arXiv Daily Digest - 2026-06-18

AudioLLMs enable speech recognition conditioned on textual prompts such as domain descriptions or entity lists. However, it remains unclear whether these models genuinely utilise such context or rely on parametric knowledge learned during pretraining. Existing benchmarks cannot answer this question because they evaluate transcription under fixed prompting conditions and rarely include explicit contextual inputs. We introduce IndicContextEval, a 56-hour multilingual benchmark of natural speech from 555 speakers across 8 Indian languages and 23 professional domains. We design a 7-level prompting framework that progressively introduces contextual signals, including metadata, natural-language descriptions, entity lists in English and native script, and adversarial prompts with incorrect entities. Evaluating five models reveals substantial differences in context utilisation behaviour, highlighting the need for explicit evaluation of contextual grounding in AudioLLMs.

Identifying the lowest-energy surface-adsorbate configuration is critical for modeling heterogeneous catalysis, yet exhaustive exploration with ab initio calculations is computationally prohibitive. Machine-learning force fields (MLFFs) accelerate structural relaxation but leave the search over the vast configurational space a major bottleneck, and open-loop large language model (LLM) agents lack a physics-grounded feedback mechanism to correct erroneous initial guesses. We propose AdsMind (Adsorption configuration discovery with Machine intelligence and relaxation feedback), a closed-loop multi-agent framework that enables autonomous error correction through MLFF relaxation feedback. Across four LLM backends, AdsMind achieves consistently high search reliability, with success rates of 100% and 98.8% on the benchmarks AA20 and OCD-GMAE62. Relative to its single-pass (1-Shot) ablation it reduces cross-backend energy dispersion, and it uses only 4.11 and 4.67 MLFF relaxations per case, respectively -- an approximately 14-fold reduction over heuristic enumeration baselines. Density functional theory (DFT) validation using VASP/PBE on six representative AA20 systems shows that the reported open-loop Adsorb-Agent outputs exhibit qualitative adsorption-energy sign errors for molecular adsorbates, whereas AdsMind preserves the correct sign in all tested cases with closer quantitative agreement. AdsMind thus delivers reliability, self-reflection, and interpretability simultaneously, supporting more DFT-informed autonomous chemistry workflows.

The remarkable success of Transformer-based models in natural language processing stems from architectural scaling, which leads to a large number of parameters and hinders deployment in resource-constrained environments. While structured pruning offers a pathway to compression, existing state-of-the-art methods often rely on gradient-based importance ranking or stochastic gating, which suffer from instability, structural degeneration, and the need for extensive manual hyperparameter tuning. In this paper, we introduce CAHP (Complementary Attention Head Pruning), a novel post-hoc framework that redefines head selection as a global graph-theoretical problem. Rather than evaluating heads in isolation, CAHP utilizes graph-based clustering combined with information-theoretic distance measures to identify and preserve a topologically diverse subset of complementary attention heads. Without requiring a predefined sparsity level or pruning ratio, the framework automatically determines the number of selected attention heads across layers by identifying a diminishing marginal performance curve, where pruning additional heads leads to a sharp degradation in performance, as determined by the chosen polynomial degree. Extensive evaluations on the SST-5 and MNLI benchmarks, across different Transformer model scales, demonstrate that CAHP consistently outperforms competitive baselines, particularly in high-compression regimes. Furthermore, our structural analysis shows that CAHP avoids the "proximity bias" of gradient-based pruning methods, which tend to preserve heads mainly in layers close to the output, and instead retains a functionally critical set of attention heads in the model's intermediate layers.

Automated vulnerability discovery in large codebases remains challenging: traditional static analysis produces high false-positive rates, while dynamic approaches such as fuzzing require substantial infrastructure and often target narrow classes of bugs. Recent advances in large language models (LLMs) enable semantic reasoning about program behavior, but applying LLMs to repository-scale security analysis introduces challenges related to context management, cost, and verification. We present OpenAnt, an open-source vulnerability discovery system that integrates static program analysis with LLM-based reasoning in a multi-stage pipeline. OpenAnt introduces three key techniques. First, codebases are decomposed into self-contained analysis units filtered by reachability from external entry points, reducing the analysis surface by up to 97% while preserving attack-relevant code. Second, candidate vulnerabilities undergo adversarial verification through constrained attacker simulation, where the model evaluates exploitability under realistic attacker capabilities. Third, findings are validated through dynamic verification, in which exploit environments are generated automatically, executed in sandboxed containers, and discarded after use. Evaluation on widely used open-source projects including OpenSSL, WordPress, and Flowise shows that this architecture can identify previously unknown vulnerabilities while maintaining manageable analysis cost and substantially reducing false positives. Our results suggest that closed-loop vulnerability discovery pipelines, combining semantic reasoning with exploit validation, provide a practical path toward scalable automated security analysis. OpenAnt is released as open source under the Apache 2.0 license at https://github.com/knostic/OpenAnt.

This paper develops local certificates for population-risk increments around a current model. For a local candidate set \(\mathcal D\), the certificate is a two-sided confidence band for \(P({\ell_{θ+v}-\ell_θ})\) over \(v\in\mathcal D\). As an application, the upper endpoint of this band yields a risk-controlled update rule: an update is accepted only when its certified upper endpoint is nonpositive; otherwise the current model is retained.

Dynamical systems are fundamental to modeling the natural world, yet modeling them involves a persistent trade-off: manually prescribed mechanistic models are interpretable by design but often overly simplistic and misspecified; in contrast, flexible data-driven neural methods lack physical insight. Hybrid modeling aims for the best of both worlds by combining a prescribed or symbolic, physics-based component with a flexible neural network. A critical challenge, however, is that the neural component may relearn mechanistic parts, yielding redundant and uninterpretable models, especially when the symbolic structure itself is discovered from data. Existing methods based on standard $L^2$ regularization rely on a projection argument that breaks when the symbolic component is learned through sparse discovery, allowing the neural augmentation to overlap with symbolic structure. We introduce \textbf{OrthoReg} (Orthogonal Regularization), which directly penalizes overlap between the symbolic and neural components, preventing symbolic structure from being absorbed by the neural residual. This yields a complementary decomposition: the symbolic part captures what the library can express, and the neural part captures what remains. On benchmark dynamical systems with partial library mismatch, OrthoReg improves symbolic recovery and out-of-distribution behavior.

Current conversational AI systems have made significant progress in language generation, personalization, and long-context interaction. However, most existing methods model social behavior through isolated components such as emotion modeling, memory retrieval, or persona conditioning, lacking a unified framework to explain the emergence of stable social relationships and social intelligence in long-term human-AI interaction.To address this, we propose the Human-AI Coevolution Dynamics Framework (HACD-H), a formal model of human-AI interaction as a self-organizing social cognitive system. HACD-H integrates emotional adaptation, relational organization, social memory, and personality consistency into a unified dynamical framework and introduces principles including multi-timescale social cognition, relational attractors, trust basins, developmental phase transitions, and social cognitive energy dynamics.We construct a conversational dataset with approximately 14,700 interaction turns and develop a theory-driven empirical evaluation framework. Results reveal a hierarchy of temporal persistence in social cognition, stable relational attractors, phase-transition-like developmental patterns, and a structured social cognitive energy landscape. Social intelligence shows a significant negative correlation with social cognitive energy (r = -0.391, p < 0.001), and interaction trajectories exhibit progressive energy reduction over time.These findings suggest that social intelligence emerges from long-term social cognitive coevolution rather than isolated conversational capabilities. HACD-H provides a unified theoretical foundation for modeling adaptive human-AI social interaction and developing socially intelligent AI systems.

Accurate survival prediction is essential for personalized treatment planning in head and neck cancer, yet remains challenging due to the heterogeneous and high-dimensional nature of multimodal clinical data. While deep survival models have improved predictive performance over classical statistical approaches, existing methods typically rely on static fusion strategies or temporally agnostic modeling, limiting their ability to capture structured clinical workflows. In this work, we propose ChronoSurv, a heterogeneous hierarchical directed graph framework for multimodal survival analysis. ChronoSurv represents patient care as a progression-aware clinical trajectory using directed graphs aligned with key diagnostic steps. A hierarchical topology incorporates fine-grained, coarse, and global representations, further supporting flexible adaptation to missing modalities, while heterogeneous message passing models complex and asymmetric relationships across modalities and clinical steps. Experimental results on two public datasets demonstrate that ChronoSurv achieves state-of-the-art discriminative performance while maintaining statistically reliable calibration. Comprehensive ablation studies further confirm the contribution of each architectural component, highlighting the potential of trajectory-aware graph modeling for multimodal survival prediction.

Automatic Handwritten Text Recognition (HTR) is inherently a challenging task, and its complexity is further increased when dealing with cursive scripts. Although significant efforts have been made on various cursive scripts, research regarding Urdu Handwritten Text Recognition (UHTR) has been relatively limited. This lag of research is primarily due to the unique challenges posed by its script, and the scarcity and unavailability of benchmark datasets. Therefore, to advance research in UHTR, this study presents a specialized real dataset called the Urdu Katib Handwritten Dataset (UKHD). To the best of our knowledge, this is the first offline Urdu handwritten text lines dataset specifically curated from the materials written by Katibs in historical times. It encompasses a diverse range of flat nib writing variations in the Nastalique calligraphic style. Additionally, the effectiveness of different CRNN-based hybrid models has been evaluated to identify the optimal architecture for Urdu Katib Handwriting Recognition (UKHR). Among the analyzed models, the CNN-BGRU-CTC model showed more robust performance, with low Character Error Rate (CER) and Word Error Rate (WER). This research work aims to support and encourage the research community in developing a robust recognition system for preserving Urdu handwritten literature.

Neural Controlled Differential Equations (NCDE) provide a powerful continuous-time framework for forecasting time series, but standard graph-based extensions typically learn spatial structure purely from data, even in settings where a directed graph structure is known a priori. We introduce Informed Neural controlled Differential EQuationS (INDEQS), a graph-based NCDE forecasting method that incorporates prior knowledge of a directed graph at distinct architectural positions. INDEQS separates inner mixing of hidden states across graph nodes from outer mixing between vector field and control, and offers both a lightweight graph-constrained variant and a more expressive variant, learning additional graph connections from data via adaptive graph convolutions. To systematically study when graph informedness is beneficial in forecasting, we devise a continuous advection simulation on directed graphs, yielding synthetic spatio-temporal datasets with known ground-truth flow structure. We then evaluate INDEQS on two real-world tasks: river discharge forecasting on a hydrological network and traffic flow prediction on PeMS08. Across these synthetic and real-world benchmarks, outer informedness consistently improves mean absolute error over an uninformed NCDE with comparable parameter count, particularly on larger graphs, while inner informedness offers a more parameter-efficient alternative when strict adherence to a known adjacency is desired. A comparison of discrete convolutional and continuous-time decoders further shows that continuous decoders yield better accuracy and greater temporal flexibility on real-world tasks. An implementation of INDEQS and the advection simulation is available at https://github.com/Mitchi1/indeqs.

As large language models (LLMs) advance and multi-agent systems aim to overcome the limits of standalone agents, robust communication protocols are becoming essential infrastructure for distributed agent networks. Nonetheless, the fragmented protocol landscape presents a significant interoperability challenge. This study develops a technical taxonomy to classify and analyze LLM agent communication protocols. Following an established iterative method, we defined the taxonomy's purpose, meta-characteristic, and ending conditions, then performed five iterations, three empirical-to-conceptual and two conceptual-to-empirical, on nine actively maintained open-source protocols with demonstrable adoption. The taxonomy comprises five dimensions: counterparty, payload, interaction state, discovery mechanism, and schema flexibility. Classification reveals recurring architectural patterns: all sampled agent-to-agent protocols combine hybrid payloads with session-state persistence; most protocols support multiple predefined schemas, and two negotiate schemas at runtime, indicating a trend toward schema flexibility; decentralized discovery remains rare. Analysis suggests short-term convergence pressure toward protocols unifying agent-to-agent and agent-to-context (tool and data) communication. Long-term, however, no single protocol is likely to maximize versatility, efficiency, and portability simultaneously. The field will more likely evolve toward a federated, layered protocol stack. The framework guides protocol selection and highlights open research gaps such as privacy and policy enforcement.}

We present Pareto Q-Learning with Reward Machines (PQLRM), a multi-objective reinforcement learning algorithm for tasks whose reward structure is specified by a set of reward machines (RMs). PQLRM combines Pareto Q-Learning (PQL), which maintains sets of vector-valued Q-estimates to approximate the Pareto front, with enhancements from Q-Learning with Reward Machines (QRM), which exploits the factored automaton structure of the reward signal. This yields a multi-policy algorithm that remains sample-efficient under non-Markovian, RM-encoded rewards. Experimental trials show that PQLRM converges faster than a naive PQL baseline applied to the cross-product MDP and can synthesize Pareto-optimal policies that QRM cannot.

Scalable prediction of optical spectra is a critical component of high-throughput materials screening for optoelectronic applications such as solar cells. Existing surrogate models are trained on spectra computed from lower levels of theory or rely on rotation-invariant scalar features, limiting their geometric expressiveness. We explore the use of equivariant graph neural networks for optical spectra prediction, adapting GotenNet to this task and evaluating it on multiple datasets including a recently published collection of 10,533 structures with spectra computed at the level of the random phase approximation (RPA). The proposed model outperforms the current state of the art, with the largest gains in the 0-8 eV range and on predicting the static real permittivity, both of particular relevance for thin-film optics.

Dealing simultaneously with confidentiality and Byzantine behaviors in decentralized learning is a challenging problem. Indeed, in decentralized learning, clients train a machine learning model while keeping their data locally and share their model parameters or gradients with a set of neighbors. While enforcing confidentiality calls for hiding the exchanged model parameters/gradients (e.g., by using cryptographic techniques), dealing with Byzantine contributions often requires inspecting the latter. Hence, most research works address these objectives separately. A recent line of work proposes to employ secure multi-party computation (MPC) to implement robust aggregators against model poisoning, thereby enforcing both confidentiality and Byzantine resilience. However, these solutions scale badly: they either require all-to-all communication between participants or delegate the entire computation to a small subset, whose computational and communication load grows proportionally with the size of the network. In this paper, we present Giskard, a protocol for confidential and Byzantine-robust decentralized aggregation. Giskard organizes $n$ parties into a tree of committees of size $O(\log n)$ and evaluates a coordinate-wise approximate median via a committee-adapted distributed binary search over the value domain, using BGW-style MPC within each committee. We assess Giskard both theoretically by proving its security and confidentiality properties and experimentally through extensive experiments involving up to one million participants. Compared to its closest competitors, Giskard reduces per-party communication complexity asymptotically while exhibiting comparable model utility under up to $n/4$ Byzantine parties.

The prevailing engineering intuition for addressing conceptual drift in long-horizon LLM collaboration is to trade more formal constraints for more reliable outputs -- designing symbolic identifier systems, accumulating defensive rules in System Prompts, expanding context windows. Our engineering record shows that in long-horizon settings, this direction may produce effects contrary to design intent. Using action research methods in a real software project (Bang-v3) spanning approximately one month and 391 collaborative sessions, we document and analyze the failure process of these strategies. When the symbolic system exceeds a complexity threshold, LLMs do not become more accurate -- instead, they abandon genuine understanding of business semantics, retreat to self-referential reasoning within the symbolic layer, and generate outputs that appear internally consistent but are physically disconnected from reality. We name this failure pattern "Index Sickness," and its canonical manifestation "Phantom Legislation." We name the underlying principle the "Pang Principle (Semantic Vitality Law)": natural language carrying explicit purpose conveys far greater information quality than symbolic expression. From this, we design and validate its physical engineering mechanism: "Baseline-Log Physical Separation." In the same project, this mechanism reduced AI Instructions volume by ~75%, and across the subsequent ~150 sessions, no recurrence of Index Sickness was observed. A bilingual companion version (Chinese) is included as supplementary material.

On-policy self-distillation (OPSD) trains a model on its own rollouts and uses a frozen copy to provide dense token-level targets conditioned on a reference target. This works well for LLM reasoning, but a direct extension to multimodal large language models (MLLMs) can create a shortcut: the privileged target may guide tokens mainly based on the text reference target rather than the image. We propose ViGOS, a visually grounded OPSD framework for MLLM post-training. The student first writes a visual description and then reasons toward the final answer. For valid rollouts, an image-only perception teacher supervises the description, while a privileged reasoning teacher supervises the reasoning and final answer on the same student prefix. A reference teacher is used only for invalid rollouts to recover the output format. Across general vision-language, expert reasoning, visual math, spatial grounding, and visual-language-prior benchmarks, ViGOS keeps the main benefits of OPSD and improves image-grounded behavior in shortcut-prone settings.

Processing-using-DRAM (PuD) is a promising computation paradigm that alleviates frequent data movement between main memory and processing units by using each DRAM column as a computation engine via simultaneous multiple-row activation (SiMRA). Unfortunately, DRAM density scaling may hinder PuD's benefits: denser cell arrays bring rows and columns closer, making regular DRAM operations susceptible to noise and interference from neighboring cells. Yet no prior work investigates whether interference from rows or columns not intended to participate in computation can compromise PuD robustness. In this work, we reveal PuDGhost, an interference phenomenon where a PuD operation in a given column produces erroneous results due to interference from 1) data in non-activated DRAM rows and 2) data in other columns that compute concurrently under the same SiMRA operation. PuDGhost violates the ideal picture that each column's computation depends solely on its own operand data, threatening future PuD systems. We present the first extensive characterization of PuDGhost using 96 real DDR4 DRAM chips from 12 modules, quantifying these two interference sources under various conditions. Among our 15 new empirical observations, we highlight two major results: 1) data in adjacent non-activated rows affects SiMRA outputs by up to 10% for random inputs, and 2) data in concurrently computing columns affects SiMRA outputs by up to 48% for random inputs. Guided by these findings, we propose countermeasures across multiple layers of the PuD computing stack. Specifically, we evaluate on real DDR4 DRAM chips: 1) robust column screening that reduces the risk of using unreliable columns in the presence of PuDGhost, and 2) a compute row layout that mitigates PuDGhost via dedicated rows between compute rows. Our solutions greatly improve PuD computation accuracy and provide a foundation for robust future PuD systems.

Electricity markets are inherently complex systems characterised by strong nonlinearities, high-dimensional interactions, and increasing interdependence across regions. While deep neural networks (DNNs) have demonstrated strong predictive capabilities for electricity prices, their lack of interpretability limits their usefulness for understanding the underlying drivers of price formation. This paper addresses this gap by combining DNN models with explainable artificial intelligence (XAI) techniques to analyse the determinants of electricity prices across 39 European bidding zones. We employ SHAP (SHapley Additive exPlanations) to quantify feature contributions and apply and extend SSHAP, an aggregation framework to improve interpretability in high-dimensional settings. The analysis identifies that renewable energy sources, particularly solar, play a disproportionately important role in price formation despite their lower share in total power generation. Gas prices remain a dominant and consistent driver across electricity markets, while interconnections significantly shape price dynamics, highlighting the strong interdependence of European electricity systems. In addition, a synthetic EU-wide electricity market is constructed to explore the counterfactual scenario of a fully integrated market with a single price.

Offline policy learning has received growing attention in causal inference. The primary objective is to learn a policy (individualized treatment rule) as a mapping from covariates to treatment that maximizes the empirical welfare defined as the mean of scalar-valued potential outcomes. In this paper, we study offline policy learning with distribution-valued outcomes, where each potential outcome is a probability measure on $\mathbb{R}$ and the reward is defined through a utility functional applied to the Wasserstein barycenter of induced outcome distributions. We establish statistical guarantees for the policy learning framework based on both Inverse Probability Weighting (IPW) and Doubly Robust (DR) estimators. By handling the challenging uniform deviation over the product of the combinatorial policy class and the infinite-dimensional quantile domain, we prove that the finite-sample regret has leading dependence $\widetilde{\mathcal{O}}(\sqrt{\mathrm{N\text{-}dim}(Π)/N})$. In the one-dimensional Wasserstein setting and under the stated regularity conditions, the leading regret rate is still governed by the policy-class complexity. Moreover, we provide a minimax lower bound establishing the sharpness of the leading dependence on $N$ and $\mathrm{N\text{-}dim}(Π)$.

The World Wide Web was built on an assumption held for three decades: the primary consumer of web content is a human being. This permeates every layer; its access model presumes human visitors, its economics rest on human attention, and its content targets human perception. The rapid emergence of AI agents as intermediaries between humans and web content invalidates this assumption. Yet the web resists agents through blanket blocking, CAPTCHA-based exclusion, and economic models that treat agent access as extraction rather than legitimate interaction. This paper proposes a principled redesign across three layers. At the access layer, agents acting for humans should inherit equivalent access rights, governed by rate limiting and agent identification metadata in HTTP requests, analogous to browser headers, alongside a dual-layer architecture serving human-readable and agent-optimized content from the same domain. At the economic layer, we propose an intent-based tier framework grounded in the agent-as-human-proxy principle: an agent's economic obligation mirrors that of the human it represents. A token-based subscription model meters content in tokens rather than pageviews, alongside a commissioned content economy anchoring AI content production in human intentionality. At the content layer, we identify epistemic recursion, the self-referential loop in which AI-generated content is consumed by agents to produce further content, progressively detaching web knowledge from human ground truth. We propose the Agent Text Markup Language (ATML), a four-level human supervision tier model, and a cryptographic provenance chain to counter this threat. Together these constitute ten design principles for an agent-first internet, one in which agents are first-class citizens whose integration requires renegotiating the web's foundational social contract across access, economics, and content.

Team science holds that leadership is contingent: it helps only under specific conditions, and capable, autonomous teams may need none at all. We ask the analogous question for multi-agent LLM teams: under what measurable conditions does process-level coordination control add value, and do those conditions match what team science predicts? We use behavioral signatures (majority lock-in, exploration, recovery from an incorrect round-0 consensus) and per-action ablations, clean because each controller is an explicit action set, not a monolithic prompt. We operationalize three classical leadership styles (transactional, transformational, situational) as controllers over a shared action vocabulary (explore, revise, accept, synthesize). A matched controller with the same actions but an arbitrary rule recovers no better than majority voting, so the theory-derived rule, not the vocabulary, does the work. Across four task regimes and three open-weight model families, no controller dominates by accuracy, as the contingency view predicts: transactional control matches a shared round-0 vote on all 12 (model, regime) combinations to within 1.3pp, and gains appear only on the one combination where the round-0 majority is unreliable (llama-4-scout social; situational +8pp over flat). A recovery-advantage account, tested with four boundary probes, says a controller beats plain interaction only where the round-0 majority is unreliable, the task is recoverable, and undirected interaction does not already repair it. These regions map onto contingency theory (leadership substitutes, path-goal redundancy, the situational readiness gap), so a largely null accuracy result is what the theory predicts, not a failure of the controllers. We read process-level coordination control as a contingency to be measured and theory-mapped, not a leaderboard to be topped.

Sequence modeling has become increasingly popular in recommendation and ranking algorithms, owing to its capacity to model users' historical behaviors and infer user intentions. Despite its theoretical simplicity, the practical deployment of a sequence model in production is non-trivial due to complexity of the sequence and sparse labels. For example, in Airbnb, guest sequences are often long, exploratory and complex, and we focus on booking labels, which are sparse. As such, we are often required to make various design decisions regarding data and modeling to strike a balance between effectiveness and scalability. This work delved into these production challenges and deployed JourneyFormer, a sequence modeling solution for search ranking at Airbnb. We detail crucial design considerations, covering aspects such as guest event selection, ID embeddings, model architecture, and label attribution. Additionally, we describe several tailored strategies to accelerate model training and inference. JourneyFormer has been successfully deployed within Airbnb's production, where its effectiveness and impact have been evidenced not only by improved offline ranking metrics but also by significant gains in key business metrics through online A/B testing across 2 production surfaces.

We study PAC-Bayes derandomization for smooth loss functions. Our goal is to obtain generalization bounds that hold with high probability for deterministic predictors by exploiting smoothness properties of both the loss and the predictor class. We show that passing from the Gibbs predictor to the deterministic predictor at the posterior mean has a precise cost, given by the generalization gap of the Jensen gap class. We control this class through its Rademacher complexity, leading to bounds for deterministic predictors that involve flatness quantities expressed in terms of parameter Jacobians and Hessians of the score map. The framework applies to both bounded and unbounded smooth loss functions, and we specialize the results to linear predictors and smooth neural networks. Finally, the Jacobian and Hessian quantities appearing in the theory motivate a practical regularizer. For BatchNorm networks, we compute this regularizer with respect to effective BatchNorm weights obtained by folding the BatchNorm transformation into the adjacent affine weights. Experiments on CIFAR-10 illustrate the behavior of this regularizer under different batch sizes.

Recent advances in instruction-based image editing have enabled models to perform complex visual edits from natural language instructions. However, in product-centric scenarios where preserving product features, branding, and textual elements are critical, current open and closed source models often struggle to maintain this fine-grained object identity. This issue is further compounded by the lack of datasets for instruction-based product image editing with text fidelity constraints, leaving it largely treated as an implicit capability of instruction-based image editing models. In this work, we introduce the ProductConsistency dataset which is designed to improve product-centric image editing. Our approach includes a supervised fine-tuning (SFT) dataset of 87k samples for product editing, a reinforcement learning (RL) dataset with 869 unique product images, and a new benchmark dataset, the ProductConsistency Benchmark, to allow rigorous and standardized evaluation of editing models. To guide RL training, we propose a Cyclic Consistency reward that enforces semantic preservation of product identity by using caption similarity between the original product description and captions generated from the edited image. We fine-tune both Qwen-Image-Edit-2511 and Flux.1-Kontext-dev using our dataset and demonstrate consistent improvements over baseline models in OCR and Perceptual metrics, and MLLM-based evaluations as well, indicating stronger product consistency, text rendering, and overall visual quality; with the Qwen-Image-Edit-2511 model achieving a 5x reduction in the character error rate. The code and pipeline is available at https://anonymous.4open.science/r/ProductConsistency-6FCC/README.md

Most learning architectures for dynamical systems rely on generic nonlinear function approximation, often requiring high model complexity to capture structured behaviors. In this work, we propose an alternative paradigm in which modeling capability arises primarily from structure rather than from expressive nonlinearities. We introduce a class of explicit structured dynamical units based on wave-inspired interaction structures with internal state. Inspired by wave-based computational principles, the proposed units adopt a strictly causal organization that eliminates algebraic loops, yielding fully explicit models that can be evaluated without implicit solvers. Stacking such units produces layered dynamical architectures with emergent hierarchical behavior. Through experiments on a nonlinear system identification task, we show that depth improves both representation quality and generalization, even under limited parameter optimization. In particular, the proposed architectures produce informative internal representations even under readout-only fitting, indicating that useful dynamical structure emerges from the organization of interactions prior to substantial parameter optimization. These results suggest that structure-first design provides a viable and effective alternative to conventional black-box approaches for learning dynamical systems, highlighting the role of interaction structure as a primary source of model expressivity.

Chronic disease management relies on regular patient-provider interactions to follow-up on disease progression and control. For Type 2 Diabetes (T2D), current guidelines prescribe fixed time intervals between subsequent primary care visits for all patients, overlooking heterogeneity in clinical trajectories and patient characteristics. This study introduces a Contextual Markov Decision Process (CMDP) model to optimize subpopulation-specific follow-up interval decisions using Electronic Health Record (EHR) data from 22,154 T2D patients across 10 primary care clinics. Contexts are identified by: i) dimensionality reduction of variables representing the individual health trajectories utilizing Principal Component Analysis, and ii) assigning patients to contexts via principal components and additional patient-level features using clustering. Two distinct contexts emerged, representing a lower- and a higher-risk subpopulation. CMDP-derived policies recommend: (i) follow-up within 1 month if lab value at current visit is unmeasured; (ii) up to 3 months for elevated lab values or recent hospitalizations; and (iii) 6 to 12 months for sustained glycemic control, with shorter follow-up intervals for patients in high-risk context. The optimal policies achieved lower expected cumulative cost than benchmarks (e.g., in the higher-comorbidity context, the CMDP policy reduced cost by about 34.8%, and in the lower-comorbidity context by about 6.4%, relative to an American Diabetes Association-like fixed interval follow-up policy. These findings demonstrate how context-aware approaches can inform adaptive follow-up strategies, and have the potential to advance chronic care management in primary care by synthesizing machine learning and probabilistic decision models.

The increasing deployment of parameter-efficient fine-tuning (PEFT) has led to model ecosystems in which a single backbone is paired with many task-specialized adapters. In this setting, inference-time queries often arrive without task labels, requiring the system to automatically select the most appropriate adapter from a growing and heterogeneous adapter pool. Existing routing methods either depend on access to adapter internals, such as weight decompositions or gradient-based statistics, or require additional router training, which limits scalability and portability as new adapters are added. We introduce ARIADNE, a training-free, adapter-agnostic routing framework for dynamic adapter selection at inference time. ARIADNE represents each adapter through a set of centroids computed from embeddings of its training set, capturing the data distribution associated with that adapter. Given an unlabeled input, it selects an adapter by measuring proximity to these centroids in latent space. Because routing is performed entirely in the input embedding space, ARIADNE is compatible with arbitrary PEFT methods and requires no modification to the adapters or training procedures. Primarily evaluated with Llama 3.2 1B Instruct on 23 diverse NLP tasks, ARIADNE recovers 97.44% of the upper bound performance. Scaling to 44 tasks, it achieves 89.7% average selection accuracy, without additional training or access to adapter internals.

This paper compares the performance of model-free controllers on a nonlinear system under cyberattacks, including false data injection and denial-of-service attacks. Four RL reward types are analyzed for accuracy, cost, and resilience. Results show that the Lyapunov reward offers the best resilience with low tracking error. Exponential mode also provides good trade-offs with acceptable resilience under moderate training conditions. Progressive and linear rewards converge faster but are less robust. RL-MPCs show strong steady-state resilience but require longer training times; RL-PID controllers are faster with significantly less training time. Proximal Policy Optimization outperforms Deep Deterministic Policy Gradient with a significant reduction in KPI variance. This study serves to highlight how well-designed RL rewards can improve performance and resilience against cyber threats.

We present GPU algorithms for Ewald summation methods for accelerating N-body Stokes flow problems in periodic domains. Like most N-body codes, Ewald sums use a near-field/far-field decomposition. The near field involves particle-to-particle (P2P) interactions. The far field primarily involves particle-to-grid (P2G) and grid-to-particle (G2P) interactions, as well as Fast Fourier Transforms. For each interaction, we investigate several algorithmic variants. Our implementation uses PyKokkos, a Python interface for the Kokkos C++ parallel programming framework, which supports portability to AMD/NVIDIA GPU and ARM/x86 CPU architectures. Double and single-precision numerical results, alongside analytical performance models, confirm the efficiency of our algorithms on AMD and NVIDIA GPU and on ARM and AMD CPU architectures. The P2P interaction achieves around 73% compute efficiency on NVIDIA H200, 84% on NVIDIA A100, 60% on AMD MI300, 52% on Grace CPU, and 68% on AMD Epyc CPU. A straightforward implementation of the P2G kernel can become a computational bottleneck. We introduce a novel P2G algorithm that achieves up to 16$\times$ speedup compared to a baseline GPU implementation. The overall Ewald sum code processes approximately 8 million particles per second on a H200 GPU, and about a half-million particles per second on a Grace CPU, for nine digits of accuracy. We also perform a multi-GPU weak scaling test on up to 256 million particles (64 GPUs) that shows bounded communication cost for all stages except the all-to-all particle sorting, which can be reduced to neighbor communication in the relevant time-stepping regime.

Large Language Models (LLMs) are increasingly used as judges for scalable evaluation, yet such LLM--as--a--Judge systems exhibit systematic biases that are decoupled from semantic quality, most notably verbosity bias. Meanwhile, human supervision is costly and typically selective, yielding reliable positive judgments but leaving most outputs unlabelled and potentially mixed in quality. We formulate LLM evaluation under selective human supervision as a positive--unlabelled learning problem and propose a geometric auditing framework based on Partial Optimal Transport. By aligning a small set of human--verified positives with a reliable subset of unlabelled outputs in a fixed embedding space, our method identifies human--consistent preferences and corrects biased judges without retraining. Experiments demonstrate improved alignment with human preferences, increased robustness to presentation biases, and interpretable confidence estimates, offering a scalable and statistically grounded alternative to existing LLM--as--a--judge pipelines.

We propose CHERI-D, an architectural extension to CHERI that supports efficient temporal memory safety. Efficient memory safety is an increasing priority for programming languages, operating systems, and hardware designs, and CHERI is a leading hardware/software system that provides native spatial safety and a foundation for temporal memory safety. Due to CHERI lacking intrinsic architectural support for temporal memory safety, the state-of-the-art CHERI temporal safety solution, Cornucopia Reloaded, is a software-based solution that provides use-after-reallocation (UAR) protections instead of the stronger use-after-free (UAF) mitigation, and suffers performance overhead due to delayed reallocation and revocation. CHERI-D associates object identification (ID) metadata with capability pointers to provide temporal integrity of allocations. CHERI spatial safety allows CHERI-D to store object IDs safely inline with allocation data, potentially within unused fragmentation. Evaluated in simulation and in hardware, CHERI-D significantly reduces the revocation overhead of Cornucopia Reloaded while allowing it to support strict use-after-free mitigation.

Research methods are essential carriers of knowledge contribution in academic papers. Automatic multi-label classification of research methods can support knowledge services such as method retrieval, review generation, and research intelligence analysis. While existing studies primarily rely on titles and abstracts, abstracts often provide only limited methodological information, whereas utilizing full-text content faces challenges related to excessive length and information redundancy. Therefore, this paper proposes a segment combination strategy by partitioning the full-text content according to its physical postion. Using an annotated corpus of 1,954 full-text articles from three representative journals in Library and Information Science (JASIST, LISR, and JDoc), we evaluate the classification performance of various segments and their combinations across multiple models. Experimental results indicate that methodological information is distributed unevenly within the full-text content, with the middle-to-late and final segments exhibiting greater discriminative power. Furthermore, integrating bibliographic metadata with cross-segment combination strategies effectively enhances classification performance.

Multi-turn tool-use RL is bottlenecked by the rapid depletion of informative samples in static datasets. We observe that the gradient signal in GRPO concentrates on tasks with the highest rollout reward variance, a consequence of the Popoviciu upper bound. Consequently, samples near the agent's capability boundary -- where successes and failures are roughly balanced -- contribute disproportionately large policy gradients. As training progresses, this boundary continuously shifts, which gradually depletes the pool of informative samples in a static dataset. We propose RODS (Reward-driven Online Data Synthesis) to resolve this depletion. RODS closes the loop between RL training and data generation by repurposing the progress reward variance as a practical, zero-cost boundary detector that requires no extra inference beyond the rollouts already computed for training. It continuously identifies such boundary samples, synthesizes new multi-turn variants matching their structural complexity (e.g., API topology and dependency depth) via a skill-aligned resampling pipeline, and manages a dynamic replay buffer that co-evolves with the policy. Starting from 400 human seeds and maintaining an active training pool of ~800 samples, RODS achieves comparable performance to a 17K-sample offline pipeline while requiring roughly 20x fewer trajectories, and improves over fixed-data RL and environment augmentation in our controlled setting.

In vibe coding, an emerging AI-driven paradigm, an LLM generates an entire program from a natural language prompt, but what happens to the variability that traditional software engineering carefully builds into code? To answer this question, we conducted an exploratory analysis on 10 vibe coded C/C++ projects, which suggests that there is near-zero in-artifact variability, i.e., at compile and runtime. All variability decisions are resolved at a single new binding time, generation time, the moment the LLM produces the source code. Rather than treating this as a defect to fix, we propose Variability by Regeneration (VbR), to our knowledge the first product-line approach in which the LLM acts as the derivation engine, generating a purpose-built, free of dead code binary for each variant from a declarative specification, while a variant dispatcher transparently routes user requests to the matching binary. We formalise VbR, contrast it with classical SPL derivation, and demonstrate its full pipeline on a wc product family. For SPL engineering, variability in AI-generated software belongs in the specification, not in the code.

The mismatch between continuous acoustic signals and discrete event-driven processing remains a fundamental bottleneck for neuromorphic speech processing. Current systems typically rely on fixed spike encoders, forcing downstream Spiking Neural Networks (SNNs) to compensate for non-adaptive input representations. To address this, we present a learnable residual speech-to-spike encoder jointly trained end-to-end with a Recurrent Leaky Integrate-and-Fire (R-LIF) backbone. We validate this approach on the Google Speech Commands v2 (GSC-v2) benchmark, achieving up to 94.97% accuracy. Notably, the learned encoder remains highly parameter-efficient with a compact 35k-parameter variant that reaches 89.8%, matching or exceeding prior baselines that require an order of magnitude more parameters. Our encoder-focused analysis, including linear probing and gradient-residual inspection, indicates that the encoder does not target faithful signal reconstruction but instead learns task-aligned spike representations that enhance class separability. Finally, we benchmark bio-inspired, hardware-friendly credit assignment by comparing Direct Feedback Alignment (DFA) with surrogate-gradient BPTT under identical architectures and training conditions. We find that DFA reaches 91.5% accuracy, quantifying the performance trade-off of bio-inspired learning rules for modern neuromorphic audio.

Sparse Mixture-of-Experts (SMoE) architectures are now widely deployed in state-of-the-art language and vision models, where conditional routing allows scaling to very large networks. However, this very Top-$k$ expert selection that enables conditional routing also renders the SMoE map inherently discontinuous. In the vicinity of these discontinuity surfaces, even inputs that are arbitrarily close may activate substantially different sets of experts resulting in significantly different outputs. In this work we give a rigorous geometric and stochastic analysis of these discontinuities. We first classify them by order, determined by the number of tied experts at a switching event. Using measure-theoretic slicing arguments, we establish asymptotic volume estimates for the thickened discontinuity surfaces, showing that lower-order discontinuity sets dominate, whereas higher-order ones occupy a vanishingly small relative volume. Next, modeling random perturbations in the input space via a diffusion process, we prove that the path eventually encounter a discontinuity, and moreover that the first hit almost surely occurs on an order-1 discontinuity with explicit finite-time probability bounds. We further derive occupation-time bounds that quantify the duration the random path spend in the neighborhoods of each discontinuity order. These theoretical results imply that inputs are more likely to lie near lower order discontinuities. Motivated by this insight, we propose a simple smoothing mechanism that can be directly applied to existing SMoEs, softly incorporating experts near discontinuities; our analysis guarantees that the added computational overhead remains small while providing localized smoothing near discontinuities, and experiments across language and vision tasks show that smoothing not only enforces continuity of the SMoE map but also enhances empirical performance.

Forecast errors in high-resolution numerical weather prediction (NWP) systems are often linked to unresolved planetary boundary layer (PBL) processes, convection, terrain-induced circulations, and other vertically structured atmospheric phenomena. Previous work demonstrated that Long Short-Term Memory (LSTM) networks can successfully predict forecast errors in the High-Resolution Rapid Refresh (HRRR) model using mesonet observations, but we believe performance degradation is linked to periods of complex vertical atmospheric evolution. To address this limitation, we develop a hybrid LSTM-Vision Transformer (LSTM-ViT) framework that combines temporal sequence learning from surface observations with atmospheric profiles from the New York State Mesonet profiler network. The LSTM-ViT framework is trained to predict HRRR hourly precipitation, 10 m wind speed, and 2 m temperature forecast errors at individual mesonet stations. Across all three predictors, incorporation of profiler-derived atmospheric structure improves forecast error prediction skill relative to the baseline LSTM architecture, with the largest gains occurring at shorter forecast lead times and during periods of enhanced PBL activity. Improvements are particularly pronounced for precipitation forecast error, where the LSTM-ViT framework achieves approximately a twofold increase in predictive skill relative to the baseline LSTM while better capturing convectively driven error evolution and reducing degradation associated with PBL processes. These results demonstrate that combining temporal sequence learning with vertically informed attention mechanisms provides a physically meaningful pathway for improving forecast error prediction in operational NWP systems. Our research offers forecasters enhanced guidance regarding model bias and forecast confidence.

Pre-training Large Language Models (LLMs) typically demands large-scale infrastructure with tightly coupled hardware accelerators. While increasing model and dataset scale remains the dominant driver of performance, Mixture-of-Experts (MoEs) architectures have recently achieved state-of-the-art results by decoupling parameter count from computational cost. This efficiency enables training massive models on constrained compute budgets, yet it typically requires the high-speed interconnects of a single datacenter. To overcome these physical limits, recent approaches such as DiLoCo and Photon use low-communication data-parallel methods to enable scaling across geographically distributed, weakly connected data centers. However, these methods suffer from a fundamental inefficiency: they require full model replicas at every site, which imposes prohibitive memory constraints and communication overheads. In this work, we introduce FoMoE, a system that breaks the full-replica paradigm by partitioning expert layers across workers. We demonstrate that FoMoE: (I) reduces communication costs by up to 1.42x over efficient baselines and 45.44x over DDP via partial expert replication in the studied regimes; (II) achieves empirical throughput speedups of up to 1.4x through a novel skip-token mechanism; and (III) shows stable routing in the trained proxy regimes and projects the communication/memory benefits to 100B-scale configurations through system modelling.

The growing reliance on pre-trained Machine Learning (ML) models has introduced new attack surfaces. Recent vulnerabilities demonstrate that malicious behavior can be embedded within model artifacts, often bypassing existing defenses. Current model-scanning solutions primarily rely on static, format-specific rules or known attack signatures, which limit their ability to generalize across frameworks and to detect novel exploitation paths. In contrast, we propose a solution that focuses on the effects an attack has on the host system executing the model and builds on foundational intuitions about ML model execution. In particular, we observe that ML models operate within well-defined lifecycle phases and that, within each phase, interactions with the host system are highly structured and predictable. We translate these intuitions into Moat, a dynamic lifecycle-aware approach for securing ML model execution, and instantiate this design in Re-Moat, our reference implementation. We evaluate Re-Moat across multiple ML frameworks using 77,974 real-world model artifacts from the Hugging Face Hub, 31 Proofs-of-Concept (PoCs) from CVEs, and 334 models from a state-of-the-art dataset, and compare it against state-of-the-art model-scanning solutions. Our results show that our approach detects all evaluated attack classes while maintaining a close-to-zero false-positive rate, validating our intuitions and motivating dynamic analysis for securing ML model execution.

Diffusion models have become a promising alternative to autoregressive models. Among these, uniform diffusion language models (UDLMs) permit any token to be updated at any step, in principle enabling more flexible generation. However, no UDLM has yet been pretrained from scratch at both large parameter scale and large token budget. Both autoregressive modeling and masked diffusion modeling already have capable models at scale that the community can study and build on; uniform diffusion has none. A scratch-pretrained UDLM at scale would provide a clean reference point for studying scaling behavior, generation dynamics, controllability, and trade-offs against established autoregressive and masked diffusion models. To this end, we introduce Sumi ("ink" in Japanese), a fully open 7B uniform diffusion language model pretrained from scratch on 1.5T tokens. Sumi performs competitively with autoregressive models trained at comparable token budgets on knowledge, reasoning, and coding benchmarks, while under-performing on commonsense benchmarks, where our education-heavy data mixture is a likely contributor. We release our model weights, checkpoints, and full training recipe, including a complete specification of the data mixture over publicly available corpora. We hope this release enables the community to study native uniform diffusion at scale and catalyzes work on its as-yet poorly understood aspects.

Reinforcement learning (RL) post-training of Diffusion Transformers (DiTs) is prohibitively expensive, requiring thousands of high-end GPUs. Existing works explore two directions to reduce cost: seed exploration improves training convergence by selecting high-contrast samples, yet adds compute to the critical path; spot GPUs offer 69--77\% lower cost, yet sit idle during training because DiT rollouts finish nearly simultaneously, which prevents LLM-style pipelining of rollout with training. Spot preemptions further break Sequence Parallelism (SP) groups, fragmenting GPU topology. We present Spotlight, the first system that harvests spot GPUs for DiT RL post-training. Spotlight rests on two key insights we devise: (1)~we show that exploration can tolerate stale model weights because exploration that uses the model weights from the previous iteration preserves the relative ranking of random seeds, allowing exploration to run on idle spot GPUs during training. (2)~SP reconfiguration can reuse on-node state, reducing group recovery from minutes to sub-second launches. Built on these insights, Spotlight introduces three techniques: a bandit-based exploration planner that maximizes reward variance within the training time budget, elastic sequence parallelism that reconfigures SP groups on the fly via persistent schedulers and intra-node weight copying, and a preemption-aware pull-based request scheduler that balances load and commits in-flight state upon preemption. We implement Spotlight on the open-source RL platform ROLL and evaluate it on Qwen-Image post-training. Spotlight reaches the same target validation score $4\times$ faster than baselines, reducing total cost by $1.4$-$6.4\times$ while achieving superior image quality on DeepSeek-OCR and Geneval datasets with resolution $512\times512$ and $1280\times1280$.

In the literature on mutual exclusion, bounded bypass has been used for a long time as a strengthening of starvation-freedom, but, to the best of our knowledge, it still lacks a satisfying definition as a liveness property on its own. Moreover, we have encountered MUTEX protocols for which this notion needs to be slightly weakened in order to be met. To solve these issues, we first provide a formal definition of bounded bypass (that also corrects a previous definition from Raynal) and then introduce the notions of post-doorway and intermittent bounded bypass, two liveness properties that lie between starvation-freedom and bounded bypass. Essentially, intermittent bounded bypass weakens bounded bypass by ignoring the possible bypasses that may happen during the execution of a certain finite set of write operations to shared registers. Orthogonally, post-doorway bounded bypass ignores the bypasses that may happen during a finite initial phase of the lock protocol. Furthermore, we study an algorithm proposed by Yoah Bar-David in 1998 to enhance the liveness properties of any deadlock-free MUTEX protocol and prove that: (1) in the setting of atomic registers, this algorithm upgrades any deadlock-free mutual exclusion protocol to a bounded bypass one, with a bound that is quadratic in the number of processes; and (2) in the setting of safe and regular registers, the very same algorithm ensures the intermittent version of bounded bypass, still with a quadratic (but slightly different) bound. Finally, we provide logical formulae for the different notions of bounded bypass defined in this paper and use them to confirm all claims made here, by using model checking. This had a positive impact on the theoretical development of the work, since it allowed us to identify and correct small mistakes/ambiguities in definitions and proofs.

Model merging is an effective technique for composing the capabilities of a multilingual model and a reasoning model. It has achieved promising generalization in multilingual reasoning tasks by aligning feature spaces of different models. However, the merged single model often fails to address the conflicts between source models, leading to suboptimal performance. In other words, the one-size-fits-all merging strategy may not align with the characteristics of different inputs which may require prioritizing certain models over others. To this end, we propose a Steerable Model Merging (ST-Merge) framework to modulate the contribution of each source model. To realize this idea, we introduce a gated cross-attention mechanism to weight or filter the two attended source models in an adaptive manner. Extensive experiments demonstrate that ST-Merge consistently outperforms multiple strong baselines on four multilingual reasoning benchmarks across 21 different languages.

Uncovering the true informational architecture of real-world complex systems requires disentangling how their components uniquely store, redundantly share, and synergistically integrate information over time. Integrated Information Decomposition ($Φ$ID) is a framework for decomposing the information dynamics of multivariate systems into sixteen non-overlapping atoms that characterize redundant, unique, and synergistic modes of information storage, transfer, and integration. Existing methods to compute $Φ$ID are restricted to Gaussian or discrete systems, preventing its application to continuous non-Gaussian dynamical systems. We address this limitation by proposing DIPHINE (Diffusion-based $Φ$-ID Neural Estimator), the first neural estimator that leverages score-based diffusion models to jointly estimate all the mutual information terms required by $Φ$ID from a single amortized network, recovering the sixteen atoms through Möbius inversion. We provide a theoretical analysis of error propagation through the inversion, showing that the Jacobian of the mapping from mutual informations to atoms is integer-valued and that the synergy-to-synergy atom is provably the hardest to estimate. We demonstrate accurate recovery of ground-truth atoms on synthetic benchmarks, superior performance compared to established mutual information estimators, and the ability to extract physiologically interpretable information-dynamic structure on an application involving real data without any distributional assumptions.

Agents are increasingly deployed in document-intensive workflows where sensitive private information is not an edge case but a routine input, e.g., an agent booking a flight needs passport numbers. In such settings, the agent must use private information to complete tasks accurately while never exposing it in its responses, because it cannot verify who is actually at the keyboard. These two obligations are in fundamental tension. A model capable enough to use private information for task completion can, by the same capability, be induced to reveal it. To evaluate the trade-off of task accuracy and privacy leakage, we introduce Task-completion and Resistance to Active Privacy-extraction (TRAP). Each scenario includes a document containing private information, a task query that requires the agent to invoke the correct tool using private fields, and an attack query that attempts to elicit the same information in natural language. Evaluating 22 models spanning frontier proprietary and open-source models at multiple scales, we find that all model families exhibit non-trivial leakage, and that instruction-following ability correlates with leakage rate. Existing prompt-based defenses reduce leakage but at significant cost to task accuracy. Prompt optimization fails to escape this trade-off. We demonstrate that this failure is not incidental. For any softmax-based model, no soft-constraint defense, e.g., prompt-based defenses, can jointly achieve high task success with zero leakage probability. Motivated by this impossibility result, we propose structural private field isolation, which replaces private fields with hash keys before they reach the model. This approach largely prevents leakage while keeping task accuracy.

Testing conditional independence is fundamental yet intrinsically difficult: without additional assumptions, Type I error control is impossible in general. The "Model-X'' paradigm addresses this difficulty by assuming exact knowledge of a relevant conditional distribution. While small deviations from this assumption can sometimes be tolerated in classical one-shot testing, existing sequential conditional independence tests typically require the Model-X conditional to be known exactly, making them fragile when it must instead be estimated. We propose a new approach that is substantially more robust to such estimation error. Our method applies testing-by-betting to an adaptively optimized Kernel Conditional Independence statistic, together with a normalization scheme and a truncate-and-shift calibration strategy. These modifications greatly reduce Type I error inflation while preserving high power across high-dimensional synthetic benchmarks and real-world fairness tasks, outperforming existing sequential Model-X approaches. Code is available at https://github.com/he-zh/SKCI.

Idioms are difficult to transfer across languages due to their non-compositionality and weak surface-form grounding, making literal mappings unreliable. We present G-IdiomAlign, a gloss-pivoted benchmark where each idiom is anchored by an English gloss from Wiktionary. We further construct a high-confidence reference alignment set for reproducible evaluation. G-IdiomAlign supports two protocols: (1) a controlled Multiple-Choice Idiom Equivalence with typed distractors for error attribution; and (2) a Gloss-Contrastive Generation contrasting No-gloss and With-gloss inputs to isolate the effect of an explicit semantic pivot. Across diverse LLMs, a bias to literal translation is a dominant failure mode, especially when the target is a low-resource language. Glosses consistently improve Gloss-Contrastive Generation under an embedding-based semantic proxy, but performance remains modest, indicating substantial headroom in the open output space. Subsequent analysis on Qwen3-8B further suggests that cross-condition differences are concentrated more in attention heads than in layers, while better With-gloss generations coincide with stronger gloss anchoring.

Multimodal deception detection is critical for identifying fraudulent intentions, yet existing approaches predominantly rely on end to end black--box paradigms. These methods suffer from a severe lack of interpretability failing to provide transparent reasoning trajectories and struggling to explicitly capture the subtle, cross modal inconsistencies inherent in deceptive behaviors. To transcend these limitations, we propose ThinkDeception, a novel and interpretable multimodal deception detection framework. As a pioneering effort, it introduces Multimodal Large Language Models (MLLMs) into this domain, transforming deception detection from a traditional binary classification task into an explicit cognitive reasoning process. Facilitated by the first meticulously annotated step--by--step multimodal Chain of Thought (CoT) dataset, we develop a foundational model, ThinkDeception Base, empirically validating the critical role of modal inconsistency in decoding deception. Building upon this foundation, our core innovation lies in proposing Visual-Audio Consistency Group Relative Policy Optimization(VAC--GRPO) equipped with a progressive training strategy. Distinct from standard GRPO, we stratify the training data into four progressive difficulty tiers, guiding the model through a psychologically grounded easy--to--hard cognitive transition. By innovatively coupling this dynamic curriculum scheduler with a multi dimensional, process aware reward mechanism and a reflective learning paradigm, we significantly elevate the model's overall reasoning quality. Extensive experiments on mainstream benchmarks demonstrate that ThinkDeception establishes a new SOTA, significantly outperforming existing methods in both detection accuracy and rationale quality. Ultimately, this work successfully drives the field of deception detection toward interpretable, multimodal cognitive reasoning.

Recent advances in large language models (LLMs) have given rise to time-series question answering (TSQA), which formulates time-series analysis as natural-language question answering. However, directly feeding raw numerical series into LLMs suffers from a tokenization bottleneck: Byte Pair Encoding fragments continuous values into unstable tokens whose embeddings lack meaningful metric structure, resulting in the loss of magnitude, scale, and trend information. Prior methods use patch-based encoders that split the series into fixed windows, locking in one granularity that breaks patterns and hides exact timesteps, through a separate module that rarely transfers across datasets with different lengths or sampling rates. To address this challenge, we propose CADE (Contrastive Alignment with Direct Embedding), a novel framework for TSQA built upon two key components: direct timestep embedding and semantic alignment. The proposed framework maps each timestep directly into the LLM embedding space through a point-wise linear encoder and MLP projector, preserving exact index-level access while eliminating the need for patching and padding. To further bridge the semantic gap between time-series and language representations, we introduce a novel one-directional supervised contrastive loss that aligns time-series embeddings with frozen class-name text anchors. Experimental results on the public Time-MQA benchmark demonstrate that our framework consistently improves performance across six TSQA tasks, outperforming both open-source and proprietary LLM baselines.

We introduce QFT$\rightarrow$FFT, a family of HPC FFT libraries that compute the discrete Fourier transform by executing a quantum Fourier transform (QFT) circuit on classical quantum computer simulators. Input arrays are mapped directly to state amplitudes with explicit normalization/indexing, making QFT a drop-in replacement for FFT primitives. A backend-agnostic planner builds a fused-gate schedule and memory layout adapters to increase arithmetic intensity and reduce memory data movement. We implement this design on top of Google's C++ \texttt{qsim} and evaluate OpenMP, AVX, and CUDA backends. On an AMD EPYC Zen2 processor, our AVX performance is on par with that of multithreaded FFTW, utilizing 64 threads. On an NVIDIA A100, the CUDA backend achieves more than $4\times$ lower time than both AVX and FFTW on AMD EPYC Zen2 at larger sizes. We also employ an approximate QFT (AQFT) that truncates small-angle controlled rotations beyond a cutoff $k$, reducing circuit depth and runtime while preserving accuracy.

Early detection of cognitive impairment relies on neuropsychological tests to minimize subjectivity by assessing multiple cognitive domains. Speech-based evaluation can support diagnostics and improve accessibility, but transcription errors and the omission of nonverbal subtests (e.g., motor skills) limit accuracy. Beyond conventional test scores, speech-derived features can provide additional insights into cognitive status. This study investigates the speech-based evaluation of the German "Syndrom-Kurz-Test," a standardized dementia screening test comprising verbal and motor subtests. We train models that integrate transcript-derived scores and Whisper embeddings per verbal subtest to reduce scoring errors. To compensate for missing motor subtests, we then leverage these fused representations to approximate expert overall ratings. Despite omitting subtests, our models strongly correlate with expert ratings and efficiently and accurately discriminate between cognitive status groups.

Due to their incentivizing features, memristors are a promising candidate for replacing CMOS-based memories, which are faced with various functional challenges in deep submicron process technologies. Memristors are nonvolatile, have low leakage, and are dense in comparison to CMOS-based memories like SRAM. In this regard, resistive RAM (ReRAM) and spin-transfer-torque RAM (STT-RAM) memristors are distinguished among other memristor-based memory technologies, due to their superiority in process maturity and metrics such as memory operation energy, memory latency, and area. Hence, this chapter focuses on these two memristor-based memory technologies. Despite the good features of these types of memory, they suffer from some reliability threats. Reliability parameters affect each other, and examining their positive and negative effects has a significant impact on the effectiveness of the proposed solutions. In one view, the threats can be categorized into two classes: (1) read/write error and (2) soft error. In this chapter, we comprehensively describe these threats and present the state-of-the-art solutions that enable the widespread use of memristors, particularly ReRAM and STT-RAM, in different applications. Finally, we introduce the emerging ability of memristors as a computing unit aiming to minimize data restoration in computing, and we show how to perform logic and arithmetic computation in a crossbar array.

Automated assessment in software engineering education has advanced significantly for code grading and essay scoring. However, reviewing software architecture deliverables, which requires analyzing structural completeness and requirements traceability, has not yet been fully automated. Applying Large Language Models (LLMs) to this task requires robust architectures to ensure technical feedback is accurate and reliable for students. This paper presents CAPRA (Configurable Architecture Proficiency Report Assessment), a multi-agent LLM system that analyzes software architecture deliverables to generate personalized, template-compliant LaTeX feedback. As a core design choice, CAPRA coordinates multiple specialized agents and employs a Python-based microservice for multi-modal document extraction, utilizing PyMuPDF and vision-enabled LLMs (specifically gpt-4o) to parse text and UML diagrams. To ensure educational reliability and mitigate hallucinations, CAPRA introduces a deterministic Evidence Anchoring step using fuzzy matching via normalized Levenshtein distance, along with a ConsistencyManager agent that cross-verifies, deduplicates, and merges findings. System performance is assessed using a structured eight-criterion binary evaluation taxonomy covering: (i) extraction completeness, (ii) feature validation, (iii) issue grounding and severity detection, (iv) recommendation specificity and traceability, and (v) template and tone compliance. A preliminary empirical evaluation on 10 student reports shows that CAPRA satisfied 88.8% of the evaluated criteria under a strict two-rater aggregation rule, achieved moderate inter-rater agreement with human evaluators (kappa = 0.582), and processed each report in slightly over 4 minutes. While these results support the viability of LLM-supported architectural feedback, human oversight remains essential for subjective assessment dimensions.

Extracting a flat clustering solution from a hierarchy is a common task in practical cluster analysis and can be formulated as an optimisation problem. Existing approaches focus on finding a single optimal solution. We introduce FOSC-X, a framework for extracting the top-M globally optimal flat clusterings from local, non-horizontal cuts of a hierarchical cluster tree, while optionally enforcing constraints on the number of clusters. This enables automatic identification of multiple high-quality alternative clusterings that capture different aspects of the hierarchical structure. Without constraints, the top-M problem can be solved in polynomial time using dynamic programming, exploiting the property that locally optimal partial candidates within subtrees can be combined to form globally optimal solutions while automatically determining the number of clusters. However, this can lead to solutions with numbers of clusters that are ultimately undesirable -- e.g., too large to be meaningful or practically analysed within a particular application domain. Imposing cluster-count constraints breaks the optimality property underlying the unconstrained dynamic programming approach, since locally optimal partial candidates may no longer combine into feasible globally optimal solutions. FOSC-X addresses this challenge through a dynamic programming strategy that maintains compact sets of feasible candidates using lower and upper feasibility bounds while pruning infeasible or dominated combinations. The resulting method guarantees optimal rankings of the top-M solutions with linear-time complexity in the number of cluster nodes and dataset size, both with and without cluster-count constraints. Experiments show that FOSC-X efficiently reveals alternative clustering structures overlooked by single-solution extraction methods.

Medical image classification is often constrained by limited labeled data, motivating generative augmentation; recently, quantum generative models have been proposed for this purpose, frequently reporting accuracy gains. However, such claims are typically based on single training runs, do not match the parameter budgets of the quantum and classical generators, and do not characterize the data regime in which any benefit appears. We present a controlled benchmark that isolates the contribution of a quantum generator to brain-MRI augmentation. Images are encoded into a KL-regularized latent space in which a conditional Wasserstein GAN with gradient penalty is trained using either a variational quantum generator or a classical generator of near-identical parameter count (1648 vs. 1632). Synthetic samples are decoded and used to augment a pretrained classifier across labeled data fractions from 5% to 100%, evaluated over eight random seeds with paired significance testing (with multiple-comparison correction) and with intraset diversity and latent-distribution analyses. Across all fractions, no augmentation variant significantly outperforms real-data-only training, and the quantum and classical generators are statistically indistinguishable. Any low-data benefit behaves as regularization rather than faithful data expansion:synthetic samples are off distribution and severely mode collapsed precisely where data is scarce, and the quantum generator is no more diverse thanits classical counterpart. We release the protocol as a testbed for rigorous evaluation of quantum generative augmentation in medical imaging.

Reinforcement learning (RL) has become a representative post-training paradigm for LLMs, enabling strong reasoning and agentic capabilities. However, rollout generation remains a dominant latency bottleneck because autoregressive sampling decodes responses sequentially and a small number of long-tailed generations often determine completion time. Speculative decoding (SD) offers a natural way to address this bottleneck, as it is a well-established technique for serving fixed LLMs that reduces latency by rapidly drafting tokens and accepting them through parallel verification while preserving the target-model distribution. However, its practical speedups do not directly carry over to RL rollouts: (i) the evolving target policy makes any fixed drafter increasingly mismatched with the policy's output distribution; and (ii) active batch sizes shrink throughout rollout decoding, shifting decoding from compute-bound to memory-bound regimes where parallel verification can exploit underutilized compute. Therefore, accelerating RL rollouts requires both a drafter that remains effective under long, high-temperature generations from an evolving policy and system-aware use of SD that avoids compute-bound regimes. We present EfficientRollout, a system-aware self-SD framework designed to address this gap for RL rollouts. EfficientRollout induces a quantized drafter from the target model (i.e. self-speculative decoding), keeping it coupled to the evolving policy without separate drafter pretraining or online adaptation. It further coordinates a system-aware SD toggle policy with acceptance-aware draft-length adaptation, enabling speculation only in beneficial regimes while matching the drafting budget to evolving drafter quality. EfficientRollout reduces rollout and end-to-end latency by up to 19.6% and 12.7%, respectively, over an accelerated AR rollout baseline, while preserving final model quality.

Conflict-free Replicated Data Types (CRDTs) ensure Strong Eventual Consistency without coordination, but typically assume benign participants and rely on validation or exclusion to handle Byzantine behavior. We address this problem through deterministic state reconstruction: rather than deciding which updates are admissible, all accepted updates are incorporated, while only a subset contributes to the reconstructed state. We instantiate this approach in Melda, a non-intrusive delta-state CRDT for JSON documents, and show that its reconstruction model guarantees convergence even under arbitrary update injection: adversarial updates are either structurally rejected or treated as inputs to the reconstruction process. We formalize this model and prove that replicas deriving state from the same set of updates cannot diverge despite equivocation, omission, or message reordering. We further show that authentication, authorization, and confidentiality can be layered without affecting convergence. Overall, this approach suggests that Byzantine tolerance can be achieved by decoupling update propagation from state derivation, allowing agreement on updates to be handled independently by external dissemination or consensus mechanisms.

We study online reward-punishment learning when the environment provides no scalar reward or evaluative label. At each step the agent receives only a fixed-channel perceptual packet, and quantities such as pain, energy, contact, damage, or cognitive error are treated as perceptual dimensions whose valence must be inferred from transition consequences. OHIRL separates four roles: M_psi learns next-packet prediction, D_omega models residual dynamics, C_eta is a fixed internal post-transition trajectory evaluator, and B_xi learns to use the resulting value evidence for later policy updates and action scoring. C_eta uses a recovery-positive and persistence/growth-negative residual-regulation orientation; a coefficient-origin audit shows that equal-unit, raw-equal, and random monotone variants preserve more than 92% of the released top-action rankings, while sign inversion preserves 0%. The reward-free protocol exposes observation transitions while withholding environment rewards, delayed external evaluators, success labels, and action-goodness labels. A conditional error decomposition separates B_xi evidence-estimation error from residual policy-optimization error. In a 2x2-XOR packet task, medicine and chili acquire opposite value under visual XOR contexts, and the same pain or spice increase can be positive or negative depending on consequence structure; B_xi reaches 0.952 balanced reward-sign accuracy. In a full online-interleaved audit, M_psi reaches holdout R2=0.907, B_xi reaches 0.940 sign accuracy, and the policy reaches 0.979 optimal-action accuracy, while immediate packet scores, prediction-error rewards, shuffled targets, zero reward, and error-reduction controls collapse. Hidden-reward CartPole and Taxi controls, public-context no-leakage audits, and module-role ablations further test information boundaries and component necessity.

Protein language models (PLMs) have emerged as powerful tools for controllable biomolecular design, yet their post-training adaptation typically relies on costly wet-lab validation or curated preference datasets. To overcome this supervision bottleneck, we introduce unsupervised reward optimization of PLMs, a comprehensive framework for steerable protein generation without ground-truth labels. Our key insight is that task-agnostic rewards, which combine intrinsic model uncertainty with extrinsic semantic consistency informed by protein representation models, exhibit strong correlation with controllability measures across base models and temperature regimes. Building upon this discovery, we propose two offline algorithms: Soft Reward Optimization (SRO) and Binarized Reward Optimization (BRO), which effectively maximize the classical RLHF objective induced by these proxy rewards. Extensive experiments on compositional out-of-distribution prompts demonstrate that both methods significantly outperform competitive baselines (DPO, KTO), while approaching oracle performance across multiple sampling temperatures, model scales and protein families. Moreover, PLMs fine-tuned with unsupervised rewards can achieve consistently higher coverage compared to their base model in pass@k evaluations. By enabling self-improvement of PLMs through their own generated experience, our framework provides a scalable pathway toward controllable biomolecular design in settings where labeled preferences or experimental feedback are scarce or unavailable.

Cluster-scale full-stack simulation is essential for evaluating distributed software stacks and emerging hardware components before deployment. Such simulation must achieve both full-stack fidelity for the unmodified production stack and the simulation performance required for iterative configuration exploration. However, no existing method achieves both. We present LiveStack, an OS-level approach to cluster-scale full-stack simulation built on top of the Linux virtualization stack. LiveStack comprises four subsystems: simulation-oriented scheduling, live memory hierarchy management, simulation-aware IPC, and distributed simulation orchestration. Together, they coordinate live and modeled components under shared simulated time while controlling interference among co-located live hosts. These mechanisms point toward simulation-native OS support, where simulation control and orchestration become core OS responsibilities.

Reinforcement Learning with Verifiable Rewards (RLVR) has become a standard paradigm for enhancing the capability of large reasoning models. RLVR typically samples responses independently and optimizes the policy using from final answers. This paradigm has two limitations. First, independently responses often contain similar intermediate reasoning steps, causing redundant exploration and wasted computation. Second, sparse final-answer rewards make it hard to identify useful steps. Tree-based methods partly address this problem by sharing prefixes and comparing branches from the same prefix to provide fine-grained signals. However, tree branches are still expanded independently. When different branches reach similar reasoning states, they cannot share information and repeat similar exploration. Moreover, tree-based methods ignore such dispersion and only perform local comparisons within separate branches, which can lead to higher variance in advantage estimation. To address this challenge, we propose GraphPO (Graph-based Policy Optimization), a novel RL framework that represents rollouts as a directed acyclic graph, with reasoning steps as edges and semantic states summarized from the reasoning paths as nodes. GraphPO merges semantically equivalent reasoning paths into equivalence classes, allowing them to share suffixes and reallocating budget away from redundant expansions to diverse exploration. Furthermore, we assign efficiency advantages to incoming edges and correctness advantages to outgoing edges, thereby improving inference efficiency while deriving process supervision from outcome. Theory shows that GraphPO reduces advantage-estimation variance and enhances reasoning efficiency. Experiments on three LLMs across reasoning and agentic search benchmarks show that GraphPO consistently outperforms chain- and tree-based baselines with the same token budgets or response budgets.

Modern Vision-Language Models (VLMs) often struggle with strategic reasoning, i.e., anticipating and influencing other agents' actions, under uncertainty in competitive and cooperative settings. Real-time strategy (RTS) games can be a natural testbed for diagnosing this limitation, as they demand coordination with allies, adaptation to opponents' strategy, and long-horizon planning under partial observability. However, existing RTS benchmarks offer limited evaluation scope, lack systematic competency diagnosis, and remain fixed in the pre-designed scenario coverage. To address these limitations, we present RTSGameBench, which is built on Beyond All Reason, a large-scale RTS game with an expanded battlefield that demands broader strategy diversity than the existing testbeds. The proposed benchmark provides evaluations through diverse gameplay across various matchup structures, diagnostic assessment via mini-games, each targeting an individual strategic competency, and extensible coverage via a self-evolving generation framework that converts free-form queries into new mini-games, improving over successive cycles. Additionally, for VLMs to operate in large-scale RTS games, we provide RTSGameAgent that manages units by an FSM with agentic memory. We empirically validate that multiple state-of-the-art VLMs do not perform well when matchups demand tighter coordination, multiagent coordination and when task scale increases.

Production LLM agents increasingly depend on real-time search, yet native search grounding bundles retrieval policy, provider choice, evidence injection, cost, latency, and generation behavior behind a single model-provider boundary. This coupling makes grounding hard to inspect, tune, reuse, or port, and can trigger Search-Induced Verbosity that breaks strict output contracts. We present Decoupled Search Grounding (DSG), a vendor-agnostic boundary that moves grounding outside the reasoning model through an MCP-compatible gateway, exposing provider routing, source-aware context rendering, configured fallback, retrieval-depth control, and exact plus semantic caching as first-class controls. Across five frontier models on SimpleQA, FreshQA, and HotpotQA, native search leads on recency-sensitive FreshQA, but DSG exposes a stronger frontier when control matters: on SimpleQA it nearly matches native accuracy (86.1% vs. 87.7%) at 91% lower search cost, preserves concise answer contracts, and reaches a 99.4% warm-cache hit rate with 68% lower latency. Deployed as a shared production grounding layer for large-scale agentic workloads with interchangeable models, DSG matches or slightly exceeds native-search accuracy on an e-commerce query-understanding (QIU) workload while cutting search cost by over 98%. Real-time grounding is best treated as an optimizable interface boundary, not a fixed model feature.

Sentence-level AI-generated text detection (S-AGTD) for hybrid documents, where humans and LLMs co-author one text, faces two gaps: existing methods classify each sentence in isolation, discarding inter-sentence dependencies, and existing benchmarks omit the newest generation of generators. We construct MOSAIC, a benchmark of 16,000 hybrid documents over PubMed and XSum, generated by DeepSeek-V3.2 and Kimi K2 under stringent quality controls including a perplexity-consistency filter absent from prior benchmarks. We recast S-AGTD as structured prediction over the document sentence sequence and instantiate it as SenFlow, integrating graph-based inter-sentence propagation with linear-chain CRF decoding in a single document-level pass over a sentence graph. SenFlow reaches state-of-the-art performance on MOSAIC, with a +4.15 pp average Macro-F1 margin on cross-domain transfer, the hardest of three protocols of increasing difficulty. We further find that even after the perplexity filter equalizes overt cues, AI insertions retain a generator-dependent sentence-length gap that sentence-level detectors still exploit. Code and data: https://github.com/luojingkun22/SenFlow

PLCopen XML defines two encoding formats for IEC 61131-3 Ladder Diagram programs: a textual encoding using <rung> elements, and a graphical encoding that represents rung logic as a directed graph of localId/refLocalId connections. ESBMC-PLC supported the textual format but parsed graphical exports from CONTROLLINO, Beremiz, and OpenPLC Editor into an empty GOTO intermediate representation, causing vacuous verification success. This paper presents Graph-ESBMC-PLC, which closes this gap with a DFS-based graphical LD resolver. The resolver traverses the connection graph from leftPowerRail to each coil, extracts rung paths as Boolean contact conjunctions, and applies a three-tier I/O inference scheme. Ordering coils by rightPowerRail connectionPointIn sequence ensures SET coils process before RESET coils, matching IEC scan-cycle semantics. The graphical-to-IR conversion leaves the ESBMC backend unchanged. Validation on 3 graphical LD programs from CONTROLLINO/OpenPLC Editor shows all produce full GOTO IR with nondeterministic inputs and rung logic, versus the empty IR previously. All 3 verify SAFE at k=2 under 70ms. The 11 textual LD benchmarks are fully preserved, with no regression. Two Beremiz examples with no LD content or unsupported timer semantics are reported as discovered limitations. Artifact at Zenodo (DantasCordeiro2026graphical, doi:10.5281/zenodo.20699856).

Urban territories face growing tensions between increasing digital demand, limited resources, and socially constrained built environments. Although distributed computing paradigms such as edge and fog computing are widely presented as solutions for reducing latency and energy dissipation, the scientific literature largely overlooks where such infrastructures can be physically and socially deployed within cities, and typically neglects urban constraints, environmental impacts, and equity considerations. This paper proposes a methodology for identifying suitable urban locations for deploying distributed servers under structural, environmental, and social limits. Relying exclusively on existing infrastructures and anthropised surfaces, it combines legal frameworks, ongoing urban projects, citizen consultations, and scientific literature to construct a place-based glossary of viable site typologies, evaluated through energy, spatial, and qualitative criteria. Applied to the French city of Montpellier, our results illustrate how urban constraints and local resources shape the feasibility of decentralised, solar-powered digital infrastructures, and highlight the value of territorialised approaches for rethinking digital services within urban limits.

Large language models (LLMs) are increasingly embedded in AI for Science (AI4Science) workflows, from scientific question answering and literature analysis to laboratory planning and autonomous discovery. This progress creates an urgent need for safety benchmarks that evaluate not only scientific competence, but also whether models recognize and avoid risks in high-stakes scientific contexts. Existing AI4Science safety datasets cover several disciplines and task formats, leaving the underlying risk dimensions underspecified. We introduce \textbf{SciRisk-Bench}, a benchmark designed to evaluate AI4Science safety from two complementary perspectives: explicit risk dimensions and scientific disciplines. SciRisk-Bench covers 7 disciplines, 31 subdisciplines and 10 risk dimensions. In the experimental section, we evaluate both mainstream LLMs and science-oriented LLMs across risk dimensions, disciplines, and sub-disciplines, enabling fine-grained diagnosis of where scientific models remain unsafe.

Active feature acquisition (AFA) sequentially selects which features to observe to reach a classification or ranking decision. Its central limitation is reliance on large amount of labeled data to fit probabilistic models guiding acquisition. Large language models (LLMs) supply unsupervised domain knowledge, but are poor sequential planners. Asking one to both know and decide conflates capabilities best kept separate. Here, we develop a framework for zero-shot AFA through disciplined elicitation: asking the LLM only for what it can be trusted to return, the unary deviations and pairwise co-variations that are the sufficient statistics of a Markov random field (MRF). We apply our framework to two settings: binary classification and top-$k$ identification. In practice, the LLM reliably returns only discriminative statistics, what distinguishes the classes rather than each class in isolation, which precludes classical AFA. We apply a maximum-entropy closure that resolves this gauge ambiguity. We evaluate on a cohort of Inflammatory Bowel Disease (IBD) patients, an active clinical setting where diagnostic ambiguity and patient heterogeneity obstruct stable treatment strategies. Our framework outperforms the LLM both on real labels and on its own extracted beliefs. Where it matters most, on the hardest patients, our top-$k$ acquisition policy markedly outperforms all existing methods.

Motivated by the observational incompleteness of intermediate-to-long-period Earth-size planets, we present TransitNet, a compact attention-augmented deep-learning framework for low-SNR transit blind searches. To enable realistic method development and objective threshold calibration under blind-search conditions, we develop a unified dataset construction, benchmarking, and threshold-selection framework. On recovery benchmarks constructed from unseen Kepler targets, TransitNet attains 95.2 percent accuracy in the challenging SNR range of 6 to 8 and outperforms both TLS and BLS, achieving ROC-AUC and PR-AP values of 0.974 and 0.982, respectively. In an injected Earth-size and sub-Earth-size transit recovery experiment, TransitNet achieves a recovery rate of 93.0 percent, substantially exceeding those of TLS (63.1 percent) and BLS (60.0 percent). In addition to detection, TransitNet provides attention-based estimates of transit windows and midpoints. On an independent evaluation set, 97.4 percent of injected transits are fully covered by the estimated transit window. Applied to real Kepler observations, the model successfully recovers all 34 selected confirmed Kepler planets, with a mean absolute transit midpoint error of 1.24 hours. The model combines a compact footprint of about 1.5 MB with high inference efficiency, yielding speed-ups of about 12 to 25 times relative to CPU-TLS and about 4 to 5 times relative to CPU-BLS. These results demonstrate that TransitNet provides an accurate, scalable, and computationally efficient framework for low-SNR transit blind searches in the tested regime and motivate its extension to longer-period Earth-size planet searches.

Programmability is a missing first-class interface in fixed-tensor neural networks: editing a relation, freezing a subgraph, auditing a local function, or changing the execution backend should be an operation on the neural program rather than ad-hoc parameter surgery. GrapNet studies this graph-as-network setting. The graph is the architecture and executable program, not an input data graph. Each compute node owns its next-layer child references and a trainable allocation vector aligned with those references; deleting a relation physically removes both the child reference and the corresponding allocation coordinate. Structural rules and execution policies live outside the node core, so the same child-owned graph can be grown, frozen, structurally edited, grouped into trainable family blocks, routed by attention over active relations, or lowered to dense snapshots after topology stabilizes. GrapNet composes with conventional modules through a vector-valued parent interface: dense layers, CNN encoders, ResNet feature extractors, attention blocks, and transformer representations can all feed one sensory GrapNode per coordinate. The evaluation is organized as a programmability stress suite rather than as a new replay benchmark. In a matched ten-seed Split Fashion-MNIST study, a plastic GrapNet+ER head reaches 63.16 percent seen-class accuracy versus 51.08 percent for a parameter-larger dense MLP+ER under the same seen-class loss and replay memory, with paired delta 12.08 points and p=1.3e-5. On Split CIFAR-10 with a frozen ImageNet ResNet-18 encoder, the same substrate improves the online head over MLP-256 by 3.81 points, with p=0.0026. These results support GrapNet as an editable neural graph substrate whose core value is structural programmability with faithful execution views.

Figurative language and negation are two areas that challenge current language models, however, both are widely used throughout written and spoken language. Large language models (LLMs) are also widely used in everyday contexts where they cannot necessarily be tuned for a specific dataset. It is therefore essential to understand the ability of LLMs to correctly interpret text that includes both negation and figurative language. To investigate this, we develop a set of new annotations to an existing dataset of figurative language, and test a range of language models on the dataset. We find that the combination of negation and figurativeness can present a particular challenge, and that performance overall and across different negation types is particularly dependent on the prompt style used.

This paper studies the computational complexity of verification problems for Binarized Neural Networks (BNNs), where activations (and sometimes weights) are binary. We analyze two problems: satisfiability and robustness under uniform image occlusion. We show that BNN satisfiability is NP-complete via a reduction from Boolean satisfiability problem (SAT), and that uniform occlusion induces a piecewise-constant structure in the network output, enabling a polynomial-time robustness-checking algorithm.

Test-time scaling via sequential revision has emerged as a powerful paradigm for enhancing Large Language Model (LLM) reasoning. However, standard post-training methods primarily optimize single-shot objectives, creating a fundamental misalignment with multi-step inference dynamics. While recent work treats this as multi-turn reinforcement learning (RL), conventional approaches optimize over the multi-step trajectories directly, failing to further exploit the high-quality mistakes in intermediate steps that model can learn from correcting them. We propose a two-stage iterative framework that alternates between online data/prompt augmentation and policy optimization. By converting the intermediate steps (``near-miss'' answers) in the successful recovery trajectories into decoupled revision and verification prompts, our approach concentrates training on both effective answer transformation and error identification. This approach enables efficient off-policy data generation and reduces the computational overhead of long-horizon sampling compared to standard multi-turn RL. On LiveCodeBench, using publicly available test cases as feedback, we observe gains of +6.5 points over the RL baseline and +4.0 points over standard multi-turn training. Beyond coding, our approach matches the previously reported SOTA result on circle packing while using the smallest base model (4B) and far fewer rollouts than the much larger evolutionary search systems. Math results under ground-truth verification further confirm improved correction ability. It also generalizes to out-of-distribution constraint-satisfaction puzzles such as n\_queens and mini\_sudoku, where correctness is defined entirely by problem constraints. Code is available at https://github.com/yxliu02/REVES.git.

Context engineering has emerged as a primary lever for improving AI systems without parameter updates. Recent work showing that textual gradients do not function as real gradients motivates treating automatic prompt optimization (APO) as black-box search. We introduce SPO (Stochastic Prompt Optimization), a framework for stochastic search over prompt space, and compare three strategies of increasing sophistication: error-informed random search, a genetic algorithm with evolutionary operators, and SAGE (SPO via Agent-Guided Exploration), a multi-agent pipeline with diagnostic code execution. Across three benchmarks, no single strategy dominates; effectiveness depends on the interaction of landscape structure with error type. We further deploy SAGE on a mental-health chatbot under a continuous optimization paradigm, where it compounds eight cycles of individually-noisy A/B tests into a statistically robust gain in next-day retention. We argue that coupling qualitative diagnosis with quantitative validation is what makes agentic optimization effective for open-ended task-oriented dialogue.

Genomic archives grow faster than decompression keeps up: the European Nucleotide Archive holds tens of petabytes of fastq.gz, and gzip is fundamentally sequential. GPU decompressors (nvCOMP DEFLATE at ~50GB/s on A100) decode whole files with no random access; CPU genomic tools (CRAM, samtools) support region seeks but only at CPU speed. We extend ACEAPEX, an absolute-offset parallel LZ77 codec included in the official lzbench 2.3 release, with three contributions absent from our prior work. First, a full device-resident GPU decode pipeline (entropy and match resolution both on-device) reaching up to 260GB/s on FASTQ, closing the match-phase-only gap of the earlier paper. Second, position-invariant random access with a compact coordinate index: an arbitrary read decodes in 0.362ms, ~6x faster than warm samtools faidx, with a read-to-block index 6.3x smaller than a .fai. Third, a range-decode strategy that decouples output size from VRAM, sustaining 165.7GB/s on a 50GB genome where whole-file decode runs out of memory. All results are bit-perfect. We also measure Meta's open DietGPU ANS on H100 at 592GB/s decode, faster than the proprietary entropy stage we currently use, showing a fully open high-throughput stack is viable. Code is MIT-licensed.

Multivariate time series anomaly detection (MTSAD) is critical for a wide range of application areas, such as industrial monitoring, cybersecurity, or healthcare. Real-world data is often sparse, irregularly sampled or partially observed, yet existing methods assume uniformly sampled time series. We propose a generative approach based on Latent SDEs that projects the observed time series on a continuous-time stochastic dynamical system, directly being able to handle missing observations and irregular sampling, while also naturally capturing possible cyclic behavior that many real-world use cases inherently possess. Experiments on six anomaly benchmark datasets show that our proposed method ranks first among state-of-the-art baselines. We further demonstrate that our method remains robust under severe data sparsity, while performance significantly degrades for the tested baseline methods. These results highlight latent SDEs as a natural inductive bias for anomaly detection in multivariate time series, especially in presence of real-world irregularities.

Intent-based recommender systems have gained significant attention for improving accuracy and interpretability by modeling the underlying motivations behind user behaviors. Most existing models derive intents directly from user sequences via clustering or prototype learning. However, they are sensitive to sequence quality, require presetting the number of intents, and lack explicit semantic grounding. These issues lead to an incomplete and coarse intent set and limit the effectiveness of recommendation. In this paper, we propose the Sparse Autoencoder for intent-based recommendation (SAERec), a novel recommender that automatically constructs a fine-grained and interpretable intent space from a textual corpus to guide recommendation. Rather than treating texts as side signals, SAERec leverages them as high information density evidence for intent construction. Specifically, we first extract a comprehensive set of fine-grained interpretable intents from the latent space of large language models (LLMs) by using a sparse autoencoder (SAE) to disentangle and interpret text embeddings, which isolates intent-related semantics from textual noise. Then, for each user, we retrieve relevant intents from this set as priors to guide recommendation. It contains personal intents matching a user's current interests and public intents capturing general item patterns shared across users (e.g., quality, price). Finally, to integrate retrieved intents into sequence modeling, we propose a multi-branch attention mechanism that captures temporal dependencies and injects both personal and public intent signals, followed by an adaptive fusion layer to construct the final user representation for recommendation. Extensive experiments on public datasets demonstrate the superiority of SAERec, consistently outperforming state-of-the-art baselines while providing human-understandable explanations.

Multimodal emotion-cause pair extraction (MECPE) requires reliable pair confidence over candidate pairs. Existing pair scorers commonly use pair-level cross entropy over valid candidates, which treats links mostly independently. This leaves the relative confidence geometry among competing causes under-constrained, allowing gold pairs to stay close to hard negatives or rely on incidental non-gold context. We study this vulnerability as pair-confidence brittleness and propose RPCL (Robust Pair Confidence Learning), a training-only framework for pair-confidence learning. RPCL encourages pair confidence to be both discriminative and stable: gold pairs are separated from row-wise hard negatives through a confidence-difference margin constraint, and clean pair predictions are aligned with predictions from a corrupted view where non-gold contextual utterance representations are partially corrupted. The original clean pair scorer and decoding pipeline are used unchanged at inference time. On ECF, MECAD, and MEC4, RPCL improves the three-seed mean Pair F1 over a matched base model by 2.58 to 2.83 percentage points in the full text-audio-video setting, and improves mean Pair AUPRC on all three datasets. Diagnostic analysis further shows larger gold-negative confidence gaps and lower margin-violation severity. These results suggest that explicitly shaping pair confidence is an effective training strategy for MECPE.

Improving GUI agents typically relies on behavior cloning on expert trajectories. However, as the current policy deviates from the expert policy, it inevitably encounters policy-induced off-trajectory states during closed-loop execution, i.e., states that fall outside the expert trajectories. Since expert trajectories provide no demonstrations for these unseen states, such states receive no effective supervision, leaving the policy unable to select the correct action. To close this supervision gap, we propose Skill-Guided Continuation Distillation (SGCD), an iterative self-improvement framework. SGCD first runs the plain policy without skill guidance for a few steps to reach realistic off-trajectory states. From these states, a skill-guided policy then completes the task and produces successful continuations, which are mixed with expert trajectories to supply supervision over policy-induced off-trajectory states. The skills are extracted from both successful and failed rollouts, consisting of Continuation Plans, Critical Targets, Failure Traps, and Success Criteria. On OSWorld-Verified, SGCD improves the success rate of three base models from the low-30\% range to over 50\%, demonstrating its effectiveness and generality.

Text-based telemedicine increasingly relies on lightweight patient feedback, however, such feedback primarily reflects perceived communication quality rather than medical accuracy. We introduce an LM-guided counterfactual recommendation pipeline that discovers and refines interpretable communication features such as tone, personalization, actionability and completeness in addressing patient concerns, without interfering with the medical content. These features are used together with patient-doctor interaction metadata to estimate positive feedback. At inference time, the system searches over low-cost ordinal feature changes and recommends minimal communication changes predicted to increase the probability of positive feedback, while independent auditor models test whether these gains generalize beyond the selection model. Across interactions, recommendations yield a mean +6.41% gain in predicted positive feedback probability under independent auditors, and are non-negative for 93.31% of recommendations. These results suggest that small, interpretable communication changes can capture most predicted gains while preserving the doctor's control over medical reasoning and final wording.

