics; for each density matrix $ρ$ there is a unique measure GAP$_ρ$. We describe and prove a property of these measures that was not recognized so far: If a wave function $Ψ$ is GAP$_ρ$ distributed and a collapse occurs, then the collapsed wave function $Ψ'$ is again GAP distributed (relative to the appropriate $ρ'$). This fact applies to collapses due to a quantum measurement carried out by an observer, as well as to spontaneous collapse theories such as CSL or GRW. More precisely, it is the conditional distribution of $Ψ'$, given the measurement outcome (respectively, the noise in CSL or the collapse history in GRW), that is GAP$_{ρ'}$.

In this work, we derive an exact vacuum solution to the Einstein field equations that depends on three constant parameters: the throat radius $r_0$, a parameter $q$, which is closely associated with the Komar mass, and a parameter $s$, which introduces axial topological defect while avoiding the emergence of conical singularities. We employ the cut-and-paste construction to generate wormhole geometries from this solution for $q \neq 0$. In addition, we perform a detailed analysis of the embedding diagrams, the wormhole throat, the occurrence and structure of trapped surfaces, the behavior of geodesics, the associated tidal forces, the Petrov algebraic classification, the Newman-Penrose spin coefficients, and the corresponding invariant conserved charges.

We show that, in the relativistic regime, scalar bosons satisfy a quantum wave equation which is quite analogous to the Dirac equation. In contrast with the Klein-Gordon equation it is first order with respect to time derivation. It leads in a regular way to the standard Schrödinger equation in the non-relativistic limit. There are two components for the wave function in this representation for the scalar boson, in a way completely analogous to the four components for the spin $1/2$ fermion in the Dirac equation.

Local Hilbert spaces with large dimension are of key interest for quantum information with applications in quantum computing and memories, quantum simulations and metrology. Thanks to its weak coupling to external perturbations, the large ground-state nuclear spin manifold of fermionic alkaline-earth atoms is an exciting resource to explore for quantum information. Simultaneous single atom and state-resolved detection however remains an outstanding challenge limiting the development of novel quantum computing and simulation schemes beyond qubits. Here, we report on a new imaging technique enabling the simultaneous detection of up to four quantum states encoded in the nuclear spin manifold of a single fermionic strontium atom within 100 microseconds, with state-resolved detection fidelities ranging from 0.936 to 0.997. This technique is further used to track the highly coherent nuclear spin dynamics after a quench highlighting the potential of this system for quantum information. These results offer fascinating perspectives for quantum science with multi-electron atoms ranging from qudit-based quantum computing to quantum simulations of the SU(N) Fermi-Hubbard model.

We investigate the quasinormal modes (QNMs) of Planck stars within the framework of scale-dependent gravity (SDG). In our setup, the running parameter $α$ is fixed to a negative value by matching the effective Newtonian potential to the one-loop EFT result. As a consequence, the associated running Newton coupling does not realise the ultraviolet fixed point of asymptotically safe gravity, and the geometry should be interpreted as an SDG-inspired effective metric rather than a realisation of asymptotically safe gravity itself. We focus on the resulting renormalisation-group-improved Schwarzschild metric, which naturally yields a finite-size Planck-density core. Building on this background, we compute the QNM spectrum for scalar, electromagnetic, and gravitational perturbations using the Spectral Method (SM). This approach, known for its superior accuracy over high-order WKB schemes, enables the detection of fundamental modes, large families of overtones, and purely imaginary overdamped modes that are entirely missed in previous analysis. Our results reveal a robust Martini glass morphology of the oscillatory spectrum across perturbation sectors, nearly equally spaced overdamped modes with characteristic anomalous gaps, and the emergence, in the gravitational sector, of isolated overdamped modes separated from the main sequence by exceptionally large frequency intervals. These features, resolved here for the first time in the Planck-star context, underscore the importance of high-precision spectral techniques in probing subtle signatures of quantum-gravity-inspired black hole models.

Primordial black holes formed in the early universe are compelling candidates for dark matter. We investigate their production in Rastall gravity, a modification of general relativity that introduces a non-minimal coupling between matter and geometry through the non-conservation of the energy-momentum tensor. Analyz- ing cosmological perturbations during radiation domination, we demonstrate that the Rastall parameter fundamentally alters the collapse dynamics, modifying the growth of density fluctuations, the critical threshold for black hole formation, and the fluctuation amplitude at horizon crossing. Current cosmological constraints from Big Bang Nucleosynthesis, the cosmic microwave background, and large-scale structure restrict the Rastall parameter to small values, yet within this allowed range PBH production can be altered by orders of magnitude compared to general relativity. Our results establish primordial black holes as novel probes of modified gravity.

Highly excited Rydberg states of lanthanides are a promising, yet largely unexplored, playground for quantum studies. Here, we report on the first high-resolution spectroscopy of 162Dy obtained by two-color trap depletion spectroscopy in a magneto-optical trap. The absolute excitation frequency of over 700 states with effective principal quantum number n between 21 and 130 is measured with an accuracy of 20 MHz. Most states are assigned to the 8 different series converging to the first 4f10(5I8)6s(2S1/2) J = 17/2 ionization potential. This energy is measured at EIP = 47901.8265 +/- 0.0008 cm-1, improving the precision of the literature value by over an order of magnitude. A multichannel quantum defect theory approach is used to benchmark and refine the assignments and to characterize six observed perturbing states belonging to higher ionization limits. These results pave the way for using dysprosium in Rydberg-based quantum architectures, leveraging the unique properties arising from its complex electronic structure. They also represent a compelling benchmark for ab-initio calculations of open-shell atomic systems.

We study the Euclidean path integral of higher spin gravity on $S^4$. Based on a one-loop analysis, we are led to a gluing formula expressing the $S^4$ path integral in terms of an underlying $S^3$ path integral. We view the three-sphere as a boundary hypersurface splitting the four-sphere into two halves. For a higher spin spectrum containing even spins only, the resulting boundary theory living on the $S^3$ cut is the $\mathrm{Sp}(N)$ invariant sector of $N\in \mathbb{Z}^+$ anti-commuting, conformally coupled free scalars, with conformal higher spin sources mediating the gluing. This boundary $\mathrm{Sp}(N)$ theory was previously shown to compute the Hartle-Hawking wavefunction at $\mathcal{I}^+$ in the higher spin dS$_4$/CFT$_3$ correspondence. In contrast to the infinite spatial volume of $\mathcal{I}^+$, here the conformal fields populate a finite size $S^3$ hypersurface of $S^4$. For theories with both bosonic and fermionic higher spin fields, the gluing formula is instead built from an $\mathcal{N}=2$ superconformal boundary field theory coupled to $U(N)$ invariant superconformal sources. Under this assumption, the leading contribution to the four-sphere partition function is $2^N$, and we observe exact cancellations at one-loop.

We study the robustness of the Floquet quantum Ising model against integrability-breaking perturbations, focusing on the phase hosting both Majorana zero and $π$ modes. A recent work [Phys. Rev. B 110, 075117, (2024)] observed that the Floquet product mode, a composite edge mode constructed from both Majorana operators, is considerably more robust than the individual Majorana edge modes. We analyze these strong modes from the point of view of the eigenphase order present in finite chains with open boundary conditions. As a result of the Majorana modes, all Floquet eigenstates come in quadruplets in the integrable limit. We show that the robustness of the various modes as well as the behavior of the boundary spin correlation functions can be understood in terms of the spectral statistics of these quadruplets in the presence of integrability-breaking perturbations.

To operate quantum sensors at their quantum limit in real time, it is crucial to identify efficient data inference tools for rapid parameter estimation. In photodetection, the key challenge is the fast interpretation of click-patterns that exhibit non-classical statistics -- the very features responsible for the quantum enhancement of precision. We achieve this goal by comparing Bayesian likelihood-free methods with ones based on deep learning (DL). While the former are more conceptually intuitive, the latter, once trained, provide significantly faster estimates with comparable precision and yield similar predictions of the associated errors, challenging a common misconception that DL lacks such capabilities. We first verify both approaches for an analytically tractable, yet multiparameter, scenario of a two-level system emitting uncorrelated photons. Our main result, however, is the application to a driven nonlinear optomechanical device emitting non-classical light with complex multiclick correlations; in this case, our methods are essential for fast inference and, hence, unlock the possibility of distinguishing different photon statistics in real time. Our results pave the way for dynamical control of quantum sensors that leverage non-classical effects in photodetection.

We present a quantum optical scheme for imaging transversely displaced thermal sources of arbitrary intensities by employing multiphoton interference with a reference single-photon Fock state at a beamsplitter. Obtaining an analytical form for transverse momenta-resolved $L$-photon probabilities in either output, we show via Fisher information analysis that separation estimators built using interference sampling of multiphoton events exhibit significantly enhanced precision vis-à-vis existing imaging schemes over a wide range of separations and brightness. Even-photon-number coincidences exhibit constant precision in the sub-Rayleigh regime, demonstrating quantum superresolution of our scheme beyond the diffraction limit. For sources emitting on average $N_s\sim1$ photon per frame (such as in IR emission of thermal sources), precision bounds for our scheme scale linearly in $N_s$, exemplifying an enhanced precision of estimators in relation to weak sources $N_s\ll1$, and matching the ultimate quantum scaling. Finally, transverse momenta resolution in the Fourier plane produces finite imaging precisions for intermediate and large source separations using coarse pixel sizes of order $δy\sim100\,μ\mathrm{m}$ for exemplary image spot sizes $σ_x \sim 0.1\, μ\mathrm{m}$, in contrast with existing schemes of diffraction-limited direct imaging and superresolved inversion interferometric imaging that are severely degraded by coarse pixel sizes and have limited use. Combining the relatively straightforward sensing operation of Hong-Ou-Mandel interferometers with multiphoton coincidence detection of arbitrarily bright thermal sources and inner variable resolution of transverse photonic momenta, our scheme offers a robust alternative to non-invasive single-particle tracking and imaging of bright sources in nanoscopic chemical and biological systems.

Determining the ground state of a many-body Hamiltonian is a central problem across physics, chemistry, and combinatorial optimization, yet it is often classically intractable due to the exponential growth of Hilbert space with system size. Even on fault-tolerant quantum computers, quantum algorithms with convergence guarantees -- such as quantum phase estimation and quantum subspace methods -- require an initial state with sufficiently large overlap with the true ground state to be effective. Variational quantum eigensolvers (VQEs) are natural candidates for preparing such states; however, standard VQEs typically exhibit poor generalization, requiring retraining for each Hamiltonian instance, and often suffer from barren plateaus, where gradients can vanish exponentially with circuit depth and system size. To address these limitations, we propose an end-to-end representation learning framework that combines a graph autoencoder with a classical neural network to generate VQE parameters that generalize across Hamiltonian instances. By encoding interaction topology and coupling structure, the proposed model produces high-overlap initial states without instance-specific optimization. Through extensive numerical experiments on families of one- and two-local Hamiltonians, we demonstrate improved generalization and trainability, manifested as reduced test error and a significantly milder decay of gradient variance. We further show that our method substantially accelerates convergence in quantum subspace-based eigensolvers, highlighting its practical impact for downstream quantum algorithms.

Spin-boson models are the canonical benchmark for quantum dissipation. We show the symmetry structure of general spin-boson Hamiltonians and obtain their spectra explicitly by exploiting the symmetry. As an illustration of the general case, we numerically demonstrate the exact solution for the two-mode case.

A procedure to find static axially symmetric solutions to the Einstein field equations is presented. We obtained two general solutions and five particular solutions, which depend on the existence conditions for circular and $z$ direction motion. Our endeavour consists making a thoroughrowly analysis of all the possible geodesics solutions stemming from this spacetime.

