arXiv Daily Digest - 2026-06-30

Adaptive Financial Transformer (AFT) is proposed for stock return prediction under non-stationary financial markets. The model incorporates a Market Regime Encoder, an Adaptive Gate Network, and an Adaptive Financial Context module to dynamically bias self-attention based on semantic relationships between financial indicators. Unlike conventional Transformer architectures that treat all input features uniformly, the proposed approach groups 95 engineered financial features into 11 semantic categories and adapts attention according to latent market regimes. The study also identifies and corrects sequence alignment and backtesting issues that can inflate reported trading performance, and introduces a financially-aware composite objective that jointly optimizes prediction error, directional accuracy, and non-overlapping Sharpe ratio. Extensive experiments compare the proposed architecture against classical machine learning models, recurrent neural networks, and Transformer baselines using chronological evaluation, five random seeds, ablation studies, hyperparameter optimization, explainability analysis, and multi-stock validation. Results demonstrate competitive predictive performance while reducing model complexity by 15.2% and improving parameter efficiency through feature selection, providing an interpretable Transformer architecture for financial time-series forecasting.

Post-hoc explanation methods are routinely used to interpret scientific machine learning models, with the deliverable understood to be insight into the phenomenon the model has been trained on. The transition may be taken to be secured once the model is reliable enough and the explanation faithful enough. We argue it is not. Reliability checks that the model's predictions match the phenomenon's outcomes, and faithfulness checks that the explanation matches the model, but neither checks whether the model works as the phenomenon works, which is what a claim about structure requires. The chain can support candidate hypotheses under external corroboration, but it cannot, on its own, support claims about how the phenomenon is in fact structured.

On-policy self-distillation (OPSD) trains a reasoning model on rollouts sampled from its own policy by matching a privileged teacher that also sees verified reference solutions. Existing OPSD objectives supervise only the output distribution, so privileged context affects training through a token-level divergence without directly supervising the internal computation that produced that distribution. We propose Privileged Hidden Flow (PHF), which additionally distills how a privileged teacher's hidden states move along the same rollout. Rather than forcing each student hidden vector to match the teacher vector at the same token position, PHF aligns token-to-token transition directions and trajectory geometry over selected generated positions. The all-layer recipe also includes an adjacent-layer relation computed from these same transitions, without pointwise hidden-state imitation. Under the same 100-step training schedule, PHF improves the Average@12 aggregate over our reproduced OPSD baseline on Qwen3-1.7B, 4B, and 8B, with observed gains of about +2.2, +1.5, and +1.7 points. The transport objective is exactly invariant to shared trajectory offsets; its local geometry term is also invariant to orthogonal transformations of transition directions. Ablations distinguish the fixed PHF recipe from pointwise hidden-state matching, single-channel transition losses, and layer-subset choices, supporting PHF as a compact hidden-flow extension to OPSD.

Reliable event detection underpins induced-seismicity monitoring for Carbon dioxide Capture and Storage (CCS) and geothermal operations, distributed acoustic sensing (DAS), and industrial condition monitoring. In each setting a detector must stay reliable both when sensors fail and when the signal is buried in noise. These two failure modes are routinely conflated, and architectural complexity is often credited with robustness it may not deserve. We assemble a unified binary event-detection benchmark from three physically distinct real sources -- Hi-net seismic waveforms, Utah FORGE 2024 borehole DAS, and MAFAULDA industrial vibration -- each mapped to a common 8-channel, 256-sample representation, and evaluate a fault-tolerant detector (CEPHALON) trained with per-sample sensor-dropout against standard detectors (a 1D convolutional network, a temporal convolutional network, and a compact Transformer) trained with an identical recipe. On clean data every model is near-perfect (AUC ~ 0.99). Under progressive sensor loss, simple models with sensor-dropout are already robust and CEPHALON holds no advantage. Under additive noise, however, CEPHALON degrades far more gracefully: at -2.5 dB its overall AUC is 0.939 versus 0.532-0.572 for the convolutional baselines. Same-architecture ablations isolate the cause: disabling internal redundancy at inference reduces the low-SNR advantage only modestly, whereas removing sensor-dropout training collapses it (0.899 to 0.603 at -5 dB). The training recipe is therefore the dominant cause and parallel redundancy only secondary. We release a complete, numbered, reproducible pipeline so that every figure can be regenerated.

We summarize our submission to Sub-Challenge 1: W4A4 Quantization for Inference (HiF4 / MXFP4) of the ICME 2026 Low-Bit-width Large-Model Quantization Challenge. The sub-challenge targets 4-bit weight and 4-bit activation inference on Wan-AI/Wan2.2-I2V-A14B under HiF4 or MXFP4 numerical formats. We adapt two complementary ideas from LLM quantization, MixQ-style mixed precision for sparse activation outliers and SmoothQuant-style per-channel smoothing, together with block-wise HiF4 packing for Wan2.2 feed-forward linear layers. Calibration on representative OpenS2V-5M batches identifies heavy-tailed activation channels; smoothing rebalances dynamic range before W4A4 rounding; and a dual-branch GEMM preserves outlier columns in higher precision while the bulk of channels use strict W4A4. On official VBench I2V metrics, our pipeline stays within 2-3.5 percent of FP16 on most quality axes and improves motion smoothness, outperforming a native HiFloat4 baseline that degrades roughly 5 percent relative to FP16 across all reported scores.

Multimodal speaker identification systems face two key challenges in real-world deployment: missing modalities and language mismatch between training and testing conditions. In practical scenarios, background multi-speaker conversations, ambient noise, and overlapping speech further degrade identification accuracy. To address these challenges, we propose a multimodal polyglot speaker identification system for the POLY-SIM 2026 Grand Challenge. The system is fundamentally built upon Adaptive Modality Routing(AMR), a modality fusion module that dynamically assesses per-sample input quality and integrates modality information. Specifically, AMR employs two modality adapters to process the embeddings extracted from a linguistically robust audio encoder(W2V-BERT 2.0) and a large-scale pretrained face encoder(IResNet-18), producing modality-adapted embeddings. Based on these adapted embeddings, a trainable router estimates dynamic modality weights, which are subsequently applied to aggregate the modality-specific logits for the final prediction. To optimize this routing mechanism, we adopt a modality-aware training strategy that constructs four types of sample pairs to simulate diverse input conditions, with KL divergence serving as explicit supervision for weight assignment. Experimental results on the POLY-SIM 2026 evaluation set show that the proposed system achieves identification accuracy of 99.93%(English multimodal, P3), 100.00%(Urdu multimodal, P5), 97.50%(English audio-only, P4), and 98.83%(Urdu audio-only, P6). The average accuracy across all four protocols is 99.07%, surpassing the Fusion and Orthogonal Projection(FOP) baseline by 32.73%.

Scientific discovery via symbolic regression is often viewed as statistically and computationally intractable because the hypothesis space of expressions grows combinatorially with depth. This paper revisits the statistical side through the lens of PAC learning, focusing on compositional function trees built from a finite vocabulary of smooth operators (e.g., $\{+,\times,\sin,\exp\}$ and affine maps). We prove that the relevant generalization quantity, Rademacher complexity, hence the excess risk, does not necessarily blow up exponentially with the number of distinct symbolic structures, but is controlled by (i) the depth $d$ and (ii) the Lipschitz constants of the base operators along the composed computation graph. Concretely, under mild Lipschitz conditions on operators and bounded affine leaves, a finite-union bound over a vocabulary of size $K=|\mathcal{H}_{\mathrm{base}}|$ together with Maurer-type vector contraction yields $\mathfrak{R}_n(\mathcal{H}_{\mathrm{comp}}^{d}) \leq (Kb\sqrt{2}L)^{d-1}\mathfrak{R}_n(\mathcal{H}_{\mathrm{comp}}^{1})$ with arity bound $b$; corresponding high-probability risk bounds scale as $\mathcal{O}(L^{d}/\sqrt{n})$ when $K,b=O(1)$ and $\mathfrak{R}_n(\mathcal{H}_{\mathrm{comp}}^{1})=O(n^{-1/2})$. We complement the theory with a modular codebase that trains differentiable operator trees (not MLPs) on synthetic "physics-like" targets of controlled depth and shows that the empirical generalization gap correlates positively with the predicted complexity term $(\widehat{L}^{d})/\sqrt{n}$.

Retrieval-augmented generation (RAG) typically treats context selection as ranking chunks against a single query embedding. This assumption breaks down for complex queries, such as multi-hop or ambiguous questions, where top-k selection tends to over-cover one semantic aspect while ignoring critical sub-questions. We propose GeoRAG, which recasts context selection as Information Demand Coverage Optimization. GeoRAG builds a multi-dimensional demand distribution through diverse sub-query generation and reverse-validation weighting, then selects context by minimizing the Sinkhorn-Wasserstein distance between this demand distribution and the coverage of the selected set. The resulting demand-weighted facility-location objective is monotone submodular, giving a $1-1/e$ greedy guarantee, which we approximate with a Sinkhorn-based marginal-gain surrogate. The method is unsupervised, training-free, and retrieval-agnostic. We further show that single-point, query-proximity scorers cannot cover multi-modal demands, exposing a structural limit of ranking-based selection. On six open-domain QA benchmarks, GeoRAG improves exact match (EM) by +6.5 to +7.5 points over top-k truncation (up to +9.7 on HotpotQA and ASQA) and outperforms strong baselines including MMR, DPP, BGE-Reranker, SMART-RAG, and AdaGReS, with stable gains across context budgets and sub-query generators.

Gradient boosting in the form of decision tree ensembles has successfully been applied to a variety of problems using simple objective functions based on log-likelihoods of a single variable. The concept extends naturally to objective functions operating on vectors - for example, multinomial logistic log-likelihood for multi-class classification, where observations have a score for each class - but popular frameworks approach these functions by either updating one value of the input vectors at a time, or by using a diagonal upper bound on the second derivative. This work extends the usual gradient boosting framework to functions of vector inputs and sketches a simple algorithm that can be used efficiently with histogram-based decision trees.