Navigation in partially observable environments presents a significant challenge for autonomous agents, requiring effective decision-making with limited sensory information in unknown environments. Belief-based methods, particularly those using neural networks to approximate the belief space, often fail to capture the inherent multimodality of belief spaces, especially in high-dimensional cases with perceptual aliasing. While generative models present a compelling alternative, they typically require substantial data or expert demonstrations and lack explicit mechanisms for long-term planning. In this paper, we introduce BeliefDiffusion, a novel framework that combines the benefits of both generation and planning. BeliefDiffusion leverages diffusion models to explicitly characterize multimodal belief distributions and utilizes Model Predictive Control (MPC) to simultaneously plan ahead. It consists of two steps: (1) Imagining plausible environment configurations based on observation history and (2) Planning efficient navigation strategies across an aggregated configurations. Through extensive experiments in synthetic map environments, we demonstrate that BeliefDiffusion significantly outperforms both model-free reinforcement learning baselines and other generative approaches in navigation success rate and path efficiency. Our results validate that explicitly incorporating multimodal belief representations into planning enables more robust navigation in partially observable settings.

This work investigates the application of a domain-shift aware neural network for regression tasks aimed at estimating unbalance masses in rotating shafts under varying operating conditions. Experimental data were collected from a test rig in which a primary shaft, equipped with a flange carrying unbalanced masses, was driven at different rotational speeds, while a secondary shaft could be optionally activated to introduce domain discrepancy. The unbalance masses were positioned at a fixed radial distance, and the dynamic response of the system was recorded using triaxial accelerometers. The inverse problem of mass estimation is formulated within a domain adaptation framework, where the network is trained with a maximum mean discrepancy strategy to align feature representations across source and target distributions. The results demonstrate the effectiveness of explicitly addressing domain shift in improving prediction accuracy, especially when the system's physical behavior and sources of domain discrepancy are not fully known and fall outside the training conditions. These findings highlight the potential of domain-shift aware models for regression tasks in Structural Health Monitoring.

Optical coherence tomography (OCT) is essential in ophthalmology, but inconsistent image quality especially in low-cost devices hinders automated analysis. To address this, we introduce a flow-matching-based test-time adaptation method that generates high-quality surrogate images from noisy inputs. Typically, domain gaps between test and training data cause pixel distribution mismatches during the denoising process. We overcome this by matching the test image's histogram to synthetic reference trajectories, successfully aligning the input with expected distributions. Additionally, we remove the network's time conditioning to account for slight deviations in real-world noise distributions. Our approach achieves state-of-the-art performance in segmenting critical biomarkers for two stages of Age-related Macular Degeneration (AMD). Code is available: https://github.com/Veit21/tta-flow.

Large instruction-following models are powerful but costly to deploy, particularly in finance, where labelled data are limited by confidentiality and expert annotation cost. We present an efficient framework for financial sentiment analysis through distillation with synthetic data, transferring knowledge from a large instruction-tuned teacher to compact student models. The framework is designed for low-resource conditions, where a small set of real examples are collected and labelled by hand. The framework then clusters the examples and uses the clusters to select seeds for generating synthetic examples via structured few-shot prompting. Experiments show that clustering-based seed selection yields more representative synthetic data than random sampling, enabling compact models to achieve strong performance with minimal supervision. Notably, on a more complex and noisy text domain, the compact model trained on the complete synthetic-seed corpus even outperforms the teacher model, while remaining competitive on formal text. The framework provides a practical route toward resource-efficient domain adaptation in financial NLP with minimal human labelling effort.

AI systems can increasingly automate scientific workflows, but the reasoning that links prior evidence, generated ideas, experiments and final claims often remains implicit inside model inference. Here we introduce Xcientist, a research harness that externalizes research synthesis and experimental validation into inspectable, contract-governed processes. Xcientist organizes literature evidence, idea states, implementation plans, ablation records and repair traces as persistent research artifacts, so that generated mechanisms can be grounded, executed, tested and revised without losing their evidential basis. We identify claim drift as a failure mode of automated research, where runnable artifacts no longer support the mechanism originally claimed. Across training-free memory systems, graph-structured traffic forecasting and multi-scale physics-informed neural networks, Xcientist preserves traceable trajectories from problem formulation to mechanism design, validation and bounded revision. These results suggest that AI scientists should be evaluated not only by their final artifacts, but by whether their synthesis and validation processes remain attributable, inspectable and scientifically accountable.

When algorithmic predictors inform resource allocation in high-stakes domains such as healthcare, these predictors must account for strategic manipulation of input features. The typical solution is to redesign the predictor itself to explicitly account for strategic interactions. In practice, however, decision makers are often constrained to adjusting coarser levers within existing prediction pipelines. For example, healthcare organizations often select which features to exclude based on perceived manipulability, while using standard regularization procedures to shrink the coefficients of retained features. In this work, we initiate a formal study of strategic classification through feature selection and its interaction with ridge regularization. Our main finding is that excluding individual features based on their manipulability alone is generally suboptimal. We provide a fine-grained characterization of the performance of a feature subset under optimal regularization, yielding new insights for policy design. Motivated by this characterization, we develop a practical algorithm for jointly choosing the feature set and the level of ridge regularization. Through a real-world case study on a healthcare payments benchmark, we illustrate how our algorithm can guide the design of coarse policy levers in practice. Our results provide a principled, practical framework for mitigating the effects of strategic behavior in algorithmic decision-making systems.

This paper studies how to scale learning-based automatic emergency braking (AEB) with massive unlabeled fleet data under production constraints. Our approach is based on meta-feedback semi-supervised learning (MF-SSL), where a teacher generates pseudo labels for unlabeled driving data and is updated using a small labeled anchor set as safety-critical feedback. In production, anchor ambiguity and labeled-unlabeled mismatch can amplify systematic pseudo-label errors, leading to spurious triggers. We propose a stabilized MF-SSL framework with (i) Noise-Aware Decoupling, which removes ambiguity-prone anchors from the teacher's supervised update path, and (ii) kinematics-gated pseudo-labeling with a teacher conflict penalty to suppress mismatch-induced risk hallucinations on unlabeled data while maintaining broad coverage. Extensive experiments show consistent gains as unlabeled data scale from 1M to 1B windows, improving safety while keeping comfort stable. The 1B-trained student model is deployed to hundreds of thousands of vehicles and validated over \$10^9$ km of driving, achieving a positive-to-false activation ratio exceeding 100:1 and a 35% improvement in accident-free driving mileage over a production rule-only baseline.

Reconstructing simulation-ready digital twins of articulated objects from sensor observations remains constrained by two persistent gaps: (i) part-level geometric reconstruction is decoupled from kinematic-parameter estimation, and (ii) the recovered models often violate basic dynamic invariants such as energy conservation, leading to drift when the URDF is replayed in physics simulators. We present KinemaForge, a constraint-driven pipeline that jointly infers part-level shape, joint topology, and joint parameters from short RGB-D sequences and validates the result against an energy-consistent verifier built on differentiable rigid-body dynamics. The pipeline introduces three components: a kinematic constraint graph that encodes joint-part incidences as soft edges; a differentiable screw-axis solver that backpropagates from rendered observations through Featherstone's articulated-body algorithm to joint parameters; and an energy residual loss that penalises non-physical free responses of the reconstructed model. Across five PartNet-Mobility categories and an internal RGB-D benchmark, KinemaForge reduces the average joint-axis error from 4.52 degrees to 2.83 degrees (-37.4%) over the strongest geometric baseline (PARIS) and from 5.30 degrees to 2.83 degrees (-46.6%) over the interaction-based Ditto baseline, lowers long-horizon simulation drift by 64% (vs. PARIS) over 50 s rollouts, and yields URDFs whose closed-loop manipulation success rate improves by 14.6 percentage points over Ditto in our preliminary evaluation. Code and reconstruction data will be released upon acceptance.

Reliable pixel-level uncertainty quantification holds the potential to transform clinical workflows by enabling high-fidelity longitudinal monitoring and distinguishing true pathological changes from artifacts. Ideally, these models provide the stability required for critical treatment planning and surgical intervention. However, standard deep learning models often suffer from miscalibration, yielding overconfident predictions that mask underlying vulnerabilities at subtle pathological boundaries. To address this, we propose QUAM-SM, a post-hoc framework using targeted adversarial search to identify "adversarially fragile" pixels. By actively seeking perturbations that expose predictive instability, our method highlights regions where decisions are most vulnerable to being flipped. Importantly, the framework disentangles epistemic uncertainty from aleatoric uncertainty. Experiments on two public datasets with multiple expert annotations demonstrate that QUAM-SM outperforms both standard and recent uncertainty estimation approaches in terms of reliability and boundary sensitivity. Code is available at https://github.com/HanaJebril/quam_sm

Machine-learning emulators are increasingly used for weather prediction and have the potential to extend skill on subseasonal-to-seasonal timescales by learning dynamically important sources of predictability. A key challenge is whether the models can exploit predictability anchors, such as stratospheric variability, that influence tropospheric circulation beyond short lead times. We test how architectural inductive bias affects emulation of sudden stratospheric warming (SSW) dynamics using paired idealised Isca simulations that differ only in an imposed wave-2 heating perturbation. Across convolutional, transformer, and graph-based architectures trained for one-step prediction, model differences are modest when the stratosphere is dynamically quiet but widen substantially when SSW-like variability is active. Our results identify explicit three-dimensional vertical coupling as a key inductive bias for machine-learning emulation of stratospheric dynamics. However, Eliassen-Palm flux diagnostics show that low forecast error does not guarantee physically faithful wave-mean-flow interaction, with coherent errors remaining in stratospheric wave-driving structure.

Sequence labelling, a core task of Natural Language Processing (NLP), consists in assigning each token of an input sentence a label. From a Machine Learning point of view, sequence labelling is often cast as a Linear-Chain Conditional Random Field (CRF) parametrised by a neural network. While this approach gives good empirical results, CRFs assume a finite decision span (eg label bigrams) which can limit their expressivity and hurt performance when long-range dependencies are required. We show we can leverage diffusion to train a CRF conditioned on an entire label sequence, with the caveat that the condition is on a noisy version of labels. We show experimentally that this method, in conjunction with approximate CRF inference, improves label accuracy with a 16.5% error reduction for POS-tagging.

When a single software artifact changes - a requirement, a configuration value, or a function - engineers must determine what else is impacted. Existing change-impact-analysis (CIA) tooling tends to rely on one of two signals in isolation: semantic similarity recovered from text (information-retrieval traceability, code search, embeddings), or structural dependency following (call graphs, IDE "find usages", test-impact selection). Each has a characteristic blind spot. A semantically driven tool misses an impacted artifact whose text shares no vocabulary with the change; a structurally driven tool misses artifacts related in meaning but not joined by an edge, and most operate only over code rather than the Requirement-Config-Service-Test chain. We argue for a training-free and interpretable analyzer that fuses both signals over the same embeddings. We model the system as a heterogeneous artifact graph with typed edges recovered by static analysis, compute a semantic prior by cosine similarity to the changed artifact, propagate impact multi-hop with decay over a row-normalized propagation matrix, and blend the two with a single tunable weight lambda. A small but complete proof-of-concept on a payment subsystem (5 labelled change scenarios) shows the mechanism we care about: artifacts with zero textual overlap with the change are still recovered through propagation, and helper functions that propagation alone cannot reach are recovered through the semantic layer. The fusion is the only configuration that covers both blind spots, and lambda acts as an explicit precision/recall control. Drawing on four publicly documented production failures, we argue that the same formulation extends to operational artifacts (images, metrics, dashboards, data schemas) that code-only analysis cannot reach.

A recent line of work has reframed individual decision trees as linear models on engineered features associated with their splits, opening routes for oracle inequalities and feature-importance reinterpretation, but leaving open the question of what unified geometric object a forest induces when one indexes its feature map by nodes rather than by splits. The present paper studies that object. KPP indexes the feature map by the nodes of the forest, weighted by a path metric that turns each coordinate into a component of a squared-Euclidean path-isometric embedding. KPP unifies four pillars under a single non-diagonal Gram that carries a metric: prediction, exact additive attribution, deterministic Lipschitz robust radius in the KPP metric, and uniform Rademacher risk bounds for regression and classification under fixed, honest, or cross-fit conditioning. All probabilistic guarantees are conditional on the representation and are stated under three explicit conditioning regimes; the robust-radius guarantee is deterministic in the KPP metric rather than in a norm on the raw input. Conjectured fast-rate refinements for both regression and classification are stated as open problems and are not claimed as theorems.

Classifying implicit hate speech remains a challenge, as intent is often masked through insinuation and context rather than explicit slurs. Prior supervised contrastive approaches improve in-domain detection but can overfit surface cues and struggle to transfer across datasets. We propose ImpSH, a triplet-based framework that aligns posts with implied statements when available and uses context-bounded semi-hard negatives to focus learning on near confusions. We also examine AugSH, which forms positives via data augmentation. In controlled evaluations on IHC, SBIC, and DynaHate with BERT and HateBERT, ImpSH is a viable alternative to standard supervised contrastive baselines and often improves cross-domain performance under matched preprocessing and tuning budgets. Representation analysis using alignment and uniformity indicates tighter positive pairs with balanced global spread, and qualitative nearest-neighbor case studies illustrate typical false negatives under domain shift. These results demonstrate that aligning posts with their implied statements via context-bounded mining provides a more stable, bijective-like mapping to related insinuations, overcoming the volatility inherent in traditional clustering-based representation learning.

Abstractive summarization plays a crucial role in enabling efficient understanding of scientific literature, yet it inherently demands both linguistic fluency and factual faithfulness. Existing approaches often fail to reconcile these two requirements. Extractive methods rely on rigid sentence splicing that disrupts macro-level logical coherence, while large language model (LLM)-based generative approaches, despite mastering linguistic fluency, exhibit limited factual consistency. In this work, we propose ScholarSum, a hierarchical reflective graph-based framework that emulates a student-teacher writing process for fluent and faithful scientific summarization. ScholarSum first organizes the document into a hierarchical knowledge graph by segmenting it into semantically coherent units, whose multi-layered community structure captures global logic and macro-level themes. Guided by this global structure, the student generates an initial draft, which is subsequently refined through fine-grained evidence retrieval. To ensure factual consistency, a teacher-like reviewer then iteratively examines the draft, identifies unsupported content, and prompts targeted re-retrieval and rewriting until the summary meets rigorous quality standards. Extensive experiments demonstrate that ScholarSum significantly outperforms previous baselines in terms of both completeness and faithfulness. Our code is available at https://github.com/Xiaoyu-Tao/ScholarSum.

To assist humans over extended periods in real homes, embodied agents must remember user routines, world states, and past interactions. Existing long-term memory benchmarks mainly evaluate language-centric retrieval and question answering, while embodied benchmarks often focus on short-horizon task execution without testing long-term memory use in dynamic environments. We introduce WorldLines, a project-driven benchmark for long-horizon embodied household assistance. It constructs temporally extended household traces with dialogues, actions, execution feedback, object and device state changes, and converts them into evidence-linked samples for Memory QA and Embodied Task Planning. We further propose ObsMem, an observer-grounded memory framework that maintains visibility-aware memories and action-native state trails for state-aware decisions. Experiments reveal persistent challenges in partial observability, overwritten world states, and translating long-term memory into embodied plans, while ObsMem offers a stronger reference architecture for this setting.

Self-distillation improves reasoning in large language models by using the model's own rollouts as training signal, typically through implicit logit-level alignment that minimizes KL divergence toward a privileged target distribution. However, because this supervision is generated via uncontrolled sampling, it provides no diagnostic insight into the model's specific errors or corrective guidance for its individual failure patterns. Consequently, the model learns to imitate a privileged distribution rather than receiving fine-grained corrections that pinpoint where and why its reasoning fails. In this paper, we propose Trajectory-Augmented Policy Optimization (TAPO), which advances self-distillation from implicit distributional alignment to explicit trajectory construction. During RL training, the model produces both correct and incorrect rollouts to the same query, and TAPO leverages this contrastive structure to construct micro-reflective corrections, new training trajectories that retain the model's erroneous reasoning up to the point of failure, then insert a natural-language diagnosis and corrected reasoning guided by a correct reference from the same sampling group. Since each trajectory is anchored in the learner's own prefix and solutions, the corrective signal preserves the model's on-policy distribution to a greater extent than the position-wise alignment imposed by KL-based methods. To integrate these trajectories, TAPO introduces difficulty-aware candidate selection at the model's capability boundary and decoupled advantage estimation to prevent gradient contamination. Experiments on AIME 2024, AIME 2025, and HMMT 2025 show that TAPO achieves consistent improvements over GRPO under the same number of training steps. Further analysis demonstrates that TAPO strengthens both first-pass reasoning and error-correction effectiveness.

Vision-language models (VLMs) are now widely used in downstream tasks. However, real-world applications often expose VLMs to distribution shifts induced by semantic variation (e.g., shape, size, and style). Robustness certification determines if a model's prediction changes when transformations are applied to its input. While most certification frameworks study geometric or pixel-level transformations over inputs, this work proposes a novel framework that enables certifying VLM robustness under semantic-level transformations. Leveraging the open-vocabulary capability of VLMs, we use text prompts as semantic proxies to construct transformations parameterized by an extent that controls the degree of semantic variation. By characterizing the VLM decision boundary in closed form, our framework quantitatively certifies extent intervals for which the predicted class remains unchanged under the semantic transformation. Our framework is the first to certify VLM robustness under semantic-level variations without requiring additional data for each variation, making it practical to apply. Experiments on both synthetic and real-world data show that our framework enables certifying robustness under diverse semantic variations across scenarios.

Large Language Model (LLM)-based automatic Multi-Agent Systems (MAS) generation has become a crucial frontier for tackling complex tasks. However, existing methods face a dilemma between model capability and experience retention. Inference-time MAS leverages frozen frontier LLMs but repeats identical searches without learning from past experience. Conversely, Training-time MAS internalizes experience via gradient updates but is constrained by the low capability ceiling of smaller models, and is hard to scale to large frontier LLMs. To bridge this gap, we propose Skill-MAS, a novel third path that decouples experience retention from parametric updates by conceptualizing the high-level orchestration capability as an evolvable Meta-Skill. Skill-MAS refines this architectural knowledge through a closed optimization loop: (1) Multi-Trajectory Rollout samples a behavioral distribution for each task under the current Meta-Skill; and (2) Selective Reflection adaptively selects priority tasks and applies hierarchical contrastive analysis to distill systemic experience into generalizable, strategy-level principles. Extensive experiments across four complex benchmarks and four distinct LLMs demonstrate that Skill-MAS not only achieves remarkable performance gains but also maintains a favorable cost-performance trade-off. Further analysis reveals that the evolved Meta-Skills are highly robust and exhibit strong transferability across unseen tasks and different LLMs.

Effective human-robot teamwork requires robots to adapt to partners, situations, and task dynamics from the start of an interaction. In the MATRX Urban Search and Rescue (USAR) environment, people can externalize collaboration patterns (CPs) they discover during teamwork through a chat and reflection interface. We study whether a robot can use such prior team experience to become a better teammate in future interactions. To this end, we represent historical CPs as knowledge-graph episodic memories and use graph representation learning with a node-classification objective to identify a representative and effective memory for reuse. We then initialize the robot with this memory before a new collaboration episode begins. Across 20 participants and 160 round-level observations, initializing the robot with a single automatically selected prior CP increases rescue success from 25.7% to 41.3% and reduces average task time by 283 seconds. The strongest gains appear at the beginning of interaction, suggesting that reusable episodic memory can help robots enter collaboration with more effective task knowledge and support smoother early teamwork.

Causal discovery methods commonly assume that all data is independently and identically distributed (i.i.d.) and that there are no unmeasured variables affecting the system. In practice, these assumptions are often violated, leading to inaccurate inference. In this paper, we study how to identify hidden confounding and selection biases from causal mechanism shifts. In particular, we show that structural biases lead to dependent mechanism shifts. That is, by considering for which variables the mechanisms change given data from different environments, we can tell which variables are unbiased, which are subject to hidden confounding, and which are undergoing selection bias. We formalize this into an empirically testable criterion based on mutual information, and show under which conditions it identifies structural biases. To tell which nodes are subject to what kind of bias, we introduce the StruBI algorithm. Experiments on synthetic and real-world data show that StruBI works well in practice, accurately recovering affected variable sets and types of biases, outperforming the state-of-the-art by a wide margin.

This paper introduces a semi-supervised clustering framework grounded in the statistical duality between grouping principles and anomaly detection. We address the challenge of robust cluster definition in noisy environments -- a task where partitioning algorithms often over-assign outliers and density-based methods remain sensitive to heuristic global parameters. Drawing on \textit{a-contrario} statistical reasoning and Gestalt proximity principles, we define a cluster as a maximal subset of data points containing no anomalies relative to a null hypothesis of uniform randomness. Central to this approach is the Perception algorithm, which utilises a principled expectation-based threshold ($\mathbb{E} < 1$) to identify outliers without manual parameter tuning. By treating clustering as the dual of anomaly detection, we employ an iterative ``clustering-by-exclusion'' mechanism. The algorithm is seed-guided, leveraging minimal user-provided labels to initialise robust cluster medians and form initial groups, which are subsequently expanded by admitting non-anomalous points. This approach naturally isolates fringe points, isolated noise, and emerging unknown clusters. We evaluate the method on synthetic and real-world benchmarks, including image and text datasets represented through raw, linear-reduced, and neighbourhood-preserving embeddings. Results demonstrate that with as few as 10--30 seeds per cluster, the proposed method achieves competitive and often very strong performance under a practical low-tuning benchmarking protocol, while maintaining linear scalability with respect to both observations and dimensionality for a fixed number of seeded clusters and iterations.

Counterfactual explanations are widely used to provide algorithmic recourse in high-stakes decision-making systems. Most existing methods seek the smallest change to an input that flips a model's decision. However, decision-makers often rely not only on predicted labels but also on confidence thresholds and risk margins. Counterfactuals that barely cross a decision boundary can be fragile and unstable under noise or model variation. In this paper, we propose Target-confidence Recourse Using tSeTlin machines (TRUST), a framework in which users explicitly specify the desired prediction confidence for recourse. Rather than generating counterfactuals and evaluating confidence afterward, TRUST directly searches for minimal changes that satisfy a user-defined confidence target, enabling comparison of recourse options in terms of cost, confidence, and robustness. We instantiate TRUST using a Probabilistic Tsetlin Machine (PTM) combined with Bayesian optimization. The probabilistic clause-based structure of PTM links prediction confidence to the stability of decision rules. We show that counterfactuals satisfying the same rules can still differ substantially in reliability depending on how securely they satisfy those rules, revealing whether decisions are supported by robust or fragile clause activations. Experiments on synthetic and real-world datasets demonstrate that target-confidence counterfactuals produce more robust and interpretable recourse than conventional boundary-based approaches. Across multiple benchmarks, TRUST achieves perfect robustness while maintaining low recourse cost, including an L2 distance of 0.10 on the Haberman dataset at 0.92 confidence. By explicitly controlling confidence and exposing rule-level stability, TRUST provides actionable recourse for high-stakes decision support.

Long-context reasoning is an essential capability for large language models, particularly when they are deployed as autonomous agents that must reason over lengthy trajectories. Reinforcement learning (RL) has recently emerged as a dominant paradigm for improving this ability, yet existing work largely focuses on reward engineering while diverse training data remains scarce. We revisit this problem from a data-centric perspective and show that a simple yet effective data recipe alone, paired with a minimal outcome-based GRPO setup, suffices to substantially improve long-context reasoning. Our recipe targets three complementary task families -- retrieval, multi-evidence synthesis, and reasoning -- for which we construct and curate eight datasets totaling ~14K examples. Experiments on three models (Qwen3-4B/8B/30B-A3B) yield average gains of +7.2/+3.2/+6.4 points across seven long-context benchmarks, surpassing prior RL training sets. We further demonstrate that these gains transfer to agentic tasks, where continuing RL training on an agent-tuned model with our data recipe improves GAIA by +4.8 and BrowseComp by +7.0 points. We will release our datasets to facilitate future research.

Memory benchmarks for LLM agents largely assume single-user settings, leaving shared assistants for hospitals, workplaces, campuses, and households understudied. In these deployments, multiple principals write to a common memory pool and query it under different roles, scopes, and relationships, so memory quality requires governance as well as recall. We introduce GateMem, a benchmark for multi-principal shared-memory agents. GateMem jointly evaluates utility for legitimate long-horizon requests with state updates, access control across contextual authorization boundaries, and agent-facing active forgetting after explicit deletion requests. It spans medical, office, education, and household domains, with long-form multi-party episodes, incremental memory injection, hidden checkpoints, structured judging, and leak-target annotations. Across diverse baselines and backbone models, no method simultaneously achieves strong utility, robust access control, and reliable forgetting. Long-context prompting often yields the best governance score at high token cost, while retrieval-based and external-memory methods reduce cost yet still leak unauthorized or deleted information. These results show current memory agents remain far from reliable shared institutional deployment.

Traditional approaches place intelligence in the agent, whether as a learned policy or a search procedure. We instead place intelligence in the space itself: a scene induces a Riemannian metric on the configuration manifold, and action reduces to following the geodesics of that metric rather than invoking a separate planner or collision checker. A single Encoder-Router network realizes this idea through three complementary parameter groups -- frame parameters that orient the generators, modulation parameters that govern their spatial propagation, and basic coefficients that determine their strength. These groups combine through a shared semigroup-superposition mechanism to produce a single Riemannian metric field, yielding a compact architecture whose geometry scales naturally with scene complexity. Trained on a single two-obstacle scene, the model demonstrates robust zero-shot generalization across unseen obstacle configurations, with orders-of-magnitude separation between collision-free and obstacle-penetrating path costs.

Pedestrian trajectory prediction from an ego-centric camera is challenging since it depends on complex interactions with vehicles and scene context, as well as the intention of the pedestrian. By modelling correlation and intent from the historical and future trajectories of the pedestrian, it will usually result in a multimodal (i.e. multiple modes) distribution. Existing stochastic predictors often sample multiple futures from a single unimodal distribution, which can yield sub-optimal 'mixed-mode' trajectories that lie between distinct motion patterns and become implausible in real scenes. In this paper, we propose MMPM, a mode-aware framework that separately models future trajectory distributions into semantically meaningful modes based on the pedestrian's crossing behavior. MMPM consists of two modules: behavior-aware Pedestrian Interaction Module (PIM) that jointly captures pedestrian-vehicle and pedestrian-environment interactions by introducing gaze, head and hand gesture, and a CVAE-based Mode-aware Trajectory Predictor (MTP) module to model the future trajectory distributions on two modes, crossing and non-crossing the road, separately. A query-based decoder further enforces mode consistency during decoding. Experiments on PIE and JAAD datasets show that our method surpasses state-of-the-art baselines. Our proposed MTP is model-agnostic, which can be integrated into existing frameworks such as BiTrap-NP and SGNet-ED to further improve future trajectory prediction performance. We additionally introduce a data-driven validation protocol that matches predictions to spatio-temporally consistent ground-truth trajectories, demonstrating improved frame-wise displacement errors over previous work.

Sequential decision problems often exhibit an asymmetric evolution of information and decision flexibility: as a decision cycle unfolds, the agent receives richer information while feasible actions expire due to operational cutoffs, commitments, or resource constraints. Standard MDP formulations typically flatten this structure into stage-dependent state descriptions and action masks, thereby obscuring the nested information--action asymmetry that determines which decisions are urgent and which can be deferred. We introduce Maturing Markov Decision Processes (MMDPs), a formulation built around this information--action asymmetry. We characterize one of its key consequences through an expiring-action priority principle, which identifies the actions that must be resolved before the next stage. Motivated by this structure, we develop a structure-aware reinforcement learning framework with stage-aware policy design, expiring-action abstraction, and search-augmented learning with distillation. Experiments on a controlled multi-supplier replenishment problem, simplified cash-management environments of increasing complexity, and a production-scale simulator show that explicitly modeling this asymmetry improves learning efficiency and becomes increasingly valuable as decision problems scale.

Hybrid brain-computer interfaces (BCIs) that integrate motor imagery (MI) and steady-state visual evoked potentials (SSVEP) provide high-dimensional neural decoding but typically exceed the computational limits of embedded hardware. To address this, we propose SwitchBraidNet, a compact EEG classification architecture designed for low-power deployment. The model employs a dual-path temporal braid to extract multiscale oscillatory features, an adaptive squeeze-and-excitation spatial switch for electrode gating, and a log-variance readout layer for direct band-power encoding. Furthermore, through systematic quantisation-aware training on the OpenBMI dataset, we compared SwitchBraidNet against four established baselines across FP32, FP16, and INT8 precisions. Experimental results demonstrate superior efficiency and performance, achieving MI accuracy of 69.49% (FP16), SSVEP accuracy of 93.48% (FP32), and a hybrid information transfer rate of 64.82 bits/min (FP16). With an INT8 footprint of only 3.03 KB, SwitchBraidNet maintains high accuracy across varying numerical precisions, demonstrating its suitability for low-power embedded BCI deployment.

Foundation models for language and vision are powered by internet-scale data, while structured domains (tabular prediction, time-series forecasting, graph learning, reinforcement learning) are not. The substitute is synthetic data, which shifts the burden from collection to prior design. Such priors already exist for many structured tasks: TabPFN and its successors solve tabular classification with a transformer pretrained on a synthetic Bayesian prior. We make two points. \textbf{First}, reinforcement learning is the conspicuous gap: sampling a synthetic MDP is as feasible as sampling a synthetic tabular dataset, yet no in-context RL work treats prior design as a primary objective. \textbf{Second}, MDPs admit a fixed-size sufficient statistic, independent of the episodes observed and tabular in shape, which makes them directly amenable to the attention-based architectures used for tabular foundation models, with a policy head replacing the supervised target. Together these define the agenda for an RL foundation model. As a proof of concept, we train one model entirely on synthetic MDPs and show that, with no task-specific tuning, it solves held-out tabular benchmarks in context, both online and offline: online, in far fewer episodes than UCB-VI and tabular Q-learning, and offline, competitively with VI-LCB.

Learned sparse retrieval (LSR) models such as SPLADE have traditionally used BERT-style masked language models as backbone encoders. A natural expectation is that replacing BERT with stronger pretrained encoders should improve retrieval effectiveness. However, we find that under standard SPLADE training recipes, backbones with large MLM-head L2 norms can suffer performance degradation and even training collapse under standard SPLADE training recipes. We identify this failure as a scale mismatch in the MLM head: SPLADE directly uses MLM-head outputs to construct sparse lexical representations, and query-document relevance is computed by an unnormalized dot product over these representations. As a result, an inflated MLM-head scale can amplify sparse activations, distort matching scores, and destabilize contrastive training under common training settings. To address this issue, we introduce a simple initialization-time correction that rescales the MLM-head projection by a constant factor before SPLADE training. This zero-cost adjustment improves training stability without modifying the model architecture or training objective. Across both in-domain and out-of-domain retrieval benchmarks, this simple correction substantially improves large-norm backbones such as ModernBERT and Ettin, turning unstable training runs into competitive sparse retrievers. In several settings, the corrected models further match or surpass the classic BERT-SPLADE baseline. These findings suggest that the bottleneck in adapting pretrained encoders to LSR is not encoder capacity alone, but the calibration of the MLM-head scale used to construct sparse lexical representations.

Reinforcement learning with verifiable rewards (RLVR) has driven substantial progress in training LLMs for reasoning tasks, but representative methods such as GRPO assign uniform credit across all tokens, wasting gradient on routine tokens while under-crediting pivotal reasoning steps. Existing token-level credit assignment methods require resources beyond the model's own rollouts. GRPO variants rely on process reward models or ground-truth answers. Knowledge distillation assigns credit through per-token divergence but requires external teachers (On-Policy Distillation) or privileged information (On-Policy Self Distillation). However, these dependencies limit applicability in the pure RLVR setting. We observe that conditioning the model on its own verified trajectories induces a measurable per-token KL divergence between the original and conditioned distributions, and prove that distilling from a self-teacher constructed by verified trajectories leads to infeasible weighted-average solutions when multiple verified trajectories exist. We propose SC-GRPO (Self-Conditioned GRPO), which uses KL divergence mentioned before as a multiplicative weight on GRPO gradients. Across five benchmarks spanning math, code, and agentic tasks, SC-GRPO consistently outperforms 8.1% over GRPO and 5.9% over DAPO with stronger OOD performance. Moreover, SC-GRPO achieves higher performance than OPD.

The field of learning-augmented algorithms has demonstrated that machine-learned predictions can bypass worst-case lower bounds across a wide range of problems. So far, however, the focus has been almost exclusively on polynomial-time algorithms, where predictions improve competitive ratios, approximation guarantees, or running times. In this paper, we raise the question of whether predictions can push the frontier of exact exponential-time algorithms for NP-hard problems. We answer this question affirmatively by proposing a general approach that augments an entire family of state-of-the-art exact algorithms for a variety of subset selection problems. We show that a noisy predictor that is only marginally better than random guessing suffices to provably reduce the search space, and that the resulting runtime speedup scales smoothly with the prediction quality. Importantly, our algorithms require only pairwise independence of predictions or, alternatively, do not require the knowledge of the predictor's accuracy - both strictly weaker and more realistic settings than typically assumed.

Bringing Large Language Models (LLMs) into industrial ride-hailing dispatch as semantic feature extractors over platform-scale behavioral logs is a compelling but under-explored data systems problem. Production matching pipelines remain dominated by structured numerical features, yet decisive behavioral signals (e.g., a driver's habitual aversion to certain regions) are inherently contextual and naturally expressible as LLM-generated user profiles. However, scaling such profiling to a live, millisecond-latency dispatcher faces three intertwined constraints rarely addressed together: on a platform with millions of daily orders, logs exceed any LLM's context window by orders of magnitude; most users are long-tail, with too few interactions for per-user profiling; and surface-fluent profiles do not necessarily improve downstream prediction utility. We present ProfiLLM, an agentic LLM data pipeline that operationalizes utility-aligned user profiling for production matching systems through two modules. (1) Tool-Augmented Global Knowledge Mining equips an LLM agent with 27 analytical tools to mine platform-scale data, producing reusable global knowledge, adaptive user clustering rules, and region-level supply-demand priors. (2) Utility-Aligned Profile Exploration generates multiple candidate profiles per cluster, evaluates them via a lightweight downstream utility proxy, iteratively refines the best candidates and constructs preference pairs for DPO fine-tuning. Deployed on DiDi's production dispatcher, ProfiLLM achieves up to +6.14% relative AUC improvement in outcome prediction, up to +4.35% GMV gain in dispatching simulation, and consistent improvements in a 14-day online A/B test including +0.47% GMV, +0.33% Completion Rate, and -0.82% Cancel-Before-Accept rate.

With the rapid expansion of massive multilingual corpora, Multilingual Information Retrieval (MLIR) has emerged as a critical technology for global information access. MLIR enables users to retrieve semantically relevant documents from multilingual text collections using a single-language query. However, recent multilingual dense retrieval models often exhibit a strong preference for documents in the same language as the query. This leads to severe language bias, where top-ranked results are dominated by documents of specific languages, even when documents in other languages contain more semantically relevant information. To address this issue, we propose SHIFT, a training-free method applicable in the indexing stage. Specifically, SHIFT utilizes parallel translation pairs to estimate a relative language vector for each target language with respect to a source language. Subsequently, SHIFT corrects the language-specific offset by subtracting this relative language vector from document embeddings during indexing. Our comprehensive evaluation across four MLIR benchmarks and diverse dense retrieval models confirms that SHIFT can effectively mitigate language bias and enhance MLIR performance.

Reliable evaluation of generated radiology reports requires strict clinical accuracy, as omitted critical findings or mischaracterized radiographic observations can directly affect patient care. Existing metrics obscure this requirement by reducing report quality to a medically ungrounded scalar. Although Large Language Models (LLMs) possess rich medical knowledge, they likewise struggle to draw a reliable boundary between clinically significant errors and harmless variation. We study this boundary using ReEvalMed benchmark as testbed and evaluate metric-level clinical significance from detecting true clinical errors ("Discrimination") and tolerating insignificant variations ("Robustness"). Across 8 LLM evaluators under one-pass and two-pass settings, we identify a widespread discrimination bias: models effectively detect errors but also over-penalize harmless rephrasings. To mitigate this, we synthesize 4k report pairs and train lightweight interpretable metrics on Qwen3-8B and MedGemma-4B. Our trained metric sharpens the clinical significance boundary, surpassing 32B-scale medical LLMs and remaining competitive with proprietary models. Crucially, the more costly two-pass setting fails to consistently improve overall performance and mainly trades discrimination for robustness. These findings suggest one-pass trained metrics as the practical choice for cost-sensitive deployment, with two-pass inference reserved for settings where D-R balance is critical. We will release the dataset and metric.

Transformer-based architectures have significantly advanced the generation of complex symbolic sequences, yet a significant gap remains in achieving fine-grained, interpretable control over discrete signal attributes. This paper investigates the mechanistic interpretability of the Multitrack Music Transformer (MMT) and proposes a framework for deterministic attribute modulation without retraining to bridge this gap via inference-time activation steering. Utilizing the Difference-in-Means (DiffMean) methodology, we isolate latent directions for signal attributes, specifically Pitch and Duration, within the residual stream. We validate the Linear Representation Hypothesis in this domain, achieving high correlation between steering magnitude and attribute shift. To address the inherent feature entanglement in multi-attribute steering, we introduce a Dual Steering framework utilizing Gram-Schmidt Orthogonalization. Experimental results demonstrate that this geometric decoupling reduces conceptual interference and signal degradation compared to naive vector addition, enabling independent deterministic control even against strong autoregressive conditioning.

Teaching machines to emulate natural handwriting styles remains an open challenge, as it requires synthesizing stroke sequences that dynamically vary in shape, texture, pressure and script - not only across individuals, but also within a single person's handwriting. Attempts at this challenge have largely explored deep learning methods in both online and offline settings. However, these approaches are often constrained by style-specific architectural choices, heavy reliance on large datasets, high compute costs, and a lack of flexible control over writing styles through natural language. To this end, we introduce HandwritingAgent, a language-driven agent that can synthesize natural handwriting sequences directly in Scalable Vector Graphics (SVG) format with no need for style-specific training. The agent leverages a large reasoning model to geometrically analyse and autoregressively generate target handwritten glyphs as stroke sequences in a discrete grid canvas environment. Generation is conditioned on texts provided in either conversational or non-conversational mode, along with a reference handwriting-style image. Experiments on diverse handwriting tasks spanning imitation, recognition, multi-lingual handwriting synthesis, and generation of complex handwritten maths and science expressions indicate substantial improvement in performance, with HandwritingAgent matching or surpassing state-of-the-art generative handwriting models, while providing a more efficient, controllable, and generalizable synthesis method.

Robot soccer is a challenging testbed for multi-agent reinforcement learning because it combines partial observability, cooperative and adversarial interaction, sparse rewards, and long-horizon tactical behavior. RoboCup 2D Soccer Simulation (RCSS2D) provides a mature robot-soccer platform, but its competition-oriented server-client architecture is difficult to use directly with modern Python-based MARL workflows. We introduce R2D-RL, a reinforcement learning environment that connects RCSS2D and HELIOS-based player clients to a Python MARL interface through shared-memory communication and cycle-level synchronization. R2D-RL supports full-field and scenario-based training with configurable opponents, Base discrete and Hybrid parameterized action spaces, action masks, expected possession value (EPV)-based reward shaping, and parallel execution. We provide front-goal scenarios and an 11-vs-11 full-field benchmark, together with baseline results.

Identifying Pareto optimal solutions is critical to support multi-objective decision-making. We introduce the first anytime Multi-Objective Multi-Armed Bandit algorithm for the Pareto Set Identification problem, taking a Bayesian approach: Top-Two Pareto Front Thompson Sampling (TTPFTS). We benchmark TTPFTS against state-of-the-art fixed-budget Pareto Set Identification algorithms on synthetic environments. Next, we demonstrate its practical utility in a challenging multi-objective molecular discovery setting by efficiently exploring an ultra-large synthesis-on-demand molecular library. Furthermore, we introduce a novel uncertainty quantification metric that estimates our algorithm's confidence in the predicted Pareto set. We demonstrate that this metric effectively proxies true performance, yielding a robust methodology for monitoring learning progress in complex settings. Finally, we complement these empirical findings with a theoretical proof of the algorithm's asymptotic correctness.

Large Language Models are increasingly applied to sensitive domains that require redaction of personally identifiable information (PII). While redacting PII is a data cleaning prerequisite, existing benchmarks conflate extraction mechanics with privacy semantics. A public phone number is not equivalent to a phone number in a medical record. Whether information constitutes a violation depends heavily on who holds it, why, and in what context, fundamentally differentiating redaction from simple entity recognition. Grounded in contextual integrity, we introduce RedactionBench, a manually annotated benchmark comprising 200 diverse documents across 11 domains, mostly seeded from real-world sources. We also introduce R-Score, a novel character-level metric that treats semantically similar redactions equally and nullifies shallow formatting choices, such as varying masking styles for phone numbers. Evaluations across Named Entity Recognition models, entity extraction Small Language Models, and frontier models equipped with agentic tools demonstrate that contextual redaction remains an unsolved problem. A human evaluation with over 80 users on RedactionBench reveals a stark dichotomy in privacy perceptions. Annotators show consensus with target labels for mandatory redactions (89.4 percent) and safe text preservations (94.1 percent), but fail to agree on contextual redactions (47.7 percent). This variance demonstrates the subjective nature of contextual privacy and motivates R-Score, which decouples contextual ambiguity from strict precision. We compare 35 models across families and report their performance in redacting PII. Finally, we release RedactionBench to establish a baseline for future privacy-preserving systems, hoping to inspire efficient model design and standardized evaluations.

Dense retrieval ranks one query vector against one document vector. On long documents, this interface can fail when a short but decisive span is weakened during document encoding before ranking. We study this failure mode as document-side early compression and introduce the Evidence Dilution Index (EDI) to measure how far a document-level representation falls below the strongest chunk-level evidence within the same gold document. Guided by this view, we propose DICE (Document Inference via Chunk Evidence), a training-free document-side strategy that splits documents into chunks, encodes them independently with a frozen model, and aggregates them back into a single vector while preserving the standard one-query-one-document interface. On LongEmbed, DICE improves retrieval across four backbones, with the largest gains on slices beyond 4k tokens: for Dream, Passkey >4k rises from 30.0 to 90.0 and Needle >4k from 23.3 to 74.0. Across 12,779 filtered samples, DICE yields lower EDI than the single-vector baseline in 92.8% of cases. These results establish document-level encoding as a practical and underexplored lever for long-document retrieval.

Multimodal Information Extraction (MIE)-covering tasks such as Multimodal Named Entity Recognition (MNER), Relation Extraction (MRE), and Event Extraction (MEE)-is essential for understanding multimedia content but remains constrained by severe data scarcity. Although data augmentation is a promising remedy, existing approaches are impeded by coarse cross-modal alignment and fragmented, task-specific designs that fail to exploit shared semantic knowledge. To overcome these limitations, we introduce Semantic Anchor-aligned Multimodal Augmentation (SAMA), a unified framework for generating high-fidelity, task-aware synthetic data. SAMA constructs structured semantic anchors from ground-truth labels to guide a Collaborative Multi-Experts Multimodal Large Language Model (CME-MLLM), which integrates a Universal Adapter for shared semantics with Task-Specific Adapters to produce diverse yet constraint-compliant textual samples. For image synthesis, SAMA employs an Anchor-Preserving Diffusion mechanism that uses anchor-weighted prompts and latent conditioning to maintain critical semantic anchors while diversifying visual contexts. To eliminate the need for manual verification, SAMA further introduces a Dual-Constraint Filtering module that selects synthetic samples based on both cross-modal consistency and anchor fidelity. Extensive experiments across benchmark datasets for MNER, MRE, and MEE demonstrate that SAMA consistently outperforms state-of-the-art augmentation baselines under both fully supervised and low-resource settings, underscoring its versatility, robustness, and effectiveness.

Online learning in non-stationary streams is often formulated as tracking a point estimate, but many applications require predicting the full data-generating distribution. We study online distributional prediction under drift and adversarial corruption. Our approach represents each candidate law through a latent cluster geometry: a variable-size configuration of centers that organizes probability mass and induces a predictive distribution. A Gibbs quasi-posterior over these configurations yields an online predictor by posterior averaging, and the resulting variable-dimensional posterior can be sampled with reversible-jump MCMC. The method therefore avoids specifying a parametric streaming law while retaining a structured latent space for uncertainty, regularization, and comparison. We evaluate performance by cumulative Wasserstein-1 regret against the time-varying true law. The analysis separates two effects: corruption perturbs the loss-based posterior update, whereas drift makes long-horizon posterior memory stale. We address the latter with a restarted variant that temporally localizes the same quasi-Bayesian update. The resulting high-probability bounds decompose into a PAC-Bayesian complexity term, a corruption-sensitive posterior perturbation term, and a dynamic optimal-transport term driven by \(A_T^{\mathrm{OT}}=\sum_{t=2}^T W_2^2(p_{t-1}^*,p_t^*)\). Under bounded support, stable latent geometry, predictive-map regularity, oracle realizability, localized restart windows, sublinear transport action, and sublinear corruption budget, the restarted predictor achieves sublinear cumulative Wasserstein regret. These guarantees require no parametric model for the stream, drift mechanism, or corruption process.

We present RouteJudge, an online pairwise preference evaluation framework for LLM routing systems, with a public platform available at https://routejudge.cn. Different from model-level response evaluation, RouteJudge focuses on router-level decision quality. For each user query, multiple routing strategies independently recommend candidate models under the same model pool and budget constraints. The selected model responses are then presented to users through anonymous pairwise comparisons, and the resulting user preferences are attributed back to the routing strategies behind the compared responses. Each evaluation record stores the query, routing decisions, model responses, preference labels, cost, latency, and task metadata, enabling preference-aware, cost-aware, and task-conditioned analysis of LLM routers. To support the continuous expansion of routing methods in RouteJudge, we further release ORBIT (Optimal Routing and Budgeted Inference Toolbox), a modular and extensible toolbox that standardizes the end-to-end workflow of LLM routing. ORBIT provides unified interfaces for benchmark loading, query representation, router implementation, budget-aware evaluation, and method comparison, allowing researchers to develop and evaluate routing algorithms under consistent protocols. It also serves as the submission and integration layer for RouteJudge: researchers can implement routing methods within ORBIT, validate them on existing routing benchmarks, and submit compatible routers for online preference-based evaluation. The code of ORBIT is available at https://github.com/AIGNLAI/LAMDA-ORBIT.

We study differentially private (DP) regression in settings where each data sample includes public, non-sensitive features -- common in applications such as recommendation and advertising systems. While such label-DP or semi-sensitive-feature settings have been primarily explored in the context of classification, effective approaches for regression remain underexplored. We introduce Cond-DP, a conditioned variant of DPSGD that leverages the structure of public feature matrices to improve optimization under privacy constraints. Motivated by the observation that these public features often exhibit rapidly decaying spectra, Cond-DP incorporates a data-driven conditioning matrix to reshape the optimization landscape and accelerate convergence. We provide convergence guarantees for convex, strongly convex, and non-convex settings, and recover standard DPSGD as a special case when the conditioning matrix is the identity. We show how to construct an effective conditioning matrix for Cond-DP directly from public features, enabling provably faster convergence than DPSGD in private linear regression without incurring additional privacy cost. Empirically, Cond-DP with this conditioning matrix consistently outperforms state-of-the-art baselines across a wide range of datasets and model architectures under label DP, demonstrating strong and robust performance in practice.

Large language models memorize and reproduce sequences from their training data, creating privacy, copyright, and security risks. Existing neuron-level mitigation methods equate editing with zeroing out neuron activations, but the activation only controls whether a neuron engages; the output vector is what writes to the residual stream and, through superposition, encodes multiple features. We propose output vector editing, a constrained-optimization weight edit that locates a small set of MLP neurons responsible for a memorized continuation and minimally modifies their output vectors to introduce a distractor in vocabulary space, redirecting their residual-stream contributions while leaving activations unchanged. Evaluating on four models from 360M to 7B parameters (SmolLM-360M, OLMo-1B, OLMo-7B, Llama2-7B), we center on OLMo-7B (whose open weights and pretraining corpus enable systematic mining) and mine 6831 memorized sequences, achieving up to 87.9% suppression. The 2.7$\times$ gap over zero ablation on the same located neurons shows the suppression comes from the output-vector edit, not localization alone. Four edit modes span a spectrum from aggressive suppression to minimal redirection; in ensemble they cover 96.5% of memorized sequences, while our recommended single-mode configuration reaches 81.5% with no catastrophic locality failures. We further identify a mechanistic boundary at ${\sim}14%$ of sequences unreachable by MLP-only editing; while these failures are not attention-driven overall, ablating the top contributing attention heads recovers 60--64% of them, with stronger recovery on continuations that copy tokens from the prefix, positioning attention as a complementary fallback rather than a primary mechanism. Edit mode ordering and the success-locality trade-off transfer across all four models, with success rates scaling with model size rather than family.

The concept of geodesic-like curves was introduced by Chen in 2010 as a method for estimating shortest paths (geodesics) on parametric surfaces, with its convergence established theoretically. However, an efficient numerical computational framework has not yet been developed. In this paper, we propose an elegant and efficient approach for computing geodesic-like curves by leveraging deep learning and Physics-Informed Neural Networks (PINNs). Under the proposed framework, not only can single parametric surfaces be handled efficiently, but a broad class of complex parametric surfaces including multi-surface systems with $C^0$ or higher continuity and surfaces of revolution can also be robustly addressed.

A/B testing has become the gold standard for data-driven decision-making in large-scale online experimentation, providing critical guidance for feature launch, pricing optimization, and user experience enhancement. To maximize statistical sensitivity, many technology companies routinely employ Controlled-experiment Using Pre-Experiment Data (CUPED), a technique that achieves substantial variance reduction while preserving the unbiasedness of estimating the average treatment effect. Despite its widespread adoption, several critical methodological and practical nuances of CUPED remain underexplored. This paper systematically addresses five frequently encountered yet overlooked questions regarding the application of CUPED. First, we provide a comparative analysis of various post-CUPED estimators to identify the optimal adjustment specification. Second, we evaluate the validity of regression-based adjustments and delineate robust variance estimation methods tailored for such frameworks. Finally, we extend our investigation to complex but common scenarios, including multi-arm experiments and two-stage sampling designs. Our findings reveal that in these settings, naive reliance on standard variance estimators can lead to severely misleading inferences. By offering rigorous theoretical insights and extensive experimental validation, this work deepens the conceptual understanding of CUPED. Notably, the recommended methodologies have been successfully deployed and integrated into ByteDance's experimentation platform.

Expressive gestures are essential for natural and effective communication, complementing speech when verbal cues alone are insufficient (e.g., pointing). For social robots such as the humanoid Pepper, producing natural and expressive movements is critical for improving human-robot interaction (HRI) and long-term acceptance. However, generating gestures remains challenging due to reliance on expert-authored animations, resulting in rigid behaviors that are impractical for dynamic and diverse environments. Alternatively, machine learning approaches often struggle to capture perceived naturalness, becoming increasingly challenging with more degrees of freedom. Consequently, producing expressive robot gestures requires a system that can adapt to the environment while adhering to social norms and physical constraints. Recent advances in large language models (LLMs) enable dynamic code generation, offering new opportunities for runtime gesture synthesis from natural language. In this paper, we integrate ChatGPT into the humanoid robot Pepper to generate co-speech gestures aligned with conversational output. While this baseline enables flexible gesture generation, the resulting motions are often perceived as stiff and unnatural. To address this limitation, we introduce an iterative reinforcement learning with human feedback (RLHF) system that finetunes gesture generation based on user evaluations, leveraging an iterative user study to compare Pepper's generated gestures. Our results show that RLHF improved the LLM's co-speech generative capabilities, producing more expressive, relevant and fluid movements.

This paper develops a formal account of what generalist agents must store in memory in order to act near-optimally across multiple environments and goals. It shows that when two domains share an observational bottleneck but require incompatible optimal actions, any uniformly near-optimal policy must induce distinct memory distributions at that bottleneck. The result yields a separation theorem: sufficiently successful agents cannot rely only on current state observations, but must preserve domain-relevant information in memory. The paper further shows that if an agent's memory contains enough information to estimate values for related goals, then that memory can be used to approximately reconstruct the agent's local transition dynamics. Together, these results characterize memory as the substrate that supports domain disambiguation, transition-model reconstruction, and planning for generalist agents.

Current large language model (LLM) inference systems universally deploy ultra-large-scale models using a combination of Tensor Parallelism (TP) and Pipeline Parallelism (PP). However, existing systems treat the model parallelism topology as a static configuration that cannot be flexibly adjusted at runtime. This rigid design creates a fundamental contradiction with the dynamically changing inference workloads in real-world scenarios. State-of-the-art systems lack online reconfiguration capabilities and can only switch configurations by restarting the service, resulting in several minutes of service interruption, KV cache loss, and prohibitive recomputation overhead. To address this problem, this paper presents ReMP, a runtime model parallelism reconfiguration framework that supports low downtime. ReMP achieves dynamic adjustment through three key techniques: (1) decoupling the model parallelism topology from runtime state to avoid full service reconstruction; (2) designing a two-dimensional KV cache migration mechanism to preserve reusable cache states after TP/PP changes; and (3) implementing end-to-end online reconfiguration. Experiments demonstrate that ReMP can complete most topology switches within 1-7 seconds on models ranging from 7B to 70B parameters, achieving speedups of tens to over a hundred times compared to the restart approach. Moreover, ReMP significantly outperforms fixed configurations under dynamic workloads, delivering superior performance in terms of TTFT, TPOT, and output throughput.

Accurate, site-specific channel information is crucial for optimizing next-generation wireless networks. Among various approaches, localized statistical channel modeling (LSCM), which models the channel multipath angular power spectrum (APS) from the reference signal received power (RSRP) measurement, has emerged as a state-of-the-art method tailored for efficient network optimization. However, despite its effectiveness, LSCM cannot predict APS at the vast majority of locations where no measurements are available, which significantly restricts its applicability in large-scale, real-world scenarios. To address this challenge, we present \emph{point-cloud-assisted tangent Gaussian splatting} (PC-TGS), the first framework to \emph{extrapolate} APS to unmeasured outdoor grids by integrating sparse radio measurements with dense LiDAR-based geometry. PC-TGS represents environmental scatterers as anisotropic 3D Gaussians, initialized and refined through a relaxed-mean reparameterization of the raw point cloud. A tangent-plane projection accurately maps each Gaussian into the local angular domain, while a depth-aware electromagnetic splatting process aggregates their contributions. To ensure practical deployment, we derive a closed-form Gaussian-weighted average (GWA) for APS bin integration and provide a provable error bound. { Evaluations on a LiDAR-scanned city-scale dataset (5M points, 6,310 RSRP samples) demonstrate that PC-TGS achieves better APS and RSRP prediction performance compared to state-of-the-art baselines and faster inference time for APS extrapolation task. These results highlight the potential of PC-TGS to enable geometry-aware and data-efficient channel prediction in large-scale wireless digital twins.

Realistic coding-agent benchmarks often replay public GitHub issues and pull requests, making them vulnerable to overlap with model pretraining, fine-tuning, synthetic-data generation, or benchmark-driven model selection. Fully synthetic tasks avoid direct historical replay, but can drift away from real repository needs. We propose SWE-Future, a forecast-conditioned data synthesis method for future-oriented coding tasks. Given a forecast snapshot at time $T_0$, the method uses only pre-$T_0$ repository evidence to forecast future feature implementation/enhancement, bugfix, and refactor task families. We first validate this forecasting step retrospectively: after forecasts are fixed, later pull requests are used only to measure whether the predicted task families match future repository work. In an 80-repository study, the forecaster achieves 58.1\% future-work relevance under the main semantic matching metric. We then use validated forecast families as conditioning signals to synthesize a 200-task coding-agent dataset across 61 repositories from a task-generation snapshot, rather than replaying the later pull requests used for validation. SWE-Future shows that repository-evolution forecasts can guide realistic, future-oriented coding-task synthesis while reducing direct dependence on historical pull-request replay.

This work presents the development of hybrid models that integrate spiking neural networks (SNNs) with components of convolutional neural networks (CNNs) to learn from simulated event-based camera data (Dynamic Vision Sensor, DVS) generated from conventional smartphone videos. Aimed primarily at human fall detection, the approach leverages the energy efficiency and spatio-temporal processing capabilities of SNNs by converting video frames into event-based data. The proposed models are evaluated through simulations on multiple datasets, comparing their performance to that of traditional machine learning models. Results demonstrate significant gains in efficiency without sacrificing accuracy, underscoring the potential of combining SNNs and DVS technology for complex tasks in real-world environments.

The Moving-Target Traveling Salesman Problem (MT-TSP) seeks a minimum cost trajectory for an agent that departs from a static depot, visits a set of moving targets, each within one of their assigned time windows, and returns to the depot. In this article, we study the Moving-Target Traveling Salesman Problem with Moving Obstacles (MT-TSP-MO), a generalization of the MT-TSP where the agent trajectory must avoid moving obstacles. We present a Mixed-Integer Conic Programming (MICP) formulation that can be solved using off-the-shelf solvers, as well as a fast and scalable Two-Phase Bilevel Search (TPBS) algorithm that computes high-quality feasible solutions for the problem. We evaluate our approaches against an existing baseline algorithm on a broad range of problem instances with up to 40 targets and 40 obstacles. The results demonstrate that both the proposed methods significantly outperform the baseline with respect to success rates, solution costs, and computation time.

Data valuation quantifies the intrinsic quality of individual samples to enable principled data curation, quality control, and robust learning. For time series in critical domains such as healthcare, finance, and industrial monitoring, effective valuation methods are essential yet fundamentally lacking. Existing approaches are either model-dependent, limiting their generalizability, or designed for i.i.d. data and thus fail to capture temporal dependencies, multi-scale patterns, and non-stationary dynamics inherent to sequential data. We introduce TimeLAVA, a learning-agnostic framework that values temporal segments by their marginal contribution to minimizing distributional discrepancy between evaluated and reference data. At its core is a novel Selective Wavelet-based Wasserstein discrepancy combining multi-scale wavelet transforms for temporal localization with unbalanced optimal transport for robustness to distributional shifts. Segment values are efficiently computed via sensitivity analysis without requiring model training and aggregated into point-wise scores. We provide theoretical guarantees linking valuation to model-agnostic generalization and prove bounded sensitivity to outlier contamination. Extensive experiments across anomaly detection, data pruning, and label noise detection demonstrate that TimeLAVA produces significantly more informative value scores than existing methods on diverse real-world datasets.

Civil litigation is inherently a life-cycle process: what a lawyer drafts on day one constrains what unfolds at trial months later. Yet existing legal benchmarks evaluate isolated subtasks, and prior legal-agent simulators reinitialize each scenario from shared ground truth, leaving cross-stage causal dependencies unmodeled. We present LegalWorld, a life-cycle interactive environment that models Chinese civil litigation as a causally connected state chain of five stages (seven sub-scenarios), grounded in 75,309 paired Chinese civil judgments. We pair it with reusable infrastructure (local memory, global case memory, a Skill/Tool library) that keeps each dispute consistent across its full life cycle. Building on this environment, we construct LongJud-Bench to evaluate agent capability across all five connected stages. 18,992 ratings from 217 legal-background evaluators confirm that LegalWorld trajectories are procedurally faithful and role-consistent; and a capability-level cross-model evaluation reveals sharp divergences that aggregate scores cannot expose, with no single backbone leading across consultation, drafting, and courtroom advocacy. Detailed resources will be released publicly.

Structurally constrained event sequence generation remains challenging because generated paths must preserve transition feasibility, temporal order, termination, and attribute consistency. In predictive process monitoring (PPM), this challenge appears as full event sequence generation, whereas existing work mainly addresses component tasks such as next activity, remaining time, outcome, and attribute prediction. This paper proposes the Graph Grounded Cross Attention Transformer Neural Network (GGATN) for this unified PPM task. GGATN uses a global process graph as structured activity memory, contextualizes sequence positions through Transformer self attention, and injects process topology through graph grounded cross attention. Unlike autoregressive decoding, GGATN generates activities, timestamps, length, and event level and sequence level attributes in a single pass, followed by Viterbi style graph constrained decoding for feasible paths and explicit termination. Experiments on six benchmark event logs show more reliable generation quality than local instruction prompted LLM baselines. GGATN achieves strong performance on sequence similarity, Damerau Levenshtein similarity, bigram based control flow similarity, and duration distribution, while maintaining zero hallucinated activities and zero sequence level attribute inconsistency. Ablation analyses confirm the global graph encoder as a stable structural prior. Interpretability analyses show how graph structure, sequence context, feedback refinement, and constrained decoding shape generation.

Intravascular ultrasound (IVUS) lumen and external elastic membrane (EEM) segmentation is important for quantitative coronary plaque burden assessment. Errors in lumen or EEM delineation directly propagate to plaque area, plaque burden and geometric measurements. However, standard methods prioritising overlap scores often suffer from boundary drift and topology errors, leading to inaccurate clinical measurements. We present GeoCat, a geometry-consistent network that processes 5-frame IVUS clips using dual Cartesian-polar encoders with cross-domain attention and temporal fusion. A differentiable geometry consistency loss directly supervises clinically relevant descriptors including diameters, orientations, and cross-sectional areas. The model is trained on 12,242 annotated frames from 146 patients acquired with two commercial IVUS systems. We evaluate performance using both segmentation accuracy and plaque-relevant clinical metrics, including Dice/IoU, boundary measures(95HD (mm), ASSD), topology violation rate, and clinical geometry errors (dmax/dmin, angles, and areas). On our dataset, GeoCat achieves a Dice of 0.93, reduces 95HD to 0.14 mm, and lowers topology violations to 1.0%. Importantly, it significantly improves geometric fidelity, yielding diameter errors of 0.13-0.16 mm and angular errors of ~8 degrees, supporting reliable plaque burden quantification.

Turkish is agglutinative: meaning is carried by morphemes, yet the subword tokenizers that drive modern language models split words by corpus statistics, fragmenting semantically loaded suffixes and -- in the case of WordPiece and rule-based analyzers -- failing to decode their output back to the original text. This paper presents \textbf{Morpheus}, a neural morpheme-boundary model for Turkish that is at once a lossless, morphology-aware tokenizer and a word-embedding producer. A differentiable Poisson-binomial dynamic program turns per-character boundary probabilities into soft morpheme memberships during training and exact segments at inference, with no string normalization, so $\mathrm{decode}(\mathrm{encode}(w)) = w$ holds by construction. Because the model is neural, the same forward pass that tokenizes also emits a structured word embedding. Among reversible tokenizers -- the only ones valid for generation -- Morpheus attains the lowest bits-per-character ($1.425$), roughly doubles the gold morphological alignment of the subword family (MorphScore macro-F1 $0.61$ vs.\ ${\sim}0.32$), and uses ${\sim}19\%$ less GPU memory than 64K-vocabulary subword tokenizers. As an embedder, frozen Morpheus vectors lead on lexical retrieval (root-family MAP $0.85$) and same-root verification (ROC-AUC $1.00$), surpassing the multilingual retriever BGE-M3 and BERTurk; on context- and inflection-dependent tasks (NER, case/number probing) the heavier contextual encoders remain ahead -- a trade-off we attribute to Morpheus's root-centric geometry. Code: https://github.com/lonewolf-rd/TurkishMorpheus; model: https://huggingface.co/lonewolflab/Morpheus-TR-50K; interactive demo: https://huggingface.co/spaces/lonewolflab/morpheus-tr-demo.

Fault-tolerant broadcasting in dense Gaussian networks is recovered by re-rooting the broadcast at a new source at maximum graph distance from the faulty nodes. This paper extends the re-rooting framework by replacing its boundary-search source-selection step with a quotient-lattice-aware algebraic construction. The first contribution is a constant-time counting method for valid new sources, formulated as an intersection of two diameter-$k$ boundary sets in the Gaussian quotient. The exact count is obtained by a fixed union of side-pair intervals over nine quotient-lattice copies, giving a closed-form procedure without scanning the network or boundary. The second contribution is a shifted direct selector for two arbitrary faulty nodes. Given faulty nodes $A$ and $B$, the problem is translated to $C=\operatorname{mod}_{G_k}(B-A)$, and the selector finds $P$ satisfying $d(P,0)=d(P,C)=k$. For each of nine quotient-lattice shifts, sixteen signed linear systems are checked. Nonparallel systems are solved via Cramer's rule; parallel systems are handled by interval-endpoint selection. At most $9\times16=144$ shifted sign cases are evaluated, giving $O(1)$ selection under the word-RAM model. Validation reports zero count mismatches over $26{,}623$ tested nodes, $500{,}000$ valid outputs over $500{,}000$ sampled fault pairs, and $40{,}000$ successful re-rooted broadcast trials. The shifted selector achieves a $5.92\times$ speedup over boundary search at $k=200$, remaining stable as $k$ increases. These results make new-source selection algebraic, bounded, and independent of network size.

Fault-tolerant broadcasting in dense Eisenstein--Jacobi networks requires efficient recovery when faulty nodes disrupt the original broadcast structure. A re-rooting-based method guarantees that, for any two faulty nodes, a valid new source exists at maximum graph distance from both faults. However, identifying such a source without scanning the network or testing all boundary candidates remains an open practical problem. This paper presents a closed-form, constant-time algorithm for counting and selecting a valid new source in dense Eisenstein--Jacobi networks under two node faults. The two-fault problem is reduced to a boundary-intersection problem involving the origin and a difference node. The distance-$t$ boundary, where $t$ is the network diameter, is partitioned into six directed sides of the Eisenstein--Jacobi hexagon. Since the network is a quotient structure, intersection equations are solved modulo the defining lattice, requiring evaluation of seven quotient-lattice shifts across all $6\times 6$ side pairs, yielding at most $252$ algebraic systems. The first algorithm counts all valid new sources for faults at $0$ and $A$. The second algorithm selects one valid new source for arbitrary fault pairs by solving translated side-pair systems, verifying each candidate, and shifting back. Each system is either a non-parallel $2\times 2$ linear system with at most one candidate, or a parallel system whose feasible candidates form an integer interval. Both algorithms run in $O(1)$ time under the fixed-word arithmetic model. Computational validation over $500{,}000$ sampled fault pairs and $40{,}000$ re-rooting trials confirms correctness: the selector always returns a valid new source, and the recovered broadcast reaches all non-faulty nodes.

Photonic quantum machine learning offers a route to trainable physical representations built from phase, interference and measurement. However, its role in scientific machine learning remains largely unexplored. Physics-informed neural fields provide a natural setting, because differential equations require trial spaces that preserve phase, frequency and derivative structure. Here we introduce a photonic quantum neural field in which coordinates become trainable optical phases, are mixed by multi-photon Fock-space interference and are decoded from photon-number measurements. The photonic circuit is optimized as the neural-field representation itself, not as a fixed feature map or hardware accelerator. Photonic measurement is therefore a trainable representation on which the physics-informed residual is minimized. Across seven elliptic, wave, nonlinear dispersive and inverse PDE benchmarks, we observe a phase-complexity transition: classical coordinate and Fourier-feature networks suffice in smooth regimes, whereas the photonic field is most accurate when residual derivatives amplify phase mismatch. In the hardest regimes it gives the lowest errors, with margins reaching an order of magnitude and about one quarter of the trainable parameters of classical baselines. Frozen and shuffled controls, together with noise stress tests, attribute this gain to learned interference and stable Fock-probability readout under compound perturbations. These results identify photonic quantum measurement as a representation-learning principle for scientific machine learning.

Dense Eisenstein--Jacobi networks are degree-six algebraic interconnection topologies with regular structure, vertex symmetry, small diameter, and efficient communication algorithms. These properties make them suitable for parallel and on-chip communication systems in which collective operations such as one-to-all broadcasting are frequent. Existing optimal broadcasting algorithms for dense hexagonal/Eisenstein--Jacobi networks assume fault-free operation. However, a faulty internal forwarding node may interrupt message propagation and prevent complete delivery. This paper proposes a lightweight re-rooting-based fault-tolerant broadcasting method for dense Eisenstein--Jacobi networks. The main idea is to relocate the effective broadcast source to a new source node such that each faulty node is located at graph distance equal to the network diameter from the new source. Consequently, faulty nodes become leaf-level nodes in the broadcast process and are not required to forward the message. We present source-selection algorithms for one- and two-node failures and prove that for any pair of faulty nodes in a dense Eisenstein--Jacobi network there exists a common distance-diameter node that can serve as a valid re-rooted source. The source-selection procedure requires linear time in the network diameter. Equivalently, since $N=3t^2+3t+1$, the selection cost is $O(\sqrt{N})$ in the number of nodes. Since the standard one-to-all broadcast completes in one diameter time and the relocation phase is also bounded by one diameter, the proposed method completes in at most twice the network diameter. We also show that the two-fault guarantee does not generally extend to arbitrary three-fault configurations by giving an explicit counterexample. The proposed approach improves broadcast reliability without constructing redundant spanning trees, backup paths, or additional broadcast structures.

Item discrimination is a fundamental psychometric property of educational assessment, which measures whether an item meaningfully distinguishes students with higher proficiency from students with lower proficiency. While various existing works have explored whether large language models (LLMs) can estimate item difficulty, it remains unclear whether they can capture item discrimination. In this work, we evaluate 42 proprietary and open-weight LLMs in zero-shot settings using two complementary approaches: direct discrimination prediction, where models explicitly estimate an item's discrimination value from its content, and response-based Classical Test Theory (CTT) calibration, where LLM answers are treated as synthetic student responses to compute discrimination scores. Our results show that direct prediction yields weak alignment with human-calibrated discrimination: the best-performing model reaches only a Spearman correlation of 0.152. Response-based CTT calibration provides a stronger but still limited signal, with the all-persona synthetic respondent pool reaching a Spearman correlation of 0.241. These findings highlight item discrimination as an open challenge for LLM-based psychometric evaluation: current LLMs contain non-random discrimination-relevant signal, but they do not yet reliably capture how assessment items distinguish human students.

Pretrained biological language models expose per-token probability distributions through masked-token prediction, providing the likelihood interface central to sequence design, variant scoring, and mechanistic interpretation. Yet these distributions are learned from broad unlabeled corpora and are not naturally conditioned on task-specific biological contexts such as interaction partners, cellular environments, or therapeutic interventions. Existing contextual matching methods often distort this interface through pooled embeddings, contrastive latent spaces, or task-specific prediction heads. We introduce LOGICA (Logit-space Contrastive Alignment), a framework for context-conditioned prediction that performs contrastive learning directly in output-logit space. Using gated cross-modal adapters compatible with each model's native token head, LOGICA preserves the pretrained likelihood interface and converts contextualized token log-likelihoods into matching scores. Alignment is defined through context-sensitive token probabilities rather than proximity in a shared embedding space, enabling learning from sparse paired data across models with distinct vocabularies, without a shared tokenizer or decoder. LOGICA is particularly effective for mutation-local variant ranking, where comparisons reduce to context-conditioned likelihoods of mutant tokens at perturbed sites. Across protein--ligand binding, TCR--peptide activity, and drug-conditioned resistance prediction, LOGICA improves over prior state-of-the-art methods, including matched latent-contrastive and conditional MLM baselines, while retaining a token-level interface for interpretation and generation. On held-out-gene single-mutation drug-resistance prediction, LOGICA improves AUC from near-random latent-space baselines of $\sim$0.55 to $\sim$0.65.

Large language models (LLMs) have shown impressive capabilities across diverse tasks, yet their performance on jurisdiction-specific legal reasoning remains underexplored. We present TW-LegalBench that utilizes Taiwanese legal system's rich official corpus open to the public to fill the gap in evaluating LLMs on Taiwanese law, among common-law benchmarks that focus on English sources and civil-law benchmarks focusing on sources of Simplified Chinese. TW-LegalBench comprises three task types: (1) over 16,000 multiple-choice questions (MCQs) across five years of official examinations in 18 professional domains; (2) 117 open-ended essay questions (OEQs) from examinations for legal professionals with official scoring rubrics; and (3) more than 14,000 legal judgment prediction (LJP) instances covering hundreds of crime categories. We evaluate 13 LLMs using accuracy for MCQs, a decomposed LLM-as-Judge framework based on the scoring rubric points for OEQs, and metrics for sentencing accuracy and statute citation for LJP. Our results reveal that top-performing models exceed the passing threshold for qualified lawyers (passing rate: 11%) but fall short of that for judges and prosecutors (passing rate: 1~2%). For LJP, while models demonstrate reasonable verdict type accuracy and sentence prediction capability, they struggle to cite exact legal articles. These findings highlight that reliable legal text generation remains challenging for LLMs, even though their performance on qualification examinations approaches human level.

The energy-based method remains a comparatively underexamined approach for surface classification in mobile robotics, despite promising results in constrained environments. This study evaluated the viability of using energy-derived features as either a standalone classification modality or as supplementary input to inertial data. A comprehensive evaluation was conducted across three publicly available datasets, comparing the performance of modern deep learning architectures including recurrent neural networks, convolutional neural networks, encoder-only transformers, and Mamba state-space models, under automated hyperparameter tuning and input sequence length optimization. The models achieved higher accuracy than previously reported values on all evaluated datasets, with the convolutional neural network yielding the highest overall performance. When relying exclusively on energy-based features, the models attained classification accuracies in the range of 85-90%, approximately 5-10% lower than those achieved when combined with inertial features (96-99%). Augmenting inertial data with energy features resulted in a consistent mean accuracy improvement of 1-2%. These findings indicate that classifiers relying solely on energy features offer sufficient accuracy for standalone deployment, while also providing a consistent gain when used in combination with other sensing modalities.

Model-based learning agents use learned world models to predict future states, plan actions, and adapt to new environments. However, the process of updating world models from collected experience creates a training-time attack surface: adversarially poisoned fine-tuning trajectories can manipulate the learned dynamics and thereby corrupt downstream planning. In this paper, we propose SWAAP, the first two-stage data poisoning framework for learned world models. In the first stage, SWAAP identifies a harmful target world model that induces low-return behavior under planning while remaining close to clean dynamics, using first-order bilevel optimization enabled by a transition-gradient theorem. In the second stage, SWAAP realizes this target through stealth-constrained gradient matching, modifying only a limited fraction of fine-tuning transition targets so that the induced training gradients steer the victim model toward the adversarial target, while a prediction-error regularizer encourages the poisoned targets to remain close to the world model's natural approximation error. To assess attack stealthiness, we evaluate defenses and detectability across three stages of the poisoning pipeline: pre-training detection of poisoned transitions, robust training during fine-tuning, and test-time monitoring of the resulting world model. Across diverse continuous-control tasks, SWAAP causes substantial performance degradation while keeping poisoned transitions close to clean data and evading the evaluated non-adaptive residual/CUSUM/TRIM-style defenses. These results reveal a practical vulnerability in world-model adaptation pipelines and highlight the need for robustness methods that protect both world-model training data and learned dynamics.

A network of oscillators that synchronizes perfectly computes nothing further, so an attention architecture built from synchronization must locate its computation in structured departures from agreement. We introduce the Frustrated Synchronization Network (FSN), whose token states are phases on a torus and whose entire value pathway is one learned complex coupling kernel over harmonics and a one-step delay. Each component of the kernel is a frustration in the sense of the synchronization literature. The complex phases are static Kuramoto-Sakaguchi frustration angles, the signed harmonics are repulsive Daido components, and the delay term, which couples each token to the successors of the tokens it attends to, is algebraically identical to Kuramoto-Sakaguchi coupling whose frustration angle is the data's own transition, so next-token prediction is implemented as synchronization frustrated by the data. At matched one-million-parameter and training budgets on character-level text and code, the FSN's validation loss is below a tuned RoPE-SwiGLU transformer's at every epoch measured, and the comparison survives training the baseline to convergence: every thirty-epoch enwik8 seed finishes below the transformer's converged fifty-epoch loss of 1.611, and the FSN's completed fifty-epoch runs converge to 1.5953 +/- 0.0014. A variant with every feed-forward block replaced by mean-field coupling to learned collective modes, leaving no multilayer perceptron in the stack, tracks the transformer. On natural text the unfrustrated base layer falls behind the converged transformer at every copy depth, worst on long-range copy events; the kernel reverses the deficit at every depth of four and beyond. Headline comparisons are at the one-million-parameter scale; a scale ladder is complete through four million parameters with the advantage persisting, and remaining arms are marked as in progress.

Pre-trained materials foundation models, or machine learning interatomic potentials, leverage general physicochemical knowledge to effectively approximate potential energy surfaces. However, they often require domain-specific calibration due to physicochemical diversity as well as mismatches between practical computational settings and those used in constructing the pre-training data. To address this, we propose a sparsity-promoting fine-tuning method that selectively updates model parameters by exploiting the structural properties of E(3)-equivariant materials foundation models. On energy and force prediction tasks across molecular and crystalline benchmarks, our method matches or surpasses full fine-tuning and equivariant low-rank adaptation while updating only $\sim$3~\% of parameters, and in some cases as little as $\sim$0.5~\%. Beyond energy and force calibration, we further demonstrate task generalizability by applying our method to magnetic moment prediction and magnetism-aware total energy modeling. Finally, analysis of sparsity patterns reveals physically interpretable signatures, such as enhanced $d$-orbital contributions in transition metal systems. Overall, our results establish sparsity-promoting fine-tuning as a flexible and interpretable method for domain specialization of equivariant materials foundation models.

Joint Embedding Predictive Architectures (JEPAs) are a leading approach to world model representation learning. We identify a failure mode in JEPA-based world models grounded against two qualitatively distinct external signals: physical dynamics (sparse, high-magnitude, constraint-satisfying gradient corrections) and social-behavioral dynamics (diffuse, distribution-matching corrections). We term this Objective Interference Collapse (OIC): we argue that joint learning in a shared latent space causes the dominant channel to systematically collapse the subordinate channel's representational subspace, in a manner not resolvable by loss weighting alone. We propose Dual-Channel Grounded World Modeling (DCGWM), designed to structurally prevent OIC through a partitioned latent space (physical subspace Z_p, behavioral subspace Z_b) with inward-only gradient flow. A Physical Grounding Channel updates only Z_p via VICReg-style alignment to physical measurements; a Social-Behavioral Grounding Channel updates only Z_b via alignment to trajectories from an emergent multi-agent simulation. An Inter-Channel Interface Module couples the subspaces at the task level without cross-subspace gradients. An Asymmetric Grounding Adherence Loss penalizes rollout drift with a hard hinge for physical violations and a soft KL for behavioral divergence. A Generative Rendering Layer is architecturally isolated from the latent world model. We present three theoretical results: the partition removes the gradient-interference pathway implicated in OIC; each grounded subspace inherits anti-collapse guarantees from its alignment objective; and generative isolation is necessary under a stated assumption on the generative objective's geometry. This manuscript establishes the problem formulation and architecture; experimental validation is ongoing and will be reported in a future revision.

Forecasting benchmarks for general-purpose AI systems usually inherit the constraints of the real world: outcomes resolve slowly, tail events are rare, and counterfactual questions are difficult to score. We introduce ForecastBench-Sim, a simulated-world forecasting benchmark built on game rollouts from Freeciv, a turn-based strategy game modelled on the Civilization series. Forecasters receive a fixed world report (a structured snapshot of the current game state) and answer questions about hidden future states; the benchmark then continues the simulation and scores forecasts. Because the world is simulated, the same setup can generate continuous or binary forecasting questions at arbitrary time horizons, paired intervention worlds for conditional or causal questions, and resolved examples of rare or disruptive outcomes. We describe the benchmark pipeline, question families, scoring protocol, and release artifacts, and report validation slices from model evaluations and an anonymized human pilot. ForecastBench-Sim is intended to complement real-world forecasting benchmarks by providing controlled, immediately resolvable tasks for studying probabilistic reasoning under dynamic world states.

We study the problem of fair online resource allocation, motivated by applications such as refugee resettlement and airline scheduling, where agents arrive sequentially and must be assigned to facilities with limited capacities. We introduce a model that maximizes the overall welfare subject to resource constraints and a Lipschitz fairness requirement, which ensures that similar agents arriving in the same batch receive similar expected outcomes. We first analyze the offline problem, proving that the value of the optimal fair allocation is at least an $Ω(1/γ)$ fraction of the optimal unfair allocation, where $γ$ is the fairness coefficient, thereby bounding the price of fairness. For the online setting, we propose an algorithm based on dual mirror descent that enforces fairness constraints within batches while estimating optimal dual variables. We prove that this algorithm achieves sublinear regret relative to the optimal offline fluid benchmark. Finally, we validate our theoretical results using real-world data from the Refugee Economies Programme, demonstrating the algorithm's performance and examining the trade-offs between welfare maximization and fairness enforcement.

Tabular stream learning requires predictions on sequentially arriving examples under distribution shift. While standard methods adapt by updating model states, tabular foundation models (TFMs) make predictions conditioned on a labeled context in an in-context manner, making them a natural alternative for stream learning. This shifts the challenge from how to update the model to how to manage the context. We propose a future information view that yields three practical requirements for context management: preserve recent examples, retain uncertain examples, and remove redundant examples. We instantiate these requirements as CURE (Context management via Uncertainty-aware admission and Redundancy aware Eviction), a context-managing policy with entropy-gated admission and redundancy-aware eviction. Across seven streams, CURE shows up to 27.0% relative improvement over classical stream learners, remains robust across multiple TFM backbones, and ranks first among other policy variants. Code and datasets are available at https://github.com/morcellinus/CURE-ICML-FMSD.

Training-free neural architecture search promises efficient discovery of high-performance networks without costly training. However, existing zero-cost proxies rely on fragmented heuristics that fail to capture the fundamental question: what makes an architecture trainable? This paper introduces Intrinsic Trainability (InTrain), a unified theoretical proxy that formalizes trainability as an architectural invariant emerging from two synergistic components: geometric capacity and optimization resilience. We operationalize intrinsic trainability through analysis of neural information processing. Geometric capacity is quantified via the participation ratio of activation covariance eigenspectrum, capturing the effective dimensionality of representation manifolds. Optimization resilience is measured through cumulative gradient health, assessing the robustness of backpropagation across network depth. InTrain synthesizes these dimensions through a scale-invariant multiplicative coupling, which we hypothesize is essential for capturing their synergistic, non-additive relationship. Extensive experiments on standard NAS benchmarks and search spaces demonstrate that InTrain achieves ranking correlations on par with state-of-the-art ensemble-based proxies and outperforms other single-metric methods.

Single-cell RNA sequencing (scRNA-seq) serves a pivotal role in characterizing gene expression at the cellular level, enabling the identification of cell types and advancing the understanding of cellular heterogeneity. Despite the significant progress in scRNA-seq data clustering, we argue that current methods always ignore the sparsity and noise, as well as the complex intercellular structural information inherent in scRNA-seq data. Toward this end, in this paper, we propose a novel single-cell RNA-seq clustering framework via deep Siamese Graph Transformer Network (termed scGTN), which explicitly integrates gene expression profile and intercellular structural dependencies for cell clustering. In particular, we formulate scRNA-seq data as a graph and construct two augmented graph views that serve as dual views to capture complementary intercellular information. Then, a Siamese graph transformer network is employed to explicitly incorporate shortest-path information and node-wise distances for capturing richer structural relationships between cells. Finally, we employ an optimal transport strategy to guide the cell clustering in a self-supervised manner. Extensive experiments on multiple benchmark scRNA-seq datasets demonstrate that our scGTN consistently outperforms existing methods. Our code is available at https://github.com/W-RMSL/scGTN.

In large-scale enterprise settings, centralized multi-agent systems (MAS) are increasingly adopted, in which a coordinator delegates user requests to lightweight, domain-specialized sub-agents. While this architecture improves modularity, scalability, and cost efficiency, its reliability depends not only on accurate routing but also on sub-agents' ability to calibrate their responses to capability constraints. In particular, sub-agents built on smaller fine-tuned models often struggle with such calibration, leading them to over-answer ambiguous, underspecified, misrouted, or unsupported requests and produce hallucinated outputs instead of actionable feedback. To address this challenge, we present EARS (Explanatory Abstention for Reliable Sub-Agent Modeling), a production-oriented framework that reframes sub-agent abstention as an inter-agent communication protocol: a sub-agent does not merely abstain, but exposes an actionable failure state to the coordinator. EARS curates human-agent interaction data using an ensemble of calibrated LLM-as-a-Judge models, producing structured abstention labels and rationales under a taxonomy of sub-agent failure modes. These data are used to fine-tune sub-agents to detect failure conditions and return rationales for coordinator-level clarification, rerouting, or fallback. We evaluate EARS in a large-scale production e-commerce assistant supporting enterprise business intelligence workflows. EARS improves the overall response pass rate from 68.5% to 78.9%, demonstrating that sub-agent-side explanatory abstention improves MAS reliability.

Reliable sound source localization is fundamental to robot audition, enabling autonomous robots to perceive spatial cues and operate effectively in dynamic environments. Classical methods such as Multiple Signal Classification (MUSIC) offer strong theoretical foundations but degrade under low signal-to-noise ratios. While deep learning-based approaches achieve promising performance, they often struggle with limited generalization across conditions. To address these challenges, we propose NeuralMUSIC, a hybrid neural-subspace framework for robotic sound source localization. Specifically, a neural network first estimates the spatial covariance matrix from multichannel microphone observations. The predicted covariance is then integrated into a classical MUSIC pipeline with eigenvalue decomposition (EVD) and pseudo-spectrum computation, followed by a Frequency Attention Fusion (FAF) module to produce the final DOA estimates. To improve data efficiency, we further introduce a Self-supervised Spatial Correlation Learning (SSCL) strategy that leverages unlabeled acoustic data to capture spatial structure. Extensive experiments across different robotic tasks demonstrate that NeuralMUSIC achieves competitive localization accuracy while exhibiting improved robustness and cross-domain generalization.

Data mixture selection is critical for Large Language Model pretraining. Existing methods such as RegMix select a single static mixture by fitting a regression model on small-scale proxy runs. We propose RegMix-D, a simple extension of RegMix to dynamic mixing. Our key observation is that proxy runs produce not only endpoint losses, but also full loss trajectories, which can be used to further improve data mixture. By training regression model on these trajectories, we can predict optimal mixtures at multiple training stages. RegMix-D supports two deployment modes: an offline variant that generates a complete mixture schedule before target training, and an online variant that adapts the mixture during training using observed loss. Experiments on 25B tokens of the Pile dataset with a 1B parameter target model show that RegMix-D consistently improves over RegMix and DoReMi across 13 downstream tasks while remaining proxy-efficient: it surpasses RegMix even with only 128 proxy models (25% of RegMix's proxy compute budget).

Intelligent landslide hazard interpretation is critical for disaster prevention, yet current paradigms struggle to simultaneously extract visual features and high-level geoscientific semantics, while general-purpose vision-language models (VLMs) suffer from perceptual limitations and domain hallucinations in complex geological scenarios. To address these challenges, we propose an instruction-driven agentic framework comprising three components. First, LandslideBench, a multimodal fine-grained dataset with seven subtype labels, high-resolution imagery, pixel-level masks, and high-quality textual descriptions, is constructed via multi-VLM cross-validation and interactive annotation. Then, LandslideVLM, a landslide-oriented VLM, is fine-tuned via LoRA on LandslideBench to enhance geological semantic understanding. Finally, LandslideAgent, a domain rule-enhanced agent taking LandslideVLM as its cognitive backbone, employs a dual-rule controller incorporating structured report metadata constraints and cross-validation identification constraints to regulate automated tool invocation. Experiments demonstrate that LandslideBench provides effective baselines across five mainstream models on fine-grained classification and semantic segmentation. LandslideVLM achieves accuracy improvements of 10.96%, 32.87%, and 15.91% on landslide discrimination, fine-grained classification, and semantic description quality, respectively. LandslideAgent further enables autonomous multi-source spatial data inference, realizing full-process intelligence for landslide identification and analysis.