We present a quantization of unimodular gravity in the connection representation for a homogeneous, isotropic, and spatially flat cosmological model without matter. In this model, the wave function is governed by a Schrödinger-type equation derived from a reduced phase space approach. Our analysis suggests that, within this minimal setting, the regularity of the operators and the self-adjointness of the Hamiltonian operator are incompatible with a negative cosmological constant. For a positive cosmological constant, the wave functions vanish at zero spatial volume. This behavior emerges as a consequence of enforcing the unimodular condition at the quantum level. Semiclassical fluctuations of the geometry are evaluated and discussed in relation to the cosmological constant problem.

We introduce a method for the measurement of the lower bound on the initial polarization of spinful nuclei in a diamond by following the coherence evolution of an NV center spin qubit after a simple scheme is operated on the qubit to facilitate the transfer of information from the environment into the qubit state. Current polarization measurement techniques are challenging to implement due to the need for direct access to the environment. In our method, information is obtained by measuring the difference of the evolution of the qubit coherence resulting from preparation phase when the environment evolution is conditional on the qubit pointer state. We find that the method does not depend strongly on the applied magnetic field, but rather on the number of spinfull nuclei that lead to decoherence, and gives a reasonable estimate if the environment is polarized. The key advantage of this approach is its simplicity and minimal experimental requirements, allowing the inference of initial nuclear polarizations without direct access to the environment. We demonstrate the efficacy of this method using a simulated environment of up to fifteen randomly placed nuclear spins.

Quantum reservoir computing (QRC) exploits fixed quantum dynamics and a trainable linear readout to process temporal data, yet reversing the transformation -- reconstructing the input from the reservoir output -- has been considered intractable owing to the recursive nonlinearity of sequential quantum state evolution. Here we propose a four-equation encode-decode protocol with cross-key pairing and constructively show that quantum reservoir and key combinations satisfying all four equations exist. Using a full XYZ Hamiltonian reservoir with 10 data qubits, we expand the feature dimension to 76 without increasing qubit count and achieve machine-precision reconstruction (mean-squared error $\mathrm{MSE} \sim 10^{-17}$) for data lengths up to 30 under ideal conditions; the rank condition $\mathrm{dim}(V) \geq N_c$ is identified as a necessary criterion. A comprehensive noise analysis across seven conditions and four baseline methods reveals a clear hierarchy: shot noise dominates, depolarizing noise adds a moderate factor, and asymmetric resource allocation -- 10 shots for encoding, $10^5$ for decoding -- yields approximately two orders of magnitude MSE improvement by exploiting the asymmetric noise roles of the encryption and decryption feature matrices. Under realistic noise the MSE degrades to $10^{-3}$-$10^{-1}$, indicating that error mitigation is needed before practical deployment, but our results establish the feasibility of bidirectional reversible information transformation within QRC.

Squeezed states, crucial for quantum metrology and emerging quantum technologies, have been demonstrated in various platforms, but quantum squeezing of magnons in macroscopic spin systems remains elusive. Here we report the experimental observation of quantum-level magnon squeezing in a millimeter-scale yttrium iron garnet (YIG) sphere. By engineering a strong dispersive magnon-superconducting qubit coupling via a microwave cavity, we implement a significant self-Kerr nonlinearity to generate squeezed magnon states with their mean magnon number less than one. Harnessing a magnon-assisted Raman process, we perform Wigner tomography, revealing quadrature variances of $\sim\!0.8$ ($\sim\!1.0$~dB squeezing) relative to the vacuum. These results lay the groundwork for quantum nonlinear magnonics and promise potential applications in quantum metrology.

We study whether latent motivation signals in short Spanish admission responses predict engagement and performance in an early quantum computing pathway run by QuantumHub Peru. We analyze N=241 applicants' open responses and link them to outcomes from two selective modules: Module 1 (secondary; mathematics and computing foundations; n=23) and Module 2 (secondary + early undergraduate; quantum fundamentals; n=36, including M1 continuers). To ensure baseline comparability, the M2 university entrance exam matched the difficulty of the M1 final. Final grades followed the program's official cohort-specific weightings (attendance/assignments/exam), which we retain to preserve ecological validity. Methodologically, we model text with Latent Dirichlet Allocation (LDA, k=8) and, for robustness, with sentence embeddings from a small multilingual language model, EmbeddingGemma-300M, projected via UMAP and clustered with HDBSCAN. This combination leverages the transparency of bag-of-words topics and the semantic richness of small language model embeddings. Descriptively, curiosity/learning topics show higher grades and attendance than technology/career-oriented topics; inferential tests are underpowered (e.g., linear R2 ~ 0.03; logistic pseudo-R2 ~ 0.04) so effect-size estimates should be viewed as preliminary rather than confirmatory. Embedding-based clustering yields seven clusters with 11.2% noise and modest agreement with LDA (ARI=0.068; NMI=0.163). Results suggest that brief motivation responses encode promising signals that could support early mentoring in rigorous STEM pipelines, while highlighting the need for larger, pre-registered studies.

Understanding the properties of strongly interacting matter at extreme densities is a central problem in fundamental physics, but neutron star mergers provide a natural laboratory for probing this regime. However, the complexity of the merger process complicates the interpretation of the associated gravitational-wave and electromagnetic signals. This picture becomes even more complex in the potential scenario in which dark matter accumulates around and in neutron stars, altering their structure and the associated observables. In this work, we study synthetic gravitational-wave observations of binary neutron star mergers with next-generation detectors, investigating their potential to extract both nuclear physics and dark-matter constraints. We also examine how the potential presence of fermionic, non-interacting dark matter inside neutron stars affects the inference of nuclear empirical parameters. We find that combining observations can tighten constraints on nuclear empirical parameters. However, the inferred values remain sensitive to systematic modeling biases and intrinsic degeneracies among the parameters. Conversely, our analysis reveals that even in the presence of dark matter, it will be unlikely to find decisive evidence for dark matter when analyzing gravitational-wave signals. Consequently, systematic biases in nuclear empirical parameter inference potentially resulting from the presence of dark matter are expected to be negligible even for observations with next-generation gravitational-wave detectors.

The stability of motional-mode frequency is essential for realizing high-fidelity quantum gates in trapped-ion quantum computing. While broadband Gaussian noise has been extensively studied and mitigated using pulse shaping techniques, the impact of coherent periodic noise has remained largely unexplored. Here we report a systematic investigation of 60-Hz power-line noise and its effect on the secular frequencies of a single ${}^{171}\mathrm{Yb}^{+}$ ion. Using spin-echo Ramsey spectroscopy, we characterize the amplitude and phase of the resulting secular-frequency modulation and validate this characterization via passive phase correction of the Ramsey sequence. Building on this, we implement active cancellation by injecting a compensation tone into the set-point of a PI controller that stabilizes the trap RF drive amplitude. A phasor-fitting procedure optimizes the amplitude and phase of the compensation signal, enabling near-complete suppression of the 60-Hz component. With active cancellation engaged, the coherence time of a radial motional mode is extended from approximately 10 ms to 35 ms, consistent with the limit set by motional heating. Our results provide both a clear characterization of periodic motional-mode noise and a practical framework for its suppression in trapped-ion quantum computing platforms.

Recent cosmological tensions, notably the Hubble and $S_{8}$ tensions, necessitate extensions of the conventional $Λ$CDM framework, wherein additional dynamical fields alter the effective spacetime encountered by matter and radiation. In K-essence cosmology, the scalar field induces an emergent FLRW geometry that is disformally linked to the gravitational metric, resulting in a \emph{tilted causal structure} where the light cone propagation differs from that of gravity. This study develops a covariant Boltzmann formalism inside a homogeneous K-essence framework and derives the modified mass-shell condition, geodesic equations, and collision integrals for both massless and massive particles. We demonstrate that the photon distribution retains its thermal properties in the emergent frame, while it seems geometrically rescaled in the gravitational frame. The Thomson and Compton processes maintain their microscopic structure while obtaining effective masses and interaction rates governed by the scalar field. During the tightly coupled epoch, the photon-baryon fluid experiences acoustic oscillations characterized by a modified sound horizon. For the kinetic K-essence DBI-type Lagrangian, the interaction rate scales as $n_{e}σ_{T}^{\rm eff}a\propto a^{-8}$, indicating a strong coupling in the early universe. Additionally, the diffusion damping scale scales as $k_{D}^{-2}\propto a^{29/2}$, indicating that small-scale anisotropies become increasingly sensitive to the evolving geometry. The results provide a coherent kinetic description of particle transport in a tilted spacetime and demonstrate that CMB propagation effects may serve as an observational probe of K-essence and emergent gravity frameworks.

Within the framework of quantum theory, we review the Chrestenson operator, the Weyl operator basis, and the Kronecker-Pauli operator basis in $d$-dimensional Hilbert spaces using Dirac notation, where $d$ is a prime integer strictly greater than 2. We establish a new algebraic relation connecting these operators and present the cases $d=3$ and $d=5$ as illustrative examples.

Calderbank-Shor-Steane (CSS) codes are a versatile quantum error-correcting family built out of commuting $X$- and $Z$-type checks. We introduce CSS-like codes on $G$-valued qudits for any finite group $G$ that reduce to qubit CSS codes for $G = \mathbb{Z}_2$ yet generalize the Kitaev quantum double model for general groups. The $X$-checks of our group-CSS codes correspond to left and/or right multiplication by group elements, while $Z$-checks project onto solutions to group word equations. We describe quantum-double models on oriented two-dimensional CW complexes (which need not cellulate a manifold) and prove that, when $G$ is non-Abelian and simple, every $G$-covariant group-CSS code with suitably upper-bounded $Z$-check weight and lower-bounded $Z$-distance reduces to a CW quantum double. We describe the codespace and logical operators of CW quantum doubles via the same intuition used to obtain logical structure of surface codes. We obtain distance bounds for codes on non-Abelian simple groups from the graph underlying the CW complex, and construct intrinsically non-Abelian code families with asymptotically optimal rate and distances. Adding "ghost vertices" to the CW complex generalizes quantum double models with defects and rough boundary conditions whose logical structure can be understood without reference to non-Abelian anyons or defects. Several non-invertible symmetry-protected topological states, both with ordinary and higher-form symmetries, are the unique codewords of simply-connected CW quantum doubles with a single ghost vertex.

The deterministic preparation of quantum many-body ground states is essential for advanced quantum simulation, yet optimal algorithms often require prohibitive hardware resources. Here, we propose a highly efficient, non-variational protocol for ground state preparation using a Power-Cosine quantum signal processing (QSP) filter. By eschewing complex block-encoding techniques, our method directly utilizes coherent time-evolution operators controlled by a single ancillary qubit. The integration of mid-circuit measurement and reset (MCMR) drastically minimizes spatial overhead, translating iterative non-unitary filtering into deep temporal coherence. We analytically demonstrate that this approach achieves exponential suppression of excited states with a circuit depth scaling of $\mathcal{O}(Δ^{-2}\log(1/ε))$, prioritizing implementational simplicity over optimal asymptotic complexity. Numerical simulations on the 1D Heisenberg XYZ model validate the theoretical soundness and shot-noise resilience of our method. Furthermore, an advantage analysis reveals that our protocol exponentially outperforms standard Trotterized Adiabatic State Preparation (TASP) at equivalent circuit depths. This single-ancilla framework provides a highly practical and deterministic pathway for many-body ground state preparation on Early Fault-Tolerant (EFT) quantum architectures.