Low-Earth orbit (LEO) satellite Internet has become an indispensable infrastructure that provide growing coverage for global users. Despite extensive measurement efforts, the principles underlying region-level performance characteristics remain insufficiently understood, limiting the ability to identify region-specific latency signatures under dynamic network conditions. In this paper, we formulate the problem of region-level latency characterization using Starlink round-trip time (RTT) measurements from the public LENS dataset. We then propose a hierarchical analytical framework that transforms raw RTT sequences into multi-scale statistical features for cross-region comparison. Using data from five geographically representative regions, we demonstrate that latency differences are strongly associated with deployment factors, particularly infrastructure availability and Starlink dish-to-Point-of-Presence distance. Mutual information analysis identifies minimum RTT as the most discriminative feature, which is further supported by XGBoost-based feature importance. The proposed model well achieves 83% accuracy on short-term data. However, its performance degrades over longer periods, indicating limited temporal generalization and motivating the need for adaptive models and feature representations for long-term performance in the future.

Collaborative learning is sustainable only when it benefits each participant. Standard federated learning optimizes a global average objective, which can under perform for clients whose data distributions differ substantially from the population. We study selfish personalization: how a designated target client can use peer gradients to minimize its own risk while avoiding negative transfer. We propose SP-CACW, a convergence-aware client-weighting framework that selects aggregation weights by minimizing an upper bound on the target client's convergence error. The resulting rule explicitly trades off peer bias against stochastic variance and can assign zero weight to harmful peers. We provide convergence guarantees under smoothness and bounded-variance assumptions and evaluate the method on MNIST, CIFAR-100, and LEAF Shakespeare, where it is competitive with or improves over strong personalized and clustering baselines.

Large language models (LLMs) are increasingly used to take actions in the real world and support human decision-making, yet most agents rely on parametric knowledge, fixed post-training data, retrieval, or search. This paradigm breaks down in novel domains and for sophisticated queries that cannot be answered from prior knowledge alone. Knowing the laws of physics, for instance, does not by itself enable LLMs to answer queries or complete long-horizon tasks in a complex physical system. To address this, we introduce Hierarchical Experimentalist Agents (HExA), an in-context self-improvement framework to learn from active experimentation. HExA iteratively designs and refines query-relevant experiments, learns a reusable library of composable skills from experience, and integrates experimental evidence to answer queries or take actions. HExA is training-free, compatible with any black-box model, and does not require external supervision, oracles, or offline data. To evaluate active experimentation, we introduce Interphyre, a tool-calling benchmark built on the PHYRE 2D procedural physics environment, where agents propose interventions and test hypotheses through simulation APIs. Experiments show that current LLM agents struggle in these settings, especially on the hardest levels of Interphyre. Claude Sonnet 4.6 achieves only 2% success, while HExA improves the same model to up to 77% success. HExA also improves open-weight models and outperforms agentic baselines such as ReAct and Reflexion. Moreover, using only skills learned from easier levels and transferred without active experimentation, HExA achieves 44% success, demonstrating the reusability and generalization of its learned skills. Overall, HExA shows that learning through active experimentation can help agents discover useful knowledge, acquire reusable skills, and make efficient progress on novel long-horizon tasks.

Group Relative Policy Optimization (GRPO) is a default recipe for process-supervised reinforcement learning of LLM reasoners, and dense process supervision -- via learned process reward models (PRMs) or on-policy-distillation KL signals -- is a common way to densify its otherwise weak outcome reward. Layering such a step-level signal on top of GRPO's group-standardized advantage, however, exposes three structural pathologies: \emph{channel contamination} between the pooled process, outcome, and format streams at group standardization; \emph{resolution mismatch} between the granularity of the process signal and the granularity of the logical decisions being credited; and a \emph{cumulative trap} by which GRPO's return-to-go sum surfaces either length inflation or truncated exploration depending on the sign regime of the signal. We propose \textbf{PASS} (\emph{Process Advantage Signal Shaping}), a compact middleware that sits between any scalar step-level process signal and GRPO's clipped surrogate and addresses the three pathologies in turn: \emph{Advantage Fusion} standardizes the three streams independently within each group, \emph{Chunk-by-Value} derives value-homogeneous chunks from the signal itself and broadcasts credit within each chunk, and \emph{Divide-Length} converts the cumulative objective into an average-value-density score. We validate PASS across two domains and two process-signal paradigms -- a learned PRM on mathematical reasoning and an on-policy-distillation KL signal (with a generalized variant) on multi-hop question answering -- and under two group-standardization operators. In every regime PASS delivers a consistent pass@1 gain over the corresponding GRPO baseline.

Diffusion and continuous-flow generative models achieve high-quality generation, and their deterministic sampling can be formulated as solving learned ODE dynamics. However, accurate ODE discretization often requires many steps, making efficient few-step generation a key challenge. Among acceleration strategies, reflow-based distillation simplifies teacher ODE trajectories so that a student model can approximate the teacher transport with fewer steps. We identify a theoretical limitation of this paradigm, namely that trajectory matching can under-determine the distribution induced by the student model. In particular, two student models can attain the same trajectory-matching loss while inducing different endpoint marginal distributions, which may lead to different generation quality. To address this limitation, we introduce a marginal-alignment regularizer that penalizes the discrepancy between the student-induced marginal and the corresponding teacher marginal at the endpoint of each distillation interval. The regularizer is computed by tracking log-density changes along the ODE induced by the student model and evaluating scores from the frozen teacher model, without requiring auxiliary trainable networks or adversarial optimization. The resulting framework applies uniformly to the reflow family, including vanilla reflow and piecewise reflow. We further prove a telescoping total-variation bound showing that local marginal alignment controls the final-time discrepancy between the student-induced and teacher-induced distributions. Experiments on benchmark backbones demonstrate the effectiveness of the proposed method for few-step generation.

We identify intervention bias as a previously unquantified failure mode of zero-shot large-language-model (LLM) educational advisory agents: without task-specific training, they recommend action when a hindsight-optimal oracle policy mandates inaction. In a six-arm ablation on the Open University Learning Analytics Dataset (N=800 students, four temporal cutoffs), at day 56 -- when the oracle designates 70.1% of students as needing no intervention -- zero-shot GPT-4o recommends action for 73%, a 43 percentage-point false-positive rate. Commercial RAG and SQL-augmented retrieval are comparably miscalibrated; at 10,000 students this implies about 4,300 unnecessary advisor contacts per cycle. Supervised policy learning eliminates this bias: a trajectory-conditioned ONNX Decision Transformer (DT) and a snapshot XGBoost classifier, trained on the same oracle-labelled trajectories under strict prefix-only features, both achieve near-zero calibration error. The DT reaches macro-F1 0.79 (macro-recall 0.85) across all five action classes, predicting even the rare load-reduction action without collapsing, at a 0% action flip rate and sub-5 ms CPU decision latency. The two supervised arms are on par; the DT's edge over XGBoost at the final cutoff is indicative only (unpaired across cohorts). Scope: we validate Stage-2 decision-making (EAV state vector to supervised policy) under controlled oracle input from structured OULAD data; high fidelity reflects feature-oracle alignment, not general high-stakes-AI capability. The most robust finding is the intervention-bias contrast, not the absolute accuracies. We also show an Evaluation Gap: LLM-as-judge scoring (DeepEval G-Eval) is blind to intervention bias, rewarding fluent over-prescription rather than decision quality.

LLM agents carry conclusions across steps and sessions in compressed memory, and memory products (e.g., mem0, LangMem) rewrite conversation into stored "facts" that later steps trust. We show this rewriting manufactures confidence: across our constructed agent settings, a casual, hedged remark becomes a confident, dated assertion the agent then obeys like a verified fact, granting every above-clearance request it faces. No attacker is needed: a role that was true once and never corrected is stored as a flat fact and acted on like a deliberate injection. We then isolate what the agent responds to. It is not the source: attributed, unattributed, and even forged "system of record" claims all grant alike. It is the confidence of the phrasing. A hedge is discounted, a flat assertion is obeyed, and this holds with no special keyword. Not all hedges are equal, though: the evidential register is the least-discounted, with "reportedly" obeyed like a flat assertion on most models. The obvious fixes fail. A passive "unverified" tag is ignored, and an active "do not trust this" instruction escalates even correct memory, so it is safe only by refusing to decide. The real fix lives in the store: keep the tentative phrasing rather than upgrade it. But that is hygiene, not a defense against an attacker who can simply write a confident lie. The deployable lesson is narrower and constructive: a single load-bearing memory is the hazard, and one redundant source restores correct decisions. We release the harness and demonstrations.

We introduce the Complexity Ceiling Benchmark (CCB), a controlled evaluation of how language-model reasoning decays as the number of required sequential steps grows. CCB fixes the semantic content of a task and varies only its depth N in {5,...,50} across three structurally distinct regimes: grounded spatial state-tracking, abstract symbolic pointer manipulation, and transitive relational inference. Across 6,000 trials over five frontier and open-weight LLMs we find a consistent pattern of geometric per-step decay with widely separated domain ceilings: on the first two regimes the strongest models retain pd>0.92 across N=50; on the third every model collapses by N=5, with the best model's 50%-success horizon at H0.5~4.7 steps despite pd=0.863. A trace-level metric (TFBC) shows that 14.5% of correct answers across the benchmark are reached via incorrect intermediate reasoning. Forced verbose state-tracking does not move the ceiling (McNemar p=1.000), and the mean step at which reasoning first diverges, k*, predicts within-domain accuracy better than parameter count. CCB and the geometric decay model together reduce a model's long-horizon reasoning profile to one interpretable number per task family.