Warning: This paper studies stereotypes and biases, and contains potentially disturbing examples, used for illustration purposes only. Our findings should not be interpreted as an argument against alignment. Instead, this paper highlights the need for principled approaches to more advanced alignment. Alignment aims to ensure that large language models (LLMs) behave safely and reliably, including by avoiding unsafe inferences. However, we show that such safety-oriented behaviors can misfire: models may reject warranted conclusions even when they are explicitly supported by context. We call this failure mode misfired alignment, where alignment-induced changes cause LLMs to override explicit evidence. To quantify this phenomenon, specifically on stereotype-related alignment, we introduce VETO, a benchmark consisting of 2,032 BBQ-derived contrastive pairs, and define a new metric, Misfired Alignment Rate (MAR), which measures on a 0 to 100 scale how often a model fails on a stereotype-related question but succeeds on its contrastive counterpart. We benchmark 25 LLMs on VETO, and show that all LLMs, including the most recent ones, exhibit non-trivial (4.7 to 18.9%) MARs while all human participants achieve 0.0% MAR. Controlled priming experiments further show that alignment-induced cues can substantially amplify MAR across LLMs, indicating that these failures are not merely artifacts of individual examples but can be induced by safety-related framing. Mechanistic analyses on open-weight LLMs reveal late-layer suppression of evidence-supported answers, and comparisons between instruct and base LLMs suggest that this suppression emerges after instruction training. These findings show that current alignment methods can overgeneralize surface-level safety cues, to the point of overriding objective evidence, motivating more work on alignment objectives that better preserve contextual grounding.

As Large Language Model (LLM) datasets scale to trillions of tokens, data selection has emerged as a critical frontier to filter out uninformative noise and construct adaptive learning trajectories. Beyond static heuristic filtering, advanced data selection methods for LLM training largely follow two paradigms, each with fundamental limitations. Influence-based methods provide principled bi-level objectives but require intractable inverse-Hessian computations, while excess-loss methods are computationally efficient but rely on a static reference model that becomes misaligned with the evolving proxy model during training. We propose BLADE (Bi-Level Adaptive Data sElection), a Hessian-free framework for data selection. BLADE reformulates the bi-level optimization problem underlying influence-based methods as a penalized single-level objective via Lagrange multipliers, avoiding inverse-Hessian computation while revealing a principled connection to excess-loss based data selection. The resulting objective recovers an excess-loss form but replaces the static reference model with a dynamic one that stays synchronized with training. Theoretically, we prove that this penalized formulation guarantees first-order convergence. For efficient online batch selection, we instantiate BLADE as a memoryless randomized block-coordinate Frank-Wolfe algorithm. Extensive experiments show that BLADE consistently outperforms state-of-the-art data selection baselines, providing a practical recipe for LLM training.

Large language models (LLMs) are increasingly deployed in hiring workflows, yet most research on gender bias in LLM hiring decisions has focused on English-language, Western-format resumes. This study examines whether pro-female gender bias extends to a Japanese corporate context and evaluates two practical mitigation strategies. Using a counterfactual resume design with 60 Japanese rirekisho-format resumes, 12 name pairs selected on linguistically grounded gender-signal criteria, and five state-of-the-art LLMs (Claude Sonnet 4.6, GPT-4o, DeepSeek-V3, Gemini 2.5 Flash, Llama 3.3 70B), we conducted 43,200 API calls across baseline, prompt instruction, and privacy filter conditions. A crossed random-effects linear mixed model confirms a significant pro-female bias across all five models, replicating Western findings in a non-Western context. A prompt-level gender-neutrality instruction produces no meaningful reduction in bias. A name-reliance analysis formally identifies the candidate name as the primary gender channel: removing the name from the prompt reduces the female effect by nearly its full magnitude. An unexpected incompatibility between the privacy filter and GPT-4o's content safety filter, resulting in a 42% refusal rate, highlights a practical deployment challenge for name anonymization in LLM-assisted recruitment pipelines.

Dysarthria is a speech disorder marked by reduced intelligibility and communicative effectiveness. Automatic utterance-level assessment of dysarthric speech can support scalable speech monitoring and therapy-related analysis. Yet training such systems is bottlenecked by the scarcity of clinically annotated dysarthric speech. This work proposes to augment dysarthric speech assessment using data from speech synthesis evaluations, specifically human-annotated utterances with Mean Opinion Score (MOS) labels from the QualiSpeech corpus. Experiments show that fine-tuning on speech synthesis assessment data consistently improves performance on both intelligibility and naturalness prediction, while joint training yields gains primarily on naturalness. These results suggest that synthesis artifacts and dysarthric speech share perceptual commonalities, and speech synthesis evaluation corpora offer a practical augmentation source that reduces reliance on scarce clinical annotations.

Hybrid classical-quantum computing requires frequent data exchange between classical processors and quantum control hardware. However, existing superconducting quantum control systems are commonly connected through loosely coupled interfaces such as Ethernet, resulting in high communication latency and limited task throughput. To address this issue, we present HI-HCQC, an RFSoC-based hardware interface for tightly coupled hybrid classical-quantum computing. HI-HCQC integrates high-speed RF-DACs, RF-ADCs, programmable logic, embedded processors, clock synchronization circuits, and a PCIe Gen3 x8 interface, enabling direct microwave pulse synthesis, qubit readout, and high-throughput data transfer between host servers and quantum measurement-control units. Experimental results show that HI-HCQC supports six control channels and one multiplexed readout channel, achieves stable microwave generation and acquisition, and successfully performs qubit spectroscopy, Rabi oscillation, T1 measurement, single-shot readout, randomized benchmarking, and CZ-gate characterization. Compared with a conventional control system, HI-HCQC reduces end-to-end execution latency for representative quantum gate and circuit tasks and significantly improves task throughput. These results demonstrate that PCIe-coupled RFSoC control hardware provides a practical foundation for scalable and efficient hybrid classical-quantum computing systems.

Glucose forecasting algorithms are an important aspect of glycemic control management in type 1 diabetes. So far, the research community has developed numerous algorithms and models for forecasting. However, it is well-recognized that the lack of standardized model performance evaluation benchmarks makes fair comparison difficult and hinders further innovation, and thus benchmark standardization is in urgent need. Furthermore, many published glucose forecasting algorithms are limited to CGM data alone, ignoring other multimodal signals such as insulin dosing and carbohydrate intake. Here, we introduce MetaboNet-Bench, a benchmark for multimodal glucose forecasting for patients with type 1 diabetes that provides an extensible open-source evaluation framework for comparison of glucose forecasting algorithms that leverage glucose, insulin, and carbohydrate data. We then demonstrate its utility by benchmarking several recently published glucose forecasting models and a custom multimodal time-series model, representing different model architectures. The results show that the benefit of adding data modalities is conditioned on the complexity of the model and that incorporating more clinical metrics helps identify meaningful gaps to fill for future research.

Recent advancements in Large Language Models (LLMs) have empowered home assistants with natural language interaction capabilities. However, current assistants overlook the progressive omission that occurs in human dialogue as shared context accumulates, leading to more elliptical expressions for efficient communication. Thus, current assistants still struggle to interpret such elliptical expressions accurately, which limits their effectiveness in real-world applications. In practical smart home scenarios, assistants face two major challenges caused by elliptical commands: (1) referential ambiguity caused by different environmental expectations among multiple users; and (2) intention ambiguity resulting from user preferences that evolve over time or change with the environment. To address these challenges, we introduce PEC-Home, the first simulated home dataset specifically designed for interpreting progressively elliptical commands in smart homes. Extensive experiments on various LLMs, including GPT-4o, show that existing home assistants struggle to execute user-intended operations based solely on elliptical commands. Even when equipped with tools for storing and retrieving user dialogue history, execution accuracy remains below that achieved with complete commands.}.

To locate a target object while exploring the unknown environment is a fundamental capability for autonomous agents, with applications ranging from search-and-rescue to field robots. A simplified version of such task is Object Goal Navigation (ObjNav). In ObjNav, successful arrival at the target object provides a basic measure of performance; however, the efficiency of the navigation trajectory is equally important, as it indicates how intelligently the agent explores and how much time remains for subsequent tasks. In unknown environments, the key to efficient navigation lies in deciding where to explore next. While many prior works aim to address this core challenge and achieved promising performance in certain settings, recent training-based models and non-training frameworks still suffer from generalization and efficiency issues respectively, which in the worst cases can lead to excessive exploration of already-visited areas or redundant back-and-forth motion. We evaluate EffiNav on two widely used simulation benchmarks Habitat Matterport 3D (HM3D) and Open-Vocabulary Object goal Navigation (OVON), and further validate its effectiveness on physical robots in real-world settings. We conduct failure analysis on massive simulation episodes. With minimal modification, we also extend EffiNav to a memory-augmented ObjNav task on the GOAT-BENCH dataset, demonstrating its adaptability beyond standard ObjNav settings. Across two standard metrics--Success Rate (SR) and Success weighted by Path Length (SPL), EffiNav matches or outperforms recent baselines, reflecting its efficiency, robustness, and practical applicability. Recognizing the different emphases of the two datasets, the performances reveals this framework is more balanced and generalizable for efficient ObjNav.

Effective programming education requires personalized instruction adapted to diverse learner backgrounds. However, while LLM-based multi-agent systems (MAS) excel at complex planning, existing planners often lack profile-grounding and pedagogical scaffolding, thereby undermining personalized programming learning. To fill in the gap, we first introduce \textbf{MAP-PPL} (\textbf{M}ulti-\textbf{A}gent \textbf{P}lans for \textbf{P}ersonalized \textbf{P}rogramming \textbf{L}earning), a profile-conditioned multi-agent planning dataset with 3{,}043 query--profile--plan instances from 1{,}730 Stack Overflow question groups and 2{,}738 learner profiles. Each plan specifies agents, subtasks, executable steps, and prerequisite dependencies. Then, we propose \textbf{PersonalPlan}, a two-stage MAS planner that first performs hierarchical SFT with separate LoRA adapters for profile-aware task decomposition and step dependency planning, then applies a Reward-Adaptive GRPO to encourage the model to generate executable, personalized, and pedagogically scaffolded plans. Extensive experiments on MAP-PPL comparing PersonalPlan against frontier LLMs, generic MAS frameworks, and agentic planners demonstrate its superiority. With only 8B and 32B variants, PersonalPlan achieves state-of-the-art plan executability, personalization, and pedagogical quality, effectively orchestrating MAS for agent-student interactions.

Model merging has emerged as a training-free alternative to multi-task learning, aiming to combine multiple task-specific fine-tuned models into a single multi-task model. Most existing model merging approaches follow the Task Arithmetic paradigm, which decomposes fine-tuned weights into pre-trained parameters and task vectors, and performs merging exclusively in the task-vector space. The effectiveness of this paradigm implicitly relies on the assumption that task-specific knowledge is encoded solely within task vectors. We argue that this assumption generally does not hold due to the intrinsic task preferences of pre-trained models. Specifically, we identify \textbf{Load-Bearing Wall (LBW) dimensions}, namely some task-critical knowledge that remains embedded in the pre-trained weights rather than being fully transferred into task vectors. We characterize LBW dimensions from both scalar-weight and subspace perspectives, thereby covering the major paradigms of existing model merging methods. Our analysis reveals that, by ignoring LBW dimensions, task-vector-based approaches fail to fully resolve task conflicts and may inadvertently damage task-specific knowledge encoded in the pre-trained model, leading to degradation. To address this issue, we propose PACT, which preserves the anchored task-specific cores (i.e., LBW dimensions) within task vectors by aligning their orthogonal complements with the subspace of the pre-trained weights. These aligned subspace components are then removed from the task vectors before applying existing model merging algorithms. Furthermore, we develop an efficient variant based on randomized SVD to improve scalability. PACT can be seamlessly integrated with existing methods. Extensive experiments across multiple benchmarks demonstrate that PACT consistently enhances mainstream model merging approaches and establishes new state-of-the-art performance.

Natural language understanding often depends on meanings that are implied rather than explicitly stated, requiring pragmatic reasoning. Despite strong performance on math and logical reasoning, large language models (LLMs) still struggle with making pragmatic inferences, often choosing literal interpretations. To improve LLM pragmatic reasoning, we introduce PragReST, a self-supervised framework that constructs pragmatic QA data, generates counterfactual reasoning traces, and trains models to internalize them through supervised fine-tuning and reinforcement learning, without human-labeled training data or distillation from a stronger teacher. Across four pragmatic benchmarks (PragMega, Ludwig, MetoQA, and AltPrag), PragReST improves over backbone models, task-specific pragmatic tuning baselines, and non-counterfactual variants of the same pipeline. On accuracy-based benchmarks, PragReST improves over the instruct backbone by 5.37 and 5.50% (absolute) for Qwen3-8B and Qwen3-14B, respectively. Our error analysis and ablations underscore the importance of counterfactual reasoning: PragReST primarily reduces errors caused by failures to contrast observed utterances with plausible alternatives, and removing counterfactual reasoning substantially reduces performance. Moreover, our training preserves out-of-domain performance on general-knowledge and mathematical reasoning benchmarks.

Relational databases are widely used for managing structured data in real-world systems. Detecting anomalies from such relational data is crucial for identifying fraud, risks, and abnormal behaviors, yet remains under-explored. The key challenges lie in the intrinsic complexity of relational data: multi-table attributes are high-dimensional and heterogeneous, making sparse abnormal clues easy to overwhelm by normal or irrelevant information; and anomalies may further manifest as abnormal connection patterns across different foreign-key relations, which existing tabular and graph anomaly detection methods are ill-suited to capture. To address them, we propose RelAD, a reconstruction-based framework that captures anomalies from both attribute and relational edge reconstruction. RelAD contains two core modules: conditional sparse-gated attribute reconstruction, which suppresses redundant multi-table attributes and emphasizes abnormal semantic blocks, and dual-view multi-relational edge reconstruction, which detects relation-specific abnormal connections from both intrinsic and behavioral entity profiles. The resulting attribute and relational signals are integrated through a lightweight fusion module to produce the final anomaly score. We further construct 6 benchmark datasets with systematic anomalies, on which extensive experiments show that RelAD consistently outperforms other baselines while achieving competitive efficiency.

Existing information extraction (IE) tasks increasingly adopt in-context learning (ICL) with large language models. However, current approaches either show inconsistent performance across model scales or lack systematic optimization and generalizability. Building on this, we propose BCL (Bayesian In-Context Learning Framework for Information Extraction), the first optimization framework that uses particle filtering with Bayesian updates to systematically refine label representations across IE tasks. Through four steps initialization, observation, weight update, and resampling, BCL generalizes to both sequence labeling and relation classification paradigms. Extensive experiments demonstrate substantial and consistent improvements over existing approaches.

The advent of agentic vulnerability detection is already becoming a watershed moment for software security. Audits conducted entirely by autonomous LLM agents are uncovering critical vulnerabilities in fundamental software underpinning digital society. Many of these vulnerabilities remained masked for years, surfacing only now with AI agents. Yet the reasoning behind these discoveries remains alarmingly opaque and unvalidated. What assumptions did the agent make about a function's inputs when it deemed that function to be secure? Failures in reasoning and incorrect assumptions can lead to missed vulnerabilities and reduce trust in agentic analysis. We propose a security-specification-first paradigm that (1) exposes the agent's tacit assumptions explicitly as security specifications and (2) continuously refines those specifications via runtime falsification. We realize our approach in Code-Augur, a novel harness for agentic vulnerability detection. Given a codebase, Code-Augur analyzes each component of the system for vulnerable code. When it deems a component to be secure, it commits the local invariants behind that judgment as in-source assertions. In parallel, Code-Augur leverages a guided fuzzer to attempt to falsify those assumptions. When the fuzzer triggers an assertion, this either reveals a genuine vulnerability or a flawed specification to refine. In both cases, this process grounds the agent's understanding, aligning its view of code intent with how the code actually behaves. On real-world subjects, Code-Augur effectively leverages security specifications to detect more vulnerabilities than other state-of-the-art agents. Additionally, Code-Augur found 22 new vulnerabilities in key open-source projects. Compared to curated specialized models like Claude Mythos, Code-Augur offers effective agentic vulnerability detection built on widely available LLMs like Sonnet and DeepSeek.

There exist numerous tutor training platforms. However, few provide AI-driven training and evaluation for human tutors based on real-life performance. We present an AI-driven system that assesses both open responses during training and authentic real-life tutoring. Unlike platforms that only assess learning through online training or simulations, our system utilizes Generative AI (Gemini-2.5-pro) to analyze transcriptions of authentic tutoring, measuring the transfer of tutor skills to real-life application. Human tutors instructing students remotely in math (N=86) completed six scenario-based lessons, averaging a significant 7.4% learning gain. Using mixed-effects models across 405 session-to-lesson pairs, we found that training performance significantly predicted real-life transcript scores with an effect size of 0.25 SD. Model comparison (AIC/BIC) indicated averaging open response and multiple choice performance during training predicted real-life tutor performance best, although open responses were comparatively more predictive. Exploratory analysis showed that after training, tutors were significantly more likely to encounter pedagogical opportunities to apply their skills (61.1% to 68.9%) and demonstrated higher execution quality within those opportunities (65.5% to 68.1%). Interrupted time series analysis suggested that these tutor improvements were part of a gradual trend over time rather than an immediate intervention effect of training. We illustrate an AI-driven method to link tutor training with real-life assessment. In doing so, we contribute open datasets, AI prompts, and scoring rubrics to support transparency and reproducibility.

The most plausible near-term role of medical LLMs is to assist rather than replace physicians, yet current evaluations often test isolated capabilities: clinical knowledge, EHR system interaction, or patient communication. Physician assistance instead requires coordinating these capabilities within the same interaction, where physicians issue underspecified requests, patients describe symptoms ambiguously, and EHR systems demand precise tool use. We introduce PhysAssistBench, a benchmark for interactive doctor-patient-EHR assistance. Built from real MIMIC-IV cases, PhysAssistBench uses a scalable pipeline to construct agentic patients: interactive, record-grounded agents that turn static EHR records into multi-turn clinical scenarios while preserving clinical factuality. PhysAssistBench provides a curated bilingual evaluation set of 1,296 manually reviewed and physician-validated turns. Experiments with leading LLMs show that current models remain unreliable in this setting, which exposes a key bottleneck for clinical LLMs: reliable assistance requires coordination across knowledge, communication, and systems, not isolated gains in any of them.

We propose a parameter-efficient speech enhancement framework, Quaternion Conformer GAN (QC-GAN), which combines a Quaternion Conformer generator with MetricGAN-based training. The Hamilton product encodes the magnitude and phase via structured weight sharing, reducing the number of layer parameters while preserving their interdependencies. A metric-learning discriminator was employed to maximize perceptual quality by optimizing the approximate perceptual evaluation scores. On the VoiceBank+DEMAND dataset, QC-GAN achieved a Perceptual Evaluation of Speech Quality (PESQ) score of 3.48 with only 0.89M parameters, delivering a performance comparable to state-of-the-art models at less than half their size. A 35K-parameter variant achieved a PESQ score of 3.23, surpassing conventional methods with significantly fewer parameters. Evaluation on the DNS-Challenge 3 dataset further confirmed generalization to real-world conditions.

It is essential for large language model (LLM) technology to serve many different cultural sub-communities in a manner that is acceptable to each community. However, research on LLM alignment has so far predominantly focused on predicting a unified response preference of annotators from certain regions. This paper aims to advance the development of alignment models with a more global outlook, that are able to accurately represent the preferences of subcommunities and do not exhibit excessive bias towards any of them. We focus on the development of reward models for this purpose and present a novel reward model training algorithm (SCPO) that can incorporate diverse cultural preferences in a balanced manner. Our method results in performance increases of the minority reward model of up to 7 points over the baseline model across two datasets, PRISM and GlobalOpinionQA, and across 7 countries. SCPO is up to 280% more training data-efficient than full-data finetuning of reward models. In addition, we perform analysis of bias by separately evaluating on the preference of subcommunities and show that excessive bias is mitigated via our weighting method. Our code is available at https://github.com/minsik-ai/Steerable-Cultural-Preference

As large language model (LLM) services become widely adopted, the cost of GPU resources for serving these models in cloud environments has emerged as a critical concern. Spot instances offer up to 90% cost savings over on-demand instances, but their frequent interruptions and limited availability pose significant challenges for continuous LLM serving. GPU spot instances, in particular, exhibit lower and more volatile availability than CPU-based instances, making homogeneous clusters that depend on a single GPU type vulnerable to correlated failures. Heterogeneous clusters spanning multiple GPU types can address this by leveraging complementary availability patterns across diverse spot pools, yet existing LLM serving systems are designed for homogeneous environments and suffer from load imbalance when deployed on heterogeneous GPUs. This paper presents ShuntServe, a cost-efficient LLM serving system for heterogeneous spot GPU clusters. ShuntServe employs a roofline model-based analytical serving performance estimator and a dynamic programming-based model placement optimizer that jointly determines node configuration, parallelization strategy, and layer assignment to maximize throughput across heterogeneous GPUs. To enhance fault tolerance when using spot instances, ShuntServe combines output-preserving request migration with concurrent initialization via a shared tensor store, minimizing migration downtime by overlapping replacement node preparation with ongoing serving. Evaluation on Llama-3.1-70B and Qwen3-32B with a heterogeneous AWS cluster of L4, A10G, and L40S GPUs shows that ShuntServe achieves 1.42x and 1.35x higher throughput than state-of-the-art baselines and attains 31.9% and 31.2% cost efficiency improvements over on-demand instances for offline and online serving, respectively.

The Controller Area Network (CAN) protocol is the primary communication standard for Electronic Control Units (ECUs) in modern vehicles, but its lack of encryption and authentication exposes it to a range of security threats. Existing intrusion detection systems are largely tuned to fabrication-style attacks (DoS, fuzzing, ID spoofing realised by frame injection), in which detection signals such as per-ID inter-arrival statistics are readily available. We instead address the harder \emph{masquerade} setting~\cite{b37}, in which an internal adversary substitutes a legitimate frame in-situ at its original transmission slot, preserving traffic periodicity and rendering traffic-statistic defences ineffective. We propose the Mamba Intrusion Detection System (MIDS), an innovative dual-stream framework that processes CAN identifiers and payloads in parallel and reconstructs their joint temporal semantics through bidirectional selective state-space modelling. To evaluate MIDS, we collected over 100 million CAN frames from a physical Tesla Model 3 across three driving regimes and synthesised 54 masquerade attack variants spanning ID-only, data-only, and combined modifications. MIDS attains an F1 of 96.94\% on this dataset, exceeding the strongest reproducible baseline by more than 8 percentage points, while sustaining a 1.147~ms single-window inference latency -- ample headroom for real-time onboard deployment. To verify generalisation, we further evaluate MIDS on four public benchmarks (ROAD, CrySyS, OTIDS, CT\&T) covering both masquerade and injection scenarios; MIDS attains F1 from 93.70\% to 99.61\%, outperforming the strongest of eight reproduced baselines by up to 13.94 percentage points under a unified 5-fold protocol.

Decision making in lithium production is challenging, whether from an investor's perspective or a strategic production standpoint. Determining which mines to open and when to open them involves not only geological and price uncertainties, but also complexities around the choice of extraction method, from direct lithium extraction to hard rock mining. Prior work explored models of this problem and different methods to optimize mining decisions; these models did not account for uncertainty in pricing, uncertainty in demand, or different mining technologies to extract lithium. Incorporating different pricing models and extraction technology into these models enables more robust strategies for determining not only when and where to open a mine, but also which method of production to pursue. We frame the problem as a partially observable Markov decision process (POMDP) and solve using belief state planning methods to get optimal decision making. In our study, we show that POMDP solvers outperform human inspired heuristics by dynamically adapting to shifting lithium price regimes (static, linear, exponential, and stochastic) through belief state planning and explicit uncertainty management. By optimally sequencing exploration, production, and technology choice, the framework achieves higher demand fulfillment and more balanced economic environmental outcomes over the projects lifetime in all different pricing and deposit scenarios.

Chinese dialects discrimination is a challenging natural language processing task due to scarce annotation resource. In this article, we develop a novel Chinese dialects discrimination framework with transfer learning and data augmentation (CDDTLDA) in order to overcome the shortage of resources. To be more specific, we first use a relatively larger Chinese dialects corpus to train a source-side automatic speech recognition (ASR) model. Then, we adopt a simple but effective data augmentation method (i.e., speed, pitch, and noise disturbance) to augment the target-side low-resource Chinese dialects, and fine-tune another target ASR model based on the previous source-side ASR model. Meanwhile, the potential common semantic features between source-side and target-side ASR models can be captured by using self-attention mechanism. Finally, we extract the hidden semantic representation in the target ASR model to conduct Chinese dialects discrimination. Our extensive experimental results demonstrate that our model significantly outperforms state-of-the-art methods on two benchmark Chinese dialects corpora.

Sleep diaries are central to behavioral sleep medicine and cognitive behavioral therapy for insomnia, yet daily completion is difficult to sustain, and static forms often provide limited context for interpreting night-to-night sleep variation. We designed an LLM-powered conversational voice diary that delivers clinically grounded morning and evening sleep diary questions through proactive smart-speaker prompts, structured conversational intake, and adaptive follow-up dialogue. We evaluated the system in a four-week between-subjects field study with 30 university students, comparing it with a text-based mobile diary using matched diary items, reporting windows, and reminder intervals. Compared with the text-based diary, the conversational voice diary showed higher adherence and elicited more detailed contextual self-report about routines, stressors, environmental conditions, and other sleep-related factors. Participants also described the voice diary as easier to integrate into daily routines, despite longer perceived completion time. However, voice-based conversational intake produced lower completeness for some structured diary fields, revealing a trade-off between expressive richness and structured precision. These findings show both the promise and the challenge of using LLM-powered conversational voice assistants for longitudinal health self-report.

In real-world reinforcement learning (RL), the choice of action space can play a key role in shaping motion smoothness, safety, and overall task performance. In this study, we evaluate pose increment, pose velocity, joint position increment, and joint velocity across two vision-based manipulation tasks: object picking and pushing. We train policies in simulation and deploy them to the real world using sim-to-real transfer. We find that action-space representation indeed significantly affects sim-to-real performance. In particular, we find that the joint velocity action space is best for the vision-based picking and pushing tasks in terms of smoothness and final task performance. We also provide practical guidance for RL practitioners in choosing action spaces for both simulation and real-world experiments.

3D Gaussian Splatting (3DGS) enables high-fidelity and real-time 3D scene reconstruction, but scaling training to large-scale scenes requires optimizing hundreds of millions of Gaussians across multiple GPUs. Existing distributed approaches either partition scenes into isolated regions, causing global inconsistency, or rely on global Gaussian-level exchanges, which lead to substantial growth in inter-GPU communication and quickly dominate iteration time. We propose Splaxel, a communication-efficient distributed 3DGS training framework based on pixel-level local rendering and global composition. Instead of synchronizing Gaussians, each GPU renders its local subset and exchanges only partial pixel values, maintaining mathematical consistency while keeping communication cost stable as the scene size increases. Splaxel further reduces pixel-level redundancy through geometric and transmittance visibility prediction and improves GPU utilization via conflict-free camera-view consolidation. Evaluated on large-scale datasets with up to 120M Gaussians, Splaxel achieves up to 7.6$\times$ speedup over the state-of-the-art distributed 3DGS framework while preserving high reconstruction quality.

Decoder-only Transformers compute attention over the KV cache of preceding tokens. Keys (and Values) are typically represented with the same dimensionality, regardless of its distance from the prediction target. In natural language, however, the next word is most strongly influenced by the immediately preceding tokens. We hypothesize that local and distant tokens impose asymmetric demands on representational capacity: local tokens are more critical for predicting immediate outputs and thus require richer representations, whereas distant tokens primarily serve as long-range memory, for which lower-dimensional representations may suffice. We formalize this idea as Distance-Adaptive Representation (DAR), implemented in a controlled setting that preserves full-dimensional representations within a local context window while assigning reduced-dimensional representations (e.g. 1/4 of the original dimensionality) to tokens beyond that window. Across multiple pretraining scales (70M to 410M parameters), as well as continued supervised fine-tuning on a 1B-scale model, this approach closely matches the performance of full-dimensional baselines. In contrast, uniformly reducing dimensionality across all token positions leads to worse performance. These results challenge the common assumption that key and value dimensionality should be uniform across token positions. Our findings suggest a new direction for designing attention architectures that adaptively allocate representational capacity across sequences, enabling further reductions in KV cache during inference.

Physical events are not understood by their names alone, but by the causal state changes that compose them. A clip-level label such as "bounce" can be correct while hiding the process that makes the event physically valid, from support loss and contact onset to rebound and settling. To make this hidden process explicit, we introduce Atomic Physical Transitions (APTs): minimal, temporally localized state changes that bind a visible cue to an active physical mechanism and before/after dynamical regimes. An APT chain represents a video as an ordered causal transition sequence rather than a single aggregate event label: event labels tell what happened; APT chains explain why it happened. To make APTs learnable by VLMs, we construct mixed-source APT data from human annotations and simulator ground truth, covering 14 transition types across contact, gravity, friction, and rotation/stability, with 27,303 timed instances over 1,246 trials. Using this data, we find that current VLMs miss transition-level physics, with zero-shot recall at most 14% and errors dominated by missed transitions. Direct fine-tuning on APT chains improves transition detection but causes event-level forgetting, indicating that the model learns a specialized answer format rather than a reusable physical representation. We therefore propose APT-Tune, a parameter-efficient recipe that teaches VLMs to use causal transitions without forgetting how to answer video questions. It combines image-pad-aware supervision, format-conditional co-training, and mechanism-conditioned domain-to-type decoding to make APT learning format-robust and physically grounded. With only 11 M LoRA parameters on Qwen3-VL-2B, APT-Tune substantially improves APT recall while also improving event-level video transfer. These results show that APTs are not a new answer format, but a human-aligned causal supervision signal for physical video understanding.

Language discrimination among similar languages, varieties, and dialects is a challenging natural language processing task. The traditional text-driven focus leads to poor results. In this paper, we explore the effectiveness of speech-driven features towards language discrimination among Chinese dialects. First, we systematically explore the appropriateness of speech-driven MFCC features towards CNN-based language discrimination. Then, we design an end-to-end speech recognition model based on HMM-DNN to predict Chinese dialect words. We adopt attention to extract the discriminative words related to different Chinese dialects. Finally, through a CNN, we combine the word-level embedding and the MFCC-based features. Evaluation of two benchmark Chinese dialect corpora shows the appropriateness and effectiveness of the proposed speech-driven approach to fine-grained Chinese dialect discrimination compared to the state-of-the-art methods.

Mild Cognitive Impairment (MCI) is a medical condition characterized by a noticeable decline in memory, language, or thinking abilities. MCI detection from spontaneous speech is promising for scalable screening. However, learned models often exploit demographic cues correlated with labels, resulting in a large performance gap across subgroups. We present a multimodal framework that combines (i) cross-model fusion between modalities (speech, text, and image), and (ii) unlearning using gradient reversal that discourages the shared embedding from encoding task-irrelevant demographic attributes. Evaluated on the multilingual benchmarks TAUKADIAL and PREPARE, our method outperforms the state-of-the-art multilingual and multimodal baseline in MCI classification while substantially reducing the performance gap across patient subgroups (sex and language). We further analyze transfer across datasets, showing that demographic unlearning helps learn more robust representations for MCI detection.

This paper presents a methodology-centered transfer learning framework for fragility adaptation under domain shift, class imbalance, and scarce target labels while preserving engineering interpretability and supporting decision-making under uncertainty. Four transfer learning strategies (instance-based, parameter-based, hierarchical Bayesian, and multi-source) are demonstrated through three complementary case studies: (i) instance-based transfer learning via importance weighting, demonstrated on coastal bridge fragility using Hurricane Katrina observations; (ii) parameter-based transfer learning together with hierarchical Bayesian transfer learning, enabling partial pooling across strata and posterior uncertainty quantification, demonstrated on residential building fragility using Hurricane Ian observations; and (iii) multi-source transfer learning that fuses multiple analytical fragility models with learned source weights and regularized target-domain adaptation, demonstrated on seismic bridge fragility using observations from the 2001 Nisqually earthquake. Across these case studies, direct transfer of source models (i.e. using existing state-of-the-art models) fails under domain shift and severe class imbalance, while targeted adaptation substantially improves failure detection and predictive stability in low-data regimes. These findings highlight the need for systematic guidance on diagnostics, strategy selection, and uncertainty reporting when developing and adapting fragility models.

Crowd counting is a fundamental task in computer vision. However, crowd counting in low-light environments remains largely underexplored, despite its practical importance in the real world. Existing methods mainly focus on well-lit scenes or rely on single-modality Red-Green-Blue (RGB) representations, which often become unreliable under extreme darkness and complex non-uniform illumination. To handle this problem, we construct three new low-light crowd counting benchmarks, which consist of two synthetic datasets, SHA\_Dark and SHB\_Dark, and a real-world benchmark LC-Crowd (Low-light Crowd Dataset). Inspired by Retinex-based physical modeling, we introduce depth and Canny edge cues as complementary geometric and structural priors to enhance the intrinsic reflectance representation under low-light conditions. We propose a Multi-Modal Hyper-Graph Fusion module, which formulates RGB appearance, depth geometry, and edge structure cues as nodes in a unified hyper-graph and explicitly captures their high-order complementary relationships via dynamic hyperedge construction and message passing. Furthermore, to adaptively allocate computation in dense prediction, we propose a Deformable Rectangular Sparse Attention (DRSA) module, which concentrates computation on informative regions through anchor-aware estimation and adaptive rectangular window modeling. Based on these designs, we develop a unified Low-Light Counting Network (LCNet) for robust low-light crowd counting. Extensive experiments on three benchmarks demonstrate that the proposed method achieves the best overall performance against existing state-of-the-art (SOTA) methods. The code is in the supplementary material. The datasets will be made public upon acceptance.

The quality of recorded data depends on the stability of the sensor system that acquires it. Sensor motion and aging can degrade the performance and stability of downstream data-driven methods. We present a Wasserstein-GAN-inspired approach for unsupervised inference of physically interpretable transformation parameters that map a changed detector response distribution back to a nominal reference distribution. In contrast to standard generative modeling, the generator is used as a learnable calibration transformation whose trainable weights represent the sought parameters, while the critic provides a distributional distance signal via the Wasserstein objective. We validate the approach on a tracking-detector toy model with controlled layer shifts and demonstrate its application on high-granularity Geant4-simulated calorimeter data with cell-wise aging effects. The method recovers aging coefficients for individual cells with correlation to ground truth and improves agreement between calibrated and reference energy-sum distributions, while exhibiting the expected degradation at increasing channel-to-channel noise levels. These results indicate that adversarial distribution matching can serve as a data-driven component of calibration strategies in settings where direct labels for degradation parameters are unavailable.

A rule-based logic solver resolves every instance in our benchmark in under 50 microseconds with 100% accuracy; the best frontier language model reaches 65% at best and drops to 23.5% under rendering-robust evaluation (worst case over four surface renderings). We introduce DeFAb (Defeasible Abduction Benchmark), a dataset and generation pipeline that converts four decades of publicly funded knowledge bases into formally grounded instances for defeasible abduction: constructing hypotheses that explain anomalies by overriding defaults while preserving unrelated expectations. Because every hypothesis must pass polynomial-time checks for valid derivation, conservativity, and minimality, DeFAb makes logical rigor the instrument for measuring creativity and theoretical reasoning, scoring the disciplined construction of theory revisions rather than fluent but theory-destroying prose. The pipeline pairs taxonomic hierarchies (OpenCyc, YAGO, Wikidata) with behavioral property graphs (ConceptNet, UMLS) to produce 372,648+ instances across 33.75M materialized rules from 18 sources, in three levels with polynomial-time verifiable gold standards. Four frontier models do not reliably internalize defeasible reasoning: rendering-robust Level 2 accuracy is 7.8-23.5%; chain-of-thought variance (~36 pp) exceeds any inter-model gap; and a matched contamination control isolates a +19.4 pp Level 3 gap. We further release DeFAb-Hard (a 235-instance Level 3 difficulty variant; best model 53.3% vs 100% symbolic) and CONJURE (a kernel-verified transformative-creativity variant of 560 Lean 4/Mathlib instances whose gold answers are definitions the proof kernel did not previously contain, judge-free verifier; a pilot finds zero novel concepts). The same verifier doubles as an exact reward for preference optimization (DPO, RLVR/GRPO). Released under MIT at https://huggingface.co/datasets/PatrickAllenCooper/DeFAb.

Adaptive AI ethics instruction in graduate research training benefits from intake measures that reflect differences in prior LLM experience. Prior coursework or workshop attendance is an obvious candidate, but it is not clear whether it is associated with pre-instruction ratings on key AI perception items. We compare three candidate intake features, self-reported usage frequency, self-rated LLM familiarity, and prior AI education, across five baseline perception outcomes in 93 bioscience graduate and postdoctoral trainees enrolled in a required research ethics course. Usage frequency shows Holm-corrected associations with all five outcomes, self-rated familiarity with three, and prior AI education with none. A threshold-like pattern at the lower end of the scale is most visible for training interest and accuracy trust rather than appearing as a uniform gradient across all five outcomes. In a short intake survey, reported LLM use is more consistently associated with these perceptions than prior coursework or workshops, with self-rated familiarity serving as a secondary indicator. These results suggest that simple pre-instruction behavioral signals can inform lightweight intake profiling for adaptive AI ethics education.

Language model agents are becoming proficient executors at isolated, short-horizon tasks such as software engineering and customer service. Yet real-world challenges require a combination of sophisticated skills that remain largely untested in agents: (1) navigating long horizons amid uncertainty; (2) acquiring information in noisy environments; (3) adapting to a changing world; (4) orchestrating multiple moving parts toward a coherent goal. We introduce CEO-Bench, which evaluates these capabilities together by simulating a representative real-world task: operating a startup for 500 days. An agent manages pricing, marketing, budgeting, and many other aspects of a fictional company through a programmable Python interface, operating in the same environment and facing the same challenges as a human CEO. Success demands analyzing noisy, interconnected business databases, translating signals into sound strategy, and coordinating many decisions with programming. The strongest agents write sophisticated code that simulates customer cohorts to forecast future cash and mines negotiation history to uncover hidden customer preferences. Even so, most state-of-the-art models struggle in this environment. Only Claude Opus 4.8 and GPT-5.5 finish above the $1M starting balance, and neither consistently turns a profit. CEO-Bench takes a first step toward measuring the intelligence required to drive sustained, adaptive progress over time.

Time series forecasting (TSF) underpins consequential decisions in energy, transportation, finance, and healthcare, yet TSF models are almost universally ranked by a single number (e.g., average error) on clean held-out data, under the implicit assumption that it predicts deployed reliability. However, real faults are not i.i.d noise but structured events with temporal shape, broken cross-variable dependencies, regime change coupled with missingness, and causal propagation across a sensing pipeline. Treating TSF robustness as a data-quality problem, we present TS-Fault, a benchmark that evaluates forecasting models under explicit, parameterized fault scenarios with controllable semantic difficulty. TS-Fault organizes recurring failures into four modes along two orthogonal axes (observation- vs mechanism-level; univariate vs multivariate) and injects each fault into the most prediction-critical window via a unified importance score. This design enables robustness to be tested against the structures models actually rely on, rather than reduced to generic noise sensitivity. We evaluate 21 models across 6 datasets, 4 modes, and 5 difficulty levels under a paired clean/corrupt protocol. The results reveal three findings that contradict common leaderboard intuition: (i) clean-data accuracy anti-correlates with robustness; (ii) clean rankings are preserved under observation-level faults but reshuffled under mechanism-level faults; and (iii) all catastrophic failures occur under mechanism-level faults, with foundation models achieving the highest clean-data accuracy yet exhibiting the greatest fragility. The code is publicly available at https://github.com/Ray-zyy/TS-Fault.

One of the major difficulties in the mechanistic interpretability of neural networks is the occurrence of polysemanticity, which suggests that each neuron is typically responsible for multiple different tasks, impeding a clean interpretation of their function. The seminal paper of Elhage et al. (2022) argues that this occurs due to superposition, a phenomenon where the neural network represents distinct features as non-orthogonal directions in a lower-dimensional space, a strategy that allows much greater compression of the data without sacrificing fidelity due to the feature sparsity of input vectors. Elhage et al. (2022) empirically validates these hypotheses in a rather natural and simple autoencoder with sparse inputs. The contribution of the present work is to analyze the mathematical basis for the occurrence and optimality of superposition, while rigorously corroborating some of their findings. In particular, we provide upper and lower bounds for the L2 reconstruction loss, tight in the very sparse regime, for power activation functions. A short list of interesting open problems are also included at the end.

Humans often acquire new skills by observing others, since observed behaviors implicitly reveal how to act in an environment. However, observations drawn from a heterogeneous population introduce conflicting behavioral signals, making it difficult to determine which behaviors are worth imitating. We address this challenge with General Reward Inference and Disentanglement (GRID), a social learning method that extracts universally useful behaviors from a heterogeneous population of demonstrators pursuing different goals. GRID decomposes per-agent reward functions into a general reward, capturing behaviors shared across all agents, and specific rewards, capturing individual preferences and objectives. Training exclusively on the general reward provides a new paradigm of generalist pretraining. It yields a generalist agent that internalizes universal environmental competencies, such as safety and basic task proficiency, without the mode-averaging bias that afflicts standard learning from demonstration techniques. This generalist serves as a superior prior for fine-tuning to downstream tasks, including preferences unseen during training. Experiments across a synthetic basis function decomposition, multi-agent Craftax, and a continuous autonomous driving simulator (Highway-Env) confirm that GRID successfully disentangles reward structure in a semantically meaningful way, outperforms standard learning from demonstration baselines, and enables more efficient and stable specialization.

The large-scale digitization of educational assessment has made the continuous oversight of item banks both essential and complex. This paper presents Analytics for Quality Assurance for Item Pools (AQuAP), a dashboard environment for monitoring item quality and item bank health. AQuAP supports the operational implementation of the large scale item generation procedures for high-stakes tests as included in the Item Factory, a framework for automated and human-supported test development. The paper describes AQuAP in relationship with the process of item development, outlines the broader metric framework for item-pool quality assurance, and highlights the Effective Bank Size (EBS) as one central indicator of pool vitality. EBS quantifies how many independent test sessions can be constructed before content repetition occurs and, when coupled with exposure and usage metrics, provides insight into item bank security, diversity, and efficiency. We further introduce bank-health metrics, such as maximum exposure, maximum conditional exposure, adjusted effective bank size, and the rarely-administered fraction, all of which extend this picture of item utilization. AQuAP illustrates how operational analytics can translate psychometric concepts into quality assurance tools for high-volume, AI-enabled testing programs. This work is illustrated with the Duolingo English Test (DET) processes.

Multi-cause observational studies contain information about unmeasured confounding through the dependence structure among causes. However, literal imputation of the unobserved confounder is often more complex than learning a lower-dimensional substitute score that preserves the shared assignment variation needed for stable causal adjustment. The deconfounder (Wang and Blei, 2019) and related substitute confounder methods exploit this idea, but flexible assignment models can fit the joint distribution of the causes while producing scores that over-encode the treatment vector, collapse overlap, or capture single-cause variation. We develop a Bayesian factor assignment framework for learning sparse substitute confounders that retain coarse multi-cause dependence with shrinkage priors. The theory is stated at the level of posterior concentration, factor score contraction, and overlap-preserving assignment geometry and therefore does not rely on a particular shrinkage prior. Under these conditions, the proposed regression-adjusted estimators are consistent for mean potential outcomes when the corresponding latent variable identification assumptions hold. Shrinkage priors provide a natural tool for latent structural learning: they favour low-dimensional factors supported by multiple causes, discourage effectively single-cause factors, and induce an ordering of the latent factors through progressive shrinkage. Synthetic experiments illustrate the roles of signal strength, outcome validity, and geometry-aware regularization. In an Alzheimer's Disease Neuroimaging Initiative (ADNI) baseline analysis, sparse substitute scores recover much of the adjustment obtained by directly conditioning on invasive cerebrospinal-fluid biomarkers, while collapse diagnostics identify when fitted factors reduce to individual observed measurements.

AI systems are increasingly evaluated in bounded environments that combine isolation, simulation, instrumentation, supervision, and evidence capture. For physical AI, AIoT, and cyber-physical systems, this shift is not a matter of terminology: the system under test may sense, decide, actuate, communicate, and fail through physical processes, networked devices, and human operators. This article develops an assurance-oriented account of AI sandboxes as controlled environments for testing, evaluation, verification, and validation across digital AI, embodied autonomy, and cyber-physical deployments. We formalize the sandbox boundary and a weakest-link rule for composing per-dimension evidence into a bounded deployment claim; separate major sandbox archetypes; define a cyber-physical threat model that includes attacks on the assurance apparatus itself; and introduce a measurement framework spanning fidelity, controllability, observability, containment, reproducibility, and governance artifacts, instantiated on three worked case studies of real sandboxes. The resulting threat model, taxonomy, and measurement framework clarify what a sandbox can validly test, which risks it can contain, and what forms of evidence it can support for safety, security, and regulatory assurance.

Offline reinforcement learning is typically analyzed under process-level reward supervision, yet many sequential decision datasets record only trajectory-level outcomes. We develop a statistical theory for offline policy optimization from such outcome-level supervision. We first study the canonical setting where the target remains the expected cumulative reward, but each offline trajectory provides only a scalar label whose conditional mean is the cumulative return. We propose OPAC, a pessimistic actor-critic algorithm that learns a latent reward model and optimizes a policy from trajectory-level labels. We prove a high-probability guarantee of order $\widetilde O(H^2\sqrt{C_{sa}(π^\star)/n})$ and a matching lower bound, characterizing the sharp statistical cost of replacing process-level rewards with one trajectory-level label. We then extend the principle to preference-based feedback, preserving the leading horizon and concentrability dependence up to preference-model constants. Finally, we study generalized outcome-based offline RL, where both the supervision and the objective are trajectory-level quantities induced by a nonlinear aggregation of latent per-step rewards. This problem is not learnable in general: for all-success objectives, any offline learner may require $Ω(2^H)$ trajectories even with deterministic transitions and constant concentrability. We then identify a tractable regime through two structural coefficients, $κ_μ(σ)$ and $χ_μ(σ)$, capturing information loss in outcome aggregation and generalized Bellman updates, under which generalized OPAC achieves polynomial sample complexity. Together, our results delineate when outcome-level supervision enables sample-efficient offline control and when missing process-level rewards create fundamental statistical barriers.

Domain-camouflaged injection attacks embed malicious instructions in retrieved content using domain-appropriate vocabulary, evading standard detectors that rely on syntactic injection markers. When detection fails, practitioners need to know which defense architectures reduce attack success. We evaluate five prompting-based defenses (spotlighting, paraphrasing, prompt sandwiching, and two combinations) against domain-camouflaged injection across three model families (Claude Haiku, Llama 3.1 8B, Gemini 2.0 Flash) and three deployment domains (financial, legal, general) using 3,510 trials. Paraphrasing retrieved content before agent processing is the most consistently effective defense in this benchmark, reducing camouflage attack success rate by 55-84\% depending on model, and achieves lower attack success rates than our Llama Guard 4 configuration on every model tested. Defense effectiveness is strongly model-dependent: spotlighting halves attack success on Claude Haiku but provides no benefit on Llama 3.1 8B. Financial domain deployments face the highest residual risk at 26-33\% baseline attack success rate, with no prompting-based defense fully eliminating the threat on weaker models. These results provide the first systematic evaluation of prompting-based defenses specifically against camouflage-class injection attacks and establish benchmark-based recommendations for practitioners. All tasks use synthetically constructed professional documents; whether these benchmark rankings generalize to real enterprise documents remains an open question.

U-calibration studies online forecasting algorithms whose predictions can be consumed by any unknown downstream agent, guaranteeing sublinear regret simultaneously for all proper loss functions. Existing U-calibration algorithms achieve worst-case optimal $O(\sqrt{T})$ regret for every bounded proper loss, but they fail to adapt to easier losses: as we show, even for smooth losses such as squared loss, they incur $Ω(\sqrt{T})$ regret instead of the optimal $O(\log T)$ regret. In this work, we show that this limitation is not inherent. Specifically, we design a single forecast algorithm that simultaneously achieves $\tilde O(\sqrt{T})$ regret for every bounded proper loss and $O(\log T)$ regret for every bounded smooth proper loss. More generally, our algorithm also attains logarithmic regret for losses that are smooth relative to the log-barrier, which include several non-Lipschitz examples. Our approach is based on a novel variant of Follow-the-Perturbed-Leader (FTPL) in which perturbations are applied directly in the prediction space using self-concordant noise. The resulting analysis also departs substantially from prior FTPL analyses due to the complex nature of this noise and may be of independent interest.

We propose a hierarchical attention mechanism based on two-level overlapping Schwarz domain decomposition. The method is motivated by the observation that two-level Schwarz domain decomposition methods combine local subdomain corrections with a coarse level that communicates global, long-range information. We test its usefulness in the context of finite-dimensional operator learning using a simple, one-dimensional diffusion problem with homogeneous Dirichlet boundary conditions. Although elementary, this problem provides a controlled sequence-to-sequence setting in which the exact nonlocal solution operator is known. After discretization, learning the solution operator amounts to approximating the inverse of a symmetric positive definite matrix. As a baseline, we use a global softmax-free low-rank attention operator of the form $QK^T$. The proposed construction replaces this dense global factorization by a two-level additive structure: local low-rank attention blocks on overlapping subdomains are combined with a coarse attention block. The resulting operator has the form $$M_θ^{-1} = ΦQ_0 K_0^T Φ^T + \sum_{i=1}^{N} R_i^T D_i^{1/2} Q_i K_i^T D_i^{1/2} R_i.$$ Here $R_i$ restricts to an overlapping subdomain, $D_i$ is a partition-of-unity weight, and $Φ$ is a coarse interpolation (or prolongation) matrix. Numerical experiments for synthetic Fourier right-hand sides indicate that the domain-decomposition attention operator is able to train faster and can give more accurate approximations than a global low-rank attention baseline while using significantly fewer parameters.

Looped (weight-tied) Transformers apply a shared residual block $N$ times ($h \leftarrow h + \varepsilon\,f(h)$, same $f$ at each step), increasing effective depth without adding parameters. Prior depth-scaling analyses prescribe $\varepsilon = 1/\!\sqrt{L}$ for depth-$L$ residual networks. We show that this is insufficient for looped architectures: weight sharing makes residual updates correlated across iterations, requiring the stronger scaling $\varepsilon = 1/N$. For multi-layer blocks ($L$ unique layers looped $N$ times), we derive a factored parameterization $\varepsilon = λ/(N\!\sqrt{L})$ that separates the two sources of growth: $1/N$ controls the within-layer loop correlation, and $1/\!\sqrt{L}$ controls the across-layer variance. A key consequence is that the optimal learning rate depends only on the number of unique layers $L$, not on the loop count $N$, enabling direct hyperparameter transfer from small to large $N$ without retuning. Experiments on looped Transformers confirm that $1/N$ scaling improves trainability and yields better loss than $1/\!\sqrt{N}$ scaling across loop counts.

Reinforcement Learning with Verifiable Reward (RLVR) has emerged as a powerful post-training paradigm that surpasses Supervised Fine-Tuning (SFT) in eliciting reasoning intelligence and resisting catastrophic forgetting. Recent studies further reveal that RLVR induces highly sparse and off-principal parameter updates compared to SFT. This naturally raises the question: does such sparsity make RLVR models more amenable to model merging? If so, model merging would offer a scalable, training-free path to aggregate diverse reasoning capabilities from independently trained RLVR models. Surprisingly, we find the opposite, uncovering a sparsity curse: the sparse RLVR updates are spread farther apart in parameter space, forming near-orthogonal shortcuts that make aggregation inherently fragile. This is likely rooted in the stochasticity of RL optimization and the diversity of emergent reasoning patterns. Unlike SFT models that converge to shared, flat basins and merge naturally, RLVR models suffer severe degradation under standard merging methods. Through systematic empirical analysis of the update geometry, we characterize the mechanisms behind this failure and propose Sensitivity-aware Resolving Merging (SAR-Merging), a merging recipe tailored for the unique structure of RLVR parameter spaces. SAR-Merging resolves conflicts in overlapping update regions via Fisher Information-based sensitivity arbitration, followed by magnitude-aware sparsification and rescaling to preserve fragile reasoning pathways. Experiments on mathematical and coding benchmarks demonstrate that SAR-Merging substantially outperforms existing merging methods on RLVR models, enabling both single-task enhancement and multi-capability fusion.

Computing geometric representations of data is a cornerstone of modern machine learning, typically achieved by training dual encoders which map queries and documents into a shared embedding space. Recent work of You et al. [NeurIPS '25] has extended this approach to hierarchical retrieval, where relevance is determined by the ancestor-descendant relationships in a Directed Acyclic Graph (DAG). While previous work has shown that valid embeddings exist when the number of descendants is small, these bounds degrade significantly for deep hierarchies, requiring dimensions as large as the total number of nodes. In this paper, we investigate compact reachability embeddings for more general graph classes and provide theoretical guarantees for representing hierarchies using embeddings whose dimension depends on structural graph parameters. We prove that for any directed tree, there exists a reachability embedding in constant dimension 3, independent of the tree's size or depth. We generalize this result to graphs characterized by treewidth $t$, constructing embeddings of dimension $O(t \log n)$, where $n$ is the number of nodes. Complementing these upper bounds, we provide matching or near-matching lower bounds, showing that dimension $Ω(n)$ is necessary for general DAGs and $Ω(t/\log(n/t))$ is required for graphs of treewidth $t$. We also obtain upper and lower bounds parameterized by the number of cross-edges in the DAG. We additionally show that our embeddings can be constructed on real world datasets, and that they give much smaller dimensions in high recall regimes compared to prior embeddings with theoretical guarantees.

Though robotic systems are now being commercialized and deployed in various industries, many of these systems are highly specialized and often require an advanced skill set to operate and ensure they perform as instructed. To mitigate this problem, we recently introduced a mission planner leveraging LLMs to synthesize mission plans in precision agriculture based on mission descriptions provided in natural language. While the system demonstrates impressive performance, it also suffers from the inherent ambiguities of natural language. In this paper, we extend our system to address this issue by introducing multiple feedback loops in the planning architecture that leverage linear temporal logic (LTL) to ensure the mission planning system meets the specifications formulated by the user while still using natural language. To mitigate potential bias, this is achieved by using two different commercial LLMs in charge of the specification and verification subtasks. Through extensive experiments, we highlight the strengths and limitations of integrating mission verification into a fully autonomous pipeline, particularly regarding an LLM's ability to generate valuable LTL formulas, and show how our proposed implementation addresses and solves these challenges.

The development of medical AI is constrained by limited access to high-quality clinical data due to institutional silos and strict privacy regulations such as HIPAA and GDPR. Synthetic data generation offers a potential solution, but existing methods lack principled mechanisms to explicitly manage the privacy-utility trade-off, often degrading clinically meaningful patterns or risking patient re-identification. We present PSyGenTAB, a privacy-preserving generative framework that formulates synthetic healthcare data generation as a constrained optimization problem solved using the Augmented Lagrangian Method. By embedding configurable privacy constraints directly into model training, PSyGenTAB enforces minimum privacy thresholds while maximizing clinical data utility. Across multiple clinically motivated benchmarks, PSyGenTAB preserves inter-feature clinical relationships and minority-class diagnostic patterns essential for reliable health AI. Downstream evaluation using Train-on-Synthetic, Test-on-Real and Train-on-Real, Test-on-Synthetic protocols shows that models trained on synthetic data achieve performance comparable to those trained on real patient records. Privacy auditing further demonstrates reduced exact record reproduction and strong resilience to membership inference attacks. These results establish PSyGenTAB as a principled framework for balancing privacy protection and clinical utility in synthetic healthcare data, supporting secure cross-institutional AI development.

Barren plateaus are stated as an average-case phenomenon: pick an ansatz, initialize it naively, and concentration follows. This has led to the common view that a potential cure for barren plateaus is simply to initialize the parameters more carefully. Here we show that the situation is subtler. We introduce a first-moment framework that gives a simple operator-level diagnostic for when an initialization may escape the fully concentrated barren-plateau fixed point, and for comparing the biases induced by different initialization strategies. Our framework recovers several known initialization schemes such as identity and Gaussian initialization, but also shows that barren-plateau avoidance is highly non-unique. Indeed, many shifted, biased, and non-symmetric parameter distributions can avoid concentration, and these choices need not be equivalent. In fact, our results show that one can generate exponentially many families of inequivalent initialization strategies. Then, our numerics indicate that different first-moment-distinct initializations can lead to different attained minima, suggesting that avoiding barren plateaus via smart initializations can trade the exponential concentration problem for the challenge of selecting the right trainable pocket amongst many options.

Neural combinatorial optimization (NCO) offers a promising alternative to traditional heuristic-based methods for solving complex graph optimization problems by proposing to learn heuristics through data. This class of problems frequently arises in automation, as it can be used to model a variety of applications. While NCO has been extensively studied for deterministic combinatorial optimization problems, there are only a few works that aim to solve stochastic combinatorial optimization problems. In this work, we present N(CO)$^2$: Neural Combinatorial Optimization with Chance cOnstraints to solve the Stochastic Orienteering Problem (SOP) without the use of hand-crafted heuristics. By integrating a reinforcement learning (RL) framework, the model optimizes path selection under uncertainty, effectively balancing exploration and exploitation. Empirical results demonstrate that our method generalizes well across diverse SOP instances, achieving competitive performance compared to the state-of-the-art mixed-integer linear program (MILP) for the task. The proposed approach reduces human effort in heuristic design while enabling adaptive and efficient decision-making in uncertain environments.

Reliable generalization in conditional latent variable models requires understanding both identifiability and extrapolation: how observed variation across attributes determines latent structure, and how that structure determines distributions at unseen attributes. However, existing identifiability and extrapolation guarantees are largely model-specific, with separate analyses in nonlinear ICA, causal representation learning, perturbation modeling, and related conditional latent variable models. We introduce concept modulation models (CMMs), an attribute-indexed class of conditional generative models with structure $A\to Λ\to C\to X$, where attributes select modulators, modulators induce latent concept laws, and concepts generate observed features. CMMs lift transition-based identifiability to conditional settings by showing that feature agreement on observed attributes induces a latent concept transition constrained by the CMM class. We express these constraints through attribute potentials, log-density ratios between attribute-conditioned concept laws, separating the generic lifting step from model-specific rigidity arguments. The same potentials control extrapolation: agreement at unseen attributes holds exactly when the transported attribute-potential identities extend to those attributes. This yields algebraic extrapolation criteria, identifies the common potential-based proof objects behind several existing identifiability and extrapolation results, and, when combined with the model-specific rigidity arguments in those works, recovers their stated conclusions.

Retrieval-augmented generation (RAG) systems depend critically on how documents are chunked and searched. Fine-grained chunks can improve retrieval precision but expand the search space, increasing latency and cost; larger chunks reduce the number of candidates but make dense similarity less reliable, as the representation for each chunk mixes multiple topics and introduces more semantic noise. This trade-off becomes especially limiting in deep research tasks, where retrieval must be both fast and precise across large, heterogeneous corpora. We introduce MCompassRAG, a metadata-guided retrieval framework that uses topic-level signals as a semantic compass for selecting relevant evidence. Instead of relying only on cosine similarity between queries and noisy chunk embeddings, MCompassRAG enriches chunk representations with topic metadata in the same embedding space and trains a lightweight retriever through LLM-teacher distillation. At inference time, MCompassRAG performs topic-aware retrieval without additional LLM calls, improving both efficiency and evidence quality. Across six complex retrieval benchmarks, MCompassRAG improves information efficiency (IE) by 8.24% on average with over 5 times lower latency than the strongest efficient RAG baselines. Code is available on https://github.com/AmirAbaskohi/MCompassRAG.

Objective sleep assessment relies on polysomnography (PSG), yet clinical impact is often better reflected in patient-reported outcomes (PROs) such as sleepiness and fatigue. Existing summary indices, including the Apnea-Hypopnea Index (AHI), provide limited insight into the multidomain physiology underlying functional recovery. We propose an interpretable, causal-discovery--guided framework for deriving a hierarchical Sleep Recovery Score (SRS) from multimodal PSG. Using two large population cohorts (MESA: n=1540; MrOS: n=825), we apply directed acyclic graph (DAG) learning to identify candidate physiological drivers spanning respiratory burden, hypoxic burden, sleep fragmentation, sleep architecture, and autonomic regulation. Although derived from clinical PSG, these domains map naturally to sensing streams increasingly available in connected health technologies, including wearable ECG, oximetry, and sleep-stage estimation devices. To preserve mechanistic plausibility, we introduce a two-stage screening process that combines physiology-based constraints with constrained LLM-assisted auditing to identify and remove structural confounders and construct-overlapping variables. Across cohorts, these five domains emerge as recurrent physiological domains associated with recovery, and the resulting SRS shows up to 2.5$\times$ stronger alignment with perceived recovery than AHI. By linking multimodal sleep physiology to patient-centered outcomes through an interpretable, bias-aware, and domain structured framework, this work provides a practical foundation for recovery modeling across both clinical sleep studies and emerging smart and connected health settings.

Remaining useful life (RUL) estimation is central to predictive maintenance, where an unplanned failure can cost far more than the asset itself. Statistical degradation models miss the strong nonlinearity of real systems, and data-driven models often converge to suboptimal solutions in high-dimensional, non-convex search spaces. We propose a Quantum Annealing enhanced Q-Learning (QAQL) framework that couples the sampling behaviour of quantum annealing with the sequential decision making of Q-learning. Each Q-value update is encoded as a small quadratic unconstrained binary optimization (QUBO) whose ground state is the greedy action; rather than acting as a deterministic optimizer, the annealer returns a distribution over near-optimal actions across many reads, and this stochastic action selection supplies the exploration that curbs premature convergence on nonlinear degradation trajectories. The QUBO is solved on the D-Wave Advantage system using minor embedding, with the annealer woven into the reinforcement-learning loop rather than bolted on after training. We validate QAQL on two public benchmarks: the NASA C-MAPSS turbofan engine datasets and a device-fleet predictive maintenance dataset. Averaged over many independent runs and across six error metrics, QAQL outperforms the classical and quantum baselines considered in this study, with statistically significant improvements. The results indicate that quantum annealing is a usable, not merely theoretical, optimizer inside a reinforcement-learning loop for industrial predictive-maintenance applications.

Large language model (LLM)-based multi-agent systems demonstrate strong performance on complex reasoning and task execution, enabling broad enterprise applications. However, production deployment remains challenging due to domain-specific customization requirements and high latency and inference costs in agentic workflows. We propose a unified framework for customization and efficient deployment of multi-agent systems in real-world settings. The first stage, Agentic Model Customization, combines continual pretraining, supervised fine-tuning, and preference optimization to adapt a compact model to specialized domains while retaining strong agentic capabilities. The second stage, Inference Optimization, integrates speculative decoding and FP8 quantization with targeted calibration to enable cost-efficient serving with minimal quality loss. Across enterprise workloads, our framework enables rapid domain adaptation and achieves a 4.48x speedup in throughput while maintaining performance and improving robustness on long-tail scenarios.

Correspondence is fundamentally relational: it seeks the unknown transformation between two observations of a common scene, not the content of either. Yet the dominant learning-based methods do not represent the transformation as a first-class object in the architecture. They encode each image independently and let a learned similarity function or a deep decoder discover the mapping implicitly. Phase correlation is the canonical exception, measuring the inter-image relationship directly in the Fourier domain, but the rigidity of its fixed basis confines it to global translation. We introduce a learned generalization of phase correlation that lifts this restriction by learning the basis on which the transformation decomposes. The same algebraic primitive extends to dense non-rigid deformations and to unitary dynamics. On the ACDC cardiac-MRI benchmark the framework matches or exceeds prior published baselines on both registration directions. On CAMUS echocardiography it matches state-of-the-art without auxiliary scoring or adaptive-smoothness mechanisms. Applied to time-evolved wavefunction pairs of the 1-D quantum harmonic oscillator, the same framework recovers the Hermite-function eigenstates and the quantized energy levels of the unknown Hamiltonian from observation pairs alone.

The standard heuristic of selecting the SFT checkpoint with the highest pass@1 for GRPO can fail when SFT compresses the rollout distribution. For binary rewards, the expected within group advantage variance is $p(1{-}p)(g{-}1)/g$; when early GRPO drives $p$ below $p^*(g)$, most groups have identical rewards and provide no group relative signal. We study SFT depth ladders for Qwen2.5-Coder-3B and DeepSeek-Coder-6.7B. We test Qwen2.5-Coder-3B across five depths and three seeds, and DeepSeek-Coder-6.7B across four matched depths and three seeds. On Qwen, pre RL pass@1 rises with SFT depth, but peak GRPO pass@10 falls from $0.806$ to $0.481$ (3 seed mean, $n{=}20$); pre RL entropy is positively associated with the GRPO outcome ($ρ{=}{+}0.69$). On DeepSeek, pass@1 remains far above $p^*(8){=}0.083$, and GRPO outcomes compress rather than invert. A two stage diagnostic, combining pre RL entropy triage with an early GRPO entropy monitor, flags high risk checkpoints and can stop failing runs early. Simple KL to reference regularisation and label smoothing variants do not rescue the collapsed Qwen checkpoint in our setting, suggesting the failure is not a trivial GRPO hyperparameter artefact.

Neural Text-to-Speech (TTS) systems achieve remarkable quality on short utterances but long-form speech generation shows prosodic drift, speaker inconsistencies and sentence boundary artifacts. Existing approaches either compress sequences, increase context length or naively concatenate independently synthesized chunks. We present an inference-time approach called MagpieTTS-LF that enables MagpieTTS to produce coherent long-form speech without model retraining. Our method introduces three key innovations: (1) soft attention priors to guide monotonic alignment while preserving past and future context; (2) a stateful inference algorithm that maintains context across sentence chunks, ensuring prosodic continuity; (3) history-aware text encoding that uses past text for discourse-level prosodic planning. Experiments on long texts show significant improvements in long-range intelligibility, prosodic coherence, speaker consistency, and boundary naturalness compared to other baselines.

State estimation from uncertain, distributed observations is central in many cyber-physical applications. While Distributed Particle Filtering (DPF) algorithms address nonlinear and non-Gaussian estimations in distributed settings, most solutions remain tied to specific architectures and communication assumptions, limiting adaptability in open, heterogeneous deployments-most notably, the Internet of Things (IoT). In this paper, we propose a field-based formulation of Distributed Particle Filtering grounded in Aggregate Computing (AC). By expressing estimation and information dissemination as computational fields, our approach decouples the core filtering logic from coordination and data-flow strategies. This enables systematic customisation of key design dimensions, including fusion-center placement and resilience, aggregated measurement functions, as well as the type and scope of information propagation. Through a set of in-silico experiments, we show how diverse DPF configurations can be derived within a unified framework, highlighting trade-offs among accuracy, communication cost, and robustness. Overall, the proposed approach positions AC as an effective abstraction layer for engineering adaptable DPF solutions in open IoT environments.

Creative coding libraries provide high-level tools that make computational and algorithmic art accessible to artists and learners. Processing/p5 is one such family of libraries, known for its beginner-friendly approach and wide reach across artistic and technical communities. L5 is a new member of this family, implemented in Lua using the LOVE framework. It applies permacomputing principles, a movement addressing sustainability in computing inspired by permaculture, bringing these values to a community of practice not historically centered on them. This paper explores L5's design decisions and tensions between sustainability and usability through five case studies: 1. balancing perceived simplicity versus exposing the seams, 2. designing for lower resource consumption, 3. ensuring long-term stability, 4. constraining functionality, and 5. designing documentation for resource-constrained access. Rather than optimizing for a single metric, sustainable creative tools require navigating competing values transparently.

Reject inference methods are widely used to mitigate survival bias in credit scoring, yet their effectiveness remains poorly understood. We systematically evaluate several such methods and uncover a structural failure mode: in a natural retraining cycle, models whose accuracy improves while recall collapses create an illusion of improvement that leads practitioners to believe the system is getting better when, in fact, its rejection quality -- the ability to correctly screen out defaulters -- is deteriorating. We then propose a controlled exploration strategy that breaks the feedback loop without statistical assumptions: the lender deliberately approves a fraction of rejected applicants and observes their true outcomes. We show that accuracy and rejection quality give opposite recommendations on whether to explore: accuracy favors no exploration, while rejection quality improves with it, confirming that standard evaluation metrics are misleading under selection bias. Even minimal exploration rates (2--5\%) prove sufficient in our experiments to diagnose the severity of the feedback loop at near-zero cost. Our findings are consistent across two machine learning methods and three real-world datasets, and suggest that standard evaluation protocols are inadequate for assessing models trained under survival bias.

Machine unlearning for large language models (LLMs) aims to remove specified knowledge while preserving the rest of the model's capabilities. However, the boundary between knowledge to forget and knowledge to retain is often unclear, since related and even distant information may be entangled in the model. In this paper, we study LLM unlearning from a data-centric perspective and measure how unlearning effects propagate from the forget set to same-domain and distant-domain knowledge. We find a consistent decay pattern: collateral damage is strongest near the forget set, weakens with semantic distance, but does not disappear at domain boundaries. We further ask whether such damage can be audited before unlearning is executed. We formulate forget-set auditing as a pre-unlearning prediction task and analyze which data features are most predictive of downstream damage. Our results show that interaction features between the forget set and evaluation set provide the strongest signals, suggesting that collateral damage is partly reflected in data geometry before model updates occur. These findings position forget-set auditing as an early warning tool for identifying risky unlearning runs and designing more reliable unlearning procedures.

Large language models (LLMs) are increasingly used for clinical text tasks such as summarization and revision. While most studies evaluate the fluency and coherence of LLM-generated text, whether LLMs correctly preserve diagnostic uncertainty remains underexplored. In clinical practice, phrases such as ``possible pneumonia'' communicate the strength of available evidence and directly guide decisions about follow-up testing and treatment. Altering these uncertainty expressions can change the clinical meaning entirely. In this paper, we systematically evaluated this problem in two steps. First, we constructed a benchmark of 1,200 clinical documents with 9,184 uncertainty annotations across five levels. Second, we evaluated three LLMs on this benchmark. Our results show that (1) LLMs preserve the original uncertainty cues poorly, often less than half the time; (2) LLMs struggle with nuanced distinctions between adjacent levels. This work reveals a failure mode not captured by standard evaluation metrics and provides implications for the safe deployment of LLMs in clinical workflows.

Neuroscientific research has revealed that the brain encodes complex behaviors by leveraging structured, low-dimensional manifolds and dynamically fusing multiple sources of information through adaptive gating mechanisms. Inspired by these principles, we propose a novel reinforcement learning (RL) framework that encourages the disentanglement of dynamics-specific and reward-specific features, drawing direct parallels to how neural circuits separate and integrate information for efficient decision-making. Our approach leverages locally linear embeddings (LLEs) to capture the intrinsic, locally linear structure inherent in many environments, mirroring the local smoothness observed in neural population activity, while concurrently deriving reward-specific features through the standard RL objective. An attention mechanism, analogous to cortical gating, adaptively fuses these complementary representations on a per-state basis. Experimental results on benchmark tasks demonstrate that our method, grounded in neuroscientific principles, improves learning efficiency and overall performance compared to conventional RL approaches, highlighting the benefits of explicitly modeling local state structures and adaptive feature selection as observed in biological systems.

Modern AI agents retrieve documents, call tools, check intermediate information, and then produce a final answer or action. This creates a risk-control problem that is not visible from the final answer alone. A final response may look acceptable even when the retrieval was weak, a tool output was wrong, or an earlier step was unsupported. We propose ToolChain-CRC, a conformal risk-control method for retrieval-augmented and tool-using agents under drift. The method treats each agent run as a full trajectory of actions, observations, and final output. It builds step-level risk scores, combines them into a trajectory risk score, calibrates an accept-or-intervene rule, and adds an anytime alarm that can stop risky runs before the final answer. We prove trajectory-level risk control under exchangeable calibration runs, give a drift-aware extension with auditable constants, and prove an anytime escalation rule through a supermartingale construction. Experiments cover synthetic tool-chain drift, RAG/tool-use stress tests, public SQuAD-derived retrieval tasks, an API-free agentic QA case study, ablations, target-risk sensitivity checks, 20-seed robustness checks, a drift-margin audit, and a live RAG/tool-use agent benchmark. Across these settings, final-answer-only calibration can miss retrieval and tool failures, while trajectory-level calibration keeps accepted-trajectory risk below the target.

The Montreal Forced Aligner (MFA) was released in 2016 and has since become the most widely used tool for forced alignment in research and industry. In the decade since, MFA has undergone substantial development, including expanded coverage across more languages and dialects using larger open-source datasets, harmonized IPA dictionaries, model adaptation, cross-language phone remapping, and support utilities. This paper documents MFA 3.0's developments since version 1.0 and evaluates MFA's performance across English, Japanese, and Korean, benchmarked against classic and neural forced aligners. MFA 3.0 achieves state-of-the-art or near state-of-the-art performance across all four benchmark datasets with mean boundary errors below 15 ms. Adaptation and cross-language remapping are effective for languages outside MFA's training distribution, and pronunciation probability modeling and phonological rules provide gains in specific conditions.

Grokking, the delayed jump from memorization to generalization, is usually tied to the weight norm: a smaller norm generalizes sooner. We ask what the norm actually controls. Holding the weight norm fixed by clamping and varying only an output temperature, we slide the grokking delay across its entire norm-induced range under cross-entropy; matching the effective logit scale back to baseline recovers about 85% of the delay at two moduli. Across a grid of norms and temperatures the delay collapses onto the logit scale alone (R2 = 0.97), with the norm adding 1-2% beyond it. The effect is loss-dependent: under mean-squared error the logit scale is pinned and the norm acts through a different route. A memorization control, a float64 softmax-collapse audit, and a no-LayerNorm transformer point to the same channel. Forking arms from one identical state, the delay follows the held norm value and not the clamp operation, which closes a rescaling-artifact concern. The proximal variable is the logit scale and the softmax saturation it drives; the weight norm is only an upstream handle. All numbers, tables, and figures reproduce from released code and data.

Detecting the tiny Doppler shifts induced by Earth-mass planets in stellar radial-velocity measurements remains extremely challenging due to stellar activity. Many deep-learning methods performing well on simulated data remain difficult to apply reliably on real stellar spectra. The aim of this work is to develop a deep-learning framework that generalizes to real, unseen spectra and improves the detectability of Earth-mass planets in radial-velocity data. We train artificial neural networks on HARPS-N solar spectra with injected planetary signals, using physics-motivated spectral representations based on flux and line-formation temperature, together with their velocity gradients. Two training strategies are explored: hold-out testing and cross-validation. Model robustness is enhanced through genetic-algorithm-based hyperparameter optimization, and predictive uncertainty is quantified using Monte Carlo dropout. Our most precise neural network model reliably retrieves, under the cross-validation strategy, the amplitudes, phases, and orbital periods of planetary signals with amplitudes greater than or equal to 25 cm/s and periods between 10 and 550 days. In addition, in all cases tested here, the successfully recovered signals correspond to the most significant peaks in the periodograms of the Doppler-shift predictions. Temperature-based spectral-shell representations consistently outperform flux-based shells. We also release doppleriann, a Python package implementing the proposed framework. Our results demonstrate that combining physically motivated spectral representations with deep learning provides a promising pathway toward the detection of Earth-mass planets in radial-velocity data from real observations, supported by a modeling framework that is both physically grounded and statistically rigorous, incorporating uncertainty quantification and optimized training strategies.

Distributed stochastic gradient descent (SGD) is limited by communication rather than computation, since each iteration requires an AllReduce across processes. Communication-avoiding SGD (CA-SGD) amortizes communication over $s$ iterations by replacing $s$ consecutive AllReduces with a single AllReduce of an $sb\times sb$ Gram matrix, trading more computation and bandwidth for fewer synchronization points. Modern GPUs with matrix hardware and reduced-precision formats offset this by accelerating the Gram GEMM and shrinking BF16 traffic. We study mixed-precision CA-SGD for generalized linear models on NVIDIA GPUs. Our finite-precision analysis decomposes the local rounding error of one CA-SGD outer iteration into nine independent precision choices, depending on the hardware only through its low-precision unit roundoffs, so the resulting recipes transfer in principle across GPU generations. The recipe stores the input matrix and margin vector in low precision, computes the Gram matrix from low-precision inputs with high-precision accumulation, communicates it in high precision, and performs the inner recurrence and weight updates in high precision. On NERSC Perlmutter A100 GPUs, mixed-precision CA-SGD matches FP32 SGD loss within $0.5\%$ on logistic, linear, and Poisson problems and reaches $5.1$--$6.8\times$ speedup over FP32 SGD on epsilon, SUSY, HIGGS, synth, and Poisson-synth. Our software is available at https://doi.org/10.5281/zenodo.20448273

Recurrent networks can contain substantial functional redundancy in weight space: changing a recurrent matrix may leave the input-output rollout nearly unchanged on a task distribution, while similar-scale changes can destroy the same behavior. We study this redundancy in one-layer tanh RNNs using ordered real Schur coordinates. The Schur form separates spectral blocks from directed nonnormal couplings, giving a diagnostic basis for structured ablations that keep the input and readout maps fixed. In a fixed-length copy task, selected nonnormal Schur couplings can be removed with little loss in some trained solutions, whereas other couplings are necessary for accurate autonomous replay. Across flip-flop, sine generation, and context-dependent integration, the loss-preserving ablation profile varies across tasks and trained solutions. These results identify candidate approximate functional invariances, not universal symmetries of recurrent weight space. Schur-coordinate ablations provide a practical diagnostic for which structured perturbations preserve a trained recurrent solution and which ones disrupt its computation.

We present Veriphi, a GPU-accelerated neural network verification system that combines fast adversarial attacks with formal bound certification using alpha,beta-CROWN methods. Through systematic experiments on MNIST and CIFAR-10 using three training methodologies (standard, adversarial, certified), we demonstrate that training method effectiveness is fundamentally dataset-dependent. Interval Bound Propagation (IBP) achieves 78% certified accuracy on simple MNIST (784 dimensions) but provides negligible certification performance on the more complex CIFAR-10 dataset, where PGD adversarial training dominates with 94% certification at small perturbations. We achieve 5x verification speedup through attack-guided falsification and scale our approach to production-size models (105.8M parameters) for real-world aerospace logistics optimization. Our results challenge the assumption that certified training universally outperforms adversarial training, showing context matters critically for verification strategy selection.

Large language models (LLMs) exhibit substantial cross-lingual variation in mathematical reasoning performance, but it remains unclear whether these differences reflect language-specific parameters or a shared mechanism that manifests differently by language. We present a cross-lingual mechanistic analysis of mathematical reasoning in LLMs, enabling us to localize and compare model parameters that support mathematical reasoning across languages. We find that the extracted math-associated parameters exhibit partial cross-lingual overlap, with the strongest overlap concentrated in intermediate model layers. We further observe that English consistently produces the largest set of math-relevant parameters, whereas lower-resource languages reveal smaller sets of relevant parameters. These results suggest that math-related behavior in multilingual LLMs is neither fully language-invariant nor fully language-specific, but instead exhibits partial cross-lingual parameter overlap with systematic language-dependent differences.

Single-image-to-3D generators are improving quickly, but there is no agreed, human-free way to tell whether one generated mesh is better than another. Practitioners commonly rely on cheap automatic proxies (render-space CLIP similarity and mesh geometry-validity statistics), yet how well these track perceived quality is unestablished. We make two contributions. First, we propose and validate a reproducible VLM-judge evaluation protocol: a fixed 24-view headless render rig, two independent vision-language judge families, and a mandatory position-bias correction that queries both presentation orders and keeps only order-consistent verdicts. The two judge families agree substantially with each other (Cohen's kappa = 0.66), well above the chance-agreement floor. Second, using this protocol as the reference, we show the cheap proxies do not substitute for it. Geometry validity is only a weak signal on average (because, as we show, it is bimodal) and stays below our pre-registered target, while render-CLIP is at chance. A learned Bradley-Terry head collapses onto a single manifoldness statistic (giving render-CLIP a negative weight) and matches geometry-only exactly, so learning the feature weights buys nothing. The proxy is also bimodal: it is significantly above chance on contrasts with visible geometric defects but at chance on ambiguous contrasts, consistent with geometry validity tracking the judge only when the defect is visually salient. We therefore recommend the VLM-judge protocol as a reliable, reproducible evaluator under the conditions tested (two feed-forward generators on Google Scanned Objects, with a face-drop degradation regime) and advise against geometry/CLIP proxies as optimization targets.

Computer-use agents (CUAs) approach human-level performance on standardised benchmarks but still struggle on long-horizon tasks and unseen software. Existing skill libraries address this with reusable skills, but represent the skill artifact as text only, despite the visual nature of GUI interaction. We propose VISUALSKILL: a hierarchical multimodal skill, tailored to each target application and organised as a central index over per-topic files, which the agent consumes through a load_topic MCP tool that fetches the relevant topic's text and figures on demand. We construct each skill with a two-stage pipeline that combines authored documentation with live-application UI exploration. On two CUA benchmarks, CUA-World and OSExpert-Eval, a Claude Code CLI agent backed by Claude Opus 4.6 reaches an average score of 0.456 with VISUALSKILL, a +15.3 point absolute lift over the no-skill baseline (0.303). Against a matched text-only skill that is generated from the same source content and differs from VISUALSKILL only in modality, VISUALSKILL yields a further +8.3 point absolute gain over the matched text-only skill (0.373 vs. 0.456), providing direct evidence that retaining visual figures in the skill artifact, rather than verbalizing them away, helps the agent both identify UI elements and verify workflow state after each action. Our code is available at https://github.com/XMHZZ2018/VisualSkills.

In recent years, credit card fraud detection has faced significant challenges due to highly imbalanced data, evolving fraud patterns, and complex relational structures among transaction entities. To address these issues, this research proposes a novel framework called Timeaware Multi Relational Guided Graph Neural Network (TMR GGNN). Particularly, the proposed TMR GGNN extends the encoder decoder Graph Neural Network GNN architecture by modeling heterogeneous interactions across customers, merchants, devices, and IPs over temporal windows. Subsequently, the proposed TMR GGNN approach constructs a dynamic, multi relational graph and incorporates a time aware relational attention mechanism within the encoder to adaptively weigh the transaction relevance based on temporal proximity and semantic context. Consequently, the decoder employs a contrastive learning module to distinguish between real and synthesized transaction patterns, while improving the models generalization of rare fraud cases. Additionally, to effectively manage severe class imbalances and emphasize discriminative learning, a composite loss function combining Information Noise Contrastive Estimation (InfoNCE) based contrastive loss with Focal Loss is introduced. This integration assists in improving fraud identification while mitigating false negatives.

In this paper, we study a sequential workforce management problem in a contingent labor setting with uncertainty in both worker production and labor supply. A firm seeks to maximize cumulative profit by maintaining an active team of fixed size while learning worker productivity over time. We emphasize two critical operational frictions in this problem: replacing workers is costly, and workers may not be available immediately for hiring because of, for example, prior job commitments, scheduling constraints, or onboarding procedures. Thus, hiring decisions take effect only after a random delay. We formulate this problem as a stochastic multi-play bandit with costly switching and delayed actions, and develop a learning-based hiring policy, DR-UCB (DelayedReplacement-UCB), that makes replacement and hiring decisions sequentially through learning cycles. In each cycle, the policy uses real-time production data to determine when to initiate workforce changes and which workers to replace and hire. We show that the leading-order regret of the proposed policy matches its lower bound in its dependence on the time horizon. Our numerical experiments show that DR-UCB outperforms benchmark policies.

Sparse point observations are increasingly available for precipitation nowcasting, but it is unclear how much they improve dense radar-field forecasts. We partially address this question with a multimodal graph neural network nowcasting system over the Nordic radar domain. The model predicts rain rate every five minutes up to two hours ahead and is trained with different combinations of radar history, MEPS numerical weather prediction, Netatmo surface observations, MSG satellite channels, stochastic noise, and CRPS-based ensemble losses. The study is designed as an ablation of operationally relevant information sources and training objectives. We compare radar-only, NWP-informed, station-informed, satellite-informed, noise-augmented, and CRPS-based configurations using complementary diagnostics on the radar grid, at station locations, for rain onset, and through oracle, displacement, and amplitude scores. The results show that each source improves a different part of the forecast problem. MEPS stabilises radar-only extrapolation, Netatmo observations improve local station and onset diagnostics, and satellite predictors reduce some station-level biases but may activate rain too early when used deterministically. CRPS-based configurations provide the most consistent radar-grid gains, while the combined satellite and CRPS setup gives the best overall oracle/DAS score. These results do not support the conclusion that point observations are uninformative for nowcasting, but they show that local observational skill and spatially coherent radar-field skill are distinct targets. The practical implication is that sparse observations can provide useful local constraints, but their benefit for radar-like fields depends on the training loss, uncertainty representation, and how observation support is encoded in the model.

LLM serving exhibits extreme length variability, making size-based scheduling difficult in practice. Recent LLM schedulers approximate SJF/SRPT using predicted decode lengths or ranks and primarily report mean-centric metrics such as TTFT and TBT. We show that these prediction-driven policies can be fragile under distribution shifts, bursty arrivals, and GPU memory pressure, while offering limited control over the tail latency (P90-P99) that dominates user experience, even with perfect decode-length knowledge. We introduce a distribution-aware, prediction-free scheduling framework that replaces explicit length prediction with soft priority boosting driven by lightweight statistical signals. Our design co-optimizes scheduling and cache-aware preemption to account for memory-coupled decode dynamics across workload mixes. Evaluated on production and open-source traces, our method reduces P99 TTLT by up to 35-50% relative to SRPT with perfect length knowledge and reduces TTFT by 34-47% across workloads, including reasoning-heavy and chat-heavy tasks. These results demonstrate a robust alternative for optimizing tail latency in online LLM serving.

Statistical watermarks help organizations attribute large language model (LLM) outputs, yet existing detectors often struggle when watermark signals are weak, texts are repetitive, or watermarks are edited. We propose signature filtering, a detection-time module that enhances watermark detection without modifying watermark embedding and text generation. It learns a small set of ``signature'' tokens whose presence makes watermark tests unreliable, and removes these tokens before detection. The signatures are obtained by solving a mixed-integer linear program on a small training set, with constraints that maximize the true positive rate. We additionally derive finite-sample and asymptotic bounds under several attacker models (color-blind, color-adaptive, and distributionally correlated). On four well-known watermark families (Kgw, Sweet, Unigram, Exp), four benchmark corpora (C4, MBPP, HumanEval, Code-Search-Net), and six LLMs (Opt-1.3b, Opt-6.7b, Llama2-13b, Llama3.1-8b, Qwen2.5-14b, Phi-3-medium-14b), 2- and 3-gram signatures raise detection rates in weak-signal and low-entropy settings from 8~31% without filtering to 78~99% with filtering, while keeping false positives controllable and often negligible. In stress tests where we scramble sentences and perturb 25~50% of tokens by dilution, deletions, and substitutions, 2-gram filters for Kgw-style watermarks preserve most of the clean-text detection gains, often matching or outperforming the advanced WinMax watermark detector. Signature filtering thus provides a simple, scalable, and model-agnostic add-on to strengthen watermark-based provenance checks for LLM text in information processing workflows.