The Galactic Centre black hole provides a naive environment for understanding unknown matter distribution and new gravitational physics. For this stellar orbits in the nuclear star cluster are reliable probes. We investigate different dark matter mass profiles through pericentre shift of stellar orbits near the black hole. We also study capability of existing and upcoming astrometric facilities to detect dark matter induced precession and to distinguish between several dark matter profiles. Parameters of different dark matter density profiles are estimated by using the most recent upper bound on dark mass near the black hole. These profiles are then used for calculating the gravitational potential and hence the relativistic pericentre shift of both low and high eccentricity orbits of 13 S-stars. We use the recently measured deviation parameter $f_{sp}$ for investigating competition between dark matter and gravitational physics within S2's orbit. The astrometric shift of the pericentres has been calculated and compared with existing and upcoming astrometric capabilities of large and extremely large telescopes. The orbit of S2 is found to be insensitive to dark matter induced precession. Low eccentricity and wider orbits are prominent probes for measuring dark matter induced precession which is accessible to present and upcoming astrometric facilities such as Keck, GRAVITY and TMT. The existing and upcoming facilities can distinguish between different dark matter profiles for some stars and hence they posses the capability to distinguish between possible formation histories of the central region of our Galaxy.

Timing observation of pulsars orbiting around a supermassive black hole (SMBH) can measure the spacetime around the SMBH to a high precision and thus be a novel probe of the gravity theory. Future high-frequency surveys of the Galactic Centre (GC) region to be performed by the next-generation radio telescopes, such as the SKA, may discover pulsars that orbit around Sagittarius A* (Sgr A*), the SMBH dwelling in our GC. In this paper, we present a realistic pulsar-SMBH timing model based on the post-Newtonian equations of motion of the pulsar. Considering the expected timing precision in the future, we take into account several next-to-leading order light propagation time delays in the timing model. For the first time, we include the effects of proper motion of Sgr A*, which were expected to break the spin measurement degeneracy. We forecast the measurement precision of various parameters of Sgr A*, and discuss the data analysis procedure in the presence of red noise, which can be strong if the pulsar is a normal pulsar. The realistic timing model constructed in this study will serve as a useful tool in future searching and timing of pulsar-SMBH systems in the GC.

Extreme mass-ratio inspirals (EMRIs) are among the most powerful probes of strong-field gravity and of the environments surrounding supermassive compact objects. Motivated by the expected presence of dark matter near galactic centers, we investigate the emergence and gravitational-wave imprints of chaotic dynamics in EMRIs evolving in non-vacuum spacetimes. Within a unified dynamical framework, we analyze test-particle motion in a broad class of dark-matter-embedded geometries, including singular black holes, regular black holes, naked singularities, and Einstein-cluster configurations. We show that environmental perturbations generically break integrability in the strong-field regime, giving rise to chaotic motion whose onset, duration, and termination depend sensitively on horizon structure, core regularization, and matter distribution. Using the numerical Kludge approach, we demonstrate that chaotic trajectories produce systematic qualitative modifications of the emitted gravitational radiation, such as irregular amplitude modulation and loss of phase coherence, in contrast to the smooth, quasi-periodic waveforms generated by regular motion. Our results establish the robustness of chaos in environmentally perturbed EMRIs and provide a clear conceptual link between nonlinear orbital dynamics, spacetime structure, and observable gravitational-wave signatures.

An all-orders worldline effective action for Kerr black hole is achieved in twistor particle theory.

We analytically derive the bit-string probability distributions of subsystems of random pure states and depolarized random states using the Dirichlet distribution. We identify the exact Beta distribution as the universal statistical law of random quantum states, providing a unified finite-size description of full-system, subsystem, and conditional statistics. In the presence of depolarizing noise, these distributions are scaled and shifted by the noise strength, producing a noise-induced gap in their support. Remarkably, we prove that random states exhibit exact conditional self-similarity: the distribution of subsystem bit-string probabilities conditioned on specific outcomes of the complementary subsystem is identical to that of the full system. This hidden scale invariance enables the exact restoration of the full-system statistics from the marginalized Beta distribution via post-selection, and persists under depolarizing noise. Our results uncover a fundamental symmetry of Hilbert space and provide a scalable, rigorous framework for validating random circuit sampling via subsystem or conditional cross-entropy benchmarking.

A fundamental challenge in low-frequency gravitational-wave detectors is the limited signal-to-noise ratio (SNR) of calibration lines, particularly in torsion-bar systems where the response is governed by rotational dynamics. In this work, we resolve this issue by optimizing the geometrical configuration of a torque-coupled gravity field calibrator (GCal), achieving an improvement in calibration-line SNR by more than an order of magnitude compared to conventional layouts. For the Cryogenic sub-Hz cROss torsion-bar detector with quantum NOn-demolition Speed-meter (CHRONOS), the calibration signal appears as a monochromatic line within the $0.1$--$10~\mathrm{Hz}$ band. At $1~\mathrm{Hz}$, the strain-equivalent calibration amplitude reaches $|h_{\rm GCal}| = 1.18 \times 10^{-14}$, corresponding to an SNR density of $|h_{\rm GCal}|/S_h = 4.25 \times 10^{3}$. This demonstrates for the first time that a high-SNR calibration line can be directly injected into the sub-Hz band of a torsion-bar detector. A first-order perturbative error propagation analysis yields a total fractional systematic uncertainty of $δh_{\rm GCal}/h_{\rm GCal} = 0.24\%$, dominated by geometric alignment uncertainties, while contributions from mass uncertainties and the gravitational constant remain subdominant. The corresponding absolute systematic uncertainty is $δh_{\rm GCal} \sim 10^{-17}$ at $1~\mathrm{Hz}$. These results establish torque-coupled gravitational calibration as a practical solution to the longstanding low-SNR problem in sub-Hz torsion-bar detectors and provide a robust pathway toward precision absolute calibration in the low-frequency regime.

Preparing high-fidelity Greenberger-Horne-Zeilinger (GHZ) states on noisy quantum hardware remains challenging due to cumulative gate errors and decoherence. We introduce Group-Majority-Voting (Group-MV), a dynamic-circuit protocol that partitions arbitrary coupling graphs, prepares local GHZ states in parallel, and fuses them via majority-voted mid-circuit measurements. The majority vote over redundant boundary links mitigates measurement errors that would otherwise propagate through classical feedforward. We evaluate Group-MV on simulated Heavy-hex and Grid topologies for 30 through 60 qubits under a realistic noise regime. Group-MV generalizes to arbitrary GHZ sizes on arbitrary coupling topologies, achieving 2.4x higher fidelity than the Line Dynamic method while tracking the unitary baseline within 3%.

Quantum state discrimination enables the accurate identification of quantum states, which are generally nonorthogonal. Among various strategies, minimum-error discrimination and unambiguous state discrimination exhibit contextuality-enhanced success probabilities that surpass classical bounds, offering significant advantages for quantum sensing and communication. However, in practice, both error and failure outcomes can occur, suggesting the need for a unified strategy that incorporates both aspects while exploring the potential for contextuality enhancement. In this work, we theoretically demonstrate contextuality enhancement in quantum state discrimination under a fixed failure probability. We show that this enhancement disappears within a certain intermediate range of failure probabilities--a phenomenon absent in conventional strategies, where both minimum-error and unambiguous discrimination consistently outperform the noncontextual bound for equal priors. Moreover, we analyze how the existence of this non-enhancement region depends on the confusability of the quantum states, which corresponds to their fidelity in a quantum model. We further extend the discussion to the noisy state discrimination, which even encompasses the maximal-confidence discrimination. In this extended discussion, we observe that the non-enhancement region tends to disappear with increasing noise strength.

The Gravitational Poissonian Spontaneous Localization (GPSL) model is a hybrid classical-quantum framework in which Newtonian gravity emerges from stochastic collapses of a smeared mass-density operator. Consistency of the hybrid dynamics entails momentum diffusion and, hence, spontaneous heating. Without smearing, which enters both the collapse (measurement) and gravitational-feedback components of the dynamics, the heating rate would be divergent. Previous work assumed identical smearings for both components. Here, we treat the general case of distinct spatial smearings $g_{r_C} (\mathbf{x})$ and $g_{r_G} (\mathbf{x})$, characterized, respectively, by length scales $r_C$ and $r_G$. We characterize the spontaneous heating rate for arbitrary $g_{r_C} (\mathbf{x})$ and $g_{r_G} (\mathbf{x})$, and then discuss which smearing profiles minimize the spontaneous heating rate in relevant physical situations. Remarkably, there are situations in which, while the measurement noise remains the same, allowing $g_{r_G} (\mathbf{x}) \neq g_{r_C} (\mathbf{x})$ may reduce the feedback-induced spontaneous heating by more than 60 orders of magnitude already for $r_G = 10 r_C$. Finally, we use our results to estimate the spontaneous heating rate of neutron stars and to set new lower bounds on the model's parameters by comparing the theoretical predictions with astronomical data on temperature, radius, and mass of neutron stars.

We investigate the global well-posedness and modified scattering for the one-dimensional Schrödinger equation with gauge-invariant polynomial nonlinearity. For small localized initial data of finite energy in a low-regularity class, we establish global existence of solution together with persistence of the localization of the associated profile. We further provide a rigorous derivation of the asymptotic expansion at arbitrary order of such solutions, taking into account long-range effects induced by the cubic component of the nonlinearity. Our analysis relies on the space-time resonance method.

In this paper, we solve the relativistic Euler equations with a linear barotropic equation of state on a large class of background spacetimes with Kasner big bang asymptotics. Building on previous work in the asymptotically non-tilted regime, which applies when the speed of sound of the fluid is large in comparison to the Kasner exponents, we now consider the asymptotically extremely-tilted regime when the speed of sound is small. We solve the Cauchy problem for the Euler equations towards the big bang singularity and prove, without any symmetry assumptions or smallness conditions on the Cauchy data, that the solutions exist globally in time provided the mean curvature of the initial hypersurface is sufficiently large. Finally, we prove that the solutions exhibit the asymptotics expected from standard heuristic arguments in this regime; in particular, fluid particles are driven towards the speed of light in the direction of the largest Kasner exponent as the big bang singularity is approached.

Higher-dimensional quantum systems are attracting interest for improving quantum protocol performance by increasing memory space. Characterizing quantum resources of such systems is fundamental but experimentally costly. We tackle the first non-trivial example: a qubit-ququart system, focusing on partial-transpose spectral classification. Entanglement distillation extracts maximally entangled states from noisy resources, but determining distillability typically requires full state tomography, experimentally prohibitive for high-dimensional systems. We explore a machine learning framework to classify distillable bipartite quantum states using fewer measurements than complete tomography. Our approach employs the PPT criterion, categorizing states by negative eigenvalues in the partial transpose. We use various ML algorithms, including Support Vector Machines, Random Forest, and Artificial Neural Networks, with features from fixed measurements and learnable observables. Results show learnable observables consistently outperform Collective Measurement Witnesses methods. While all models distinguish between non-distillable (PPT) and distillable (NPT) states, differentiating NPT subclasses remains challenging, underscoring the intricate Hilbert space geometry. This work provides an experimentally friendly tool for distillability verification in high-dimensional quantum systems without full state reconstruction

We analytically solve the entangling quantum dynamics of two interacting Stern-Gerlach Interferometers~(SGI). Each SGI exploits an operator-valued force applied by a qubit to create and recombine a non-Gaussian state of matter. The entangling phase between the two qubits generated by the leading-order gravitational interaction of the massive degrees of freedom is found to be mass-independent, both for unitary and open dynamics, irrespective of the temperature and squeezing of the initial states. Further, we show that the solution of the four interferometric paths reveals that the mere presence of the interaction does not allow for a perfect recombination of the centre of mass. This second-order effect, alongside higher-order interaction terms, can be used to bound the mass from above and below, thus restricting the experiment's regime to mesoscopic masses. By solving the open dynamics which includes diffusion and dephasing with initial squeezed thermal states, the bounds are tightened by the inclusion of realistic experimental noise. We discuss diamagnetic levitated masses with embedded NV-centres as a specific physical implementation.