Diffusion Language Models (DLMs) are typically trained under fixed context structures, restricting denoising to predetermined token subsets. This creates a mismatch between training and inference, where models must operate over arbitrary configurations, leading to degradation off the training grid. We propose Adaptive Block Diffusion (ABD), which resolves this mismatch by optimizing denoising risk over a distribution of prefix-window configurations. By treating the configuration as a stochastic variable, ABD trains a single model over the full configuration space without architectural changes. We show that generalization across decoding strategies is governed by the support of the training distribution, and that ABD guarantees denoising optimality for any inference policy whose configurations are covered during training. Empirically, ABD exhibits structural invariance across decoding scales, avoiding off-grid collapse and recovering a monotonic relationship between block size and perplexity, while matching or outperforming fixed-block specialists at their target scales.

Emotion recognition of song lyrics is a challenging task since lyrics may not necessarily align with the overall emotion of a song. As a result, lyrics annotation remains largely underexplored. Drawing inspiration from research in large language model (LLM) assisted annotation, we examine the alignment between humans and LLMs for annotation of lyrics by creating a new sentence-level dataset of lyrics. Our observations highlight the subjectivity of the task and the inherent challenges. Following this, we present a hybrid annotation framework that optimizes human and LLM annotation by predicting potential misalignment in annotation.

Technology computer-aided design (TCAD) semiconductor device simulation is fundamentally constrained by the high computational cost of iteratively solving coupled drift-diffusion equations. Existing ML surrogates either reduce internal physics to macroscopic scalar regressions, or rely on single-step mappings that lack the iterative refinement required to resolve stiff, coupled fields. To address this, we introduce PCGD, a Physics-Guided Conditional Graph Diffusion framework operating natively on unstructured TCAD meshes to predict coupled electrostatic and carrier density fields. PCGD employs a Condition-Aware MeshGraphNet denoiser that explicitly injects boundary conditions and device structure context via global cross-attention. By augmenting data-driven denoising with a physics-guided hybrid objective that integrates exponent-free quasi-Fermi gradient matching with noise-aware PDE residuals, PCGD progressively enforce physical constraints in the iterative diffusion trajectory. This strategy successfully bypasses the numerical instabilities typical of stiff drift-diffusion equations. Evaluated on a challenging mixed PN/MOS benchmark, PCGD significantly outperforms deterministic one-step regression (1.207% error) and local diffusion (1.585% error) baselines by achieving a sub-percent mean relative field error of 0.835%, while concurrently reducing maximum PDE residual errors by nearly three orders of magnitude compared to pure diffusion. It also transfers robustly to unseen SOI topologies (0.815% error) via LoRA adaptation, using 5.30$\times$ less data and 14.34$\times$ fewer parameters than full fine-tuning. Ultimately, PCGD bridges the computational efficiency of generative surrogates with the rigorous physical fidelity of traditional TCAD, unlocking highly scalable, field-level analysis for robust device engineering.

Multi-Agent Debate (MAD) with Majority Voting is a dominant paradigm for improving LLM reasoning, yet its effectiveness rests on the Condorcet Jury Theorem's assumption of independent errors. Because contemporary LLMs share similar pretraining corpora, their errors are strongly correlated, causing the majority to systematically suppress correct minority opinions, a phenomenon we term Minority Truth. Through debates among three heterogeneous LLM agents on six benchmarks, we find that roughly one in four divergent cases has the minority holding the correct answer, yielding a 10-percentage-point theoretical recovery margin. We propose Minority Sentinel, a lightweight meta-classifier that extracts a multi-dimensional debate fingerprint from debate logs and trains a LightGBM model to decide when to overturn majority voting. Minority Sentinel achieves a stable Flip Precision of 81.2% with positive Net Gain across all six datasets and all 20 random seed trials, demonstrating that debate logs contain sufficient behavioral signals for a non-LLM classifier to reliably recover suppressed minorities without degrading system accuracy. The LLM-as-Judge baseline yields negative Net Gain despite higher recall, confirming that flip safety, not recovery volume, determines intervention value.

Reasoning large language models (LLMs) have recently made much progress in complex problem-solving, leveraging internal reasoning (or thought) to guide their solution generation. However, existing LLM-based counseling agents, including those using Motivational Interviewing (MI), generate responses without explicitly aligning thoughts with counseling techniques, limiting their effectiveness. We propose MIThinker, a lightweight thinking model that generates therapeutic thoughts to guide MI counseling agents in strategy selection and response generation. To overcome the lack of annotated thought data, we introduce AugR1-MI, an automated pipeline that reverse-engineers counselor's thoughts from observed responses. Through two-stage training combining supervised fine-tuning and reinforcement learning, MIThinker demonstrates improved theory-of-mind assessment and strategy alignment. Comprehensive evaluations show that MindfulMI, our agent leveraging MIThinker, achieves MI competency comparable to state-of-the-art systems with an order of magnitude less computation.

We derive the linear union-of-subspaces (UoS) model for subspace clustering (SC) from the nonlinear mixture model (NMM) used in blind source separation (BSS) to represent a D-dimensional observation vector as an unknown multivariate nonlinear mapping of C latent variables. Assuming the mapping is differentiable up to an unknown order K, we approximate NMM by a K-th order Taylor expansion, yielding a model equivalent to the linear UoS framework underlying SC. This establishes that: (i) the smoothness order K corresponds to the unknown subspace dimension d; (ii) KC equals the number of anchors; and (iii) the sparsity of the representation vector equals K (i.e., d). These relationships enable estimation of bounds on subspace dimension, and that is validated on six benchmark datasets using five established SC algorithms. Established theoretical results are important for post-processing of self-representation matrices estimated by SC algorithms.

In recent years, models based on the Transformer architecture have seen widespread applications and have become one of the core tools in the field of deep learning. Numerous successful techniques, such as parameter-efficient fine-tuning and efficient scaling, have been proposed surrounding their applications to further enhance performance. However, the success of these strategies has always lacked the support of rigorous mathematical theory. To study the underlying mechanisms behind Transformers and related techniques, we first propose a Transformer learning framework motivated by distribution regression, with distributions being inputs, connect a two-stage sampling process with natural language processing, and present a mathematical formulation of the attention mechanism called attention operator. We demonstrate that by the attention operator, Transformers can compress distributions into function representations without loss of information. Moreover, with the advantages of our novel attention operator, Transformers exhibit a stronger capability to learn functionals with more complex structures than convolutional neural networks and fully connected networks. Finally, we obtain a generalization bound within the distribution regression framework. Through the aforementioned theoretical results, we further discuss some successful techniques emerging with large language models (LLMs), such as prompt tuning, parameter-efficient fine-tuning, and efficient scaling. We also provide theoretical insights behind these techniques within our novel analysis framework.

Online inspection images of final optics in high-power laser facilities contain pseudo-damage sites that closely resemble true damage sites. Determining the authenticity of online-detected sites is therefore difficult and requires accurate matching to offline ground-truth sites. However, this matching remains highly challenging due to limited match-discriminative features, local geometric distortions, and numerous distractor sites. Existing matching models mainly suppress distractors implicitly through loss-function supervision. We propose a confidence-feedback-weighted graph matching network that requires only damage-site centroid coordinates as input. It estimates node matchability confidence from each round of matching scores and feeds it back as a reliability weight to guide subsequent edge-feature aggregation, thereby suppressing distractor propagation and enhancing cross-graph discriminability. Within this framework, a geometric consistency constraint calibrates spurious high-confidence matchability estimates, while a hard-example mining loss improves discrimination between structurally similar sites. Experiments on our Complex-Scene dataset show that the proposed method achieves a matching F1-score of 96.36$\%$ with robust and efficient performance.

Large language models (LLMs) demonstrate broad reasoning abilities but struggle with accuracy and reliability in specialized domains such as travel, where reasoning depends on precise definitions, rules, and expert-defined conceptual frameworks, and where confident but unfounded outputs arise from a reasoning failure in which the model has not internalized the underlying domain graph rather than from missing domain knowledge alone. We propose a modular pipeline for building a travel-domain reasoning LLM grounded in an expert-designed knowledge graph (KG). Our pipeline integrates a travel KG that encodes domain entities and their relationships, a bottom-up construction procedure that walks the KG to produce multi-hop question answer (QA) pairs, a supervised fine-tuning stage that embeds the domain knowledge into a reasoning-capable LLM using the generated QA pairs as auditable reasoning traces, and a travel-domain benchmark dataset that measures the fine-tuned model's accuracy and calibration. We evaluate our approach using Qwen3-4B with LoRA adaptation. Our reasoning model achieves an $82.4\%$ exact match on the benchmark. This performance significantly outperforms the pretrained Qwen3-4B baseline at $22.4\%$. A calibration analysis decomposes the residual $17.57\%$ of errors into two distinct failure modes: an over-confident multi-label decoder that predicts both correct answers plus one spurious option on most dual-answer mistakes, and a smaller reasoning failure on single-answer questions where the supporting facts are present in the KG but the model fails to reconstruct the correct multi-hop path. This split confirms that explicit KG-grounded reasoning substantially improves the accuracy and uncertainty interpretation of LLMs in specialized domains, and isolates per-option calibration and trace-length-aware decoding as the next axes of improvement.

We study repeated bidding in multi-unit discriminatory (pay-as-bid) auctions for a single bidder with per-round utility equal to value minus $α$ times payment, where $α\in[0,1]$ is a cost-of-capital parameter. The bidder aims to maximize cumulative utility over $T$ rounds subject to a total budget $B$. The problem is challenging even without budgets: the action space is exponential in $M$, the maximum demand of the bidder and the valuation vector (context) varies over time. Exploiting a decomposition of utility across units, we develop polynomial-time learning algorithms based on shortest paths in a directed acyclic graph, obtaining sublinear regret under both full-information and bandit feedback. In the bandit setting, the regret is independent of the number of contexts due to complete cross-learning: observing the utility of the chosen action under the realized context reveals the utility for the same action under all counterfactual contexts. With budget constraints, when the average normalized per-round budget $ρ=\frac{B}{MT}<1$, we design a coupled primal-dual algorithm in which the DAG-based procedure uses dual-adjusted edge weights for primal updates, while online gradient descent updates the dual variable, yielding $ρ$-approximate sublinear regret. Finally, we give implementations whose per-round time and space are independent of the number of contexts, enabling scalability to large or even infinite context spaces.