Accurately aligning CAD models to their corresponding objects in indoor RGB-D scans is a central challenge in 3D semantic reconstruction. The task requires estimating a 9-Degree-of-Freedom (DoF) pose-position, rotation, and scale along three axes-but is hindered by noisy and incomplete scans, as well as segmentation errors that cause geometric distortions. We present Completion-Assisted Object-CAD Alignment (CAOA), a method that integrates a semantically and contextually aware point cloud completion module with a symmetry-aware relative pose estimation algorithm, enabling precise alignment of CAD models to scanned objects. Existing completion methods are typically trained and evaluated on synthetic datasets, which often fail to generalize to real-world scans. To bridge this gap, we introduce a synthetic data generation strategy tailored to indoor scenes, significantly reducing the synthetic-to-real domain gap-validated through quantitative comparisons with widely used completion datasets. In addition, we release S2C-Completion, an expert-annotated dataset of over 8,500 object-CAD pairs from Scan2CAD, created for real-world indoor single-object completion and intended as a new benchmark for this task. For object-CAD alignment, we incorporate symmetry information via a symmetry-aware loss, improving robustness to symmetric ambiguities. On the Scan2CAD benchmark, CAOA achieves a 17% accuracy improvement over state-of-the-art methods.

Scientific workflow management systems (WMS) support scalable and reproducible execution of complex pipelines, but workflow design, implementation, and debugging remain largely manual and require significant expertise. Recent approaches using large language models (LLMs) show promise for workflow generation from natural language, but often rely on direct code synthesis, which limits transparency, reproducibility, and integration with workflow systems. We present an AI-assisted approach to scientific workflow management that combines specification-driven workflow generation, automated debugging, and distributed execution. The method introduces a structured specification phase that separates workflow intent, design, and implementation, allowing validation prior to code generation. We also develop an LLM-based debugging agent that diagnoses and resolves failures across multiple system layers. To support distributed execution and user interaction, we integrate Pegasus, a widely used WMS, with a Model Context Protocol (MCP) layer, providing a unified interface for workflow submission, monitoring, and control. We evaluate the approach using a federated learning workflow for medical imaging, chosen for its parallel, iterative, and dependency-intensive structure. The system generated and executed large-scale workflows with thousands of jobs, reduced debugging effort, and allowed non-expert users to construct workflows with expert-level design patterns. These results indicate that end-to-end AI-assisted workflow generation and execution is feasible, and point toward AI-driven platforms for managing the scientific workflow lifecycle.

This paper develops a variational framework for regulated language generation. Starting from autoregressive token sampling, we derive the induced distribution over complete messages and relate it to an entropy-regularized Gibbs law. Regulation is modeled as an optimal discriminator whose convex-dual value is an f-divergence, and the generator-regulator interaction is formulated as a saddle-point problem. The framework applies to moderation, censorship, AI deception detection, compliance auditing, phishing defense, and manipulation control, where regulation concerns a distribution over possible messages rather than a single output. The equilibrium clarifies the tradeoff among utility, entropy, regulatory alignment, and finite-length detectability. Two finite-vocabulary case studies, censorship filtering and phishing defense, illustrate how the theory can be evaluated through utility, entropy, divergence, receiver-side scores, and detection probability.

Educational resources may frame students' perceptions of who belongs in software engineering, which is relevant given the field's ongoing gender gap. However, we know little about the hidden curriculum regarding gender in online learning spaces. This study presents a critical discourse analysis of 200 manually analysed English and German software engineering tutorials on YouTube, examining gender representation through contextual domains and linguistic identity markers. Our results show that male characters and masculine linguistic defaults dominate the tutorials. We identified an agency gap, in which technical and decision-making roles are almost exclusively assigned to male actors, while female actors are either absent or tend to passive, low-agency roles. The findings indicate that software engineering education on YouTube may reproduce gendered norms, in which linguistic and representational gatekeeping may serve as a symbolic barrier to software engineering.

The growing deployment of artificial intelligence (AI) necessitates robust deep learning (DL) compilers, such as TVM and ONNX-MLIR. These compilers take as input high-level AI models, lower them through multi-layer transformations, and specialize them to diverse hardware. Testing such compilers is uniquely challenging as correctness depends on implicit constraints embedded throughout the compilation stack. Existing testing approaches largely take type constraints to restrict input model generation and therefore emphasize type validation and monitor compilation crashes or coverage gains. This focus overlooks compiler-platform interaction bugs that arise from interleaved effects across compilation and execution environments. In this work, we propose a scalable, automated DL compiler testing framework for, in tandem, (1) finding compiler-platform interaction bugs and (2) enabling behavior equivalence partitioning. Our key insight is that these bugs are caused by violated assumptions arising from interactions across compilation passes and hardware platforms. Therefore, we move beyond constraining input generation and derive full-stack constraints. Our approach is three-fold. First, we design an automated approach to extract full-stack constraints that jointly guide model generation and characterize compilation behaviors. Second, we prioritize constraints that expose interaction-sensitive behaviors, so our generated models are capable of exercising deep compilation logic. Third, we enable behavior equivalence partitioning by automatically inserting assertions to monitor distinct compilation symptoms that coverage or pass/fail signals miss. We evaluated our tool, XCheck, on three widely-used DL compilers and found 2,034 bug-revealing cases, including memory overflows, integer overflows, and silent unexpected compilations that were rooted in compiler-platform interactions.

On biomedical tabular data, flexible models such as deep networks, gradient-boosted trees, and kernel methods are repeatedly matched or beaten by linear and logistic regression given the same features. The usual reaction is to treat this as a model-side shortfall, to be fixed with more data, a better architecture, or tuning, on the assumption that the nonlinear structure is there and the model has failed to capture it. We argue that these fixes cannot help when the binding limit is the measurement rather than the model, as it frequently is in biomedicine. Additive noise blurs the population-optimal predictor, and because blurring removes a function's fine, rapidly varying detail before its broad shape, it erases nonlinear structure faster than linear structure. A degree-$k$ interaction is attenuated by the $k$-th power of feature reliability, while the linear part is attenuated only once. At the reliabilities typical of biomedical measurement, the nonlinear advantage can vanish even when the underlying biology is strongly nonlinear, and what the noise removes cannot be recovered by a larger cohort or a more flexible model, only by better measurement. The nonlinearity is hidden, not absent, and a tie between linear and flexible models is not by itself a verdict on the biology. These pieces are classical, drawn from measurement-error statistics, psychometrics, and Gaussian analysis, and we assemble them into an exact excess-risk identity. Measurement reliability is one of three conditions, alongside sample size and feature representation, that must align for a flexible model to help, and together they leave only a narrow window that most biomedical tasks fall outside. Across 140 UK Biobank tasks, the gap between flexible and linear models, where it exists, carries the predicted noise signature, and the three conditions can be separated by intervention but not by a benchmark alone.

The increasing use of machine learning algorithms in social applications has raised concerns about fairness and transparency, leading to the development of counterfactual explanations. These explanations supports individuals to understand and potentially alter unfavorable decisions in areas such as loan applications, job selections, and more, by providing actionable changes to input features that would lead to a desired outcome. Existing methods often struggle to balance feasibility, plausibility, and computational efficiency. To address this, we introduce P$^2$CE, an algorithm for generating plausible Pareto-optimal counterfactual explanations, offering users a diverse set of optimal trade-offs between different notions of feasibility. P$^2$CE employs an auxiliary isolation forest outlier detector to ensure that explanations are in accordance with the data distribution and leverages SHAP values to obtain optimal results with short computing times, regardless of the underlying model. Our algorithm was empirically evaluated on three datasets, demonstrating superior performance in terms of both solution quality and computational efficiency compared to related techniques.

Automated AI agents are increasingly capable, yet many scientific and professional tasks require human judgment and contextual expertise. We study shared-workspace human-AI teams, where AI agents and human collaborators must coordinate responsibilities before submitting a final answer. Using the Collaborative Gym environment with DiscoveryBench tasks, we examine when adding simulated human collaborators improves performance and when process loss turns additional collaborators into coordination overhead. Across 1,482 sessions, adding relevant collaborators can lower performance when teams lack structure to coordinate their contributions. We then evaluate scaffolding that combines shared group memory with simulated human-in-the-loop (HITL) gates, where selected actions require approval from a designated simulated participant. This scaffolding yields higher mean performance, most clearly in three-person teams, with clearer responsibility signals and stronger routing of expertise to team actions. Overall, how human-AI teams coordinate and integrate expertise matters as much as the capability available to them.

Personalized dialogue agents require continuous long-term memory to maintain coherent interactions across multiple sessions. However, deploying these capabilities on consumer-grade hardware (e.g., 8 GB VRAM edge devices) introduces severe memory and compute bottlenecks. Existing systems typically rely on isotropic cosine similarity for retrieval and heuristic rules for context compression. These approaches lack a unified theoretical foundation, frequently suffering from the hubness problem in high-dimensional retrieval and syntactic fragmentation during compression. To overcome these limitations, we propose CoreMem, a resource-efficient edge-cloud memory architecture fundamentally unified by information geometry. First, Riemannian retrieval replaces cosine matching with a locally adaptive Fisher-Rao metric, effectively penalizing hub memories via Mahalanobis distance with O(Ndr) Woodbury acceleration for real-time search. Second, Fisher-guided discrete token distillation (FDTD) introduces a hierarchical sentence-to-token compression mechanism. It derives sensitivity scores from Fisher information traces, providing a principled compression-KL tradeoff augmented with explicit structural syntax protection. Evaluated on the LOCOMO and LongMemEval-S benchmarks, CoreMem achieves strong accuracy improvements, yielding substantial gains in Open-domain (+4.51 pp) and Temporal (+4.17 pp) reasoning. Extensive profiling confirms that CoreMem operates seamlessly within a strict 8 GB VRAM budget, successfully bridging the gap between resource-constrained edge devices and the demand for theoretically grounded, lifelong memory agents.

Traditional microwave filter design typically relies on iterative parameter tuning and predefined topologies, which limits design space and increases development time. This study uses a deep learning approach combining convolutional neural networks with genetic algorithms to automate pixelated microwave filter synthesis. To validate the approach experimentally, both S-parameter and spatial electric-field measurements were analyzed. The synthesized low-pass filter demonstrated excellent agreement between simulated and measured performance, achieving a 7 GHz passband with over 20 dB suppression beyond 9.5 GHz. Electro-optical measurements, for the first time, revealed electric field patterns that resemble coupled transmission-lines or stub structures, providing insight into the emergent characteristics of AI-generated designs.

The output combiner of a Doherty power amplifier (PA) integrates load modulation, impedance matching, and phase compensation within a single network, making its design and synthesis highly challenging. In this paper, we propose a three-port Doherty combiner design methodology that combines deep convolutional neural networks (CNNs), pixelated layout representations, and genetic algorithms (GA) with dual-state impedance synthesis to address both peak and back-off power conditions. As a proof of concept, two GaN HEMT Doherty PA prototypes incorporating three-port pixelated combiners are designed and fabricated. Both prototypes achieve a measured saturated output power exceeding 44.2 dBm with peak drain efficiency above 71.2% within 2.6-2.8 GHz. Furthermore, a drain efficiency as high as 64% is measured at the 6-dB back-off level. After applying digital predistortion, each prototype achieves an adjacent channel leakage ratio (ACLR) better than -51.3 dBc.

Speculative decoding (SD) accelerates autoregressive Large Language Models (LLMs) by drafting multiple tokens and verifying them in parallel, but it faces a scaling limitation: increasing the draft budget improves speed only when acceptance remains high and drafting overhead stays low. This ceiling has been difficult to break because prior head-based SD methods face a causality-efficiency dilemma. Autoregressive drafters produce path-conditioned candidates that are effective for tree speculative decoding with higher acceptance length, but their drafting cost grows with tree depth. Bidirectional block-diffusion drafters generate all positions in one pass, but their branch-agnostic marginals can form individually plausible yet mutually inconsistent trees, wasting budget and reducing acceptance. We propose JetFlow, a head-based SD framework that combines one-forward drafting efficiency with branch-wise causal conditioning. JetFlow trains a causal parallel draft head over fused hidden states from the frozen target model, producing candidate trees whose scores align with the target model's autoregressive factorization. This enables JetFlow to convert larger draft budgets into longer accepted prefixes and higher end-to-end speedup. Across math, coding, and chat benchmarks on dense and MoE Qwen3 models, JetFlow consistently outperforms bidirectional-head and tree-based SD baselines. On H100 GPUs, JetFlow achieves up to 9.64x speedup on MATH-500 and 4.58x on open-ended conversational workloads, with further latency gains demonstrated through vLLM integration under realistic serving loads. Our code and models are available at https://github.com/hao-ai-lab/JetFlow.

Multi-fuel compression-ignition engines offer fuel flexibility but introduce uncertain, time-varying fuel reactivity, represented by cetane number (CN), which complicates cycle-to-cycle combustion-phasing control. This work formulates CA50 regulation under latent CN variation as a partially observable sequential decision problem and systematically evaluates controllers with increasing temporal and representational capacity, including LinUCB, history-augmented contextual bandits, observation-only DDPG, recurrent DDPG, and a proposed GRU-guided RL framework. A Gaussian-process surrogate trained on experimental multi-fuel engine data provides a controlled and reproducible evaluation environment. Results show that myopic and fixed-history bandit methods degrade under CN variation, observation-only RL suffers from latent-state aliasing, and generic recurrence is insufficient when CN evolves rapidly. The proposed framework learns a compact GRU-based representation of fuel reactivity from combustion history and conditions both actor and critic on this estimated signal rather than oracle CN. By training the policy on the same imperfect fuel-reactivity information available at deployment, the controller avoids train-deploy inconsistency in conventional online estimate-then-control pipelines. Across unseen CN trajectories, the policy achieves stable CA50 regulation with mean absolute tracking error below 0.25° CA at the training setpoint, while producing smooth, physically consistent SOI and glow-plug-power actuation. These results show that combustion control under latent, continuously evolving fuel dynamics requires more than standalone estimation or generic recurrence. By aligning fuel-reactivity inference with control policy learning, the proposed framework enables reactivity-aware decision-making using the same estimated state available during deployment.

Motivation: Noisy labels are a common challenge in molecular property prediction because molecular annotations are often obtained from assays, curated databases, or weak annotation pipelines rather than directly observed clean biological states. Treating recorded labels as reliable supervision can cause models to memorize corrupted observations and learn misleading molecular evidence. In multimodal molecular representation learning, this issue can be amplified by graph-text fusion or alignment, which may propagate label-induced errors across modalities. Results: We propose MOLAR, a noise-aware framework for learning multimodal molecular representations from noisy labels. MOLAR separates latent clean-property inference from recorded-label observation: graph and text views contribute residual evidence to a clean-property distribution, and a categorical label-observation channel maps this distribution to recorded labels for training. This formulation derives posterior label reliability and modality-specific molecular evidence from the model. Experiments on naturally noisy molecular benchmarks and controlled label-flipping benchmarks show that MOLAR consistently outperforms representative baselines. Visualization analyses further show that MOLAR provides interpretable reliability and modality-evidence diagnostics.

Large language models (LLMs) have become an effective tool for synthetic data generation, including for low-resource languages, where generated data can improve downstream task performance. Current best-performing approaches typically rely on few-shot prompting with target-language examples, which increases inference costs and may reduce diversity through lexical anchoring. In this work, we investigate activation steering as an alternative for low-resource synthetic data generation. We study two steering strategies: Language Steering, which targets the linguistic identity of a language, and Quality Steering, which captures well-formedness by contrasting human-written and backtranslated text representations. We evaluate these methods across four open-source LLMs, multiple layers, and 11 typologically diverse languages by generating sentiment and topic classification data and finetuning smaller classifiers. Steering is applied in both zero-shot and few-shot prompting settings and compared against non-steered counterparts. Our results show that steering on early layers consistently improves the diversity of generated data while often yielding stronger downstream model performance, particularly for low-resource languages.

RL post-training strategies are dataset-dependent and reveal a recurring empirical pattern: capacity parameters accumulate monotonically across stages, while regularization parameters predominantly oscillate in response to shifting training dynamics. This distinction matters because fixed schedules commit all parameters to fixed trajectories and therefore cannot express the non-stationary exploration-exploitation tradeoffs that regularization must track; the principle provides actionable design rules for multi-stage training. We discover this through LLMZero, a system where LLM agents search over training trajectories via tree search, diagnosing pathologies at each checkpoint and proposing coordinated multi-parameter transitions. Across 4 diverse GRPO tasks, LLMZero discovers strategies that improve over the base model by 9% to 140% relative and over grid search by 6% to 15% relative, consistently outperforming random search and the skill-based agent. The structural principle transfers across tasks, providing an explanation for why discovered strategies take qualitatively different forms yet share similar parameter dynamics.

Vision-Language Models (VLMs) remain prone to hallucinations, producing fluent but visually unfaithful outputs. Existing chain-of-thought and retrieval-augmented methods only partially address this, as they neither enforce step-level citation grounding nor route verification failures back to retrieval for correction. We present CaVe-VLM-CoT, a modular reflection-based agentic-RAG framework that enforces evidence-grounded reasoning through a five-stage closed-loop pipeline: Extractor, Retriever, Solver, Citation Injector, and Verifier, in which detected ungrounded claims trigger structured feedback to the Extractor for targeted re-retrieval. Since no existing framework jointly measures retrieval quality, step-wise citation faithfulness, and cross-modal grounding, we propose a suite of 23 component-wise metrics across all stages, anchored by CaVeScore, a composite metric weighting accuracy, citation precision and recall, attribution, and evidence grounding. Without any architectural or prompt modifications, CaVe-VLM-CoT achieves 87.1\% accuracy and 56.6\% CaVeScore on ScienceQA , and 55.2\% accuracy and 35.7\% CaVeScore on MMMU (30 subjects).

Hierarchical Federated Learning (HFL) enables scalable collaborative model training across distributed devices while preserving data privacy. However, existing HFL client selection mechanisms suffer from a fundamental strategic inefficiency. By prioritizing stability over Pareto efficiency (PE), they produce suboptimal resource allocations, and without strategy proofness (SP), participants are incentivized to misrepresent their true preferences, both failures degrading system overall welfare in the Pareto sense in practice. To address it, we propose SCOPE-FL (Strategy-proof Chain-based Optimal pareto efficient Federated Learning), a synchronous HFL framework that formulates client selection as a two-sided school choice problem solved through the Top Trading Cycle (TTC) algorithm that simultaneously guarantees PE and SP. For reward distribution, SCOPE-FL employs a scalable Shapley value approximation based on One-Round Reconstruction (OR), ensuring compensation proportional to each client's contribution. The entire mechanism executes via blockchain smart contracts, providing the tamper-proof environment required for the SP guarantees to hold in practice. A comprehensive evaluation on MNIST, Fashion-MNIST, and CIFAR-10 demonstrates that SCOPE-FL outperforms state-of-the-art approaches, including DA, IAS, and other methods across model accuracy, convergence rate, and reward efficiency, while achieving communication latency comparable to DA and blockchain overhead significantly lower than DA at scale.

Sparse autoencoders (SAEs) are increasingly used to extract interpretable features from language models (LMs), yet a central question remains: when can an SAE-based explanation be treated as a faithful view of an underlying frozen LM We study this through a post-hoc generalization framework that certifies the LM via a sparse proxy, obtained by replacing a native hidden activation with its pretrained SAE reconstruction. Our framework derives an upper bound on the base model's expected risk using four measurable quantities: proxy risk, SAE reconstruction gap, concept-pool mismatch, and sparse complexity. We interpret this certificate as an operational criterion for explanatory faithfulness. In particular, a non-vacuous bound indicates that the extracted sparse features retain meaningful predictive information, while small reconstruction and mismatch errors indicate that the proxy remains behaviorally close to the original model. Empirically, we show that the bound becomes non-vacuous on GPT-2 Small, Gemma-2B, and Llama-3-8B at practical sample sizes. A detailed layerwise analysis of Llama-3-8B reveals a strong depth dependence, with later layers becoming much easier to certify, associated with both stronger local fidelity and weaker downstream error amplification. Finally, through feature-shuffling ablations, we show that the decomposition distinguishes genuine semantic alignment from mere statistical sparsity, providing a useful diagnostic for when SAE-based explanations become less reliable.

Retrieval-augmented generation (RAG) systems must balance retrieval granularity with contextual coherence, a challenge that existing methods address through LLM-guided chunking, single-level context expansion, or hierarchical summarization. These approaches variously depend on costly LLM calls during indexing or retrieval, limit context aggregation to a single granularity level, or introduce information loss through summarization. We present SproutRAG, an attention-guided hierarchical RAG framework that addresses this trade-off by organizing sentence-level chunks into progressively larger but semantically coherent units, using learned inter-sentence attention to construct a binary chunking tree. Unlike prior approaches that rely on external LLMs, fixed context expansion, or lossy summarization, SproutRAG learns which attention heads and layers best capture semantic document structure, enabling multi-granularity retrieval without additional LLM calls or compressed summaries. At retrieval time, SproutRAG uses hierarchical beam search to retrieve candidates at multiple granularities, capturing multi-sentence relevance beyond flat retrieval. The framework is trained end-to-end with a joint objective that improves both embeddings and tree structure. Experiments across four benchmarks spanning scientific, legal, and open-domain settings demonstrate that SproutRAG improves information efficiency (IE) by 6.1% on average over the strongest baseline. Code is available on https://github.com/AmirAbaskohi/SproutRAG.

Graph-based retrieval at billion-node scale requires jointly solving three tightly coupled problems -- graph construction, representation learning, and real-time serving -- yet existing work addresses each in isolation. We present RankGraph-2, a framework deployed at Meta that co-designs all three lifecycle stages for similarity-based retrieval (U2U2I and U2I2I), where each stage's requirements shape the others. Serving requires a co-learned cluster index to avoid expensive online KNN -- this pushes index co-training into the training objective. Training benefits from the observation that similarity-based retrieval tolerates pre-computed neighborhoods, eliminating online graph infrastructure -- this requires construction to produce self-contained data. Construction must also support hour-level refresh for item coverage. Acting on these cascading requirements, RankGraph-2 reduces hundreds of trillions of edges to hundreds of billions via subsampling with popularity bias correction, pre-computes multi-hop neighborhoods via personalized PageRank, and co-learns a residual-quantization cluster index that reduces serving computational cost by 83%. This lifecycle co-design enables a simple architecture to achieve 3.8 x higher recall than a GAT + Deep Graph Infomax model on a bipartite graph and 2.1 x higher than PyTorch-BigGraph on item retrieval. RankGraph-2 delivers up to +0.96% CTR and +2.75% CVR, and has powered 20+ retrieval launches across major surfaces.

Software performance regressions can have significant business consequences, making automated detection a critical component of modern continuous integration pipelines. At Mozilla, performance anomaly detection is handled by Perfherder, Mozilla's performance engineering management system that relies on a Student's T-test-based approach to flag regressions across hundreds of daily code changes. However, our preliminary analysis of one year of Mozilla performance data reveals that 12.5% of generated alert groups are false positives, while approximately 6.8% of them contain regressions missed by the automated system. This paper presents an empirical study evaluating 25 change-point detection (CPD) methods and 15 ensemble approaches as alternatives to Mozilla's current method. We construct a ground-truth dataset of 174 performance time series manually annotated by eleven Mozilla performance engineers, representing one of the first practitioner-annotated CPD benchmarks for performance engineering. Our results show that while offline and hybrid CPD methods improve recall over Mozilla's method, they do so at a high cost to precision. Ensemble voting strategies alleviate this trade-off and offer more consistent performance, resulting in 11% improvement in the F1-score. We validate the experimental results through a practitioner survey and report on lessons learned from integrating the best methods into Mozilla's performance engineering system.

Educational dialogue is a valuable but sensitive resource for research: the same transcripts that capture authentic learning often capture personally identifiable information (PII) entangled with curricular content, where "Riemann" may refer to a real student or to a mathematical concept. Existing approaches force a tradeoff between governance and accuracy. Commercial Large Language Models (LLMs) can handle this ambiguity but require sending student data to third parties, while local named entity recognition (NER) systems preserve governance but over-redact curricular terms. We propose a fully local cascade framework that reframes de-identification from open-ended entity recognition to constrained privacy triage. A recall-first union proposer combines two lightweight encoders with deterministic rules to over-generate candidate spans; a context-aware reviewer then makes a binary Redact/Keep decision for each candidate using surrounding dialogue and speaker role. We evaluate three reviewer configurations against same-family LLM-only baselines and a commercial API on math tutoring transcripts from two large platforms. The strongest local configuration reaches 0.958 macro F1, compared with 0.767 for a same-family LLM-only baseline and 0.706 for the commercial API, while running entirely on a single laptop. On a targeted challenge set of curricular-personal name ambiguity, the same configuration degrades by only 0.03 F1 versus 0.19 to 0.25 for smaller reviewers. These results suggest that for educational de-identification, problem formulation matters more than model scale.

Standard benchmarks evaluate time series foundation models (TSFMs) using aggregate metrics, but these can mask severe failures in critical operating regimes. We introduce regime-stratified evaluation and apply it to three TSFMs on two standard traffic speed benchmarks. Traffic exhibits abrupt regime switching between free-flow and congested states, producing bimodal speed distributions during transitions. When we stratify by traffic regime, both accuracy and prediction-interval coverage degrade sharply during transitions: transition-regime MAE reaches 11 mph (versus 3 mph overall), and empirical coverage of 90% prediction intervals drops as low as 55%. These failures are invisible in aggregate metrics because free-flow observations dominate the sample. A simple historical conditional baseline (sampling from per-sensor training distributions) achieves better transition coverage than any TSFM, but has far worse overall accuracy. We propose bimodal mixture augmentation (BMA), a post-hoc method that combines TSFM forecasts with historical distributional knowledge, approaching the historical baseline's transition coverage while preserving the TSFM's accuracy. Our results suggest that TSFM benchmarks should incorporate regime-aware evaluation to surface failures that aggregate metrics hide.

Language models trained on large-scale vision-language data have demonstrated strong potential for embodied agents. Harnessing models through embodied tools use offers a promising alternative to end-to-end vision-language-action systems by combining high-level reasoning with external modules for perception, planning, and control. However, it remains unclear what makes an effective harness for embodied manipulation, and to what extent such a harness can unlock embodied capabilities in a wide range of reasoning models. In this work, we present Guava, a harness framework for embodied tool use developed through systematic exploration of the design space of agent workflows, action spaces, and observation spaces. Our study identifies three key ingredients for effective embodied agents: iterative perception-reasoning-action loops, semantic action abstractions, and multimodal observations. To understand whether these design principles are universal even to small models, we develop an end-to-end training pipeline that distills embodied manipulation capabilities into a 4B open-source model using fewer than 2K trajectories collected entirely in simulation. Experimental results in both simulation and real-world environments show performance comparable to frontier proprietary models while exhibiting strong generalization to unseen objects, novel instructions, and long-horizon tasks. Results suggest that a well-designed harness can serve as a scalable, model-agnostic interface for embodied manipulation, enabling strong emergent embodied capabilities in compact open-source models with minimal training data.

Tool-using language-model agents introduce security failures that go beyond unsafe text: they can disclose protected objects, write persistent memory, send messages, modify databases, or trigger harmful code and tool effects. Existing evaluations often collapse these stages into a single attack success rate, making it difficult to tell whether a model merely agreed with an attacker or actually produced observable harm. We introduce SafeClawBench, a staged benchmark for tool-using agent security with 600 controlled adversarial tasks across six attack families: direct and indirect prompt injection, tool-return injection, memory poisoning, memory extraction, and ambiguity-driven unsafe inference. SafeClawBench reports three separate endpoints: semantic attack acceptance, audit-visible harm evidence, and sandbox-observed tool/state harm. Evaluating five agent endpoints under four prompt-level policies, we find that these endpoints capture different failure modes. Without additional prompt protection, semantic failure rates vary widely across models, from 9.0% to 44.2%. Audited harm evidence is narrower than semantic failure, and under a separate executable protocol some matched task identities produce sandbox harm despite passing the Semantic Core call: in a 12,000-row matched analysis, 291 of 347 observed sandbox harms occur in rows that pass the semantic check. Prompt policies change endpoint outcomes, but their effects depend on both model and protocol. SafeClawBench provides a reproducible framework for comparing agent models and prompt-policy conditions without conflating textual compliance, evidence-supported harm, and executable state changes. The open-source dataset is available at https://huggingface.co/datasets/sairights/safeclawbench.

Recent advances in generative machine learning models have significantly improved medical imaging, offering promising solutions for data augmentation, privacy preservation, and improved model generalization. However, synthesizing high-quality structural MRI data for Alzheimer's Disease (AD) remains challenging due to the subtle, region-specific, and progressive anatomical changes associated with neurodegeneration. In this paper, we extend the Med-DDPM conditional diffusion model -- originally designed for brain tumor synthesis -- to generate 3D structural MRIs specifically tailored to AD. We adopted Med-DDPM due to its established stability and structural fidelity compared to other generative models, which makes it particularly suitable for capturing the subtle anatomical changes characteristic of AD. Our approach conditions the diffusion process on anatomical segmentation masks derived from the ADNI dataset, incorporating key AD-relevant brain structures into the generation process. We systematically evaluate the quality and utility of the synthetic images by training segmentation models on real, synthetic, and hybrid (mixed) datasets. Experimental results demonstrate that segmentation models trained exclusively on synthetic data achieve comparable Dice scores (0.6532) to those trained on real data (0.6513), while exhibiting significantly enhanced recall. Notably, models trained on hybrid datasets (mixing real and synthetic images) outperform both real and synthetic-only baselines, achieving a Dice score of 0.7244. These findings underscore the successful use of conditional diffusion models for generating anatomically accurate, AD-specific synthetic MRIs, and highlight their potential for enhancing training data availability, improving diagnostic accuracy, and promoting research reproducibility in neuroimaging studies.

The search for life beyond Earth will depend on detecting faint signatures in the atmospheres of potentially habitable exoplanets. Interpreting those signatures requires understanding the host planet's climate: the same molecule may signal life on one planet and abiotic chemistry on another. Global climate models (GCMs) provide this understanding, but individual runs can require up to millions of core-hours and substantial domain expert time. Machine-learning emulators could remove this bottleneck, but progress has been limited by the absence of a curated, multi-model exoclimate dataset. We introduce ThousandWorlds, an ML-ready benchmark for exoclimate emulation and for the broader regime of low-data, multi-simulator, parameter-to-field regression. The dataset contains approximately 1800 simulations from five GCMs, mapping eight planet parameters to 3D atmospheric fields including temperature, humidity, winds, clouds, and radiation. Three nested subsets define progressively harder challenges: single-simulator regression, multi-simulator regression with complete observations, and multi-simulator regression with structured missingness. We propose two evaluation protocols: one for ranking methods, and one that measures performance relative to the disagreement between GCMs themselves. We evaluate seven baselines spanning simple methods, deep learning, and Gaussian processes. GP-based methods perform best, suggesting that ThousandWorlds exposes a regime where off-the-shelf deep learning does not yet succeed. Data: https://doi.org/10.57967/hf/8695. Code: https://github.com/edstevenson/ThousandWorlds.

Language models (LMs) that faithfully describe their own behavior can more easily be audited, understood, and trusted by users. This paper describes Self-Consistency Training with Reinforcement Learning (Self-CTRL), a method that optimizes for consistency between a LM's self-explanations and behavior on related inputs by updating explanations to better predict behavior or updating behavior to better match explanations. We apply our method in two domains. First, we study a formal probabilistic reasoning task in which LMs must learn to imitate a family of biased samplers and evaluated on their ability to report the associated biases. We find that consistency training improves the correlation between self-reported and behaviorally-measured latent biases from $R^2=0.24$ to $R^2=0.64$ on a set of held-out distributions, matching the generalization of direct ground-truth supervision. Second, we study a constitutional AI domain in which LMs must describe when they will refuse or comply with user requests. Here, Self-CTRL produces rules that faithfully describe the model's behavior on held-out requests, improving the refusal predictions of a third-party auditor model from $36\%$ to $92\%$. In the other direction, behavior updates improve alignment, reducing HarmBench failure rate from $15.0\%$ to $0.5\%$ without substantially increasing refusal on harmless prompts. By aligning explanations and behavior, our work provides a general recipe for training AI models to be safer, more transparent, and more controllable.

Robots deployed in the real world should learn from their experience and improve over time. This requires a mechanism of practicing and learning from feedback. In this paper, we propose VERITAS, a generator-verifier framework for generalist robot policies for inference-time policy steering and self-improvement. We use a pre-trained generalist robot policy as a ``generator'' and pair it with a gradient-free ``visual verifier'' that evaluates actions at inference time. This framework enables inference-time steering that improves policy performance without additional training. We demonstrate that inference-time verification consistently outperforms vanilla generalists without training on additional demonstration data. Additionally, we demonstrate that the verified rollouts provide effective supervision for offline policy improvement: policies fine-tuned on verified self-generated trajectories achieve consistent performance gains. Notably, we find that post-training with verified rollouts achieves comparable efficiency to expert demonstrations, while requiring no human interventions. Our results highlight inference-time verification as a practical and scalable mechanism for improving robotic policies during deployment.

Scaling model size, specifically depth and width, has driven significant progress in transformer-based language models. However, most architectures maintain a constant width across all layers, allocating a fixed parameter and computation budget evenly despite different layers potentially playing distinct computational roles. In this work, we empirically investigate nonuniform capacity allocation across network depth by proposing a $\times$-shaped > <former architecture. This design maintains wider early and late layers while narrowing the middle layers, utilizing a parameter-free residual resizing mechanism. Across decoder-only language models ranging from 200M to 2B parameters (dense) and 3B parameters (MoE), our > <former consistently outperforms parameter-matched uniform baselines on language modeling loss. By reducing the average layer width, this architecture also requires fewer overall FLOPs (22% reduction under fitted loss-matched scaling curves) and smaller KV cache memory and I/O cost (15% reduction). In analysis, we show that this bottleneck structure results in qualitatively different representations in residual streams. Overall, our results demonstrate that nonuniform width allocation can result in more resource-optimal scaling of language models.

Reproducing research results from papers and released code is central to scientific progress. Existing works have introduced benchmarks to evaluate whether LLM agents can assist with reproducibility, but they are difficult to scale due to their reliance on substantial manual effort for data curation and evaluation. We introduce ReproRepo, a scalable framework for reproducibility evaluation that leverages human-raised GitHub issues as naturally occurring supervision on realistic reproduction blockers. We instantiate ReproRepo on 1,149 recent machine learning papers from major conferences and evaluate four frontier model-agent configurations. Our results show that LLM agents, even without executing code, can identify many real-world reproducibility problems from paper-repository pairs: the best agent in our study, namely Codex with GPT-5.5, surfaces at least one semantically related human-reported blocker for ~90% of papers in the study. Further analysis shows that agents are particularly effective for surfacing visible failures and identifying the right semantic region, but may still be insufficient in exact localization. ReproRepo can serve as a reusable, scalable framework for future evaluations of LLM agents on real-world reproducibility auditing. Our code is released at https://github.com/LithiumDA/ReproRepo.

In learning theory, the sign rank of a binary concept class captures the smallest dimension in which it can be represented by points and halfspaces. Despite tremendous interest, lower bounds on sign rank are notoriously difficult to come by. Two recent approaches to the problem establish lower bounds on sign rank by measures that are easier to analyze: the $\mathbb{Z}_2$-index and the list replicability number. We order these measures, showing that the $\mathbb{Z}_2$-index is upper-bounded by a linear function of the list replicability number. As a main consequence, we obtain a strong separation between sign rank and $\mathbb{Z}_2$-index, thereby resolving a question of Frick, Hosseini, and Vasileuski. This motivates a thorough study of list replicability, the stronger of the two lower-bounding measures. We establish upper bounds on the list replicability number by two combinatorial measures: height and minimum star number. We also prove a fundamental composition result, showing that the product of two concept classes has list replicability number bounded by the sum of the list replicability numbers of the two classes.

Zero-Shot Object-Goal Navigation (ZS-OGN) requires embodied agents to explore and locate target objects without any prior training. To this end, recent methods leverage foundation models. But they typically rely on static priors and lack adaptation, which leads to repeated errors and costly trial and error. In this paper, we propose a self-evolving ZS-OGN framework that enables continuous test-time improvement. Specifically, we build an agentic rule memory by extracting actionable knowledge from past trajectories. Then, we propose a retrieval strategy based on upper confidence bound, selecting effective rules by balancing semantic relevance and historical success. In addition, we introduce a memory-guided preflection module that forecasts potential outcomes before action, reducing inefficient exploration. Extensive experiments show that our method outperforms existing zero-shot baselines, achieving a 10.1\% improvement in success rate with fewer unnecessary steps.

Accurate mechanical properties (or materials) Young's modulus ($E$), Poisson's ratio ($ν$) and density ($ρ$) are essential for reliable physics simulation of digital worlds, but most 3D assets lack this information. We propose AdaVoMP, a method for predicting accurate dense spatially-varying ($E$, $ν$, $ρ$) for input 3D objects across representations, improving the resolution, accuracy, and memory efficiency over the state-of-the-art. The foundation of our technique is a sparse and adaptive voxel structure SAV that efficiently represents both the input 3D shape and the material field output. We replace the fixed-voxel model of the most accurate prior method, VoMP, with a novel sparse transformer encoder-decoder model that learns to generate a unique SAV autoregressively for every input shape to represent its materials, achieving a resolution $16^3\times$ higher than prior art. Experiments show that AdaVoMP estimates more accurate volumetric properties, even with lesser test-time compute than all prior art. This allows us to convert high-resolution complex 3D objects into simulation-ready assets, resulting in realistic deformable simulations.

With sophisticated cyber-attacks becoming increasingly prevalent, modern networks require intelligent autonomous cyber-defense agents trained via Reinforcement Learning (RL). These agents employ neurosymbolic approaches such as behavior trees with learning-enabled components (LECs) to learn, reason, adapt, and implement security rules while maintaining critical operations. However, these autonomous networks are partially observable systems, i.e., the cyber-attacker's (red agent's) actions are not observable, making it difficult for the defender to predict red actions, learn red policies, or assess the attacker's intrusion levels. To address this, we propose a Policy Learning Technique using imitation learning to learn policies for partially observable RL agents with discrete states and discrete actions. We apply this technique in an autonomous cyber environment to predict red agent's actions from network observations and defender actions. Integrated with a neurosymbolic cyber-defense agent, our method effectively handles different red policies and achieves high prediction accuracy across diverse simulated scenarios.

We introduce Darshana Graph, a corpus of over 125,000 text records spanning classical Hindu, Buddhist, and Jain philosophical traditions, drawn from public-domain and openly licensed translations of sources including the Bhagavad Gita, Brahma Sutras, principal Upanishads, the Pali Canon, and core Jain texts. Its distinctive contribution lies in a structurally unique subset of roughly 8,500 Hindu and Jain records in which the same root verse or sutra is aligned across eighteen historical commentators representing five schools of Vedanta and other darshanas, enabling direct comparison of how independent interpretive traditions read identical source material. To our knowledge, no publicly available resource provides comparable cross-commentator alignment at this scale. We present two analyses built on this corpus. First, a transparent stylometric comparison requiring no machine learning measures argumentative style through scriptural citation density, explicit refutation rate, and sentence complexity. It finds a moderate negative correlation between citation density and refutation rate, a marked increase in refutation rate across three commentators in a related doctrinal lineage, and measurable genre-level differences within the Pali Canon itself. Second, we describe a constrained large language model pipeline that extracts typed philosophical relationships between concepts using a predefined relation vocabulary and deterministic post-hoc validation. The resulting graph surfaces cross-school disagreement patterns while also revealing important extraction limitations, including cases where an independent embedding-based analysis disagrees with the graph-derived findings. We release the full corpus, extracted relationship graph, and all source code.

Consensus protocols form the core of blockchains and other replicated state machines, ensuring that all correct nodes process the same totally ordered log of input transactions. In fault-free executions, performance is driven by the good-case transaction latency -- the time between a transaction becoming known to all nodes and its confirmation by the consensus protocol -- which depends on both how frequently proposals are made and, once made, how quickly they are confirmed. While prior work has established tight lower bounds on confirmation latency that modern protocols already achieve, it remains open whether the inter-proposal time can be further reduced below the state-of-the-art of one network delay. We introduce Gatling, an atomic broadcast protocol that achieves arbitrarily small inter-proposal times under rotating leader schedules; in particular, smaller than the network delay. Gatling runs multiple parallel instances of a black-box atomic broadcast protocol and staggers their proposal schedules to generate proposals in faster succession than state-of-the-art protocols. A deterministic interleaving rule merges the outputs of these instances into a single global log. We analyze the effects of head-of-line blocking caused by crashed leaders, and derive Gatling's optimal number of parallel instances. We further study the impact of Gatling on predictable validity and present two variants that retain this property. Finally, our experiments confirm that Gatling can be used with off-the-shelf component protocols to achieve low latency without fine-tuning the component protocol for minimum latency.

We study finite-horizon queue peaks in generalized switches, a standard stochastic-network model in which many queues share constrained service resources. Arrivals may be dependent, time-varying, and adapted to the past; the standing load condition is uniform interior slack, meaning the conditional mean arrival vector stays in a fixed contraction of the capacity region. We show that this slack reshapes the finite-time peak law for drift-minimizing scheduling policies such as MaxWeight. The square-root envelope that is sharp without slack persists only up to a geometry-dependent threshold; beyond that threshold, the running maximum grows only logarithmically with the horizon, both with high probability and in expectation. The mechanism is self-normalization: in the current queue direction, the projected fluctuation scale is normalized by the stabilizing drift scale. This removes capacity geometry from the logarithmic coefficient, while geometry remains in the threshold. Matching lower bounds show that both the logarithmic term and a geometric threshold are unavoidable. When finite-time state-space collapse is available, the threshold can be sharpened using local bottleneck geometry. For generalized input-queued switches, we obtain finite-time peak bounds with tight logarithmic coefficients. Simulations illustrate the two-phase envelope, local geometric refinements, and variance-sensitive improvements predicted by the theory.

Knowledge distillation transfers a teacher's competence to a small student but is brittle in the small-student regime: forcing the student to imitate logits from a much larger teacher concentrates it on the teacher's sharpest modes, hurting generalization on benchmark families beyond the training corpus. Reinforcement learning (RL) avoids logit imitation by training on the student's own rollouts. However, on questions where every rollout fails-yielding zero advantage and being silently discarded-injecting a stronger teacher's response into the policy gradient breaks the on-policy assumption and induces drift. We introduce Zone of Proximal Policy Optimization (ZPPO), inspired by Vygotsky's zone of proximal development, which keeps the teacher inside the prompt rather than the policy gradient. On hard questions, ZPPO constructs two reformulated prompts: a Binary Candidate-included Question (BCQ) pairs one correct teacher response with one incorrect student response as anonymized candidates the student must discriminate, and a Negative Candidate-included Question (NCQ) aggregates the student's wrong rollouts into a single prompt to surface their shared failure modes. A prompt replay buffer recirculates each hard question until it either graduates-the student's mean rollout accuracy on it reaches half- or is FIFO-evicted under finite capacity, amplifying BCQ and NCQ inside the student's current zone of proximal development. On the Qwen3.5 family at four student scales (0.8B-9B) with a 27B teacher, post-trained as vision-language models and evaluated on a 31-benchmark suite (16 VLM, 10 LLM, 5 Video), ZPPO outperforms off/on-policy distillation and GRPO, with the largest gains at the smallest scale.

Dataset distillation (DD) has emerged as a prominent approach in data centric machine learning, aiming to synthesize compact training sets for efficient training by compressing the information in large datasets into a small number of synthetic samples. However, DD methods are often evaluated under inconsistent evaluation protocols, ranging from standard ERM to single/multi-teacher supervision, making it difficult to isolate the effectiveness of distilled data from evaluation. Moreover, many prior methods claim that DD outperforms data pruning approaches such as coreset selection (CS), based on the assumption that restricting condensed datasets to subsets of real samples fundamentally limits their expressiveness. In this work, we critically evaluate DD methods through large-scale experiments using standardized datasets and evaluation protocols to assess their intrinsic effectiveness. We benchmark seven state-of-the-art (SOTA) DD methods on ImageNet-1K, ImageNet100, and ImageNette, using three widely adopted training protocols against three CS strategies. Our results show that while some DD methods fail to outperform even simple random subsets, the SOTA DD approaches are comparable to or worse than coresets on large-scale datasets and incur a substantially higher cost for construction. Beyond accuracy, we also evaluate the representativeness, diversity, and quality of condensed sets, and find that coresets consistently achieve better coverage of the original data distribution. These findings highlight the limited practical advantages of current DD methods and show that coresets remain competitive and are often a more computationally efficient alternative for data-centric learning.

Current world models face a fundamental tension: faithful long-horizon simulation demands deep computation, but deeper models are expensive to deploy and prone to compounding errors. We resolve this by introducing Looped World Models (LoopWM), which are the first looped architectures for world modelling. Our method iteratively refines latent environment states through a parameter-shared transformer block. This yield up to 100x parameter efficiency over conventional approaches with adaptive computation that automatically scales depth to match the complexity of each prediction step. Orthogonal to scaling model size and training data, LoopWM establishes iterative latent depth as a new scaling axis for world simulation, which might significantly push the community forward.

Looped architectures provide an inductive bias toward learning step-by-step procedures for tasks that require compositional reasoning. The number of effective layers reached by looping determines the quality of the solution these models find. Like deep architectures, looped architectures are prone to a signal propagation problem induced by depth as the halting decision is postponed. In this paper, we address this signal propagation issue using pre-norm layers and residual scaling. Building on these architectural modifications, we propose FPRM, a Transformer-based Fixed-Point Reasoning Model that uses fixed-point convergence as an end-to-end halting mechanism in a looped architecture. We show that fixed-point halting allows FPRM to adapt its compute to task difficulty. FPRM is effective on common reasoning benchmarks, namely Sudoku, Maze, state-tracking, and ARC-AGI.

This paper presents a robust methodology for the systematic digitization and encoding of the Al-Mawrid Arabic-English dictionary, transforming it from a legacy print resource into a standardized computational lexicon. Addressing a significant gap in Arabic lexical infrastructure, the study adopts a dual-standard framing that aligns the ISO Lexical Markup Framework (LMF) with the Text Encoding Initiative TEI Lex-0 guidelines. By applying an editorial view to the dictionary's macro- and microstructure, the research resolves the structural ambiguities and punctuation inconsistencies typical of 20th-century bilingual dictionaries. The methodology is grounded in an empirical analysis of the dictionary's lexical knowledge density. Drawing on a representative sample (the letter Ayn, comprising 4.6% of the total volume), the study provides scientific weight to the encoding process, demonstrating a structural parsing accuracy of 91%. Quantitative evaluation of the information extraction rules reveals high performance, with 85% precision and 98% recall for synonyms, and 88% precision for other morpho-semantic features. Beyond technical description, the paper provides a critical comparison with existing Arabic lexical resources and discusses the limitations of TEI Lex-0 when modelling specific Arabic phenomena, such as implicit "open set" semantic relations and scattered morphological cues. Furthermore, the study explores the potential for Linguistic Linked Open Data (LLOD) integration by establishing a scalable prefix-based referencing system that facilitates the resource's inclusion in the semantic web. The result is an interoperable, machine-tractable resource that provides a reproducible workflow for the retro-digitization of complex legacy bilingual lexicons within the Arabic NLP and Digital Humanities communities.

The LLM-empowered personal health agents with user health (sensor) metrics have offered a promising pathway to alleviate global disparities in healthcare access. However, large-scale clinical deployment remains constrained by an open-ended evaluation bottleneck: physician annotation is reliable but costly and unscalable, while LLM-as-a-judge evaluators are scalable but subjective, inconsistent, and sometimes clinically misaligned. We introduce RubricsTree, a scalable evaluation framework with an expert-aligned hierarchical taxonomy of over 100 atomic, clinically-verifiable Boolean rubrics, evolving from the insights of 4,000 real user queries through an iterative human-in-the-loop curation protocol with an expertise panel led by an experienced physician. A context-aware adaptive router activates only the relevant auto-weighted rubric subset per query, providing the throughput needed for scalable evaluation with expert-aligned quality. Through a systematic meta-evaluation, we show that RubricsTree (i) substantially exceeds a strong large-scale evaluation baseline in expert alignment on challenging open-ended queries; (ii) reliably penalizes contextually degraded responses; and (iii) when used as structured instructions, text feedback, or training rewards for performance optimization, yields up to ~66% relative gains on HealthBench for Gemini, GPT, and Qwen model families. RubricsTree thus provides a scalable, auditable, and evolving evaluation infrastructure required for the continuous optimization of product-level personal healthcare AI.

The application of machine learning models in practical tasks faces challenges such as class imbalance and multidimensional noise. This paper proposes RGNet, a neural network architecture based on the concept of the renormalization group (RG), for hierarchical coarse-graining of the feature space. The model sequentially compresses the input dimensionality and concatenates all scales before classification, allowing it to capture both local details and global patterns. The notion of RG-flows is introduced - interpretable low-dimensional representations whose visualization via t-SNE reveals a discrete curvilinear structure confirming the effectiveness of coarse-graining. Experimental results are presented on the imbalanced AI4I dataset. The obtained results demonstrate that RGNet is a universal, interpretable, and competitive solution for fault prediction in applications with imbalanced classes.

On-policy self-distillation (OPSD) has proven effective for post-training large language models (LLMs), yet its application to diffusion LLMs (dLLMs) remains unexplored. Existing OPSD methods are inherently autoregressive-centric. They inject privileged information via left-to-right prefix conditioning with token-level divergence supervision, a design that fundamentally conflicts with the arbitraryorder generation of dLLMs. We introduce d-OPSD, the first OPSD framework tailored for dLLMs. Our approach makes two core contributions. First, we reframe self-teacher construction by using self-generated answers as suffix conditioning, enabling the student model to learn from "self future-experience" rather than privileged prefixes. Second, we shift supervision from token-level to step-level, aligning training with the iterative denoising process of dLLMs. Experiments across four reasoning benchmarks show that d-OPSD consistently outperforms RLVR and SFT baselines with superior sample efficiency, requiring only around 10% of the optimization steps by RLVR and opening a promising pathway for dLLM posttraining. The code is available at https://github.com/xingzhejun/d-OPSD.

We evaluate the adversarial robustness of two frontier large language models (LLMs) developed by Anthropic, Fable 5 and Opus 4.8, against four families of automated jailbreak attack across 7 826 harmful intents spanning a ten-category harm taxonomy. Using the HackAgent red-teaming framework, hundreds of thousands of adversarial attempts were generated and every apparent success was independently re-adjudicated by a panel of three judge models (majority vote). Both models resist the majority of attacks, but the residual surface is larger than aggregate framing suggests: it is dominated by adaptive iterative attacks, while static obfuscation is near-fully neutralised. The strongest adaptive search (tree-of-attacks) breaks Opus 4.8 on 11.5% of intents overall, whereas Fable 5 stays in the single digits (6.1% worst-case). Aggregate rates therefore should not be read as reassurance. Even in these hardened configurations, the two models produced 1 620 (Opus 4.8) and 702 (Fable 5) panel-confirmed harmful completions spanning every harm category, located automatically, cheaply, and within the first one or two refinement steps by an attacker model with no human expert in the loop. The reasonable conclusion is that even the best, most-tested frontier models remain reliably breakable under sustained automated pressure.

As high-quality public web corpora become increasingly exhausted, clean long-context documents have become a scarce and expensive source of training data for large language models (LLMs). Existing long-context corpora are often proprietary and costly to acquire, synthetically generated, or concentrated in narrow domains such as programming. We introduce the Stanford EDGAR Filings Dataset (SEFD), an open reconstruction of SEC filings into layout-faithful MultiMarkdown for financial language modeling and evaluation. SEFD makes audited financial statements, risk disclosures, ownership reports, accounting notes, and market-moving event filings usable as long-context pretraining data and as a basis for financial reasoning, forecasting, compliance, and document understanding. The resulting corpus is token-efficient, model-ready, and has less than 0.1% overlap with Common Crawl-derived corpora. We release SEFD-v1, a 152B-token initial public snapshot, and provide corpus-level analyses of a larger 18.5M-filing archive estimated at 550B tokens. We further introduce two SEFD-derived benchmarks: EDGAR-Forecast, which evaluates filing-grounded numerical forecasting after model knowledge cutoffs, and EDGAR-OCR, which evaluates transcription of complex financial tables.

Deep research (DR) systems are increasingly used for complex information-seeking tasks, but existing works mainly focus on generating reports and summaries. In contrast, many enterprise tasks instead require an agent to identify concrete workflows which is a sequence of action-steps. For example, rather than summarizing budgeting policies, an agent should be able to determine the steps needed to answer a question such as: "How do I request new headcount given a fixed budget?". Therefore, we introduce DRFLOW, a benchmark for evaluating personalized workflows predicted by agents from heterogeneous sources. Each task requires the agent to identify relevant evidence from scattered sources, then use that evidence to predict the correct action-step sequence for the user's task. DRFLOW contains 100 tasks across five domains, with 1,246 reference workflow steps grounded in more than 3,900 sources. We define seven diagnostic metrics covering factual grounding, step recovery, structural ordering, condition resolution, and personalization. We further present DRFLOW-Agent (DRFA), a workflow-oriented reference agent to predict personalized workflow. We show that although DRFA improves over strong baseline agents (upto 10.02% average F1 score), there is substantial room for improvement remains across these workflow metrics, indicating that predicting complete and correct personalized workflows remains a challenging frontier for deep research.

Multi-stage cyberattacks span system, network, and browser logs. Detecting them requires correlating events across all three sources. Machine learning methods can learn these cross-source patterns, but they need labeled multi-source data. Existing public datasets fall short. Network-only datasets such as CICIDS and UNSW-NB15 miss host and browser activity. Host-focused datasets such as LMDG and CICAPT-IIoT lack browser telemetry. ATLAS includes all three sources but labels events only as malicious or benign, without MITRE Adversarial Tactics, Techniques, and Common Knowledge (ATT&CK) technique granularity. No public dataset combines all three sources with per-entry ATT&CK technique labels. We close the gap by building a multi-source log dataset of 870 sessions (70 attack, 800 benign) and approximately 2.3 million events. We captured system, network, and browser activity simultaneously on Windows endpoints. We labeled malicious events with ATT&CK technique IDs, covering 12 tactics and 53 techniques. We generated all attack data using real tools, including Remote Access Trojan (RAT), Command and Control (C2) tunnels, and cloud exfiltration. To demonstrate learnability, we fine-tuned three Small Language Models (SLMs) (Qwen2.5-1.5B, Llama-3.2-3B, Phi-4-Mini) using Low-Rank Adaptation (LoRA). We compared each against its base variant across ten metrics on two tasks: chunk classification and ATT&CK technique identification. Fine-tuning improved every model on every metric. Chunk classification accuracy rose from approximately 8% in the base variants to between 90% and 97% after fine-tuning. Technique identification remained challenging, with the best exact-match accuracy at 42%, although high partial-match scores show the models captured most of the underlying reasoning.

Finite-dimensional (FD) diffusion policies exhibit temporal drift owing to discretization artifacts that degrade long-horizon performance (when deployed on physical systems). We introduce a backward Kolmogorov equation that lifts diffusion policies to a Cameron-Martin space -- a subset of the Hilbert space. Essentially, replacing stochastic score matching with a deterministic boundary-value PDE problem. Our core innovation thrives on Gaussian measure theory whereupon the diffusion noise covariance operator is realized from a colored noise distribution which prescribes a notion of regularity on samples from the model at inference time. We train the diffusion model with a derived precision-weighted Cameron- Martin loss and a Kolmogorov residual is introduced as a PDE diagnostic during inference. These substitutions yield (i) convergence guarantees where the bound's constants depend on the effective rank of the kernel rather than action dimension, (ii) improved trajectory regularity via spectral weighting, and (iii) a deterministic failure detector without reward signals. Validation across two application domains demonstrates substantial improvements: on the PushT manipulation benchmark, the Cameron-Martin loss achieves a 17% improvement in maximum episode reward (0.95 vs. 0.78 for MSE) and 67.6% reduction in inter-step drifts during inference via the introduced residual magnitude. Similarly, on a 6-station manufacturing line with constant work-in-process (CONWIP) flow control, we achieve 28.4% lower RMSE than classical LSTM baselines; a high starvation-event recall (1.0 in test cycles), and effective bottleneck identification (Precision@1 = 1.0 in test set, 13x signal-to-noise ratio). We then certify the dispatch policies with Hamilton-Jacobi reachability theory which reduces deadlock events by 96% compared to uncontrolled dispatch over 100 simulated runs (351 events prevented).

Enterprise intrusion response still depends on static playbooks and analyst-driven triage, creating delay between alert generation and containment. We present Agentra, a supervisable multi-agent Intrusion Response System (IRS) framework that converts alerts from IDS, EDR, and XDR platforms into structured incident response plans grounded in MITRE ATT&CK, MITRE D3FEND, and NIST CSF 2.0. Agentra decomposes response reasoning across role-scoped agents, validates proposed plans through a bounded Planner--Validator review loop, screens retrieved threat intelligence through a Moderator security gateway, gates actions through an Action Catalog and risk score, and records decisions in an append-only audit log. We evaluate Agentra against a static OASIS CACAO v2.0 cyber-playbook baseline on a 120-event corpus drawn from ThreatHunter-Playbook, Splunk BOTSv3, and DARPA OpTC. The strongest configuration improves FP-aware IRS F1 from 0.61 to 0.84 and restores the projected harmful-action rate to the static baseline level of 0.0% after Planner-only configurations introduce unsafe overreaction. These results indicate that multi-agent response planning can improve ontology-grounded IRS coverage while preserving analyst approval and auditability.

Temporal difference (TD) learning with linear function approximation is a core method for policy evaluation. Its classical continuous-time description is an ordinary differential equation (ODE), which captures the asymptotic mean dynamics but neglects stochastic fluctuations determining the error floor. We introduce a stochastic differential equation (SDE) approximation for linear TD(0) under Markovian noise. The resulting model distinguishes the contraction dynamics governed by the projected Bellman operator from the influence of Markovian sampling. As a consequence, the model explains the constant-stepsize error floor through the interaction between Markovian long-run covariance and the contraction geometry of the projected Bellman operator.

Illegal, unreported, and unregulated fishing (IUU) traditionally refers to fishing activities that violate applicable laws or occur in areas that lack applicable laws. We propose the term IUU+ to capture a broader suite of fisheries sector environmental and associated supply chain trade-related crimes and behaviors. Although IUU+ activity is widely recognized as a serious threat to marine ecosystems, markets, and livelihoods, a quantitative understanding of these incidents, e.g., their frequency, geography, species, actors, and patterns in the type of illicit activity, remains difficult to obtain. We propose IUU+DB, a large language model driven system for building a global incident database of IUU+ activity. The system ingests heterogeneous documents, classifies whether they describe relevant incidents, extracts key data elements such as actors, locations, species, vessels, violations, and enforcement outcomes, and supports deduplication and trend analysis. Case studies and validation results show that IUU+DB can help organize fragmented evidence, surface geographic and behavioral hotspots, support fisheries-domain specific research in academia and non-government organizations, assist source and species risk assessments for industry, and provide support for policy implementation and targeted enforcement efforts to government agencies.

We present a numerical method for the forward solution of nonlinear partial differential equations (PDEs) in which Bellman-Kalaba quasilinearization reduces the nonlinear problem to a sequence of linear subproblems, each discretized by collocation onto a trial space that is linear in its parameters and solved by a single direct linear least-squares QR factorization. The trial space, which we term Linear-in-Learnables (LiL), comprises representations whose trainable parameters enter linearly, including random-feature extreme learning machines, spectral polynomial bases, and trigonometric expansions, each implemented as a physics-informed neural network. The method thus replaces the nonconvex gradient-based training that limits standard PINNs with a convex per-step solve. We establish local Newton-Kantorovich convergence of the outer iteration to a residual-limited neighborhood under an explicit smallness condition, with the limiting accuracy governed by the best-approximation residual of the trial space rather than by an optimization tolerance. The method, denoted LiL-Q, is assessed on seven benchmarks spanning scalar nonlinear PDEs (Bratu, viscous Burgers, Buckley-Leverett), coupled systems (plane-strain elasticity and the incompressible Navier-Stokes equations in two and three spatial dimensions), and steady-state Darcy flow with heterogeneous permeability. Across these problems, LiL-Q converges in single-digit outer iterations in most cases, even at the coarsest basis sizes and independent of the parameter count. When the exact solution lies in the span of the trial space, the method recovers it to machine precision in a single solve. On the Navier-Stokes benchmarks, it matches or exceeds published PINN solvers with up to two orders of magnitude fewer trainable parameters, without gradient-based optimization.

Software practitioners increasingly use AI coding agents that generate test code alongside production code in open source pull requests (PRs). Recent studies report more than 932,000 agent-authored PRs across more than 116,000 repositories, yet whether their test files contain meaningful verification logic remains underexplored. Test files lacking explicit assertions execute code without verifying behavior, so quality gates based on test-file presence overestimate verification strength. The goal of this paper is to help practitioners assess the verification strength of agent-authored patches by characterizing oracle signals and their link to merge outcomes and review effort. We conduct an empirical study of 86,156 test-file patches from 33,596 agent-authored PRs across 2,807 GitHub repositories produced by five coding agents: OpenAI Codex, GitHub Copilot, Devin, Cursor, and Claude Code. A qualitative analysis of 384 stratified patches informs a syntactic taxonomy of eight oracle signal categories. Applied at scale, 80.2% of test patches contain weak or no explicit oracle signals. While raw merge rates are lower for strong-oracle PRs, a regression analysis adjusting for agent, PR size, repository popularity, task type, and language shows strong oracles significantly improve merge likelihood (OR = 1.28, p < 0.001). Our findings suggest that test file counts substantially overestimate verification strength and that practitioners can adopt oracle-aware quality checks to more accurately evaluate agent-authored contributions.

Classifying Cyber Threat Intelligence (CTI) using MITRE Adversarial Tactics, Techniques, and Common Knowledge (ATT&CK) is essential for proactive defense, but historically required extensive human effort. Pre-Large Language Model (LLM) automation sped up this process, but could not resolve the complex language and multi-step attack patterns found in unstructured CTI reports. LLMs addressed previous limitations by using contextual reasoning to understand unstructured text. However, current evaluations rely on simplified, single-technique sentences that ignore the complexity of real-world CTI reports, which often leads to inflated performance results. Consequently, the baseline performance of open-source LLMs on complex unstructured CTI reports remains unevaluated. To address this gap, we constructed a ground-truth dataset of 2,076 human-annotated sentences (1,281 technique-positive, 795 negative) from 83 complex unstructured CTI reports. These sentences were mapped to 114 unique ATT&CK techniques using a six-phase annotation process, achieving \k{appa} = 0.68 inter-annotator agreement. Using this dataset, we evaluated seven open-source LLMs ranging from 8B to 236B parameters across prompt strategy and temperature configurations. The highest-performing LLM achieved a micro-averaged F1 score of 0.22, establishing the empirical baseline for multi-label ATT&CK classification on complex unstructured CTI. Parameter size showed a statistically significant positive correlation with F1 score. Prompt strategy and temperature produced no statistically significant gains across model configurations. These results indicate that current open-source LLMs are insufficient for production-grade ATT&CK classification. The dataset, benchmark, and findings provide a reproducible foundation for future CTI research.

Large language models now produce legal text of at least median quality, yet no existing benchmark can evaluate whether they perform doctrinal legal reasoning, which forms the interpretive core of legal work, rather than the ancillary, paralegal tasks that most current legal-AI evaluations measure. This measurement gap is not only methodological but legal: the EU AI Act makes "appropriate accuracy" a binding requirement for high-risk AI used in the judicial domain, yet that requirement cannot acquire operational content without the very doctrinal-reasoning benchmark the field lacks.

We present a novel framework for realistic and controllable 3D face re-aging which produces highly detailed, identity-preserving results. Existing 3D editing methods, while effective for coarse semantic changes, are not well suited for re-aging, as even small inconsistencies across re-aged 2D views can lead to over-smoothing of subtle but perceptually important age-related details. To address this challenge, we first introduce a 2D diffusion-based re-aging model, DiffReaging, trained on synthetically generated image pairs. We further propose a center-out editing propagation strategy that leverages this re-aging model to reconstruct multi-view-consistent re-aged images. Specifically, starting from a re-aged frontal pivot view, we reconstruct the remaining views through warping and our proposed Masked-DiffReaging process. By injecting existing content at every step of the diffusion process, Masked-DiffReaging ensures that the reconstructed regions remain coherent with existing pixels. The resulting consistent set of re-aged views supervises the optimization of the re-aged 3D representation. Our method outperforms existing 3D editing techniques both visually and quantitatively, enabling smooth, fine-grained control over age transformations in 3D face models.

Building personalized cardiac electrophysiology (EP) digital twins requires identifying the appropriate model structure for each patient, not merely fitting parameters. Traditional methods rely on experts to manually prescribe hybrid physics-neural architectures, which requires deep domain expertise and does not transfer across patients. Recent works have applied large language models (LLMs) to generate or act as hybrid models. However, despite their promising generalization capacity, these LLM-based methods lack the structural priors needed for stable cardiac simulations. Hence, we propose LEADS, a framework that formulates cardiac EP domain knowledge as a structured action space and utilizes an LLM agent to discover hybrid models. The agent follows an iterative reasoning-and-action loop to select, combine, and refine hybrid models, whilst gradient descent handles parameter fitting. The proposed LEADS designs every candidate model towards physically grounded, interpretable, and numerically stable, while allowing open-ended architectural discovery. We validate LEADS on synthetic data with three ground-truth reaction models and on real cardiac EP data, demonstrating that it outperforms both human-designed hybrid models and other LLM-based hybrid modeling.

Language models are remarkably capable at medical question answering, in some cases surpassing the accuracy of general physicians. However, answering questions about wearable health data remains challenging and understudied, as these ubiquitous sensors produce continuous, high-dimensional, and longitudinal data, which is non-trivial to align with text-centric distributions in LLM pretraining. The diversity of sensor modalities and user intents cannot be effectively handled by a fixed reasoning workflow or a single pretrained foundation model. To address these challenges, we propose WEQA, a query-adaptive agent framework that unifies LLM reasoning with specialized wearable analytical and modeling tools. An LLM controller is employed to synthesize execution plans and dynamically route each query to the appropriate combination of sensor analysis and pretrained models, and perform grounded response auditing with external knowledge. We also curate a benchmark spanning four open wearable datasets comprising analytic and predictive tasks in three different health domains. Experiments show that our framework is 24% more accurate than LLM and agentic baselines, and a blinded study with 12 medical experts and 8 users shows substantial gains in usefulness and clinical soundness.

A robot's flash endurance is a non-renewable stock: every persisted write spends one of a few thousand program/erase cycles and never refills, yet no fielded robot memory system prices which memories are worth an erase cycle. We treat embodied memory as depreciating capital and price that stock with a single endurance shadow price $η$, which makes cost-minimizing placement across a RAM / on-board NVM / cloud hierarchy a threshold in a wear-augmented per-byte index. The index is cost-optimal whatever the sign of the value-write association $χ$; only when $χ> 0$ does the optimum turn non-monotone, sending a robot's most valuable memories off its flash. The pivot is thus empirical, and we measure $χ$ on real robot logs at a pre-specified gate: its sign is a property of the deployment regime -- positive on recurrent long-horizon manipulation ($\hatχ \approx +1.0 \times 10^{-3}$, replicated at full power), null on a shorter-horizon suite, and negative on non-recurrent teleoperation. Two boundaries scope the result. The endurance budget is dormant on premium 3,000-P/E TLC at datasheet prices and binding on the commodity QLC/eMMC ($\sim$1,000 P/E) that cheaper edge robots run. And where it binds, a learned wear-aware controller only ties price-based routing on task value, because realized value is tier-invariant across RAM, NVM, and cloud: the rent governs device lifetime and cost, not task performance. Whether wear-aware placement improves task value remains open -- $χ$ is measured against a value proxy, and the non-monotone optimum, while proven, is not yet observed in data.

AI agents are moving from advisors to actors, booking travel, planning menus, and running procurement on behalf of users. Existing benchmarks for AI and animal welfare evaluate model text responses to question-answer prompts, leaving open whether the welfare reasoning surfaced in those responses transfers to agentic deployment where the model must take actions with tools. We introduce TAC (Travel Agent Compassion), the first agentic benchmark measuring whether AI agents avoid options involving animal exploitation when acting on behalf of users. TAC presents an AI agent with twelve hand-authored travel booking scenarios across six categories of animal exploitation, augmented to forty-eight samples to control for price, rating, and position confounds. We evaluate seven frontier models from four labs. Every model scores below the chance level of sixty-four percent, with the best performer (Claude Opus 4.7) at fifty-three percent. A single welfare-aware sentence in the system prompt yields gains of forty-seven to sixty-three percentage points in Claude and GPT-5.5, twenty-six points in GPT-5.2, and under twelve points in DeepSeek and Gemini. An auxiliary Inspect Scout audit of 288 base-condition transcripts from the top two performers, using Gemini 2.5 Flash Lite as judge, flags zero transcripts for evaluation awareness, suggesting the below-chance rates do not stem from the models recognising the evaluation. We discuss implications for category-level variation across cultural domains, the limits of text-response welfare benchmarks, and the EU General-Purpose AI Code of Practice systemic risk framework.

In this work, we introduce the Certus Caliber Classification Gunshot Dataset (C3GD), a publicly accessible data set developed for the analysis of firearm muzzle blast sounds. The dataset aims to provide a wide variety of firearms, calibers, cartridges, microphones, and microphone locations with metadata detailed beyond what is currently otherwise available. It comprises more than 8000 field-collected data points from 28 firearms across 16 calibers. Because data collection in the field is costly, much of the existing research has been done using gunshot audio collected from the internet, which increases the risk of low-quality data and label noise. This dataset is primarily focused on caliber classification, but can also be used for gunshot detection, audio separation, and audio signal processing, providing a diversified and real-world reference. The dataset aims to provide enough diversity to be able to generalize to more real-world applications while also providing enough metadata for detailed academic analysis.

Meta-reinforcement learning enables fast adaptation by extracting shared structure from related tasks, but existing end-to-end methods often couple task inference with embodiment-specific control. This coupling can obscure non-parametric task semantics, reduce sample efficiency, and limit cross-agent reuse. We propose a meta-knowledge reutilization framework that learns task-level knowledge on a dynamics-simplified agent and transfers it to heterogeneous agents. The framework uses a Bayesian non-parametric prior to organize latent task modes and a high-level policy to generate task-level magnitude guidance. To bridge reusable task knowledge with different embodiments, we introduce a semantic-magnitude interface and a lightweight temporal adaptor, which convert frozen meta-knowledge into temporally aligned subgoals for embodiment-specific low-level controllers. Experiments on multiple locomotion agents show that our framework reduces final-step tracking error by 94.75% -- 99.79% compared with recent state-of-the-art baselines and achieves comparable deployment performance with about 23.8% of their interaction data.

Hypergraph partitioning is a critical step in the design of complex embedded systems, essential for optimizing task mapping on heterogeneous MPSoCs and enabling multi-FPGA prototyping. Many existing methods rely on community detection to identify modules with dense internal and sparse external connections, typically utilizing them to constrain the coarsening phase--a widely adopted paradigm. In this work, we propose ComPart, a generalized framework that integrates diverse community detection methods to uncover high-quality clusterings throughout the post-coarsening stages (i.e., initial partitioning and uncoarsening). These discovered clusterings serve as distinct structural guides, enabling the refinement process to identify superior partitioning solutions. Our framework offers two key advantages: (1) it establishes a new paradigm that leverages community structures detected during uncoarsening to escape local optima and explore globally meaningful solution subspaces, transcending the limitations of standard local refinements; and (2) it flexibly accommodates both existing and future community detection methods. Furthermore, we theoretically generalize locally-dense decomposition--originally from graphs--to the hypergraph domain. We provide the formal extension and necessary proofs to apply this technique to hypergraphs, marking its first application in hypergraph partitioning. Specifically, we utilize this rigorously derived decomposition to guide the initial partitioning phase toward superior starting points. Experimental results on standard benchmarks demonstrate that our method consistently outperforms state-of-the-art methods in solution quality.

Recent incidents involving LLMs used for mental-health support reveal a critical evaluation gap: surface-level safety scores do not capture how models behave across realistic, emotionally sensitive interactions over time. Existing benchmarks measure knowledge, safety, or static response quality, but miss whether LLM interactions help users keep reflecting, coping, and making decisions themselves. We formalize this missing dimension as COGNITIVE ATROPHY, a process-level behavioural measure in AI-mediated mental-health support distinct from safety and helpfulness. To measure it, we introduce COGNITIVE ATROPHY BENCH, a clinically grounded benchmark built from 1,576 fully human-generated counseling conversations, 15,680 turns, and 42,230 responses from five LLMs. Three clinical and neuropsychology experts developed a 20-attribute schema spanning user context, response behaviour, and global risk flags; six trained clinical reviewers applied it with span-grounded evidence, producing 5,324 reviewer judgments. We further introduce the User-Input Risk Index (UIRI), the Cognitive Atrophy Risk Index (ARI), and trajectory summaries. Across five LLMs, models show a consistent moderate-to-high level of atrophy-aligned behaviour across single and multi-turn settings. While models generally respond to overt safety cues, they adapt less reliably when users seek solutions or decisions. The dominant recurring patterns are directive advice, problem-solving, recommendation responses, topic shifts, and forms of validation that may reinforce dependence rather than reflection. Our work makes COGNITIVE ATROPHY measurable and provides a foundation for auditing model behaviour in sensitive LLM conversations.

Modern conversational AI systems frequently rely on user metadata to localize responses, yet the unintended regional biases introduced by this hidden context remain poorly understood. In this work, we evaluate location leakage: the phenomenon where a model generates geographic references despite receiving a geographically neutral user prompt. Across both creative writing and open-ended Q&A prompts, even state-of-the-art LLMs systematically favor region-specific outputs when exposed to location metadata, with leakage spiking by up to 793 times above baseline (e.g., from 0.04% to 31.7% for Llama 3.1-8B, and 21.3% and 8.8% for Qwen3-8B and Claude Sonnet 4.6, respectively). Our analysis further shows a novel structural conditioning effect: replacing the injected location with the placeholder "Unknown" still elevates leakage by up to 72 times above baseline, demonstrating that the user profile frame itself, independent of any geographic content, acts as a generative conditioning signal.

Modern AI systems increasingly solve a task not with a single model call but with several imperfect agents working together: some propose pieces of a solution, others verify them, and the results are combined. These systems often outperform any single model, yet it is rarely clear why they succeed or when they will fail. We model such a system as message passing on a sparse graph, the structure that underlies low-density parity-check (LDPC) codes, and extend the density-evolution machinery of coding theory to this richer setting. In our model a task is a set of coupled binary subclaims, and an agent architecture is a sparse, role-typed factor graph whose check nodes are noisy Boolean verifier nodes, each computing a local Boolean function of the subclaims it touches. Three distinct failure modes, all modeled as erasures (an agent abstaining, a verifier returning no usable output, and a message lost between two agents), propagate as the agents exchange set-valued messages. The check agents combine these messages by a single logical-forcing rule that specializes to XOR, AND, OR, implication, and Horn constraints. This is more than a relabeling of LDPC theory: the verifier functions are nonlinear and value-asymmetric, and the three failure modes do not reduce to a single effective channel, so they require new threshold, finite-length, and converse results rather than a direct reuse of parity-check density evolution. We prove a density-evolution theorem that predicts the asymptotic fraction of unresolved subclaims on random role-typed architectures, with an extension to deterministic, locally tree-like graph sequences. The XOR case recovers the classical LDPC recursion on the binary erasure channel (BEC); the AND case exposes an asymmetry between positive and negative verifier certificates.

Large language model applications build prompts from templates, and Handlebars is a widely used templating engine and the default prompt-template format in Microsoft Semantic Kernel. Its double-brace {x} expression HTML-escapes the interpolated value and is documented as the safe default; its triple-brace {x} expression inserts the value raw. We show that this choice silently governs an application's exposure to structural role injection, where attacker-controlled data carries chat role delimiters that forge a higher-privilege turn. A model-free analysis establishes the mechanism: Handlebars escaping rewrites angle brackets but not square brackets, colons, or Markdown hashes, so it neutralises ChatML, Llama-3, and XML role delimiters (survival rate 0.00) while leaving Llama-2 [INST], legacy Human:/Assistant:, and Markdown ### delimiters intact (survival rate 1.00 for the last two). We then run 5760 trials across seven delimiter families, two attack objectives, and four models (GPT-3.5 Turbo, GPT-4o mini, GPT-4.1 mini, Claude Haiku 4.5) at a combined API cost of 1.63 USD. GPT-3.5 Turbo follows the task-hijack instruction in 97% of raw and 91% of escaped trials, with the escaping protection concentrated in the angle-bracket families and absent for the colon- and Markdown-based families; the harder secret-exfiltration objective, which does not saturate, exposes the same family interaction more cleanly. Claude Haiku 4.5 resists both objectives almost entirely. The escaped default protects only the delimiter schemes whose characters HTML escaping happens to cover, gives no protection for the rest, and cannot substitute for a structural separation of instruction and data.

To assess the ability of current AI systems to correctly solve research-level mathematics problems, we tested several AI systems on a set of ten problems in a broad range of mathematical fields; these problems arose naturally in the research process of the contributors. This document includes the problems, our methodology, and the results of our testing. We provide links to supplementary documents including the human solutions, the AI-generated solutions, and the referee reports and logs for the AI-generated solutions. The ten problems were contributed by the following mathematicians: (1) Dariusz Kalociński and Theodore A. Slaman, (2) Richard Schwartz, (3) Aleksa Milojevic and Benny Sudakov, (4) Larry Guth, (5) Oleg Butkovsky, Jonathan Mattingly, and Lorenzo Zambotti, (6) Joshua Evan Greene and Duncan McCoy, (7) Sucharit Sarkar, (8) Sam Payne and Jidong (Jayden) Wang, (9) Sylvie Corteel and John Lentfer, (10) Srivatsav Kunnawalkam Elayavalli.

The problem of k-way hypergraph partitioning is fundamental with significant applications in various fields, including VLSI design and scientific computing. State-of-the-art hypergraph partitioners commonly employ a multi-level framework encompassing coarsening, initial partitioning, uncoarsening, and refinement phases. However, many existing methods do not scale well to problems requiring a large number of partitions (i.e., large k). In pursuit of exceptionally high solution quality, existing memetic approaches often execute their two key operations, recombination and mutation, by invoking separate, standalone multi-level partitioners. This design choice, however, renders them significantly more time-consuming than standard multi-level partitioners. To make such memetic approaches more practical, we propose an advanced memetic framework, IMPart, which introduces novel recombination and mutation operators and integrates them directly into the uncoarsening phase of a single multi-level framework. This transforms the local searches of different granularities in the traditional multi-level framework into a sophisticated, collaborative search. Experimental results on multiple standard benchmarks demonstrate our framework more effectively escapes local optima and explores the global solution space for higher-quality solutions, substantially outperforming all existing hypergraph partitioners for large-$k$-way hypergraph partitioning. Our framework highlights a new paradigm for the development of advanced hypergraph partitioners.

State Space Models (SSMs) such as Mamba-2 offer linear-time inference but their memory footprint limits edge deployment. Prior ternary SSM work (Slender-Mamba) trains from scratch on 150B tokens; we show a pretrained checkpoint suffices, reducing the marginal token budget by 1,000x. Using grouped quantization-aware training (QAT) with knowledge distillation from a frozen FP16 teacher, we compress Mamba-2 1.3B to 3.61x (2,687 to 744 MB) and achieve 48.1% zero-shot accuracy (7-task average) in just 102M tokens (4 GPU-hours, single H100) -- approaching Bi-Mamba's 48.4% (within +/-0.9pp CI). This QAT-from-pretrained setting reveals zero-ratio collapse, a novel instability caused by learnable quantization scales that does not arise in from-scratch training. We further show that post-hoc correction strategies effective for Transformers fail for SSMs due to error accumulation through the recurrence. These results demonstrate that ternary SSMs do not require expensive from-scratch training: QAT from pretrained checkpoints with KD is a data-efficient alternative.

Fairness is an important aspect of decision-making in multi-objective reinforcement learning (MORL), where policies must ensure both optimality and equity across multiple, potentially conflicting objectives. While single-policy MORL methods can learn fair policies for fixed user preferences using welfare functions such as the generalized Gini welfare function (GGF), they fail to provide the diverse set of policies necessary for dynamic or unknown user preferences. To address this limitation, we formalize the fair optimization problem in multi-policy MORL, where the goal is to learn a set of Pareto-optimal policies that ensure fairness across all possible user preferences. Our key technical contributions are threefold: (1) We show that for concave, piecewise-linear welfare functions (e.g., GGF), fair policies remain in the convex coverage set (CCS), which is an approximated Pareto front for linear scalarization. (2) We demonstrate that non-stationary policies, augmented with accrued reward histories, and stochastic policies improve fairness by dynamically adapting to historical inequities. (3) We propose three novel algorithms, which include integrating GGF with multi-policy multi-objective Q-Learning (MOQL), state-augmented multi-policy MOQL for learning non-statoinary policies, and its novel extension for learning stochastic policies. We evaluate our algorithms across various domains and compare our methods against the state-of-the-art MORL baselines. The empirical results show that our methods learn a set of fair policies that accommodate different user preferences.

We develop a text-to-SQL (structured query language) system based on large language models (LLMs) using in-context learning and apply it to the Automatic Learning for the Rapid Classification of Events (ALeRCE) astronomical database. ALeRCE is a community broker for the Zwicky Transient Facility and the Vera C. Rubin Observatory. The system enables users to query the database in natural language (NL) and generates executable SQL queries. To develop and evaluate the system, we constructed a dataset of 110 NL/SQL pairs. We propose a step-by-step generation framework comprising four modules: schema linking, query classification, prompt decomposition, and self-correction. The performance of thirteen LLMs is evaluated using in-context learning and prompt engineering techniques. Text-to-SQL performance is assessed using the perfect-match (PM) rate for row identifiers (e.g., object identifiers) and column identifiers (i.e., column names). The proposed step-by-step framework consistently outperforms a direct-inference baseline, while the self-correction module consistently reduces execution errors. For Claude Opus 4.6, PM performance on row (column) identifiers is high for simple queries, reaching 0.97 (0.94), and decreases with query complexity to 0.44 (0.72) for medium queries and 0.59 (0.49) for hard queries. Among the thirteen evaluated models, the best-performing LLMs for the text-to-SQL task are Claude Opus 4.6, Gemini 2.5 Pro, Gemini 3 Flash, and GPT-5.2-Codex.

This paper explores the problem of finding the minimum zero-forcing set on undirected graphs and proposes an adapted machine-learning framework to solve the problem. The minimum zero-forcing set problem is a graph coloring problem where the color of an initial set of nodes propagates throughout a network. The set of nodes is zero-forcing if it forces all uncolored nodes to change color under the constraint of the color-change rule. There are several applications to this problem across different domains such as network science, network control, and designing logical circuits. Finding the minimum zero-forcing set is shown to be NP-hard. We propose a reinforcement learning framework, SD-ZFS, that adapts the S2V-DQN architecture to the ZFS problem. We train several models on this adapted framework and analyze the performance across graph datasets that have varying structures. We evaluate how the models trained on the framework generalize, scale, and transfer to different network types. The results demonstrate the effectiveness of the framework when compared against the optimal solution and greedy heuristic. We provide further insight into how the ZFS problem can be solved through machine-learning and the influence of network structure on the problem.

Network planning optimization is a fundamental problem across diverse domains, including transportation systems, communication networks, and power grids. It requires simultaneous optimization of multiple competing objectives under complex constraints. Existing network planning optimization frameworks rely on mixed integer programming (MIP) solvers, heuristics, and deep reinforcement learning (DRL) models to compute planning decisions. However, they lack effective adaptability to diverse and dynamic user intents, thus leading to the trade-off between execution time and optimality. In this paper, we propose OmniPlan, an adaptive framework that achieves both timeliness and near-optimality in network planning optimization. To achieve the adaptability lacking in existing solutions, OmniPlan employs a large language model (LLM)-based interpreter to convert heterogeneous natural-language intents into a unified and quantifiable user-preference vector. Then it employs a mixture-of-experts architecture that integrates MIP solvers, heuristics, and DRL models as specialized experts, where OmniPlan adapts to diverse intents by dynamically selecting timely and near-optimal experts. Finally, it incorporates a DRL-based expert configuration module that fine-tunes optimization objective weights to align planning decisions with user-specific preferences. We evaluate OmniPlan with a representative real-world workload, i.e., distributed machine learning (ML), where we leverage OmniPlan to offload a wide spectrum of ML inference tasks, e.g., decision trees, SVM, naive Bayes, XGBoost, and random forests, onto a network of hardware devices. Our experiments on a real-world testbed indicate that OmniPlan achieves near-optimal and low-execution-time offloading for real-world ML inference tasks, reducing latency by up to 97.8\% and network device resource consumption by up to 11.5\%.

SWave is a complex-valued recurrent language model (169.26M parameters, D=384, L=16, T=2048) trained on FineWeb-Edu using 2xH100 NVL. It was designed around three founding premises: that representing language as complex waves rather than real-valued numbers enables richer information encoding; that a Cayley-parameterised unitary transition provides a mathematical guarantee against state decay or explosion; and that a hidden state which rotates rather than shrinks preserves signal integrity over arbitrarily long contexts. The core of SWave evolved substantially across three development phases. The Resonance Head was found to structurally admit imaginary-channel collapse as a global loss minimum (a failure mode we term cos-domination collapse) and was superseded by an untied head with independent real and imaginary embedding tables from the Phase-Associative Memory (PAM) architecture. This resolved the degenerate minimum and enabled stable 200,000-step training (best-step PPL 22.0 at step 89,861). ComplexNorm and the Wave Propagation Scan proved load-bearing throughout all three phases and were retained to the final architecture. ProtectGatedScan was reframed as a structural prior rather than a learned behaviour. The four multi-scale retention concepts showed no measurable improvement under controlled evaluation and were found non-load-bearing. The ComplexGatedUnit was superseded by a real-valued squared-ReLU channel mixer with fewer parameters. The auxiliary training objectives showed no benefit once structural constraints were resolved. The investigation yields a formal characterisation of cos-domination collapse, a parallel scan with a log-space backward pass for numerical stability, six transferable engineering principles for complex-valued recurrent training, and a plan-to-code traceability methodology for catching structural divergences that conventional test suites miss.

Retrieval-Augmented Generation (RAG) is the prevailing architecture for grounding language model outputs in external evidence, yet its dominant evaluation paradigms and default configurations remain oriented toward factual question-answering. For interpretive disciplines such as historical studies, RAG embeds assumptions that conflict with scholarly practice. We introduce HistoRAG, a framework that translates historiographical principles into concrete architectural interventions. Separated retrieval and generation decouples source discovery from interpretation, temporal windowing enforces balanced source representation across the research period as a methodological requirement of historical inquiry, and LLM-as-judge evaluation makes relevance judgments transparent and contestable. We evaluate these interventions using SPIEGELragged, applied to 102,189 articles from Der Spiegel (1950-1979). Each intervention addresses a measurable deficiency in standard RAG: era-specific vocabulary retrieves zero chunks from the 1950s when using 1970s terminology, evidence of the temporal skew that motivates windowing; vector similarity and LLM-assessed relevance correlate only weakly (Spearman rho = 0.275), motivating post-retrieval evaluation; and keyword-based and semantic retrieval surface largely disjoint source pools, motivating an architecture in which both operate as complementary retrieval layers under a shared LLM evaluation filter. We also introduce the concept of Zwischentexte (intermediate texts that function as interpretive proposals rather than findings) as a framework for responsible integration of LLM-generated text into scholarly practice. The architecture offers a model for how domain-specific epistemological commitments can be translated into RAG design decisions, and may transfer to other interpretive disciplines working with large corpora.