Motivated by Johnson--Lindenstrauss dimension reduction, amplitude encoding, and the view of measurements as hash-like primitives, one might hope to compress an $n$-point approximate nearest neighbor (ANN) data structure into $O(\log n)$ qubits. We rule out this possibility in a broad quantum sketch model, the dataset $P$ is encoded as an $m$-qubit state $ρ_P$, and each query is answered by an arbitrary query-dependent measurement on a fresh copy of $ρ_P$. For every approximation factor $c\ge 1$ and constant success probability $p>1/2$, we exhibit $n$-point instances in Hamming space $\{0,1\}^d$ with $d=Θ(\log n)$ for which any such sketch requires $m=Ω(n)$ qubits, via a reduction to quantum random access codes and Nayak's lower bound. These memory lower bounds coexist with potential quantum query-time gains and in candidate-scanning abstractions of hashing-based ANN, amplitude amplification yields a quadratic reduction in candidate checks, which is essentially optimal by Grover/BBBV-type bounds.

Controlled unitary gates are a basic element in many quantum algorithms. Converting a general unitary $U$ with a known decomposition into its controlled version, controlled-$U$, can introduce a large overhead in terms of the depth of the circuit. We present a general method to take any unitary $U$ into controlled-$U$ using a fixed circuit with 4 CNOT gates and 2 Toffoli gates per qubit. For $n$-qubit unitaries and one control qubit, we require $2n+1$ qubits and a circuit that can generate an eigenstate of $U$, for which there are many cost-effective known algorithms. The method also works for any black block implementation of $U$, achieving a constant-depth realization independent of its decomposition. We illustrate its use in the Hadamard test and discuss applications to variational and quantum machine-learning algorithms.

While the Teukolsky equation plays a central role in traditional treatments of perturbations of algebraically special spacetimes, its relation to Friedrich's conformal Einstein field equations (CEFEs) remains largely unexplored. Here we develop a conformal formulation of black-hole perturbation theory based on the CEFEs and derive the conformal Teukolsky equation. Starting from a transparent review of Friedrich's regularisation strategy, this work establishes a direct connection between mainstream curvature-based linear perturbation theory and conformal formulations of general relativity. This perspective is timely given the growing relevance of hyperboloidal frameworks in black-hole perturbation theory, where conformal compactification is introduced at the level of an already linearised effective wave equation. Here instead, the conformal factor is a dynamical variable within the field equations. In the non-linear equations there is a coupling between conformal and curvature perturbations; however, when linearised around a Petrov-type D background, the conformal factor decouples from the equations governing the Newman-Penrose components $φ_0$ and $φ_4$ of the rescaled Weyl tensor. The resulting equation preserves the structural form of the classical Teukolsky equation while remaining regular at the conformal boundary. This provides a geometric interpretation of the hyperboloidal master variable and an entry point into the CEFE framework. We further derive the conformal Teukolsky equation for a conformal representation of Kerr spacetime where spatial infinity is realised as a blown-up cylinder. By bridging conformal and traditional approaches to black-hole perturbation theory, the framework highlights a geometrically regular representation of perturbative dynamics that may inform extensions beyond the linear regime.

In a previous paper [S. Mudli {\it et al.} Phys. Rev. A 110, 062618 (2024)], it was shown that a trapped ion can mediate interaction between two largely separated Rydberg atoms, and this mediated interaction can be leveraged to perform a universal two-qubit gate operation between neutral atom qubits in optical tweezers. In this paper, we demonstrate the universal two-qubit CNOT gate with high fidelity between an ionic and an atomic qubit relying on Rydberg excitation of the atom and the resulting phonon blockade in the motional states of the harmonically trapped ion. The phonon blockade arises due to strong ion-atom interaction when the atom is excited to a Rydberg state. These demonstrations suggest that an ion-atom hybrid system can serve as a resourceful platform or module for quantum computing and quantum networking as it can utilize the best features of charged as well as neutral atom qubits.

It is noteworthy that symmetric teleparallel gravity has attracted considerable attention in recent years. A survey of the literature indicates that this surge of interest became particularly prominent around 2017 and 2018. However, together with my students and collaborators, we published a series of systematic and pioneering papers on this subject between 2004 and 2013. In this letter, we will summarize our overlooked contributions to symmetric teleparallel gravity produced during that period. For the sake of completeness and coherence, we will also briefly review our work on this topic carried out after 2018. In the final paragraph, we will write our personal perspectives on the future of symmetric teleparallel geometry.

We investigate the propagation of gravitational waves in the presence of Lorentz- and diffeomorphism-violating operators within the linearized gravitational sector of the Standard Model Extension. Focusing on isotropic contributions, we analyze the combined effects of the nondispersive CPT-even dimension-four coefficient $\mathring{k}^{(4)}_{(I)}$ and the CPT-odd dimension-five coefficient $\mathring{k}^{(5)}_{(V)}$ on tensorial gravitational radiation. The modified dispersion relation induces both a rescaling of the propagation speed and helicity-dependent dispersive corrections, leading to birefringence and polarization mixing without introducing additional propagating degrees of freedom. We derive the retarded Green function associated with the modified wave operator and obtain explicit expressions for the gravitational waveform generated by matter sources. As a concrete application, we examine a binary black hole system and show how Lorentz violation alters the observed strain through shifted retarded times, amplitude rescaling, and higher derivative corrections to the quadrupole formula. The CPT-odd term produces characteristic attenuation and distortions in the waveform, which can be interpreted as energy exchange between the gravitational wave and the Lorentz-violating background rather than a violation of energy conservation. Using published LIGO-Virgo-KAGRA propagation tests and polarization consistency arguments, we translate current observational constraints into bounds on $\mathring{k}^{(4)}_{(I)}$ and $\mathring{k}^{(5)}_{(V)}$.

Our goal is to provide precise effective operators for monolayer graphene at Fermi energy. We consider the microscopic potential created by a lattice, and add a macroscopic potential with the same periodicity but varying at a scale $\varepsilon^{-1} \in \mathbb{N}$, creating a superlattice. Our approach consists in coupling the variational approximation, perturbation theory together with a multiscale method. At the effective level the usual massless Dirac operator is replaced by other operators, and we provide simulations in the case of graphene.

An $SO(3,3)$ BF-type gauge theory is formulated on a six-dimensional spacetime of split signature $(3,3)$, interpreted as the pre-electroweak-symmetry-breaking phase. A MacDowell--Mansouri-type symmetry breaking to $SU(2)\times SU(2)$ is implemented, and the corresponding stabilizer and coset structures are computed. The curvature decomposes into chiral sectors, and effective tetrads are introduced using components of the higher-dimensional connection. The resulting left and right sectors are formulated as constrained BF/Plebanski-like theories with appropriate simplicity and reality conditions. The six-dimensional theory yields two overlapping four-dimensional Lorentzian sectors of opposite signature, related via gluing constraints across their intersection. In the first sector, the selfdual two-forms ($Σ^{(+)}$) satisfy simplicity constraints that select the non-degenerate branch and reproduce Einstein gravity. Subsequently, the $SU(2)_R\times U(1)_{Y{\rm dem}}\to U(1)_{\rm dem}$ breaking pattern is outlined which admits an ultra-soft regime consistent with current phenomenological bounds under sufficiently suppressed couplings. In the second sector, the antiself dual two-forms ($Σ^{(-)}$) satisfy analogous simplicity constraints, realizing weak gauge dynamics as gravity on the opposite-signature sector. Subsequently, the $SU(2)_L\otimes U(1)_Y$ electroweak symmetry is realized within the Yang--Mills branch of the BF theory which incorporates the standard Higgs mechanism $SU(2)_L\otimes U(1)_Y \to U(1)_{\mathrm{EM}}$, recovering the conventional electroweak $W^\pm$, $Z$, and photon spectrum.

We investigate the effect of higher-order curvature terms, specifically Gauss-Bonnet terms, on spacetime singularities in five dimensions. For FLRW cosmologies, we demonstrate that Gauss-Bonnet terms can replace the Big Bang/Crunch with a "sudden" singularity, characterized by a finite scale factor and Hubble rate but diverging higher-order derivatives. Investigating various branches of solutions shows the possibility of explicit extension of non-spacelike geodesics beyond the singular point. Furthermore, we employ the Gauss-Bonnet junction conditions to verify the consistency of the extension with the field equations. The whole solution describes a contracting phase prior to the expansion phase with a well-defined surface stress-energy tensor. Regarding the Boulware-Deser black hole, we find that Gauss-Bonnet terms soften the central singularity for radial geodesics--rendering them "weak" according to the Tipler and Krolak criteria--whereas non-radial geodesics remain strongly singular. Junction condition analysis of this solution shows that although higher-curvature corrections alter the nature of the singularity, geodesics are still inextendible as a result of divergent extrinsic curvature. Our results are consistent with the Penrose-Hawking singularity theorems since in Gauss-Bonnet black holes, geodesics suffer from focusing (expansion parameter diverges), while in cosmology, there is no focusing since the expansion parameter remains finite at the singularity.

One of the most enduring and unresolved challenges in modern theoretical and observational cosmology is the fine-tuning and coincidence problems associated with the cosmological constant. Rather than attempting to reconcile these issues within the standard LCDM framework where they remain effectively frozen, we adopt a fundamentally different viewpoint based on alternative theories of gravity. We argue that the root of these problems lies in a deep misinterpretation of the cosmological constant, particularly its identification with the quantum ground-state energy of spacetime. In this work, we propose a novel physical interpretation of the cosmological constant and introduce a new mechanism, termed Breaking Energy-Momentum Symmetry. This framework provides a natural and unified route toward alleviating, and potentially resolving, both the fine-tuning and coincidence problems, offering a compelling alternative to the conventional cosmological paradigm.

In this paper, we investigate a quintessence field with an exponential potential motivated by the suggestion of time-varying dark energy from the DESI galaxy survey. Assuming a kination epoch in the early Universe, we analytically derive constraints on initial conditions that are consistent with Big Bang Nucleosynthesis and the current dark energy density. Compared to the severe 120-digit fine-tuning required for dark energy to be a cosmological constant, our result suggests that the degree of fine-tuning is naturally relaxed by dozens of orders of magnitude. Furthermore, we discuss the method for testing this model through future observations of the gravitational wave background.

Observations of supermassive black holes by the Event Horizon Telescope reveal significant inhomogeneities, most likely related to density and magnetic field perturbations. To model these features, we conduct high-resolution 2D general-relativistic magnetohydrodynamics (GRMHD) simulations of a Fishbone-Moncrief torus around a Kerr black hole using the Black Hole Accretion Code $\texttt{BHAC}$. We compare unperturbed accretion with a case featuring plasma density bubbles with pressure balanced magnetic islands of different amplitudes. Power spectrum analysis of accretion time series, performed via the Blackman-Tukey method, shows that the perturbed case exhibits (1) steeper spectral indices compared to the unperturbed case, deviating from the characteristic $1/ω$ noise spectrum, and (2) increased correlation times, providing evidence for absorption of macro-structures at the event horizon. Spatial auto-correlation analysis of near-horizon turbulence confirms larger energy-containing coherent structures in the perturbed case altering the accretion rate. These results provide new insights for interpreting observations of supermassive black hole environments, where near-horizon turbulence may play a key role in the accretion process.