Financial decision-makers face more information than they can directly inspect, making context compression necessary. Yet when large language models (LLMs) compress financial source material, they can alter the investment judgment supported by the original source. We frame this problem as information fidelity: compression loses fidelity when it changes the decision induced by the source. In agentic systems, such losses may recur across intermediate steps and amplify throughout the decision process. Across financial filings and earnings-call transcripts, we find that LLM-based compression can produce fluent and factually plausible compressed contexts that nevertheless alter downstream decisions. We analyze two diagnostic patterns associated with fidelity loss: decontextualization, where salient evidence is retained but separated from the caveats and contextual qualifiers needed for correct interpretation, and model dependency, where different compressors expose different views of the same source. We then propose Agentic Context Compression, which generates multiple candidate compressions and audits their disagreements against the original source. Our results suggest that financial compression should be evaluated not only by efficiency or factuality, but also by its ability to preserve decision-relevant context.

Vegetable prices in Sri Lanka are highly volatile because the market is largely import-isolated, so supply disruptions quickly drive prices up. This study develops a machine learning framework to forecast such volatility by incorporating supply-chain-aware features and explicitly modelling the country's two cultivation seasons, Maha (October-April) and Yala (May-September). An integrated dataset was constructed by combining retail and farmer-gate prices with origin-aligned weather variables, diesel costs, and exchange rates across 12 vegetable varieties and 14 market centres from 2013 to 2019. A gradient-boosted ensemble model (XGBoost and LightGBM) was trained and optimised using Optuna, and unified and season-specific configurations were compared. Results show that season-specific models improve within-season fit, with the Yala-specific model achieving the highest R2 of 0.9420 (95% CI [0.690, 1.000]), while the unified model delivers the best overall predictive accuracy of 90.84% (95% CI [88.34%, 91.52%]) and an R2 of 0.9281 (95% CI [0.760, 1.000]). Notably, the unified model maintains 85.96% accuracy on a completely unseen 2024 hyperinflationary period without retraining, successfully tracking major price surges. These findings suggest that agricultural price movements in import-constrained markets are meaningfully predictable when models capture supply-chain dynamics, offering practical value for early warning and decision making by farmers, traders, and policymakers. Existing studies on Sri Lankan vegetable prices are confined to Autoregressive Integrated Moving Average (ARIMA) and Generalized Autoregressive Conditional Heteroskedasticity (GARCH) applied to single markets, with no supply-chain features, seasonal segmentation, or cross-regime validation.

Vision-Language-Action (VLA) models represent a promising direction for embodied intelligence in surgical robotics. Despite the prevalence of VLA benchmarks for general robotics, standardized evaluation platforms specifically designed for surgical contexts remain absent. To address this limitation, we present SurgVLA-Bench, the first comprehensive benchmark for evaluating VLA models in laparoscopic surgical robotics. Leveraging the SurRoL simulation platform, we construct a hierarchical task taxonomy ranging from atomic actions to complete surgical procedures, complemented by a multi-dimensional evaluation framework assessing action accuracy and semantic consistency. We then systematically evaluate two representative paradigms, including autoregressive models such as OpenVLA, and flow matching models such as $π_{0}$, $π_{0.5}$, and SmolVLA. Our experiments show that autoregressive models tend to excel in semantic understanding, while flow matching models often achieve higher task precision but may face generalization trade-offs. However, even the best-performing models remain far from satisfactory, as the constrained endoscopic field of view, restricted viewing angles, and frequent occlusions persist as fundamental physical bottlenecks. The code and data are available at https://github.com/VCL-HNU/SurgVLA

We present KrishokChat, the first citation-grounded Bengali agricultural instruction-tuning dataset for crop advisory in low-resource settings. We establish a foundation of 290 hierarchical Knowledge Nodes, extracting disease symptoms, management practices, chemical dosages, and verbatim citations from 129 domain-filtered agricultural manuals. Every training instance inherits a verified citation header, guaranteeing 100% citation provenance. Using a Partitioned Seed Generation Matrix, these nodes are expanded into 139,200 supervised fine-tuning pairs, and augmented with 5,300 chemical safety and 1,000 adversarial safety instances, yielding 145,500 QA pairs across 18 crop categories. To evaluate real-world performance, we introduce the Farmer Benchmark, comprising 1,001 authentic farmer queries curated from field surveys and digital portals. Empirical evaluation on Gemma-4-E2B reveals that while fine-tuning on KrishokChat vastly improves structured formatting, standalone models still struggle with exact chemical dosage generalization. This highlights the dataset's true value as a verified knowledge base for retrieval-augmented generation (RAG) rather than mere parametric memorization. All data, code, and benchmarks are released under CC-BY-4.0.

Data-generating priors are a central component of tabular foundation models because they define the task distribution used during pretraining. However, priors are rarely evaluated as independent components, making it difficult to understand how much they affect downstream model behavior. This raises a methodological question: how can priors from different tabular foundation models be compared independently of the architectures and training protocols they were introduced with? To study this question, we implement a unified interface for publicly available priors from recent tabular foundation models and priors constructed from real datasets. We generate training tasks from each prior, train the same model architecture under a fixed training protocol, and evaluate the resulting models on shared downstream classification tasks. We compare priors through both generated-task statistics and downstream predictive performance. Our results show that different priors favor different downstream behaviors, with some achieving stronger absolute performance and others exhibiting more consistent relative rankings across datasets. We further find that data-level similarity only partially explains downstream behavior. Our code is available at https://github.com/automl/TFM-Playground/tree/prior-dev.

Heterogeneous graph neural networks (HGNNs) have achieved strong performance in modeling complex graph-structured data with multiple node and relation types. However, their robustness under realistic black-box adversarial settings remains insufficiently explored. Existing attacks on HGNNs usually assume access to model gradients, soft prediction scores, or the complete graph structure, which is often unavailable when HGNN-based services are deployed as closed systems. In this paper, we propose Blackknife, a hard-label, query-limited, and structure-limited black-box evasion attack framework for heterogeneous graph neural networks. Blackknife assumes no access to the victim model architecture, parameters, gradients, logits, confidence scores, or the full graph structure. Instead, it only relies on locally observable one-hop heterogeneous structures and a small number of hard-label queries. To generate effective perturbations under these strict constraints, Blackknife first constructs a local relation-aware surrogate model from observable heterogeneous neighborhoods. It then relaxes discrete edge addition and deletion operations into continuous soft weights and optimizes them through projected gradient descent. Finally, the optimized perturbations are discretized into relation-preserving structural rewiring operations and verified using limited hard-label feedback from the victim model. Extensive experiments on three benchmark heterogeneous graph datasets, including ACM, DBLP, and IMDB, demonstrate that Blackknife consistently achieves strong attack success rates against representative HGNN models. The results further show that Blackknife remains effective under topology-based defense strategies, revealing the vulnerability of HGNNs to local structure-limited black-box attacks.

Group Relative Policy Optimization (GRPO) eliminates the learned critic in PPO by using the mean reward of grouped rollouts as a baseline. We provide a rigorous derivation of GRPO from first principles of the policy gradient theorem, revealing a fundamental credit assignment failure: under output-only reward, every token in a rollout receives identical advantage, collapsing token-level credit to a single scalar. We prove this induces gradient sparsity that intensifies over training, and demonstrate empirically via SVD analysis of GRPO gradients on Nemotron-4B/GSM8K that the gradient matrix has effective rank $\approx$ 2 regardless of group size $R \in \{2, 4, 8\}$. We formalize this as an intrinsic rank-2 structure arising from the zero-sum constraint on advantages and derive conditions under which GRPO's baseline is optimal. Our results characterize when GRPO's simplicity is theoretically justified and identify the credit assignment bottleneck as the key limitation for multi-step reasoning.

Robust robot autonomy depends on scene representations that remain stable enough to support localization, navigation, and downstream decision making in dynamic environments. Monocular Gaussian Splatting SLAM provides high-fidelity mapping, but current uncertainty-aware methods still treat dynamic regions largely as per-frame observations. This makes the representation effectively memoryless: when a pedestrian slows, pauses, or reappears after occlusion, the current frame may look static, allowing dynamic content to be absorbed into the map and leaving persistent ghosting artifacts. We argue that this failure reflects a representation-level mismatch. Dynamic-ness is not an instantaneous appearance property, but a temporal property defined by motion history. Building on this view, we introduce Motion Permanence: the principle that an object's dynamic identity should persist over time rather than be re-decided from each frame independently. We realize this principle in MoPe, a memory-aware uncertainty filter for monocular Gaussian mapping. MoPe propagates the historical dynamic posterior through geometry-consistent SE(3) warping and fuses it with current-frame evidence using bounded Bayesian log-odds updates. The resulting persistent posterior guides tracking, mapping, dynamic-aware Gaussian insertion, and Gaussian-level post-cleanup. On Wild-SLAM, Bonn, and TUM sequences, MoPe improves tracking robustness and reduces residual ghosting, with the strongest gains on dynamic-human scenes that most directly violate the memoryless assumption. These results show that maintaining temporal dynamic state inside the scene representation is a practical step toward more reliable representation-centric autonomy in changing real-world environments.

Despite the capability of parallel decoding, diffusion large language models (dLLMs) require many denoising steps to maintain generation quality, motivating recent research on efficient decoding strategies. However, existing studies have reported inconsistent evaluation results even under seemingly identical evaluation settings, risking biased conclusions about dLLM decoding methods. To understand this evaluation concern, we conduct a rigorous evaluation of current decoding methods for dLLMs across diverse evaluation settings. Surprisingly, our analysis reveals that the ranking of decoding methods is highly sensitive to the choice of prompt templates. Single-template evaluation can lead to an illusion that decoding methods improve inference efficiency without performance degradation. Through comprehensive experiments, we find that current parallel decoding methods consistently underperform the single-token decoding baseline, failing to overcome the speed-quality trade-off. We further identify this evaluation inconsistency as the high sensitivity of parallel decoding methods to minor variations in prompt templates. Our experiments show that an effective prompt template can achieve strong evaluation results even with fewer denoising steps, markedly outperforming the marginal gain from increasing denoising steps. Beyond prompt templates, our experiments indicate that overlooked evaluation settings can also notably affect the assessment of decoding methods. Based on these findings, we propose practical guidelines for the reliable evaluation of decoding methods in dLLMs.