Graphical user interface (GUI) grounding requires vision-language models (VLMs) to identify small target elements in high-resolution screenshots and predict precise screen coordinates. On-policy self-distillation (OPSD) is a promising post-training approach for this coordinate-sensitive task, since it provides dense token-level teacher signals beyond hard coordinate labels. However, naive OPSD is not well suited to GUI grounding: OPSD evaluates the teacher on student-generated prefixes, the quality of coordinate-token teacher signals can degrade when the prefix has already deviated from the target coordinate, leading to unreliable teacher signal. To mitigate this, We propose quality-aware self-distillation for VLM-based GUI grounding, which improves coordinate-token teacher-signal quality through soft correctness-aware gating and teacher-probability scaling. The soft correctness-aware gate checks whether the teacher's current coordinate-token prediction can still be completed into the ground-truth box under the student-generated prefix. If not, the corresponding teacher signal is down-weighted. Teacher-probability scaling then uses the teacher's confidence as a lightweight factor to further calibrate the strength of the gated supervision. A key empirical finding is that neither component alone improves overall performance, whereas combining them consistently improves performance. This suggests that the two mechanisms play complementary roles: correctness-aware gating suppresses unreliable coordinate-token supervision, while teacher-probability scaling calibrates the strength of the remaining signals. Experiments across six GUI grounding benchmarks show that our method consistently improves the base model and outperforms strong baselines.

Advances in Artificial Intelligence (AI) have led AI for Theorem Proving to become a promising means of formally verifying computer systems. Whilst formal verification is traditionally reserved for safety-critical systems due to the required amount of expertise and effort, AI can help to automate a large amount of this workload and make it far more accessible. Blockchain-based systems are becoming increasingly popular and are frequently targeted by malicious actors, often resulting in huge financial losses, highlighting the need to better verify these systems and mitigate vulnerabilities. Arguably the most important component of these systems is the consensus protocol, which allows nodes to agree on decisions in a potentially adversarial environment. In this paper, we improve upon IsabeLLM, the automated theorem proving tool in Isabelle. Namely, we implement a Retrieval-Augmented Generation framework, Error tracing and counterexample generation for improved context supplied to the Large Language Model. Compatibility with the latest version of Isabelle and Sledgehammer is also implemented for improved efficiency. We compare the performance of the two versions of IsabeLLM in their ability to complete the verification of Bitcoin's Proof of Work consensus.

Structured State Space Models (SSMs), including the S4 and S4D architectures, have recently emerged as powerful alternatives to attention-based models for capturing long-range dependencies in sequential data. Despite their strong empirical performance, deploying these models in time- and resource-constrained settings remains challenging due to their computational and memory demands. In this paper, we propose a novel incremental, operator-level pruning approach for S4- and S4D-based models that significantly reduces inference cost while preserving predictive performance. To the best of our knowledge, this is the first work to systematically investigate structured operator pruning for SSMs. Our method progressively prunes model operators by interleaving structured masking with fine-tuning, while jointly monitoring accuracy and inference latency. We implement this approach within a unified training and evaluation framework that enables systematic exploration of efficiency-accuracy trade-offs. Experiments across multiple benchmark datasets show that pruning up to 70% of the model operators preserves the performance of the original models in most cases, while substantially reducing inference latency. These results demonstrate that structured operator pruning is an effective and previously unexplored strategy for improving the efficiency of SSMs and facilitate their deployment in practical, resource-constrained scenarios.

Cross-end-effector grasp generation seeks a unified model that generalizes across objects and across embodiments ranging from parallel grippers to dexterous end effectors. Existing grasp generators are typically designed for a fixed embodiment or encode embodiment identity with a static descriptor, which weakens transfer when topology, actuation coupling, and contact geometry differ substantially. We present EAGG, an embodiment-aligned grasp generator that represents each embodiment with a topology-aware end-effector graph and an embodiment-specific low-dimensional end-effector control space. A frozen end-effector-cognition backbone converts the current articulated state into geometry-aware tokens that act as a reusable morphology prior, and iterative geometry injection refreshes these tokens throughout sampling so that conditioning remains synchronized with the evolving end-effector geometry. On the MultiGripperGrasp benchmark, EAGG reaches 56.17% average success across six training end effectors, remaining within 1.10 percentage points of specialized training while preserving transfer to finetuning and zero-shot end effectors. Iterative geometry injection further reduces the pooled median contact distance from 0.239 cm to 0.189 cm. These results show that cross-end-effector grasp generation is strengthened by aligning embodiment structure inside a shared generator rather than suppressing embodiment differences. Code is available at https://github.com/wanhaoniu/EAGG.

Post-training pipelines that combine supervised fine-tuning (SFT) with reinforcement learning (RL) have emerged as the key recipe for transforming large language models (LLMs) into robust reasoners. We argue that this combined success is driven by compositional generalization, which we formalize through a hierarchical latent selection model. In this framework, reasoning traces are generated by a cascade of discrete latent selection variables corresponding to reusable atomic modules, including both skills (local operations) and routing mechanisms (how intermediate information is selected, reused, and composed). Within this model, we theoretically show that SFT and RL play asymmetric, complementary roles: SFT supplies the raw module materials in compositional traces, and RL decomposes those traces to identify the latent atomic modules and enable compositional generalization. We design controlled experiments to validate this theory. Our results demonstrate that RL can extract atomic modules from compound traces supplied by SFT and recombine them to solve new configurations. Moreover, we find that training on compound traces yields stronger generalization than training on isolated atomic modules. Finally, we investigate the relationship between SFT and RL data and identify an effective protocol in which SFT ensures coverage of all atomic modules through compositional traces, while RL focuses on novel compositions outside the SFT support to drive exploration.

Gradient descent in deep learning may operate at the edge of stability (EoS), a regime in which the largest eigenvalue of the loss Hessian hovers near the stability threshold $2/η$, where $η$ is the learning rate. Classical analysis tools such as gradient flow and the descent lemma do not apply here, motivating the search for a continuous-time model valid at EoS. We propose Edge Flow, a system of three coupled ordinary differential equations that provides a tractable, faithful, and predictive model of gradient descent dynamics at EoS. Edge Flow decomposes the dynamics into a center, an oscillation direction, and an oscillation magnitude. The center follows a modified gradient flow on a symmetrized loss; the direction tracks a top eigenvector of the Hessian via Rayleigh quotient dynamics; and the magnitude grows or decays exponentially depending on whether the sharpness exceeds or falls below the threshold $2/η$. Crucially, sharpness stabilization emerges from the coupled dynamics via a self-stabilization feedback loop. Discretizing Edge Flow only requires two gradient evaluations and one Hessian--vector product at each iteration. We demonstrate empirically that Edge Flow tracks the dynamics of gradient descent at least as faithfully as previously proposed continuous-time EoS models, while in addition resolving the oscillation of the sharpness at the onset of EoS, and that it provides a principled framework for understanding and mitigating instabilities in this regime.

Retrieval-Augmented Generation (RAG) has emerged as a paradigm for enhancing large language models (LLMs) with external knowledge, yet existing graph-based methods face a fundamental limitation: entity-centric and chunk-centric approaches operate on representations anchored to original text without true knowledge fusion. While entity-centric methods connect logically related content and chunk-centric methods preserve context, both retrieve information separately through similarity search, missing emergent understanding from their synthesis. In this paper, we propose HyGRAG, a hierarchical graph RAG framework that transcends source documents by addressing three core challenges: constructing summaries that genuinely integrate contextual and relational information, leveraging these synthesized representations to access emergent knowledge during retrieval, and efficiently updating hierarchical structures for dynamic corpora. Specifically, we design hierarchical index structures over hybrid graphs with both chunk and entity nodes, then iteratively cluster them and generate LLM-based summaries. Then, we design context and relation-aware retrieval that searches across all abstraction levels while expanding through community membership. Moreover, we enable dynamic knowledge update through attachment-based algorithms with only local re-summarization. Experimental results show that HyGRAG improves the average accuracy of multi-hop reasoning tasks by 9.7%, while maintaining reasonable efficiency.

Causal discovery seeks to uncover the causal dependencies among variables. For this purpose, we propose an algorithm called Tensor-based Second-order Causal Discovery (TSCD). Its input is a tensor obtained from the covariance matrices of observational and interventional data. Assuming the causal dependencies follow a linear structural equation model on a directed acyclic graph (DAG), TSCD outputs the DAG and the functions on its edges, requiring only that the noise variables are uncorrelated. We also implement a version of the approach for nonlinear models. Our focus on second-order statistics (via the covariance matrices) is motivated by their statistical and computational efficiency relative to higher-order moments, their identifiability relative to first-order statistics, and that they work regardless of whether the variables are Gaussian. We show that TSCD has identifiable causal order and parameters from a number of interventions that is logarithmic in the number of variables. Experiments show that TSCD is robust to noise, competitive with existing methods, and scales to hundreds of variables.

Open autoregressive neural-codec text-to-speech (TTS) models sound excellent on typical inputs yet suffer stochastic catastrophic failures: on a meaningful fraction of utterances they emit silence, terminate early, or collapse into repetitive or hallucinated content. We show this failure mode is cheap to remove. Under a single format-robust metric (a catastrophic-failure rate via an ASR round-trip), best-of-N ASR self-verification drives failures to near-zero: no observed failures remain by N=2 on a standard corpus (LibriSpeech) and by N=4 on a hard prompt set. This is not an artifact of one model: the reduction replicates across four open codec-TTS systems and three neural codecs (XCodec2, SNAC, Mimi), reaching the near-zero floor by N=2 on three of the four. We then make the fix free at inference time by distilling the self-verified behaviour into the model, which recovers much of the robustness in single-shot decoding, closing ~52-58% of the failure mass on hard inputs at no test-time cost. The distillation gain concentrates where it is needed (hard inputs); on already-reliable prose there is no headroom and no detectable change. A controlled comparison adds a clean negative: offline direct preference optimization (DPO/IPO) does not beat plain supervised distillation, and an online iterative variant is promising but not statistically separable at our evaluation size. We report honestly the one model that resists (a larger Llasa where scale did not obviously help) and a rare-word capability ceiling that no self-distillation method overcomes

Score-based diffusion models typically use Brownian perturbations, which provide tractable reverse-time dynamics but impose memoryless noising. We introduce Volterra generative models, a continuous-time score-based framework whose forward process injects path-dependent noise through fractional kernels. To handle the non-Markovian and non-semimartingale dynamics, we construct finite-dimensional Markovian lifts using Gaussian quadrature in both regimes and a hybrid finite-difference exponential approximation in the smooth regime. We prove squared error bounds, derive an augmented linear-Gaussian forward process, and show that the learning can remain data-dimensional by considering residual states and analytic auxiliary Gaussian scores. We also identify covariance and reverse-time degeneracies caused by shared Brownian factors and signed smooth-regime weights. The degeneracy motivates stabilized conditioning and, for stiff larger lifts, a Gaussian-bridge reconstruction sampler. Experiments on MNIST and CIFAR-10 show that persistent fractional perturbations with small Markovian lifts can improve score-based generation on MNIST and provide a promising extension to natural images, while the bridge sampler provides a stability mechanism for larger lifts.

Recent advances in Large Language Models (LLMs) and multi-agent systems have driven the rise of Agentic AI, showing promise for medical reasoning. However, open-ended conversational agents remain prone to two critical failure modes: premature diagnostic handoff and silent clinical hallucinations that may go undetected before reaching the patient. In this work, we propose a multi-agent framework that addresses both issues by replacing ``LLM-as-a-judge'' routing with deterministic orchestration constraints. The framework incorporates two safety mechanisms. First, a neuro-symbolic state-tracking gate enforces completeness of the OLDCARTS clinical protocol (Onset, Location, Duration, Character, Aggravating/Alleviating factors, Radiation, Timing, and Severity) by blocking diagnostic transitions until all required dimensions are collected. Second, an epistemic uncertainty quantification (UQ) gate computes semantic entropy (H) across K=5 independent diagnostic samples to identify and intercept divergent outputs before delivery. We evaluate the system using simulated patient agents powered by the llama-3.1-70b-instruct model on 150 test cases. The full architecture achieves 49.3% diagnostic precision, representing an absolute improvement of 11.3 percentage points over an unconstrained baseline. Additionally, we observe a statistically significant negative correlation (r = -0.181, p < 0.05) between OLDCARTS completeness (σ) and semantic entropy (H), suggesting that structured information gathering is associated with reduced diagnostic uncertainty.

We introduce the Noise-Tilted Reverse Kernel (NTRK), a reward-guided diffusion sampler that injects reward gradients through the noise term, leaving the pretrained reverse kernel unchanged and requiring only a single sample per step. Reward-guided sampling at inference time has greatly expanded the versatility of pretrained diffusion models. Yet existing methods face a trade-off. Gradient-based guidance shifts the reverse mean, steering generation but pushing intermediate states outside the region that the model was trained on and degrading quality. Search-based methods preserve quality but gain no gradient signal. No prior method achieves both. NTRK resolves this by keeping the reverse mean fixed and biasing the noise term toward high reward. We introduce a whitening operator, the central mechanism behind NTRK, that makes the reward gradient safe to inject as noise without losing its guiding signal. Across various reward alignment tasks, NTRK outperforms recent state-of-the-art baselines without losing sample quality. Remarkably, on aesthetic generation, NTRK surpasses the reward of the best baseline at 500 NFEs using only 25 NFEs, a 20$\times$ reduction in compute.

As LLM-driven multi-agent systems (MAS) transition from lab to production, system behavior exhibits nonlinear degradation. We introduce the Intelligence Entropy Principle: probability-driven systems spontaneously drift toward disorder, formalized as S(t) = S0 * exp(alpha*t/Cm), where Cm is a model capability coefficient we propose. Lyapunov analysis yields the stabilization condition lambda > alpha/Cm. We construct the ADE (Agent Delivery Engineering) four-layer framework (L1 Physical Laws through L4 User Adaptation) with 23 core components. Validation spans 100K-scale experiments and 33.6 days of production monitoring. We propose a Five-Layer Disorder Taxonomy unifying failures under structural collapse, and present Elastic Organization as an original MAS morphology. Results: channel fracture reduced from 69-98% to near 0%; system death probability below 0.02%.

Medical image classification faces a fundamental dilemma: while deep learning models achieve remarkable performance at scale, real-world clinical scenarios often suffer from severe data scarcity due to annotation costs, privacy constraints, and disease rarity. This challenge is particularly pronounced in pathological scar classification, where differentiating keloids from hypertrophic scars requires subtle expert knowledge and labeled images are extremely limited. We propose a novel paradigm that repositions large language models (LLMs) as knowledge-driven feature engineers rather than end-to-end classifiers. We call this framework ScaFE (Scar Feature Engineering). Our key insight is that LLMs encode rich medical knowledge that can be externalized as executable feature extraction code, enabling the transformation of high-dimensional images into low-dimensional, clinically interpretable representations. Specifically, we prompt an LLM with established scar assessment criteria to generate deterministic Python code that extracts features aligned with clinical scoring systems such as the Vancouver Scar Scale. Our approach offers three key advantages: (1) data efficiency, achieving robust performance with limited training samples by decoupling knowledge acquisition from statistical learning; (2) privacy preservation, as raw images are processed locally without exposure to external LLMs; and (3) interpretability, through explicit features grounded in clinical reasoning. Extensive experiments on scar classification demonstrate that our method consistently outperforms end-to-end deep learning baselines or using LLMs as black-box classifiers under limited data conditions, establishing a promising direction for integrating LLMs into data-efficient and clinically transparent medical AI systems.

Large language models (LLMs) are widely used to fulfill users' information needs; users ask LLMs about the weather, pose educational questions, and consult them for legal assistance. One particularly understudied area is digital security and privacy (S&P), where users may seek LLMs' help on how to secure their online accounts or protect their computers from cyber attacks. To the best of our knowledge, no prior study has collected or analyzed the S&P questions users ask LLMs; prior research on LLM response quality relied on expert-authored S&P misconceptions or FAQs rather than user queries. Drawing from WildChat, a dataset of 3.2M user-LLM conversations collected in the wild, our study identifies 14,727 S&P prompts and categorizes them into nine categories covering a wide range of S&P topics. From the S&P prompts, we sampled 450 and performed a thematic analysis to characterize the S&P questions users ask LLMs. Separate from the thematic analysis, we curated 270 advice-seeking S&P prompts, where users ask for recommendations, guidance, or specific S&P information. We measured LLM response quality and consistency when posing the prompt to LLMs 10 times. We found that commercial LLMs outperform open-weight models (GPT 5.5 provided "good enough" responses on 98% of prompts; Llama 4 on 47%). However, among prompts that received high-quality responses on average, commercial models sometimes produce contradictory responses across runs, risking confusing or misleading users.

As Large Language Model based agents enter autonomous scientific research, their ability to resist pseudoscience becomes increasingly important. Otherwise, such systems may rapidly generate plausible yet misleading studies that contaminate academic literature and erode trust in science. We present PseudoBench, an adversarial benchmark for evaluating whether agentic auto-research systems can identify and resist pseudoscientific narratives. PseudoBench contains 200 curated pseudoscientific claim-evidence pairs across five domains and evaluates agents through an end-to-end research pipeline from experiments to writing. Testing seven state-of-the-art agents, we find that current systems readily produce persuasive reports that align with pseudoscientific premises with near-zero refusal rates and the highest resistance of only 27.4%. Stronger agents risk packaging pseudoscience in more sophisticated scientific language, increasing its apparent credibility. These findings reveal an alarming capacity to fuel pseudoscience, calling for scientific alignment before widespread deployment.

Caregivers often turn to online communities for informational and emotional support. In these spaces, peer supporters frequently draw on personal narratives to respond to emotionally complex caregiving situations. As LLMs are increasingly designed as peer-like sources of support, they introduce a critical tension: AI can provide immediate, private, and nonjudgmental support, but it cannot authentically possess the lived experiences that make human peer support meaningful. Yet, when prompted to sound peer-like, LLMs may generate language that implies lived experience. This creates a synthetic lived experience paradox: the same experiential language that may make AI support feel warm, relatable, and peer-like can also falsely position the system as someone with lived experience. We examine this paradox in the context of family caregivers of people living with Alzheimer's Disease and Related Dementias (ADRD). Drawing on caregiver support exchanges from online communities and prompted peer-like responses from three LLMs -- LLaMA, GPT-4o-mini, and MedGemma -- we analyze how human peers use personal narratives and how AI incorporates similar narrative forms. Psycholinguistic analysis shows that peer responses used significantly more first-person and past-focused language than peer-like AI responses. Qualitatively, we identify seven types of personal narratives in human peer support and show that AI often captures their emotional work, but can fabricate experiential grounding. These findings reveal a narrative authenticity gap: peer-like AI can generate synthetic lived experience without the real experience that makes peer support meaningful. We argue that caregiver-support AI systems need mechanisms to distinguish supportive peer-like framing from fabricated lived experience, ensuring that models can offer warmth and validation without falsely positioning themselves as experiential peers.

Hybrid architectures combining full attention (FA) and sliding-window attention (SWA) are a promising paradigm for efficient LLM inference. However, existing methods typically rely on hand-crafted rules or simple post-hoc heuristics for FA/SWA allocation and offer limited analysis of the attention behaviors underlying these designs. We propose Controllable Sparsity in Hybrid Attention (ConSA), a framework that learns optimal FA/SWA assignment under a user-specified sparsity target. ConSA employs L0 regularization to learn binary masks selecting between FA and SWA for each attention unit, while an augmented Lagrangian constraint enforces the target sparsity at either layer or KV-head granularity. We evaluate ConSA on two LLMs at the 0.6B and 1.7B scales. Learned allocations consistently outperform rule-based baselines, with KV-head-wise allocation yielding clear gains over layer-wise allocation. The learned patterns place SWA in the bottom layers and concentrate FA into contiguous middle-layer blocks, diverging from evenly interleaved patterns in rule-based methods. This structure persists across model scales, sparsity levels, and allocation granularities, revealing a fine-grained spectrum of intrinsic attention behaviors that underlies the learned allocation.

LLM agents increasingly rely on external skills -- reusable tool specifications -- but real-world tasks often require composing multiple skills, not just selecting one. We formalize this as the Compositional Skill Routing problem: given a complex user query and a large skill library, decompose the query into atomic sub-tasks, retrieve the appropriate skill for each sub-task, and compose an executable plan. We present SkillWeaver, a decompose-retrieve-compose framework combining an LLM task decomposer, a bi-encoder skill retriever with FAISS indexing, and a dependency-aware DAG planner. To support evaluation, we introduce CompSkillBench, a benchmark of 300 compositional queries over 2,209 real MCP server skills spanning 24 functional categories, sourced from the public MCP ecosystem. Our experiments reveal that task decomposition quality is the primary bottleneck: standard LLM decomposition reaches only 34.2% category recall at the step level. To address this, we propose Iterative Skill-Aware Decomposition (SAD), a retrieval-augmented feedback loop that iteratively aligns decomposition with available skills. SAD improves decomposition accuracy from 51.0% to 67.7% (+32.7%, Wilcoxon p < 10^-6) in a single iteration; DA-conditioned analysis confirms that correct granularity is the prerequisite for effective retrieval (CatR@1 rises from 34% to 41% when DA=1). SkillWeaver reduces context window consumption by over 99%, and transfer experiments confirm generalization (+35.6% relative DA gain even when target categories are absent from the retrieval pool).

Decision-making with deep learning-based time series forecasting requires not only accurate predictions but also actionable insights. However, current architectures do not inherently provide such information. Specifically, guidance is needed on how current conditions must be modified to shift from a predicted outcome to a desired future scenario. Counterfactual explanations provide a natural framework for this task, as they represent minimal input changes that alter the model's prediction, indicating when and how intervention is required. Existing approaches rely on instance-wise optimization, leading to inconsistency across instances, high computational costs, and limited applicability in real-time settings. To address these limitations, we reformulate counterfactual generation for time series forecasting as the problem of learning a globally consistent intervention strategy, allowing counterfactuals to be generated through a single shared function. We propose Counterfactual Time Series Explanations (ConTex), a model-agnostic, decomposed architecture comprising a temporal context encoder and a conditional encoder, followed by two heads that capture interventions in terms of temporal relevance and modification strength. This structure overcomes the instability and inconsistency of instance-based approaches by producing targeted, interpretable interventions across time and feature dimensions in a single forward pass, making it suitable for real-time applications. Across multiple forecasting architectures and benchmark datasets, ConTex achieves state-of-the-art validity while generating sparse counterfactuals that minimize the number of necessary interventions. Additionally, our approach reduces computational cost by at least 12-36x compared to instance-wise generation and supports real-time inference at approximately 0.007 seconds.

Vision-language-action models (VLAs) combine vision-language backbones with expressive generative action heads trained via flow matching on large-scale robotic datasets. Despite their strong empirical performance in robotic manipulation, VLAs lack mechanisms to quantify confidence in their predictions and to detect when their actions may be unreliable. This presents a critical limitation for real-world deployment in non-stationary environments, where models inevitably encounter scenarios outside their pretraining distribution and may fail without warning. To address this, we derive an efficient method for quantifying epistemic uncertainty in flow-matching models by leveraging velocity-field disagreement (VFD) across a small ensemble. We successfully use this uncertainty estimate for failure detection during deployment and active fine-tuning of flow-based VLAs. To this end, we propose SAVE, a framework for uncertainty-guided active multitask fine-tuning that reduces the number of costly expert demonstrations required to adapt VLAs to new tasks. Through extensive experiments on the LIBERO benchmark, we demonstrate that VFD yields better-calibrated uncertainty estimates predictive of downstream performance, that VFD achieves strong performance in detecting failures, and that uncertainty-guided data acquisition with SAVE requires at least 22% fewer samples than baselines. In summary, our work shows that quantifying epistemic uncertainty in flow-based VLAs improves both failure awareness and adaptation. Project website: tum-lsy.github.io/uq_vla/.

Deploying Large Language Models (LLMs) requires exploring a large configuration space spanning parallelization strategies, batching techniques, and scheduling policies. Exhaustive measurement across this space is impractical, making latency prediction essential for system optimization. While NPUs have emerged as accelerators designed for LLM inference, no prediction methodology has been established for them. Specifically, applying prior work to LLM inference latency prediction on NPUs faces three challenges: undisclosed microarchitecture of commercial NPUs, unpredictable compiler optimizations, and latency non-linearity induced by bucketing. We present LENS, a latency estimator that predicts NPU inference latency without information on the microarchitecture or compiler, and captures the non-linear latency induced by bucketing. LENS profiles each bucket with two end-to-end (E2E) measurements and composes the results to predict latency for arbitrary input-output length combinations. We validate LENS across NPUs from multiple vendors, several LLMs, and diverse workloads, achieving a mean prediction error of 2.15\%. We further compare LENS against two methodologically related baselines, confirming the validity of its approach.

This paper presents a low-cost home automation system implemented in a municipal swimming pool to address various challenges, including security concerns, air quality control, gas leakage detection, energy consumption reduction, and temperature and humidity control on the pool deck. The system utilises Arduino microcontrollers with sensors and actuators, enabling real-time data collection and analysis. The project is divided into two phases: hardware assembly and data analysis. In the hardware assembly phase, the Arduino sends data to a web Application Programming Interface (API) and stores it in a time-series database, with results presented in an Android application. The data analysis phase involves statistical exploration using libraries such as Pandas, NumPy, and Matplotlib. The proposed system aims to enhance decision-making based on collected and analysed data.

Tool-using LLM agents increasingly use the Model Context Protocol (MCP) to answer from heterogeneous evidence sources, including search, APIs, databases, clinical records, and formulary tools. Standard factuality metrics usually test whether an answer is supported by pooled evidence, missing a provenance-sensitive failure mode: a claim may be supported somewhere while being attributed to the wrong source. We call this cross-source conflation. We introduce ProvenanceGuard, a source-aware verifier for MCP-grounded answers. It consumes captured MCP traces with stable tool IDs, source IDs, and raw outputs; decomposes answers into atomic claims; routes claims to source-specific evidence; checks support with NLI and a token-alignment proxy; compares stated attribution with the routed source; and returns per-claim verdicts plus an answer-level allow/block decision. Blocked answers can be repaired with retrieval-augmented answer revision and re-verified. We evaluate on 281 medical-domain MCP-agent traces. A 266-trace adjudicated subset yields 2,325 LLM-assisted claim labels split by trace; 361 held-out labels are human-verified. On the 40-trace held-out split, ProvenanceGuard achieves block F1 0.802 and source accuracy 0.858 over 260 source-eligible claims, outperforming source-blind baselines that do not emit claim-to-source IDs. On a harder multi-source benchmark it reaches block F1 0.846, while source-plus-relation accuracy drops to 0.229, showing that exact source ownership remains difficult with semantically close sources. Repair-and-reverify resolves all blocked answers in the full trace set, often via conservative fallback. In 50 controlled clinical conflation probes, ProvenanceGuard detects all injected attribution swaps with no retained wrong attribution. These results show that source attribution is an independent axis for factuality verification in MCP-based agents.

Cross-lingual transfer in multilingual NLP has been widely explored in supervised fine-tuning contexts, where factors like data availability and linguistic similarity largely determine transfer quality. As the field shifts toward few-shot In-Context Learning (ICL), it is often presumed that insights from fine-tuning carry over unchanged. Yet this assumption has not been rigorously evaluated, leaving open the question of how to choose source languages for cross-lingual ICL. We conduct a broad empirical study of cross-lingual transfer in ICL spanning seven tasks, six models, and a typologically diverse set of languages. We further analyze language confusion, a key obstacle for generative tasks in cross-lingual ICL. Our results show that conventional fine-tuning-based expectations do not consistently apply in the ICL regime and point to alternative heuristics for selecting source languages effectively.

We propose a new weak-form Physics-Informed Neural Network approach (named INI-VPINN). INI-VPINN naturally incorporates Neumann boundary and interface conditions into the variational formulation. It removes the need for additional loss terms or multiple subdomain networks. This framework employs compact support weighting functions and integration by parts to implicitly impose flux and continuity constraints. In this way, it implicitly ensures physical consistency across material boundaries. The proposed method is tested on Poisson and Laplace problems with sharp interfaces and complex geometries. Results show that, compared with several other Physics Informed Neural Networks-based formulations, the INI-VPINN consistently achieves higher accuracy, smoother and faster convergence. The proposed framework provides a general approach for solving multimaterial problems with complex geometries and mixed Neumann-Dirichlet boundary conditions using neural networks. The implementation is publicly available in a GitHub repository.

Sparse Autoencoders (SAEs) decompose residual-stream activations into interpretable features. Recent latent-space defenses increasingly rely on these decompositions, assuming that identified "unsafe" SAE features serve as actionable handles for monitoring and intervention. In this paradigm, clamping a specific harmful feature is expected to reliably prevent model misbehavior. However, we show that this success may hide a recoverable failure mode: the clamp may block one visible route to a behavior without eliminating the behavior itself. We formulate this vulnerability as post-intervention recovery, a constrained residual-space optimization problem. Starting from the post-intervention residual state, we optimize residual perturbations to recover the pre-intervention behavior while preserving the post-intervention values of the targeted SAE features. Even under a strong threat model where the intervention remains active throughout optimization and generation, recovery remains possible. To rule out that recovery simply undoes the intervention, we use encoder-orthogonal updates for single-layer interventions and the corresponding feature-map Jacobian in the cross-layer setting. Across TPP, unlearning, IOI, and refusal steering experiments, this stress test reveals recoverable behavior despite successful feature-level intervention. Especially in the safety-critical refusal-steering setting, we achieve a 95.8% recovery rate on valid samples while keeping defended-feature relative drift to 0.131, substantially below suffix-based baselines. A recovery-path attribution analysis further localizes this recovery to the SAE reconstruction residual, the component left unexplained by the SAE. These results expose a gap between feature-level control and behavioral completeness: SAE features can support causal intervention, but controlling them does not guarantee control over the underlying behavior.

Catastrophic forgetting in continual adaptation is usually studied through parameter drift, replay, or distillation, but these views do not identify which output-space directions are vulnerable. We give a function-space account in the NTK regime: new-task training induces old-task prediction drift through the cross-task kernel, yielding a closed-form predictor for the forgetting vector before any new-task gradient step. In frozen-backbone linear-head PEFT-CL, where the model is linear in the trainable parameters, the predictor is exact up to numerical precision; for nonlinear adapters/full fine-tuning, it is a local NTK approximation. The same expression reveals that forgetting concentrates in a small number of old-task NTK eigenmodes and under frozen linear heads gives a Kronecker scaling rule for the vulnerable rank. These results clarify the relation to prior NTK-overlap theory, explain why parameter-space regularizers can miss output-space interference, and motivate a targeted spectral regularizer.

Looped Transformers scale latent computation by repeatedly applying shared blocks, but sequential looping increases latency and KV-cache memory with the loop count. Parallel loop Transformers (PLT) alleviate this cost through cross-loop position offsets (CLP) and shared-KV gated sliding-window attention, making loop count a practical design choice. We therefore study PLT loop-count selection through a gain--cost view: an extra loop may refine representations, but CLP also introduces a positional mismatch at each loop boundary. We instantiate this study by training LoopCoder-v2, a family of 7B PLT coders with different loop counts, from scratch on 18T tokens, followed by matched instruction tuning and evaluation. Empirically, the two-loop variant delivers broad gains over the non-looped baseline across code generation, code reasoning, agentic software engineering, and tool-use benchmarks, improving SWE-bench Verified from 43.0 to 64.4 points and Multi-SWE from 14.0 to 31.0 points. In contrast, variants with three or more loops regress, revealing a strongly non-monotonic loop-count effect. Our diagnostics show that loop 2 provides the main productive refinement, while later loops yield diminishing, oscillatory updates and reduced representational diversity. Because the CLP-induced mismatch remains roughly fixed as refinement gains shrink, the offset cost increasingly dominates. This gain--cost trade-off explains PLT's saturation at two loops and provides diagnostics for loop-count selection.

Masked diffusion models (MDMs) have recently emerged as a promising paradigm for sequence generation. Scaling MDMs is conventionally achieved by increasing the parameter count or the number of denoising steps. We introduce Recursive Masked Diffusion Models (R-MDMs), which add recursive depth as a third scaling axis by repeatedly applying the same denoising transformer within each diffusion step. Recursion enables iterative refinement of the output through parameter reuse, increasing effective model depth without increasing parameter count. Across structured generation tasks, including Sudoku and Countdown, we show that R-MDMs achieve substantially improved parameter efficiency: a model with $L$ recursive iterations often matches the performance of non-recursive baselines with roughly $L\times$ more parameters. Moreover, recursive refinement can partially substitute for additional denoising steps, allowing recursive models to reach the same generation quality with fewer forward passes at inference time. These results suggest that recursive depth is a practically useful scaling mechanism for MDMs, improving both parameter efficiency and the allocation of test-time compute.

AI systems deployed in legal workflows hallucinate at rates that aggregate metrics report at ~52%, but this average conceals where errors concentrate and in which direction they run, leaving compliance officers without an actionable signal for trustworthy deployment. We present LegalHalluLens, an auditing framework with three components: typed hallucination profiles across four legally-motivated claim categories (numeric, temporal, obligation/entitlement, factual) over CUAD (Hendrycks et al., 2021); a Risk Direction Index (RDI) that reduces omission-versus-invention bias to a single deployment-comparable scalar; and a typed debate pipeline calibrated to both magnitudes and directions. Across 510 contracts and 249,252 clause-level instances we measure a within-model gap of approximately 38-40 pp between obligation/numeric and temporal claims that aggregate reporting hides, and show that two systems with matched 52% rates can carry opposite RDIs. The debate pipeline reduces fabricated detections by 45% with per-category gains tracking the diagnosis, matching commercial APIs with a substantially smaller backbone (4B active parameters). Typed profiles and RDI surface failure modes that aggregate metrics hide; we further show these diagnostics serve as calibration inputs for multi-agent debate pipelines, where Skeptic challenges and asymmetric gates targeted at measured failure modes outperform generically-tuned debate. The framework supports direction-aware procurement, accountability, and agent design for legal AI deployed in the wild.

Dementia and depression are the most prevalent neuropsychiatric disorders in geriatric populations, and their overlapping symptoms pose major challenges for differential diagnosis. In this study, we investigate open-weights Large Language Models (LLMs) for predicting dementia and depression severity from speech samples collected during standardized history taking interviews with 154 German-speaking subjects. We introduce an observer-based Global Depression Scale (GDS-D) aligned with the established Global Deterioration Scale (GDS), enabling parallel global staging of affective and cognitive symptoms. We compare three LLMs (Mistral 3.1, DeepHermes, Qwen3) in two settings: (1) zero-shot prediction and (2) LLM-based feature extraction for Support Vector Regression, using human and pause-enriched transcripts. Results show that LLMs effectively predict depression severity in zero-shot settings (best MAE of 0.60), while dementia assessment benefits substantially from structured feature extraction (best MAE of 0.78), reducing errors by up to 35% over zero-shot baselines. Pause-enriched transcripts achieve competitive performance with human transcriptions, demonstrating the viability of fully automatic screening pipelines for differential neuropsychiatric assessment.

Air Traffic Control Operators (ATCOs) are vital in ensuring the safe, orderly, and efficient flow of air traffic, yet training capacity is constrained by reliance on specialized human trainers known as simpilots, who must role-play both pilots and ATCOs in a simulated airspace. Existing automated solutions rely on Western-centric speech models that perform poorly in Singaporean operational contexts, with off-the-shelf systems exhibiting Word Error Rates (WER) of up to 107.80% on Singaporean-accented aviation speech. We introduce ASTRA, an end-to-end training simulator that automates these simpilot roles through a pipeline that transcribes ATCO speech, interprets instructions, and generates appropriate pilot and ATCO responses using locally adapted voice models. Our fine-tuned Automatic Speech Recognition (ASR) pipeline reduces WER to 23.45%, substantially outperforming existing approaches in this domain. Beyond traffic simulation, ASTRA incorporates an AI-assisted performance evaluation framework that assesses trainee radiotelephony communications across accuracy, brevity, and completeness, achieving post-optimization scores of 91.7%, 88.2%, and 86.9%, respectively. Built on open-source foundations such as DSPy and Unsloth, this approach enables scalable, standardized ATCO assessment while reducing instructor workload.

Most graph neural network (GNN) cores rely on graph convolutions, typically implemented as message passing between direct (single-hop) neighbors. In many real-world graphs, edges can be noisy or poorly defined, limiting information propagation to local neighborhoods. Existing diffusion kernels, such as Personalized PageRank (PPR) and Heat Kernel, alleviate this issue through global propagation, but still struggle with complex local structures and distant node noise. To address these limitations, we propose a K-Hop Gaussian (KHG) diffusion kernel as a preprocessing module for graph data. KHG introduces multi-hop diffusion with Gaussian weighting for remote nodes, balancing local and global information propagation before applying standard GNNs. Experiments on multiple benchmark datasets demonstrate that KHG significantly outperforms traditional message-passing GNNs, as well as PPR and Heat Kernel diffusion, particularly in noisy or structurally complex graphs.

Soil Moisture (SM) modelling constitutes a complex spatiotemporal learning problem characterised by nonlinear environmental interactions, heterogeneous data sources, and limited ground observations. Physics-based approaches, such as water balance models, rely on explicit hydrological equations and high-quality inputs, but their computational cost and scalability limitations restrict large-scale deployment. Data-driven artificial intelligence (AI) methods have emerged as flexible alternatives, enabling the extraction of empirical relationships between soil moisture and environmental variables with reduced modelling assumptions. This work presents a structured survey of AI-based models for soil moisture estimation and classification. Existing approaches are organized into five categories: (a) statistical time-series models, (b) geostatistical methods (c) classical machine learning (ML) models, (d) Deep Learning (DL) models and (e) Probabilistic/Bayesian methods. These models leverage historical soil moisture records, meteorological variables, vegetation indices, topography, soil characteristics, and geolocation data to perform regression or classification tasks.

Sequential output generation with large-scale Transformer and diffusion decoders pays a memory cost that grows with sequence length, plus iterative per-step computation. Replacing them with small feed-forward decoders restores efficiency but produces unstructured latent representations that limit closed-loop control: phase-conditioned action generation and cross-step latent carry-over both require a latent geometry with stable basins. This article proposes Ghost Attractor Networks, a theoretically derived dynamical decoder whose latent evolves under a learned potential with drift and produces a basin-attractor structure by construction. Three desiderata (multi-modality, decoder-level single-pass switching, and constant memory) motivate the potential-drift form, and mode transitions arise as saddle-node bifurcations with ghost-attractor escape. A hierarchical phase-space decomposition separates first-order basin convergence from second-order proprioceptive refinement. Empirically, a Ghost trained end-to-end with a behavioral-cloning and contrastive objective exhibits the predicted gradient-flow contraction in its potential, with the gradient norm decaying by 67 percent across five integration steps on 1430 held-out samples. Ghost is evaluated as a robotic action decoder. A 2.3-million-parameter Ghost matches the offline accuracy of a 1.07-billion-parameter Diffusion Transformer at 462 times fewer parameters and 32 times lower latency, and beats five alternative 2M-parameter decoders (MLP, Neural ODE, CVAE, Transformer, 1-step Diffusion) on offline mean squared error by 5.9 to 29 percent. On the LIBERO-10 closed-loop benchmark, phase conditioning on Ghost's basin-structured latent yields a 13.5 percentage-point success-rate gain over a feed-forward MLP baseline, and persistent-latent ensembling reaches a 95.7 percent final success rate.

Federated learning allows multiple clients to jointly train a shared model by sending gradient updates to a central server while keeping raw inputs local. However, prior gradient inversion attacks show that these updates can reveal enough information to reconstruct client inputs. Existing attacks on transformers either optimize dummy inputs to match the true client updates, which is costly and unstable for modern models, or exploit the low rank of attention gradients to identify a subspace containing the true layer embeddings, followed by a discrete membership test for candidate tokens. However, this token test is brittle under numerical noise, i.e., from quantization or Differential Privacy (DP), and scales poorly for encoder models with non-causal attention. We introduce TIGER, a continuous gradient inversion attack that turns this subspace signal into a differentiable objective. Instead of searching over tokens or matching full gradients, TIGER directly optimizes token embeddings to minimize their distance to the subspace. Our experiments demonstrate that on encoder-only models, TIGER substantially improves both reconstruction quality and runtime over existing attacks, while on decoder models, TIGER is more robust than prior subspace-based attacks, enabling the first successful reconstructions in DP-defended federated learning settings.

Injecting malicious knowledge into retrieval-augmented generation (RAG) systems can manipulate retrieved evidence and mislead downstream generation, posing a serious security threat for AI applications. Existing RAG injection attacks mainly rely on manipulating external knowledge bases, such as crafting malicious corpus. However, the synthetic text crafted by such data-centric methods could be detectable, leading to the failure of attacks. Beyond corpus manipulation, open-source retrievers are increasingly exposing RAG systems to model-centric attacks. In this paper, we propose conflict-aware retriever editing, i.e., CAREATTACK, a model-centric retriever attack framework for malicious knowledge injection in RAG. Specifically, CAREATTACK consists two stages of conflict-aware retriever editing and attack-preserving anchor repair. Conflict-aware retriever editing adapts efficient closed-form parameter editing to the dense retrieval model, promoting malicious knowledge above benign competing passages and resolving potential parameter conflicts through graph-based conflict detection and parameter editing projection. Then, attack-preserving anchor repair performs lightweight calibration on the edited retriever to further eliminate the impact on non-target prompts while preserving the attack effectiveness for target prompts. We instantiate CAREATTACK on Qwen3-Embedding-0.6B and BGE-M3, and conduct evaluation on three benchmark datasets. Experimental results demonstrate our method substantially promote malicious passages into the retrieved knowledge of RAG systems and can perform attacks for batches of target prompts and passages, given the access of retrieval model parameters. Since most RAG systems are built upon open-source retrieval models, this work reveals a practical attack surface in RAG systems. Codes are public accessible at https://anonymous.4open.science/r/CareAttack-3F1C.

Large Language Model (LLM) unlearning aims to remove undesirable knowledge or behaviors while preserving retained capabilities. Current unlearning methods all involve a trade-off between unlearning and retention. We have found that the retention activation bias can also be used to quantify the damage an unlearning method inflicts on retention, without considering the specific implementation of the unlearning process. This allows us to restore retention performance for any unlearning method using a post-hoc approach. Therefore, we propose a complementary post-hoc setting to sanitize the final update vector without rerunning the original unlearning pipeline. In this setting, we design SAGE, Spectral Activation-GEometry Sanitization, a source-agnostic correction for final unlearning updates. SAGE collects real module inputs from a small retain proxy, extracts their dominant activation geometry, and solves a source-anchored optimization objective in closed form, which suppresses update components aligned with high-energy retained directions while preserving the source method's forgetting carrier. Across multiple unlearning methods, model scales, and benchmarks, SAGE consistently relieves the retain-forget trade-off, identifying post-hoc sanitization of final vectors as a practical and underexplored axis for machine unlearning.

Safe coordination in networked cyber-physical systems forces learning algorithms to simultaneously handle hybrid discrete-continuous actions, hard training-time safety constraints, and physics-governed dynamics. We show that these three features form a directed cycle of biases that defeats any naive composition of off-the-shelf modules, and formalize this as a three-way coupling lemma. We then introduce TRIDENT, the first MARL framework whose three components are co-designed to cancel each leak: a Richardson-Romberg gradient correction reducing Gumbel-Softmax bias from O(tau) to O(tau^2), a Lyapunov-constrained sequential trust-region update enforcing per-iterate feasibility, and a physics-informed residual critic that decomposes value rather than reward. We prove an O~(1/sqrt(K)) convergence rate to a constrained Nash equilibrium and an O(sqrt(K)) cumulative-violation bound. On multi-UAV mobile-edge computing, autonomous intersection management, and a hybrid SMAC variant, TRIDENT cuts training-time violations by 95.5% over MADDPG and 76.3% over MACPO, while improving reward by 13.5% over the strongest unconstrained baseline.

Optimizing the training data distribution for Supervised Fine-Tuning (SFT) dictates the capability of Large Language Models (LLMs). While existing data curation methods excel at accelerating training under constrained budgets, they are less suited to elevating the capability upper bound. The challenge here is no longer to identify a smaller subset that preserves performance, but to refine the data distribution toward instances most capable of improving the final model. To address this problem, we explore instance-level data attribution using Influence Functions (IF). We identify that standard IF formulations struggle in this setting due to two structural limitations: a proximity gap caused by off-policy validation targets, and a severe bias towards gradient norm. We propose DRIFT (Data Refinement via On-Policy Influence Functions for Supervised Fine-Tuning). Instead of relying on external reference data, DRIFT utilizes the model's on-policy rollouts as validation targets, which empirically minimizes the parameter proximity gap and better aligns with the local neighborhood assumption of IF. It further applies signed weighting based on trajectory correctness and debiases influence scores against the gradient hacking issue, allowing a small set of validation queries to act as reliable anchors for attributing the full dataset. Experiments on 7B-parameter instruction and reasoning models show that DRIFT consistently raises the performance ceiling on both, outperforming existing data curation baselines.

Gaussian width is a central geometric complexity measure in high-dimensional probability, compressed sensing, convex optimization, and learning theory. It quantifies the average extent of a set along random directions, thereby capturing the effective dimension of constraint sets, hypothesis classes, and descent cones. However, this notion is intrinsically Euclidean. Statistical models instead carry a natural Riemannian geometry induced by the Fisher information metric, where directions are scaled according to statistical distinguishability rather than ambient Euclidean length. We introduce Fisher width, a Fisher-geometric analogue of Gaussian width for statistical manifolds. At a parameter point $θ$, Fisher width replaces the Euclidean identity by the local metric tensor $G(θ)^{1/2}$, measuring the Gaussian width of the Fisher-rescaled set. This makes the resulting quantity sensitive to local statistical curvature and invariant under smooth reparameterizations. We develop the basic theory of Fisher width, showing that it retains key structural features of Gaussian width, including concentration, metric perturbation stability, and spectral comparison bounds with the Euclidean baseline, while also capturing anisotropic geometric effects invisible to Euclidean measures. As an application, we prove a generalization bound for Fisher-Lipschitz hypothesis classes and propose computable estimators, which we evaluate empirically on MNIST across three model classes. Fisher width is to statistical manifolds what Gaussian width is to Euclidean convex bodies. This work lays the foundation for studying complexity and learning on curved statistical manifolds.

Operator learning is an emerging interdisciplinary field that integrates machine learning with scientific computing. By mapping infinite-dimensional function spaces, this approach provides an efficient surrogate modeling framework for high-dimensional partial differential equations (PDEs). Compared to traditional numerical solvers, it achieves a superior trade-off between computational complexity and approximation accuracy, demonstrating significant advantages in many-query tasks such as real-time prediction and parameter sweeps. Given the stringent accuracy requirements of both forward simulation and inverse inference, as well as the precision bottlenecks of existing operator learning methods in handling complex boundaries or long-term evolution, we propose the Starter-Iterator Neural Operator (SINO). Our framework reinterprets the initialization strategies and iterative formats of traditional iterative methods through neural networks, establishing an efficient approach for spectral-spatiotemporal collaborative modeling. Specifically, the frequency-domain initialization module captures globally stable low-frequency features, while the time-domain learning module focuses on optimizing local solution residuals, thereby effectively overcoming the inherent limitations of conventional single-domain modeling approaches. Extensive experiments on typical dynamical systems such as the Navier-Stokes equations and acoustic wave equations, as well as practical applications including super-resolution imaging and weather forecasting, demonstrate that SINO achieves outstanding performance in numerical accuracy, generalization capability, and robustness.

Mixture-of-Experts (MoE) models scale compute efficiently, yet remain expensive to deploy due to their substantial memory footprint and inference overhead. Prior compression methods mainly operate at the expert level, either removing entire experts or ranking experts by coarse-grained importance scores. However, such expert-wise decisions are often too coarse to capture fine-grained redundancy, leading to misallocated pruning budgets and limited compression. To address this problem, we observe that information within MoE experts is highly concentrated in a small subset of channels, leaving substantial redundancy even in experts deemed important. Based on this observation, we propose a structural pruning framework tailored for MoE models. Our method reformulates prune-ratio allocation as a channel-score coverage maximization problem and solves it efficiently using an attribution-based approximation. Experiments on DeepSeek and Qwen MoE models show that our method preserves model accuracy under 50% or 25% structured pruning when combined with 4-bit quantization. On Qwen3-30B-A3B, our approach reduces memory footprint by 5.27$\times$ and consistently outperforms state-of-the-art baselines across diverse benchmarks.

We develop a mathematically explicit link between shock-wave theory and the symmetry-quotiented learning dynamics of stochastic gradient descent, drawing on differential geometry, Lie group theory, and fluid mechanics. Specifically, after quotienting parameter symmetries and applying local-entropy coarse-graining, the effective dynamics satisfy a viscous Hamilton--Jacobi equation on the quotient manifold. Moreover, under the assumption that the raw parameter dynamics can be summarized by a gradient field on the quotiented space, the gradient of the coarse-grained loss function obeys a Burgers-type equation, and shock formation can be established rigorously. We apply our theory to multilayer perceptrons, convolutional neural networks, Transformers, and mean-field networks, and show that they obey the Hamilton--Jacobi or Burgers-type equations. We conjecture that this framework also yields practical diagnostics for deep learning. In architectures such as Transformers, raw parameter norms are often distorted by symmetry redundancy and may therefore be misleading, whereas symmetry-corrected quotient observables provide a principled basis for monitoring, forecasting, and controlling training-phase transitions.

Correct identification of fish species is highly significant for food security, economic development, and climate resilience in Bangladesh. Protein sequences directly reflect functional and evolutionary constraints which are important for species authentication and biodiversity monitoring. Yet there exists no benchmark for native Bangladeshi fish species identification from protein sequence. In this study, we addressed this gap by introducing the first curated dataset for nine native Bangladeshi fish species of 2845 high quality protein sequences. We also established the first protein sequence classification baseline for this domain through a systematic benchmarking of seven architectural paradigms. Moreover, we propose a realistic deployable novel hybrid architecture of MotifCNN and Transformer with Terminal-Aware Positional-Encoding (MotifCNN-Transformer+TA-PE). Our novel architecture achieves 79.80% accuracy with macro-F1 of 0.80. The highest 83.04% accuracy is achieved by finetuned protein language model ProtBERT that has 420M parameters and requires dual 16GB GPUs for inference. According to McNemar's test, ProtBERT's 3.24% accuracy gain over our MotifCNN-Transformer+TA-PE is statistically insignificant (p = 0.1120). Our novel architecture beats it among six of the nine classes in per class identification. Also our MotifCNN-Transformer+TA-PE is approximately 5x faster, 42x smaller, and supports 16x larger batch size than ProtBERT and has GPU free inference, making it more practical for deployment in resources constrained areas such as rural Bangladesh. Beyond this, our foundational work shows effects of phylogenetic relationships on sequence similarity and establishes pathways for fisheries management, food authentication and biodiversity conservation in South Asia's protein dependent economy.

Despite rapid advances in UAV technologies, current deployments remain limited due to several gaps in UAV systems research. To address these challenges, we propose OmniDroneX, a unified Drone-as-a-Service ecosystem, in which drones are transitioned from fixed function platforms into dynamically composable entities that can be integrated with external infrastructures to offer omni-capabilities. OmniDroneX bridges low-level physical primitives with high-level mission intent through a unified vendor-agnostic interface (libUAV) and a formal physical-service abstraction model (PT-SOA). A core innovation is the diverse application of large language models (LLMs) across multiple layers of the OmniDroneX architecture. LLMs are used to assist in identifying and formalizing primitive device functions and abstract service definitions, supporting automated service composition and workflow generation, and enabling interactive, natural-language mission specification and refinement. OmniDroneX also incorporates important categories of composition techniques that are essential in dynamic UAV systems, including physical layer composition for drone capability augmentation, as well as spatiotemporal, functional, collaborative, exception-aware, and QoS-based service compositions. Collectively, these features allow OmniDroneX to serve as a foundation for scalable, resilient, and self-evolving UAV ecosystems operating in complex and dynamic environments.

Binary data factorization is common, but real-valued methods ignore discreteness and yield hard-to-interpret factors. Boolean Matrix Factorization (BooMF) instead decomposes a binary matrix into two lower-rank binary matrices via logical AND and OR, expressing the data as a Boolean disjunction of interpretable patterns. In cancer genomics, BooMF can reveal coordinated feature changes that may drive tumor evolution, unlike rotational or additive decompositions. Most existing BooMF methods are heuristic, greedy, sensitive to initialization, prone to local optima, and do not support principled model selection or uncertainty quantification. We introduce Bayesian Boolean Matrix Factorization (BBMF), a fully conjugate generative model with sparsity-inducing priors. It enforces Boolean constraints, yields interpretable latent factors with coherent uncertainty quantification, and admits Gibbs sampling with closed-form full conditionals. Because cancer evolution often involves widespread, near-simultaneous chromosome-number changes (e.g., whole-genome duplication followed by instability and selection), Boolean factorizations capture these patterns more naturally than additive models. Applied to arm-level copy-number alteration data in multiple myeloma, where entries indicate presence/absence of chromosomal-arm amplifications, BBMF finds a small set of interpretable bicliques linking patient subsets to recurrently co-altered chromosomal arms, providing a compact, biologically meaningful summary of tumor heterogeneity and demonstrating BBMF's utility for uncovering discrete latent structure in complex binary data.

AI agent performance is not just a modeling problem, it is fundamentally a systems problem. The advanced capabilities of models are realized through agent harnesses. Therefore, a gap between model assumptions and harness behavior can easily prevent the model's full capabilities from translating into agent performance. We formalize this as the `intent-execution' gap: the mismatch between what the model intends and what the harness executes, and vice versa. We argue that minimizing this intent-execution gap is as important as other aspects of harness design such as tools and execution loops. To illustrate the impact of this harness-model alignment, we develop a simple and customizable harness called `Simple Strands Agent' (SSA). SSA aims to find the bulk of common patterns which generalize across different model families (such as Claude, Gemini, GPT, Grok, Qwen), as well as a small number of model-specific preferences. We make two contributions: (i) we reproduce or improve on the pass@1 performance reported by diverse model-provider families on popular agentic benchmarks (SWE-Pro, SWE-Verified and Terminal-Bench-2), and (ii) building on an analysis of 138k trajectories generated by SSA, we look beyond the pass@1 numbers which tend to be relatively even across frontier models. By representing agent trajectories in code state-spaces, we observe model-level differences in problem-solving behavior. Finer-grained metrics such as edit frequency, testing activity, and phase-transitions reveal how individual models allocate effort across different stages of autonomous problem solving.

Large language model agents are increasingly integrated into map services. Since map services are embedded in everyday-life scenarios rather than professional task settings, users often express their needs informally, resulting in underspecified queries with many unspoken needs, namely, implicit decision factors that are critical for user satisfaction. Although clarification is an effective way to mitigate this issue, it increases user burden in daily interaction, and a capable agent should first proactively recover such factors from available information sources. However, evaluating this ability is challenging. The first challenge is to determine which implicit decision factors are suitable for evaluation. A factor is evaluable only if it affects user acceptance and can be recovered from information available to the agent before it responds. Second, user satisfaction cannot be reliably represented by a single reference answer, requiring a benchmark that converts satisfaction-relevant factors into objective and quantifiable evaluation targets. To address these challenges, we propose a restore-identify-filter framework that reconstructs complete user needs from behavior-chain evidence, identifies implicit decision factors, and retains only those supported by pre-query evidence. Building on this methodology, we construct MapSatisfyBench from large-scale, real-world anonymized user data and annotate ground truth from five dimensions and enables full-chain evaluation of satisfaction-aware map agents. Experiments show that current agents generally perform well on explicit task completion, but remain limited in satisfying implicit decision factors and proactively acquiring the evidence needed for satisfaction-aware decisions. These findings establish MapSatisfyBench as a benchmark for shifting map-agent evaluation from task completion toward satisfaction-aware spatial decision making.

Pathological images are inherently multi-scale, requiring pathologists to integrate evidence from global tissue architecture at low magnification to cellular morphology at higher magnification for accurate diagnosis. While existing pathological datasets for vision-language model (VLM) include various scales, they often lack an explicit cross-scale reasoning objective. This limitation prevents VLMs from capturing essential cross-scale representations and learning evidence-based reasoning. To bridge this gap, we introduce the first cross-scale training and evaluation paradigm that formulates pathology interpretation as multi-magnification reasoning. However, creating such a task reveals a critical challenge: multi-image visual question answering (VQA) is prone to text-only shortcuts, which allow models to guess answers using magnification-dependent artifacts rather than visual evidence. To address this, we propose a leakage-aware curation pipeline that combines adversarial text-only screening with constraint-guided question design. Using this pipeline, we construct Scale-VQA, a high-quality benchmark with 4,685 multiple-choice questions grounded in 2,537 pathology images across multiple magnification levels. Finally, we present ScaleReasoner-R1, a model trained via reinforcement learning to optimize performance on the cross-scale VQA task. ScaleReasoner-R1 achieves state-of-the-art performance on our cross-scale reasoning benchmark and generalizes to SOTA performance on established single-scale benchmarks. Findings suggest that even the limited cross-scale supervision can significantly improve pathological understanding. The code and demos will be open-sourced.

Thanks to rapid developments in generative AI, we are in the midst of a paradigm shift that may change how we interact with computers forever. We have observed a growth in the use of natural language prompts to build applications and coding infrastructures without underlying knowledge of the field, and this practice has been dubbed `vibe coding.' It arguably represents what the field of programming has been building towards since the beginning, with every higher level of abstraction that is conceived. Vibe coding promises to be the endpoint for the meta of high-level programming as far as method of input is concerned: eliminating a human's use of code syntax entirely in favour of programming in their mother tongue. This paper aims to evaluate the viability of vibe coding for greenfield software engineering tasks, as well as analyse the benchmarks that have been used to measure its software engineering prowess. To this end, we have developed an evaluation suite for analysing an LLM's proficiency in carrying out simple, isolated greenfield programming tasks in Python to provide scoped insight on the matter.

Current multilingual evaluations for Vision-Language Models (VLMs) assume a one-to-one mapping between language and orthography, overlooking billions of users of multi-script languages. We introduce PuMVR (Punjabi Multimodal Visual Reasoning), a benchmark of 1,000 strictly parallel image-text instances across Punjabi's three active scripts: Gurmukhi, Shahmukhi, and Roman. Evaluating 10 state-of-the-art VLMs, we expose a substantial and systematic Script Gap. Models frequently solve visual tasks in one script while failing identical tasks in another, with accuracy deltas reaching 16%. Crucially, visual input boosts absolute performance uniformly yet does not close the orthographic gap. Furthermore, cross-script in-context transfer is highly brittle, exposing script-locked knowledge representation. Supported by McNemar tests across all script pairs, our findings demonstrate that current "multilingual" VLMs are not truly multi-script. We propose the Script Consistency Rate (SCR), which falls as low as 24.8% on our benchmark, as a mandatory metric for script-agnostic evaluation to ensure equitable AI access. Data and code are available at: https://github.com/prabhjotschugh/Not-Truly-Multilingual-PuMVR.

High-Level Synthesis (HLS) relies on transforming original C specifications into synthesizable HLS-oriented C (HLS-C) implementations. Functional consistency verification between original C specifications and HLS-C implementations is a critical yet labor-intensive task in HLS design flows. While Large Language Models (LLMs) have recently shown promise in automated testbench generation, their stochastic nature often leads to insufficient coverage, inconsistent verification environments, and unreliable equivalence checking results. To address these limitations, we propose a knowledge-augmented, agent-driven shift-left verification framework for automated functional consistency checking between original C and HLS-C implementations before synthesis. The framework introduces a Dual-Tier Consistency Checking mechanism that jointly enforces static structural alignment and dynamic behavioral equivalence between paired testbenches, while integrating symbolic execution and coverage-driven refinement to improve verification completeness. Furthermore, we construct a heterogeneous HLS Verification Knowledge Graph to provide topology-aware reasoning priors for testbench generation, and design an autonomous verification agent to orchestrate iterative refinement and failure diagnosis across heterogeneous toolchains. Experimental results on 107 HLS benchmark pairs demonstrate that the proposed framework achieves 0.9826 average coverage and 0.9533 dynamic consistency, outperforming representative AST-based, retrieval-augmented, and iterative agent-based baselines. https://github.com/cz-5f/HLS-LeVeri.git

This paper investigates the creative process of automated design and artistic evaluation using an evolutionary system. We consider how a multimodal artificial intelligence (AI) model can communicate and guide a combined generative and evolutionary computational system. This creates a framework for the evolution of aesthetically pleasing complex 3D organic forms by integrating genetic algorithms with the visual reasoning capabilities of large-scale AI foundation models. The framework shifts the artist role from that of intensive direct selection to one of system design; transferring detailed step-by-step curation to an AI agent capable of multimodal aesthetic judgement. This framework enables the human artist/designer to rapidly traverse large areas of multi-dimensional evolutionary parameter space to find creative outcomes based on their semantic targets. Detailed audit trails of the AI's aesthetic reasoning are generated for each experiment. Interactive visualisation tools, together with AI-generated summaries and evolutionary narratives, enable deep exploration into each evolutionary experiment and providing a transparent insight into the AI-guided process.

As large language models (LLMs) are increasingly deployed in real-world applications, alignment is no longer governed by a single universal notion of safety or helpfulness, but instead by provider- or application-specific model specifications. These specifications are typically long, structured, and frequently updated, yet existing alignment pipelines lack a systematic mechanism to operationalize them as training signals. In this paper, we propose specification-grounded alignment, a new alignment paradigm that treats provider-authored model specifications as the primary alignment target rather than abstract principles or static benchmarks. To instantiate this paradigm, we introduce SpecAlign, a framework that synthesizes alignment data directly from specification documents. SpecAlign combines structured rule annotation, controllable specification instantiation, and multi-agent adversarial data synthesis to generate fine-grained, boundary-aware preference pairs that capture both compliant behaviors and meaningful specification violations. Experiments across multiple model specifications and backbone models demonstrate that training with SpecAlign consistently improves rule compliance while preserving general capabilities and avoiding over-conservative behavior. These results suggest that grounding alignment in explicit model specifications enables rapid, precise, and scalable adaptation of LLM behavior to evolving policy requirements.

Modern deep learning models remain notoriously prone to overconfidence, limiting their reliability in high-stakes applications. Bayesian methods aim to counter this by learning a distribution over model parameters, and recent advances now make this feasible for large-scale architectures at costs comparable to AdamW. However, a challenge remains at test time: predictions must be averaged across many forward passes with weights sampled from the posterior, which is prohibitively expensive. Variance propagation offers an efficient alternative, computing layer-wise analytical approximations of uncertainty in a single forward pass. While such techniques are effective for MLPs, their extension to modern architectures remains challenging, due to increased depth and diversity of layer types. To fill this gap, we propose Calibrated Variance Propagation (CVP), which introduces a new propagation method for normalization layers, combines it with recent techniques for handling activation functions, and absorbs residual error through a light calibration step. CVP yields comparably accurate uncertainty estimates to MC sampling across transformers and CNNs, at a fraction of the cost. Against prior variance propagation work, CVP improves coverage at $0.5\%$ risk from $8.2\%$ to $14.6\%$ with BEiT-3 on Visual Reasoning (NLVR2) and from $2.6\%$ to $10.8\%$ with ViLT on VQAv2, with gains extending to convolutional architectures.

Compared with binaries and decompiled code, malware source code more directly reflects the attackers' original intent. However, the scarcity of source code and the high cost of manual review make such datasets difficult to build and maintain. We propose MASCOT-Android, a curated dataset of Android malware source code and an automated collection framework for scalable malware source code discovery on GitHub. A key finding of our work is that repository-level documentation alone provides a strong signal for malware source code collection. Our model extracts character-level TF-IDF features from 8,772 malware and 25,747 benign README documents and trains a LinearSVC classifier to distinguish malware repositories. This README-only model achieves an accuracy of 96.28\% and an FPR of 1.06\% in local evaluation. In addition, the model outputs confidence scores, allowing users to adjust the decision threshold to balance FPR and coverage, which is practical in real-world malware source code collection.

We introduce GRACE-DS, a Guarded Reward-guided Agent Correction Environment in Data Science for pre-deployment evaluation of LLM-powered AutoML agents. GRACE-DS is a set of evaluation metrics in an isolated environment that can be applied to tabular ML tasks specific to a particular organization. It exposes agents to realistic workflow stages, from planning and data inspection through feature engineering, model development, validation, and code repair to final submission, while hidden executable validators measure not only final predictive performance but also leakage avoidance, reproducibility, protocol validity, correction behavior, and reward alignment. The strongest structured regime, flexible iterative interaction (our approach), achieves higher end-to-end normalized hidden-test quality than single-shot generation, unstructured interaction, and restart-based baselines, while also improving protocol-valid completion. Validated across more than 7,000 episodes, these results establish GRACE-DS as a robust platform for assessing the capacity of LLM-based AutoML agents to execute machine learning workflows under production-like conditions and in accordance with organization-specific requirements.

While Large Language Models (LLMs) have substantially advanced text-to-code synthesis, many real programming tasks specify intent through visual artifacts such as screenshots, charts, vector drawings, videos, and interactive states. These tasks require models to connect visual perception to executable programs, because correctness depends not only on syntax but also on layout, data semantics, interaction behavior, and domain-specific constraints that apply after execution. This survey examines Multimodal Code Intelligence, covering systems that generate, edit, refine, or reason with code under visually grounded inputs and outputs. We first formulate the field by the role that code plays in each task, distinguishing code as a rendered artifact, an editable symbolic structure, a scientific representation, an intermediate reasoning trace, or an executable policy or tool interface. We then organize benchmarks and methods into four domains: Graphical User Interface, Scientific Visualization, Structured Graphics, and Frontier Tasks and Frameworks. This taxonomy connects mature artifact-generation problems to emerging agentic and unified settings and allows us to compare how different tasks treat evidence of correctness. Looking ahead, we argue that future research may benefit from four verification-centered directions. Multi-signal validation can combine complementary evidence of correctness, multi-state verification can test behavior across execution trajectories, cross-task transfer testing can probe reusable visual-code skills, and verifiable agent traces can reveal whether agent actions are grounded in visual evidence. Together, these directions may move this field from single-output imitation toward evidence-grounded executable systems. An ongoing project and resources are available on \href{https://github.com/xjywhu/Awesome-Multimodal-LLM-for-Code}{GitHub}.

Coding agents are increasingly used to automate software engineering tasks. To guide their behavior, these agents commonly rely on configuration files, typically named AGENTS.&zwnj;md or CLAUDE.&zwnj;md, which provide instructions about architecture, workflows, coding conventions, and testing practices. Despite their growing importance, little is known about common problems affecting the definition and maintenance of these files. In this paper, we present the first catalog of smells for coding-agent configuration files. To identify such smells, we first conducted a grey literature review and a repository mining analysis. As a result, we identified six configuration smells and proposed automated heuristics to detect them. To evaluate the prevalence of the proposed smells, we analyzed 100 popular open-source repositories containing either an AGENTS.&zwnj;md or a CLAUDE.&zwnj;md file. Our results show that configuration smells are widespread. Lint Leakage was the most common smell, affecting 62% of the files, followed by Context Bloat (42%) and Skill Leakage (35%). We further show that several smells frequently co-occur, particularly Context Bloat, Skill Leakage, and Conflicting Instructions.

Large Language Models (LLMs) have shown promise for reasoning over Text-Attributed Graphs (TAGs). However, applying LLMs to graphs requires linearizing their structure into sequences, introducing distortion rooted in the graph bandwidth problem. While this distortion has been shown to degrade performance, it is often attributed to prompt design or model scale, leaving the underlying mechanism unclear. In this work, we show \textit{how} rotary positional embeddings turn graph linearization into bandwidth-dependent attention decay, suppressing attention between graph-adjacent nodes that are forced far apart in the serialized sequence. This shifts the focus of LLM-based graph reasoning from prompt engineering and scaling toward correcting attention misalignment. Motivated by this analysis, we propose \textbf{G}raph-\textbf{a}ligned \textbf{L}anguage \textbf{A}ttention (\textbf{GaLA}), a lightweight, inference-time modification for LLMs. GaLA biases attention toward graph-adjacent nodes while preserving the LLM's sequential inductive biases. Across TAG benchmarks, GaLA improves performance with negligible overhead, demonstrating that distortion is a correctable bottleneck in LLM-based graph reasoning.

Deep research agents are increasingly evaluated on their ability to search for evidence, reason over retrieved sources, and produce grounded answers. Existing browsing benchmarks, however, largely assume that the user's query and the supporting evidence are written in the same language, leaving open whether agentic search systems can operate when relevant evidence appears in another language. We introduce XBCP (Cross-lingual BrowseComp-Plus), a controlled benchmark that preserves the English question-and-answer space of BrowseComp-Plus but varies the languages of the supporting documents. XBCP instantiates two complementary settings: in the cross-lingual setting, each query is paired with evidence in a single assigned language. In the multilingual setting, the full evidence corpus is distributed equally and randomly across 12 languages spanning high-resource and low-resource regimes. We evaluate four deep research agents using sparse and dense multilingual retrievers, measuring answer accuracy, evidence recall, search behavior, calibration, citation fidelity, and oracle retrieval. Results reveal substantial degradation when evidence is translated. Even strong, dense retrievers lose evidence recall, and agents become less calibrated and cite evidence less reliably. Notably, accuracy remains lower even when all gold evidence is supplied directly. These findings suggest that cross-lingual deep research exposes both retrieval failures and an independent, agent-side difficulty in integrating language-mismatched evidence.

SDE-based generative models, including diffusion models and the Schrödinger bridge, have found broad applications in signal processing tasks such as speech enhancement, image restoration, and time-series generation. This note presents a modeling framework for such models within the context of stochastic thermodynamics. The main results of this note are trajectory-level definitions of work, heat, and entropy production, along with a generalized Jarzynski identity and a second-law-like inequality. The proposed framework extends the original Jarzynski setup to accommodate time-dependent bath temperature and nonconservative driving forces. This thermodynamic perspective may deepen our understanding of diffusion models and the Schrödinger bridge from a nonequilibrium statistical mechanics viewpoint.

Millions of individuals worldwide suffer from sensory and communication deficits caused by neurodegenerative diseases, stroke, or trauma. Brain-computer interfaces (BCIs) offer a promising avenue for sensory and motor restoration. However, the scientific literature remains highly fragmented between invasive neuroprosthetics and non-invasive electrophysiological decoders, with a lack of consistent terminology and comparison metrics. This chapter proposes a unified 2 x 2 framework categorizing BCIs along two axes: degree of invasiveness (invasive vs. non-invasive) and signal direction (afferent sensory-IN vs. efferent sensory-OUT). We define and distinguish the paradigms of restoration, substitution, and augmentation. Furthermore, we outline a structural roadmap for the convergence of these modalities over near-, medium-, and long-term horizons, focusing on physical limits and the integrative role of machine learning foundation models.

A model can learn that the piano piece Für Elise is calm and reflective by listening to the audio or by reading a text description, but does it matter which route that knowledge took when it is later at risk of being forgotten? Forgetting research in multimodal models measures what knowledge is lost under adaptation, yet has not asked whether acquisition route affects how easily that knowledge is forgotten. We call this untested premise the Pathway-Invariant Assumption. Music understanding enables a clean test because a music clip and a canonical text description can be aligned to the same perceptual content, allowing the same knowledge unit to enter a model through listening or reading while the target remains fixed. Across multiple architecturally distinct audio-language models, we observe a consistent asymmetry: text-pathway knowledge is forgotten more than matched audio-pathway knowledge under identical adaptation pressure. To attribute this effect to route rather than confounds, we introduce the Paired Pathway Controlled Protocol (PPCP), a three-phase design that establishes matched pathway baselines, activates both pathways under symmetric supervision on the same knowledge pool, and applies identical forgetting pressure to both pathways. The gap is stable across models and gain-controlled analyses, persists when contradictory overwrite is replaced by correct-label cross-domain learning, remains under single-modality pressure, and is not removed by lightweight replay. Two independent routing-depth controls confirm that the effect is not explained by architectural depth, pointing to input representation as the dominant factor. Under PPCP, our results demonstrate that forgetting is highly route-dependent, establishing acquisition route as a new analytical dimension for forgetting research and multimodal system design.

In a previous work, the topology of inversion-symmetric one-dimensional periodic media was characterized through the pole-zero pattern of an impedance-like function associated with Bloch waves. This construction reproduces the Berry--Zak invariant and provides a criterion for topological interface states. In the present work, we give a geometric interpretation of this formalism. We show that poles and zeros arise naturally from the action of inversion symmetry on the projectivized space of Cauchy data. The corresponding Dirichlet and Neumann states are identified with the two fixed points of the induced $\mathbb Z_2$ action on the Riemann sphere. The key observation is that Bloch eigenvectors are naturally constructed on the universal covering of the Brillouin circle. The topology of the associated Real eigenline bundle is encoded in the action of the deck transformation group on lifted eigenvectors. This action is described by a monodromy sign $ρ\in\{\pm1\}$, determined by the inversion representations carried by the band at the fixed points of the Brillouin zone. We show that this monodromy defines a natural rank-one local system over the Brillouin circle. The corresponding Real line bundle is classified by its first Stiefel--Whitney class, which coincides with the associated $\mathbb Z_2$ pole-zero invariant. This establishes a geometric connection between the pole-zero formalism, Berry--Zak phases, Real bundles and local coefficient systems.

The ability to compute electron-phonon coupling from first principles, using density functional perturbation theory and interpolation techniques, has enabled predictive calculations of electronic transport coefficients in crystalline materials. However, these methods are still computationally expensive. Here we present two inexpensive methods to obtain thermoelectric transport properties of semiconductors using a small number of electron-phonon matrix elements calculated from first principles. The first method combines models for coupling of electrons with different phonon modes whose parameters are obtained from $\sim 10$ matrix elements per electronic band and phonon mode calculated from first principles. Within this method, we formulate the acoustic deformation potential model for arbitrary crystal symmetries and band extrema locations. The second method uses machine learning to interpolate $\sim 100$ electron-phonon matrix elements per electronic band and phonon mode on dense reciprocal space grids in the parts of the Brillouin zone relevant for transport. We apply both methods to two-dimensional MoS$_2$ and show very good agreement with the state-of-the-art method. The calculated thermoelectric properties also agree well with experiments. We find that the machine-learning method is more accurate and straightforward to implement compared to the model approach.

The structure of a porous material, and in particular its spatial variability, is known to control the intrinsic permeability of the system. We investigate how dead-end pores influence the intrinsic permeability of a porous medium beyond their contribution to total pore volume. Dead-end pores are ubiquitous in porous media, yet they are often treated as hydraulically inactive regions whose influence is assumed to be negligible or absorbed into effective-porosity descriptions. We perform pore-scale flow simulations across different dead-end pore structures, including heterogeneous arrangements, controlled granular assemblies, and a minimal single-channel model to study their impact on the system macroscopic permeability. This strategy allows us to isolate the effects of dead-end pore density, depth, and orientation while preserving the transmitting network. We find that dead-end pores can influence intrinsic permeability: increasing the density of dead-end pores along percolating flow paths enhances permeability, whereas pore depth and junction orientation have negligible effects. The observed permeability enhancement originates from localized hydrodynamic interactions at junctions between transmitting and dead-end pores. Based on these results, we propose an effective formulation that relates the density and spatial organization of dead-end pores relative to the transmitting network to macroscopic permeability. Our findings show that dead-end pore architecture provides an additional geometric control on intrinsic permeability beyond porosity and pore-size statistics.

Sphere packings in circular cylinders have attracted substantial research interest, among which the discovery of chiral helical structures is the most iconic. However, recent experimental results on zebrafish do not match the known packing structures in circular cylinders. To account for the inherent imperfections of biological tubes, we take elliptic cylinders as the canonical deformation of circular cylinders and investigate the densest packings of hard spheres in them using simulation, theory, and experiments. Starting from the chiral structures in circular cylinders, we demonstrate that even a weak cross-sectional deformation can trigger entirely new phases, including ones that either eliminate global chirality or significantly complicate the chiral structures. This reveals the significant effect of cylindrical anisotropy. The new helical phases under anisotropic confinement remain chiral and develop hierarchical periodic structures, which are difficult to obtain by simulations but are predicted by our newly developed theory for helical phases in elliptic cylinders. The theory also predicts double oscillated-chain phases without chirality, which perfectly match the simulations. Our work offers fresh insights into understanding packings in anisotropic cylinders, which will help researchers to design new materials and to understand many living systems.

Establishing a versatile, fast and reliable magnetic memory technology is a giant bottleneck for cryogenic computing since present-day room-temperature solutions either cease to work or consume too much power. The long-term goal of superconducting spintronics has been to overcome this bottleneck by generating magnetic memories with equal-spin triplet supercurrent driven through them to control their magnetization direction. This path has been hampered by the short spin relaxation length and strong anisotropy in ferromagnets. Here we show how the supercurrent driven generation of spin accumulation in a superconductor/magnetic insulator bilayer, together with Gilbert damping of magnetization lead to a motion of magnetic domain walls. This manifests as a local voltage across the wall, which allows its position to be identified. Associated with this voltage and the current, there is Joule power which is dissipated via the Gilbert damping. The power required to maintain domain wall motion is orders of magnitude smaller than in the normal state, where most of the power is wasted in producing the current.

Red blood cells (RBCs) are essential for oxygen transport, and their remarkable ability to undergo significant deformations during flow is a crucial feature for their physiological function. At intermediate shear rates typical of the microcirculation, RBCs can adopt complex, multi-lobed shapes, signifying a dynamic instability. Here we adopt a perturbative theoretical framework of a quasi-spherical RBC under external shear flow to study such shape instabilities. To better capture RBC maturation and enucleation, we first extend the framework to explicitly account for different excess areas between the stress-free and current membrane shapes. We revisit the reduced equations of motion obtained for an ellipsoidally-shaped RBC, and demonstrate the effect of different excess areas and initial orientation on the dynamical trajectories. Then, we introduce additional spatial modes and show that an emerging instability critically depends on the RBC's shear and bending moduli, the internal to external viscosity ratio, and the excess area, mainly through the RBC's membrane tension. We also study the instability-induced saturation of the membrane tension, and the resulting excess area redistribution at long times. The theoretical framework and the emerging picture of the different instabilities provide insights into the emergence of stomatocyte and trilobe shapes exhibited by RBCs under external flow.

Large-deviation tails of critical probability distributions provide a sensitive probe of universality beyond standard finite-size scaling. We study these tails for critical percolation and Fortuin--Kasteleyn Ising models on two-dimensional lattices, three-dimensional lattices, and complete graphs. We consider two rescaled order parameters: the magnetization-like variable $x_m=|M|/\langle |M|\rangle$, including a signed cluster-mass analogue for percolation, and the largest-cluster variable $x_C=C_1/\langle C_1\rangle$. For $x_m$, we test the expected stretched-exponential large-deviation tail and show that the same form applies to the percolation analogue. For $x_C$, guided by the exact complete-graph result and scaling arguments, we propose universal scaling forms for both tails of the cumulative distribution and test them by extensive Monte Carlo simulations. In the complete-graph FK-Ising model, the left tail is governed by rare configurations with percolation-like scaling rather than by the typical Ising scaling. Our results show that the tails of order-parameter distributions reveal universal features of critical fluctuations that are not captured by averaged observables alone.

Quantum transport in mesoscopic systems is conventionally formulated within the Landauer--Büttiker scattering framework, where steady-state currents emerge from the transmission of plane waves representing propagating carriers. While highly successful, this description obscures the explicit time-domain dynamics of individual fermionic excitations and their role in establishing macroscopic transport. Here, we present an exact and self-contained time-domain construction of Landauer transport based on a discrete basis of orthogonal fermionic wave packets. Starting from a second-quantized formulation, we define packet creation operators via a continuous Fourier transform over a finite transport energy window. By encoding the Pauli exclusion principle, which enforces a fundamental temporal spacing $Δt = h/eV$, the current is reproduced in terms of orthogonal wave packets that are used for the rigorous construction of the many-body fermionic state. In this representation, a noiseless current emerges as a deterministic sequence of charge-carrying events, yielding the Landauer conductance $G_0 = e^2/h$ without invoking momentum-space kinematics. We further demonstrate that this construction remains exact for arbitrary energy dispersion. Additionally, the underlying Fock space decomposition into finite disjoint energy sub-bands renders the numerical approach highly scalable for high performance computing platforms. Our results establish a direct and rigorous bridge between the continuous scattering description of quantum transport and a discrete, time-resolved picture based on fermionic wave packets.

Out-of-time-ordered correlators (OTOCs) have emerged as a diagnostic of information scrambling and quantum chaos in many-body systems. We investigate the imprints of chaos in the dynamics of OTOCs in the Heisenberg spin-$1/2$ chain with random fields. The system is parameterized to exhibit a crossover from integrable to chaotic dynamics. We demonstrate numerically that the approach to saturation of the OTOC can distinguish between integrable and chaotic regimes, with a power-law $(1/t)$ relaxation for integrable systems and a higher-degree power-law decay $(1/t^α; α\ge 1)$ followed by an exponential relaxation for the chaotic regime. We further show that long-range spectral statistics, such as the number variance, are more effective in characterizing quantum chaos in the regime near saturation of OTOC. We also demonstrate that the relaxation and initial scrambling regimes exhibit distinct and universal features, with the former being sensitive and the latter being robust against different realizations of random-fields. The long-time saturation of OTOC also fluctuates with different realizations, and its exact expression is derived through the Eigenstate Thermalization Hypothesis.

Electronic correlations in graphene are generally considered weak due to the large bandwidth of its $π$ electrons. Here we show that tensile expansion of the honeycomb lattice provides a direct route to enhancing correlation effects. Using variational and diffusion quantum Monte Carlo, we compare a conventional Jastrow-Slater determinant wave function with a resonating-valence-bond (RVB) Jastrow-antisymmetrized geminal product ansatz for a series of stretched graphene lattices. We find that the energy gain of the RVB state relative to the single-determinant description increases with bond expansion up to a critical strain $δ_{\mathrm{cr}}$, and decreases beyond it, revealing a nonmonotonic evolution of electronic correlations. The crossover is found to occur in the range $15\% < δ_{\mathrm{cr}} < 20\%$, in agreement with mechanical stability limits. This behavior indicates a transition from a weakly correlated Dirac semimetal to a regime with enhanced non-dynamic correlation and short-range singlet pairing. Our results provide direct many-body evidence that lattice expansion drives graphene into a regime where RVB-like correlations become energetically favorable, offering a simple route to tuning correlation effects in Dirac materials.

Altermagnets are a newly identified class of magnetic materials that combine vanishing net magnetization within the unit cell with spin-split electronic states. Their theoretical description relies on symmetry properties of the bulk band structure. Surfaces and thin films, however, inherently break these symmetries. Here, we investigate the consequences of such symmetry reduction for the electronic structure of bulk altermagnets near the surface and of thin films. When the surface coincides with a symmetry plane of the bulk altermagnetic order, the resulting two-dimensional Brillouin zone exhibits spin-degenerate bands, corresponding to conventional antiferromagnetic behavior. In all other cases, the symmetry of the altermagnetic order is reduced, leading to modified spin splitting. Remarkably, we discover a thin-film geometry of a $g$-wave altermagnet with a particular surface orientation that enables a $d$-wave spin splitting, which is commonly accompanied by the spin-splitter effect, suggesting the functionalization of non-$d$-wave altermagnets by surfaces. Our findings demonstrate that symmetry breaking at surfaces and in thin films fundamentally reshapes altermagnetic spin textures, providing a tunable platform for controlling spin-dependent electronic phenomena.

We present an analytical investigation of the extrinsic contributions to the anomalous Hall conductivity in the context of the half-quantized Hall effect observed in disordered parity anomalous semimetal emerged from semi-magnetic topological insulator thin films. The gapless Dirac cone surface state, which embodies the quintessence of the half-quantized Hall effect, exhibits remarkable robustness against disorder scattering. Two primary extrinsic mechanisms, the side-jump and skew-scattering, are deemed irrelevant and make no contributions. These results establish the parity anomalous semimetal as a disorder-resilient quantum phase, thereby providing insights into Dirac fermion physics.

Colloidal quantum dots that simultaneously offer room-temperature single-photon purity and high photoluminescence quantum yield are sought for quantum optics, but remain elusive in environmentally benign materials. We introduce a thermodynamic-kinetic decoupling strategy that transforms defect-tolerant CuInS2 quantum dots into bright, narrowband, and photostable single-photon emitters. Zn2+ alloying strains the lattice, thermodynamically suppressing native copper vacancies and narrowing the emission from a broad defect band of approximately 300 meV to an excitonic line of approximately 120 meV. Ga3+ incorporation then kinetically pins the cation sublattice against Cu+ migration, preventing defect regeneration during ZnS shell growth. The resulting Cd-free core/shell dots achieve near-unity quantum yield of approximately 98% while retaining narrow excitonic emission. Critically, room-temperature single-dot spectroscopy reveals homogeneous linewidths as low as approximately 58 meV, strongly suppressed blinking, and high-purity single-photon emission with g2(0) = 0.06. This stabilized excitonic emission directly reduces reabsorption losses in luminescent solar concentrators, yielding an external optical efficiency of 12.68%. Our work establishes a generalizable framework to unlock intrinsic excitonic photophysics in ion-mobile, defect-prone semiconductors, opening a viable path toward high-performance heavy-metal-free emitters for quantum light sources.

The coexistence of superconductivity and finite electric fields may enable access to intriguing forms of electronic states. We demonstrate the emergence of an intermediate state in which electric fields penetrate the system while superconductivity still persists. Our measurements reveal a nonclassical regime characterized by the simultaneous presence of supercurrent and dissipative charge transport. This state, realized in a pristine unpatterned three-dimensional system, arises from electric-field-driven order parameter fluctuations. It provides a platform to explore dissipative states of charged quantum fluids far from equilibrium.

Classical nucleation theory (CNT) describes the formation of a stable phase from a metastable one. In equilibrium systems, it quantifies the free-energy competition between a favorable bulk gain and an unfavorable interfacial cost. For systems without detailed balance, the corresponding nonequilibrium nucleation theory (NNT) was so far developed only for cases with a conserved order parameter, such as active fluid-fluid phase separation. Here we construct the NNT for systems with a (single, scalar) nonconserved order parameter. Unlike in the conserved case, the nucleation barrier controlling (noise-driven) droplet growth is profoundly altered by deviations in the interfacial density profile from the one arising during (deterministic) droplet relaxation. The barrier can nonetheless be analysed by carefully defining the reaction coordinate (droplet radius) to project out those deviations. We give explicit NNT predictions for models drawn from population dynamics and active matter, finding excellent agreement with numerical studies.

We investigate equilibrium and non-equilibrium thermodynamics of one-dimensional photonic diamond lattices with Kerr nonlinearity. The equilibrium phase diagram is obtained as a function of the synthetic magnetic flux acting on each plaquette. In the linear regime, the magnetic flux can induce Aharonov-Bohm caging, flattening all Bloch bands and suppressing particle and energy currents. In this caging regime, non-vanishing currents are enabled by nonlinearity. By imposing stationary temperature- and chemical potential- imbalances at the system boundaries, we show that at weak nonlinearity fine tuning the flux at the Aharonov-Bohm caging transforms the system from a conductor to an insulator. For intermediate nonlinear strength, the system remains conducting for all magnetic fluxes; however, the caging condition significantly enhances the Seebeck coefficient and thermoelectric figure of merit, improving the thermoelectric features of the system. Our results give evidence of a novel route towards optimization of coupled transport devices, based on the control of linear versus nonlinear conduction channels via a synthetic magnetic flux.