Achieving high fidelity of quantum teleportation (QT) in a noisy environment is an essential requirement for its real-world applications. To this end, we devise a distinctive protocol for ensuring teleportation fidelity {\it close to unity}, hinging essentially on the timing of Alice's Bell-basis measurement (BM) dependent on the choice of Bob's local noise parameters, but is independent of Alice's local noise. Our scheme is enabled by Alice communicating to Bob only two of the BM outcomes corresponding to the states that are decoherence-free under common dephasing at Alice's wing. On the other hand, Bob is asked to discard the states of his qubit for the other two BM outcomes in order to maximize fidelity of the teleported state. This ensures the teleportation fidelity's independence of noise parameters in Alice's wing. We formulate the protocol in terms of a generic two-level quantum system, subjected to non-Markovian dephasing noise, applicable for any pure maximally/non-maximally entangled state as well as a Werner-type mixed state as resource. Notably, we show that high fidelity is achievable even using resource states with small values of the entanglement measure. Remarkably, even within the local regime of Werner states, where Bell-CHSH inequalities are not violated, the teleportation fidelity remains significantly high. Finally, we discuss the empirical feasibility of our scheme using photonic qubits.

Directional codes, recently introduced by Gehér--Byfield--Ruban \cite{Geher2025Directional}, constitute a hardware-motivated family of quantum low-density parity-check (qLDPC) codes. These codes are defined by stabilizers measured by ancilla qubits executing a fixed \emph{direction word} (route) on square- or hex-grid connectivity. In this work, we develop a comprehensive \emph{word-first} analysis framework for route-generated, translation-invariant CSS codes on rectangular tori. Under this framework, a direction word $W$ deterministically induces a finite support pattern $P(W)$, from which we analytically derive: (i)~a closed-form route-to-support map; (ii)~the odd-multiplicity difference lattice $L(W)$ that classifies commutation-compatible $X/Z$ layouts; and (iii)~conservative finite-torus admissibility criteria. Furthermore, we provide: (iv)~a rigorous word equivalence and canonicalization theory (incorporating dihedral lattice symmetries, reversal/inversion, and cyclic shifts) to enable symmetry-quotiented searches; (v)~an ``inverse problem'' criterion to determine when a translation-invariant support pattern is realizable by a single route, including reconstruction and non-realizability certificates; and (vi)~a quasi-cyclic (group-algebra) reduction for row-periodic layouts that explains the sensitivity of code dimension $k$ to boundary conditions. As a case study, we analyze the word $W=\texttt{NE$^2$NE$^2$N}$ end-to-end. We provide explicit stabilizer dependencies, commuting-operator motifs, and an exact criterion for dimension collapse on thin rectangles: for $(L_x, L_y) = (2d, d)$ with row alternation, we find $k=4$ if $6 \mid d$, and $k=0$ otherwise.

Quantum error correction (QEC) and dynamical decoupling (DD) are tools for protecting quantum information. A natural goal is to combine them to outperform either approach alone. Such a benefit is not automatic: physical DD can conflict with an encoded subspace, and QEC performance is governed by the errors that survive decoding, not necessarily those DD suppresses. We analyze a hybrid memory cycle where DD is implemented logically (LDD) using normalizer elements of an $[[n,k,d]]$ stabilizer code, followed by a round of syndrome measurement and recovery (or, in the detection setting, postselection on a trivial syndrome). In an effective Pauli model with physical error probability $p$, LDD suppression factor $p_{DD}$, and recovery imperfection rate $p_{QEC}$ (or $p_{QED}$), we derive closed-form entanglement-fidelity expressions for QEC-only, LDD-only, physical DD, and the hybrid LDD+QEC protocol. The formulas are expressed via a small set of code-dependent weight enumerator polynomials, making the role of the decoder and the LDD group explicit. For ideal recovery LDD+QEC outperforms QEC-only iff the conditional fraction of uncorrectable Pauli errors is larger in the LDD-suppressed sector than in the unsuppressed sector. In the low-noise regime, a sufficient design rule guaranteeing hybrid advantage is that LDD suppresses at least one minimum-weight uncorrectable Pauli error for the chosen recovery map. We show how stabilizer-equivalent choices of LDD generators can be used to enforce this condition. We supplement our analysis with numerical results for the $[[7,1,3]]$ Steane code and a $[[13,1,3]]$ code, mapping regions of hybrid-protocol advantage in parameter space beyond the small-$p$ regime. Our work illustrates the need for co-design of the code, decoder, and logical decoupling group, and clarifies the conditions under which the hybrid LDD+QEC protocol is advantageous.

Achieving superradiant lasing with an ultranarrow linewidth is crucial for enhancing atomic clock stability in quantum precision measurement. By employing the exceptional point (EP) property of the system, we demonstrate theoretically superradiant lasing with linewidths in the $μ$Hz range, sustained at the high-power level. This is achieved by incoherently pumping optical lattice clock transitions with ultracold alkaline-earth strontium-87 atoms in the EP of a $\mathcal{PT}$-symmetric system. Physically, the atomic coherence reaches a maximum in the EP, significantly amplifying the superradiance effect and resulting in superradiant lasing with an ultranarrow linewidth. This linewidth is even three orders of magnitude smaller than that of superradiant lasing in the systems without EP. Our work extends the realm of superradiant lasing by introducing the EP property, and offers promising applications for developing atomic clocks with exceptional stability and accuracy.

The co-propagation of quantum and classical signals through shared optical fibers is crucial for scalable quantum networks. However, this coexistence is fundamentally limited by spontaneous Raman scattering (SpRS) from the bright classical light, which generates overwhelming noise that disrupts the single-photon-level quantum signals. Here, we overcome this long-standing challenge by leveraging the inherently ultralow nonlinearity of hollow-core fiber (HCF) to suppress SpRS noise. By operating both the quantum time synchronization (QTS) and classical optical frequency transfer (OFT) signals within the telecom C-band, separated by only ~10 nm, we successfully demonstrate their simultaneous transmission over a 122-km HCF link. With a classical OFT power of 1 mW, the QTS performance shows negligible degradation, maintaining sub-picosecond time stability at 2000 s, while the OFT achieves a fractional frequency instability of 10^-20. Near-sub-picosecond QTS stability is preserved even when the classical power is increased to 3 mW. Furthermore, simulations based on our experimental data indicate that with next-generation low-loss HCF, the platform can tolerate classical powers beyond 10 mW and extend the QTS range to over 500 km. By realizing a unified quantum-classical time-frequency distribution framework, this work establishes HCF as a highly capable and practical platform for future scalable quantum networks.

In this work, we perform a relativistic test of the spacetime geometry of Sagittarius A* (Sgr A*) using the orbit of the S2 star. We consider a broad class of compact object models, including Schwarzschild, Reissner-Nordström, Bardeen, Hayward, and Simpson-Visser black holes, as well as the Janis-Newman-Winicour naked singularity spacetime. For each geometry, we integrate the timelike geodesic equations and consistently project the resulting trajectories onto astrometric and spectroscopic observables, incorporating Rømer time delay and relativistic redshift effects. The theoretical predictions are tested with current Very Large Telescope (VLT) observations of the S2 star, while simultaneously imposing constraints from the Event Horizon Telescope shadow size. We find that several spacetimes that are degenerate at the level of shadow imaging, most notably Schwarzschild, Reissner-Nordström, and Bardeen regular black hole geometries, remain statistically indistinguishable when tested against present S2 data. We further carry out a statistical model comparison based on the Akaike and Bayesian information criteria (AIC and BIC) to evaluate the relative performance of the alternative spacetime models. Our analysis also constrains the generalized charge like parameter $q/M$ in non-Schwarzschild spacetimes based on current S2 star observations, and identifies specific black hole and horizonless geometries that can be further tested with forthcoming high precision astrometric observations from the VLT and Keck telescopes.

We investigate inertial frame dragging and relativistic precession in the Kerr--Newman spacetime and show how gyroscopic observables can operationally discriminate between Kerr--Newman black holes and Kerr--Newman naked singularities. We study a test gyroscope attached to a stationary observer and derive closed-form expressions for the Lense--Thirring, geodetic, and general spin-precession frequencies. A sharp qualitative distinction emerges: for Kerr--Newman black holes, the spin-precession frequency generically diverges as the horizon is approached (remaining finite only for the ZAMO family), whereas for Kerr--Newman naked singularities, the precession remains finite throughout the spacetime, with divergences confined to the ring singularity on the equatorial plane. Working with physically admissible stationary observers (including ZAMOs), we first construct the timelike geodesic motion and the fundamental orbital frequencies for equatorial circular orbits. Using these, we analyse the radial and vertical epicyclic frequencies, the ISCO shift induced by the charge parameter $Q$, and the associated periastron and nodal precession frequencies relevant to quasi-periodic oscillations (QPOs). We demonstrate that nonzero $Q$ produces systematic, and in rapidly rotating regimes nontrivial, modifications to the frequency hierarchy: $ν_{\rm nod}$ can develop a finite maximum at $r=r_p=\mathcal{O}(M)$, its peak amplitude decreases with increasing $Q$, and sign reversals may occur for sufficiently large charge and high spin, signalling a reversal of nodal-precession orientation. These results establish spin-precession behaviour as a robust strong-field probe of horizons versus exposed singularities, with potential implications for testing cosmic censorship using future high-precision precession/QPO measurements.

We construct charged AdS black hole solutions in four dimensional Einstein Gauss Bonnet gravity coupled to Euler Heisenberg nonlinear electrodynamics and investigate their physical properties. The modified field equations admit black hole solutions whose horizon structure is significantly affected by higher-curvature and nonlinear electromagnetic corrections, allowing for multiple horizons depending on the model parameters. In the extended phase space, where the cosmological constant is interpreted as thermodynamic pressure, we analyze the thermodynamic behavior and show that both the Gauss Bonnet coupling and the Euler Heisenberg parameter induce notable modifications in the equation of state, critical behavior, and thermal stability. Interpreting the black hole mass as enthalpy, we study the Joule-Thomson expansion and determine the inversion temperature and pressure, demonstrating that higher curvature and nonlinear electrodynamic effects substantially influence the cooling and heating regions. Finally, we examine time-like geodesics and show that Gauss Bonnet corrections significantly modify the effective potential, orbital stability, and particle motion in the strong-field regime

The nonstabilizerness of quantum states is a necessary resource for universal quantum computation, yet its characterization is notoriously demanding. Quantifying nonstabilizerness typically requires an exponential number of measurements and a doubly exponential classical post-processing cost to evaluate its standard monotones. In this work, we show that nonstabilizerness is, to a large extent, in the eyes of the beholder: it can be witnessed and quantified using any set of $m$ $n$-qubit Pauli measurements, provided the set contains anti-commuting pairs. We introduce a general framework that projects the stabilizer polytope onto the subspace defined by these observables and provide an algorithm that estimates magic from Pauli expectation values with runtime exponential in the number of measurements $m$ and polynomial in the number of qubits $n$. By relating the problem to a stabilizer-restricted variant of the quantum marginal problem, we also prove that deciding membership in the corresponding reduced stabilizer polytope is NP-hard. In particular, unless $\mathrm{P} = \mathrm{NP}$, no algorithm polynomial in $m$ can solve the problem in full generality, thus establishing fundamental complexity-theoretic limitations. Finally, we employ our framework to compute nonstabilizerness in different Hamiltonian ground states, demonstrating the practical performance of our method in regimes beyond the reach of existing techniques.