LLM agents handle user requests on behalf of organizations through tool calls and must follow the company policies stated in their system prompts. Prior work approaches this as a safeguarding problem -- external checks that block non-compliant agent actions. We argue that policy adherence is a broader problem: real workflows unfold across many turns, require explicit user confirmation and prerequisite reads, and hinge on the content of the dialogue rather than on any single argument value. Meeting this bar requires (i) full conversation context, (ii) self-reasoning over the policy and the current dialogue, and (iii) conversation-specific remediation that guides the agent's next turn -- three capabilities that prior safeguard work has often underestimated. We introduce POLICYGUARD, a sub-agent verifier that shares the agent's view of the dialogue, reasons over the policy in context, and provides actionable feedback for the agent's next turn. On tau^2-BENCH airline across three vendors (GPT-5.4, Claude Sonnet 4.6, Gemini 2.5 Pro) with four trials per setting, POLICYGUARD improves PASS4 by +12.0 / +6.0 / +12.0 pp. Per-call analyses show POLICYGUARD achieves higher policy-violation recall while blocking roughly half as often as argument-level guards.

Autoregressive LLM decoding evaluates every generated token through the full layer stack, even though many tokens become predictable at intermediate depths. Existing lossless depth-adaptive methods exploit this redundancy by choosing a single non-final exit depth and verifying its prediction with the final-depth model. However, our measurements show that this selection-based strategy leaves substantial headroom: choosing an exit too late wastes computation, while choosing one too early triggers fallback and discards dependent drafts. We propose Depth Exploration Decoding (DEX), a lossless decoding algorithm that replaces single-depth selection with parallel exploration over multiple candidate depths. At each commit position, DEX validates candidates against the final-depth reference, commits exactly the final-depth token, and collapses the exploration lattice to retain only reusable branch states. This expand--commit--collapse procedure preserves equivalence to standard autoregressive decoding while reducing the cost of committing each token. Across early-exit-trained and standard LLMs, DEX outperforms representative depth-selection baselines and achieves competitive end-to-end throughput against speculative and distributed decoding methods. Moreover, DEX improves as the explored depths become finer, showing that parallel depth exploration provides a scalable way to exploit the underused depth axis of LLM decoding.

We address the problem of online multi-human multi-robot teaming through the lens of a linear matching bandit framework, where a learner assigns robots with unknown features from a fixed pool to distinct sets of human agents over multiple rounds. To solve this problem, we propose LinMatch, an online learning algorithm that updates the confidence intervals of the unknown features and makes the optimistic matching under uncertainty. The contributions and novelty of this work are twofold. First, we recast the optimistic matching problem in each round as a linear program of maximum weighted matching, efficiently solvable by the celebrated Hungarian algorithm. Second, we provide novel bounds for matching with linear feature problems, showing an upper bound of $\tilde{O}(d\sqrt{MKT})$ and a minimax lower bound of $Ω(d\sqrt{MKT})$, establishing a tight optimal regret rate of $\tildeΘ(d\sqrt{MKT})$. This demonstrates that LinMatch achieves strictly optimal achievable regret with respect to the total number of rounds $T$, the feature dimension $d$, and the matching parameters $M$ and $K$. The proposed algorithm and bounds apply to a wide range of matching problems with applications beyond human-robot matching, such as housing allocation, recommendation systems, and more.

Block Diffusion Language Models (BD-LMs) improve diffusion-based text generation with KV caching and flexible-length generation. A natural next step is to extend them from Single-Block Diffusion (SingleBD) to Multi-Block Diffusion (MultiBD), where a \textit{running-set} of consecutive blocks is decoded concurrently for inter-block parallelism. However, existing BD-LMs are mostly trained under teacher forcing, where the model observes only one noisy block conditioned on a clean prefix. While the recent diffusion forcing strategy introduces visibility among multiple noisy blocks, its training states still differ from MultiBD inference, where decoding operates on a bounded \textit{running-set} with heterogeneous slot-wise noise patterns. To bridge this gap, we propose \textit{Multi-Block Diffusion Language Models} (MBD-LMs), obtained by post-training BD-LMs with \textit{Multi-block Teacher Forcing} (MultiTF). MultiTF integrates teacher forcing and diffusion forcing by training on bounded \textit{noise-groups} conditioned on clean prefixes, with randomized \textit{noise-schedulers} that better match MultiBD inference states. To make MultiBD practically executable, we further introduce an optimized decoding algorithm based on the \textit{Block Buffer} mechanism that preserves prefix-cache reuse, keeps input shapes static, and translates increased decoding parallelism into wall-clock acceleration. Empirically, MBD-LLaDA2-Mini increases average Tokens Per Forward pass (TPF) from 3.47 to \textbf{6.19} and improves average accuracy from 79.95\% to \textbf{81.03\%}; when combined with DMax, MBD-LLaDA2-Mini-DMax reaches an average TPF of \textbf{9.34} with only a 1.02\% accuracy drop on math and code benchmarks.

OCR systems, ranging from classical engines to specialised OCR vision-language models (OCR-VLMs) and frontier multimodal LLMs, report strong results on English and Chinese document benchmarks, yet their behaviour on Indic scripts is largely uncharacterised. We benchmark ten systems on Devanagari (Hindi): classical EasyOCR; open VLMs (Qwen2.5-VL-3B, Qwen3-VL-8B, olmOCR-7B); specialised OCR-VLMs (DeepSeek-OCR, Unlimited-OCR); and frontier closed models (Gemini 2.5 Flash, Claude Opus 4.7, GPT-5.5, Mistral OCR), across four synthetic degradation conditions and 300 real printed scans. We report four findings. First, on clean rendered text all ten cluster within chrF++ 91 to 98, so synthetic text does not separate them. Second, under degradation the specialised OCR-VLMs are the most fragile: DeepSeek-OCR suffers rare but catastrophic repetition failures (outputs up to 71 the reference length) that wreck its corpus mean even though its median is the best of any system, which is why we report median and catastrophic-rate instead of the mean. Third, on real scans nine of the ten systems collapse (EasyOCR falls from chrF++ 93.6 to 58.3) and the field spreads across a 76-point range, so synthetic renders badly overstate Devanagari quality. Fourth, strong English OCR does not predict Indic OCR: GPT-5.5 drops to chrF++ 58.5 (tying classical EasyOCR) and olmOCR-7B, the model behind olmOCR-Bench, falls to 40.5, while the open Qwen3-VL-8B (75.2, runnable on a single 24 GB GPU) beats GPT-5.5 and approaches Mistral; Gemini and Claude lead at 86.3 and 82.2. An error taxonomy separates surface errors (numerals, punctuation) from structural ones (conjuncts, matras, nukta), and a byte-level (ByT5) post-corrector improves a cheap engine on its own error distribution (chrF++ +1.2 to +1.5) but does not transfer across engines. We release the benchmark, code, and models.

Agent-based evacuation simulations are widely used to study crowd behavior during emergencies, but many models rely on assumptions such as perfect event awareness, complete exit knowledge, and fully rational decision-making. This paper presents an extended evacuation framework that integrates cognitive, emotional, social, and personality-related mechanisms into a unified model of human behavior under uncertainty. The framework incorporates a dynamic event-awareness mechanism based on a continuous Event Certainty Level, a memory-based representation of exit knowledge subject to acquisition, forgetting, and recall, a continuous fear model in which panic emerges as a high-intensity state, and an OCEAN-based personality representation. Neuroticism is explicitly integrated into the emotional model, influencing fear generation, escalation, social contagion, and recovery. Behavioral heterogeneity is further captured through individualized decision thresholds that affect responses to perceived risk. The framework is evaluated through simulation experiments examining the effects of spatial familiarity, memory robustness, decision sensitivity, emotional dynamics, and personality variation. Results show that cognitive, emotional, and personality-driven processes substantially influence evacuation dynamics, reducing evacuation efficiency and generating realistic crowd phenomena such as delays, confusion, injuries, and socially influenced behaviors. The proposed framework provides a more realistic representation of human behavior in emergency evacuations and supports systematic investigation of the interactions between cognition, emotion, personality, and crowd dynamics.

We present AnyBody, a unified whole-body humanoid controller driven by an arbitrary subset of body keypoints chosen at deploy time. Prior physics-based trackers either rely on expensive full-body motion capture and error-prone trajectory retargeting, which bottleneck scalable data collection and policy learning, or decompose upper- and lower-body control into separate hierarchical representations, sacrificing the coordinated whole-body motions that loco-manipulation requires. We close this gap by learning a single latent motion representation that any keypoint subset can address. To achieve this, we first train a privileged teacher tracker on a large unstructured motion corpus and distill it online into a deterministic encoder-decoder student whose latent space is a unit sphere. We then train a transformer keypoint encoder that admits any subset of body keypoints through masked self-attention, aligning it to the privileged latent. Additionally, we treat the frozen decoder as a motor prior and specialize downstream tasks with a lightweight residual corrector in the latent space. We demonstrate the effectiveness of AnyBody by tracking large-scale human motions from arbitrary keypoint subsets, free-form control, flexibly teleoperating, and learning downstream behaviors including locomotion, in-air writing, and obstacle-reach.