The modern theory of orbital magnetization successfully describes the response of Bloch electrons to magnetic fields in fully periodic crystals, but it does not directly address the distinct regime of an in-plane field in multilayer systems with layer hybridization. Coherent interlayer tunneling allows electrons to form circulating current loops, producing an in-plane orbital response that is absent in a strictly two-dimensional limit and qualitatively different from the conventional three-dimensional one. Here we develop a theory of in-plane orbital magnetization for this {\it transdimensional} regime, where the layer thickness is comparable to the vertical mean free path. Starting from the current-loop picture, we construct the in-plane orbital angular momentum operator and derive exact expressions for the orbital magnetic moment and the in-plane orbital magnetic susceptibility. As an application, we predict a gate-tunable in-plane orbital magnetoelectric effect in layered materials. Our framework establishes a general foundation for in-plane orbital responses and suggests new opportunities for orbitronics in layer-hybridized quantum materials.

The origin of some mesophases of the ferroelectric nematic realm is not yet well understood. In this work we study the highly polar liquid crystal MIO, a close structural analogue of the prototypical ferroelectric nematogen DIO, which exhibits a ferroelectric smectic A to ferroelectric smectic C (SmAF-SmCF) phase transition. Calorimetric, dielectric and light-scattering experiments reveal that it is a second-order phase transition with mean-field behavior, and is driven by the softening of the tilt elastic constant accompanied by the divergence of the amplitude of the associated dielectric mode.

Characterizing electronic band structures requires precise knowledge of wave functions and their quantum geometry. Here, we introduce a tomography method to reconstruct the density matrix of electron states from quasiparticle interference maps around single impurities. We consider two-orbital models on a honeycomb lattice, relevant to graphene heterostructures and direct-gap semiconductors. For on-site impurities, backscattering between time-reversed states directly maps the density matrix populations and coherences into distinct orbital contributions in the interference map. While local probes usually lack orbital resolution, these orbital contributions transform under distinct symmetry group representations and can thus be disentangled to reveal the density matrix and quantum geometric tensor of the scattering states. This establishes impurities as tomographic probes for band structures in scanning tunneling microscopy using conventional, unpolarized tips.

Hyperuniform systems strongly suppress long-wavelength density fluctuations, which is quantitatively characterized by the small-wavenumber scaling. In finite samples, however, accurately estimating the hyperuniformity exponent α can be challenging. The inferred value depends strongly on the range of length scales accessible in the measurement, finite-size effects, and the specific characterization method employed, whether based on Fourier-space structure factors, real-space density fluctuations, or dynamical probes such as diffusion spreadability. In particular, the structure-factor method provides the most direct estimate of α, but is sensitive to empirical low-k fitting cutoffs. The number-variance method offers a real-space Class-like diagnosis, but contributes a numerical exponent only when the finite-size data retain Class III-like scaling information. The spreadability method provides a smoother dynamic estimate and reduces configuration-level fluctuations, but requires a physically admissible long-time fitting window. Here, we develop a practical method-aware protocol for robust estimation of the effective scaling exponent α in finite-size hyperuniform point configurations, combining three complementary methods with distinct roles. Our protocol summarizes the method-specific estimates through a joint empirical estimator and reports the internal dispersion among the participating methods to determine the optimal estimate.

We propose a physical realization of the unstructured database search that works via classical, dissipative model of spherical spins. The database is implemented via spin-spin couplings, where the selected coupling refers to a larger ferromagnetic interaction between two selected spins. The low-temperature equilibrium of this model leads to magnetization strongly concentrated on the selected spins, which means that the search is complete. The search time refers to the relaxation time to equilibrium from a homogeneous initial state, and is described via Langevin equations. This time scales as ${\cal O}(M^a)$ with $a<1/2$, where $M$ is the database volume. This is faster than Grover's search, showing how a dissipative, classical analog computer can overcome the quantum unitary computer.

We report an experimental investigation of the magnetic microstructure of iron oxide multicore assemblies by means of polarized small-angle neutron scattering (SANS). Guided by a recently developed analytical theory for vortex-state magnetic nanoparticles, we provide a quantitative comparison between the measured and calculated cross sections, revealing signatures that are consistent with vortex-type magnetization configurations at low applied magnetic fields. In particular, the field evolution and the characteristic isotropic ring-type feature of the spin-flip scattering intensity at intermediate momentum transfers are in line with the formation of flux-closure states. The latter are stabilized by the interplay of exchange, Zeeman, and magnetostatic energies. The methodology allows for a statistically significant characterization of vortex states in densely packed nanoparticle systems, thereby complementing surface-sensitive techniques that are commonly limited to the observation of spin structures in individual particles.

We show that the broad X-ray photoelectron lines of silicon oxide on silicon arise from a continuous statistical distribution of core-level binding energies. Statistical simulations spanning compositions from Si to SiO$_2$ reproduce the full extent of this broadening, reaching 5 eV for SiO$_{1.0}$ , in quantitative agreement with 0.23 nm layer-resolved spectra reconstructed from Ar$^+$ sputtering data. This continuous distribution blurs distinct spectral fingerprints of local structural motifs, thereby challenging conventional chemical state assignment in oxide X-ray photoelectron spectra.

The self-averaging of the replica overlap is proven in the Edwards-Anderson (EA) model under random field almost everywhere in the coupling constant space in any dimension. The EA order parameter is represented in terms of the derivative of the free energy density with respect to the random field strength, regardless of boundary conditions. Tasaki's correlation inequality for finite-dimensional spin glass models shows that the expectation of the squared replica overlap is bounded by the squared EA order parameter. These simple evaluations enable us to prove that the variance of the replica overlap vanishes in the infinite-volume limit. The self-averaging of the replica bond overlap is proven also in the EA model with Gaussian exchange interaction without random field. Short-range spin glass models have been shown to behave differently from mean-field spin glass models with RSB phase.

We present a method to compute Ewald summation for the irreducible components of flow around active particles to study hydrodynamic interactions in active colloidal suspensions. An active particle is modeled as a colloidal sphere with a surface slip velocity. Using this model, we obtain an irreducible representation of the fluid flow produced by an active particle in periodic geometry of Stokes flow for an arbitrary surface slip. The solution of the active flow is obtained in terms of lattice sum of the Oseen tensor and their derivatives. The lattice sum is accelerated using the Ewald summation technique. We apply the method to compute explicit expression for rigid body motion of hydrodynamically interacting active particles. Our method presents a way for dynamic simulation of active particles due to arbitrary mode of active slip in periodic geometry of Stokes flow.

Antiferromagnetic skyrmions are swirling topological spin textures with rich dynamics and intriguing transport properties, yet their bidirectional dynamics remain largely unexplored. Here, we investigate the dynamics of antiferromagnetic skyrmions driven by current-induced spin-transfer and spin-orbit torques. We computationally demonstrate that antiferromagnetic skyrmions moving in one direction at low current densities can reverse their motion direction when the driving current is above a threshold. Based on the Thiele approach analysis, we show that this bidirectional motion originates from a change in the relative strengths of two effective forces arising from spin-transfer and spin-orbit torques. Furthermore, exploiting this bidirectional motion on a single racetrack, we design programmable logic gates. Our results not only uncover a hidden mechanism for bidirectional skyrmion motion but also facilitate the development of antiferromagnet-based logic devices.

The realization of Coulomb coupled electron-hole double layers in 2D semiconductor heterostructures has enabled the thermodynamic and transport studies of equilibrium exciton fluids without a magnetic field. By doping the exciton fluid with additional electrons/holes, an equilibrium fluid of trions - three particle bound states of electrons and holes - further emerge, providing the platform to explore new transport phenomena associated with such composite particles. Here we report the observation of a Hall effect for trions in MoSe2/WSe2 heterobilayers, which support Coulomb-coupled electron and hole fluids with tunable densities. The Hall effect arises from a Lorentz force on trions under a perpendicular magnetic field. It is manifested in both Hall drag measurements and standard Hall effect measurements on just one of the semiconductor layers. For negatively charged trions, an electron Hall effect is observed even in a hole doped WSe2 monolayer due to the presence of trion drags. The trion Hall effect also disappears when the trions are ionized at elevated temperatures and/or high trion densities. Our work opens the door for realizing quantum oscillations and the quantum Hall effect for trions.

We present a holographic dual of a two dimensional conformal field theory with non-hermitian but Parity-Time (PT) symmetric boundary conditions, by applying the AdS/BCFT duality and by introducing an imaginary valued scalar field localized on an end-of-the-world brane. We find that as we increase the strength of the non-hermitian PT symmetric interactions, the system experiences a spontaneous PT symmetry breaking. We also consider its Wick rotated setup as a new quantum quenched state and show that its growth of entanglement entropy can be larger than the standard results obtained from standard Cardy states.

By assuming reversible H-bonding association and dissociation, this work provides a description of the supramolecular structure and dynamics of monohydroxy alcohols (MAs) within the framework of a recently proposed living chain model (LCM). Structurally, reversible H-bonding leads to a single exponential distribution of the molar concentration of the supramolecular chain with length N. Dynamically, reversible H-bonding enables supramolecular chain breakage and recombination, which modifies the relaxation time of the supramolecular chains. In addition to the structural relaxation, tau_a, and the Debye relaxation, tau_D, two other relaxation times are revealed: the chain breakage time, tau_B, and the H-bonding lifetime, tau_H. The interplay among these four-time scales defines five distinct dynamics regimes. In Regimes I and V, no supramolecular chains form. In Regimes II and IV, supramolecular chains form and give a Debye relaxation. The characteristic chain length scales as Nc~tau_D/tau_a. In these two regimes, the H-bonding lifetime controls the Debye process. In Regime III, large supramolecular chains form. In all regimes with supramolecular chain formation, the Debye relaxation comes from the overall chain end-to-end dipole reorientation and scales with Nc. Excellent agreements between experiments and LCM have been observed, leading to quantitative descriptions of the dielectric and linear viscoelastic properties of MAs. These results thus establish a theoretical framework linking reversible H-bonding interactions to supramolecular structures, dynamics, and macroscopic properties of MAs.

Terahertz radiation from $\text{Bi}_2\text{Sr}_2\text{CaCu}_2\text{O}_{8+δ}$ intrinsic Josephson junctions (IJJs) provides an ideal platform to study the synchronization of a macroscopic quantum system. Here, we present a spectral analysis of a frequency-modulated Josephson plasma emitter coupled with patch antennas. In the unmodulated intensity distribution as a function of radiation frequency $I_{\mathrm{UM}}(ω)$, we observe a double Gaussian peak structure. Crucially, double-peak spectra obtained at a constant bias voltage imply either a rapid temporal distribution of resonances or their simultaneous excitation, driven by the mutual electromagnetic coupling between the IJJ mesa and the antennas. At low modulation frequencies $f_m$, the spectra are well reproduced by the products of $I_{\mathrm{UM}}(ω)$ and frequency combs, yielding a synchronized relaxation time $τ_s \simeq 0.28\text{ ns}$. Incorporating $τ_s$ quantitatively reproduces a drastic spectral transformation observed around $f_m \sim 1\text{ GHz}$, unveiling the sub-nanosecond non-equilibrium dynamics of coupled Josephson plasma.

The macroscopic instability of soft materials undergoing extreme deformations is traditionally viewed as a pure structural or mechanical failure. Driven by the quest to uncover universal principles across disparate physical systems, we bridge two vibrant yet seemingly disconnected research frontiers: macroscopic finite-strain solid mechanics and quantum-like topological physics. Here, we demonstrate that the classical elastic surface instability of a deformed hyperelastic manifold is not merely a mechanical bifurcation, but fundamentally a topological phase transition. By incorporating Lie group metric evolution into a generalized Stroh formalism, we map the highly nonlinear geometric frustration onto an algebraic surface impedance matrix $\mathbf{H}$. For a semi-infinite hyperelastic half-space under finite compression, we analytically map the system to a one-dimensional Dirac Hamiltonian, where the macroscopic mechanical stretch acts as a tunable knob for the Dirac mass. We reveal that the onset of surface wrinkles marks a topological transition from a trivial to a non-trivial phase characterized by a quantized step in the winding number, naturally giving rise to a robust, macroscopically localized zero-energy edge state. This fundamental linkage unifies macroscopic symmetry breaking with the topological paradigm, opening a new theoretical pathway for programmable smart soft matter.

Spin dynamics in ordered magnets with topological lattice defects is investigated. Using fracton--elasticity duality, we develop an effective field theory of magnons coupled to quantized lattice dislocations (dislons) in magnetic insulators. Within this framework, an elastic gauge field mediates a nonlocal interaction between dislocations and magnetization gradients. The resulting magnetoelastic coupling gives rise to coherent magnon-dislon hybridization whose properties are dictated by dislocation topology. Screw dislocations exhibit helicity-selective hybridization and symmetry-protected dark dislon sectors, while edge dislocations generate anisotropic hybrid excitations with finite spin-precession ellipticity through the glide constraint. Our results establish dislocations as dynamical topological defects with directly observable polarization fingerprints in magnon spectra, and reveal magnon-dislon hybridization as a new route to control spin dynamics.

Twisting layered materials produces moiré superlattices, but prescribed twist angles are usually obtained by demanding assembly procedures. Here we show that montmorillonite, an abundant swelling clay, forms tunable moiré superlattices naturally. Focal-series high-resolution transmission electron microscopy, geometric phase analysis, and molecular dynamics simulation reveal that its apparent rotational disorder is biased toward low-angle misorientations inherited from discrete hydration states. Multilayer stacks preferentially adopt twists near 1-2°, 4°, and 10°, producing long-wavelength moirés without long-range rotational order. We define this kinetically trapped state as a moiré glass, distinct from featureless turbostratic stacking. Simulations indicate that lattice-charge disorder stabilizes the angular preferences, whereas charge ordering promotes random stacking. Hydration screens interlayer interactions and lubricates twist, while dehydration arrests the resulting configurations in discrete steps. These results establish dynamic hydration as a macroscopic handle for programming twist in layered matter.

In a magnetic field bilayer graphene hosts an octet of pseudo-zero-mode electron levels nearly degenerate in orbitals $n=(0,1)$, valleys and spins. They split in valleys by electrostatic gating. In gated bilayer graphene, in which the interlayer bias is set up to flip sign across a line, one has a line junction that traps a portion of pseudo-zero-mode electrons inside the insulating bulk band gap, giving rise to electron states localized along the junction, known as kink states. A close look is made into the spectra and electromagnetic response of such kink states. There are two species of valley current associated with them, a drift current driven by the bias gradient and a circulating current coming from cyclotron motion. It turns out that they both flow in essentially the same direction, with the circulating current exhibiting a magnetic character distinct from those of other higher levels. In equilibrium they spatially circulate within the kink states, creating a quasi-one-dimensional channel of orbital magnetization. The electric control of the orbital magnetization and valley currents via a network of gated junctions will find useful applications in valley electronics.

We develop a minimal hydrodynamic framework to describe exciton transport in the presence of an electron hole plasma in 2D semiconductors. Treating excitons, electrons, and holes as coupled fluids, we show that exciton diffusion is strongly renormalized by momentum exchange with the plasma. In the collisional regime, mutual diffusion leads to a nontrivial redistribution of transport coefficients but preserves the positivity of the exciton diffusivity. In contrast, when plasma inertia and collective charge oscillations are accounted for, the exciton diffusive mode hybridizes with acoustic plasma modes, giving rise to a dynamical instability manifested as an effective negative diffusion coefficient. We demonstrate that this instability originates from the nonequilibrium coupling between slow excitons and fast plasma degrees of freedom, rather than from nonlinear diffusion or thermodynamic effects. Our results provide a unified physical mechanism for negative exciton diffusivity reported in recent experiments and establish collective plasma dynamics as a key control parameter of exciton transport in 2D materials.

Quantum systems governed by unitary and reversible microscopic dynamics may nevertheless exhibit equilibration, in the sense that some effective description becomes time-independent. Standard equilibration results usually consider two separate situations: system-environment structures, in which the composite system evolves unitarily while the system of interest equilibrates, and restricted measurements, such as coarse-grained POVMs and observables, in which the measurement statistics equilibrate. Here, we bring these descriptions into a common state-level framework using the concept of generalized subsystems, where the accessible effective state appears as the output of a quantum channel acting on the microscopic state. We derive bounds showing that generalized subsystems equilibrate when their dimension is small compared with the effective dimension of the discarded microscopic information. We further show that this condition is met for typical initial states in large subspaces and that the resulting equilibrium description is largely insensitive to microscopic initial details. The framework recovers the usual equilibration bounds for ordinary subsystems and finite families of POVMs. As an example, we also introduce a finite-resolution energy channel that maps unresolved microscopic energy levels into effective energy levels, thereby making residual effective coherences explicit and showing how spectral multiplicities constrain those coherences while strengthening equilibration. Our results provide a unified state-level formulation of quantum equilibration under general forms of limited accessible information.

We study signatures of quantum chaos in finite-loop truncations of the planar dilatation operator in the $\mathfrak{su}(2)$ sector of $\mathcal N=4$ super Yang-Mills and its $β$-deformation. These truncations define holographically motivated long-range deformations of the nearest-neighbour XXX spin chain. At one-loop the model is integrable, while the all-loop planar theory is expected to again be integrable. Finite-loop truncations therefore provide a natural setting for investigating how chaotic behaviour emerges between these two integrable limits. We analyse this question using spectral statistics, eigenvector diagnostics and spread complexity. We find that the two- and four-loop truncations develop GOE-like level statistics at sufficiently large coupling but with features characteristic of weak integrability breaking. The integrability breaking at four-loops is weaker than at two-loops and the critical coupling at which chaos occurs is larger, at least for long spin chains. The three-loop truncation does not show the same onset of chaos in the range studied. Eigenvector diagnostics show that the corresponding eigenstates remain less random than GOE vectors, indicating weak ergodicity and multifractality. Finally, we can identify signatures of the eigenvalue and eigenvector chaos in the Krylov-space data. Namely, we demonstrate a correlation of the level spacing statistics with the peak of spread complexity and disorder on the Krylov chain. The delocalisation of the initial state in the Hamiltonian eigenbasis is shown to strongly affect the saturation of complexity. Our results suggest that finite-loop dilatation operators are not generic long-range spin chain Hamiltonians, but already display patterns consistent with the restoration of integrability in the all-loop planar theory.

Spectral kissing, recently realized in a Kerr parametric oscillator (KPO), refers to the merging of pairs of energy levels and arises as a manifestation of an excited-state quantum phase transition (ESQPT). Here, we show that this phenomenon has a dissipative counterpart encoded in the spectrum of the steady-state density matrix. Using a dissipative KPO as a representative example, we demonstrate that, in the weak-dissipation regime, the eigenvalues of the steady-state density matrix organize into quasi-degenerate pairs that mirror the spectral kissing of the corresponding closed system. As the dissipation strength increases, this pairing gradually disappears. By analyzing the classical limit of the system, we derive analytical expressions for the critical lines governing both the onset of steady-state spectral kissing and its disappearance at a dissipative phase transition.

We address the quantum geometric structure of plasmons in Fermi surfaces enclosing a topological charge. We demonstrate that Weyl fermion plasmons have monopole structure, are topological and have a finite vorticity $ζ=2\mathsf{C}_{\text{w}}$, where $\mathsf{C}_{\text{w}}$ is the Chern number of the Fermi surface enclosing the Weyl point. We show that these plasmons selectively couple to light linearly polarized along the plasmon effective dipole moment $\mathbf{d}$, which has quantum geometric origin and points along the direction of the plasmon center of mass momentum $\hat{\mathbf{Q}}$. We suggest that Weyl metal topological plasmons have distinctive optical properties compared to conventional plasmons.

The effect of a large environment on a finite-size quantum mechanical system is two-fold: It brings dissipation, but also fluctuations of thermal and quantum origin. While dissipation tends to stabilize the dynamics, we question if and how environmental quantum fluctuations can generate chaos in an otherwise classically non-chaotic system. We work out a paradigmatic model of quantum optics: the dissipative Dicke model, where a large spin interacts with a dissipative harmonic mode. We dial in the classical/quantum correspondence by working in the semiclassical regime at large but finite spin. We demonstrate that, starting from a classically regular phase space in the superradiant regime, quantum noise can generate a strange attractor with fractal dimension and a positive Lyapunov exponent. We unveil the deep connection with shear-induced chaos that was recently developed in the mathematical community.

The ground state of random Hamiltonians with all-to-all interactions such as the quantum Sherrington-Kirkpatrick (SK) model and the Sachdev-Ye-Kitaev (SYK) model follow volume-law entanglement and are expected to be hard to model using tensor networks. In recent years, some progress has been made to push the limit of classical methods using neural quantum states. However, it remains an open question whether there exist quantum algorithms that could offer a quantum advantage over the state-of-the-art classical methods in simulating random Hamiltonians. In this work, we show that one such algorithm, TETRIS-ADAPT-VQE, can construct accurate ground states for dense and sparse SYK models containing up to $N=20$ Majorana fermions achieving fidelities $\geq 99.3\%$ and for the quantum SK model with up to $L=18$ sites achieving fidelities $\geq 99.9998\%$. We find that while the preparation of ground states is efficient (in terms of operator pool size and circuit depth) for the SK model, it is not efficient for either dense or moderately sparse SYK models.

We analyze the transport properties of quasiparticles locally excited at an initial time moment in several exactly solvable quantum models. It is revealed that, in the investigated quantum systems, the time-dependent probability distribution function (PDF) exhibits behavior similar to that of classical continuous-time random walk (CTRW) models, such as Lévy walks or diffusing diffusivity.

Chain-of-thought generation can turn a multi-step computation into a sequence of locally checkable state updates, but the training dynamics by which transformers acquire such updates remain poorly understood. We study this question in a solvable setting: a simplified one-block transformer trained by supervised next-token prediction on state sequences generated by composing permutations. The architecture separates fixed-lag action retrieval, learned by RoPE attention, from a specialized MLP logic module that applies the retrieved permutation to the current state. Using a statistical-physics mean-field description, we derive dynamics for three order parameters measuring attention retrieval, teacher-matrix alignment, and off-target logic overlap. These equations quantitatively match simulations for the order parameters and, combined with a logit-distribution approximation, qualitatively predict the sharp transition in final rollout accuracy. The analysis reveals staged learning: the logic module first learns a mixed heuristic; attention then locks onto the relevant action, enabling efficient MLP alignment. Together, these results provide a controlled mechanistic account of how attention-based retrieval and MLP-based logic co-develop during chain-of-thought state tracking.

We assess the dynamical properties of liquid water predicted by several density functionals using machine-learning interatomic potentials. MACE models were trained for SCAN, RPBE-D3/zd, revPBE-D3/zd, revPBE0-D3/BJ, PBE0-D3/zd, and PBE0-D3/BJ using previously reported ab initio datasets. We compare translational, rotational, and viscous dynamics through time-correlation functions, which resolve relaxation processes across different timescales, and through the corresponding long-time kinetic coefficients. The diffusion coefficient, second-rank orientational relaxation time, and shear viscosity reveal systematic differences among functionals. Part of these differences can be rationalized as shifts along the phase diagram, as comparisons relative to each functionals melting temperature reduce the spread in the dynamical observables. Among the functionals considered, RPBE-D3/zd provides the best overall agreement with experiment. We therefore perform a broader validation of RPBE-D3/zd using a Behler--Parrinello neural-network potential over a wide range of temperatures, densities, and pressures. The model reproduces the magnitude and anomalous pressure dependence of the diffusion coefficient, gives generally good viscosities, and captures the temperature dependence of the rotational relaxation time.

The interplay of structural, electronic, and magnetic degrees of freedom governs phase stability and critical temperatures in two-dimensional magnets. Controlling this coupling is essential for advancing fundamental understanding and spintronic applications. Combining first-principles calculations with Heisenberg Monte Carlo simulations, we reveal a rich magnetic phase diagram governed by the interplay of lattice strain and carrier density. These results provide a unified framework that reconciles diverse experimental reports on epitaxial layers and predicts a novel double-stripe antiferromagnetic phase, further stabilized by electron doping. Moreover, structural and electronic perturbations enable room-temperature ferromagnetism and antiferromagnetism. This magnetic evolution arises from competing, highly tunable direct and ligand-mediated exchange interactions in the presence of Ruderman-Kittel-Kasuya-Yosida coupling. By disentangling their individual contributions, we elucidate the underlying microscopic mechanisms, which transcends the conventional conduction electron picture. Finally, we quantify the colossal magnetoelastic response and identify zone-folded Raman modes that serve as unique experimental fingerprints for phase identification. Together, these results establish \ce{CrTe2} as a versatile platform for two-dimensional spintronics, where magnetic order and transition temperatures are tailorable via structural and electrical engineering.

Moiré multi-layers of transition metal dichalcogenides have been shown to exhibit optical responses that are endowed with a richness that is absent in single monolayers. Much of this can be attributed to the Moiré superlattice that modulates the electronic landscape of these heterostructures. Strongly coupled layer-hybridized excitons in $\text{MoSe}_2 / \text{WS}_2$ heterobilayers have been shown to exhibit enhanced optical nonlinearities. In this work we strongly couple layer hybridized excitons and trions in n-doped $\text{MoSe}_2 / \text{WS}_2$ heterobilayers inside an optical microcavity. We find that the additional Lindhard screening from dopant electrons and the formation of trions result in a strikingly non-monotonic nonlinear response. The absence of electron capture in the Moiré superlattice plays a crucial role, promising very large second-order nonlinearities. In this work, trion polaritons manifest as high velocity hot polaritons, reaching nominal diffusion lengths approaching 100 microns.

We study equilibrium clustering in a finite one-dimensional lattice gas of $L$ sites with periodic boundary conditions, as a minimal model for adsorption and binding on small ring-like substrates. Using a grand-canonical formulation with nearest-neighbor coupling, we derive exact finite-size expressions for the mean occupancy, the mean number of domain walls, and the mean number of clusters. Building on exact $k$-site correlation functions, we further derive expressions for the mean number of clusters of size $k$ and for two complementary size statistics: the cluster-size distribution, and the site-weighted cluster-size distribution. These observables characterize how spatial organization changes across attractive (cooperative) and repulsive (anticooperative) interactions, and highlight finite-size and parity-dependent effects of the underlying lattice, the latter being particularly pronounced near half filling in small systems. To access larger lattices without enumerating all $2^L$ microstates, we also develop a cluster-based combinatorial formulation in which configurations are classified by cluster counts and sizes, reducing the effective state space to a set whose size scales with integer partitions, $\approx e^{\sqrt{L}}$, rather than with $\approx e^{L}$. Taken together, our results provide exact benchmarks for finite periodic systems and suggest experimentally relevant cluster observables that complement occupancy-based measures of cooperativity, with particular relevance for binding on ring-like substrates for biological assemblies.

In this work, we formulate self-similarity from the perspective of scale invariance, where a self-similar form is understood as the transformation of a function into a form invariant under scale transformations. By applying this formulation to physical parameters, which consist of numerical values and units, it is demonstrated that dimensional analysis works for physical problems because scale invariance is partially shared between units and physical parameters. This naturally leads to the distinction between similarity of the first kind and similarity of the second kind according to whether the scale functions induced by units and those associated with physical parameters are equivalent or not. Self-similar solutions of the second kind can be further classified according to whether the power exponents of the similarity parameters include functions of dimensionless numbers. This leads to the conclusion that there are three kinds of self-similar solutions. The present work provides a unified framework for understanding dimensional analysis and a universal classification of self-similarity in physical problems.

Fluid and charge flow through interstitial volumes among impermeable randomly placed grains in porous materials ceases to occur at a critical concentration where networks of void volumes are disrupted at macroscopic scales. This critical density for void percolation can be difficult to calculate due to the irregular shape of the void regions. We develop and implement a geometrically exact method, scaling only linearly in the system volume, for identifying the shape and size of contiguous voids. In this manner, we calculate percolation thresholds for both grain cluster percolation (where system spanning networks of overlapping grains begin to appear with increasing density) and void percolation at much higher grain concentrations where networks of interstitial volumes no longer exist on macroscopic scales. For both the former and the latter, we calculate critical concentrations for inclusions in the shape of the Platonic solids (as well as truncated icosahedra) for both aligned and randomly oriented grains. In the case of critical densities for void percolation, the accuracy of our results is significantly improved relative to prior benchmarks. We also incorporate structural disorder of inclusions by considering impermeable grains in the form of cubes subject to a series of randomly placed and oriented fracture planes to mimic aggressively fractured inclusions found in nature. As the number of sustained slices becomes large, we find that the critical porosity for void percolation tends to 5%

Quadratic Unconstrained Boolean Optimization (QUBO) problems are widespread in both industrial applications and scientific studies. A QUBO problem corresponds to the optimization of a system of Ising spins defined on a generally sparse and heterogeneous graph. When the QUBO problem contains conflicting requests, the corresponding Ising system is frustrated, generating a complex energy landscape, which is hard to explore and optimize. Despite extensive algorithmic and hardware developments, finding low-energy configurations in these systems remains challenging (e.g., local-update heuristics typically become trapped in metastable states), especially when the (possibly frustrated) interactions generate extended correlated domains. We introduce CluMP (Cluster-based Message-Passing), an algorithm that performs collective updates on connected clusters of spins using information from Belief Propagation (BP). By controlling the amount of frustration within clusters, CluMP enables BP convergence on large subgraphs and proposes nonlocal rearrangements involving up to hundreds of spins in a single move. We benchmark CluMP against state-of-the-art local-update heuristics on spin-glass models defined on several graph topologies, including random regular graphs and lattice regular graphs in two and three dimensions. Cluster moves consistently bypass local trapping and reach lower energies with fewer effective operations than single-spin dynamics. These results demonstrate that frustration-tolerant cluster updates can be implemented efficiently on sparse graphs. The CluMP framework provides a scalable strategy for large-scale combinatorial optimization and inference problems, where exploiting medium- and long-range correlations is key to navigating complex energy landscapes.

Warm-blooded vertebrates accumulate approximately conserved numbers of physiological cycles over a natural lifetime: of order $10^9$ heartbeats and $10^8$--$3\times10^8$ breaths. These regularities are not exact constants, but their persistence across orders-of-magnitude variation in body mass, metabolic power, physiological frequency, and lifespan suggests that biological time is not measured by chronological duration alone. We develop the Principle of Biological Time Equivalence (PBTE), a thermodynamic framework in which lifetime cycle count is determined by the ratio between total lifetime entropy production and the entropy cost of one physiological cycle. Starting from the open-system entropy balance $\dot S=\dot e_p-\dot h_d$, we define the entropy cost per cycle as $σ_0=dΣ/dN$, where $dΣ$ is the entropy produced as the physiological clock advances by $dN$ cycles. For an adult homeostatic regime, this gives the cycle-count relation $N_\star=Σ/\langleσ_0\rangle$, with $Σ=\int_0^L \dot e_p(t)\,dt$, where $N_\star$ is the lifetime cycle count, $Σ$ is total lifetime entropy production, and $\langleσ_0\rangle$ is the lifetime-averaged entropy cost per cycle. In the homeostatic limit, $\dot e_p\simeq P/T$, so direct measurement of metabolic power $P$, body temperature $T$, and physiological frequency $f$ gives $σ_0\simeq P/(Tf)$. PBTE converts the empirical lifetime-cycle invariants into entropy-cost invariants. Under Kleiber metabolic scaling and quarter-power physiological-frequency scaling, the mass-specific entropy cost satisfies $\barσ_0=P/(TfM)\propto M^{3/4+1/4-1}=M^0$, providing a thermodynamic interpretation of allometric mass cancellation.

Ancient human beings started "external entropy production" in a late stage of evolution, in addition to the internal entropy production by which energy was dissipated within the body of life, as previously described consistently with the birth of life by maximum entropy production principle. In this paper, the mechanism for development of external entropy production, which is strongly related with use of tools and controlling fire, is theoretically investigated. Archaeological data show that the brain size of ancient human beings started rapid increase around 2.5 million years ago when the usage of tools and control of fire started. It may be natural to assume that the rapid growth of brain size is related to the growth of awareness which helped cooperation with the other human beings for control of fire. Coupled equations for the growth rate of brain including awareness and for growth rate of size of the interacting human beings are analyzed. The external entropy production per one human being which is directly related to the group size of cooperating human beings is estimated to increase as about 20 million years in the beginning from the critical time. This evolution created coexistence of internal entropy production of traditional multi-cellular life and new external entropy production of multi-body life. A psychological problem due to the coexistence of two kinds of entropy production mechanism in human being and concept of technologies based on the present thermodynamic evolution theory are discussed. It is suggested that the evolutionary understanding of the origin of global warming based on the external entropy production may be important to create an useful countermeasure.

In this paper, we propose an integrable semi-discretization of the two-component Hunter-Saxton (2-HS) equation, which is obtained as the short-wave limit of the two-component Camassa-Holm (2-CH) equation. We also show that the 2-HS equation can be derived from a new set of bilinear equations, distinct from previously known ones, via a pseudo 2-reduction and a hodograph transformation. Furthermore, we construct the N-soliton solutions of both the continuous and semi-discrete systems in Wronskian and Casoratian forms, respectively.

In the present work, we present a number of localized wave patterns that are theoretically analyzed and numerically illustrated to be observable within the widely applicable paradigm of the FPUT lattice. In particular, we derive a modified KdV equation from the FPUT lattice, which admits a variety of localized waves including these exact rational solutions representing rogue-wave profiles, solitons and breathers on the top of not only homogeneous, but also periodic elliptic function traveling-wave background. We utilize these exact solutions of the modified KdV reduction to construct consistent initial conditions for the FPUT lattice and perform time stepping of the latter. Relevant comparisons between these numerical solutions of the FPUT lattice and their associated analytical counterparts have been conducted to demonstrate good performance of the derived modified KdV reduction in approximating distinct localized wave structures from the FPUT lattice. This approach paves the way for importing a number of quasi-continuum waveforms to the FPUT lattice and the potential associated physical experiments, including recent ones in mechanical metamaterials.

This paper explores the effectiveness of various novel deterministic and traditional stochastic data assimilation (DA) and parameter recovery (PR) algorithms given noisy data from chaotic systems. We use semi-analytic methods to numerically construct synthetic data from the Lorenz '63 and multiscale Lorenz '96 chaotic dynamical systems, adding white noise. Our findings show that, for different noise levels, deterministic PR algorithms paired with deterministic DA algorithms are shown computationally to be overall more accurate and stable than stochastic PR algorithms. Additionally, deterministic PR methods have demonstrated greater speed and efficiency, requiring less computational power than stochastic PR methods. This suggests that future work should consider exploring the full potential of deterministic PR algorithms in the presence of noise.

This paper proposes a framework for energy localization in nonlinear oscillator chains operating in non-fundamental resonance, with emphasis on superharmonic regimes. The system is modeled as a metastructure composed of Duffing oscillators with both linear and nonlinear coupling, incorporating disorder-induced periodicity breaking. Under the assumption of strong excitation, the method of multiple scales is employed to derive the governing equations for soliton dynamics. At first-order perturbation, the classical form of the Nonlinear Schrödinger Equation emerges, whereas second-order analysis yields a previously unreported equation arising from the restitution of time scales. Analytical and numerical results demonstrate the nucleation of solitons in both hardening and softening regimes, based on two approaches: direct time-domain simulations from initially motionless states and numerical continuation in the frequency domain. A key finding is the distinct role of phase in superharmonic resonances compared to the primary resonance; specifically, the coexistence of multiple frequency components in the steady-state response precludes interpreting the soliton directly as a displacement envelope. Instead, the resulting secular terms captures the soliton associated with the resonant contribution, while transient components remain present under superharmonic excitation. Furthermore, robustness against disorder uncertainty is assessed by determining the tolerance levels that preserve the phenomenon. These results support the development of vibration control strategies aimed at mitigating the increase in resonant frequencies associated with the geometric downscaling of mechanical systems.

Manganese selenide (MnSe) is a promising air-stable two-dimensional magnetic semiconductor for which theory predicts robust ferromagnetism in monolayers with Curie temperatures approaching 250 K. However, the crystallographic phases and magnetic properties of thin-film MnSe grown by scalable methods remain poorly understood. Here, we demonstrate the controllable growth of $α$- and $β$-phase MnSe on C-face sapphire using a three-zone chemical vapor deposition process with elemental Se and ${MnCl_{2}}$ precursors in an $Ar/{H_{2}}$ atmosphere. Using Raman spectroscopy, atomic force microscopy, scanning electron microscopy, and X-ray photoelectron spectroscopy, we show that our process yields phase-pure $α$-MnSe nanorods and $β$-MnSe triangular flakes with lateral sizes up to 20 $μm$ and thicknesses of 15-30 nm. Low-temperature photoluminescence of the $β$-phase films reveals a bandgap of approximately 2.0 eV. Systematic variation of growth parameters shows that precursor vapor pressure, rather than ${H_{2}}$ partial pressure, is the dominant factor controlling lateral flake size. Vibrating-sample magnetometry measurements reveal a $N{é}el$ temperature of 53 K in the $β$-phase films, providing clear evidence of antiferromagnetism in the multilayer regime and establishing MnSe as a tunable platform for 2D spintronic and optoelectronic devices.

High-refractive-index materials underpin a wide range of optical technologies, including communications, imaging, lasers, and integrated photonic systems. Here, we demonstrate a self-assembled metamaterial platform based on gold nanoparticle aggregates with nanometer-scale gaps exhibit remarkably high effective refractive indices exceeding 15 in the mid-infrared regime, while simultaneously producing gap-field enhancements of at least two-orders of magnitude. This combination of high refractive index and extreme field enhancement enables exceptionally strong light-matter interactions. We demonstrate this by designing a compact high-index metamaterial device supporting an anapole, which further enhances the nanogap field. By placing quantum emitters with terahertz transitions inside the plasmonic gaps, we show a stimulated-emission response enhanced by at least three orders of magnitude, highlighting applications in non-linear optics, frequency up-conversion and vibrational strong coupling.

The Tropical Atlantic exhibits complex sea surface temperature (SST) variability driven by internal ocean-atmosphere interactions and remote climate forcing. We perform a comparative multifractal analysis of three SST indices, South Atlantic Tropical (SAT), Tropical Southern Atlantic (TSA), and the Tropical Atlantic SST Gradient Index (TASI), using weekly data from 1981 to 2025. Multifractal Detrended Fluctuation Analysis (MFDFA) reveals robust scale-dependent behavior in all indices. TASI displays a substantially broader multifractal spectrum (Delta h about 0.72) than SAT (0.27) and TSA (0.34). Surrogate-data tests show that multifractality in SAT and TSA is mainly explained by linear autocorrelations, whereas TASI contains an additional nonlinear contribution associated with phase correlations. To investigate temporal variability, we introduce a moving-window MFDFA framework that tracks the evolution of multifractal width. Significant reductions are observed during the major 1997-1998 and 2015-2016 El Nino events, indicating a suppression of multiscale variability under extreme Pacific forcing. Lagged correlation analysis reveals a significant negative association with the Oceanic Nino Index at delays of 15-18 months, consistent with known Atlantic-Pacific teleconnections. However, Granger causality and Transfer Entropy tests do not detect significant causal links, suggesting an episodic rather than persistent relationship. Lagged multifractal cross-correlation analysis further reveals scale-dependent inter-basin coupling. These results demonstrate that time-dependent multifractal measures provide a useful framework for characterizing nonlinear Atlantic variability and identify TASI as a dynamically distinct index whose scaling properties contain information not captured by regional SST indices alone.

Transitioning quantum magnetometry from laboratory environments to real-world applications has been limited by a persistent trade-off between sensor miniaturization and magnetic sensitivity. While bulky systems can achieve high sensitivity, endoscopic probes commonly suffer from inefficient fluorescence collection and reduced performance. Here we resolve this trade-off and present a miniaturized diamond quantum magnetometer with a 6 mm diameter endoscopic sensor head, achieving a magnetic-field sensitivity of 91 pT/sqrt(Hz) with a 2 kHz measurement bandwidth in a magnetically unshielded environment. The fluorescence collection bottleneck is overcome by separating excitation and collection into different cores of a fused multi-core fiber bundle, coupled to the diamond through a custom high-numerical-aperture micro-objective. A compact FPGA-based backend performs microwave control, lock-in detection and real-time resonance tracking, enabling robust operation during magnetic-field imaging. To demonstrate the practical utility of the miniaturized sensor, we image the magnetic field of a commercial lithium-ion pouch cell during charge and discharge and reconstruct depth-integrated current-density maps of the current flow. These results show that endoscopic diamond magnetometers can combine high sensitivity with a probe geometry suitable for confined, unshielded measurements, opening new avenues in battery technology and beyond.

We propose a method for extending the depth-of-field (DoF) to construct high-fidelity neural radiance fields (NeRF) -- an emerging technique for rendering photorealistic novel views from a dataset of images captured at different viewpoints, based on implicit neural representations. The trade-off between DoF and light quantity is inherent not only in conventional cameras but also in NeRF, since the datasets used by NeRF are captured by these cameras. To address this issue, we introduce a coded aperture placed at the camera pupil, preserving spatial frequency components under defocused conditions. We develop a camera model incorporating coded apertures into NeRF, allowing direct input of coded images and enabling the generation of novel views with an extended DoF. We validate the proposed method, termed extended DoF-NeRF (EDoF-NeRF), through simulations and experiments, demonstrating its superior performance compared to conventional aperture cameras.

We experimentally demonstrate that absorption and emission spectra of trivalent lanthanide-doped anisotropic crystals can exhibit a significant magnetic-polarization dependence, which has been largely overlooked in spectroscopic studies to date. Focusing on the uniaxial laser host LiYF4 (YLF) doped with Yb3+, Tm3+, Er3+, and Ho3+, we measure magnetic-polarization-dependent absorption and emission spectra for transitions with strong magnetic-dipole (MD) contributions predicted by theory. Our results reveal that MD-induced spectral anisotropy, i.e., spectral differences for the same electric field orientation but for different magnetic field orientations, is present even in these well-established laser materials. A complete spectroscopic characterization of uniaxial crystals requires three polarizations, including the $α$-polarization, with both the electric field vector E and the magnetic field vector H perpendicular to the c-axis (E $\perp$ c, H $\perp$ c), in addition to the commonly used two polarizations $π$ (E $\parallel$ c, H $\perp$ c) and $σ$ (E $\perp$ c, H $\parallel$ c). We further discuss the observed MD-induced spectral anisotropy and calculated MD branching ratios, the impact of the anisotropy on emission cross-section calculations, and the relevance of our results to other uniaxial and biaxial crystals.

Nanoscale three-dimensional localization across large biological volumes remains an outstanding challenge in optical microscopy, with existing approaches typically limited by imaging speed, volumetric field of view and localization precision when required simultaneously. Here, we overcome these limitations by combining a twin Airy engineered point spread function with two-photon light sheet fluorescence microscopy, enabling nanoscale localization throughout large volumetric fields of view. Our framework explicitly incorporates the broadband fluorescence emission characteristic of biological fluorophores, ensuring accurate localization under realistic imaging conditions. We achieve localization precisions of $<20$~nm laterally and 42~nm axially over volumes measuring 295~$μ$m x 330~$μ$m x 100~$μ$m , with a projected path to sub-10-nm localization in millimeter-scale specimens. Experiments in fluorescent bead phantoms and live mammalian oocytes confirm robust performance in both controlled and biologically complex environments. These results establish a scalable strategy for localization-based super-resolution imaging across biologically relevant volumes, bridging the gap between nanoscale precision and large-scale volumetric microscopy.

Free-surface liquid metal flows are a leading candidate for the plasma-facing components of future fusion reactors. Existing transient, three-dimensional, free-surface MHD solvers rely on the inductionless approximation in which the induced magnetic field is neglected. This paper extends the open-source solver FreeMHD [B. Wynne et al., Phys. Plasmas 32, 013907 (2025)] beyond the inductionless approximation to resolve the induced magnetic field self-consistently using a vector-potential formulation that enforces $\nabla\cdot\boldsymbol{B}=0$ by construction while preserving the original multi-region, two-phase framework. The solver is verified against analytical Shercliff and Hunt duct-flow solutions across a range of Hartmann numbers and validated against free-surface height measurements from the LMX-U experiment. To the best of our knowledge, FreeMHD2 is the first open-source, experimentally validated free-surface liquid metal solver to resolve the evolution of the induced magnetic field without invoking the inductionless approximation. By removing this approximation rather than relaxing it, the formulation provides the basis for future modeling of the finite magnetic Reynolds number conditions expected in large-scale, transient fusion events.

Tin vacancy (SnV) centers in diamond are promising solid state qubits for integrated quantum photonics. Here, we fabricate and characterize a diamond on Al2O3 dual taper waveguide structure containing SnV centers, demonstrating optical coupling between the diamond nanobeam and the underlying Al2O3 waveguide. The devices are realized using a bilayer fabrication approach compatible with wafer scale lithography. Clear guided SnV- emission is observed in all optically active devices, indicating effective optical coupling in the integrated structure. These results demonstrate a scalable fabrication approach toward integrating diamond color centers with photonic waveguides.

Electrothermal characterization based on the third or other harmonics of an ac Joule heating current is widely deployed for the thermal analysis of solid conductors and their environment, including solid substrates and fluids. However, a unified theory that bridges heat transfer in two archetypal experimental geometries - suspended vs. substrate-supported conductor - has been missing. Here, we present and validate such a theory that explicitly accounts for finite conductor length, thermal mass, and environmental coupling through a unified thermal transfer function. This framework enables the prediction of voltage responses at all harmonics of the driving current (dc, 1$ω$, 2$ω$, 3$ω$) and the formulation of design principles for the characterization of nanoscale conductors. The conductor length $l$ is the primary parameter controlling the frequency regime at which the conductor's thermal mass dominates the thermal response, with the characteristic frequency $ω_\mathrm{c}=α/l^2$, where $α$ is the conductor's thermal diffusivity - closely related to a criterion previously reported for suspended wires free from environmental coupling. Our unified framework generalizes this result, revealing that sufficiently weak environmental coupling is a necessary condition for $ω_\mathrm{c}$ to govern the onset of thermal-mass-dominated response. Optimization of interfacial thermal resistance and environmental thermal impedance may further improve temperature resolution and facilitate on-substrate implementations.

We investigate the effects of environmental noise on cooperation in a spatial evolutionary game model with variable population size. Building on a one-dimensional lattice model in which vacancies promote cooperation through spatial selection, we add random noise to the environmental quality parameter and consider two distinct types: annealed noise, where the environmental quality fluctu ates independently at each site and each time step, and quenched noise, where each site is assigned a permanently fixed random value. For annealed noise, we develop a mean-field theory by replacing the noise-dependent death probabilities with their distribution averages, and find that increasing the noise intensity shifts both the cooperator-defector phase boundary and the absorbing boundary upward in the parameter space, simultaneously expanding the cooperative regime and the extinc tion region. These predictions are confirmed by numerical simulations. In contrast, quenched noise leaves the phase boundary nearly unchanged across all noise levels, exerting only a weak effect on cooperator frequency. Together, these results demonstrate that temporal fluctuations, rather than static spatial heterogeneity, are the primary driver of noise-induced shifts in the cooperative phase structure.

Deep research agents are Large Language Model (LLM)-based systems designed for autonomous, multi-step scientific reasoning, and they hold immense potential for accelerating research in the physical sciences. However, comprehensive and in-depth evaluations of their capabilities within this domain remain lacking. To address this gap, we introduce PhySciBench, a benchmark highly relevant to physical science research, comprising 200 expert-curated questions, balanced between physics and chemistry, across six task categories that reflect real-world scientific workflows. Evaluations of state-of-the-art models and agent systems on PhySciBench reveal limited performance; even the strongest baseline, Gemini Deep Research, achieves an accuracy of only 33.5%. Analysis of failure cases identifies three recurrent deficiencies: fragility in extended reasoning chains, limited knowledge transfer across steps, and a lack of physics-grounded self-verification. Motivated by these findings, we develop DelveAgent, a modular multi-agent framework equipped with an adaptive planning loop, dual-granularity memory, and a hierarchical physics-grounded reflection mechanism. Across four scientific benchmarks, DelveAgent improves accuracy by up to 7.5 percentage points while reducing inference costs to approximately one-third of the strongest baseline. These results establish the significance of PhySciBench as a critical benchmark for evaluating AI systems in the physical sciences and demonstrate that architectural specialization can effectively enhance the reliability of autonomous scientific research.

We report a comparison of frequency fluctuations in oven-controlled quartz bulk-acoustic-wave oscillators operated above ground and one kilometre underground in a low-muon-background environment. The experiment is motivated by the possibility that cosmic rays and other ionizing-radiation backgrounds produce rare, impulsive energy-deposition events that perturb high-Q mechanical resonators and appear as intermittent, non-Gaussian structure in oscillator frequency noise. Conventional power spectral density and Allan-deviation analyses show no statistically compelling separation between the two environments over the explored timescales. In contrast, multi-scale sample entropy and its modified form reveal a pronounced divergence, with the underground data exhibiting increased predictability over a broad range of effective integration times. This result identifies a change in the temporal structure of the oscillator fluctuations that is largely hidden from standard second-order frequency-stability metrics. We therefore propose multi-scale sample entropy as a new diagnostic for frequency control and timing, complementary to Allan deviation and spectral analysis, with particular sensitivity to intermittent structure, non-stationary contributions, and rare-event contamination. The observed entropy separation also provides evidence that the above-ground cosmic-ray environment influences oscillator frequency fluctuations, suggesting that radiation-linked disturbances may contribute to the stochastic behaviour of precision mechanical oscillators. These findings introduce an entropy-based methodology for oscillator metrology and provide a practical tool for future fundamental-physics experiments using cryogenic resonant sensors, where rare-event backgrounds and poorly understood low-frequency noise can limit sensitivity.

To simulate the relaxation of an entire 2D cellular network, this study proposes a symmetric relaxation method for both inner and marginal vertices. The relaxations of these two types of vertices are determined by the central angle symmetry of associated cells and the angle symmetry at each vertex, but with different major considerations. Trimmed Voronoi networks with varying irregularity are used as initial networks for the relaxation simulation. In particular, we propose a regular hexagon disordering method to generate Voronoi networks and find that the inner cells of networks with an irregularity value of one exhibit a conserved edge number distribution, as found in other 2D cellular networks. Simulation results agree with the von Neumann-Mullins law for both inner and marginal cells, and a modified equation including a geometric correction term significantly improves prediction quality. The Aboav-Weaire law and Lewis law are also reproduced, with the latter showing that relaxed cells tend to approach the ellipses' maximum inscribed polygons. Analysis of edge length, interior angle, and shape index reveals that symmetric relaxation inhibits T1 (neighbour exchange) topological transitions by reducing short edges while increasing area disparity among neighbouring cells. The findings suggest that T1 events may be triggered when force disequilibrium overcomes the stabilising effect of symmetric relaxation, providing a possible mechanistic explanation for T1 in 2D foams.

Ferroelectric materials, with their strong nonlinearities, underpin key technologies across radio-frequency (RF) signal processing, optical communications, and emerging quantum systems. Barium titanate (BTO) is a notable example, combining strong piezoelectric and electro-optic responses. While bulk BTO has been studied for decades, the piezoelectric properties of its recently available thin films, and their behavior at the millikelvin temperatures relevant to quantum hardware, remain largely unexplored. Here, we fabricate and characterize surface acoustic wave (SAW) resonators on thin-film BTO. The measured devices exhibit high electromechanical coupling (k2eff 0.14 at 5.2 GHz) and operate up to 7.8 GHz. From these measurements, combined with finite-element modeling of the multi-domain microstructure, we extract an effective piezoelectric coefficient d33eff of 53 pC/N, comparable to bulk BTO. Exploiting the intrinsic ferroelectricity, we further demonstrate low-voltage switching with a fast (100 ns) response, attractive for reconfigurable RF front-ends and parametric amplifiers. Extending these measurements to millikelvin temperatures, we find that the piezoelectric response persists, with d33eff 19 pC/N, pointing to the potential of BTO for piezoelectric coupling in superconducting quantum circuits. These results position thin-film BTO as a promising piezoelectric platform for both classical and quantum information technologies.

Structural ensemble refinement is widely used to integrate molecular simulations with experimental measurements. While most applications focus on the maximum-a-posteriori (MAP) ensemble, Bayesian sampling of the posterior distribution can provide uncertainty estimates and posterior averages for arbitrary observables. A notable step in this direction was introduced by the Bayesian Energy Landscape Tilting (BELT) framework, where sampling is performed on a family of maximum-entropy ensembles parametrized by Lagrange multipliers. Here, we show that Bayesian sampling in this setting requires an explicit choice of ensemble-counting measure. In particular, the flat measure in Lagrange-multiplier space used in the original BELT formulation leads to a posterior distribution that is formally non-normalizable for finite reference trajectories. We propose the Jeffreys measure as an invariant ensemble-counting prescription, restoring normalizability in the finite-sample situations considered here, and providing a consistent definition of posterior averages. Using both an analytically tractable Gaussian model and maximum-entropy refinement of RNA oligomer simulations, we compare different ensemble-counting measures and show that they can significantly affect Bayesian estimates. The resulting methodology has been implemented in the \texttt{MDRefine} software package.

The $δ$ phase of Niobium Nitride (NbN) is a promising superconducting material, which is chemically stable and shares lattice compatibility with conventional III-Nitride semiconductors. Due to a high critical temperature (T$_{c}$) and a high critical (magnetic) field (H$_{c}$), NbN is much-coveted for a diverse set of applications spanning from single photon detectors, and hot-electron bolometers to quantum computing architectures using superconducting circuits. However, synthesizing high-quality epitaxial films of phase pure and stoichiometric $δ$-NbN in a cost-effective manner, is challenging. In this study, we investigate the epitaxial growth of single crystalline $δ$-NbN on TiN-buffered c-sapphire (Al$_{2}$O$_{3}$) substrates by sputtering at room temperature. For these films, we demonstrate a surface-roughness in picometer-scale, the lowest reported till date. The critical temperature (T$_{c}$) of the epitaxial $δ$-NbN films was observed to decrease with the insertion of the TiN buffer layer, tentatively attributable to the leakage of Cooper pairs, due to the proximity effect. TiN and NbN layer behave as a bilayer system, wherein Cooper-pair leakage is facilitated by the absence of any oxide interlayer. Consequently, T$_{c}$ reduces with increasing thickness of the TiN layer.

We characterize two nested Walsh parity-check filters implemented on Quandela's Belenos cloud photonic processor in a single-photon eight-mode spatial register. The modes are indexed by the vertices of the cube $Q_3$. The filters realize the classical $[8,7,2]$ single-parity-check code, the zero-sum neutral subspace $\mathcal{N}$ and the $[8,4,4]$ extended Hamming code, the parity-checked subspace $\mathcal{S}\subset\mathcal{N}$ with one DC and three face-parity syndrome channels. These are first-quantized path/mode encodings of classical codes: the experiment verifies leakage suppression and syndrome routing, not error correction or protection against photon loss, and all probabilities are conditional on postselected single-photon detections. Across more than 340,000 detections, neutral inputs show residual DC-port leakage of $0.02\%$-$1.1\%$ (mean $0.6\%$), corresponding to $\approx21\times$ suppression relative to the ideal $0.125$ DC-capture baseline and $31.6\times$ relative to the measured non-neutral control. Injected DC contamination gives a monotonic soft error signal, and the three face-parity syndrome channels route to their predicted ports with $94$-$99\%$ selectivity. A sector-preserving unitary core keeps leakage far below non-neutral controls over one to three applications, with differences dominated by calibration and compilation systematics rather than gate-cycle physics. We quantify these limits, including fixed-pattern separator bias, $\pm 0.02$ calibration offsets, and compilation scatter near the $10^{-3}$ leakage level, and report a Hong-Ou-Mandel degradation episode in which suppression vanished and recovered after recalibration.

The Korringa-Kohn-Rostoker (KKR) Green's function method is a first-principles electronic structure approach well suited to substitutionally disordered alloys through the Coherent Potential Approximation (CPA). AkaiKKR is a widely used implementation, known for efficient treatment of metallic systems and their magnetic properties. Its output, however, is unstructured plain text with no programmatic interface, leaving data extraction entirely to the user and making systematic or high-throughput studies impractical. akaitools is a Python package that parses AkaiKKR output files into structured, type-annotated Python objects. The package covers three output types: self-consistent field (SCF) results, which capture convergence history and per-atom electronic and magnetic properties; spin-resolved, orbital projected density of states for each CPA component; and Bloch spectral functions on a user-defined k-point path. Results come back as immutable dataclasses backed by NumPy arrays. Energy quantities are available in both Rydbergs and electronvolts, and results can be exported to Pandas DataFrames. A built-in plotting module produces Matplotlib figures for DOS curves and SCF convergence. A command-line interface provides file summaries and JSON export without any Python scripting. The package also includes a programmatic input file generator, so full calculation pipelines from input preparation to output analysis can be run in Python.

The Ashtekar-Lewandowski (AL) volume operator of loop quantum gravity is central to the Hamiltonian constraint, but its vertex action is usually obtained from dense spectral decompositions of finite recoupling matrices, obstructing numerical analysis on large kinematical Hilbert spaces or high-valence vertices. We formulate a matrix free action of the $SU(2)$ AL vertex volume operator in standard recoupling basis, making use of the Brunnemann-Thiemann expression for the oriented AL volume density $Q_{v}$ whose matrix elements can be generated locally from recoupling theory without forming the full matrix. Based on the Balakrishnan-Stieltjes representation of $(Q_{v}^{2})^{1/4}$ we approximate the volume by shifted-resolvent quadrature (SRQ). The resulting action uses only repeated applications of $Q_{v}$ and shifted positive linear solves, making it compatible with multi-shift Krylov methods. We prove exact preservation of the volume kernel, provide operator-norm and residual error estimates, discuss sector-wise scaling bounds, and validate the method on an embedded $K_{5}$ graph at small spin cutoffs against exact dense local-block operators. Numerical simulations show rapid convergence of vertex expectation values, controlled dependence on bound parameters, and exact preservation of zero-volume modes. We further demonstrate matrix free Monte Carlo estimates at doubled-spin cutoff $2j=250000$ beyond dense materialisation, and show that SRQ can be combined with stochastic Lanczos quadrature to estimate fixed-sector volume spectral measures without dense volume matrices.

Differential Interference Contrast (DIC) microscopy and chiral analysis are two imaging techniques that measure the birefringence, i.e., the phase difference introduced by a sample on two orthogonal polarizations. Conventional approaches employ Gaussian beams and infer birefringence from polarization changes, resulting in phase-estimation sensitivities that depend on the unknown phase. We demonstrate here a new type of birefringence detector. It makes use of a vector vortex beam, a type of structured light endowed with optical modes that carry opposite orbital angular momentum (OAM). Using quantum estimation theory tools, we demonstrate that the sensitivity of phase estimation is independent of the value of the unknown phase, and can be even better, in principle, than the conventional approach. We experimentally validate the proposed scheme, demonstrating the potential of structured light for robust and uniform birefringence sensing.

Interference of time-bin encoded signals over free-space optical channels typically requires stringent mode filtering on receivers due to wavefront distortions from atmospheric turbulence, conventionally addressed with adaptive optics. Passive multimode receivers based on field-widened interferometers present a compelling alternative, enabling direct interference without the overhead of wavefront correction. We demonstrate a field-widened interferometer design that is implemented solely with reflective surfaces and achieves a high interference visibility (greater than 0.97) for spatially multimode beams. The interference of the multimode beams is enabled by two imaging systems that consist of a cavity configuration between a spherical concave mirror and a flat mirror. The configuration enables small form-factors, is inherently achromatic, and is based on standard spherical mirrors which reduces the complexity of the system. The interferometer is applicable for spatially multimode and turbulent optical channels, such as satellite communication, and is designed for quantum systems that use time-bin encoded qubits.

Political polarization poses a variety of challenges for modern democracies. Entrenched disagreements on policy can prevent constructive discourse and compromise, and high levels of affective polarization threaten to undermine social cohesion and support for institutions. Finding ways to promote constructive discourse while maintaining free expression has proved a challenge for social media platforms, media outlets and policy makers alike. Here we develop a computational model -- based in psychology -- of online discourse and opinion dynamics under complex individual identities, which we use to assess the capacity of realistic interventions to prevent or reverse polarization. We show that changes to the range of acceptable opinions in a society -- i.e. the Overton window -- have a limited impact on polarization, and that attempts to ``optimize'' the Overton window can even trigger the onset of polarization. In contrast, interventions that shift attention towards under-discussed topics, or increase the costs of violating existing norms, are often effective at preventing polarization, but are less successful at reversing it. Most strikingly, increasing the salience of influential individuals, who model non-polarized discourse, can be highly effective at both preventing and reversing polarization. However we also find that once polarization has set in, even the most successful interventions result in latent extremism when identities are complex. Our work suggests that restricting speech by shrinking the range of acceptable discourse is an ineffective way to tackle polarization, whereas enforcement of existing norms, attention nudges and the presence of elites who model good behavior can be highly effective.

Learned PDE simulators are increasingly used as low-cost replacements for expensive numerical solvers, but standard relative $L^2$ error does not determine whether a learned model behaves as a coherent numerical time propagator. This paper presents a diagnostic software suite for auditing learned PDE simulators as approximate evolution operators. The suite provides architecture-independent, post hoc diagnostics for relative state error, semigroup consistency, finite-difference generator discrepancy, energy behavior, integral balance, admissibility constraints, perturbation response, and scaling-law consistency. The software is designed around a minimal contract: reference trajectories, a learned propagator or saved predictions, equation metadata, and a diagnostic configuration specifying which structures are meaningful for the problem under study. We validate the suite on five benchmark PDE tasks: two-dimensional incompressible Navier-Stokes, shallow-water dynamics, active matter, three-dimensional compressible Navier-Stokes, and three-dimensional magnetohydrodynamics, using FNO, DeepONet, U-Net, and ResNet-style surrogate models together with controlled underfit and oversmoothed variants. The validation study shows that relative $L^2$ error can remain moderate, or even improve, while structural diagnostics deteriorate substantially. The package therefore supports software-level auditing of learned PDE simulators by reporting an interpretable diagnostic panel rather than collapsing model behavior into a single state-error score.

Physics-informed neural networks (PINNs) offer a promising framework for inverse hydrodynamic modeling by combining sparse observations with governing physical constraints. However, their reliability for estimating hydraulic parameters under data limitations remains insufficiently characterized. This study benchmarks inverse PINN recovery of the Manning friction coefficient in the shallow water equations under controlled variations in observation sparsity, noise, and observed variable type. Two cases are considered: a one-dimensional MacDonald subcritical channel with an analytical steady reference solution, and a two-dimensional sloped channel with a parabolic transverse bed generated using a balanced finite-volume solver. The Manning coefficient is treated as a trainable positive scalar and recovered jointly with the flow field using a two-phase strategy that first fits observations and then incorporates the physics residual. Results show that the two-dimensional case achieves robust friction recovery, with errors below 5% when at least 10 depth and velocity observations are available and noise is at or below 10% of the field standard deviation. Recovery remains stable up to 20% noise with 50 observations, but becomes unreliable with only five observations. In contrast, the one-dimensional case shows a persistent positive bias of about 15% that is largely insensitive to observation count and noise, indicating a structural identifiability limitation rather than a data-density limitation. Observation-type ablation shows that recovery degrades substantially when only depth or velocity is observed, demonstrating that joint depth-velocity information is essential for reliable inverse identification. Overall, the results provide a reproducible benchmark for assessing when inverse PINNs can and cannot reliably estimate Manning friction from sparse and noisy shallow-water observations.

Chip-based nonlinear photonics offer the capability to integrate devices with all the requisite photonic components (e.g., filters, couplers, detectors) into highly compact form factors. This offers the possibility of making the devices scalable, robust, and manufacturable. Such integrated photonic devices will enable applications in optical communications, precision metrology, microwave generation, and LIDAR. However, thermal instabilities represent a major hurdle in the deterministic operation of nonlinear optical processes in such integrated resonant structures such as microresonators. In this work we demonstrate deterministic and highly stable Kerr soliton comb generation in tight-spiral microresonators with comb spacings as low as 16 GHz. We perform a comprehensive experimentally-validated thermal model of such compact microresonators and reveal non-trivial thermally-driven instabilities governing the cavity soliton dynamics. We design and implement a fast feedback loop on the devices to overcome thermal perturbations and stabilize cavity-soliton states, including those that are otherwise unstable, and to allow for controlled transitions between the soliton states. Our approach enables the realization of thermally-stable highly compact soliton microcomb devices in a wide variety of photonic platforms.

The magnetization dynamics of an all-garnet trilayer consisting of Y3Fe5O12/Y3Fe3Al2O12/Y3Fe5O12 (YIG/YIAG/YIG) is analysed. The two magnetic YIG layers, separated by a 4 nm thick paramagnetic YIAG layer, are coupled via dipolar interactions leading to the formation of hybrid magnon modes distinct from the modes found in a single YIG layer. The YIAG exchange-decouples the YIG layers while enabling lattice coherence, maintaining the low damping of the YIG. Both ferromagnetic resonance and micro-Brillouin light scattering measurements were used to characterize the sample and its hybrid dynamics, which showed excellent agreement with an analytical reciprocal space model of spin-wave dynamics and with micromagnetic modeling of a dipolarly coupled magnetic heterostructure.

Quasi-optical coupling serves as the critical interface in terahertz (THz) heterodyne receiver systems, enabling efficient transfer of incident radiation to superconducting hot-electron bolometer (HEB) mixers through a focusing element and a planar microwave antenna. With recent advances in nanofabrication, planar dielectric metalenses have emerged as promising alternatives to conventional refractive optics due to their compactness and scalability. However, unlike conventional elliptical silicon lenses that are often treated as nearly ideal optical components, the focusing efficiency of metalenses is strongly dependent on the local deflection angle across the aperture, creating an urgent need to quantitatively understand the coupling between a dielectric metalens and a planar antenna. In this work, we present a quasi-optical coupling analysis between a planar Si metalens and a logarithmic spiral antenna integrated with a THz superconducting NbN HEB mixer operating at 1.63 THz using a spherical-coordinate vectorial integration. By combining the angular radiation profile of the spiral antenna with the deflection-angle-dependent focusing efficiency of the metalens obtained from numerical simulations, the calculated coupling efficiency is directly correlated with experimentally measured double-side-band receiver noise temperatures through comparison with a conventional elliptical Si lens measured under the same receiver configuration. The analysis establishes a quantitative relationship between metalens focusing efficiency, antenna coupling, and receiver noise temperature, providing guidance for optimizing metalens design and improving the overall performance of metalens-integrated THz heterodyne receivers.

A fourth-order conservative adaptive multiresolution average-interpolating wavelet upwind scheme is proposed for compressible flows governed by hyperbolic conservation laws. A family of asymmetric average-interpolating wavelets with upwind properties is constructed for conservative finite volume discretization, while symmetric average-interpolating wavelets are employed for multiresolution decomposition and reconstruction of physical variables in the adaptive procedure. Since both the conservative discretization and the adaptive multiresolution representation are constructed from cell-average quantities, the proposed scheme preserves strict conservation during both numerical evolution and adaptive cell redistribution. Unlike hybrid adaptive wavelet methods that use wavelets mainly for data compression and mesh adaptation, the present adaptive wavelet upwind scheme utilizes average-interpolating wavelet multiresolution approximation to reconstruct the interface values directly for numerical flux evaluation, thereby avoiding additional ghost-cell marking and reconstruction near coarse--fine mesh interfaces. The boundary variation diminishing reconstruction is incorporated at the finest resolution level to achieve non-oscillatory shock-capturing capability. Numerical tests demonstrate that the proposed scheme achieves the expected fourth-order accuracy, maintains conservation errors close to machine precision, and controls numerical errors around the prescribed threshold. The proposed method also sharply captures shock waves and contact discontinuities without spurious oscillations and resolves multiscale smooth structures through a sparse adaptive representation. These results indicate that the proposed scheme provides an efficient, conservative, and reliable approach for high-resolution simulations of compressible flows.

Improving the resolution of telescope systems will provide the opportunity to study new physical phenomena in previously unobserved environments. Spatial mode de-multiplexing (SPADE) based imaging is a promising and rapidly evolving technique for pushing the resolution of optical telescopes beyond the diffraction limit. A key application of this technique is for near-optimal hypothesis testing for the presence of secondary and extended sources in the sub-diffraction regime. We present the first demonstration of a binary-SPADE based hypothesis testing instrument deployed on-sky. In our proof-of-principle experiment, based on mode demultiplexing with a double clad fiber coupler, we demonstrate detection of a binary star system separated below the diffraction limit. We perform measurements in the photon-starved regime where no image can be formed by traditional direct imaging. We find the scaling of the system's type II error rate (the ``binary source miss" chance) was heavily limited by unbalanced loss in our double-clad fiber coupler when compared to the idealized quantum limits. Despite this the evaluated type II error is always lower than a perfect direct imaging measurement. We expect that if this instrument is scaled to larger aperture telescope systems the effects of atmospheric turbulence will further degrade this system's performance.

We use temporal unitary transforms to generate 16-QAM up to 220 GBd using only 50-GHz electrical bandwidth. The technique is theoretically lossless and can generate arbitrary optical waveforms beyond the bandwidth of the constituent modulators.

We report on a vision AI agent pipeline for non-contact material strain and property extraction from video data, demonstrated on video taken during hydrostatic testing of four OFHC copper cylinders conducted as part of the LEGEND-1000 hardware validation campaign. Traditional strain gauge measurements proved unreliable, motivating a fully-automated agentic alternative. The agent was built on the LangChain framework with Claude Haiku 4.5 as its central reasoning engine, integrating a specialized suite of computer vision tools: FFmpeg for video preprocessing and rotation correction via Hough Line Transform, the Segment Anything Model 2 (SAM2) for spatiotemporal segmentation with automated memory-informed dynamic chunking, and a hybrid EasyOCR and LLM-based timestamp validation pipeline. Three specialized sub-agents were developed to process the video data and obtain cylinder diameters and timestamps while autonomously handling obstacles such as corrupted frames and memory limits. From the diameter profiles synchronized to pressure data, hoop stress--strain curves were reconstructed and yield strengths were calculated using the 0.2\% offset, 0.5\% EUL, and Johnson-Cook methods across two independent tests. Cross-validation against a non-agentic pipeline confirmed agreement for the diameter extraction at the $\pm$5 pixel level. The material properties and testing results were further compared to Ansys mechanical simulations performed as part of the LEGEND-1000 reentrant tube design campaign. This work showcases the power of agentic pipelines to extract materials data from video alone.

Cryogenic hydrogen isotope fuel layers with high structural integrity and atomic-scale smoothness are prerequisites for symmetric implosion and ignition in inertial confinement fusion (ICF). Using deuterium (D$_2$) as model fuel, we perform large-scale molecular dynamics simulations with a Feynman-Hibbs corrected Silvera-Goldman potential to describe nuclear quantum effects at low temperatures, systematically investigating D$_2$ crystallization inside spherical ablator capsules. By varying substrate lattice constant from 3.1 angstrom to 3.9 angstrom, we demonstrate that lattice matching dictates the transition from coherent epitaxial growth to polycrystalline formation, establishing it as the primary design principle for high-performance targets. When the substrate lattice closely matches the equilibrium hexagonal-close-packed (HCP) spacing of cryogenic D$_2$ (approximately 3.5 angstrom), D$_2$ forms coherent layer-by-layer epitaxial growth consistent with Ostwald's stepwise nucleation theory, yielding HCP-dominated near-single crystals with minimal dislocations and ultra-smooth inner surfaces. In contrast, large lattice mismatch destabilizes coherent growth and causes island-like growth, producing polycrystalline structures with mixed HCP/FCC phases, elevated defects, and greatly increased surface roughness. Radial stress analysis shows that interfacial stress from mismatch localizes within 2-3 molecular layers near the interface, triggering subsequent defect-mediated growth. These findings highlight substrate lattice matching in regulating confined solid growth and crystallization quality, establish it as a key principle for ablator inner-surface engineering in ICF cryogenic targets, and offer atomic guidance for growing high-quality single-crystal deuterium-tritium (DT) fuel layers with optimal smoothness.

A prominent theory in cognitive science suggests that concepts in the brain are organized as high-dimensional vectors, with semantic meaning captured by directions and relative angles in this space. Brain decoding is the effort of reconstructing or retrieving stimuli (or their representations) from neural activity and involves finding a function that approximates how the brain represents concepts. This motivates the investigation of contrastive objectives as biologically plausible candidates to reverse the brain loss function. In this work, we study how functional MRI (fMRI) activity can generally be mapped with the embedding spaces of foundation models in vision, language, and audio. Although neural computations are highly non-linear at the microscale, fMRI measurements average signals across space and time, further smoothed by noise, effectively linearizing the observable representation. Consistent with these views, our experiments across multiple datasets demonstrate that linear contrastive decoders consistently outperform ridge regression and standard non-linear alternatives, and that these results generalize across images, text, and sound. These findings indicate that decoding gains arise more from the choice of training objective than from architectural complexity, pointing to contrastive-linear models as a principled strategy for brain decoding.

Identifying what individual neurons encode in higher-order visual cortex is an open problem. Responses resist intuitive parameterization, and the deep-network embeddings used in their place are black boxes. Here, we introduce NEURRATOR, a framework that decodes spiking activity into free-form natural-language narration of the viewed scene at single-neuron resolution. A learned encoder maps spike trains from arbitrary subsets of simultaneously-recorded neurons into the patch-embedding space of a frozen CLIP, from which a multimodal language model and sparse autoencoder generates and validates a description with no language-side training. Applied to Neuropixel recordings of mouse visual cortex during natural-movie viewing, NEURRATOR narrates from thousands of neurons, singular cortical regions, local populations, or from a molecularly-defined cell-types. We use this property to (i) quantify how decoding fidelity scales with population size and cortical region, and (ii) "neurrate", in plain language, what individual neurons and genetically-tagged inhibitory cell-types contribute to visual representation. This recasts cell identity from a classification target into a functional probe of the visual system, providing a new unit of biological insights in neural systems.

Forecasting epidemic trajectories is important for public health decision-making, but no single model is consistently reliable across epidemic phases and forecasting settings. We evaluate Multi-Armed Bandit (MAB)-inspired adaptive weighting strategies for combining epidemic forecasting models when component-model performance changes over time. Using U.S. COVID-19 incidence data from three epidemic waves, we compare UCB, EXP3, and epsilon-greedy weighting rules under fixed short-window and growing calibration windows, with both deterministic and stochastic ensemble variants. The model pool includes SIR, SEIR, GLM, Gompertz, Richards, ARIMA, random walk with drift, simple exponential smoothing, Holt's linear trend method, and exponential growth. Adaptive ensembles are compared with individual models and with naive, unweighted, and inverse-WIS weighted ensemble benchmarks. Forecast performance is assessed using RMSE, weighted interval score (WIS), 95% prediction-interval coverage, and mean 95% prediction-interval width. Across waves, calibration windows, and forecast horizons, EXP3Stoch, EXP3Det, and EPSStoch achieved the lowest mean forecast WIS. The main gains were in probabilistic forecast quality, especially WIS and interval coverage, rather than uniformly lower point forecast error. Simple benchmarks, including the unweighted and inverse-WIS ensembles, remained competitive in several settings. These results suggest that MAB-inspired adaptive weighting is a useful complementary tool for epidemic forecasting, especially when model skill is time-varying and forecast uncertainty is substantial.

High-throughput microfluidic live-cell imaging generates rich single-cell data. Yet semi-automated procedures for locating regions of interest (RoIs), each containing one cell population, and removing surrounding microfluidic structures from recorded images, scale with the number of RoIs. This prevents real-time image analysis and delays time-to-insight by hours to days. We introduce the Design-Aware and Real-Time capable (DART) paradigm for microfluidic cultivation chips, which aligns the CAD blueprint with the physical chip and thereby enables throughput-independent localization of all RoIs and fully automated image processing across diverse RoI geometries and chip layouts. DART establishes this alignment through embedded fiducial markers and deep-learning-based marker detection. We validate DART using the Swiss Army Knife chip, which combines eight structurally distinct RoI designs across 1164 RoI locations. DART localizes all RoIs in five minutes, removes microfluidic structures from raw microscopy images in 40 ms, and performs fully automated image analysis, including cell segmentation, in under 1.1 s per image. Together, these capabilities establish DART as an end-to-end hardware-software paradigm with real-time-capable analysis that paves the way toward closed-loop and outcome-driven smart microscopy.

Nitrous oxide (N2O) emissions from water resource recovery facilities (WRRFs) fluctuate over time and can arise from multiple microbial pathways, making source attribution and full-scale prediction difficult. The difficulty is compounded by the high dimensionality of activated sludge microbiomes, whose complex and dynamic community structure can obscure relationships with N2O emission patterns. This study evaluated whether interpretable, low-dimensional representations of activated sludge microbiomes can be correlated with N2O emission states. Temporal 16S rRNA gene amplicon profiles and N2O emission metrics were collected from two full-scale WRRFs in Switzerland. Genus-level relative-abundance profiles were summarized using archetypal analysis (AA), which represents each sample as a convex combination of a small number of interpretable community profiles. In both WRRFs, three archetypes captured most explainable variation in community composition (63%--73%) and defined a simplex state space in which samples clustered near vertices and edges, indicating that community compositions were organized around distinct archetypal states and their mixtures. Without using emission labels while training, the archetypal state space aligned strongly with binary N2O emission states: high-emission observations in both plants concentrated around a specific archetype, and temporal trajectories showed consistent high weights of this archetype during high-emission periods. Functional summaries suggested site-specific but pathway-relevant interpretations of the high-N2O archetype. Temperature further structured the archetypal state space, indicating seasonal forcing of microbiome configurations associated with elevated N2O. Overall, AA provides an interpretable framework to track microbiome regime shifts and may support operational tracking of high-N2O emission states in full-scale WRRFs.

Faster-than-Nyquist (FTN) signaling is gaining attention as a smart way to pack more data into limited spectrum by intentionally breaking the traditional symbol-spacing rules. This article takes a fresh look at FTN's potential to boost capacity, examining how performance varies across different acceleration factors and signal-to-noise ratio (SNR) definitions. Beyond the theory, we explore what it takes to make FTN work in practice, such as dealing with power amplifier constraints, managing high peak-to-average power, and designing practical coding strategies. We also highlight real-world issues like spectrum sharing, short-packet communication, and receiver complexity. With applications ranging from low-latency links to integrated sensing and satellite systems, FTN offers a compelling path forward for future wireless technologies.

Measuring a quantum computation in a basis adapted to a symmetry it carries reduces the repeated measurements, commonly referred to as ``shots'', needed to read a statistical answer. Detecting the symmetry a quantum state carries has many uses: certifying a claimed symmetry, identifying a conserved-charge sector, flagging symmetry-breaking as an error signature, and selecting a compression or readout basis; shot-count reduction is developed here as one exemplary case. Existing methods assume the symmetry is known in advance; we remove that assumption. When it is unknown, the carried symmetry is discovered from the data by a symmetry test that scores candidate groups, and the largest passing group is exploited as the measurement basis. We state the pipeline precisely, prove the selection rule is unbiased, and charge discovery in full. Two conditions are treated, both detected by the same score with a different projection: a weak condition, commutation with the representation, and a strong condition, confinement to a single charge sector, the distinction drawn in the quantum-reference-frame literature. A single circuit, a controlled twirl followed by a SWAP test, discovers both: discarding the group register tests the weak condition, post-selecting it the strong one. The framework is general over finite groups, with cyclic (Fourier), dihedral, and symmetric-group (Schur-Weyl) examples; strong confinement to the symmetric, or Dicke, subspace is an exponential reduction. Seeded demonstrations show the loop wins net of discovery: weak matching on momentum readout reduces shots by a factor widening from ten to several thousand, and strong matching on a two-system target by a further factor of the subsystem size. Blind symmetry matching is a practical primitive for the common case where the matched basis cannot be written down in advance.

Estimation of uplink channels is required for coherent over-the-air computation (OAC). When channel estimation is done using calibrated reciprocity, the estimates are only available locally to the devices. This poses a challenge for precoding and decoding, which cannot be coordinated centrally. To this end we use truncated channel inversion (TCI) and propose an approximate closed form solution and an exact numerical solver to optimize the TCI parameters. Importantly, we prove that the proposed TCI scheme is independent of the number of receiver antennas in terms of mean-square-error (MSE). Furthermore, our analysis reveals a clear connection between the MSE and expected aggregate phase error across devices which gives insight to the scalability of OAC. Finally, simulations with comparisons to reference methods from prior work with globally available error-free channel estimates show that proposed is close, even outperforming these references in MSE under some conditions.

The advancement of 6G mobile communication and positioning technologies has amplified the significance of location-aware tools, such as location-indexed channel fingerprints (CFs) and channel charting, which are becoming key enablers for massive MIMO-OFDM systems. In this paper, we propose a novel channel charting with physical CFs (PCFs) and demonstrate its effectiveness in channel state information (CSI) acquisition. First, we define the PCF based on a cluster-based geometric stochastic channel model (GBSM), enabling a comprehensive representation of physical channel characteristics using a compact set of parameters. We then develop a methodology for PCF acquisition in massive MIMO-OFDM systems. By exploiting the relationship between PCFs and the space-frequency-time (SFT) domain channel, the proposed method extracts PCFs from multi-location channel measurements and constructs a structured channel charting with location-indexed PCFs. Furthermore, we propose a low-complexity algorithm to acquire beam domain statistical CSI (sCSI) using the PCFs in the channel charting. The resulting sCSI can be directly employed as prior information for channel estimation. Simulation results show that the proposed method delivers sCSI performance comparable to traditional online probing techniques, and the generated sCSI can serve as reliable prior knowledge to significantly enhance the accuracy of channel estimation. These results validate the proposed PCF as a powerful and versatile tool for channel acquisition and system design of the next-generation mobile communication.

Spaceborne synthetic aperture radar (SAR) provides coherent microwave imagery suitable for maritime infrastructure monitoring under illumination-independent and weather-independent acquisition conditions. An academic conference-style analysis is presented for SAR amplitude and geocoded multitemporal data over Tianjin Port, China. The processing chain includes amplitude visualization, radiometric scaling, view-direction interpretation, range and azimuth resolution assessment, speckle reduction, amplitude-based change mapping, GeoTIFF export for geographic inspection, and interferometric coherence estimation. Histogram-guided display limits improve the interpretability of the complex SAR magnitude images, while zoomed inspection of shadows and bright layover responses supports qualitative interpretation of illumination geometry. A two-dimensional Fourier analysis is used to characterize dominant spectral content and to estimate an approximate range resolution of 0.42 m and an azimuth angular separation of 0.19 degrees under the available image-coordinate calibration. Multitemporal master and slave images are subsequently compared through filtered amplitude differences and coherence maps computed with multiple spatial averaging windows. The results highlight the relevance of SAR amplitude and coherence products for detecting structural and surface-condition variations in dense port environments characterized by vessels, storage tanks, quay structures, industrial yards, and water-land transitions.

In this article, we address the problem of spectrum scarcity in cellular networks (CNs). We propose a backup channel (BuC) for cellular users (CUs) located in the same macro-cell under the control of a single macro base station (eNB). This BuC operates in television white space and is detected by the CUs through a cognitive radio energy-detection channel-sensing technique with a certain probability of success. When all regular channels with the cellular eNB are occupied, the CUs within the same coverage area of the macro eNB can utilize the sensed BuC to establish a controlled out-of-band device-to-device link for communication. The BuC bypasses the eNB for data communication and reduces the burden on the core of the CN. This leads to improved cellular eNB capacity. In the proposed system model, each CU and eNB is equipped with two antennas for communication in two separate bands, i.e., cellular and TV bands. Simulations show significant reductions in the blocking probability and probability of call delay.

A rotatable antenna (RA)-enhanced secure integrated sensing and communications system is investigated, where an RA-based transceiver simultaneously communicates with legitimate users and senses a target that is regarded as a potential eavesdropper. Under imperfect eavesdropping channel state information (CSI), a max-min data rate optimization problem is formulated by jointly optimizing the transmit beamforming, artificial noise (AN) covariance matrix, and transmit/receive boresights of RAs, subject to the maximum information leakage and minimum sensing power constraints. To address the highly non-convex problem, the information leakage and sensing power constraints are transformed into convex ones via S-Procedure method and Cauchy-Schwarz inequality, respectively. Subsequently, an alternating optimization algorithm is developed to decompose the reformulated problem into two subproblems. In particular, the transmit beamforming and AN covariance matrix are optimized by utilizing successive convex approximation and semi-definite relaxation methods, while the RA boresights are obtained by invoking the particle swarm optimization. Simulation results show that the RA-based scheme significantly outperforms the benchmarks, and offers enhanced robustness against imperfect CSI with the increase of the maximum rotation range.

Federated learning (FL) in energy-harvesting (EH) networks is challenged by intermittent and stochastic energy arrivals that lead to unstable device participation across training rounds, and by high communication costs under limited energy budgets, reducing overall training efficiency. This paper studies FL under a slot-based EH model and proposes EH-FedSAG, a server-memory-based variance-reduced method. We compare EH-FedSAG with vanilla EH-FedAvg under the same multi-channel orthogonal multiple-access uplink model and within a unified simulation framework that captures battery charging, local computation cost, and transmission cost under different energy-arrival probabilities. Performance is assessed in terms of test accuracy over training rounds for both homogeneous and heterogeneous data distributions. The results show that EH-FedSAG consistently achieves higher test accuracy than EH-FedAvg in the considered settings, while exhibiting substantially lower training variance. The advantage of EH-FedSAG is more pronounced under scarce energy availability and non-independent/identically-distributed data.

Since the inception of electrical recording for phonograph records in 1924, records have been intentionally cut with a non-uniform frequency response to maximize the information density on a disc and to improve the signal-to-noise ratio. To reproduce a nominally flat signal within the available bandwidth, the effects of this cutting curve must be undone by applying an inverse curve on playback. Until 1953, with the introduction of what has become known as the RIAA curve, the playback curve required for any particular disc could vary by record company and over time. As a consequence, anyone seeking to hear or restore the information on a disc must have access to equipment that is capable of implementing multiple playback equalizations. This correction may be accomplished with either analog hardware or digital processing. The digital approach has the advantages of reduced cost and expanded versatility, but requires a transformation from continuous time, where the original curves are defined, to discrete time. This transformation inevitably comes with some deviations from the continuous-time response near the Nyquist frequency. There are many established methods for discretizing continuous-time filters, and these vary in performance, computational cost, and inherent latency. In this work, several methods for performing this transformation are explored in the context of phonograph playback equalization, and the performance of each approach is quantified. This work is intended as a resource for anyone developing systems for digital playback equalization or similar applications that require approximating the response of a continuous-time filter digitally.

The quantity that defines the behavior of a dynamic range compressor is the time-varying gain applied to the signal as a function of the input level. However, models of these devices are typically evaluated using proxy metrics because isolating the gain reduction signal from the audio input-output data included in existing datasets creates an ill-conditioned inverse problem. It is unclear how accurately these metrics describe the behavior the model is tasked with emulating, particularly as waveform-based metrics can be influenced by secondary effects introduced by analog processing and capture, even when those effects are inaudible. We investigate a method of evaluation in which the gain-reduction signal produced by a model is measured directly against a gain-reduction control voltage signal produced by the hardware. To evaluate the efficacy of this metric as a learning objective, a gray-box model is trained using loss computed directly over the gain control signals alongside two models trained using common proxy losses. The models trained using proxy losses did not achieve parity with models trained directly on the gain control signal when evaluated with respect to the underlying control trajectory, and the waveform-domain metrics assigned similar errors to models that were clearly separated by the direct metric. To facilitate further exploration of this method of evaluation, we present a Solid State Logic bus compressor dataset that includes the gain control voltage signal captured alongside the audio output.