Both in quantum theory and in general probabilistic theories, measurements with $n$ outcomes are modelled as $n$-tuples of \emph{effects} summing up to the unit effect. Why is this the case, and can this assumption be meaningfully relaxed? Here we develop \emph{generalized measurement theories (GMTs)} as a mathematical framework for physical theories that is complementary to general probabilistic theories, and where this kind of question can be made precise and answered. We then give a definition of \emph{probabilistic state} on a GMT, prove that measurements are made of effects in every GMT in which the probabilistic states separate the measurements, and also argue that this separation condition is physically well-motivated. Finally, we also discuss when a GMT should be considered classical and characterize GMTs corresponding to Boolean algebras as those that are strongly classical and projective.

It was recently demonstrated that imposing the condition $P_{r} = -ρ$ on the radial pressure of a galactic halo can lead to regular black-hole solutions for certain density profiles, such as the Dehnen and Einasto models. In the present work, we show that some of the most commonly used halo profiles, including the Hernquist model, do not yield regular geometries under the same condition, but instead support black-hole solutions that retain a central singularity

We investigate the decomposition of ergotropy into incoherent and coherent contributions for quantum systems subject to typical Markovian noise channels. The incoherent part originates from population inversion in the energy eigenbasis after dephasing, while the coherent part captures the role of quantum coherence in work extraction. For single-qubit systems, we derive explicit conditions for freezing and enhancement of coherent ergotropy and obtain an analytical upper bound, showing that it cannot exceed one half of the state's quantum coherence. We then study two classes of separable two-qubit states under local noise. For Bell-diagonal states, which are locally completely passive and possess no local coherence, we prove that the total extractable work equals the average of geometric quantum and classical correlations. In this case, coherent ergotropy cannot be enhanced, although freezing occurs under specific noise conditions. By contrast, for separable states with local coherence, coherent ergotropy can increase under all considered noise channels, including phase-flip and depolarizing noise. Extending the analysis to multipartite systems, we show that both the magnitude and range of noise-induced enhancement grow with the number of qubits, indicating collective reinforcement. Finally, we demonstrate through an explicit example that entanglement does not prevent this enhancement: coherent ergotropy may increase under noise even for entangled states. Our results reveal that noise can assist energy storage, challenging the conventional view of noise as purely detrimental and suggesting compatibility between noise-assisted enhancement and fast entanglement-based charging mechanisms in quantum batteries.

This work examines the cosmological implications of two functional forms of $f(R,T) = R + αT^n$ gravity: for two different value of $n $ where $n=1$ and $n\neq 1$, and $α$ and $n$ are free parameters. The modified Friedmann equations are derived, and the cosmic evolution of the Hubble parameter $H(z)$ is determined. Cosmological parameters are estimated through $χ^2$ minimization and MCMC analysis using the emcee algorithm, with model parameters constrained by various observational datasets. Both models reproduce late-time acceleration and remain observationally indistinguishable from $Λ$CDM, while allowing small deviations parameterized by $α$ and $n$. The cosmological behavior of the deceleration parameter $q(z)$, the jerk, the snap parameter $s(z)$, and the effective equation of state parameter $ω$ is also analyzed. The results indicate that the Universe transitioned from deceleration to late-time accelerated expansion, consistent with the $Λ$CDM model. Analysis of the energy conditions reveals that the NEC and DEC are satisfied, while the SEC is violated, which explains the transition from a decelerating matter-dominated epoch to the present accelerated phase. These findings indicate that the proposed $f(R,T)$ models are compatible with current observational data and provide a viable alternative to $Λ$CDM in describing cosmic acceleration.

We develop a thermodynamic characterization of multimode entanglement in pure continuous-variable systems by quantifying the gap between globally and locally extractable work (ergotropy). For arbitrary pure multimode Gaussian states, we prove that the $2$-local ergotropic gap is a faithful entanglement monotone across any bipartition and constitutes a functionally independent upper bound to the Renyi-2 entanglement entropy. We further introduce the $k$-ergotropic score, the minimum $k$-local ergotropic gap, and show that it faithfully quantifies multimode entanglement across $k$ partitions. For pure three-mode Gaussian states, we derive its closed-form relation with the geometric measure for genuine multimode entanglement $(k=2)$, and total Gaussian multimode entanglement $(k=3)$. For systems with more than three modes, the $k$-ergotropic score becomes a functionally independent measure of multimode entanglement to the standard geometric measures. Our results reveal a direct operational hierarchy linking Gaussian multimode entanglement to work extraction under locality constraints, and provide a computable and experimentally accessible thermodynamic framework for characterizing quantum correlations.

We consider the motion of a charged spinning test/probe particle -- governed by the Mathisson-Papapetrou-Dixon equations with generic, adiabatic, and conservative spin- and field-induced multipole moments -- in a background $\sqrt{\text{Kerr}}$ field on flat spacetime: the electromagnetic field of a charged spinning ring-disk singularity obtained from the $G\to 0$ limit of the Kerr-Newman solution for a charged spinning black hole. We investigate the existence of two extra hidden constants of motion, analogous to the Carter constant (for geodesic motion in a Kerr spacetime, or for its spinning-probe generalization) and Rüdiger's linear-in-spin constant for a spinning probe in a Kerr background. We find that these two constants exist only when the Wilson coefficients parameterizing the probe's multipole structure take the particular values corresponding to spin-exponentiation of the effective Compton amplitudes through second order in spin.

We investigate static and spherically symmetric vacuum solutions in the symmetric teleparallel $f(\mathbb{Q})$ modified theory of gravity. Starting from a recently proposed classification of affine connections compatible with both the symmetries of spacetime and the constraints of symmetric teleparallel geometry, we develop a systematic approach to solve the full field equations. We first identify two distinct classes of connections that satisfy the off-diagonal metric field equations and the connection constraints. For an arbitrary $f(\mathbb{Q})$ function when the non-metricity scalar $\mathbb{Q}$ vanishes, we recover exact analytical solutions equivalent to those of general relativity, including the Schwarzschild and Schwarzschild (anti)de-Sitter metrics. We then extend our analysis beyond general relativity by considering the quadratic model $f(\mathbb{Q})=\mathbb{Q}+α~\mathbb{Q}^2$ with a small parameter $α$. Using a perturbative approach, we derive asymptotically flat, analytical solutions up to second order in $α$. These solutions exhibit corrections to the standard Schwarzschild metric, characterized by new integration constants that can be interpreted as connection hair. We explore the asymptotic behavior of these solutions and disclose that the horizon radius receives corrections that can be expressed compactly using the Lambert $\mathcal{W}$ function. Our results provide new, non-trivial vacuum solutions within $f(\mathbb{Q})$ gravity and highlight the rich structure introduced by the non-metricity connection.

We write down a series of basic laws for (strict) higher-order circuit diagrams. More precisely, we define higher-order circuit theories in terms of: (a) nesting, (b) temporal and spatial composition, and (c) equivalence between lower-order bipartite processes and higher-order bipartite states. In category-theoretic terms, these laws are expressed using enrichment and cotensors in symmetric polycategories, along with a frobenius-like coherence between them. We describe how these laws capture the salient features of higher-order quantum theory, and discover an upper bound for higher-order circuits: any higher-order circuit theory embeds into the theory of strong profunctors.

We investigate the ultraviolet behavior of 4-dimensional Lorentzian covariant Loop Quantum Gravity (LQG) and address the problem of infinite ambiguities relating to the triangulation dependence of spinfoam amplitudes. We consider the complete LQG amplitude that summing spinfoam amplitudes over 2-complexes. By introducing spin-network stacks and their covariant extension, spinfoam stacks, the summation over complexes is partitioned into distinct families. We demonstrate that the theory exhibits a condensation phenomenon, where quantum geometry condenses to a dominant small spin configuration. We identify a candidate fixed point controlling the ultraviolet (small spin) regime of covariant LQG. At this fix point, the complete LQG amplitude dynamically reduces to a topological theory at leading order, and the infinite ambiguities of triangulation dependence reduces to a finite set of boundary coefficients associated with a finite basis of 3-dimensional boundary blocks. These results provide a definition for the continuum limit of spinfoam theory at the fundamental level.

Recent multi-messenger observations, including gravitational wave detections of compact objects in the neutron star-black hole mass-gap region and precise measurements of high-mass pulsars, motivate mechanisms capable of enlarging the stellar mass window without arbitrarily stiffening the equation of state (EOS) toward the causal limit. In linear $f(Q)$ gravity of the form $f(Q)=β_1 Q+β_2$, the theory is dynamically equivalent to General Relativity at the geometric level and modifies stellar structure solely through a uniform rescaling of the matter sector governed by $β_1$. Consequently, linear $f(Q)$ alone does not introduce new geometric families of stellar solutions or alter classical compactness bounds. To overcome this structural limitation, we incorporate gravitational decoupling within an embedding class-I (Karmarkar) Vaidya-Tikekar configuration in linear $f(Q)$ gravity. While similar VT-based decoupling constructions exist in GR, the present framework introduces a controlled two-parameter deformation characterized by $(ε,β_1)$: the decoupling parameter $ε$ governs geometric deformation and EOS stiffness, whereas $β_1$ independently rescales the matter sector without altering the metric structure. This separation permits a direct comparison between GR and linear $f(Q)$ gravity at fixed geometric deformation, thereby isolating pure coupling-driven mass enhancement. We determine the admissible parameter domain from regularity, matching, causality and compactness requirements and derive an analytic compactness bound for the decoupled embedding class-I configuration. The combined action of $ε$ and $β_1$ enlarges the accessible stellar mass window while preserving physical acceptability, allowing configurations compatible with recent high-mass pulsars and mass-gap candidates without exceeding causal limits.

We use the Velocity-dependent One Scale Model for topological defect evolution to explore and classify the possible scaling solutions for string networks with time-varying tension, in cosmological and non-cosmological settings and under two different phenomenological assumptions for the behavior of these variations, which rely on different stretching and damping contributions to the string dynamics. We discuss how these assumptions impact the standard scaling solutions, as well as the evolution of the string network density. In addition to simple power-law cosmological epochs solutions, we also discuss the behavior of the network during the radiation-to-matter and matter-to-acceleration transitions. Overall, our results show that for the same amount of tension variation, a change in the stretching length scale tends to have a more significant impact on the network than a change in the damping length.

We study particle creation for scalar fields in conformally flat spacetimes using resummed heat-kernel techniques. We make use of an analogy between quantum scalar fields in conformally flat spacetimes and scalar field theories with a Yukawa coupling in Minkowski space. The correspondence holds exactly at the level of the effective action and includes nonconformal curvature couplings. This framework provides access to particle creation at strong curvature. In a radiation dominated universe, the particle production rates in arbitrary dimensions are independently confirmed through explicit calculations of the Bogoliubov coefficients. We also find new exact gravitational analogues of the Schwinger effect in quantum field theory in curved spacetime.

We experimentally demonstrate transitions between electronic angular momentum states with a difference in magnetic quantum numbers $Δ\mathrm{m_J} = $ 3, 4, and 5 via resonant four- and six-photon stimulated Raman transitions in a single trapped atom. Derivation of the corresponding Rabi frequencies, which are verified experimentally, follows the standard treatment of two-photon transitions including the adiabatic elimination of intermediate states. Finally, we discuss pathways to increase the observed multi-photon transition fidelities to $>99.99\%$, providing a tool for efficient, high-fidelity control of qu\textit{d}its and single-atom logical qubits.

Symmetry constraints provide a powerful means to control the dynamics of open quantum systems. However, the set of accessible control parameters is often limited. Here, we show that a tunable phase in the collective light-matter coupling of a cavity QED system induces a phase-dependent strong symmetry of the Liouvillian, enabling dynamical control of the open quantum system evolution. We demonstrate that tuning this phase substantially reduces the critical driving strength for dissipative phase transitions between stationary and non-stationary phases. We illustrate this mechanism in two experimentally relevant cavity QED settings: a two-species ensemble of two-level atoms and a single-species ensemble of three-level atoms. Our results establish phase control as a versatile tool for engineering dissipative phase transitions, with implications for quantum state preparation.