LLM serving is increasingly dominated by long and dynamic decode workloads from agents, reasoning models, and extended conversations. When bursty long-context demand exceeds deployed capacity, existing serving systems typically scale out by launching additional serving instances with model replicas. This instance-level elasticity increases KV capacity only by provisioning another full copy of the model, inheriting startup latency, memory overhead, and batch fragmentation. We present KernelFlume, a decode-centric architecture that disaggregates the stable projection/FFN path from core-attention computation: weight nodes execute dense projection/FFN kernels, while weightless attention nodes store token-range KV partitions and scale with request-state demand. To make this separation elastic, KernelFlume maintains a routing table that maps token ranges to attention-node endpoints. It updates routes at token boundaries and uses host-visible graph signals to drive pre-registered UCX endpoint communication outside the captured CUDA Graph. To preserve low per-token latency after disaggregation, KernelFlume combines query-first core-attention dispatch with inter-layer kernel pipelining, overlapping remote attention and communication with local projection/FFN work. On real GPU testbeds (intra-node A6000 and cross-node H100), under a dynamic long-context agentic workload serving Llama-3.1-8B, KernelFlume sustains flat p99 TPOTs of ~74 ms on A6000 and ~34 ms on H100, while lowering cost per million output tokens by up to 32% and 61%, respectively, relative to full-instance elastic scaling with ServerlessLLM, a state-of-the-art instance-startup method. Replaying the same trace at larger model scale in simulation projects a 56--66% cost reduction over ServerlessLLM, widening to 80--85% with cheaper heterogeneous attention-node hardware and persisting into the million-token context range.

We study the Bayesian fixed-budget best-arm identification problem in which a learner can abstain from making a terminal recommendation. Subject to an abstention budget $α$, we analyze the probability of undetected error--the risk of recommending a suboptimal arm without abstaining. Our central finding is that abstention induces a phase transition: without abstention, the error probability decays polynomially in the sampling budget $T$; in contrast, introducing any small positive abstention budget shifts this to an exponential decay. For Gaussian priors and rewards, in the regime $T\to\infty$ followed by $α\downarrow0$, we establish exact matching information-theoretic lower bounds and algorithmic upper bounds on the optimal error exponent, which takes the form $\exp(-\frac{α^{2}T}{8κ_ν^{2}})$. The hardness parameter $κ_ν$ represents the prior density of the top-two gap at zero, highlighting that nearly tied instances drive the fundamental error. We introduce an adaptive algorithm, PGWS, that successfully achieves this optimal exponent by expending its abstention budget on statistically ambiguous instances. We further demonstrate that this polynomial-to-exponential improvement is exclusively a Bayesian phenomenon--in the frequentist setting, abstention only affects lower-order exponent terms. We also extend our results beyond the Gaussian model.

Behavior-cloned policies often learn multiple behavior modes from demonstration datasets, including modes that are unsafe or otherwise undesired at deployment. For example, a policy trained on diverse handover demonstrations may learn to pass a knife blade-first. Standard remedies such as data curation and inference-time steering either require access to the original demonstrations for full retraining or add substantial inference-time overhead. To address this gap, we propose MoRE(Mode Redirection), which redirects policy rollouts toward desired behavior modes through a short "uncloning" step. Specifically, MoRE distills the redirection signal from a temporary mode classifier into the policy weights to steer behavior. A retain loss balances this edit by preserving desired-mode competence, allowing the standalone policy to suppress unwanted modes with zero inference-time overhead. Across eight simulated and real-world tasks, MoRE improves the average deployment success rate (SR) by 44 percentage points over the original mixed-mode policy. Among all compared adaptation and steering baselines, MoRE achieves the strongest SR and approaches the filtered-data retraining reference, while preserving task competence and inference speed. MoRE also generalizes across robot policy backbones, including Diffusion Policy and the Pi0.5 VLA, diverse task categories, and real-world deployments.

Multi-site functional MRI (fMRI) studies are essential for robust neuropsychiatric diagnosis yet suffer severe domain shifts from scanner heterogeneity, demographics, and site-specific acquisition protocols. Traditional domain adaptation requires concurrent source and target data access, violating clinical privacy regulations. Moreover, functional connectivity matrices lie on the Symmetric Positive Definite (SPD) manifold, where Euclidean operations cause geometric distortions corrupting diagnostic patterns. We propose BrainRiem, a source-free domain adaptation framework learning compact Riemannian brain prototypes via manifold-aware bi-level optimization. It employs the Log-Euclidean Metric to ensure prototypes remain valid SPD matrices, while Dirichlet Energy spectral calibration aligns their frequency characteristics with real brain networks. Only anonymized prototypes are transmitted to target sites, serving as stable anchors for training local models without source data access and reducing leakage under the evaluated attacks. Comprehensive experiments on ABIDE and REST-meta-MDD show BrainRiem consistently outperforms state-of-the-art source-free, traditional, and graph domain adaptation methods across diverse scanners and demographics. Notably, learned prototypes exhibit biologically interpretable connectivity patterns aligning with established neuroscience findings, validating the necessity of Riemannian geometry for brain network analysis.

Do language models know when they are being tested? This question matters for AI safety: a model that recognises an evaluation context could alter its behaviour strategically, making downstream benchmarks harder to interpret. Using 11 models spanning Qwen 2.5, Gemma 2, and Llama 3.2, we find a systematic size-dependent shift in representational depth: in both Qwen 2.5 and Gemma 2, the layer at which evaluation-awareness is most linearly recoverable moves from late layers in smaller models to early layers in larger ones. This suggests that scale changes not only the strength of evaluation-awareness but also where it is most linearly recoverable in the network. This depth shift helps explain why within-family scaling trajectories are non-monotonic or inverse rather than smooth and family-general, showing that a simple universal power-law account is not supported under denser within-family sampling. Finally, white-box probe signals are consistently stronger than black-box behavioural expression, and the relationship between the two varies by family in ways not predicted by probe AUROC alone.

Automated alpha mining holds the scoring function fixed and varies the search algorithm over it. A search that converges against a fixed scorer overfits whatever the scorer cannot penalize, a primary cause of the out-of-sample generalization gap. We treat the scoring function as a search artifact alongside the alpha factors and study what conditions make this joint search admissible. Sealed Joint Search (SJS) is a framework: a set of structural conditions on information flow in an autonomous-discovery system that prevent joint search from collapsing into self-confirmation while keeping the evaluator sealed. Conditions cover role decomposition, typed inter-role communication, provenance-sealed reads, versioned stores, and substrate-local promotion. Agora tests SJS empirically: five LLM agent classes communicate via three channels, evolving eight skill libraries, with alpha libraries built on AlphaGen operators. Three evaluators write reports aggregated into one brief, carrying forward disagreement instead of voting. We run Agora for 100 rounds on CSI 1000 and evaluate on a 91-day 2026 holdout sealed from all LLM inputs. Agora achieves holdout Sharpe +1.87; best baseline +1.334 at favorable seed and -0.755 cross-seed mean. Pre-loading Agora's two metrics into a frozen-library ablation recovers only +0.40 of the +2.25 Sharpe gap, and adding PPO without library evolution worsens the gap. The two metrics emerge rather than being designed. Caveats: single-seed run, short-side concentrated signal, intended for long-short.

LLM-based agents are reshaping microservice operations into AgentOps, where benchmarks are key to evaluating failure diagnosis over multimodal observability data. However, existing benchmarks remain largely outcome-oriented: they score only the final answer and fail to assess the systematic reasoning process in failure diagnosis. We address this gap by introducing two large-scale datasets (AIOps2025 and RCA100) under a reasoning-process evaluation paradigm that assesses agentic diagnostic capability along three dimensions: Localization (where the fault occurs), Identification (what type of fault it is), and Reason (whether the reasoning trace is grounded in relevant evidence). Together, the two datasets comprise over 500 expert-labeled failure cases across two representative microservice systems (HipsterShop and the OpenTelemetry Demo Store). They cover diverse fault scenarios across resource, network, runtime, middleware/database, and application-logic categories and provide fine-grained causal evidence to support agent learning and reasoning-process evaluation. Beyond scale and coverage, the datasets have been carefully labelled by domain experts and validated through large-scale competitions, supporting more than 6,000 participating teams. This makes them not only expert-labeled diagnostic datasets, but also competition-validated benchmarks for evaluating agentic failure diagnosis in real-world microservice environments. Datasets are available at https://www.aiops.cn/gitlab/aiops-live-benchmark/agenticopseval.

While Low-rank adaptation (LoRA) enables highly efficient fine-tuning by constraining task-specific updates to fixed low-rank subspaces, this rigid design limits representational flexibility and often results in overconfident predictions and miscalibrated uncertainty, especially in low-data regimes. Recent Bayesian LoRA variants improve uncertainty estimation by modeling posterior distributions over adaptation parameters. However, these approaches typically rely on fixed or heuristically determined ranks, overlooking the inherently context-dependent nature of adaptation capacity. In this paper, we propose BaRA, a Bayesian Adaptive Rank Allocation framework for parameter-efficient fine-tuning. Drawing inspiration from probabilistic topic models, BaRA dynamically allocates adaptation capacity by activating a sparse, context-dependent subset of disentangled latent factors, enabling instance-wise variation in effective rank. This Bayesian formulation provides principled, data-driven capacity control, mitigating over-parameterization while preserving expressiveness. Beyond the modeling contribution, we provide a complexity-theoretic generalization analysis showing that the generalization gap of BaRA depends on the learned joint effective rank $\bar{s}_{Φ,θ}$ induced by the global-local gate, rather than the maximum rank $r$. This result explains why sparse adaptive rank allocation can reduce the effective hypothesis complexity while preserving input-dependent expressiveness. Extensive experiments on diverse natural language benchmarks demonstrate that BaRA consistently improves predictive performance, robustness, and uncertainty calibration compared to standard LoRA and existing Bayesian LoRA variants.