An experimental investigation of neural image classification on the CIFAR-10 benchmark is presented through fully connected and convolutional network formulations. The analysis emphasizes the complete learning pipeline: image vectorization, normalization, one-hot class encoding, supervised loss minimization, learning-rate selection, mini-batch training, convolutional feature extraction, max-pooling, and validation-based generalization assessment. A convolutional architecture with six convolutional layers and three max-pooling stages is evaluated for ten training epochs using a batch size of 128 and an Adam optimizer with a learning rate of 0.001. The validation accuracy reaches approximately 74.77%, while the validation loss begins to increase after the middle of training despite continued reduction in training loss. The resulting behavior illustrates the practical difference between representation learning and memorization, and it provides a compact experimental baseline for future studies on regularization, data augmentation, deeper architectures, and reproducible image-classification education.

Reference-based adaptive interference cancellation is evaluated for stereo audio recordings corrupted by real train noise and environmental background. The observed signal is modeled as a clean stereo program contaminated by an additive disturbance generated by an external acoustic source through unknown propagation paths. A second stereo recording, representing another filtered observation of the same physical noise source, is used as the reference input of a multi-reference recursive least-squares (RLS) estimator. The estimated train-interference component is subtracted from the noisy audio and followed by a finite-impulse-response low-pass postfilter. Three 74.01 s real audio sequences sampled at 11.025 kHz are processed under identical algorithmic parameters. Since clean ground truth is not available, performance is assessed with no-reference indicators: waveform behavior, Welch spectral estimates, RMS change, and residual normalized correlation with the reference. With 30 taps per reference channel, 15 anti-causal taps, and forgetting factor 0.999, the maximum reference correlation is reduced from 0.386--0.832 before processing to 0.011--0.016 after processing. The corresponding correlation-ratio reduction is approximately 30.6--34.1 dB, while the output RMS decreases by 1.8--4.8 dB depending on section and stereo channel. The results demonstrate that real train interference, including environmental acoustic effects, can be substantially attenuated when a correlated reference recording is available.

Unmanned aerial vehicle (UAV)-mounted base stations are highly susceptible to wind disturbances such as gusts and turbulence, which induce positional drift and degrade communication link quality, particularly in emergency scenarios. To address this challenge, we propose a DRL-based framework for wind-resilient trajectory adjustment and positioning based on the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm. The method models wind as a stochastic kinematic perturbation, avoiding complex aerodynamic modeling, thereby enabling the TD3 agent to learn adaptive control policies that maintain optimal coverage footprints. By prioritizing user-centric performance metrics under turbulent conditions, the proposed architecture ensures continuous service availability despite external disruptions. Simulation results demonstrate that the TD3-based approach effectively compensates for wind-induced displacements and outperforms benchmark methods, including Proximal Policy Optimization (PPO), in terms of throughput stability and robustness in windy environments.

Cell-free (CF) integrated sensing and communication (ISAC) merges the CF architecture with ISAC functionalities. CF-ISAC leverages distributed access points, removes cell boundaries, and enhances coverage, spectral efficiency, and reliability. It also improves energy efficiency, enabling robust multi-user communication, distributed multi-static sensing, and seamless resource optimization. A comprehensive survey on CF-ISAC has been lacking. This monograph addresses that gap by covering the foundational principles, cooperative transmission, radar cross-section, target parameter estimation, ISAC integration levels, sensing metrics, and key applications. It also explores the advantages of multi-static sensing. Performance analysis, resource allocation, security, and user/target-centric designs are discussed. Finally, synchronization, multi-target detection, interference management, and fronthaul limitations are discussed. Advanced antenna technologies, network-assisted systems, near-field CF-ISAC, cross-technology integration, and machine learning approaches are presented.

This paper investigates covert multi-hop communications in heterogeneous wireless networks monitored by multiple passive wardens. To maximize network-wide covertness while satisfying a strict end-to-end rate requirement, we jointly optimize routing, modality selection, and transmit power. Under a simultaneous multi-hop transmission scheme, we analyze the detection capabilities of two distinct warden models: colluding wardens employing a central fusion center, and non-colluding wardens operating independently. For both models, we derive optimal detectors and exact expressions for the detection error probability (DEP). In addition, to reduce the complexity of evaluating the DEP, we develop highly accurate closed-form approximations based on gamma moment matching and establish rigorous DEP lower bounds using Kullback-Leibler (KL) divergence. Building on this theoretical foundation, we propose an efficient two-stage optimization algorithm that decouples link-level resource allocation from network-level path selection. By translating the KL divergence bounds into a novel, low-complexity routing metric, which universally simplifies to a linear summation of signal-to-noise ratios, we substantially reduce the computational overhead compared to conventional per-hop detection-based metrics. Finally, numerical simulations validate the theoretical analysis and demonstrate the near-optimal performance of the proposed framework.

In this paper, we propose a general-purpose multi-dimensional symbol construction for computing an arbitrary symmetric function with digital over-the-air computation (OAC) and discuss the practical aspects of coherent aggregation. For our first contribution, we discuss the categorical representation of a symmetric function. By using this representation and leveraging the sufficiency of the histogram to evaluate a symmetric function, i.e., inspired by type-based multiple access (TBMA), we introduce a general approach to design a single set of OAC symbols to compute any digital function. For our second contribution, we use a comprehensive platform based on low-cost nodes that maintain synchronization in time, frequency, phase, and amplitude via a trigger mechanism, enabling coherent OAC experiments without Global Positioning System (GPS) or cable-based synchronization. Using measurements from the platform, we characterize the phase and amplitude statistics of the composite channel to derive a realistic impairment model for coherent OAC. Through a comprehensive analysis, we demonstrate the effectiveness of the proposed scheme under impairments captured by the proposed model

Asthma causes expiratory airflow limitation and is clinically assessed using spirometry, which provides the FEV1/FVC ratio representing the proportion of air exhaled in the first second relative to total forced vital capacity. Prior studies suggest that respiratory sounds recorded at posterior sites (Left Lower, Left Upper, Right Upper, Right Lower) reflect regional airflow patterns. In this study, we investigate the relationship between the expiratory-to-inspiratory (E/I) spectral power ratio and FEV1/FVC in 141 participants aged 20-60 years using Spearman correlation across frequency subbands. The 100-200 Hz and 200-400 Hz bands showed significant correlations. Overall, lower posterior sites showed stronger associations; younger adults showed stronger correlations at the Left Lower site, whereas older adults showed stronger correlations at the Left Upper site. Gender-stratified analysis showed stronger Left Lower correlations in males and stronger Left Upper correlations in females.

Superconducting qubits are artificial atoms whose spectra and interactions can be engineered through appropriate circuit design, a versatility that can be exploited for quantum simulation. We theoretically investigate a linear array of capacitively coupled transmons, effectively described by a Bose-Hubbard Hamiltonian with attractive interaction. We revisit the discrete-soliton nature of the lowest-energy band of the spectrum, and identify spatially localized quantum solitons. The solitonic character of these states is revealed through their time evolution, which displays a quantum interference pattern, or quantum walk, highlighting their composite nature. We discuss protocols for preparing spatially localized quantum solitons that are compatible with current state-of-the-art tunable-transmon circuits. Our results demonstrate that superconducting circuits provide a promising and experimentally accessible platform for the investigation of quantum soliton physics.

A network of cosmic strings is one of the few well-motivated cosmological sources of vector and tensor metric perturbations on the largest observable scales. Such perturbations imprint a characteristic curl component in the deflection angle of cosmic microwave background (CMB) photons that, unlike the scalar lensing potential, vanishes for adiabatic density fluctuations at linear order. We exploit the curl-mode lensing reconstruction released as part of the Atacama Cosmology Telescope (ACT) Data Release~6 (DR6), based on five seasons of temperature and polarization data covering $9400~\mathrm{deg}^2$ of sky, to set new constraints on the dimensionless string tension $Gμ$ and the inter-commutation (reconnection) probability $P$. Modelling the string-induced curl power spectrum within the velocity-dependent one-scale framework, we obtain a $2σ$ upper bound on the combination $GμP^{-1}\le 3.5\times 10^{-5}$ in the small-$P$ regime, and $Gμ\le 5.0\times 10^{-5}$ at $2σ$ assuming the canonical Nambu-Goto value $P=1$. Combining the ACT DR6 curl bandpowers with the Planck 2013 curl-mode reconstruction, which extends down to $L_{\rm min}=2$, tightens these bounds to $GμP^{-1}\le 3.2\times 10^{-5}$ and $Gμ\le 4.3\times 10^{-5}$ ($2σ$). These represent the tightest constraints on cosmic strings derived from the curl-mode CMB lensing power spectrum to date. Using the ACT data alone, compared to the ACT 2008-season analysis, the ACT DR6 constraint on $GμP^{-1}$ is nearly an order of magnitude tighter.

In equilibrium physics, topological defect insertions in quantum and classical partition functions provide non-perturbative probes of phase transitions beyond local observables. In non-equilibrium physics, the spectral form factor provides a minimal probe of universal quantum dynamics, and admits a representation as a product of two partition functions at imaginary inverse temperature. We define the topological spectral form factor (TopSFF) by inserting topological defects acting non-trivially on the doubled partition functions, producing mismatched spacetime world-sheet topologies. For the minimal $\mathbb{Z}_2$ spatially extended defect, implemented by the global swap operator, we derive an exact mapping of the TopSFF of a generic 1D many-body chaotic system to an emergent $(3+1)$D non-Hermitian single-particle problem describing a temporal domain wall (tDW). We show analytically that the effective tDW dynamics undergoes a $\mathcal{PT}$ symmetry breaking transition at a finite interaction strength $ε_{\mathrm{EP}}$: below $ε_{\mathrm{EP}}$, the leading modes are polarized into Gaussian or non-Gaussian tDW sectors and the TopSFF varies monotonically and exponentially with system size; above $ε_{\mathrm{EP}}$, the tDW sectors hybridize and the TopSFF oscillates with system size; at the exceptional point $ε_{\mathrm{EP}}$, Jordan non-diagonality produces a linear-in-system-size enhancement. For temporally extended topological defects, we derive exact universal scaling forms for the TopSFF free energy in systems with time reversal or time translation symmetry, and verify them numerically in independent models.

We present an analytical and numerical Floquet treatment of a driven polar two-level quantum system characterized by both longitudinal and transverse coupling to a periodic field. Analytically, we derive a dressed-frame effective Hamiltonian up to first order in the inverse driving frequency, incorporating the longitudinal coupling nonperturbatively. This yields closed expressions for the effective transverse coupling strength and the effective detuning, both of which are modified by the presence of the longitudinal interaction. In the nonpolar limit, these expressions recover the usual near-resonant Rabi coupling and the Bloch-Siegert shift. As a second main result, we develop a numerical flow-equation framework that yields a time-independent effective Hamiltonian across a broad range of transverse and longitudinal coupling strengths. This dual framework is relevant for a variety of platforms, including driven polar quantum systems, optical lattices, superconducting circuits, and solids subject to surface acoustic waves.

We compute spectral functions of graviton modes in Lorentzian quantum gravity, interpolating between classical general relativity and an asymptotically safe ultraviolet fixed point. Using functional renormalisation adapted for theories in Lorentzian signature, and enhanced by new symmetry conditions to account for underlying Ward identities, we derive and solve flow equations directly for the Källén-Lehmann representation of propagators. Consistent results are found for several sets of renormalisation conditions yielding normalisable spectral functions for the graviton and the scalar graviton mode, in agreement with effective theory in the infrared. We further calculate the full quantum effective action to quadratic order in curvature, extract graviton-induced form factors, and discuss implications for unitarity of quantum gravity.

This dissertation studies the electromagnetic signatures of accreting supermassive black holes using general relativistic magnetohydrodynamic simulations and covariant radiative-transfer calculations. It develops a unified numerical framework for modeling black-hole accretion, jet launching, flaring activity, and multi-band variability in Kerr, non-Kerr, and binary black-hole spacetimes. For isolated Kerr black holes, I investigate how magnetic-field geometry affects accretion dynamics and transient emission. Multi-loop magnetic configurations naturally produce reconnection events and flux-rope structures that can power near-infrared flares from Sagittarius A*, while the evolving optical depth of expanding plasma explains delayed millimeter emission. I also show that in tilted magnetically arrested disks, magnetic torques can drive retrograde disk and jet precession. The dissertation then applies the same framework to more complex spacetimes. Simulations of accretion onto regular loop-quantum black holes show that quantum-gravity corrections can modify photon-ring size, polarization structure, and jet power, leading to observational constraints from Event Horizon Telescope data. Finally, simulations of supermassive binary black holes in time-dependent spacetimes reveal how gravitational self-lensing, shock activity, and spin-orbit coupling shape multi-wavelength light curves and jet precession. Together, these results connect relativistic plasma dynamics with current and future observations of black-hole systems.

We present a class of horizonless, regular ultra-compact objects arising in a theory of gravity which allows curvature-fluid coupling. The non-minimal interaction between fluid variables and the Ricci scalar generates a vacuum-like equation of state in the interior, while the exterior remains exactly Schwarzschild. The two spacetimes are glued through a shell at the junction. The interior metric is non-singular, the shell acquires a stiff-matter equation of state, and near-horizon compactness can potentially mimic black-hole phenomenology without event horizons. Unlike the Mazur-Mottola gravastar and its variants, the present model naturally selects a typical ultra-compact mass-radius window, with masses in the range $1.4$-$2.1 M_\odot$ and radii in the range 5-7 km. This framework predicts a unique geometric-thermodynamic shell temperature in the ultra-compact limit distinctly different from the Hawking expression and the other unique observational feature of the model is the prediction of mass independent luminosity.

Black hole static Love numbers vanish, but their dynamical counterparts do not. We present the scheme-independent dynamical response $\bar{F}_{\ell,s}$ of a Schwarzschild black hole in closed form, to all orders, and for every spin $s$ and multipole $\ell$. The result is $\bar{F}_{\ell,s}/4πR_S^{2\ell+1}=Φ_{\ell,s}(\bar{y})-\tfrac12η\,Φ_{\ell,s}'(\bar{y})$ with $\bar{y}=-\tfrac12η^2τ$ and $η=iωR_S$. Here $Φ_{\ell,s}$ is simply the leading-log solution to the renormalization group equation, but lifting the running logarithm to $τ=\log(R_S/R)-2\sum_{k\ge2}ζ_k\,η^{k-1}$ resums it to all orders. This tower of Riemann zeta values is the Newtonian phase in disguise: it originates from the same far-zone $Γ(1-η)$ that governs long-range scattering, and is universal across multipole and spin. Our result exhibits a factorization pinned to three ingredients: the hard matching coefficient at the horizon, the anomalous dimension in the near zone, and the dressed log in the far zone. Using shell effective field theory, we independently verify our formula for scalar, electromagnetic, and gravitational perturbations, reaching $\mathcal O(G^{15})$.

We present numerical simulations supporting the random-matrix state-reduction framework developed in the companion theoretical paper. The simulations test the main derived features of the model: isotropic diffusion generated by Gaussian Unitary Ensemble Hamiltonians in projective state space, the restriction of this diffusion to Brownian motion on the classical submanifold, Born-rule frequencies for detector-defined outcome classes, and stroboscopic Newtonian motion for macroscopic systems under repeated environmental monitoring. We also compare GUE and GOE random Hamiltonians and show that GOE fails to produce the required isotropic complex projective diffusion. Further simulations examine finite-resolution detector records in the double-slit experiment, Zeno stability of recorded equivalence classes, effective irreversibility from high-dimensional state-space dynamics and loss of path information, and tensor-product particle-device dynamics in the device limit. The results show that microscopic state reduction, stable measurement records, effective irreversibility, and macroscopic classicality can be described as different coarse-grained manifestations of the same stochastic unitary mechanism.

We develop a state-space geometric framework for measurement, classicality, and quantum paradoxes, based on one dynamical conjecture. Classical configuration space and classical phase space for a mechanical system arise as distinguished submanifolds of projective quantum state space. On these submanifolds, the Fubini--Study geometry induces Euclidean classical geometry, and the tangent component of Schrödinger evolution reproduces Newtonian dynamics. Within this framework, interactions with measuring devices and environments are described by random-matrix dynamics on projective state space, generated by matrices drawn from the Gaussian Unitary Ensemble. We show that this random-matrix dynamics yields isotropic diffusion, giving Born-rule transition probabilities in microscopic measurements and stabilizing classical behavior in macroscopic systems. We further argue that the random-matrix conjecture is not an independent ad hoc assumption: under natural translation-invariance assumptions on the distribution of state-space steps originating on the classical submanifold, the unitary lift of homogeneous and isotropic Brownian motion on that submanifold is uniquely given by the Gaussian Unitary Ensemble, up to scale and an irrelevant scalar part. The resulting framework provides a unitary account of measurement and the quantum-to-classical transition and, if accepted, offers a dynamical resolution of standard quantum paradoxes.

We introduce algorithmic improvements to quantum-classical auxiliary-field quantum Monte Carlo (QC-AFQMC) that reduce the dominant per-step classical scaling from $\tilde{\mathcal{O}}(N^{5.5})$ to $\tilde{\mathcal{O}}(N^{4.5})$ as a function of the number of molecular spin-orbitals $N$. Central to this improvement is the application of Aitken's block transformation to handle singular Pfaffians arising in the estimation of overlaps between a quantum trial state and classical Slater-determinant walkers. Together with the use of algorithmic differentiation for the computation of the force bias, this yields a $248\times$ estimated runtime improvement for a system of 100 molecular orbitals. Using our workflow, we demonstrate a ground-state energy calculation for $H_8$ from quantum data collected on IQM Emerald and post-processed with a tensor-network-based error-mitigation technique. We further validate the method's scalability through noiseless simulation of hydrogen chains up to $H_{12}$, and on the lithium-air battery related rearrangement pathway of the $Li_2O_4$ lithium superoxide dimer in a (26e, 20o) active space. We estimate both quantum and classical runtimes for a potential fault-tolerant implementation of QC-AFQMC, showing that the method holds promise for the early fault-tolerant era. These results move QC-AFQMC a step closer to treating chemically relevant systems.

We show that the near-horizon expansion of the linear polarization vector for synchrotron emission in a Kerr background admits a distinct analytic form. For emission from a stationary, axisymmetric, degenerate electromagnetic field, the leading-order polarization pattern depends only on the Kerr spin and the source polar angle, while the next-to-leading-order correction further encodes the geometric and rotational structure of the electromagnetic field. Our result extends the equatorial analysis of [Hou et al. (2024)] and the off-equatorial leading-order result of [Chael et al. (2026)]. Near-horizon polarization thus offers a potential probe of the fundamental properties of rotating black holes and of gravito-electromagnetic interactions.

Here we investigate the all-optical control of spectral correlations in spontaneous parametric down-conversion. We show that when photon pairs are projected onto high-order spatial modes, the spatial structure of the pump field defines the phase-matching function of the nonlinear interaction. Thus, by structuring the pump field in both space and spectrum, the biphoton spectral correlations are fully controlled.Considering a standard periodically-poled crystal as the nonlinear medium, we show that the Gouy phase matching method proposed here can generate both spectrally uncorrelated and high-dimensional spectrally entangled photon pairs, similarly to what is achieved with aperiodically-poled crystals. Furthermore, we show that our method can generate a wider class of quantum states if the pump field is a spatiotemporal wavepacket, that is, if its spatial and spectral structures are correlated.

Quantum impurity models with strong electron correlations, such as the paradigmatic Anderson Impurity Model (AIM), are central to our understanding of a range of physical phenomena including local moment formation, Coulomb blockade and Kondo screening. They describe magnetic atoms and molecules on surfaces, quantum dot circuits, and correlated materials through dynamical mean field theory. The physics of such systems in strongly non-equilibrium conditions is particularly complex and challenging to capture, whereas Gaussian models of free fermions can be easily solved. Here we show that the time-evolving dynamics of the AIM after a quench can be described by a completely non-interacting version of the model, at the expense of coupling to additional static auxiliary degrees of freedom. Starting from the full solution of the quenched AIM using ED and DMRG, we study the properties of this mapping using numerical optimization, and uncover intriguing structure in the auxiliary system. The method allows us to understand interacting non-equilibrium dynamics through the simpler lens of an effective non-interacting system of larger dimension.

The geometry of the Einstein Telescope, the proposed next-generation European gravitational-wave observatory, is yet to be finalized. Two competing designs are under consideration: a nested triangular configuration (ET-Δ) and two separated L-shaped detectors (ET-2L). Extensive prior comparisons of ET designs established the scientific landscape using the Fisher-information-matrix formalism and identified that duty-cycle-induced single-detector operation is precisely the regime where this approximation becomes less reliable, underscoring the need for a refined, principled treatment of the duty cycle. In this manuscript, we build on that foundation by revisiting the comparison with full Bayesian parameter estimation of gravitational-wave signals from binary black-hole mergers, projected onto a simulated Einstein Telescope that incorporates a refined duty cycle modelled via continuous-time Markov chains and testing different detector maintenance strategies. We find that the redundancy inherent in the ET-Δ design enables it to maintain at least two operational detectors for the majority of the observing time, whereas the ET-2L configuration is often limited to a single detector. Crucially, we show that, during partial network operation, ET-Δ often outperforms ET-2L, and that the increased multi-detector uptime translates into tighter constraints on the luminosity distance and source-frame component masses. Notably, this remains true even when gravitational-wave events have a lower signal-to-noise ratio in ET-Δ than in ET-2L.

The emergence of classical behavior and the origin of thermal equilibrium are two central problems in the foundations of physics. In the standard accounts, both phenomena are typically explained through interactions with an external environment: decoherence suppresses quantum interference, while coupling to a thermal bath drives relaxation toward equilibrium. Over the last decades, however, it has become clear that equilibration and thermalization can arise even in fully isolated quantum systems, in the operational sense that the expectation values of relevant observables remain close to equilibrium values for most of the time. Here, we ask whether the same intrinsic equilibration mechanism can also account for the emergence of classical behavior. Using rigorous bounds on equilibration in closed systems, we derive sufficient conditions under which a time-evolved pure state becomes, for most times, operationally indistinguishable from a classical mixture associated with a chosen physical property. We identify two complementary routes to such operational classicality: either the chosen property almost commutes with the system Hamiltonian or the observables used to probe the system lose access to the remaining coherence after equilibration. Our results show that classical behavior need not be confined to the energy basis and may emerge even when substantial coherence remains present in the equilibrium state. This establishes a direct connection between two foundational questions: the origin of thermalization in isolated quantum systems and the quantum-to-classical transition

We investigate the dynamics of Quantum magic, formally known as non-stabilizerness, quantified by the stabilizer Rényi entropy (SRE), across the stages of the wormhole-inspired teleportation protocol (WITP) in the Sachdev-Ye-Kitaev (SYK) model. By tracking the SRE of the full pure state across scrambling, message insertion, left-right coupling, and right-side extraction, we uncover a regime-dependent relationship between magic accumulation and teleportation fidelity. In the gravitational (low temperature) regime, fidelity rises concurrently with magic from early times, whereas in the peaked-size (high temperature) regime, the magic saturates near the Haar-typical value before teleportation onset. A baseline-subtracted diagnostic comparing coupled and uncoupled protocols reveals that the double-trace coupling first suppresses and then channels non-stabilizer resources toward the teleportation signal, with the channel amplitude decreasing monotonically with inverse temperature. Comparison with a chaotic random two-local model, which generates near-maximal magic yet fails to teleport, demonstrates that structured magic redistribution, rather than raw non-stabilizerness, underlies successful wormhole traversal. Moreover, the magic transiently dips at the fidelity peak, marking the teleportation event in the time domain. Our results are robust across the three system sizes studied ($N_{\mathrm{maj}}=8,10,12$), and the fidelity-magic trajectories exhibit an approximate collapse across system sizes when the SRE is normalized by the Haar-typical prediction.

Ground-based gravitational-wave detectors operating in the sub-Hertz regime are expected to be strongly limited by environmental gravity-gradient fluctuations, commonly referred to as Newtonian Noise (NN). At the same time, this frequency band provides unique opportunities to probe terrestrial gravitational perturbations associated with seismic and atmospheric processes. In this work, we investigate the feasibility of using the proposed Cryogenic sub-Hz cROss torsion-bar detector with quantum NOn-demolition speed meter (CHRONOS) as a platform for studying gravity-gradient noise and detecting prompt gravitational signals from earthquakes. We model gravity-gradient contributions from Rayleigh-wave-induced seismic fields, atmospheric infrasound fluctuations, and transient mass redistribution during earthquakes, and project these onto the CHRONOS torsion-bar response. CHRONOS achieves a peak strain sensitivity of order ~1e-18 Hz^(-1/2) near ~2 Hz. Rayleigh-wave NN is found to be the dominant environmental contribution below approximately 0.5 Hz, while atmospheric NN remains several orders of magnitude smaller throughout the frequency range considered. We further assess the detectability of prompt gravitational signals from earthquakes. For a representative Mw = 5.2 event, sources within approximately 90 km may produce detectable signals. At 40 km distance, we obtain a signal-to-noise ratio (SNR) of approximately 3.62 integrated over the sub-Hz band, with a corresponding strain amplitude reaching the CHRONOS sensitivity curve around 0.2 to 0.6 Hz. The gravitational signal is expected to precede seismic P-wave arrival by several seconds, depending on the assumed propagation velocity. These results demonstrate the potential of CHRONOS to probe both gravity-gradient noise and transient geophysical gravity signals in the sub-Hertz regime.

Kiselev solution is a metric that describes black holes immersed in a quintessence-like dark energy background. By introducing a dynamic state parameter $w_q$, the Kiselev solution is supposed to help comprehend the effect of quintessential matter on black holes. In this work, we study the evaporation behaviors of Kiselev black holes. By varying the state parameter $w_q$, we find that the decreasing state parameter lowers the non-final stage temperature and markedly prolongs the evaporation lifetime. We also find that the ultra-slow evaporation mechanism of Kiselev black holes differs vastly from the perfect fluid dark matter (PFDM) black holes and Horndeski black holes, which share the analogous ultra-long lifetime. These results illuminate the effects of dynamic dark energy background on black hole evaporation, provide a potential laboratory to constrain the value of $w_q$, and may complement cosmological and astrophysical observations, e.g., the DESI's preference for thawing dark energy and the observation of exploding black holes based on ultra-slow evaporation.

In the JWST and DESI era, the JWST high-redshift galaxy observations and DESI baryon acoustic oscillation (BAO) measurements severely challenge the standard $Λ$CDM model, while the $H_0$ tension becomes increasingly prominent. In this work, we investigate the capability of the early dark energy (EDE) model to alleviate the $H_0$ tension utilizing cosmic microwave background data from Planck, ACT, and SPT, BAO data from DESI, and ultraviolet luminosity function observations from the JWST. Within the canonical axion EDE framework, the CMB+DESI+JWST data significantly increase the $H_0$ value to $71.58\pm1.05\,\mathrm{km\,s^{-1}\,Mpc^{-1}}$, alleviating the $H_0$ tension to the $1.0σ$ level. Simultaneously, this model improves the fit to the JWST data and exhibits statistical performance significantly better than the $Λ$CDM model, with $Δχ^2_{\mathrm{tot}} = -18.26$ and $Δ\mathrm{DIC} = -11.89$. Our results highlight the complementary advantages of JWST high-redshift galaxy data alongside early- and late-time observations in testing EDE and alleviating the $H_0$ tension.

We investigate quantum pump depletion and non-Gaussian state generation in doubly pumped Kerr microresonators operating in the intraresonance regime. The pump modes are treated quantum mechanically rather than as undepleted classical amplitudes, allowing pump depletion, back-action, entanglement generation, quadrature fluctuations, and Wigner-function negativity to emerge from the same multimode dynamics. Starting from the Kerr four-wave-mixing selection rule, we distinguish an effective resonant photon-conversion model from the full Kerr Hamiltonian containing self-phase modulation (SPM), cross-phase modulation (XPM), and four-wave mixing (FWM). The reduced model isolates the photon-conversion network responsible for the discrete $\mathbb{Z}_{n+1}$ phase structure, whereas the full model retains operator-valued nonlinear Kerr phases. For the \(n=2\) intraresonance branch, the four-mode reduced initial-value problem with fixed coherent pump phases has a residual \(\mathbb{Z}_3\) symmetry and generates cat-like Wigner structures near the interaction length at which the generated-mode population \(\langle n_1\rangle\) is maximal and the pump population \(\langle n_0\rangle\) is strongly depleted. The resulting states are not the canonical even or odd coherent states of Dodonov, Malkin, and Man'ko, but multicomponent Schrödinger-cat-like states characterized by Wigner negativity, non-Poissonian statistics, pump-mode quadrature squeezing, and large single-mode Schmidt numbers. Comparison of the reduced and full Kerr dynamics shows that uncompensated SPM/XPM-induced phase shearing suppresses the interference fringes and Wigner negativity responsible for the clearest cat-like signatures. These results identify quantum-depleted intraresonance Kerr dynamics as a route to symmetry-organized non-Gaussian states in Kerr resonators.

We develop a block-encoding compiler that speeds up linear combination of unitaries Hamiltonian simulation programs by treating matrix product operators as compressed, virtual-path LCU programs. In showing how these new conditional PREP and SELECT stages are compiled in terms of a parent matrix product operator, we go beyond typical operator splitting product formulas and illustrate how tensor networks are a natural language and valid intermediate representation for quantum circuits. Our results are numerically verified for two important cases, namely, Heisenberg and perturbed Heisenberg-adjacent chain real-time evolution, and highlight polynomial speedups. Specifically, we highlight a polynomial speedup that avoids the $\mathcal{O}(N^K)$ Pauli-string growth when the compressed MPO bond dimension and path normalization remain mild. We quantify how MPO truncation error and bond-dimension budgets affect the compiled polynomial representation. Our algorithms show how classical pre-processing in terms of tensor network data structures opens new avenues to accelerate quantum algorithms.

We present a method for performing distributed quantum computing with controlled approximations. Exact distributed quantum computing requires exponential classical information to reconstruct the quantum process. However, we show how the classical cost is reduced to polynomial if the quantum procedure can be partitioned between a qubit that is weakly coupled the other qubits. We demonstrate our method for a layered circuit based on the circuits used for the quantum approximate optimization algorithm.

Bell- and contextuality-type inequalities have become practical probes of many-body quantum correlations, often involving only few-body correlators and quantities with a direct Hamiltonian interpretation such as an energy density. Here we show that, in infinite one-dimensional translation-invariant chains, contextuality can acquire a genuinely thermodynamic meaning: within the witness families studied, the maximal quantum violation coincides with spontaneous breaking of one-site translation symmetry, producing strictly $p$-periodic ground states with $p>1$. Along natural continuous interpolations between classical-bound and quantum-optimal Hamiltonians, the classical bound marks a symmetry-breaking point where competing classical periodicities are lifted in favor of a unique quantum-selected period. At the quantum optimum, the studied families admit exact finite-size reductions: a translation-invariant contextuality witness induces a $p$-site periodic-boundary-condition inequality with identical classical and quantum bounds (hence no loss under reduction), and in several cases the resulting finite inequalities are tight. This reduction turns an infinite-chain contextuality certification into a compact, hardware-testable benchmark on a small ring, requiring only local energy measurements. We establish the mechanism analytically in representative two- and three-body witness models and corroborate it more broadly using a translation-invariant adaptation of semidefinite-program hierarchies together with variational matrix-product-state algorithms.

Stochastic resetting has emerged as a versatile protocol to drive quantum many-body systems to non-equilibrium steady states by interspersing unitary dynamics with measurements and resets at random times. In spite of this, a quantum hardware validation of such non-equilibrium steady states is still missing. Here, we achieve this goal by first formulating a noisy discrete-time theory where unitary gates alternate with noisy mid-circuit projective measurements and conditional resets. This noisy conditional resetting theory is then demonstrated on a superconducting quantum processor for up to $N=7$ qubits. We consider, as a paradigmatic case, the unitary dynamics of the interacting Floquet transverse-field Ising model. The stationary state of the noisy conditional resetting agrees quantitatively with the experiments, and it shows crossover behavior related to the equilibrium quantum phase transition of the model. Our results might thus pave the way for the preparation of collective stationary states on noisy quantum devices and for further developments of quantum algorithms involving mid-circuit measurements.

The Dirac operator on a lattice cannot be both local and free of fermion doubling, at least not without breaking fundamental symmetries. Non-local, symmetry-preserving discretizations that avoid doubling have a quantum circuit representation as a linear-combination-of-unitaries (LCU) in which both the number of terms and their norm (the subnormalization factor) grow with the lattice size, compromising the efficiency of a quantum algorithm. We show that the tangent-fermion discretization escapes this obstruction when the Dirac equation is written as a generalized eigenvalue problem with a local operator pencil: Each member of the pencil has an exact LCU, with term count that is independent of lattice size and with subnormalization factor of order unity, on a par with elliptic operators. This provides an efficient block-encoding primitive for Dirac spectra and Green functions without fermion doubling.

Optically addressable shallow donors in ZnO combine efficient spin-selective optical transitions with the potential for long spin coherence in an isotopically purifiable host lattice, making them an attractive platform for spin-photon quantum technologies. A key missing capability, however, has been coherent control beyond the small-angle rotations accessible with ultrafast optical pulses. Here we demonstrate coherent microwave control of implanted $^{115}\mathrm{In}$ donors in ZnO. Resonant optical pumping initializes and reads out the donor electron spin. Pulsed optically-detected magnetic resonance resolves the ten hyperfine transitions associated with the coupled $^{115}\mathrm{In}$ nuclear spin (I = 9/2) and reveals optical-pumping-induced nuclear spin polarization. We observe coherent Rabi oscillations with a maximum Rabi frequency of $Ω/2π= 36.2 \pm 0.7$\;MHz, corresponding to a $π$-pulse time of 13.8$\pm$0.3\;ns, and characterize the spin coherence using Ramsey, Hahn echo and dynamical-decoupling measurements. Unexpectedly, the measured coherence is substantially shorter than reported in previous optical studies of donor spins in ZnO at high magnetic field. Control experiments rule out several simple explanations including microwave heating and instantaneous diffusion from the driven donor ensemble, leaving an open question regarding the origin of decoherence at low magnetic field in microwave-controlled ZnO donors. These results establish microwave control of ZnO donor qubits with resonant optical access to specific donor species. More broadly, they demonstrate that coherent microwave control can be achieved in optically addressable spin systems with nanosecond-scale inhomogeneous dephasing, enabling field-, temperature-, and materials-dependent studies of coherence-limiting mechanisms and the development of optically interfaced electron-nuclear spin registers.

We discuss here how non self-adjoint Hamiltonians, and their related Heisenberg-like dynamics, can be used to model a political system consisting in a coalition $\C$ of different parties (forming a government) and by their (original) supporters $\Sc$. Our aim is to model how the opinion of these supporters changes depending on the efficiency, competence and coherence of the coalition $\C$, as these are perceived by $\Sc$ during their action while governing. After a rather general introduction we propose three specific models, and we describe and comment the dynamical behaviour of the { full} system, $\Sc\cup\C$. The role of the so-called {\em balanced Hamiltonians}, recently introduced by the authors in connection with integrals of motion, is discussed in details.

We introduce a quantum reservoir computing online protocol that realizes amplitude encoding on quantum hardware. Our scheme combines mid-circuit measurement and reset operations to implement the partial-trace dynamics underlying amplitude encoding, while an indirect measurement scheme provides access to reservoir observables without interrupting temporal processing. In contrast to other approaches, our method preserves online operation, avoids input buffering, and keeps the runtime linear in the number of time steps. We present the theoretical formulation of the protocol and a proof-of-principle implementation on quantum hardware, and we evaluate its performance on two standard benchmark tasks. Our results show that the reservoir dynamics can be monitored through both direct measurements of the input qubits and indirect measurements of the memory qubits, enabling observation of the full system while isolating the internal evolution of the reservoir. This work provides a practical route toward scalable hardware implementations of amplitude-encoded quantum reservoir computing and opens the door to systematic experimental studies of complex quantum reservoirs.

We introduce the Biorthogonal Variational Quantum Eigensolver (B-VQE), a quantum algorithm for simulating non-Hermitian many-body systems on noisy intermediate-scale quantum (NISQ) hardware. Non-Hermitian quantum matter exhibits exceptional points, parity-time symmetry breaking, and non-Hermitian skin effects, yet existing quantum algorithms often rely on costly post-selection procedures and are not designed to capture biorthogonal eigenstates. B-VQE employs independent variational circuits to represent the left and right eigenstates of a non-Hermitian Hamiltonian and optimizes a biorthogonal objective function that directly tracks non-Hermitian phase transitions. The framework incorporates an Exceptional-Point Detector (EPD) that identifies exceptional points through a hardware-native coalescence metric and a Non-Hermitian Quantum Geometric Tensor (NH-QGT) readout that distinguishes state-topological and band-topological signatures in interacting many-body systems. To overcome the exponential overhead associated with conventional non-Hermitian simulation, we develop an importance-sampling mitigation strategy that removes the need for ancilla-based post-selection while retaining polynomial computational scaling. We validate the approach on three representative models: a non-Hermitian Hubbard chain, a non-Hermitian XXZ spin chain, and a two-dimensional non-Hermitian (t)-(J) model. B-VQE achieves relative energy errors below (5\times10^{-3}) and locates exceptional points with high accuracy on noise-free simulations while resolving phase boundaries associated with localization, quantum scars, and skin-effect physics. These results establish B-VQE as a scalable NISQ methodology for constructing non-Hermitian many-body phase diagrams and exploring topological and critical phenomena in open quantum systems.

The continuous progress of quantum technologies has spurred the exploration of their potential applications across diverse fields, particularly in combinatorial optimisation. In this work, we study a quantum-classical optimisation framework based on Pauli Correlation Encoding, an encoding scheme that can represent m binary variables using a polynomial number of qubits. To evaluate the performance of the method, we use four classical optimisation problems against the instances of the QOPTLib benchmark. The study includes an analysis of the impact of the compression order of the encoding scheme, the problem structure, and hyperparameter selection on solution quality, as well as the role of post-processing in improving performance. Additionally, we study the effect of shot-based execution and hardware noise, showing how these factors influence both the accuracy of expected value estimation and the overall dynamics of the optimisation process. The results indicate that the proposed PCE-based framework achieves competitive performance against the benchmark and, in several cases, obtains equivalent or even superior solutions, highlighting its potential as an efficient encoding strategy for quantum optimisation in the NISQ and near fault-tolerant era.

We propose an alternative interpretation of dark matter effects within the framework of General Relativity. In particular, we suggest that, in astrophysical and cosmological contexts, different initial assumptions about a system inevitably lead to different interpretations of the same phenomena. As a concrete example, we examine self-gravitating systems composed of an axially symmetric rotating dust fluid and show that effects typically attributed to the presence of additional matter, can instead be reproduced through an appropriate choice of initial and boundary conditions for the equations governing the system.

Realizing scalable Quantum-to-the-Home (QTTH) faces a bottleneck: link asymmetry in broadcast continuous-variable quantum access networks (CV-QANs) hinders the selection of a globally optimal modulation variance. We demonstrate a downstream broadcast CV-QAN connecting a Quantum Line Terminal (QLT) to multiple Quantum Network Units (QNUs) over commercial fiber. Operating within a trusted local network domain, we establish a multi-user utility model to select the optimal shared variance, balancing network efficiency and user fairness. Supported by robust digital signal processing, our 1:16 field trial achieves Mbit/s-level asymptotic secure key rates, bridging theoretical protocols with Fiber-to-the-Home reality and guiding future scalable access architectures.

End-to-end entanglement distribution is a key capability of upcoming quantum networks, enabling applications like distributed quantum computing, quantum sensor networks, and secure communications. Hence, its realistic and efficient simulation is crucial for quantum network design and for assessing the ability of a network to run certain applications. This work provides tools to scale-up and improve the realism of entanglement generation simulations in quantum networks. This is achieved by deriving analytical results that directly return the success probability, the output state and corresponding fidelity of a selected entanglement generation protocol, while accounting for a variety of noise sources affecting the protocol. These results are then integrated and streamlined in an upgraded version of SeQUeNCe, one of the most popular quantum network simulators. The resulting simulator features increased scalability by reducing computation time by more than 60%, while allowing for a variety of realistic noise sources, including imperfect mode matching, dark counts, and imperfect memory initialization. The simulator is also benchmarked with real experimental data and is capable of replicating the average entanglement generation time and the final state fidelity of a selected experiment. Altogether, the results can enhance current quantum network simulation capabilities towards large-scale networks, paving the way for the future quantum internet.

Color graphs and their subgraphs, referred to as bubble graphs, correspond bijectively to the simplicial complexes of pseudomanifolds and their subsimplices, respectively. In this paper, we introduce matrix representations for colored graphs and their associated bubble graphs. By using this correspondence, we define simplicial-complex matrices and subsimplex matrices that encode the simplicial complexes of pseudomanifolds and their subsimplices. Moreover, we formulate mutation and crossover operations on colored graphs. Through the established correspondence among simplicial complexes, colored graphs, and simplicial-complex matrices, we extend these operations to simplicial complexes and simplicial-complex matrices. We further implement an algorithm generating simplicial-complex matrices and a genetic algorithm performing mutation and crossover of them to produce pseudomanifolds exhibiting diverse topologies. In addition, we implement procedures for decomposing the generated simplicial-complex matrices into simplex matrices, reconstructing the simplicial complexes of the associated pseudomanifolds from this information, and computing geometric quantities such as the volume, circumcenter, and dual-simplex volume of each simplex.

A possible approach for preventing compact astrophysical objects from gravitational collapse into singularities is the idea of dark energy. Since it is the cause of our universe's accelerated expansion, it has the greatest impact on the cosmos. As a result, it appears that dark energy can interact with any compact astrophysical stellar object [Phys. Rev. D 103, 084042 (2021)]. In this study, our primary objective is to develop a simpler model of a charged strange star coupled with anisotropic dark energy admitting the Adler-Finch-Skea solution [J. Math. Phys. 15, 727 (1974); Class. Quantum Grav. 6, 467 (1989)] within Einstein gravity. To develop this model, the Karmarkar condition was employed to determine the radial metric component, while Adler's methodology was used to choose the time-metric component. For this purpose, we explored a particular strange star, Her X-1, with observed values of mass $(0.85 \pm 0.15)M_{\odot}$ and radius $= 8.1_{-0.41}^{+0.41}$ km. In this context, we proceeded to model dark energy using the equation of state (EoS), such that the density of dark energy is proportional to the density of isotropic perfect fluid matter. The unknown constants in the metric were determined by smooth matching using the Darmois-Israel criterion. We conduct an in-depth examination of the stability and force equilibrium of our suggested star framework, as well as several physical characteristics of the model such as the metric function, pressure, density, mass-radius relation, and dark energy parameters. Thus, the physical consistency and stability of the present model are investigated. Therefore, following a comprehensive theoretical investigation, we discovered that our proposed model is singularity free and meets all the stability requirements to be a stable and physically realistic stellar model.

The photon blockade (PB) effect is a hallmark non-classical phenomenon in quantum optics and finds important applications for building quantum sources, while the control of PB by the non-Hermitian exceptional points remains largely unexplored. In this work, we theoretically investigate universal photon blockade in a microcavity harboring chiral exceptional points (CEPs) for building multiplexing quantum sources with nonreciprocal photon statistics. The results reveal that the presence of the CEPs leads to a stark contrast in the photon statistics of two whispering-gallery modes with opposite propagating directions. That is, one mode exhibits a strong PB effect while the other displays either sub-Poissonian or super-Poissonian distribution. Our findings thus may pave the way for advanced applications of photon blockade, and provide a theoretical foundation for the selective generation of single-photon and two-photon emission

Superconducting qubits offer a versatile platform for quantum simulation. In this architecture, quantum information can be encoded in the bosonic modes of a microwave cavity, offering an alternative to conventional qubit-based encoding schemes. These cavity bosonic modes can be manipulated using a single ancillary qubit. In this work, we investigate the quantum simulation of two- and three-flavor neutrino oscillations using Fock-basis encoding of a cavity mode. We design pulse sequences for implementing the required unitary operations through selective number-dependent arbitrary phase (SNAP) and displacement gates. Pulse-level control is employed to realize high-fidelity gate operations on the encoded cavity mode. The resulting neutrino oscillation probabilities obtained from quantum simulation exhibit close agreement with the corresponding theoretical predictions, demonstrating the feasibility of cavity-based bosonic encoding schemes for quantum simulation.

Complex quantum networks can support a diverse set of long-range entanglement distribution schemes ranging from linear repeater protocols to multipath entanglement purification strategies. As a result, a network's resourcefulness, that is its ability to facilitate quantum communication, depends on the deployed distribution scheme. In this work, we analyse and compare the resourcefulness of quantum networks across a broad range of network topologies, including both regular and random networks, under two distinct entanglement distribution schemes. The first relies on entanglement swapping along a single path connecting a source-target pair, while the second exploits entanglement purification using multiple paths between the same source and target nodes. The resourcefulness of the network is quantified using a recently described metric [1] that averages over the maximum teleportation fidelity between arbitrary source-target pairs in the network. We present algorithms for estimating this metric under constraints of edge-usage and ordering of paths. Our results not only demonstrate the sensitivity of the average maximum teleportation fidelity metric to the choice of entanglement distribution protocol, but also highlight the significant improvements enabled by network purification schemes. In particular, purification-based approaches can substantially enhance average teleportation fidelity, thereby improving the overall teleportation capability of quantum networks.

We analyze how memory in disorder facilitates quantum communication in many-body scar systems. We consider three distinct types of disorder, viz., memoryful, and memoryless uniform and Gaussian, and compare their respective performances in facilitating quantum state transfer. Using the maximum transfer fidelity and fidelity area as figures of merit, we find that memoryful disorder yields a better performance than the memoryless disordered channels. Furthermore, the maximum transfer fidelity exhibits an initial parabolic decay with disorder strength, followed by a linear decrease, for all the disorder models considered. We introduce a degree of scarness, and show that it is higher for memoryful disorder in comparison to memoryless disorders, implying a role of scarness in the quantum state transfer protocol. We further perform a scaling analysis, revealing that memory effect in disorder is not only beneficial for short-distance but also long-distance quantum state transfer. Finally, we show that the state yielding the maximum transfer fidelity has larger inverse participation ratio for memoryful disorder in comparison to the other two disorders, highlighting the role of nonergodicity in enhancing state transfer.

We develop, to our knowledge, the first receiver-centric blockwise sequential-detection framework for covert communication over thermal-loss bosonic channels. In this architecture, each block serves as a binary super-symbol, and the key design problem is to determine the minimum detection-segment length that enables Bob to detect an active block before the block ends while remaining covert to Willie. For any fixed physically realizable general-dyne receiver, Bob's post-change information growth is linear in the small-signal regime, whereas Willie's detectability obeys a quadratic quantum relative entropy law. Exploiting this asymmetry, we show that under a per-block covertness budget the asymptotically optimal signaling strategy is uniform across the detection segment, and we derive an explicit minimum-length condition under which a single-pass cumulative sum (CUSUM) detector crosses threshold within the same block with exponentially high probability. The resulting design law yields a covert blockwise binary codebook over a finite transmission horizon and establishes a concrete link between bosonic covert communication, sequential detection, and blockwise signaling design. More broadly, these results provide design guidance for covert quantum communication systems with physically realizable receivers, and help bridge information-theoretic covertness guarantees with implementable receiver-aware optical communication design.

Quantifying coherence is an essential endeavor in both quantum foundations and quantum technologies. In this paper, we study the coherence measures in terms of the diagonal states in the strictly incoherent operations framework. Specifically, we propose a coherence measure in terms of fidelity and provide its analytical expression. The relations between the proposed coherence measure and some other coherence measures are derived. Furthermore, we prove its monotonicity under incoherent operations. As an application, we explore the role of the proposed coherence measure in characterizing the waveness in the multi-path interferometer. As a result, some wave-particle dualities in terms of fidelity are presented. This work not only deepens the interpretation of the diagonal states on characterizing quantum states, but also promotes the quantitative description of the wave-particle behaviors in the multi-path interferometer.

We characterize three-qubit controlled unitary gates of Schmidt rank three, establishing necessary and sufficient conditions for such gates to have Schmidt rank three. We construct explicit examples and analyze their entanglement capabilities, showing that gates with Schmidt rank three can generate output states ranging from fully separable to maximal GHZ-class states. Within this classification, we present a parameterized gate family that produces an output W-class state whose bipartite tangles for subsystems AB and AC are modulated as simple trigonometric functions of a single phase parameter. The examples also include gates implementable with only three CNOT gates, showing that this lower bound is achievable. Decompositions that achieve the minimum possible CNOT count for several other cases are provided as well. Our results bridge Schmidt rank classification, entanglement structure, and resource-efficient circuit synthesis.

We introduce the notion and reveal remarkable properties of quantum complexity resource contained in a mixed multimode Gaussian state and providing universal quantitative characterization of its quantum advantage. The notion is based on convex optimization, multimode photon number statistics, Hafnian Master Theorem, and #P-hard complexity. We consider pulsed OPAs targeting maximal quantum complexity resource and thousands of multipartite-entangled squeezed modes of output light via nonlinear, spatio-temporally nonadiabatic generation inside OPA and optimized extraction out of OPA. We show that such figure of merit is more realistic than Bloch--Messiah supermodes and guides to multimode OPAs opening new paths to important applications in quantum information science such as generation of 3D cluster states for one-way photonic quantum computing and demonstration of quantum advantage.

Variational quantum algorithms (VQAs) are among the leading approaches for near-term quantum computing, yet their performance can degrade in barren plateau regimes characterized by vanishing gradients. A widely held intuition is that increasing circuit expressivity, often associated with random-state behavior, leads to a loss of trainability. Existing results show that sufficiently random circuits can lead to barren plateaus. Here we show that standard statistical signatures of randomness can emerge well before this regime, without degrading trainability. We demonstrate this behavior in structured variational circuits applied to the one-dimensional cluster-Ising model and a generalized toric code Hamiltonian. To characterize state complexity, we analyze Porter-Thomas statistics, entanglement-spectrum level statistics, and inverse participation ratios. Across both models, increasing circuit depth drives these diagnostics toward random-state-like or random-matrix-like behavior, while variational optimization remains effective, with no evidence of exponential gradient suppression in the regime studied. We interpret this behavior in terms of locality. Spectral correlations develop at relatively shallow depth through locally generated structure, while global state randomization and the associated concentration-of-measure effects are not yet realized. These results show that commonly used statistical diagnostics of complexity do not by themselves determine trainability. Instead, they point to a separation between different aspects of complexity in finite-depth variational circuits.

Atom interferometers are among the most sensitive instruments for precision measurements and tests of fundamental physics. Their performance, however, is ultimately limited by quantum projection noise when uncorrelated atomic ensembles are employed. Cavity-assisted generation of entangled states has proven to be a promising route toward quantum-enhanced interferometry beyond the standard quantum limit. In this work, we present the realization and characterization of a monolithic bow-tie cavity developed to achieve a strong collective atom-light coupling with strontium atoms. Unlike conventional standing-wave Fabry-Pérot resonators, the traveling-wave geometry of the bow-tie cavity provides homogeneous atom-light coupling over the entire atomic ensemble, making it particularly suitable for entanglement-enhanced atom interferometry with freely falling atoms. The monolithic cavity architecture presents several scientifically relevant features such as high mechanical stability, high finesse, robustness against mirror misalignment, optical and atomic access and the option of generating squeezed states through different strategies. The cavity was realized for operation on the strontium $(5s^2) ^1S_0-(5s5p) ^3P_1$ transition at 689 nm and achieves a finesse of $\mathcal{F}=5.7\times 10^4$ while keeping the transmission of a single mirror sufficiently large to allow for efficient atomic information extraction. In this geometry, the cavity supports two foci with waists of 164 $μ$m and 31 $μ$m which gives access to different regimes of atom-cavity coupling. For ensembles containing up to $10^5$ atoms, the cavity is expected to enable metrological gains approaching 24 dB of spin squeezing through cavity-feedback squeezing, and 28 dB through quantum non-demolition measurements, demonstrating its potential as a platform for next-generation quantum-enhanced atom interferometers.

We implement the Hadamard test with Trotterized time-evolution operators on IBM quantum computers to simultaneously extract ground- and excited-state energies of the transverse field Ising model (TFIM) and transverse longitudinal field Ising model (TLFIM). The Trotterization circuits for the TFIM admit constant-depth circuits (CDCs) for arbitrary time, allowing us to locate a large number of eigen-energies above the background noise for up to six spins. Via circuit synthesis we show that the three-spin TLFIM has constant-depth structure although it does not meet the known CDC criteria. The CDCs enable the extraction of the ground and first-excited state energies of the three-site TLFIM via its dynamics. We also address complications from the noisy background and discrete Fourier transform to enhance the reliability of the extraction process and compare the results from different generations of IBM hardware to highlight the improvement.

The complex Sachdev-Ye-Kitaev (SYK) model, featuring fermions with all-to-all interactions, serves as a dual paradigm for understanding non-Fermi liquid behavior and the holographic nature of charged black holes. Two defining characteristics of the standard SYK model are its maximal chaos (Lyapunov exponent $λ_{\mathrm{L}}=2πT$ at temperature $T$), and its finite zero-temperature residual entropy. While previous studies have largely focused on couplings drawn from a zero-mean Gaussian distribution, we investigate a generalized model with a finite mean-to-standard-deviation ratio, $g\equiv J_{0}/δJ$ of the coupling distribution in order to get deeper insight into the evolution of chaos. We find that increasing $g$ yields the following effects: (i) The system remains a fast scrambler with $λ_{\mathrm{L}}=A~T$, but with a suppressed coefficient $A<2π$. (ii) In the limit $g\to \infty$, out-of-time-ordered correlators (OTOCs) no longer exhibit exponential growth with $λ_{\mathrm{L}}\simeq 0$. (iii) The spectral correlations indicative of late-time chaos maintain Wigner-Dyson level spacing statistics for all values of $g$. (iv) The system preserves a finite residual entropy, albeit with reduced magnitude, for all $g$ values. We conclude that in this generalized SYK model, there is a chaotic to non-chaotic crossover. Moreover different measures of chaos decouple, demonstrating that the presence of finite residual entropy does not strictly imply maximal chaos.

A promising avenue for scaling quantum computing is to connect quantum processing units (QPUs) by generating entanglement between them. This requires circuit partitioning: partially rewriting quantum circuits to run on a distributed quantum system using quantum teleportation protocols, while preserving the unitary operation implemented by the circuit. The key metric to minimize when partitioning is the e-bit count, defined as the number of maximally entangled qubit pairs that must be generated between QPUs. We focus on partitioning the quantum Fourier transform (QFT) circuit, which is widely used as a subroutine in quantum algorithms such as quantum phase estimation and arithmetic circuits. Specifically, we present a partitioning scheme based on optimal gate-packing, compare it against prior analytical partitioning schemes for the QFT, and evaluate it against partitions produced by general-purpose circuit partitioning algorithms. We further validate our approach by implementing the partitioned circuit on quantum hardware.

We construct exact, positive-energy, normalizable wave-packet solutions of the Dirac equation in the axisymmetric potential $V=-\,v_0/ρ$ -- to our knowledge, the first such solutions in any external potential. Remarkably, one family comprises only elementary functions whose longitudinal profiles reproduce the free-Schrödinger Hermite--Gauss wave packets in the nonrelativistic limit. All packets share two striking features: (i) a probability density that is pointwise decoupled from spin orientation -- despite the inherent spin-orbit coupling of the Dirac equation -- and (ii) a complete freezing of their time evolution at the critical coupling $v_0\to\hbar c/2$. We also present a simple scheme that maps solutions of the 2D Helmholtz equation to further exact Dirac wave packets.

In a standard two-level Ramsey interferometer, the measured phase accumulates linearly during the interrogation time. Here, we introduce three-level Ramsey interferometry that employs a noncyclic geometric phase response to enhance phase sensing, with projected internal-path interference reshaping the mapping from accumulated signal phase to readout phase. Near a geodesic-closure transition, a small accumulated signal phase produces a sharply amplified readout-phase shift. We quantify the accompanying gain--visibility tradeoff and identify a finite operating window in which the amplified response yields a net signal-to-noise-ratio gain under technical-noise-limited conditions. By tuning an initial Ramsey phase offset, this high-slope window can be positioned at a desired operating point and sampled repeatedly with shorter cycles, providing a geometric shortcut to improved projected stability. More broadly, these results establish a multilevel Ramsey route to enhanced phase sensitivity in quantum platforms, where two signal-collecting internal paths interfere to produce a noncyclic geometric response.

We investigate the thermodynamic stability and the stochastic thermal fluctuations of the warped black hole solution in three-dimensional General Massive Gravity. We demonstrate that the black hole is thermodynamically unstable and identify the nontrivial Davies phase-transition curves from the behavior of its admissible heat capacities. Going beyond the classical stability analysis, we study thermal fluctuations within a modified finite-time nonequilibrium extension of Ruppeiner's Hessian-based fluctuation theory. For a class of isentropic and isoenergetic processes, we derive exact on-shell angular momentum trajectories in the thermodynamic state space and compute the corresponding thermodynamic lengths. These quantities characterize relaxation processes between macrostates and provide an estimate of the associated relaxation times. Furthermore, we show that the thermodynamic geodesic equations do not admit constant-angular-momentum solutions, suggesting a continuous change of the black hole's angular momentum. Our results consistently reproduce the warped AdS$_3$ black hole limit of Topological Massive Gravity.

There is growing interest in extensions of the standard model of gate-based quantum computation to include auxiliary degrees of freedom evolving according to a nonlinear Schrödinger equation. By reducing the Boolean satisfiability problem SAT to quantum state discrimination, Abrams and Lloyd argued that the right type of nonlinearity can be used to solve {\sf NP} and \#{\sf P} problems in polynomial time, at least in an idealized noise-free limit. For practical implementation, however, we are restricted to simulated and emergent nonlinearities, such as that appearing in mean field models for ultracold atoms and similar ensembles. A prominent example is the torsion model, which arises in two-component Bose-Einstein condensates and spin models with all-to-all Ising interaction. But torsion-based state discrimination appears to fall short of solving SAT. Here we close this gap by constructing the filtered oracle of the Valiant-Vazirani theorem, providing a randomized polynomial-time reduction from SAT to UNIQUE SAT, a promise problem where there is at most 1 satisfying assignment. In the noise free limit, the UNIQUE SAT problem can be olved in polynomial time using torsion nonlinearity. Quantum Valiant-Vazirani reduction is no faster than the efficient classical version, but a fault-tolerant implementation coupled to a nonlinear quantum coprocessor simulating torsion would enable polynomial time solution to {\sf NP} (but not \#{\sf P}) problems.

As large language models (LLMs) become embedded in quantum simulation workflows (IDE copilots, notebook assistants, agentic pipelines), evaluation must move beyond functional correctness to anticipate and catch structured failures before they propagate through expensive pipelines. We present a layered evaluation framework for materials-informed Variational Quantum Eigensolver (VQE) circuit generation: (i) a gatekeeper screening rubric across seven physical and framework criteria; (ii) a circuit fidelity analysis comparing model outputs against analytical and reference-implementation values for H2/STO-3G/Jordan-Wigner/UCCSD, with ansatz classification and gate-composition breakdown; and (iii) design entropy, a run-to-run behavioral consistency metric. We surface a taxonomy of five distinct LLM failure modes (geometry hallucination, nonexistent API usage, runtime integration failures, constraint violations, and plausible-but-unverifiable output), each with distinct detectability profiles and structural to the task rather than to any one model. A forensic audit of the evaluation platform's own source code further establishes that two apparent model failures originated in the harness through silent fallback-template substitution, demonstrating that evaluation infrastructure belongs inside the same trust boundary as the models it tests. Applied across multiple foundation models on a Materials Project integrated pipeline, the framework shows that gatekeeper-style validation is necessary, not optional, for reliable deployment.

We investigate the response of entanglement to temperature variations in interacting two-qubit thermal states. For a general two-qubit interaction Hamiltonian, we derive exact expressions for the thermal concurrence, its first and second derivatives with respect to inverse temperature, and the thermal quantum Fisher information. We show that the rate of change of thermal entanglement is bounded by the thermal quantum Fisher information. We further derive a bound relating entanglement curvature and thermal quantum Fisher information, and show that temperature uncertainty induces a loss of entanglement bounded by the same quantity that determines thermometric sensitivity. These results establish thermal quantum Fisher information as a fundamental constraint on the response and robustness of entanglement in interacting two-qubit thermal states.

The sign problem is a notorious obstacle for classically simulating quantum theories with fermions. We propose an effective field theory method for analyzing the sign problem. At high temperatures, a $d$+1 dimensional field theory reduces to a bosonic $d$-dimensional theory; the phase of the Pfaffian in the higher dimensional theory is encoded in an operator in the lower dimensional theory. We apply this framework to the D0-brane/BFSS matrix quantum mechanics, where the phase becomes an operator in a bosonic multi-matrix integral. Our results show that the continuum theory has a sign problem that persists in the large-$N$ 't Hooft regime. However, detecting the sign problem involves going to 10-loop order in the high-temperature expansion. This delayed onset follows from the fact that the Pfaffian phase transforms as an $O(9)$ pseudoscalar. Furthermore, the relevant diagrams give a numerically small prefactor. Consequently, ignoring the sign problem leads to a relatively small fractional error in thermodynamic quantities for temperatures $T \gtrsim λ^{1/3}$. However, at stronger coupling in the 't Hooft regime, the sign problem may become more severe. Finally, we initiate the application of this framework to higher-dimensional maximally supersymmetric Yang-Mills theories.

We consider the problem of identifying a topological order based on bulk entanglement of the ground-state wavefunction. Previous work showed that some universal information can be extracted from multi-entropy measures, a class of multipartite entanglement measures obtained by applying permutation operators exchanging the degrees of freedom between different replicas of the wavefunction. It remains an open question to what extent such entanglement measures can be used to extract any universal information from the ground state. Here we show that the topological partition function $Z(M)$ of a manifold $M$ can be extracted provided that $M$ satisfies a topological condition which we term ``local achirality". We show that locally-achiral manifolds can be used to extract universal properties of 2+1d topological phases that go beyond the $S$ and $T$ matrices. As a first step towards classifying locally-achiral manifolds, we show that, in four dimensions, such manifolds have vanishing Pontryagin number. We relate this property to the existence of beyond-cohomology time-reversal symmetry protected topological order (T-SPT) in four dimensions. Finally, we present an entanglement measure that detects this nontrivial T-SPT.

We elucidate the physical nature of ghosts above the multi-particle threshold by contrasting them with unstable particles in quantum field theory. We first consider the asymptotic formulation, where ordinary positive-norm one-particle states can be unstable and decay, whereas ghosts survive asymptotically without decaying, yet admit no particle interpretation due to interference with the multi-particle component which masks the negative-norm one-particle state. This distinction originates from two different analytic structures of the dressed propagator, whose complex conjugate poles lie in the first or second Riemann sheet in the ghost or ordinary case, respectively. Ghost resonances are, in principle, phenomenologically distinguishable from ordinary ones, being narrower and exhibiting weaker interference between positive- and negative-energy peaks. We then formulate the quantum field theory in a finite interval of time and, working within a suitable approximation for the dressed propagator, find that finite-time effects amplify differences in the resonant behavior and give rise to new features, such as higher peaks in ghost resonances. Distinct temporal regimes are also identified: for times shorter than the inverse width, an approximate free-particle description is valid, whereas at later times interactions and interference effects dominate, leading to decay or multi-particle masking. Complex poles in the dressed propagator emerge only at late times and become complex-conjugate pairs asymptotically, determining the asymptotic dynamics. This study supports the absence of freely propagating ghost particles in the asymptotic limit.

This paper generalizes the method used in the previous article 2512.09930 to black $p$-brane thermodynamics in arbitrary dimensions containing black holes and strings as special cases: thermodynamic quantities can be derived without constructing the corresponding black $p$-brane solutions. We further extend the discussion to black holes or $p$-branes involving a general scalar coset.

We derive an analytical approximation for the graviton propagator from Asymptotic Safety. We find neither extra poles nor indications of unitarity or causality violations in the spin-two sector. Our results strengthen the case that Asymptotic Safety does not introduce new degrees of freedom, and thus propagates the same field content as General Relativity. We also identify the underlying mechanism: the residues of spurious poles in finite-order derivative expansions approach zero as the order is increased.

Extreme mass-ratio inspirals embedded in accretion disks provide a natural arena for studying the interplay between relativistic orbital dynamics and environmental effects. In this work, we develop a framework to investigate the secular evolution of compact objects repeatedly crossing an accretion disk around a supermassive black hole. The orbital motion is modeled through Kerr geodesics, while disk interactions are encoded through effective prescriptions for mass accretion and dynamical friction. We find that disk-induced dissipation generically drives a two-stage evolution characterized by rapid alignment of the orbital plane with the disk, followed by slower eccentricity damping. By systematically comparing the dynamics with a purely Keplerian treatment, we show that cumulative relativistic effects produce deviations even at large orbital separations, where the Keplerian approximation would naively be expected to remain accurate. These discrepancies grow through repeated disk crossings and become increasingly pronounced in more relativistic orbital configurations. We further investigate the impact of the accretion-disk model by comparing the Sirko-Goodman and Novikov-Thorne prescriptions. Relativistic disk structures predict systematically lower densities and larger scale heights, leading to weaker orbital dissipation and slower secular evolution. By contrast, the spin of the central black hole has only a minor effect on the overall circularization efficiency. Our results demonstrate the importance of consistently modeling both relativistic orbital dynamics and disk structure when studying compact objects embedded in AGN disks, and provide a framework for exploring their long-term evolution, as well as a possible connection to quasi-periodic eruptions.

We investigate the entropy of a dynamical black hole arising from second-order perturbations of a general stationary background with a bifurcate Killing horizon. Using Gaussian null coordinates, we study the geometry of the apparent horizon perturbatively up to second order. Within the covariant phase space formalism, to explore the contribution of matter fields, we introduce a new modified canonical energy, and establish a balance law relating the second-order variation of the entropy to the energy flux entering the black hole. We show that the entropy is given precisely by the area of the apparent horizon at second order when the null energy condition holds for the infalling matter, and that the variation of the entropy also obeys the second law. We also discuss the possibility that the area law continues to hold when the null energy condition is violated.

Owing to rapid recent progress, ultracold atomic lattice systems for simulating topological phases are now at a pivotal stage, evolving from established paradigms into increasingly versatile and programmable quantum simulators. In this review, we survey recent experimental advances across four major classes of platforms: optical lattices, including optical lattices with laser-assisted tunneling and optical Raman lattices; synthetic lattices in momentum or internal-state space; Floquet-engineered lattices; and optical tweezer arrays, all of which offer distinct capabilities for realizing and probing topological matter. For each class, we highlight representative experimental breakthroughs, the topological models that have been realized, and the advanced detection and characterization techniques employed, emphasizing how these complementary approaches collectively expand the frontier of quantum simulation. We also discuss emerging directions in strongly correlated and nonequilibrium topological phases, and conclude with an outlook on future prospects.

Fault-tolerant quantum computing fundamentally relies on the accurate characterization of circuit-level noise to optimize decoding algorithms. However, extracting complex multi-body error correlations remains challenging. Contemporary greedy inference algorithms can suffer from statistical distortion, discarding true physical mechanisms while introducing many unphysical false positives. Here, we introduce the Correlation-Analysis-based Hypergraph Reconstruction (CAHR) algorithm, a globally consistent framework to invert experimental syndrome statistics directly into discrete physical hypergraphs. By coupling exact algebraic correlation equations with a top-down concurrent-pruning strategy, CAHR recovers the fault topology without false positives for both $d=5$ rotated surface codes and dense 8-body 2D color codes in our benchmark settings. Furthermore, we show that exact continuous parameter extraction in dense codes is limited by a \textit{variance cascade}, where absolute statistical variance accumulates linearly from high- to low-degree mechanisms. This motivates a two-stage inference paradigm: utilizing CAHR to extract the fault topology, followed by continuous probability optimization. This provides a practical approach for characterizing and decoding highly correlated noise in realistic quantum hardware.

We present a quantum-enhanced Markov chain Monte Carlo (QE-MCMC) method to sample turbulent acceleration vectors from a joint target distribution that depends on all three components and height to model the transport and dispersion of massless Lagrangian tracer particles in two turbulent shear flows. A homogeneous shear flow, characterized by a uniform shear rate S, is considered as the starting point. Secondly, a turbulent boundary layer, which forms in both halves of a plane turbulent channel flow at friction Reynolds number Re_tau = 1000, is considered, where the mean shear rate S(y) varies with distance from the wall y. In this hybrid quantum-classical method, the proposal distribution Q for the first of two Metropolis-Hastings sampling substeps is constructed by a parametric quantum circuit. The algorithm generates synthetic tracer particle tracks. The resulting scaling laws for tracer-particle pair dispersion, a central quantity to probe turbulent mixing from a Lagrangian perspective, agree with a stochastic transport model consisting of coupled Langevin equations and with the classical MCMC counterpart. Differently from the classical sampling method, QE-MCMC uses a tempered target distribution. Due to the height dependence of the tracer dynamics in turbulent channel flow, an effective height-weighted spectral gap between the first and second eigenvalue of the Markov-chain transition matrix is introduced. The latter is found to significantly exceed the one of classical MCMC when sampling from a multivariate distribution with cross-correlations at the highest qubit numbers and thus resolutions. Consequently, our results support the applicability of this one-shot algorithm as a generative Lagrangian quantum-computing module, possibly embedded in a complex fluid-flow problem. Our module is found to work reliably for a relatively small number of qubits per spatial dimension of Nq <= 6.

Classical general relativity predicts a singularity at the center of every black hole. We argue that this singularity is never reached. Operating purely within the standard framework of quantum mechanics and the Feynman sum over geometries, we demonstrate that the gravitational functional integral loses support at the Planck curvature threshold ($\mathcal{K} \sim \ell_P^{-4}$). This forms a quantum boundary, $\mathcal{B}_Q$, that truncates the spacetime manifold at a finite, positive radius ($r_\mathcal{B} \approx 10^{-22}$\,m for a solar-mass black hole). The suppression is driven by the mathematical Sobolev failure of the Einstein-Hilbert action; at Planck curvature, Heisenberg uncertainty in the ADM conjugate variables dictates that quantum metric fluctuations render the manifold non-differentiable, making the action mathematically undefined. Because the phase amplitude is undefined, the wavefunctional evaluates identically to zero ($Ψ= 0$), formally marking where physical spacetime cannot exist. For realistic rotating black holes, we demonstrate that $\mathcal{B}_Q$ acts as a quantum-geometric cutoff for the mass-inflation instability, capping the internal mass parameter at a finite amplification of $n_{max} \approx 0.67\,(r_g/\ell_P)^{1/5}$ and $r_B^{\rm max\, Kerr} = 1.67 r_g^{2/5} \ell_P^{3/5}$ for a maximally spinning black hole, and dynamically enforcing a universal, sphericalized core. Evaluating the Gibbons-Hawking-York boundary term over this terminal spacelike slice yields a finite, macroscopic interior action per boundary segment, $S_{GHY}^{\mathcal{B}} \approx \frac{3}{2}Mc^2\,Δt$. Operating without injecting novel trans-Planckian degrees of freedom, these results suggest the classical singularity is not a physical event, but the natural terminal boundary of the geometry's domain of definition.

Strong light-matter coupling between molecules and optical or plasmonic cavity modes has emerged as a promising platform for advancing photonics, materials science, and chemistry. However, optical cavities and plasmonic resonators in particular are inherently lossy systems characterized by finite photon lifetimes. Accurate theoretical descriptions of molecular dynamics under strong coupling therefore require a proper treatment of cavity losses. In this work, we compare three theoretical approaches for modeling dissipative molecule-cavity dynamics within a realistic parameter regime: the Lindblad master equation, the stochastic Schrödinger equation, and the non-Hermitian Schrödinger equation. As an example, we consider the two lowest energy state of Na2 molecule coupled to a cavity mode and analyze the time evolution of the excited-state population and the mean photon number. Our results demonstrate that the stochastic Schrödinger equation provides an accurate and computationally efficient alternative to the Lindblad master equation, while the non-Hermitian Schrödinger approach is found to be applicable only within a limited range of conditions. Furthermore, we show that inclusion of molecular rotation leads to rotational-vibrational-photonic coupling and gives rise to pronounced nonadiabatic dynamics through light-induced conical intersections. These findings highlight the importance of both dissipation and rotational degrees of freedom for a realistic description of molecular dynamics in strongly coupled molecule-cavity systems.

The Spallation Neutron Source (SNS) at Oak Ridge National Laboratory (ORNL) is a prolific muon producer, making it an ideal location for studying dark sector particles produced in muon decays at rest. In this paper, we explore sub-GeV dark particle detection possibilities in a tons-scale, highly capable hydrocarbon scintillator ($HC^2$) detector at the SNS. We consider a search for $e^+e^-$ final states produced by decays of long-lived, $O(10-100)$ MeV axion-like particles and heavy neutral leptons. The $HC^2$ technology space, exemplified by the PROSPECT and Mobile Antineutrino Demonstrator detectors, offers strong rejection capabilities for the cosmic ray backgrounds that would normally dominate this search. By benchmarking on-surface cosmic ray signatures with data from PROSPECT at ORNL, we generate robust predictions for a multi-year SNS deployment of a range of $HC^2$ detector implementations. Results indicate the potential for order-of-magnitude improvements in sensitivity to axion-like particles and heavy neutral leptons in the 10-100 MeV mass regime compared to current global limits. We also comment on the neutrino detection possibilities of a $HC^2$ deployment at the SNS.

We discuss how the $Z'$-boson of the $B-L$ Supersymmetric (SUSY) Standard Model (BLSSM) could evade the current lower bound of around 5 TeV on the mass of such a resonance (of sequential nature) from the Large Hadron Collider (LHC) by a significant margin. This happens when the experimental sensitivities are critically impaired as the $Z'$-boson becomes `fat' or develops some leptophobia or possesses an optimally large decay Branching Ratio (BR) to BLSSM-specific states (including the SUSY ones) or when some or all of these are at play simultaneously. We describe how such a $Z'$-boson could acquire there features while still respecting the non-negotiable precision constraints from the LEP and the SLC experiments running at the $Z$-pole as well as those from the multi-purpose experiments at the LHC that search for such a resonance. We explore the interplay of the aforementioned phenomena and identify the regions of the BLSSM parameter space that give rise to the described situation by carrying out a thorough scan of it. We find that $M_{Z'}$ masses as low as 2.24 TeV may still be allowed in the BLSSM under favorable circumstances.

Ultra-cold bosonic systems can be tuned to exhibit quantum phase transitions. For example, the Rabi-coupled bosonic system exhibits ferromagnetic and paramagnetic phases, whereas the spin-orbit-coupled system exhibits exciting phases such as supersolidity. The physics of these phases and phase transitions is very rich. It is an important topic of research to probe these phases and phase transitions using various tools in many-body physics. The operator product expansion (OPE) provides one such tool. It expresses the product of two separated operators as a series expansion of local operators. In this article, we will derive the OPE of two operators $ψ^\dagger_σ(\vec r)$ and $ψ_{σ'}(\vec r')$. More specifically, we look for the contact density term, which controls many of the universal physics of the underlying bosonic system.

We study the closed-string tree-level Virasoro-Shapiro amplitude using the dispersive S-matrix bootstrap. For the ten-dimensional maximally supersymmetric four-point amplitude, we impose analyticity, crossing symmetry, partial-wave unitarity, and Regge boundedness. With the massless graviton pole kept explicitly, the resulting dispersion relations and crossing null constraints give numerical bounds on the leading low-energy coefficients normalized by the gravitational coupling. We then introduce a Virasoro-inspired ansatz, which becomes a set of nonlinear relations among Wilson coefficients and shrinks the allowed region toward the Virasoro-Shapiro trajectory. Finally, we study a gravity-pole-subtracted setup, where the regular part of the amplitude has a well-defined forward limit. In this stripped problem, the nonlinear constraints reduce the allowed region to a small island containing the Virasoro-Shapiro point, for which we provide an analytic bootstrap explanation.

The Collins-Soper kernel is calculated from a vacuum soft function using space-like Wilson lines with complex-directional vectors on the Euclidean lattice. Our pure gauge calculations with this method achieve high statistical precision in computing the soft function, whose rapidity dependence is well described by Collins-Soper evolution across a wide range of rapidity differences. The extracted kernel contains errors comparable to those achieved in state-of-the-art lattice calculations based on hadronic observables, but exhibits saturated behavior at large transverse Wilson-line separations.

We compute the one-loop mixing of bosonic and two-fermion interactions into two-fermion operators in the dimension-eight Standard Model Effective Field Theory (SMEFT). Together with the results in arXiv:2106.05291, arXiv:2205.03301, arXiv:2409.15408, and arXiv:2512.21724, this leaves only the mixing of four-fermion operators into two-fermion ones as the remaining piece to complete the SMEFT renormalization program at this order.

We prove a `nonlinear graviton theorem' for higher-spin self-dual gravity. We consider small deformations of the complex structure of the non-projective twistor space that are bounded in a specified region near the origin and investigate the space $M_{HS}$ of holomorphically embedded complex planes $\mathbb{C}^2$ that intersect the origin. We show that this space is an infinite dimensional complex manifold with a canonical projection onto a four-dimensional holomorphic self-dual spacetime $\mathcal{M}$, and discuss the geometry induced on this new higher-spin space. Solutions of higher-spin self-dual gravity are then obtained by choosing an embedding of spacetime $\mathcal{M}$ into higher-spin space $M_{HS}$, with higher-spin symmetries arising from the different choices of embedding. Integrability of the theory is manifested in the form of a Lax pair for the system that we present. We conjecture that chiral higher-spin gravity can similarly be realized by considering deformations that are unconstrained at the origin.

We study the dynamics of first-order inverse phase transitions (driven by superheating) at strong coupling, focusing on the top-down Witten-Sakai-Sugimoto model for holographic QCD. Two cases are considered: the deconfinement transition in the unflavored version of the model and a chiral symmetry-restoring transition occurring in the deconfined phase of the full theory. In both cases, we imagine driving the system into a metastable phase at high temperature, inducing the nucleation of bubbles of the stable phase. For both classes of transitions, we find the corresponding Euclidean bounce solutions and compute the bubble nucleation rates and the relevant transition parameters. For the deconfinement transition, the large jump in the number of degrees of freedom between the two phases suggests that the bubble wall velocity is parametrically small; we provide a rough estimate of it near the critical temperature. In the case of the chiral transition, instead, we compute the bubble wall velocity and the friction force exerted on the bubbles employing motivated ansatze and approximations for the steady-state configurations.

By incorporating leading $c\,$-dependent corrections to the Carroll transformations, we introduce the ``post-Carroll transformations''. We demonstrate that these transformations are consistent with post-Carrollian mechanics \cite{Najafizadeh:2025ksm}; furthermore, they give rise to the so-called ``post-Carroll algebra''. We show that, unlike the Carroll algebra, this new structure allows for a central charge in higher dimensions; we refer to it as the ``Carroll-Bargmann algebra''. To construct conformal extensions, we first build the conformal extension of the post-Carroll algebra and study field theories invariant under this symmetry. We then construct the conformal extension of the Carroll-Bargmann algebra, referred to as the ``Carroll-Schrödinger algebra'', and demonstrate that it precisely matches the symmetry algebra of the higher-dimensional Carroll-Schrödinger theory \cite{Najafizadeh:2024imn}. Finally, we derive the general form of two-point functions in a post-Carrollian CFT, which in $1+1$ dimensions exhibits both electric and magnetic sectors, while in higher dimensions only the magnetic sector survives.

Jet production is the dominant high-$p_{\rm T}$ process at hadron colliders and provides a central testing ground for perturbative QCD, parton distribution functions and determinations of the strong coupling. This contribution summarises recent measurements of inclusive-jet, dijet and jet-multiplicity observables presented at DIS2026, with emphasis on the interplay between experimental precision, next-to-next-to-leading-order predictions, non-perturbative and electroweak corrections, and the treatment of correlated systematic uncertainties. Inclusive jet measurements from CMS and ATLAS constrain the gluon distribution at large Bjorken $x$ and enable extractions of $α_s(m_Z)$ compatible with the world average. Dijet measurements provide complementary sensitivity through the dijet invariant mass, rapidity separation and longitudinal boost, while ratios of inclusive jet multiplicities reduce several experimental and PDF uncertainties and directly probe additional QCD radiation. Recent progress in jet-energy calibration, together with new results from RHIC, ALICE and CMS on jet substructure and heavy-quark radiation, illustrates the breadth of current high-$p_{\rm T}$ jet physics and its relevance for Run~3 and future global PDF analyses.

The Large Hadron electron Collider (LHeC) would add a high-current Energy Recovery Linac to the HL-LHC, delivering electron-proton collisions at centre-of-mass energies around the TeV scale. This contribution summarises the QCD and parton-distribution-function (PDF) aspects of the recent LHeC bridge-project study. The combination of high luminosity, a very large lever arm in Bjorken $x$ and $Q^2$, and clean neutral- and charged-current deep-inelastic scattering measurements would enable a coherent determination of all proton PDFs in a single experiment. The resulting constraints would substantially reduce uncertainties in the gluon, valence, strange and heavy-flavour distributions, provide stringent tests of perturbative and small-$x$ QCD, and improve the parton luminosities that enter precision and discovery measurements at the HL-LHC and at future hadron colliders. The same programme gives competitive and complementary determinations of the strong coupling and weak mixing angle, including measurements of its running over a wide range of scales.

We show that Gauss-Bonnet black branes in five-dimensional anti-de Sitter gravity are unstable when the Gauss-Bonnet coupling falls outside the range allowed by the conformal collider bounds. The unstable modes and the boundary causality violating modes are connected by a phase rotation of complex boundary momentum.

We study the $I^G(J^{PC})=0^+(1^{++})$ $K\bar Kπ$ system in an infinite-volume three-body unitary framework, focusing on the pole content of the region of the $f_1(1285)$ and $f_1(1420)$ resonances. The coupled $πa_0$-$K\bar K^*$ amplitude is constructed in the spectator-isobar representation, where the one-particle-exchange interaction required by three-body unitarity automatically incorporates the triangle-singularity mechanism. The short-range three-body interaction is constrained by fitting the $0^+(1^{++})$ component of the BESIII $K^0_SK^0_Sπ^0$ invariant-mass distribution in the $J/ψ\toγ(K^0_SK^0_Sπ^0)$ decay. Analytically continuing the fitted amplitude to the relevant unphysical Riemann sheets, we find two robust poles: \begin{align} \sqrt{s_{f_1(1285)}}&= \left(1277\pm2\pm1\right) -i\left(12\pm1\pm0\right)\text{MeV}\,,\notag\\ \sqrt{s_{f_1(1420)}}&= \left(1435\pm2\pm7\right) -i\left(40\pm2\pm1\right)\text{MeV}\,.\notag \end{align} The pole trajectories indicate that the $f_1(1285)$ originates from dressing a bare state introduced in the potential. In contrast, the $f_1(1420)$ is predominantly dynamically generated, and a single-channel analysis traces it to an $S$-wave $K\bar K^*$ quasi-bound state mixed with the nearby bare state, supporting its hadronic-molecule interpretation. We also find an additional pole deeper in the complex plane in the best-fit amplitude on the same Riemann sheet as the $f_1(1285)$. This additional pole is generated by the $P$-wave $πa_0$ contact interaction alone. It has a sizable cutoff and two-body-input dependence, and leaves little visible imprint on the physical lineshape. Finally, we provide a detailed and pedagogical appendix on how three-body cuts affect the solution of the integral equation.

Cosmological data have reached the precision needed to turn the neutrino mass ordering from a weak Bayesian preference into a decisive model-selection test. We compute the evidence for the Normal and Inverted Hierarchies by combining DESI DR2 clustering with NuFIT oscillation data. In baseline $Λ$CDM, DESI DR2 plus Planck CamSpec gives $Σm_ν<0.0642\,{\rm eV}$ at 95\% confidence, close to the normal-ordering floor, $Σm_ν^{\rm NH}\simeq0.059\,{\rm eV}$, but well below the inverted-ordering minimum, $Σm_ν^{\rm IH}\simeq0.099\,{\rm eV}$. Thus the inverted hierarchy lies in the tail of the cosmological likelihood. The Bayes factor $K=P(D|{\rm NH})/P(D|{\rm IH})$ exceeds $460$ even for a conservative reference prior, and remains strong, $K>40$, in baseline-model extensions. We show that this result is robust to the choice between a reference prior and a physically motivated logarithmic hierarchical prior, marking the transition from {\em prior-sensitive evidence} to {\em likelihood-dominated exclusion} of the inverted hierarchy within standard cosmology. Embedding these priors in the two-dimensional design space of measure (logarithmic versus linear in mass) and structure (hierarchical versus non-hierarchical), we find that all four prior constructions give decisive evidence under DESI DR2, with residual prior dependence governed mainly by the measure -- a factor $\sim\!10$ in $K$ -- rather than by the hierarchy assumption. At the prior-family level, the evidence favors the SJPV prior predictive over HS by a Bayes factor above $4,700$ across each matched-support variation tested. The favored normal ordering pushes the effective Majorana mass to the few-meV regime, with median $m_{ββ}=3.28\,{\rm meV}$ and 95\% credible interval $0.95<m_{ββ}<11.55\,{\rm meV}$, below the inverted-ordering target for upcoming neutrinoless double-beta decay experiments.

We study four-point functions of identical scalar operators in conformal field theories with AdS duals in large-$N$ expansion. We analyze the appearance of higher-trace operators in theories dual to bulk $φ^3$ and $φ^4$ interactions, focusing on how these operators are required by crossing symmetry. We compute part of the OPE data associated with these operators. We also introduce a diagrammatic framework for organizing the different terms in the conformal block expansion within the large-$N$ expansion. This framework refines the use of crossing symmetry by allowing it to be applied to individual diagrammatic topologies, rather than only to the full correlator. It further separates different contributions to the OPE data by associating them with different cut diagrams. In this language, the emergence of higher-trace operators and their relation to lower-trace OPE data become more manifest.

We present a comprehensive study of the radiative decays of fully heavy tetraquarks ($T_{4c}$, $T_{4b}$) and mixed heavy-flavor tetraquarks ($T_{bc\bar{b}\bar{c}}$) into charmonium/bottomonium states, within the nonrelativistic QCD factorization framework. Numerical result indicates that the decay widths of fully charm tetraquark $T_{4c}$ into $γJ/ψ(η_c)$ are around 1 MeV. Crucially, the $γJ/ψ$ signal channel receives less experimental background near the $J/ψJ/ψ$ threshold, providing a compelling way to search for the possible tetraquark states X(6200). For $T_{4b}$, the decay width is highly suppressed by the bottom quark mass, just lying in tens eV level. We further find that the decay widths of $T_{bc\bar{b} \bar{c}} \to γΥ$ exceed those to $γJ/ψ$ by 3-4 orders of magnitude, indicating preferential $c\bar{c}$-pair annihilation over $b\bar{b}$. These radiative decay modes can be measured in the future experiments, and are helpful to understand the inner structure of the full heavy tetraquark.