Galactic binaries are expected to be the most numerous LISA sources and to produce a stochastic gravitational-wave foreground whose spectral shape encodes information about the underlying population. Extracting this information with standard hierarchical methods is challenging due to the high dimensionality of the problem and the computational cost of global-fit analyses. We present a simulation-based inference framework to measure the population properties of galactic binaries directly from the reconstructed foreground. Adopting an astrophysically agnostic parametrization in the observable space -- defined by signal amplitude, frequency, and frequency derivative -- we generate synthetic catalogs and foreground spectra using a global-fit-inspired subtraction algorithm. We then train a neural posterior estimator to map spectra to population parameters. We validate our method on simulated data and recover population parameters with good accuracy, including the total number of binaries. As a by-product, we present a GPU-accelerated version of the subtraction algorithm, which delivers a ~100X speed-up compared to previous implementations in the literature. Our results demonstrate that LISA's stochastic foreground alone carries significant information about the Galactic binary population and provide a practical step toward joint inference from resolved and unresolved sources.

Understanding emergent phenomena in out-of-equilibrium interacting many-body systems is an exciting frontier in physical science. While quantum simulators represent a promising approach to this long-standing problem, in practice it can be challenging to directly realize the required interactions, measure arbitrary observables, and mitigate errors. Here we use coherent mapping between the Rydberg and hyperfine qubits in a neutral atom array simulator to engineer and probe complex quantum dynamics. We combine efficient analog dynamics with fully programmable state preparation and measurement, leverage non-destructive readout for loss information and atomic qubit reuse, and use an atom reservoir for replacing lost atoms. With this analog-digital approach, we first demonstrate dynamical engineering of ring-exchange and particle hopping dynamics via Floquet driving and measure the spectral function of single excitations by evolving initial superposition states. Extending these techniques to a 271-site kagome lattice, we employ closed-loop optimization to target an out-of-equilibrium critical quantum spin liquid of the Rokhsar-Kivelson type. We observe the key features of such a state, including the absence of local order, many-body coherences between nearly equal-amplitude dimer configurations over up to 18 sites, and universal correlations consistent with predictions from field theory. Together, these results pave the way for using dynamical control in analog-digital quantum simulators to study complex quantum many-body systems.

We analyze the public DESI full-shape clustering data using simulation-based priors (SBPs). Our priors are obtained by fitting normalizing flows to the distribution of EFT parameters measured from field-level simulations, themselves generated using tailored halo occupation distribution (HOD) models for each tracer. Incorporating SBPs in a power spectrum analysis significantly enhances $Λ$CDM cosmological parameter constraints; in combination with BAO information from DESI DR2 and a BBN prior on the baryon density, we find the matter density parameter $Ω_m=0.2987\pm 0.0066$, the Hubble constant $H_0=68.80\pm 0.35\,\mathrm{km}\,\mathrm{s}^{-1}\mathrm{Mpc}^{-1}$, and the mass fluctuation amplitude $σ_8 = 0.766\pm 0.015$ (or the lensing parameter $S_8=0.764\pm 0.018$), which are $1\%$, $40\%$ and $50\%$ stronger than the baseline results, though with a notable downwards shift in $σ_8$. The SBPs also have a significant impact in extended models, with the dark energy figure-of-merit improving by $70\%$ ($20\%$) in a $w_0w_a$CDM analysis when combining with the CMB (and supernovae). In the SBP analysis, we do not find statistically significant evidence for dynamical dark energy: the equation of state parameters are consistent with a cosmological constant within $2.2σ$ ($1.4σ$) in analyses without (with) supernovae. The neutrino mass constraints are also enhanced, with the $95\%$ limits $M_ν<0.073\,\mathrm{eV}$ and $M_ν<0.090\,\mathrm{eV}$ in $Λ$CDM and $w_0w_a$CDM respectively. The latter is the strongest constraint obtained to date and reinforces the preference for the normal neutrino mass hierarchy, regardless of the background dynamics. While our results are sensitive to HOD modeling assumptions, they clearly demonstrate that the inclusion of small-scale information can significantly sharpen cosmological parameter constraints.

We investigate the quasi-local thermodynamics of rotating Kerr-AdS black holes enclosed by a finite timelike boundary (cavity). Extending recent work on static systems, we define the holographic pressure and volume via the trace of the Brown-York stress tensor on the cutoff surface. We demonstrate that the inclusion of angular momentum introduces a momentum flux term at the boundary, requiring a generalized first law: $dE = T_{\mathrm{loc}}dS + Ω_{\mathrm{loc}}dJ - \mathcal{P}d\mathcal{A}$. We derive the explicit expressions for these thermodynamic conjugates and analyze the equation of state. Crucially, we examine the extensivity of the system in the large-size limit. We find that while small rotating black holes exhibit non-extensive behavior typical of self-gravitating systems, large Kerr-AdS black holes recover extensivity, behaving effectively as a thermal fluid on the boundary. This result strengthens the holographic interpretation of the cutoff surface as the domain of a dual field theory.

High-sensitivity accelerometers and gravimeters, achieving the ultimate limits of measurement sensitivity are key tools for advancing both fundamental and applied physics. While numerous platforms have been proposed to achieve this goal, from atom interferometers to optomechanical systems, all of these studies neglect the effects of intrinsic quantum uncertainty in time estimation. Starting from the Hamiltonian of a generic linear quantum sensor, we derive the two-parameter quantum Fisher information matrix and establish the corresponding Cram'er-Rao bound, treating time as an uncertain (nuisance) parameter. Our analysis reveals a fundamental coupling between time and signal estimation that inherently degrades measurement sensitivity, with the standard single-parameter quantum limit recovered only at specific interrogation times or under special decoupling conditions. We then apply these results to an optomechanical gravimeter and explicitly derive an optimal decoupling condition under which the effects of time uncertainty are averaged out in a continuous measurement scheme. Our approach is general and can be readily extended to a broad class of quantum sensors.

Organic molecules embedded in an organic matrix exhibit lifetime-limited optical coherence and bright emission at cryogenic temperatures below 3 K. Here we present a simple vapor-phase growth method for synthesizing optically thin DBT-doped anthracene crystals that are compatible with integrated nanophotonics. The crystals are ~200 nm thick with sub-nm surface roughness and a tunable lateral dimension of up to 200 $μ$m. The molecular transitions remain narrow and spectrally stable, with inhomogeneous broadening below 100 GHz, comparable to DBT in bulk anthracene. The dopant density is tunable up to several hundred molecules per $μ$m$^2$, ensuring emitters within the near-field of nanophotonic structures. We demonstrate that the crystals can be micropositioned onto integrated photonic devices with the molecular dipole aligned to the optical mode. This approach opens a path toward on-chip single-photon sources and collective many-emitter effects.

We address the problem of detecting initial system--environment correlations when the environment is not directly accessible. Most existing approaches rely on full state tomography or multiple system preparations, which can be experimentally demanding. We show that, for a known interaction, it can be sufficient to monitor a single expectation value of the system. Focusing on a qubit interacting with an environment via isotropic Heisenberg exchange, we derive exact bounds on the signal $z(t)=\langleσ_z^S\rangle(t)$ that hold for all factorized initial states. These bounds define a \emph{factorized envelope}: if an observed trajectory exits this envelope at any time, initial system--environment correlations are certified. From a reduced-dynamics perspective, the envelope admits a clear operational interpretation as the admissible region generated by the standard product assignment (embedding) map, which serves as a null model for uncorrelated preparations. Envelope violations therefore rule out the entire product-assignment class using only a single calibrated observable. We illustrate the method using three families of correlated initial states and observe clear envelope violations, including cases in which the reduced system state is maximally mixed. We further show that the same single-observable logic extends to an exactly solvable pure-dephasing spin--boson model with an infinite environment, where factorized initial states generate a simple coherence envelope whose violation certifies initial correlations. Overall, our results demonstrate that single-axis measurements, combined with a one-time calibration of $ρ_S(0)$, can certify initial system--environment correlations without tomography or environment access.

Quantum optimization has gained increasing attention as advances in quantum hardware enable the exploration of problem instances approaching real-world scale. Among existing approaches, variational quantum algorithms and quantum annealing dominate current research; however, both typically rely on one-hot encodings that severely limit scalability. Pauli Correlation Encoding (PCE) was recently introduced as an alternative paradigm that reduces qubit requirements by embedding problem variables into Pauli correlations. Despite its promise, PCE has not yet been studied in the context of constrained optimization. In this work, we extend the PCE framework to constrained combinatorial optimization problems and evaluate its performance across multiple problem sizes. Our results show that the standard PCE formulation struggles to reliably enforce constraints, which motivates the introduction of the Iterative-$α$ PCE. This iterative strategy significantly improves solution quality, achieving consistent constraint satisfaction while yielding better cut sizes across a wide range of instances. These findings highlight both the limitations of current PCE formulations for constrained problems and the effectiveness of iterative strategies for advancing quantum optimization in the NISQ era.

We prove that there exists an absolute constant $α<1$ such that for every finite dimension $d$ and every quantum channel $T$ on $\mathsf{L}(\mathbb{C}^d)$, $\left\|Θ\circ(\mathrm{id}-T)\right\|_\diamond \le α\,\left\|Θ\right\|_\diamond\,\left\|\mathrm{id}-T\right\|_\diamond$, where $Θ$ is the transposition map. In fact we show the explicit choice $α=1/\sqrt{2}$ works.

Superposition is the core feature that sets quantum theory apart from classical physics. Here, we investigate whether sets of quantum measurements can be modelled by using only devices that are operationally classical, in the sense that they have no superposition properties. This leads us to propose classical measurement models, which we show to be stronger than commutative measurements but weaker than joint measurability. We determine both the exact depolarisation noise rate and the measurement loss rate at which the all projective measurements in $d$-dimensional quantum theory admit a classical model. For finite sets of quantum measurements we develop methods both for constructing classical models and for falsifying the existence of such model via prepare-and-measure setups. Furthermore, we show that this concept also has operational implications. For that, we consider whether quantum measurements with classical side-information can be implemented in sequence without causing a disturbance and we show that classical models imply an affirmative answer. Our work sheds light on superposition as a resource for quantum measurement devices.

We calculate the phase diagram of the Bose-Hubbard model on a half-filled ladder lattice including the effect of finite on-site interactions. This shows that the rung-Mott insulator (RMI) phase persists to finite interaction strength, and we calculate the RMI-superfluid phase boundary in the thermodynamic limit. We show that the phases can still be distinguished using the number and parity variances, which are observables accessible in a quantum-gas microscope. Phases analogous to the RMI were found to exist in other quasi-1D lattice structures, with the lattice connectivity modifying the phase boundaries. This shows that the the presence of these phases is the result of states with one-dimensional structures being mapped onto higher dimensional systems, driven by the reduction of hopping rates along different directions.

An industrial-scale adoption of Quantum Key Distribution (QKD) requires the development of practical, stable, resilient and cost-effective hardware that can be manufactured at large scales. In this work we present a high-speed (1.25GHz), field-deployable QKD prototype based on integrated photonics, that is consolidated into standard 19-inch rack compatible units. Through integrated photonics, the system prioritizes autonomous long-term stability in metropolitan settings. The architecture is further simplified by removing the need for chromatic dispersion compensation over metropolitan distances (below 100km). We demonstrate continuous key exchange over more than 4 km of metropolitan optical fiber, where the prototype maintained stable, uninterrupted operation across a measurement spanning more than 12 day-night cycles without manual intervention.