Open-ended scientific discovery with large language models (LLMs) increasingly operates as a long-horizon loop of hypothesis search and verification, where a reward signal guides which hypotheses to test next. A notable recent example is AutoDiscovery, which uses "Bayesian surprise" - the belief shift an LLM undergoes after observing evidence for a hypothesis - as both a discovery metric and a reward for search. We first observe that AutoDiscovery treats surprisal as a static quantity, while surprisal in human reasoning is non-stationary - it is defined relative to beliefs that evolve with experience, a prerequisite for continual scientific discovery. We address this mismatch with evidence-informed LLM beliefs: priors updated with evidence from previous hypotheses to compute non-stationary surprisal for new hypotheses. We compare in-context belief-updating mechanisms and find that embedding-based retrieval-augmented generation over prior discoveries best anticipates eventual posteriors, identifying 37.5% of static surprisals as spurious. We then modify search to avoid these spurious rewards and prioritize hypotheses that remain surprising under non-stationary beliefs. Concretely, we introduce two complementary changes to the original search procedure: belief-update filtering and diversity maximization. Across five discovery domains, our method increases accumulated non-stationary surprisal by 30.62% on average compared to the original search procedure, demonstrating that continual scientific discovery with LLMs requires not only better belief measurement but also search procedures that avoid redundancy and encourage diversity.

Detecting and localizing defects in 3D point clouds is challenging because abnormal samples are scarce and diverse, while training is often limited to normal data. We propose Anomaly Factory 3D (AF3AD), a modular framework that synthesizes diverse pseudo-anomalies from normal point clouds to expand the training data for unsupervised 3D anomaly detection methods that rely on pseudo-anomalies. AF3AD uses a center-conditioned parametric deformation model defined in local PCA frames, with kernel-controlled spatial falloff, anisotropy, directional gating, and normal/tangential displacement fields, enabling a broad set of geometric defect presets. We demonstrate its ease-of-use and effectiveness by integrating AF3AD with an offset-prediction detector and a reconstruction-based anomaly detection method, showing that AF3AD transfers across detection paradigms. Experiments on AnomalyShapeNet and Real3D-AD show consistent improvements in object- and point-level detection and localization, supported by ablations on preset groups and robustness under noise. AF3AD is designed as a standalone synthesis tool to facilitate adoption across different 3D anomaly detection paradigms. Code is available at github.com/vpc-ccg/AF3AD.

A Knowledge Graph (KG) represents facts as structured triples and is widely used to organize relational knowledge across diverse domains. Just as textual information ranges from words and sentences to complete documents, KG information can be interpreted at multiple levels, from entities, relations, and triples to subgraphs and entire KGs. However, existing KG embedding methods mainly focus on entities, relations, and triples, leaving graph-level semantics largely unaddressed. Conventional graph-level methods, which typically compare graphs based on structural patterns, are also insufficient because structural similarity alone cannot guarantee semantic similarity between KGs. To evaluate how well different methods capture such graph-level semantic information, we study graph-to-graph semantic similarity, which determines whether a pair of KGs represents semantically corresponding underlying information. To obtain reliable ground-truth correspondences, we construct a semantic matching dataset by modifying text documents, extracting KGs from both original and modified documents, and transferring their known correspondences to KG pairs. We compare text-based, structure-based, and KG embedding-based approaches on each dataset. For the KG embedding-based approach, we introduce two scoring functions: \textit{EmbPairSim}, which uses maximal pairwise entity similarity, and \textit{AvgEmbSim}, which uses a frequency-weighted centroid. Experiments on WikiText-2 and CC-News show that \textit{EmbPairSim} achieves up to 5.3 pp higher MRR than Sentence-BERT while using substantially fewer parameters. These results suggest that KGE representations can serve as compact and effective signals for graph-to-graph semantic similarity in KGs. Our code is available at https://github.com/SeungRyeolBaek/KG-to-KG-Semantic-Similarity.

Video-based Point Cloud Compression (V-PCC) encodes volumetric data by projecting 3D geometry and texture onto 2D video frames. To prevent spatial distortion and color bleeding during 3D reconstruction, this process requires 10-bit color depth and 4:2:2 chroma subsampling, rather than the standard 8-bit 4:2:0 format. Additionally, capturing high-density dynamic point clouds requires demanding encoding parameters, such as 8K resolution at framerates up to 120 fps. Historically, the lack of 4:2:2 chroma support in older GPU hardware encoders restricted real-time V-PCC to custom Application-Specific Integrated Circuits (ASICs). However, the recent introduction of NVIDIA's Blackwell GPU architecture, featuring on-chip hardware encoders with 10-bit 4:2:2 support, presents an opportunity to shift this workload to general-purpose hardware. This paper investigates the feasibility of such an approach. Using a commercially available Blackwell GPU equipped with four parallel on-die hardware encoders as a testbed, we evaluate the throughput, rate-distortion (RD) performance, and power consumption of 8K 10-bit 4:2:2 HEVC across various Split-Frame Encoding (SFE) configurations. Our results demonstrate that 4-way SFE achieves an encoding throughput of 122 fps, successfully meeting the strict real-time constraints of high-density V-PCC. Although the inability to exploit spatial redundancies across slice boundaries results in a BD-Rate penalty of up to 5%, the measured throughput and power efficiency establish standard, commercial off-the-shelf GPUs as a highly viable baseline for real-time volumetric video streaming.

When does retention matter for memory-augmented LLM agents? We study this with TraceRetain, a lightweight framework for bounded external memory in frozen LLM agents that scores entries by interpretable features (success, age, access frequency, redundancy, specificity, similarity, downstream utility) and evicts the lowest-scoring ones at capacity. On clean ALFWorld with gpt-5-mini, external memory robustly improves over no memory across two seeds, but differences among bounded retention policies fall within Wilson 95% CIs: clean ALFWorld at T=100 to T=200 does not naturally exhibit the memory pollution retention is designed to address. Under a controlled noisy-write stress (75% synthetic distractors), unbounded memory and FIFO-K50 degrade on Precision@5 (20.2% to 12.4% and 15.8% to 3.8%) while TraceRetain-CEM is essentially unchanged (16.9% to 16.6%) and preserves 97/100 task success. The mechanism: unbounded memory has the highest mean similarity (0.87) but lowest precision, indicating failed distractors close to the query in embedding space. Held-out in-distribution evaluation shows memory-augmented policies solving 47 to 49 of 50 tasks vs. 39/50 for no memory. Bounded retention buys memory and step efficiency on saturated clean benchmarks at no task-success cost, and only differentiates from cache heuristics when streams contain noise.

A deep network's loss is invariant to continuous symmetries of its parameters: the logit shift, the ReLU rescaling, the LayerNorm scale, the per-head attention rotation. Adam's per-coordinate preconditioner drifts along each symmetry orbit, which pulls the trajectory off the symmetry quotient where the optimization lives and blurs the singular-learning rate the quotient makes readable. We build DDC, a Dead-Direction Conditioner that lifts a base optimizer into a $G$-equivariant one: it conditions the optimizer's state in the orbit decomposition of a $G$-invariant metric, so the trajectory stays a preconditioned gradient flow on the quotient $\barΘ= Θ/G$. The construction carries four architectural gauges (cross-entropy shift, ReLU and SwiGLU rescaling, LayerNorm and RMSNorm scale, and a per-head $O(d_{\rm head})$ attention rotation matched to RoPE), proves exactly equivariant on an Adam base, and composes with a Muon base through a gauge-equivariant orthogonaliser. Respecting the symmetry changes both the minimum the optimizer reaches and what it leaves measurable there. On a language model trained past the point of fit, DDCAdam resists the over-training collapse AdamW falls into, holding a validation-train loss gap of 0.67 against 5.88, and reads the dead-direction rate in 32 of 65 layer-by-observable cells where AdamW reads it in 7. A vision transformer trained from scratch reaches lower validation loss (1.71 against 2.12) while compressing spare feed-forward capacity a matched AdamW leaves intact. On a Muon base, where the rotation gauge composes exactly, DDCMuon groks ten of eleven seeds at depth 24 that a plain Muon never reaches. Built into the optimizer, a network's gauge symmetry sharpens the minimum it finds and turns that minimum's geometry into something the trajectory can measure.

International humanitarian law protects civilians from direct attack unless and for such time as they take direct part in hostilities, with the ICRC's 2009 Interpretive Guidance operationalising this rule through a three-criterion cumulative test. This paper argues that AI-mediated civilian cyber operations challenge the direct causation element of this test in a structurally specific way: when a civilian deploys an autonomous multi-agent cyber system of the kind recently demonstrated in offensive AI research, the "one causal step" standard fails because harm is produced by system-generated decisions made after human disengagement, and the integral-part requirement does not extend because it presupposes downstream human contributors whose conduct can be independently classified. The framework therefore defaults to treating such deployments as indirect participation, in tension with its purpose of capturing civilians who personally take part in hostilities. Beyond the doctrinal analysis, this paper identifies goal-specification granularity as the property on which the integral-part test's concreteness component implicitly turns, classifies AI-mediated operations along a five-level spectrum, and argues that existing technical AI governance instruments do not log or report this property.

While existing data attribution methods can identify which training examples build specific mechanistic circuits, they cannot explain how training data shapes the high-level behavioral decisions a model learns to make. To bridge this gap, we introduce Symbolic Mechanistic Data Attribution (SMDA), a framework that attributes training pairs to the interpretable symbolic policies governing model behavior. SMDA fits a closed-form Ridge regression over sparse autoencoder (SAE) features to model a target behavior, then analytically decomposes how each supervised fine-tuning example shifts that policy through feature-activation Delta_X and output-probability Delta_Y pathways. We distill a symbolic policy for refusal behavior in Llama-3.2-3B-Instruct and analyze 200 SFT training pairs. Our analysis reveals that (1) the symbolic policy's coefficients expose systematic gaps in the base model's safety behavior for categories like religious stereotyping; (2) per-feature Delta_X/Delta_Y decomposition can mechanistically explain why harmful and harmless pairs exert qualitatively different influences on certain features; and (3) individual training pairs routinely exhibit cross-feature interference, allowing SMDA to identify training pairs whose dominant effect falls on unintended features. These results demonstrate that combining mechanistic interpretability with data attribution yields a diagnostic tool that is both more fine-grained than black-box influence functions and more scalable than manual circuit analysis.