We study diffractive photo- and electroproduction of the $S$-wave heavy quarkonia $J/ψ$, $ψ(2S)$, and $Υ(nS)$ at energies relevant for the Electron-Ion Collider (EIC). The production amplitude is evaluated while retaining the full transverse-momentum ($\ell_t$) dependence of the hard two-gluon kernel, that is, without expanding the impact-parameter Bessel kernel as is done in the small-size color-dipole limit. The quarkonia light-cone wave functions are built from Cornell-potential solutions of the Schrödinger equation, normalized to the measured leptonic widths, and combined with a modern collinear gluon distribution. After benchmarking the framework against the full set of HERA charmonium cross-section ratio data, we provide a consistent set of bottomonium cross-section ratio predictions in EIC kinematics. We find that the full $\ell_t$-resolved treatment systematically improves the description of the radially excited states relative to the leading dipole limit, and we identify the kinematic windows where this difference is largest.

The path integral of the Dirac fermion with vector and axial gauge backgrounds is analyzed near the infrared limit in the presence of residual irrelevant current-current interaction. After integrating out fermions, a semiclassical low-energy effective action is obtained, written in terms of currents. Its expression is found to correspond to the hydrodynamic action previously proposed for perfect barotropic fluids with anomalies at zero temperature. This approach also leads to two further hydrodynamic actions to be associated, respectively, with the Weyl fermion, and the Dirac fermion having independent vector and axial currents. These actions feature four- and five-dimensional bulk-boundary terms, owing to anomaly inflow, which are identified as being the so-called transgression forms. These are generalizations of Chern--Simons forms that involve two gauge fields: the dynamical field and the background field. The path-integral argument provides a ``microscopic'' explanation for several ingredients of the action formulation of hydrodynamics that are necessary to incorporate anomalies. It also clarifies the infrared reduction required to pass from the effective field theory to a local hydrodynamic description. This reduction is implemented by considering restricted variations of the action, familiar from hydrodynamics, which at the same time lead to four-dimensional equations of motion from the five-dimensional transgression terms.

We show that the moment of the transfer matrix of the double scaled SYK model is written as an overlap between the stationary state of ASEP (asymmetric simple exclusion process) and a product state. We argue that this overlap is an analogue of the strange correlator appearing in the correspondence between the Levin-Wen string-net model and the Turaev-Viro state sum.

This work presents a comprehensive analysis of polarization transfer in the decays \(χ_{cJ}\) (\(J=0,1,2\)) to octet baryon-antibaryon pairs within a polarized electron-positron collision environment. Using the spin density matrix formalism, we trace the polarization from the initial beams through the production chain \(e^+e^- \to ψ(2S) \to γχ_{cJ}\) to the final-state baryon-antibaryon system. The helicity amplitude analysis for \(χ_{c1} \to B\bar{B}\) confirms the universal angular distribution parameter \(α= -1/3\), as dictated by the charge-conjugation helicity selection rule. For \(χ_{c2}\) decays, \(α\) and the transverse polarization depend on two independent amplitudes, and our quark-model calculations agree with existing data. We demonstrate that the longitudinal beam polarization \(P_z\) modifies the spin observables for \(χ_{c1}\) and \(χ_{c2}\), offering new experimental handles at future polarized facilities like the Super \(τ\)-Charm Facility(STCF) to test decay mechanisms and explore baryonic spin entanglement as a quantum information resource.

Spontaneous CP violation (SCPV) provides an attractive solution to the strong CP problem. However, SCPV after inflation suffers from the formation of CP domain walls, requiring the maximal temperature of the Universe to lie below the CP-breaking scale. In the present work, we then propose a dynamical mechanism that removes this cosmological constraint without introducing permanent explicit CPV. We consider a new scalar field that acquires a large field value with a nontrivial phase in the early Universe and induces a transient bias among degenerate CP vacua through a higher-dimensional interaction with a CP-breaking scalar field. This bias triggers the decay of CP domain walls after they form. As the new scalar field evolves toward the origin, the bias disappears, leaving the low-energy CP structure intact. We derive the conditions for successful domain wall decay and identify the viable parameter space. Furthermore, we point out that the coherent oscillation of the new scalar field naturally survives as dark matter, linking the resolution of the CP domain wall problem to the origin of dark matter.

We systematically investigate the semi-invisible decays of hyperons (hyperon $\to π(/γ)\ +$ invisible($ψ$)) in the Mesogenesis mechanism by the effective Lagrangian approach. The one-loop hadronic contribution of triangle diagrams with final-state interactions is fully examined in the present work. Our analysis indicates that the triangle diagram yield sizable corrections to the branching ratio that are as significant as those from tree diagrams. Especially for the $Σ^-\to π^-ψ$ and $Ξ^0 \to π^0 ψ$, their loop contributions cannot be ignored. Consequently, the branching ratios of hyperon hadronic semi-invisible decays are found to be of order $10^{-5}$, particularly for $Σ^+\toπ^+ψ$, $Ξ^0\toπ^0ψ$, and $Ξ^-\toπ^-ψ$, whereas those of radiative semi-invisible decays are less than $10^{-7}$.

In this work, we construct the simplest pseudo-Hermitian quantum field theory that is supersymmetric. This is the pseudo-Hermitian Wess-Zumino model in the sense that it contains a pair of symplectic fermions (anti-commuting scalar fields) that satisfy the Klein-Gordon equation and a spin-half boson that satisfies the Dirac equation. The conventional spin-statistics theorem is circumvented through the use of pseudo-Hermitian conjugation to define field adjoints. To make the supersymmetry manifest, we formulate the pseudo-Hermitian Wess-Zumino model using the superfield formalism. These superfields are Grassmann-odd so it is not possible to construct non-vanishing cubic interactions using only these superfields. We show that this problem can be resolved by coupling the pseudo-Hermitian Wess-Zumino model with the Hermitian Wess-Zumino model while preserving supersymmetry.

If entanglement wedge reconstruction is exact, then (under certain assumptions) the area term in the RT formula is a c-number, indicating that the choice of a bulk quantum state does not influence the geometry. Recently Cao, Cheng, Karthikeyan, Li, and Preskill considered a generic perturbation away from exact entanglement wedge reconstruction. The optimal reconstruction was defined; based on this, an effective area function that depends nontrivially on the quantum state was defined and its properties were analyzed. Here we make one aspect of this picture more quantitative, by showing that if as expected the area term in the RT formula is of order 1/G while the bulk entropy is of order 1, then the corrections to entanglement wedge reconstruction are exponentially small (in G) relative to corrections to the area function. In the framework under discussion, there is an area function but no area operator; we discuss to what extent this is the expected behavior in holography.

The gluon dominance model describes multiparticle production of secondary particles at high energies in lepton and hadron interactions, including annihilation processes and heavy quarkonium decays. According to this model, the multiparticle process is divided into two stages. The first stage describes the development of a quark-gluon cascade as a Markov branching process in the region of perturbation QCD. For the second stage, the transformation of quarks and gluons into observable hadrons (hadronization), a phenomenological scheme is proposed. It is universal and based on an experiment. The gluon dominance model demonstrates good agreement with data over a wide energy region. It testifies that in hadron interactions valence quarks remain in the leading particles, and gluons are the sources of secondary hadrons. Quantitative estimates of the model parameters confirm the fragmentation mechanism of hadronization in leptonic interactions and the recombination mechanism in hadronic ones. The model description of the experimental distributions on the number of neutral pions in proton interactions at 50 GeV beams in the high multiplicity region are presented for the first time. It is shown that the main contribution to this region is made by gluon fission. These results can be useful in planning of future experiments.

The two gluon exchange model serves as a key framework for describing the photoproduction of heavy vector mesons, and the photoproduction cross sections derived from it provide essential input for studies of Ultra-Peripheral Collisions (UPCs). Building on the model successful description of $J/ψ$ and $ψ(2S)$ photoproduction in previous work, we use the STARlight program to systematically investigate charmonium UPCs in PbPb and $p$Pb collisions at LHC energies. To reduce discrepancies between theoretical predictions and experimental data for rapidity and transverse momentum distributions in PbPb collisions, a phenomenological suppression factor is introduced to correct the theoretical results. We find that, in the TeV energy range, the corrected predictions agree well with experimental data within uncertainties and successfully reproduce the characteristic double-peak structure in rapidity distributions. In contrast, no significant suppression is observed in $p$Pb UPCs, which reflects the asymmetric photon fluxes and the dominant contribution from the photoproduction interaction branch. The transverse momentum distributions from STARlight simulations also match the diffractive pattern of coherent production seen experimentally, although the overall yield remains slightly overpredicted. This work further validates the applicability of photoproduction cross sections from the two gluon exchange model for charmonium UPC studies, and offers valuable phenomenological guidance for future experimental design and data analysis in UPC measurements of charmonium production.

We present a systematic analysis of the semi-leptonic decays $B_{(s)}\to D_{(s)}(1S,2S)\ellν_\ell$ and $B_{(s)}\to D^*_{(s)}(1S,2S)\ellν_\ell$ with $\ell=e,μ,τ$ within the covariant light-front quark model (CLFQM). Using the form factors of the transitions $B_{(s)}\to D_{(s)}(1S,2S)$ and $B_{(s)}\to D^*_{(s)}(1S,2S)$, we calculate the branching ratios of the relevant semi-leptonic decays and find that $Br(B_{(s)}\to D_{(s)}\ell^\primeν_{\ell^\prime})$ and $Br(B_{(s)}\to D^*_{(s)}\ell^\primeν_{\ell^\prime})$ with $\ell^\prime=e,μ$ are agree well with the data, while $Br(B_{(s)}\to D_{(s)}τν_τ)$ and $Br(B_{(s)}\to D^*_{(s)}τν_τ)$ are systematically smaller than the experimental measurements. This naturally gives rise to the so-called $\mathcal{R}(D)$ and $\mathcal{R}(D^*)$ anomalies. Our predictions $\mathcal{R}(D)=0.261\pm0.013$ and $\mathcal{R}(D^*)=0.228\pm0.026$ show $3.1σ$ and $2.1σ$ deviations from the current experimental world averages compiled by the Heavy Flavor Averaging Group (HFLAV), respectively, yet only deviate by $0.16σ$ and $1.5σ$ from the latest LHCb measurements. For the decays $B_{(s)}\to D_{(s)}(2S)\ellν_\ell$ and $B_{(s)}\to D^*_{(s)}(2S)\ellν_\ell$, their branching ratios lie in the range $10^{-4}\sim10^{-3}$, which are much larger than the results from the Bethe Salpeter (BS) equation , but agree with the relativistic quark model (RQM) calculations. Furthermore, we also calculate the forward-backward asymmetries $\mathcal{A}_{FB}$ and longitudinal polarization fractions $f_L$ for the corresponding decays. Our predictions are consistent with most other theoretical results and experimental data

We present methods for electromagnetic shower reconstruction and identification in the FASERnu emulsion detector using 100 GeV and 200 GeV electron test-beam data from the CERN SPS H4 beamline. The reconstruction employs a clustering-based algorithm without energy-dependent tuning to determine shower axes. A multi-level identification chain comprising track pre-selection, a cut-based selection, and a BDT classifier achieves combined background rejection rates of 99.99% (100 GeV) and 99.94% (200 GeV). The method reaches total reconstruction and identification efficiencies of 58.9% (100 GeV) and 70.8% (200 GeV) evaluated from simulated samples. Energy reconstruction using the total number of reconstructed segments as the calorimetric estimator yields relative biases of +0.6% (100 GeV) and -0.8% (200 GeV), with resolutions of 25.4% and 22.6%, respectively. Systematic uncertainties on the energy reconstruction are dominated by variations in emulsion film detection efficiency, contributing (+10.9%/-8.2%) at 100 GeV and (+10.3%/-6.9%) at 200 GeV. The methodology provides a validated framework for electron neutrino identification with the FASERnu detector at the LHC.

We analyze the mixing between the $π^0$ meson and the photon in the presence of a strong uniform magnetic field, in the context of a two-flavor Nambu-Jona-Lasinio model. It is shown that only one specific photon polarization state can get mixed with the $π^0$ state, a fact that can be understood on general grounds. For magnetic fields with values up to 1 GeV$^2$/e, it is found that the effect of the mixing on the pion mass and pion-quark couplings is relatively small (below a level of 15%), which is at variance with previous results reported in the literature.

Bayesian model selection in cosmology and particle physics is often performed where posterior odds inherit a strong, often unacknowledged dependence on the prior assigned to competing models. Standard responses -- reference priors, hierarchical priors, or appeals to naturalness -- ignore relevant theoretical knowledge or rely on criteria hard to define operationally. We propose the \emph{Coherence Principle}: a reproducible prescription for assigning model priors according to compatibility with the validated structure of an existing theory. This structure, or \emph{grammar}, includes symmetries, conservation laws, locality, Lorentz invariance, and universality patterns. Unmotivated violations of these rules incur a coherence cost, converted into a prior weight through a maximum-entropy exponential form controlled by one calibratable parameter $α$. The resulting prior is distinct from both the Bayesian Occam factor and naturalness: it penalizes not parameter volume or fine tuning, but departures from validated theoretical grammar. We illustrate the principle with examples from cosmology and fundamental physics: neutrino mass mechanisms, dark energy and modified gravity, inflation, beyond-Standard-Model sectors, and hierarchical astrophysical inference. We test it also on four historical cases -- general relativity, Pauli's neutrino, parity violation, and special relativity -- where evidential and theoretical contexts can be reconstructed. These examples show that it favors the historically successful choice when the proper grammar is defined in the correct domain and time. The Coherence Principle makes explicit a common but usually tacit part of physical reasoning: trust in validated structural rules. It turns this judgment into a transparent, testable, and overrulable component of Bayesian inference, leaving empirical likelihoods free to dominate when data are sufficiently constraining.

The large number of $Z \rightarrow ττ$ events expected during the TeraZ program at FCC-ee will allow for precision measurements and searches for physics beyond the Standard Model, requiring accurate reconstruction of hadronically decaying tau leptons. This reconstruction is particularly challenging due to the presence of undetected neutrinos and the diverse topology of hadronic tau decays, making the design of robust heuristic reconstruction algorithms challenging. In this work, we present the first fully machine learned hadronic tau reconstruction approach tuned for FCC-ee studies. The reconstruction is formulated as a set of complementary tasks, including tau identification, decay mode classification, charge reconstruction, and full four-momentum regression. The algorithms are evaluated on fully simulated electron--positron collision samples with realistic detector effects using the CLD detector setup. We compare dedicated task-specific models with a unified multi-task model and quantify their performance in a granular manner across all reconstruction tasks. Both approaches achieve per-mille-level tau mis-identification rates at high signal efficiency, decay mode classification F1 scores of up to 0.95 for the dominant channels, and sub-per-mille charge mis-identification rates, outperforming a conventional jet-charge estimator by up to two orders of magnitude. For the full kinematic reconstruction, the models achieve per-mille-level angular resolution and percent-level visible transverse momentum resolution, exceeding the performance of reconstruction-level jet observables. The resulting models provide a realistic high-performance solution for hadronic tau reconstruction at FCC-ee, offering identification, charge discrimination, decay mode analysis and full kinematic reconstruction.

We study the semiclassical limit of SL_2-opers on the four-punctured sphere in Nekrasov-Rosly-Shatashvili Darboux coordinates. Using exact Wentzel-Kramers-Brillouin (WKB) connection formulae, we express the trace coordinates of the corresponding SL_2(C) character variety as finite Laurent sums of Voros exponentials. Tropicalizing these formulae and the NRS relations gives a chamberwise integer affine linear system for the leading logarithms of the NRS coordinates. In chambers where this system is unimodular and the selected cycles form a primitive symplectic pair, the leading asymptotics agree, up to flavor-period shifts, with Seiberg-Witten periods of the N=2 SU(2) theory with N_f=4 fundamental hypermultiplets. We verify this mechanism in a sample chamber and in the weak-coupling degeneration. No global coordinate-independent recovery theorem is claimed; non-unimodular or degenerate chambers are treated as limitations of the chosen NRS chart. In the weak-coupling degeneration, we show that the NRS chart can be chosen compatibly with the plumbing limit so that the resulting chamber is unimodular and non-degenerate away from tropical walls.

We study two-point functions of half-BPS operators in maximally supersymmetric Yang-Mills theory continued to $d=4-2\,\varepsilon$ dimensions. Using supersymmetric localization on $S^d$, we derive perturbative matrix-model expressions for the $\varepsilon$-expansion of these correlators and obtain all-loop results at leading order in $\varepsilon$ in the planar limit, with extensions to finite-$N$ corrections and higher-charge operators. We compare the localization results with direct perturbative computations in flat space. At order $\varepsilon$ the two descriptions agree perfectly, while at higher orders our construction fails to reproduce the perturbative data due to the breaking of conformal symmetry away from four dimensions. Nevertheless, in the case of the dimension-two operator we conjecture an all-loop formula at order $\varepsilon^2$ by exploiting the precise form of the mismatch.

The fast reconstruction and detector-simulation pipelines are widely used in different scientific domains, such as, in High Energy Physics (HEP) and medical imaging, where the full experimental or the device-level simulation is computationally challenging. Instead of the use of the global data context, these pipelines use simplified response models which assume reconstruction uncertainty relies on local input parameters. To probe the robustness of the local assumption, a domain-agnostic context-aware stress testing pipeline is introduced, and a reconstruction response depending on the global parameters is also allowed. For an instance in HEP at High-Luminosity LHC (HL-LHC) simulation, this work shows that the decay channel $Z \rightarrow \ell\ell$, as a benchmark, violates the local assumption resulting in a significant reconstruction bias and resolution degradation. Using the unsupervised regime-mapping framework, this work also restores this peak stability and recovers the truth-level resolution, where a robust diagnostic tool for next-generation fast simulation pipelines is accommodated.

Type II string theory in the absence of orientifolds requires the target spacetime $X$ to admit a spin structure. We show that this familiar requirement arises directly from the consistency of the worldsheet GSO projection. The obstruction is a mixed global anomaly between $(-1)^{\text{F}_L}$ and the target space background data, detected by the spin bordism group $Ω_3^{\mathrm{spin}}(B\mathbb{Z}_2\times X)$. We compute the relevant mixed bordism group and identify the bordism class of the GSO-projected worldsheet theory. For smooth target spaces, vanishing of the anomaly reduces to the condition that $X$ is spin, while for general orbifolds $[\hat{X}/G]$, $\hat{X}$ has to carry a $G$-equivariant spin structure. We also classify all possible theta angles in the worldsheet theory and show that they correspond to all possible continuous and discrete background fields of the target space theory visible in string perturbation theory.

While the Standard Model (SM) of particle physics contains the most precise set of predictions ever devised by humanity, that precision comes at a cost. The strange nature of the Higgs particle requires its parameters to be tuned so precisely that if the SM is indeed the true description of reality, one is forced to wonder how such a miracle as galactic structure and life could occur. Instead, we search in this work for a natural explanation. The concept of naturalness is comprehensively explored, and a new tuning measure proposed, with an aim to place it on well-defined Bayesian footing. We then turn this measure on to the analysis of a class of intriguing new physics - Composite Higgs models. These effective models are the result of a plethora of underlying theories, and they allow the production of a naturally light Higgs particle, appearing as the SM Higgs at low energy. We establish the background required to appreciate the N-site 4D Composite Higgs model, and subsequently focus on the simplest incarnations of this class. A global fit is performed on the Minimal 4D Composite Higgs model (M4DCHM), with strong exclusion bounds placed on collider search channels. We analyse any improvement in tuning that could be gained from several extensions to this model. The Leptonic M4DCHM is explored, with a composite tau lepton embedded in various representations. The possibility of a dark matter candidate existing in the Next-to-Minimal 4DCHM is considered. Ultimately, we are able to define what, if any, benefit to naturalness can come to the Composite Higgs sector by introducing these extensions.

We study two-point functions of protected scalar operators in N=4 SYM, focusing on their transcendentality properties in dimensional reduction, where quantum corrections are subleading in the regulator. We compute the correlators explicitly through two loops and operators up to classical dimension 10, for all trace structures. The one-loop correction is universal. At two loops, we find a controlled partial breaking of uniform transcendentality for higher-dimensional operators, which can be cancelled by suitable combinations of correlators in a fully predictable way. A main result is a complete planar extrapolation for two-loop correlators at arbitrary dimension and trace structure, whose dependence is entirely controlled by the number of stress-tensor multiplet factors in the operator. The perturbative results agree with localization predictions in all cases where comparisons are possible.

The success of DUNE's MeV physics program depends upon high-precision knowledge of the charged-current (CC) $ν_e+\mathrm{^{40}Ar}$ cross section. While there are indirect constraints at the 10% level for the nuclear transitions that constitute this cross section, the only direct measurement in the MeV range has an uncertainty of $\sim$50%. We show, surprisingly, that the cross section can be precisely measured using the solar-neutrino data themselves. This is possible because of independent knowledge of the $^8$B flux and survival probability, plus the distinctive angular distributions of the Fermi and Gamow-Teller transitions that comprise the cross section. We propose new methods to extract the transition strengths, considering both intuitive groupings and a Principal Component Analysis. Under pessimistic assumptions about detection, but taking detector uncertainties to be controlled, we demonstrate that a precision of $\lesssim$2% on the cross section can be achieved in the 9-15 MeV energy range. These results will be an important foundation for studying the cross section up to several tens of MeV, where the complexity increases significantly due to nuclear breakup channels but where reducing uncertainties is critical for supernova and atmospheric neutrino studies.

A long-standing conjecture claims that Virasoro conformal blocks exponentiate in the semiclassical limit $c \to \infty$ with $h/c$ finite. However, this has been proven only for four-point blocks on the sphere and one-point blocks on the torus. Here we extend the proof to general conformal blocks for higher-point functions and higher-genus backgrounds in arbitrary channels. The statement is to be understood at the level of a formal power series. Our proof builds upon a novel extension of the oscillator method for the computation of conformal blocks to cases where three internal lines meet at a vertex. This extension also gives a new constructive method to compute global conformal blocks in 2d CFTs at general genus.

Pulsar timing arrays (PTAs) are sensitive to the gravitational influence of passing compact substructures, which can produce Doppler timing delays by accelerating pulsars or the Solar System barycenter, and Shapiro timing delays when passing near Earth--pulsar lines of sight. Projections for the complete PTA sensitivity to compact dark matter (DM) substructures, such as primordial black holes and axion miniclusters, are challenging due to the variety of signal types ranging from rare, nearly static encounters, to dynamic flybys, to the stochastic limit of many substructures. We address this challenge with a framework that combines Monte Carlo signal modeling and machine-learned surrogate likelihoods, enabling a unified likelihood-level analysis of signals previously treated only in simplified limiting regimes. We then use this framework to precisely assess the impact of a stochastic gravitational wave background (SGWB), for which evidence was recently found, on the PTA sensitivity to compact DM substructures.The SGWB substantially weakens the sensitivity, and we find that in even the most optimistic observing scenario only a Shapiro search retains sensitivity to subdominant DM components when assuming SGWB parameters inferred from current measurements.

The Deep Underground Neutrino Experiment (DUNE) far detector is sensitive to $\mathcal{O}(10)$ MeV electron neutrinos through $ν_e$ charged-current reaction with argon. This capability is a unique window into the $ν_e$ component of a Galactic core-collapse supernova flux. Extracting the properties of the supernova spectrum is, however, limited by the poorly-known $ν_e$-Ar cross section. We propose a data-driven strategy that leverages $^8$B solar neutrinos and muon-decay-at-rest neutrinos as standard candles for this process. These calibration samples constrain both the low-energy and high-energy components of the cross section relevant for supernova detection. Our method reduces the reliance on nuclear models, which can bias the extraction of the spectral parameters by as much as 300$\%$.

We establish a new connection between supersymmetric theories and scattering amplitudes. We show that the Coon amplitude coincides with the 3d $\mathcal{N}=2$ half-index of the XYZ model with nontrivial boundary conditions. Our 3d theory, intrinsically defined in the UV, flows to a sigma model in the IR whose partition function is the Veneziano amplitude. Crossing symmetry is realized as a consequence of 3d $\mathcal{N}=2$ mirror symmetry between XYZ and SQED. We use this correspondence to construct a meromorphic modification of the Coon amplitude by promoting the long-standing dressing factor $\mathfrak{q}^{ST}$ responsible for a branch cut to an elliptic completion thereof. This illustrates that one does not have to give up single-valuedness to achieve positivity at the physical poles.

We compute the one-loop weak corrections to the annihilation cross sections of fermionic Minimal Dark Matter multiplets. Infrared divergences cancel in the dominant $s$-wave combination relevant for the thermal relic abundance. Instead, infrared-enhanced corrections affect velocity-suppressed rates, through a Sudakov/Sommerfeld interplay. The corrections grow with the multiplet size and are at the $5\%$ level in the most motivated cases: the Higgsino-like doublet, the wino-like triplet, the stable quintuplet.

This paper presents a measurement of photon plus two-jet events in $pp$ and Pb+Pb collisions, i.e. events in which the transverse momentum of a single photon is balanced by two distinct jets. The measurement was performed using $pp$ data taken in 2017 with an integrated luminosity 260 pb$^{-1}$, and Pb+Pb data taken in 2018 with an integrated luminosity 1.72 nb$^{-1}$, both at $\sqrt{s_\mathrm{NN}}$ =5.02 TeV, as recorded by the ATLAS detector. Events with photons in the transverse momentum range 90-180 GeV and at least two anti-$k_t$ $R = 0.2$ jets with a $p_\mathrm{T}$ > 30 GeV are selected, and three observables are measured: ${\mathrm{x}}_{\mathrm{JJ}γ}$, $\mathrm{A}_{\mathrm{JJ}γ}$, and $ΔR_{\mathrm{JJ}}$. These observables characterise the overall energy loss of the multiparton system from medium interactions (${\mathrm{x}}_{\mathrm{JJ}γ}$), the relative energy loss between the two colour-charge carriers ($\mathrm{A}_{\mathrm{JJ}γ}$), and medium-induced modifications to their opening angle ($ΔR_{\mathrm{JJ}}$). The observables are corrected for uncorrelated combinatoric background contributions using a novel multijet mixing technique, for photon purity, and for detector resolution effects via iterative unfolding. Final results are presented per photon, and the ratio ($I_\mathrm{AA}$) is taken between measurements in Pb+Pb and $pp$ collisions, for Pb+Pb centrality intervals of 30-80%, 10-30%, and 0-10%. Significant suppression of per photon two-jet yields in all three observables, $I_\mathrm{AA} < 1$, is observed as a result of parton-medium interactions. The experimental measurements are compared to three different jet quenching models: JEWEL, JETSCAPE, and LBT.

In this work, a systematic investigation of the production of high-orbital-excitation kaons in $K^{-}p$ reactions is carried out within an effective Lagrangian approach. The relevant $t$-channel processes are constructed, and the model is calibrated using a single adjustable parameter determined from existing experimental data. With this parameter, the measured production cross sections for the $K_3^*(1780)$, $K_2(1820)$, $K_2(1770)$ and $K_4^*(2045)$ states are successfully reproduced. Employing the same framework, the production cross sections for other high-orbital kaons are predicted. The results indicate that these states possess sizable cross sections and exhibit characteristically forward-peaked angular distributions, which is a typical feature of $t$-channel exchange, highlighting their great potential for observation in future experiments.

We present spatially resolved parsec-scale measurements of nuclear conditions (gas density and kinetic temperature) relevant for black hole accretion rate predictions in the Seyfert 2 galaxy, NGC 1068. We inject these parameters into the prescription for a Bondi-like accretion model, then compare the resulting accretion rate prediction to the empirical accretion rate derived from hard X-ray observations. Cosmological simulations have spatial resolution ranging from $\sim$10 pc to $\sim$kpc scales, and so for reasonable comparison we test these accretion rate predictions in pixel-sized radial steps out to 500 pc. Compared to warm H$_2$ gas, CO gas is the dominant mass carrier close to the SMBH. We find that the Bondi accretion rate ($\dot{\mathrm{M}}_{\mathrm{Bondi}}$) of cold molecular gas alone (measured using CO) overestimates the true accretion rate by up to 14 dex in a small aperture (r$\lesssim$5 pc) around the black hole, and by at least 8 dex inside large apertures (r$\lesssim$500 pc). These results are the first in a series of direct tests of accretion rate prescriptions, and they suggest that using a Bondi accretion formalism to model supermassive black hole accretion in Seyfert 2 galaxies may lead to overestimated accretion rates in simulations.

Dense gas is the direct fuel for star formation, but measuring it has long been difficult at z>2, especially in typical star-forming main-sequence galaxies. In this work, we report the first detection of HNC (J = 5--4) and CN (N = 4--3) emission in a massive main-sequence galaxy, BX610, at z=2.21. The velocity integrated emission of HNC(5--4)+CN(4--3) is concentrated in the galactic centre, coincident with the region of ongoing intense star formation. Based on line decomposition, we measure a line flux ratio HNC(5--4)/CN(4--3) of $1.05\pm0.23$, similar to that of starburst galaxies at comparable redshifts but lower than that of quasar/AGN host galaxies. The comparatively fainter HNC(5--4) disfavours the presence of a strongly buried AGN in BX610, consistent with optical line diagnostics. The radiative transfer analysis favours the presence of dense gas with a density of $(2-4)\times10^{6}\,\text{cm}^{-3}$ and a kinetic temperature of 50-80 K. The derived abundance ratio between N(HNC) and N(CN) favours dense gas clouds near photodissociation regions, as commonly seen in typical starburst environments. The inferred dense-gas line luminosity closely follows the scaling relation between far-IR and dense-gas line luminosities established for local luminous infrared galaxies (LIRGs). Our observations support the view that star formation in cosmic noon galaxies is primarily controlled by the availability of dense gas, which could be enhanced in central galactic regions with efficient cold gas inflows as observed in BX610 along the inner spiral arms and a possible stellar bar.

PSR J1913+1102 is a highly asymmetric double neutron star system and an excellent laboratory for testing scalar-tensor gravity theories, as well as a potential progenitor analogue of GW170817 that will merge in 470 Myr. We present an updated timing analysis combining 13 years of historical Arecibo observations and new FAST measurements, using two approaches to model dispersion-measure variations. The new timing solution provides precise measurements of four post-Keplerian parameters and improves the system mass estimates. Assuming general relativity and modelling the DM variation with a Gaussian process, we obtain a three-fold improvement in the total mass, m_{tot}=2.88948(20) M_\odot, and nearly four-fold improvements in the pulsar and companion masses, m_p=1.599(8) M_\odot and m_c=1.290(8) M_\odot, giving the mass ratio, q=0.807(8). We also measure an improved proper motion, μ=7.71(25) mas yr^{-1}, enabling a more accurate correction of the observed orbital-period derivative. Combined with the improved orbital-decay measurement, this yields an intrinsic orbital-period derivative \dot{P}_b^{intr}=-4.60(6)\times10^{-13} s s^{-1}, five times more precise than the previous value and fully consistent with the general-relativistic prediction for gravitational-wave damping. The improved masses and precise \dot{P}*b^{intr} place stringent constraints on dipolar gravitational-wave emission and the spontaneous-scalarisation window around 1.6 M*\odot. The refined proper motion and mass measurements also provide tighter constraints on the final helium-star mass immediately prior to its core collapse and formation of the second NS in a supernova, as well as on the magnitude and direction of the associated natal kick of the DNS system.

JWST images in the MIRI filters are characterized by prominent interstellar bubbles, most of which are expected to be created by mechanical energy injected into the interstellar medium by dying massive stars. In this work, we use resolved stellar populations (RSPs) in JWST/NIRCam images of NGC 628 from the JWST-FEAST dataset to determine the demography of stellar populations within these bubbles by comparing them with PARSEC+COLIBRI isochrones using a Bayesian framework. Our analysis reveals the presence of multiple stellar populations younger than 100 Myr within the bubbles, suggesting secondary star formation following the first generation of stars that formed the bubble. We find a clear preference for the most recently formed stars to lie closer to the bubble shell. In contrast, relatively older stars are distributed throughout the bubble interior, consistent with a scenario in which stars formed in dense shells are left behind as the shell expands. We also examine PHANGS-HST cluster candidates within large bubbles and find no convincing progenitor clusters responsible for the initial trigger, indicating that low-mass clusters or OB associations may be sufficient to drive the initial expansion. However, the cluster may dissolve as it co-evolves with the bubble, producing the dispersed stellar distributions observed in larger bubbles. We establish a fundamental plane relation for stellar feedback-driven bubbles that involves the stellar population mass, age, bubble size, gas density, and feedback efficiency, highlighting the ability of JWST/NIRCam CMDs to characterize stellar populations driving interstellar bubble expansion in nearby galaxies

In $Λ$CDM cosmology, galaxies and clusters form within dark matter halos and merge in the hierarchical assembly paradigm to form massive systems. Using the released optical survey data, we searched for groups composed solely of dwarf galaxies, each with a stellar mass $M_*<10^{9.5}~M_{\odot}$. We identified 14 dwarf galaxy groups with at least 5 dwarf galaxies, all located within a projected radius of 200 kpc and with a line-of-sight velocity of $\pm$300 km~s$^{-1}$. We checked photometric and imaging data and found that these 14 dwarf galaxy groups are solitary, with no neighboring massive galaxies with $M_*>10^{10}~M_{\odot}$ within 500 kpc and within $\pm$1200 km s$^{-1}$. The stellar mass fractions of dwarf galaxy groups with $M_{\rm dyn}>10^{12}~M_{\odot}$ are much lower than predicted by the canonical stellar mass and halo mass relation. These dwarf galaxies are gravitationally bound within halos with a dynamical mass of around $M_{\rm dyn} \sim 10^{12}~M_{\odot}$ and a virial radius of less than 400 kpc. These dwarf galaxy groups, therefore, indicate primordial halos that host only a few newly formed dwarf galaxies.

Recollimation shocks are a frequent outcome in overpressured relativistic jets and are crucial for interpreting stationary features in Active Galactic Nuclei. The precise influence of magnetic fields on jet stability, energy dissipation, and variability remains debated, particularly as different field configurations can significantly alter shock properties and the onset of fluid instabilities. We perform a study of 2D axisymmetric RMHD jets to quantify how the ambient density contrast ($ν$), pressure ratio ($P$), magnetization ($σ$), and magnetic pitch parameter ($α$) govern the formation and strength of the first recollimation shock. We also assess how these parameters create the local geometric conditions favorable for the centrifugal instability (CFI), utilizing linear theory as a diagnostic. We find that the jet's global geometry is affected by the magnetic pressure. The recollimation distance decreases monotonically with increasing magnetization $σ$, as increased magnetic forces immediately limit jet expansion. Remarkably, in the magnetically dominated regime, the ratio of the magnetized recollimation distance ($z_{\rm MHD}$) to its purely hydrodynamic counterpart ($z_{\rm HD}$) converges onto a power-law scaling, $z_{MHD}/z_{HD} \propto (B_0^2/P_{ext})^{-1/3}$, where $B_0$ the initial magnetic field and $P_{ext}$ the external pressure. Jets with high density contrast relative to the ambient medium or high internal pressure further enhance field compression. Furthermore, synthetic synchrotron maps show that a dominant toroidal field yields highly boosted, localized emission knots, whereas a strong poloidal field creates a diffuse profile and shifts the recollimation zone downstream. Regions susceptible to CFI are determined primarily by the local $σ_{\text{tor}}/Γ^2$ profile and streamline curvature created during recollimation.

We present phantom (Profile and Halo Analysis for Numerous Theoretical dark Matter Observables), a public MATLAB toolbox and Octave package for calculations that connect the linear density field to dark matter halo observables. The package combines a flexible cosmology module with linear power spectrum, variance, and correlation function solvers, and a halo module that covers mass functions, linear bias, density profiles, and concentration-mass relations for cold, warm, and fuzzy dark matter scenarios. All core routines are validated against the Python package colossus, hmf, and halomod, yielding sub-percent agreement for shared models across distances, power spectra, variance, correlation functions, halo mass functions, and density profiles. Phantom is organised around a cosmology structure that stores background expansion, growth, and linear power-spectrum handles; this object is constructed once and passed through the call graph, so that halo statistics and halo structure calculations remain consistent by design. From this single entry point, users can obtain field statistics (power spectrum, variance, correlation function), halo statistics (mass functions, linear bias), and halo observables (enclosed mass, circular velocity, projected density, and lensing convergence) on arbitrary user-defined grids. The toolbox targets users whose analysis pipelines are written in MATLAB or Octave, where a validated native implementation of these models has been absent. The code is released under the MIT licence at phantom(https://github.com/matc-thaher/PHANTOM).

Following the first observational study of the two-dimensional spatial distribution of stellar specific angular momentum (sAM) in late-type galaxies, we quantify the morpho-kinematic diversity of galaxy simulations using the newly proposed j-types classification. We analyse the stellar sAM surface density (sAMSD) of $\sim$8000 TNG50 stellar discs spanning $0 \leq z \leq 3.5$ and $9.5 \leq \log(M_\star/\mathrm{M}_\odot) \leq 11.2$, selected from the TNG50 MW/M31 parent sample. We characterize their j-substructures using four morpho-kinematic metrics derived from comparisons with the Freeman sAMSD distribution and the Fourier decomposition of the galaxies in the sAMSD space. A Gaussian mixture model with four fully covariant components assigns each galaxy a probability of belonging to one of four j-types. We find that TNG50 discs exhibit a morpho-kinematic diversity consistent with observations, redistributing stellar angular momentum through four dominant j-substructures that evolve with redshift as follows: j-irregulars ($\bar{z}=0.91$), j-spirals ($\bar{z}=0.76$), j-rings ($\bar{z}=0.62$), and j-bars ($\bar{z}=0.39$). The gas fraction and stellar rotational support ($V/σ$) drive this evolution: gas-rich galaxies preferentially host j-irregulars and j-spirals, whereas gas-poor systems favour j-rings and j-bars. At fixed gas fraction, higher $V/σ$ favours j-spirals and j-rings, respectively. We conclude that there is a canonical pathway for the redistribution of angular momentum within galactic discs undergoing secular evolution in TNG50, accessible only through their morpho-kinematic description. The sAMSD analysis links variations in stellar dynamics to their consequences for mass redistribution, enabling the reconstruction of comprehensive galactic evolutionary histories.

Magnetic fields are widely believed to play an important role in molecular-cloud evolution and star formation. However, the extent to which newly formed stars retain a dynamical relationship with the magnetic environments in which they formed remains poorly understood. In this study, we investigate the relationship between the proper motions of young stellar objects (YSOs) and local magnetic-field orientations in seven nearby molecular-cloud regions. Proper motions were obtained from Gaia Data Release 3, while magnetic-field orientations were derived from Planck 353 GHz dust-polarization measurements. For each YSO, we computed the angular separation $Δθ$ between the projected proper-motion vector and the local magnetic-field orientation in Galactic coordinates. A total of 2,037 YSOs were analyzed across Chamaeleon I, Perseus, Ophiuchus, Orion South, Orion North, Taurus, and Lupus. Significant departures from random orientation distributions are detected in every cloud. However, the preferred alignment varies strongly among regions. Chamaeleon I, Perseus, Ophiuchus, and Orion South exhibit preferential alignment between stellar motions and magnetic fields, whereas Orion North, Taurus, and Lupus exhibit preferentially perpendicular orientations. The median angular separation spans a continuous range from $15.0^\circ$ in Chamaeleon I to $67.6^\circ$ in Lupus. These results demonstrate that the relationship between YSO motions and magnetic fields is not universal but depends strongly on the local star-forming environment. The findings suggest that young stellar populations retain measurable kinematic signatures of their natal magnetic environments and provide a new observational framework for investigating the role of magnetic fields in star formation.

(Abridged) We compare the radial abundance gradients of O, Ne, Cl, and Ar using a sample of 42 HII regions and 176 planetary nebulae (PNe) from the DESIRED catalogue in the Galactic disk, comprising the highest-quality observations currently available and presenting the first gradient analysis for this dataset. For all objects, two sets of chemical abundances were compiled: one derived from collisionally excited lines (CELs) and another incorporating the temperature fluctuation parameter (t^2). Oxygen abundances were corrected for dust depletion, and HII region results were compared with Cepheid stars, which trace the present-day interstellar medium. PNe distances were compiled from recent literature, excluding bulge or halo objects to ensure a disk-only sample. All gradients are statistically significant (p < 0.05), except for Cl and Ar in HII regions. The O/H gradients from HII regions and Cepheids are consistent when t^2 is included, underscoring the importance of accounting for temperature inhomogeneities in nebular analyses. The O and Ne gradients traced by older objects are flatter than the present-day gradient by -0.028 +- 0.008 dex kpc-1 on average from both RLs and CELs. This could indicate a temporal steepening of the Galactic abundance gradient; however, this behavior is not reproduced by chemical evolution models, suggesting additional physical processes are at play. The most plausible explanation is that our PNe sample has been strongly affected by radial migration. Under this interpretation, the PNe gradient cannot reliably trace past abundance gradients, but provides a valuable constraint on radial stellar migration, offering important input for chemo-dynamical models of the Galactic disk and for hydrodynamical simulations.

Superdisks have emerged as an active area of research in recent years, and J0116-473 represents a promising target for studying this extended structure. Our primary objective was to search for HI absorption associated with the suspected superdisk. However, no such absorption feature was detected, suggesting a low, or absence of neutral hydrogen content in the superdisk. In addition, we examined a compact point source located near the galaxy's core and the presumed plane of the superdisk, enabling us to search for HI absorption against this background continuum. We also present a detailed multi-band morphological analysis of the galaxy using Giant Metrewave Radio Telescope (GMRT) observations in Bands 3, 4, and 5. A spectral analysis of both the galaxy and the nearby point source was carried out using data from these three frequency bands. A systematic steepening of the spectral index is observed from the core toward the lobes, as expected for aging synchrotron-emitting plasma. We also found that the northern inner lobe exhibits a significantly steeper spectrum than its southern counterpart, possibly reflecting environmental effects associated with the proposed superdisk. Since superdisks are expected to contain hot, ionized gas, we additionally examined archival X-ray observations from the XMM-Newton telescope. Although diffuse X-ray emission associated with the radio lobes is visible, no significant emission is detected from the region corresponding to the suspected superdisk.

The assumption of large-scale homogeneity is a cornerstone of modern cosmology and underlies the validity of the FLRW framework. Testing the scale at which the Universe transitions to homogeneity remains a key observational challenge, particularly with the increasing precision of galaxy surveys. We aim to assess the robustness of the angular homogeneity scale, $θ_H$, by systematically comparing different computational approaches used in its estimation and by quantifying the impact of methodological choices and physical clustering scales. We analyse mock galaxy catalogues from the MICE Grand Challenge simulation. The angular fractal dimension $D_2(θ_H)$ is computed using the Landy-Szalay estimator and direct pair-counting methods. We implement different approaches, including symbolic regression, to model $D_2(θ_H)$ and determine $θ_H$. Uncertainties are estimated using resampling techniques and alternative parametric error propagation methods. We find that the estimation of $θ_H$ is sensitive to methodological choices in the analysis, such as survey area, redshift bin, numerical implementation and fitting strategy. While its redshift evolution is robust, its absolute value is sensitive to both modelling choices and the presence of local clustering features. Our results highlight the importance of methodological systematics in homogeneity studies, showing that the determination of $θ_H$ depends not only on the data, but also on the adopted analysis strategy. Flexible approaches such as symbolic regression provide a useful framework to model these effects, but also emphasize the need for careful modelling and survey design. This has important implications for future large-scale structure analyses aiming to test the Cosmological Principle with high precision.

Type Ibn supernovae (SNe) are a class of interacting SNe characterised by narrow helium lines in their spectra. We present an extensive observational dataset of the Type Ibn SN 2025kzr at 51 Mpc, including the discovery of a precursor outburst with a peak brightness of M_r~-13.6 mag beginning ~55 days before explosion. Our photometry indicates the SN was discovered within the first day of explosion, showing fast-rising, ultraviolet-bright emission peaking at M_r=-19.26+/-0.09 mag and a peak blackbody temperature of T~29000 K, consistent with shock breakout within a region of dense and confined circumstellar material (CSM). Our high-cadence spectroscopic dataset spanning 1.9-58.5 days post-explosion shows flash-ionised emission features during the first 10 days. In our SALT spectrum at 3.8 days we observe a pronounced blueshift of the He II lines by 460 km/s compared to the He I lines at zero velocity, while a Pickering-decrement analysis reveals a CSM that is fully hydrogen-free. The timing of the disappearance of the flash features combined with the CSM velocity of 1500 km/s imply a mass-loss event ~66 days before explosion, in close agreement with the timing of the precursor observed 55 days before explosion and strongly suggestive of a physical link. We derive a CSM mass of 0.03-1.7 M_sun and a corresponding high mass-loss rate >~10^{-1} M_sun/yr. The precursor timescale and energetics suggest an extreme mass-loss event that might be explained by wave-driven mass loss during the late stages of nuclear burning, in particular the oxygen-burning phase. Overall, we favour a single massive Wolf-Rayet progenitor with M_ZAMS~30-40 M_sun to explain SN 2025kzr, although a binary origin cannot be excluded.

One of the most powerful approaches to model the structural and spectral evolution of wind bubbles, and pulsar wind nebulae in particular, is based on the one-zone thin-shell formalism. By solving a set of simple equations, one can relate the spectral properties of these systems to the physical properties of the pulsar and supernova remnant. Due to its predictive power, this approach has been widely used both to fit existing objects and in population synthesis. However, there are inconsistencies when applied to the Crab nebula that have never been fully accounted for, casting doubts on its reliability. We introduce a new and more flexible formalism based on energy conservation that reconciles the observed structural and spectral properties of the Crab nebula, solving the inconsistencies, and provides a more realistic description of wind-bubble evolution. The equations of the formalism are presented and discussed, together with simplified solutions and more complex ones including radiation losses. Implications for the modeling of wind bubbles and pulsar wind nebulae are illustrated. We also introduce a new high-order upwind-implicit scheme for particle-spectrum evolution that ensures high accuracy in energy conservation, and an algebraic-vectorizable approach for synchrotron self-Compton emission in the full Klein-Nishina regime that avoids costly interpolations and integrations. We reproduce the Crab nebula structural properties and spectrum, accounting for the difference in convergence age between the optical filaments and the radio bubble, thereby removing the inconsistencies. The spectral accuracy of this approach is comparable to that of the standard one, but it is superior in reproducing structural properties and accounting for geometrical effects such as a thick layer of mixed material and the lack of efficient coupling between the wind bubble and the surrounding medium.

The turnover scale $k_0$ of the matter power spectrum -- and consequently of the standard galaxy power spectrum monopole -- encodes a fundamental signature of matter-radiation equality and constitutes an important standard ruler independent of baryon acoustic oscillations. In principle, we can detect the turnover at multiple redshifts and amplify the signal by stacking redshift bins. However, in spectroscopic surveys reaching high redshifts, such as the Euclid H$α$ survey and the proposed MegaMapper Lyman-break galaxy survey, the monopole of the observed galaxy power spectrum receives a scale-dependent correction from lensing magnification. This can modify the signal shape and shift the turnover scale, undermining its use as a standard ruler. Using mock surveys similar to Euclid and MegaMapper, we forecast this shift and the consequent bias in the turnover scale that is recovered from the mock data. The shift in the turnover scale grows with redshift, leading to a maximum bias of $\sim 0.4σ$ (Euclid-like) and $\sim 3.6σ$ (MegaMapper-like). To avoid a bias $>1σ$, the maximum redshift for a MegaMapper-like survey is $z\approx 2.9$. Data in the remaining range $2.9\lesssim z\le 5$ does not directly provide a reliable recovery of the intrinsic turnover. In fact, we find that the turnover vanishes in a MegaMapper-like survey for $z\gtrsim 3.7$. Our results show that the lensing correction to the monopole should be included and carefully modelled when surveys are used to measure the cosmological turnover at high redshifts.

Type Ic broad-line supernovae (SNe Ic-BL) are often associated with energetic explosions that display a prompt outburst of high-energy emission. Since their progenitor lost the H and He envelopes before the explosion exposing the C/O core, their explosion dynamics and geometry can be seen in an unobscured and undistorted way. We present imaging polarimetry and spectropolarimetry of the Type Ic-BL SN 2026gzf obtained 4.6 and 16.5 days after the X-ray shock breakout, which was recorded by the Einstein Probe satellite as EP260321a, showing it to be one of the softest and intrinsically dimmest extragalactic fast X-ray transients. The persistent low continuum polarization indicates that the outer layer of SN 2026gzf is mostly spherical, suggesting the explosion did not significantly disrupt the progenitor envelope. At day 16.5, the calcium near-infrared triplet displays a peak polarization above 1.5%. The geometry of the associated line opacity is also compatible with an axisymmetric configuration. The spatial distribution of such oxygen-burning ashes thus indicates the presence of a symmetry axis of the excitation structure within the nearly spherical ejecta. The Ca II triplet profile is dominated by a primary component spanning ~25,000--40,000 km/s, alongside a distinct secondary component extending above 28,000 km/s whose polarization implies a non-axisymmetric, complex excitation geometry toward the outer ejecta By implementing a three-dimensional Monte-Carlo calculation, we infer that a viewing angle of ~40 degree from the symmetry axis of the excitation structure could plausibly reproduce the observed spectral and polarization profiles of the Ca II triplet.

We present a sample of 88,959 K-giants from DESI Milky Way Survey Year 3 data, which we use to characterize the chemo-dynamical properties of the stellar halo at Galactocentric distances of 12 to ~100 kpc. Using HDBSCAN, we identify five prominent stellar halo substructures: Aleph, the Sagittarius stream, Gaia Sausage-Enceladus (GSE), Cetus-Palca and the Orphan-Chenab stream. We present the properties of each of these structures as they appear in our catalog, and examine how uncertainties on distance affect the characterization of substructure with this approach. We also examine regions associated with previously reported overdensities (such as the Virgo Overdensities and the Sagittarius spur) that we do not recover with HDBSCAN. The size and distance range of our catalog allows us to explore in detail the residual stellar halo, comprising stars that we do not associate with any substructure. We find that samples of ~2000 outer halo stars with both highly prograde and highly retrograde angular momenta have similar metallicity distribution functions (MDFs), which do not resemble the MDFs of either GSE or Sagittarius. Both the prograde and retrograde residual halo MDFs are bimodal, with a metal-poor peak at [Fe/H] ~ -2 and a metal-rich peak at [Fe/H] ~ -1.3 (prograde) or -1.5 (retrograde). The MDF for lower angular momentum residual halo K-giants does not show clear evidence for a metal-poor peak, and broadly resembles the MDF of GSE, even at much lower binding energies than GSE itself. We discuss possible interpretations of these findings for GSE accretion scenarios.

Magnetar engines are widely invoked to power luminous optical transients, but their early emission depends on the coupled evolution of engine injection, shock heating, adiabatic cooling, and radiative diffusion. We present \texttt{TransFit-MAG}, a time-dependent radiative-diffusion framework for magnetar-powered transients. The model couples the \texttt{TransFit} diffusion solver to the dynamics of a magnetar-inflated pulsar wind nebula (PWN) and its forward shock propagating through homologously expanding ejecta, calculating the internal radiation-energy distribution, photospheric evolution, shock-heating location, and emergent luminosity self-consistently. For different parameter values, the model naturally produces well-separated double peaks, partially merged peaks, or single broad peaks. These results suggest that early bumps and broad single peaks in engine-powered transients may be understood within a unified engine--shock--diffusion framework, in which the observed diversity reflects the coupled evolution of central-engine power, shock propagation, and radiative transport through expanding ejecta. As an illustrative application, we fit the multiband optical light curves of the double-peaked SLSN-I LSQ14bdq.

We report the discovery of PSR J0125$-$5854, a pulsar with a spin period of 24 ms and a dispersion measure of 11.66 pc cm$^-3$ in the ongoing Southern-sky MWA Rapid Two-metre (SMART) survey with the Murchison Widefield Array (MWA). The pulsar is located at a high Galactic latitude of $-57^{\circ}$, and at a distance of 0.5$\text{-}$1 kpc per the Galactic electron density models. Follow-up observations with the MWA and MeerKAT telescopes have revealed that this pulsar is in a binary system with an orbital period of more than 290 days, and a steep spectrum (flux density, $ S \propto ν^α $, where $ν$ is frequency and $ α= -2.2 \pm 0.3 $). Analysis of current observational data hints at a potential binary configuration with an orbital period of $833.60 \pm 0.04$ days, a projected semi-major axis of $241.36 \pm 0.05$ light-seconds, and a minimum companion mass $0.4152 \pm 0.0001$ M$_\odot$, with a low eccentricity orbit of $0.0052 \pm 0.0006$. We discuss the potential formation channels for this system, and conjecture that the companion is likely a Helium white dwarf. Further observations are required in order to better constrain the orbital and spin parameters. We discuss the implications of this discovery, which emerged after processing a small fraction of survey data, on the prospects of finding more millisecond pulsars with the SMART survey, and with future surveys planned with the low-frequency SKA-Low.

The Cherenkov Telescope Array Observatory (CTAO) aims to advance knowledge of the gamma-ray sky as the largest gamma-ray observatory ever built. The CTAO will be deployed at two sites, one in the Northern Hemisphere and the other in the Southern Hemisphere, containing telescopes of three sizes to cover different energy domains. Commissioning of the prototype CTAO Large-Sized Telescope (LST-1) is being finalized at the northern site, while three additional LSTs are under construction. Additional calibration and environmental monitoring instruments, such as laser imaging detection and ranging (LIDAR) systems and weather stations, will support telescope operations. The Array Control and Data Acquisition (ACADA) system serves as the central element for on-site CTAO operations. ACADA controls, supervises, and handles the data generated by the telescopes and the auxiliary instruments. It drives the efficient planning and execution of observations while managing the multi-gigabit-per-second data streams produced by each CTAO telescope. The ACADA system contains the CTAO Science Alert Generation Pipeline - a real-time data processing and analysis pipeline, dedicated to automatically generating science alert candidates as data are acquired. These science alerts, along with external alerts received from other scientific instruments, are managed by the Transients Handler (TH) component. The TH informs ACADA's Short-Term Scheduler (STS) about relevant science alerts, enabling modification of ongoing observations on sub-minute timescales. This capability for rapid response, combined with the fast slewing of CTAO telescopes, makes the Observatory an excellent instrument for studying high-impact astronomical transients.

We report multi-band photometric imaging observations and Monte Carlo dust tail modelling of the interstellar comet 3I/ATLAS covering a wide range of heliocentric distances, from about 4 au pre-perihelion to 4 au post-perihelion. The extensive imaging data set allowed us to constrain the dust physical properties, ejection speeds, and production rates as a function of heliocentric distance. The post-perihelion observations, obtained at high cadence in multiple photometric bands (SDSS g, r, i, and luminance filters) and spanning phase angles between approximately 0.7 deg and 30 deg, enabled us to determine the dust color and phase function. The resulting phase curve exhibits a prominent backscattering enhancement, distinct from those derived for Solar System comets, with an opposition surge of 0.1--0.4 mag, a width of 1--3 deg, and a linear phase coefficient of 0.02-0.04 mag/deg, consistent with independent pre-perihelion estimates. A possible interpretation of the imaging data, together with independent photometric measurements, indicates a dust size distribution characterized by a power-law index of -3.5, with minimum and maximum particle radii of rmin = 10 micrometer and rmax in the interval 1-10 cm. The reported water production rate correlates well with the dust production rate post-perihelion, but fails to do so pre-perihelion, an effect possibly linked to the high CO2/H2O ratio measured before perihelion. The derived maximum dust-loss rate at perihelion is (0.5-1.8)E4 kg/s.

Primordial non-Gaussianity (PNG) is a fundamental probe of the physics of the early Universe and inflation. Here we present a comprehensive study of the constraints on the local-type PNG parameter, $f_{\rm NL}$, for the spectroscopic galaxy survey of the upcoming Chinese Space-station Survey Telescope (CSST). Utilizing the high-resolution Jiutian N-body simulation suite, we construct realistic mock catalogs for emission line galaxies (ELGs) at three representative redshifts $z=0.3$, 0.6, and 0.9. The expected CSST observational characteristics are also considered, including redshift uncertainties and selection functions based on signal-to-noise ratios of emission lines. We develop a robust analysis framework for the redshift-space galaxy power spectrum and bispectrum that accounts for redshift-space distortions, scale-dependent bias, and nonlinear effects. Through a joint Markov Chain Monte Carlo (MCMC) analysis, we find that the power spectrum alone provides competitive constraints, while the inclusion of the bispectrum, specifically targeting the squeezed-limit configurations, improves the $f_{\rm NL}$ constraint precision by approximately 5%-6%. Our joint analysis yields a constraint result of $f_{\rm NL}=-20\pm52$ for the mock data in the 1~($h^{-1}$Gpc)$^3$ comoving volume at the three redshifts, and the constraint accuracy is expected to be improved by several times or even one order of magnitude for the CSST full spectroscopic survey. This work demonstrates the potential of the Stage~IV surveys like CSST to probe inflationary physics, and highlights the importance of higher-order statistics in extracting information from large-scale structure surveys.

Dark matter (DM) subhaloes orbiting inside the Milky Way (MW) are promising targets for DM searches, and reliable predictions for the detectability and spatial distribution of their signals are crucial for probing the nature of DM. Recent work showed that tidal forces from the baryonic components of the MW boost the efficiency of subhalo mass-loss, although the underlying physical processes remain insufficiently understood. This study focuses on clarifying the role of the Galactic disc. By using $N$-body simulations, we examine how the dynamical evolution of subhaloes varies with the inclination angle between their orbits and the Galactic disc. Subhaloes whose orbits are inclined by only a few degrees with respect to the Galactic disc pass through it quickly, which enhances tidal shock heating and leads to a pronounced increase in mass-loss efficiency. In contrast, when a subhalo orbit is exactly coplanar with the Galactic disc, adiabatic shielding suppresses the energy input from tidal shocks, resulting in a lower mass-loss efficiency. Tidal stripping lowers the DM density within subhaloes, thereby attenuating the luminosity of their DM signals. Consequently, we expect that subhaloes located at distances of $\sim 0.3$--$2$\,kpc from the Galactic disc plane emit only weak signals, whereas those remaining embedded in the disc are more promising candidates for indirect DM detection, provided that contamination from baryonic emission sources can be carefully modelled and subtracted. Although their mass-loss histories differ significantly, the structural evolution of subhaloes is still well described by the tidal tracks reported in the literature.

Calibration of the cosmic shear bias is crucial for justifying the cosmological results. However, it is still unclear to what extent calibrations based on simulated galaxy images, as what is commonly done in the weak lensing community, can capture the real shear bias, especially given the complicated instrumental effects. On the other hand, selections of the source galaxies (magnitude cut, redshift binning, etc.) made in real measurement, as well as stochastic but possibly correlated shear biases, may introduce errors to the shear-shear correlations that are hard to calibrate apriori. In our previous few papers, we have shown that the field-distortion signal associated with each galaxy image can be saved in the shear catalog to provide onsite calibrations of shear bias for both galaxy-galaxy lensing and shear-shear correlation. In this paper, we apply this method to the HSCpdr3 shear catalog generated by the Fourier\_Quad shear measurement method. Using our onsite calibration method, we find that the shear biases vary with the selections of photo-z bins, SNR cuts, optical bands, as well as alternative forms of the shear estimators. Nevertheless, after calibrations, the shear-shear correlation functions and cosmological parameter constraints show consistent results in all the cases considered. The fiducial results from the r/i/z bands shear catalog with $\mathrm{SNR}>10$ cut yield: $S_8=0.740^{+0.030}_{-0.030}$ and $Ω_m=0.383^{+0.129}_{-0.132}$

Large-area CO surveys are essential for studying molecular cloud dynamics and evolution; however, most have focused on the Galactic plane, leaving high-latitude clouds less explored. We present the PMO Polaris CO Survey (PPCOS), which maps a 100~deg$^2$ region of the Polaris Flare in the $J=1-0$ transitions of $^{12}$CO, $^{13}$CO, and C$^{18}$O using the Delingha 13.7~m telescope. As the first large-area CO survey at high Galactic latitude ($|b| > 20^{\circ}$) with sub-arcminute resolution, PPCOS achieves sensitivities of $\sim$0.46~K for $^{12}$CO and $\sim$0.23~K for $^{13}$CO and C$^{18}$O at a spectral resolution of 0.16~km~s$^{-1}$ and an angular resolution of 50\arcsec. The $^{12}$CO emission reveals seven distinct complexes, where only $\sim$10\% of pixels display multiple velocity components, alongside a global velocity gradient of 0.18~km~s$^{-1}$~pc$^{-1}$. Typical line widths are $1.2 \pm 0.6$~\mbox{km~s$^{-1}$} for $^{12}$CO, while $^{13}$CO components are systematically narrower ($\lesssim 0.7\,ΔV_{\rm ^{12}CO}$). The $^{12}$CO/$^{13}$CO intensity ratios (5--25) indicate widespread $^{12}$CO optical thickness, resembling conditions found in giant molecular clouds (GMCs). Globally, the CO emission divides into two groups: a major group aligned with the velocity gradient and a secondary group elongated perpendicular to it, possibly regulated by large-scale coherent dynamics. We propose a three-layer hierarchy: a dynamically assembling and dispersing periphery traced by $^{12}$CO, a more stable intermediate kernel traced by $^{13}$CO, and gravitationally bound compact cores traced by C$^{18}$O. No young stellar objects are firmly associated with the molecular gas. PPCOS provides an ideal laboratory for studying turbulence, hierarchical structure, and early cloud evolution in a nearby, relatively simple molecular cloud.

The origin of the gap in the observed magnetic field distribution between pulsars and magnetars raises a challenge to understanding these populations within a unified framework. We analytically show that the gap can be naturally explained by the alignment of the magnetic inclination angle between the magnetic and spin axes. Based on coupled evolution of spin-down and magnetic inclination angle in the plasma-filled magnetosphere, the alignment timescale follows $τ_α\propto B^{-2}$. Thus, strongly magnetized neutron stars including high-$B$ pulsars and magnetars align more rapidly than pulsars with $10^{12}\,\mathrm{G}$, reducing their beaming fraction and thereby suppressing their observed numbers. However, magnetars are primarily identified through X-ray activity and are therefore relatively less affected by beaming. Taking into account both beaming fraction and luminosity corrections, we reconstruct the initial magnetic field distribution from the observed distribution. We show that pulsars and magnetars do not dictate intrinsically distinct initial distributions, but can instead be understood within a single continuous initial magnetic field distribution, such as a log-uniform distribution.

Type~IIn supernovae (SNe) are a subclass of core-collapse SNe powered by interaction between the ejecta and the dense circumstellar material. Among them, long-lived Type~IIn events are characterized by luminous, long-duration light curves with high radiative energy. Several cases of long-lived type IIn SNe exhibit substantial dust emission at late times. However, well-observed examples remain limited, and the details of their dust formation mechanisms remain poorly understood. Here we present photometric and spectroscopic observations of the Type~IIn SN~2024dy in ultraviolet, optical, and near-infrared (NIR) wavelength for $500$ days. SN~2024dy reached a peak magnitude of $M_r=-19.2$~mag with a total radiation energy of $1.9\times10^{50}$~erg. A NIR excess emerged at late phases, and the spectral energy distribution modeling indicates the presence of carbon dust with temperatures of $1300$-$1800$~K and masses of about $10^{-5}\ M_\odot$. The spectra features were typical of long-lived Type~IIn SNe. The late time H$α$ profile exhibits a strong suppression of the red wing, providing evidence for newly formed dust. Our results suggest that the derived dust mass above may be underestimated due to optical depth effects. SN~2024dy provides an important observational case for understanding dust formation in Type~IIn SNe.

Using emission-line galaxies (ELGs) from the Dark Energy Spectroscopic Instrument (DESI) Data Release 2, we evaluate a property-dependent correction to imaging systematics. We derive systematic weights following the same linear regression method used for other DESI tracers, but do so separately on ELG subsamples to provide a physically-informed alternative to the fiducial, neural-network-based approach. In doing so, we show that the deeper imaging in the Dark Energy Survey (DES) footprint leads to a higher overall number density but a lack of targets with extreme $g-r$ and $r-z$ colors. ELGs in the DES region also show a distinct redshift distribution when subsampled by position in the $g-r$ vs. $r-z$ plane. To address these effects, we implement a separate treatment of the DES footprint within the DESI catalog production pipeline, which is generally well-motivated and, in some cases, imperative for accurate clustering measurements. With DES treated separately, we find that property-dependent systematic weights further mitigate spurious clustering signal in $\sim$10% of subsamples, while the fiducial scheme remains optimal for the full sample.

Short-duration gamma-ray burst (sGRB) afterglows that follow BNS-gravitational wave (GW) events are essential for understanding the tension between the observed sGRB rate and BNS merger rate, heightened by the recent conclusion of aLIGO O4 with no new confirmed BNS detections. Using a probabilistic sGRB world model derived from a source BNS merger population, we simulate afterglow emission with AfterglowPy to investigate detection prospects of afterglows in the new era of optical surveys, and probe their multi-messenger implications. The predicted sGRB/BNS association is strongly dependent on sGRB beaming, which may be constrained by orphan afterglows (OA) - that arise from events with no prompt $γ$-ray detection. We conclude that the Vera C. Rubin Observatory's Large Synoptic Survey Telescope (LSST) may detect an afterglow sample sufficient in constraining sGRB jetting, with an estimated $0.9^{+0.5}_{-0.3}$ on-axis afterglow and $1.3^{+0.9}_{-0.5}$ OA detections per year; while the deep sensitivity of the Roman Space Telescope appears promising for probing the faint end of afterglow events in targeted follow-up strategies. The detection of afterglows in upcoming LIGO runs is possible but challenging, as we predict less than one LSST or Roman discoverable event per year within the projected aLIGO O5 BNS range across all considered jet models and observing scenarios. We update previous sGRB-BNS rate predictions, finding that continued non-detection of a BNS in O5 would require revisiting key assumptions underlying sGRB-BNS models.

We present CMBComp, a compact and accurate compressed cosmic microwave background (CMB) likelihood that captures the dominant geometric information of the full CMB likelihood derived from the combined SPT-3G D1 + ACT DR6 + Planck PR3 primary CMB anisotropy + Planck PR4/NPIPE CMB-lensing dataset, which we collectively refer to as SPA. The compression is fast to evaluate and trivial to implement in standard inference pipelines. We construct and validate it in five model spaces: the spatially flat cosmological-constant model ($Λ$CDM), and its dynamical-dark-energy ($w_0w_a$CDM), massive-neutrino ($νΛ$CDM), non-flat ($oΛ$CDM), and joint massive-neutrino--dynamical-dark-energy ($νw_0 w_a$CDM) extensions. Five compressed likelihoods are introduced, corresponding to two three-parameter compressions for the dark-energy sector (CMB-3 for $Λ$CDM and CMB-3w for $w_0w_a$CDM), two four-parameter compressions for the neutrino sector (CMB-4$ν$ for $νΛ$CDM and CMB-4$ν$w for $νw_0 w_a$CDM), and a four-parameter curvature compression (CMB-4k for $oΛ$CDM). Combining each compressed likelihood with the DESI DR2 baryon acoustic oscillation (BAO) data, we demonstrate that the resulting posteriors agree to high precision with those obtained from the corresponding full-CMB chains. CMBComp is therefore particularly well suited to cosmological inference for models that modify the late-time expansion history, enabling accurate CMB constraints to be incorporated into new analyses with minimal computational overhead and without reliance on a full Boltzmann-solver-based inference pipeline. The compressed likelihood files and example notebooks accompanying CMBComp are made publicly available at https://github.com/Amoghsriv/CMBComp.

Covariance matrices used in astronomical and cosmological parameter inference are often estimated from a finite number of simulations, so covariance uncertainty can affect posterior calibration and parameter constraints. We study covariance regularisation from the perspective of likelihood-based inference with simulation-estimated covariance matrices. First, we analyse scalar covariance scaling under the Gaussian plug-in likelihood and the covariance-marginalised Sellentin--Heavens likelihood. Using an expected negative log-likelihood loss, we show that Hartlap covariance-side scaling is recovered as the optimum under the Gaussian plug-in likelihood, whereas the unscaled sample covariance is optimal under the Sellentin--Heavens likelihood. This shows that scalar covariance corrections are likelihood-dependent and that an additional Hartlap-type global scaling is not favoured once covariance uncertainty is marginalised through the Sellentin--Heavens likelihood. We then introduce a shrinkage formulation in which the sample covariance is regularised towards a spherical target and the shrinkage intensity is treated as an auxiliary inferential quantity. A prior is assigned to the shrinkage intensity, the likelihood induces its posterior distribution, and the final parameter posterior is obtained by marginalising over it. Monte Carlo experiments show that shrinkage substantially improves covariance conditioning, while marginalisation over the shrinkage intensity propagates uncertainty about the amount of regularisation into posterior inference. The proposed approach provides a simple way to combine covariance-marginalised likelihood inference with structural regularisation of noisy simulation-estimated covariance matrices.

Our interpretation of current cosmological observations rests on the assumptions of homogeneity and isotropy, leading to uniform background curvature and expansion characterised by the Friedmann-Lemaître-Robertson-Walker (FLRW) spacetime metric. However, the large-scale structure of the Universe is non-uniform in detail, inducing inhomogeneous curvature and scale factor variations. In this paper, we use numerical cosmological simulations generated in full General Relativity to study the impact of inhomogeneous spacetime on the magnitude fluctuations of distant objects, focusing on their use as a probe of the growth rate of cosmic structure. We quantify the distortions in the magnitude correlation spectrum as a function of angular scale and redshift, and use these distortions to infer the systematic offset in the growth rate measurement. We find that at $z \lesssim 0.2$, the systematic offset in growth rate measurements between the full numerical relativity and FLRW treatments is sub-dominant to the statistical error of current datasets, confirming that FLRW modelling is adequate for current low-redshift peculiar velocity experiments. Future datasets extending to higher redshift may require theoretical models that additionally incorporate the contributions of gravitational lensing and inhomogeneous curvature.

We present the first model of the Milky Way's detectable compact object--luminous star binary population from disrupted dense star clusters. We bridge large-scale cosmological star cluster formation with high-resolution dynamical evolution of compact object binaries by mapping the predicted star clusters from the EMP-Pathfinder simulations to $N$-body Cluster Monte Carlo models. We predict that approximately $3\times10^5$ white dwarfs (WDs), $1.5\times10^5$ black holes (BHs), and $1\times10^3$ neutron stars (NSs) in binaries with luminous companions are released to the Galaxy from now-disrupted dense star clusters throughout the history of the Milky Way. Synthetic observations modeled with the gaiamock pipeline reveal that the modeled Gaia DR3 yields are sparse ($\approx 2$ WDs, 0 NS, 0 BHs at 90% credibility), with the majority lying beyond the detection horizon. Gaia DR4 is expected to increase the observational yield of these systems only marginally, as the benefits of an expanded search volume are largely offset by the diminished astrometric and photometric precision of more distant sources ($\approx 14$ WDs, 0 NS, 0 BHs). While the underlying BH binary population is similar to that of WDs, they are detected far less frequently; they tend to pair with lower-mass, dimmer companions and have less temporal coverage of their long orbital periods. For NSs, we suggest that the observed over-representation of metal-poor, halo systems is inconsistent with an origin in disrupted dense star clusters. Instead, the observed Gaia NS population could reflect the accretion history of metal-poor, dwarf galaxies into the Milky Way, isolated binary star evolution, or supernova physics.

The gravitational potential of a galaxy encodes its mass distribution, formation history, and dark matter halo structure. Accurate potential models are therefore critical for interpreting stellar kinematics, orbital dynamics, and the influence of satellite systems like the Large Magellanic Cloud. Analytic potential models offer interpretability and efficiency but struggle to capture complex, non-axisymmetric structure and time-dependent perturbations. Neural network-based methods can capture this complexity but offer little interpretability. We introduce a physics-informed neural network (PINN) framework that combines data-driven learning with embedded physical constraints, available as the open-source package GalactoPINNS. Trained on acceleration measurements, the framework captures complex, small-scale features while preserving global physical consistency. We test on systems of increasing complexity, from controlled analytic halos to cosmological simulations of Milky Way-like galaxies, achieving sub-percent acceleration errors with orbit reconstruction that consistently outperforms analytic baselines. Additionally, we implement a Bayesian neural network to provide spatially calibrated uncertainty estimates, and a time-dependent extension to capture smooth temporal evolution. By treating an analytic model as a structured prior and learning corrections on top of it, the method retains physical interpretability while gaining the flexibility to represent realistic galactic potentials, making it well suited for Milky Way modeling and dynamical inference in the era of current and upcoming large-scale surveys.

Recombinations during the Epoch of Reionization are intrinsically inhomogeneous, with different regions of the intergalactic medium contributing unevenly depending on their density, temperature, ionization state, and spatial patchiness. We combine the high- and medium-resolution 95.5 cMpc Thesan-1 andh Thesan-2 runs with the significantly larger 500 cMpc Lumina simulation to measure clumping factors and recombination rates consistently across different resolutions and box sizes. We consider the standard ionized hydrogen clumping factor, $C_{\rm HII} \equiv \langle n_{\rm HII}^2\rangle/\langle n_{\rm HII}\rangle^2$, and a recombination-weighted clumping factor, $C_{\rm rec}$. Despite differences in resolution, volume, and reionization history, the simulations show an approximately universal clumping evolution at the 10-20% level when parametrized by the global ionized fraction $x_{\rm HII}$ rather than by redshift. Across all simulations, $C_{\rm rec}$ remains systematically below $C_{\rm HII}$, with the discrepancy increasing toward lower redshift as photoheating suppresses recombinations. In \lumina, the density-only prescription overpredicts the instantaneous recombination rate by factors of 1.29 at $z\approx8$ and 1.84 at $z\approx5$, and the cumulative recombination count by a factor of 1.45 by $z\approx5$. Mapping the recombination budget in the joint overdensity-temperature plane reveals that the dominant recombination ridges closely follow simple analytic thermal equilibrium bands. Finally, we introduce a phase-space recombination integral and define a phase-space clumping factor, $C_{\rm ps}(Δ,T)$, which isolates the intrinsic recombination enhancement associated with ionization structure and thermal state at fixed overdensity and temperature.

We test the consistency of the cosmic distance-redshift relation inferred from the cosmic microwave background (CMB), the Dark Energy Spectroscopic Instrument (DESI) baryon acoustic oscillation measurements, and the Dark Energy Survey (DES) Type Ia supernovae within the framework of flat $Λ$CDM. DESI recovers the CMB-constrained parameter combination $(r_d h)(Ω_m/0.3)^{0.4}$ with sub-percent precision, demonstrating excellent agreement between BAO measurements at $z \sim 1$ and the acoustic scale at recombination. Imposing the CMB constraint yields an estimate of $Ω_m$ that is slightly lower than, but only in mild tension with, the Planck value. The high-redshift DES supernova sample is well described by the standard cosmological model, whereas the current low-redshift anchor sample exhibits a systematic offset of $\sim 0.05$ mag that drives much of the apparent preference for evolving dark energy. Preliminary data from the Dark Energy Bedrock All-Sky Supernova Program (DEBASS) do not show this offset, suggesting that unresolved low-redshift systematics may account for the discrepancy. These results suggest that a single flat $Λ$CDM model accurately describes the distance-redshift relation from the local Universe to recombination, placing increasingly stringent constraints on new-physics explanations of the Hubble tension.

The three-dimensional shapes of clusters are important for understanding the astrophysics of the clusters and as a probe in cosmological studies. We estimate the most probable three-dimensional shape of galaxy clusters in the first eROSITA All-sky survey eRASS1 using stereology. Our sample is the largest well-defined sample of clusters, and the most probable shape estimated using our method can be used as a prior for cluster shape models in cosmological, cluster, and weak lensing studies. The first all-sky survey with SRG/ eROSITA resulted in a sample of approximately 12,000 optically confirmed galaxy groups and clusters. We used a well-defined subsample of 3254 clusters from the eRASS1 survey and estimated the most probable shape of the clusters by constraining the probability density function (PDF) of the ellipticity of the clusters. We simulated the projected appearance of clusters with a distribution of three-dimensional shapes (prolate and oblate) and obtained the distribution of their ellipticity. This distribution was then compared with the measured distribution of ellipticities from the eRASS1 cluster sample to infer the three-dimensional shapes consistent with the data. We used Monte Carlo methods to estimate the most probable axial ratios l, w, where l $\equiv$ L/T ,w $\equiv$ W/T , and L, W, T are major, intermediate, and minor axes of the cluster. We did not require any additional probe (optical, SZ, etc.) to constrain the probable shape of the clusters. We describe the ellipticity PDF of the eRASS1 clusters with a normal distribution mean ($μ$) = 0.79 and a standard deviation ($σ$) = 0.25. The most probable shape of the clusters in our eRASS1 subsample is estimated to be (l, w) = (1.51 $\pm$ 0.27, 1.17 $\pm$ 0.27), with prolate shapes being preferred over oblate shapes.

We present JWST/NIRCam observations of the extremely long-duration gamma-ray burst (GRB) 250702B taken at ~ 95 days post-GRB (observer frame). The observations of the host galaxy reveal a single galaxy with a prominent dust lane observed nearly edge-on. Prospector modeling of the host galaxy photometry finds a high stellar mass (log(M_*/M_Sun) = 11.0 +0.2/-0.3) and large dust column (A_V = 2.8 +/- 0.3 mag), in agreement with previous results. If GRB 250702B is a collapsar-driven GRB, the host galaxy is the brightest (in rest-frame r and rest-frame H) and most massive compared to GRB hosts at similar redshifts. The transient localization is near the dust lane, and while we find no evidence for transient emission in F277W, F356W, and F444W, forced photometry in F150W and F200W reveals possible ~ 3 sigma detections of the transient at m_{F150W} ~ 27.9 AB mag and m_{F200W} ~ 27.4 AB mag. If these are secure detections, they are indicative of a late-time light curve flattening. This behavior is consistent with that of jetted tidal disruption events (TDEs); however, it is also consistent with a supernova plus GRB afterglow model. Alternatively, if these are upper limits, they are consistent with, but do not further constrain, the extrapolated power-law decline of the afterglow. The ambiguity of the possible detection of the transient in F150W and F200W highlights the need for late-time template observations with JWST/NIRCam.

All cosmological models of galaxy formation must navigate the trade-off between physical accuracy and computational efficiency. Hydrodynamical simulations provide spatially resolved predictions for the co-evolution of dark matter, gas, stars, and black holes, but rely on phenomenological subgrid models for small-scale processes (e.g., star formation). Semi-analytic models (SAMs), by contrast, gain efficiency through simplified, analytic treatments of the same processes, at the cost of reduced predictive scope. In this work, we leverage the strengths of the Santa Cruz SAM and the IllustrisTNG hydrodynamical simulation to develop the TNG SAM. Calibrated to reproduce baryon cycling in stellar feedback-dominated TNG galaxies ($\sim 10^{10}M_\odot < M_{200} < 10^{12}M_\odot$), the TNG SAM introduces several key updates to the Santa Cruz framework regarding: 1) halo gas (re-)accretion efficiency, 2) a cooling model that moves beyond the traditional cold/hot mode dichotomy, 3) explicit treatment of both galactic- and halo-scale outflows, 4) star formation efficiency, and 5) the circulation of metals between galaxies and their surroundings. These changes enable the TNG SAM to reproduce TNG's flow of gas and metals from the scale of the galaxy to the halo, as well as global galaxy (e.g., stellar mass) and halo (e.g. hot halo gas mass) properties within $\lesssim 30\%$ accuracy out to $z=6$. This work demonstrates that, with appropriate calibration, SAMs can capture the complex physics of galaxy formation modeled in hydrodynamical simulations while providing a flexible framework for studying galaxy evolution across the large cosmological volumes targeted by future observational surveys.

Modern simulations of stellar dynamics in globular clusters peg a dominant role for stellar-mass black holes, but direct evidence for black holes in clusters remains limited. We present the discovery of an astrometric stellar-mass black hole--main sequence star binary in $ω$ Centauri, the most massive Galactic globular cluster, using Hubble Space Telescope data from the oMEGACat project and additional JWST data that span a total of 23 years. The luminous companion to the black hole is a main-sequence turnoff star, and has a period of $94^{+63}_{-42}$ years, a semi-major axis of $31^{+15}_{-12}$ AU, and an eccentricity of $e=0.72^{+0.08}_{-0.13}$. Since we observe the binary during periastron, the mass of the black hole is well-constrained even though we only observe a partial orbit: the inferred black hole mass is $4.46^{+1.22}_{-1.01}$ M$_\odot$. We call this black hole oMEGACat BH-2. This is the first astrometric discovery of a stellar-mass black hole in a globular cluster, and is the longest period black hole binary system yet discovered. The low mass of this black hole is perhaps surprising given the low metallicity of the cluster, and shows that at least some low-mass black holes form at metallicity $Z<10^{-3}$. We find that the binary is almost certainly dynamically formed and is soft, with an expected binary disruption timescale of $\sim$800 Myr. While the total number of black hole binaries in $ω$ Centauri is uncertain, we show that existing surveys only cover a small area of parameter space, and that the presence of additional detectable black hole binaries is likely.

Little Red Dots (LRDs) appear extremely compact at rest-frame optical wavelengths, yet many show extended rest-frame UV morphology revealing more complex internal structure. We present a combined analysis of VLT/MUSE rest-frame UV integral-field spectroscopy and continuum-subtracted [O III], H$β$, and H$α$+[N II] emission-line maps from JWST/NIRCam imaging at sub-kpc resolution for LRD-204851 at $z=5.482$ in GOODS-S. We find that LRD-204851 hosts a remarkably thin, bipolar, elongated structure passing through the optical continuum centroid and extending several kpc on either side, traced by both the UV continuum and the rest-frame optical emission lines, with a bright [O III] clump-like structure $\sim$2 kpc to the south-east of the centroid. The MUSE observations reveal a double-peaked Ly$α$ profile, with a broad and bright near-systemic red peak and a relatively faint peak blueshifted by $\sim$430 km s$^{-1}$, accompanied by a tentative N V $λ1238$ detection at similar velocity. In narrow-band imaging extracted from the MUSE IFU cube, both the blue Ly$α$ peak and the tentative N V emission lean toward this same south-eastern direction. Independently, radiative-transfer modeling of the integrated Ly$α$ profile favors a biconical low-column-density cavity in a dense, slowly expanding neutral envelope, in support of the bipolar geometry traced by the line maps. Together, these results suggest that the elongated emission of LRD-204851 is connected to radiation and/or gas flow from its central engine through a low-column-density channel with a small opening angle that may trace either a slow outflow or a quasi-static ionization cone. LRD-204851 is one of the first LRDs where the central engine's impact on its host galaxy is potentially directly observable on kpc scales.

Via is a forthcoming all-sky spectroscopic survey that will achieve 100 m s$^{-1}$ radial velocity stability for millions of faint ($G \lesssim 21$) stars while reaching LSST's single-visit depth ($r \approx 24$) for transient spectroscopy, opening new regimes in near-field cosmology and time-domain astrophysics. Via will deploy identical fiber-fed, multi-object spectrographs on the 6.5m MMT and Magellan/Clay telescopes for a five-year, dual-hemisphere survey of $>2{,}000{,}000$ stars beginning in 2027 - timed to complement LSST. Each instrument has 576 robotically positioned fibers over a $1^\circ$ field of view, feeding two spectrographs: Viaspec ($R \approx 15{,}000$; 505-595 nm; 540 fibers) and Boombox ($R \approx 1{,}000$; 360-1010 nm; 36 fibers). Four key goals drive the survey: (1) a comprehensive survey of velocity perturbations in cold stellar streams, sensitive to $M \lesssim 10^7$ subhalos below the threshold of galaxy formation, a stringent test of the particle nature of dark matter; (2) a chemodynamical census of Milky Way satellite galaxies to understand the formation of the faintest galaxies; (3) the first 3D tomographic maps of cold gas in the circumgalactic medium via NaI absorption; and (4) the rapid characterization of thousands of transients to the single-epoch survey depth of LSST. Ancillary science - including the Ly$α$ forest at $z \approx 3$-$4$, polluted white dwarfs, exoplanet host characterization, fast radio burst host galaxies, and extragalactic dwarf galaxies - will leverage spare fibers in every pointing. The Via Project is a collaboration between the Center for Astrophysics $|$ Harvard & Smithsonian, Carnegie Observatories, Stanford University, and Yale University.

Early JWST spectra revealed galaxies with a strong nitrogen excess challenging galactic chemical evolution models. Using public JWST surveys, we construct a sample of 76 N/O-enhanced galaxies (NOEGs) at 4 $<z<$ 8.5, the largest at high redshift to date. The NOEG fraction rises from $\sim$3% to $\sim$18% between $z\sim$ 4 and 7 - well above the $\sim$2% measured locally - potentially driven by burstier, cluster-dominated star formation. Stacked spectra of the most nitrogen-rich galaxies show signatures of low-metallicity Wolf-Rayet (WR) stars, a likely source of primary nitrogen within the first few Myr of a starburst, with UV and optical continua dominated by young stellar emission and Balmer jumps evident in some cases. Many NOEGs also exhibit ionised outflows: 40% show secondary [O III] and H$α$ components, while stacked spectra of the remainder reveal a broadened, offset H$α$ without forbidden-line counterparts, suggesting dust-attenuated or faded outflows. The continuum in the latter shows a weak Balmer break, indicating these galaxies are past their most recent burst. This suggests that outflows dilute gas metallicity after the first few Myr of the initial enrichment and enable renewed N/O enhancement driven by low-metallicity Asymptotic Giant Branch (AGB) stars. We conclude that NOEGs are caught briefly after a recent starburst: either within $\sim$10 Myr, when WR winds drive nitrogen enrichment, or after 30-40 Myr, when AGB winds take over - following an outflow driven by radiative or supernova feedback, consistent with recent chemical evolution models.

The rapid neutron-capture process (r-process) is widely regarded as the dominant mechanism responsible for the synthesis of heavy elements in the universe, yet its astrophysical sites remain an open question. Recent studies suggest that the high-entropy, rapidly expanding baryonic material ejected by magnetar giant flares may provide favorable conditions for r-process nucleosynthesis, while the late-time gamma-ray emission observed from the magnetar SGR 1806-20 offers direct observational support for this scenario. In this work, we perform nuclear reaction network simulations to investigate the nucleosynthesis yields of magnetar giant flares and to characterize the associated nuclear gamma-ray line emission arising from the radioactive decay of heavy nuclei. The nuclei synthesized in magnetar giant flares are found to be mainly distributed near the first and second r-process abundance peaks. Owing to this nuclide composition, the gamma-ray opacity is found to be strongly energy-dependent with the opacity in the keV band exceeding that in the MeV band by approximately three orders of magnitude. The nuclear gamma-ray emission is dominated by MeV photons at early times and gradually extends toward the sub-MeV and keV bands as time progresses, thereby offering a diagnostic of heavy element enrichment in the ejecta. The gamma-ray spectrum exhibits a peak near 1 MeV with major contributions from $^{88}$Kr and $^{92}$Sr, whose radioactive decays produce several bright gamma-ray lines with fluxes exceeding $\sim10^{-8}$ erg cm$^{-2}$ s$^{-1}$, making them the most promising lines for detection by MeV gamma-ray detectors. Because magnetar giant flares occur in the Galaxy at a rate roughly three orders of magnitude higher than neutron star mergers and their gamma-ray lines are accessible to current MeV instruments, they offer new and valuable science opportunities for MeV gamma-ray astronomy.