High-energy multimessenger background analyses over the past decade have provided evidence for a population of hidden neutrino sources that are opaque to GeV-TeV gamma rays, a picture bolstered by recent observations of the nearby active galaxy NGC 1068. The coronal regions in the hearts of active galactic nuclei (AGNs) have been proposed as the most promising sites for such hidden nonthermal particle production, and NGC 1068 is expected to be the most neutrino-active galaxy for IceCube. We demonstrate that the latest all-sky neutrino spectrum, exhibiting a spectral bend around 3-30 TeV, is consistent with predictions of the magnetically powered corona scenario, and the models for the all-sky neutrino flux can simultaneously explain the multimessenger data from NGC 1068 within observational and modeling uncertainties. We further show, in a largely model-independent way, that the contribution from NGC 1068-like sources does not overshoot the observed medium-energy neutrino flux. Finally, we highlight the key role of the Eddington ratio, which can drive substantial variations in the predicted neutrino fluxes of nearby AGNs, and we encourage systematic multimessenger searches for the neutrino-brightest AGNs.

The aim of this work is to develop high precision Time-of-Flight (TOF) devices based on high refractive index solid Cherenkov radiators read out by silicon photomultipliers (SiPMs). Cherenkov light is prompt and therefore ideal for reaching the intrinsic timing limits of TOF systems. By utilizing a thin, high-refractive-index radiator a nearly instantaneous signal is generated by particles exceeding the Cherenkov threshold. In order to achieve the ultimate time resolution, we carried out a rigorous optimization of the radiator material and geometry, alongside the efficiency of the optical coupling to the SiPM sensors. The key factors limiting the time resolution were characterized by comprehensive Monte Carlo simulations, subsequently validated against experimental beam test data. We assembled small-scale prototypes instrumented with various Hamamatsu SiPM arrays sensors with pitches ranging from 1.3 to 3 mm coupled with various window materials, such as fused silica and MgF2, featuring various thickness values. The prototypes were successfully tested in beam test campaigns at the CERN-PS T10 beam line. The data were collected with a complete chain of front-end and readout electronics based on either the Petiroc 2A or the Radioroc 2 interfaced to a picoTDC to measure charges and times. By comparing the time measurements with two SiPM arrays we were able to measure a time resolution better than 33.2 ps at the full system level with a charged particle detection efficiency of 100%. Our results demonstrate the expected performance benchmarks for the charged particle detection efficiency and time resolution and highlight the potential of the developed Cherenkov-based TOF detectors for next-generation particle identification systems.

In this work we study the production of a dark matter (DM) particle in association with a Higgs boson via a central exclusive photon-fusion initiated process. We explore this type of production through the Inert Doublet Model plus a complex Singlet (IDMS), where an extension of the Standard Model by an additional $U(1)_X$ gauge symmetry and a $SU(2)$ inert scalar doublet gives rise to a DM candidate $χ$. This particular process involves the collision of two protons exchanging two colorless particles (in our case, photons), from which a central process occurs. Such interaction can be detected in the LHC using forward proton detectors, where the resulting missing mass spectrum can be observed after proton reconstruction, thus allowing a search for physics Beyond the Standard Model (BSM). We present results for different values of the difference of masses of a heavy scalar coming from the complex singlet, the DM candidate and the Higgs boson, $Δ= M_S - M_χ- M_h$, which is the phase space available for the final state in the central exclusive process.

We analyse CMB constraints on stable networks of cosmic strings and domain walls using for the first time full Planck 2018 data together with BICEP/Keck 2018 $B$-mode measurements. The defect-induced anisotropies are computed using the Unconnected Segment Model for Nambu-Goto and Abelian-Higgs strings, as well as for stable domain walls, and included in a full Markov Chain Monte Carlo analysis jointly varying all $Λ$CDM parameters, the tensor-to-scalar ratio, and the string/domain wall tension. No statistically significant evidence for defects is found, although we observe a mild preference for non-zero cosmic string tension. Our results improve previous constraints on the defect power spectrum by up to a factor of two. In the particular case of strings, the improvement is driven by the $B$-mode data, and is especially pronounced for Abelian-Higgs strings. We also present forecasts for upcoming Simons Observatory data and find that, with the baseline noise configuration, the constraints on the string tension could improve by about a factor of three. Finally, we assess the impact of Nambu-Goto string loops on CMB anisotropies in light of both current and future observations.

This paper studies the asymptotic gauge charges of the Curtright mixed-symmetry rank-3 field $φ_{[ρσ]ν}$ in Minkowski spacetime, interpreted in $ D = 5 $ as the dual graviton. In Bondi coordinates at future null infinity, we impose radiation fall-offs and fix a de Donder-like gauge together with an on-shell traceless condition, similarly to what happens in linearized gravity. Surface charges associated with the residual gauge transformations are constructed as boundary integrals via Nöther's 2-form. In $ D = 5 $, exploiting Hodge/Hodge-like decompositions on $ S^{3} $, the charge splits into a scalar sector $ Q_Φ $, a vector sector $ Q_{V} $ and a TT sector $Q_{y^{\text{TT}}}$. $ Q_Φ $ is parametrized by a single arbitrary scalar function $ Φ$ (interpreted as the supertranslation-like parameter), $ Q_{V} $ is parametrized by a vector field $ V^{i} \in \mathfrak{Diff}(S^{3}) $ and the TT sector $Q_{y^{\text{TT}}}$ is parametrized by a trasverse-traceless rank-2 tensor $y_{ij}^{\text{TT}} \in \mathfrak{TT}(S^3)$. The corresponding charge algebra closes only if $V_i \in \mathfrak{o}(4)$ as semidirect sum $ \mathfrak{o}(4) \loplus (C^{\infty}(S^3) \oplus \mathfrak{TT}(S^3)) $, i.e an abelian extension of a $\mathfrak{BMS}$-like algebra featuring a higher-spin-like supertranslation sector.

Energy dependence of global scattering parameters, mostly of total cross section, is studied for proton-proton and antiproton-proton collisions. Results are presented for physical analysis updated with taken into account the recent data from accelerator experiments as well as from cosmic ray measurements. The analytic parameterizations suggested within Axiomatic Quantum Field Theory (AQFT) provide the quantitative description of energy dependence of global scattering parameters for rather wide energy range. Detailed scan on low boundary of the fitting range for energy dependence of global scattering parameters allows the observation of the onsets for regions in which Pomeranchuk theorem and / or Froissart-Martin one is valid. It is obtained that global scattering parameters show the behavior corresponded to any formulations of Pomeranchuk theorem and closed to (modified) Froissart-Martin limit in functional sense in multi-TeV energy region. Bosonic condensation is considered as one of the possible dynamical mechanisms which would be provide the total cross section approaches to (modified) Froissart-Martin limit at quantitative level but not functionally only.

A topological A-model constructed from supertwistor variables and a worldsheet U(1) gauge field was recently proposed to describe the $AdS_5\times S^5$ superstring at zero radius which is dual to free N=4 d=4 super-Yang-Mills. In this note, holomorphic worldsheet instanton solutions of the supertwistor variables are constructed where the U(1) worldsheet gauge field is identified with the square-root of the Strebel differential that describes the ribbon graph on the worldsheet.

Massless type IIB superstring amplitudes are organized according to the number of external states and their ${\mathrm U}(1)$ charge under the R-symmetry of type IIB supergravity. In this work, we analyze the low-energy expansion of one-loop five-point amplitudes in all charge sectors, focusing on the representative processes involving five gravitons and four gravitons with one dilaton. We compute the one-loop contributions to the moduli-dependent couplings in the type IIB effective action up to the $D^{12}R^5$ and $D^{14}φR^4$ interactions. The results are consistent with $S$-duality constraints in every charge sector and exhibit rich arithmetic structure, including single-valued multiple zeta values, affine linear combinations of logarithmic derivatives of the Riemann zeta function at odd integers, and a new constant of currently unknown nature.

LHCb's Run 3 upgrade introduced a fully software-based trigger system operating at 30~MHz, processing an average of 5.6 proton-proton collision vertices per bunch crossing (event). This work presents the development of an inference engine for PVFinder, a hybrid deep neural network for finding primary vertices, the proton-proton collision points from which all subsequent particle decays originate into Allen, LHCb's High Level Trigger (HLT1) framework. The integration addresses critical real-time constraints including fixed memory pools, single-stream execution, and sub-400~$μ$s per-event processing budgets on NVIDIA GPUs. We introduce a translation layer that bridges Allen's Structure-of-Arrays (SoA) data layout with cuDNN's tensor format while maintaining zero-copy semantics and deterministic behavior. Current performance shows the CNN stage contributes significant throughput overhead. We present a roadmap targeting order-of-magnitude improvements through mixed-precision computing, model compression and other techniques.

Leptophilic dark matter (DM) is a well-motivated thermal WIMP framework that can evade stringent nuclear-recoil searches while remaining testable via DM-induced heating of neutron stars (NS). In this work, we study leptophilic scalar DM in a $\mathrm{U(1)}_{L_μ-L_τ}$ gauge extension of the Standard Model, which provides a common leptophilic portal for all scenarios considered. To reproduce the observed relic abundance while suppressing direct-detection signals, we investigate three benchmark realizations: (i) a secluded DM scenario in which the relic density is set by annihilation into $\mathrm{U(1)}_{L_μ-L_τ}$ gauge bosons, and two pseudo-Nambu-Goldstone boson (pNGB) DM models based on (ii) an SO(4) symmetry and (iii) an SO(3) symmetry. In the SO(4) pNGB model, the DM mass arises at tree level from a soft breaking term, while the elastic scattering amplitude is suppressed by a symmetry-protected cancellation. In the SO(3) pNGB model, the DM mass is generated radiatively at one loop via the $\mathrm{U(1)}_{L_μ-L_τ}$ gauge interaction, and we show that this gauging preserves the same cancellation mechanism, maintaining compatibility with direct-detection null results. We perform a systematic parameter scan imposing relic density, direct- and indirect-detection, and neutrino trident constraints, and identify viable sub-TeV to TeV DM candidates. Assuming maximal capture in NSs, we find that the remaining parameter space can be tested by near-infrared observations of NSs, providing sensitivity complementary to terrestrial searches in regions that are currently weakly constrained.

Longitudinal electroweak gauge bosons are the most technically delicate ingredient of electroweak parton showers: in the broken Standard Model, the gauge component of a longitudinal polarisation does not cancel diagram by diagram, but is related by Ward identities to amplitudes with an insertion of the associated would-be Goldstone field. Building on the default Herwig 7 treatment based on the Dawson-subtracted longitudinal current, we construct a gauge-invariant longitudinal scheme in which the Dawson remainder is retained as a well-defined contribution and completed by a Ward-identity-fixed Goldstone-matching term. We derive helicity-resolved building blocks and compact quasi-collinear splitting kernels for $q\to q'V$ and $V\to V'V''$ branchings, and implement the scheme in Herwig 7 as a switchable alternative to the default longitudinal basis. The completion leaves the transverse sector unchanged and modifies only longitudinal entries through controlled symmetry-breaking terms, including Yukawa-sensitive contributions in massive-fermion channels. In shower-level studies, we find that the Dawson and gauge-invariant prescriptions coincide at high evolution scales, while differing at lower scales precisely in channels where symmetry breaking is active. Exclusive single-emission observables can moreover display non-monotonic scheme dependence once Sudakov suppression and kinematic constraints are accounted for. A first multi-emission LHC-like study confirms numerical stability and yields controlled, interpretable shifts in observables that probe propagating electroweak shower currents, whereas quantities dominated by promptly contracted, near on-shell vector-boson matrix elements remain largely unchanged.