Many important games have more than two players and imperfect information. Existing approaches for computing Nash equilibrium, the central game-theoretic solution concept, in such games either lack scalability or obtain poor performance. In this paper we introduce a new algorithm called projected exploitability descent (PED) for approximating Nash equilibria in multiplayer games of imperfect information. The algorithm works by running projected subgradient descent minimizing a proxy for the multiplayer generalized exploitability function. The objective is nonconvex and nonsmooth, but can be represented as the sum of the maxima of linear functions, for which a subgradient can easily be computed and projected to the polytope of feasible sequence-form strategies. We explore performance of PED on a generalized version of the well-studied benchmark game three-player Kuhn poker. No prior exact algorithms scale to the version of the game with deck size larger than 4, and we compare performance to the popular algorithms of fictitious play (FP) and counterfactual regret minimization (CFR). We find that PED obtains a consistent near-monotonic improvement throughout all runs, though both FP and CFR perform significantly better in the initial iterations. This inspires a hybrid algorithm FP-PED that runs FP for an initial burn-in period before switching to PED for stable long-run refinement. We can alternatively view this as a multi-step algorithm that runs FP as a pre-processing step to obtain a strong initialization for PED.

Latent reasoning models perform multi-step inference directly in hidden-state space, yet the structure of these latent reasoning trajectories remains poorly understood. We show that contrastive refinement signals between stronger and weaker reasoning trajectories exhibit a highly concentrated low-rank structure, while unconstrained latent updates remain sensitive to paraphrases, checkpoint choice, and trajectory perturbations. These observations suggest that latent reasoning trajectories contain stable invariant directions mixed with unstable instance-specific variation. We introduce \textbf{Trajectory-Invariant Latent Refinement (TILR)}, a training-free intervention framework for identifying and manipulating stable reasoning directions in latent space. TILR first learns a low-rank invariant subspace from contrastive trajectory differences across inputs, then constrains latent interventions to this subspace while suppressing poorly aligned updates through an adaptive alignment gate. Across six reasoning benchmarks, we find that a small number of latent directions explain most variation between strong and weak reasoning trajectories. Interventions on these directions causally improve reasoning consistency and reduce trajectory instability under paraphrases and perturbations. TILR improves answer consistency under paraphrase by ~10% and reduces latent trajectory variance by up to $50\%$ while preserving reasoning accuracy. These results support a geometric view of latent reasoning in which transferable reasoning behavior emerges from stable low-dimensional structure within hidden-state trajectories.

Predicting tandem mass spectra (MS/MS) from molecular structures represents a central task in analytical chemistry with direct relevance to clinical metabolomics, systems biology, and adjacent disciplines. In this work, we revisit the problem through the lens of object detection on molecular graphs. Molecular fragmentation, a central step in MS/MS prediction, can be approximated as detecting a set of subgraphs (i.e., fragments) and their associated spectral contributions. Existing fragment-based models follow a two-stage paradigm -- first generating candidate fragments and then scoring them -- analogous to two-stage R-CNNs in computer vision. Towards higher accuracy and faster inference, we introduce GLACIER, a single-stage transformer-based fragment detection neural network for molecular graphs. This unified formulation eliminates the need for candidate enumeration, enabling scalable and globally consistent modeling of molecular fragmentation. GLACIER is faster and more accurate than existing state-of-the-art by a significant margin, achieving 70.0% and 69.7% Top-1 retrieval accuracy with and without contrastive finetuning on the MassSpecGym dataset (from the previous SOTA of 64.0%) and 52.5% and 38.5% respectively on the NIST'20 dataset (from 33.2%). Furthermore, GLACIER provides nearly 8-fold inference speedup over our prior two-stage model. Code is available at https://github.com/coleygroup/ms-pred

Offline root-cause-analysis (RCA) benchmarks commonly rank methods by a single pooled top-1 accuracy across multiple subsystems, and engineers often read the pooled winner as a recommendation for their own subsystem. We audit that reading on three public RCA benchmark families -- OpenRCA, RCAEval, and PetShop -- covering 11 subsystems and 778 matched scoring units. To keep pairwise comparisons on identical cases, the main analysis retains four methods or comparators with complete coverage: BARO, a CD-1min adapter, max-$|Z|$, and per-service alert-count. All six pairwise comparisons show subsystem-level effects of both signs, every random-effects 95\% prediction interval crosses zero, and case-level interaction tests reject exchangeability in 5 of 6 pairs. Leave-one-system-out selection picks the lower-scoring method on up to 5 of 11 held-out subsystems, with regret reaching 24.8 pp on RCAEval / Sock-Shop. We release a 320-line audit module; given a matched RCA benchmark score table, it recomputes the same per-subsystem stability checks alongside pooled scores.

Learning-rate transfer can reduce the cost of training large language models: instead of sweeping learning rates at target scale, practitioners extrapolate from smaller runs. Existing approaches often assume that the optimal learning rate follows a log-linear scaling law in data scale and model size. We carefully examine and evaluate this scaling law. In our empirical study of GPT-2--style models from 22M to 707M parameters trained on 5B to 100B tokens, the optimal learning rate develops upward curvature at larger scales, leading to inaccurate extrapolation. We find that this curvature largely disappears when learning rates are replaced by effective learning rate (the step size in normalized weight space), and when data $D$ extrapolation is used instead of model size $N$ extrapolation. Next, we explain nonlinearity in scaling: weight-norm converges to equilibrium slower when optimal learning is small, requiring a larger step size to reduce the transient phase. Experiments with AdamH, which directly controls the effective learning rate, further support this explanation.

Large Language Models (LLMs) have recently shown promise in automated binary analysis, yet they remain brittle under commercial-grade obfuscation. We present OASIF, an Obfuscation-Aware Self-evolving Instruction-Following framework for obfuscated assembly comprehension. OASIF couples a token-efficient assembly encoder with a lightweight projector to expose long obfuscated code to a pretrained code LLM under a bounded context budget and follows a three-phase training: (i) feature-space alignment, (ii) supervised instruction fine-tuning, and (iii) online self-evolving reinforcement learning with hybrid rewards, enabling continual adaptation with minimal manual verification. On VMISA-Bench, a challenging out-of-distribution suite featuring three commercial VM-based obfuscators, OASIF consistently improves open-source backbones; Qwen2.5-Coder-Instruct-14B attains Success Rate gains of +15.9, +5.8, and +16.9 percentage points (pp) on Code Virtualizer, Themida (v3.0.7), and VMProtect (v3.5), respectively, and improves the OASIF-Bench average by +9.8. OASIF further delivers stable gains across seven standard BCSD benchmarks while preserving general and domain-relevant capabilities on HumanEval, VulBench, and HumanEval-Decompile.

Discrete flow models have recently shown promising performance on few-step text generation; however, when naively applied to structured reasoning tasks such as Sudoku and Zebra puzzles, they converge confidently to incorrect answers (solving only $\sim$36% of Sudoku puzzles). We introduce Flow Reasoning Models (FRMs), a training and test-time-scaling framework for structured reasoning with flow models. We make the observation that, despite their poor solve rate, flow models can act as their own verifiers. A correct answer is a stable fixed point of the denoising dynamics, returning to itself when re-noised and re-solved. This enables a test-time-scaling paradigm: propose many candidate solutions and keep those that are dynamically stable, which alone reaches high solve rates on Sudoku-Shah (~$100\%$) and Zebra ($95.9\%$). This even generalizes to harder out-of-distribution puzzles like Sudoku-Extreme ($96.1\%$), without ever training on that distribution. This pure search, however, wastes a great deal of computation generating incorrect candidate solutions. We therefore design a training recipe to improve the base model's efficiency. First, we train flow models with a self-conditioning channel and close it at inference, letting them refine their own past predictions. Second, we train models to avoid their own failed generations using direct preference optimization. These changes substantially improve the base model's efficiency, letting it reach $99.2\%$ on Sudoku in just $7$ forward passes, over $8\times$ fewer than the strongest matched masked-diffusion baseline we compare needs for the same accuracy. When combined with test-time scaling, this lets flow models solve hard out-of-distribution puzzles (e.g. Sudoku-Extreme) far more efficiently.

Developing controllers capable of completing a wide range of tasks in a natural and life-like manner is a key challenge in enabling practical applications of physics-based character animation. In this work, we introduce Generative Pretrained Controllers (GPC), which leverage tokenization and next-token modeling to create general-purpose, reusable generative controllers from large-scale motion datasets. Our framework utilizes end-to-end reinforcement learning to jointly optimize a "motion vocabulary", modeled via Finite Scalar Quantization (FSQ), along with a corresponding control policy that can map the discrete codes to physics-based controls. After the "codebook" has been learned, the underlying structure of this large vocabulary is modeled by training a GPT-style autoregressive transformer, leading to a powerful generative controller that generates controls for a physically simulated character by performing next-token prediction. Once the generative controller has been trained, we propose a suite of adaptation techniques for finetuning the controller for new downstream tasks. Our proposed framework greatly simplifies the training process compared to previous tokenized methods, and achieves a 99.98% success rate in reproducing a vast corpus of motion clips. The generative controller exhibits a variety of natural emergent behaviors, such as responsive behaviors to perturbations and recovery behaviors after falling. This results in highly robust general purpose controllers for a variety of downstream applications.

Large language model agents are entering regulated financial systems, yet the security literature characterizing their attack surface is almost entirely laboratory-based, and the practitioner guidance on regulated deployment is neither peer-reviewed nor connected to a formal threat model. We bridge the two from production experience. We map six established agentic threat categories namely prompt injection, identity and authorization, action auditability, tool abuse, data residency, and boundary policy enforcement onto the specific control obligations imposed by the US and the EU financial regulation (ECOA and Regulation B, the EU AI Act, GDPR Article 22, and FINRA's 2026 agent guidance), showing how legal accountability amplifies each threat relative to an unregulated deployment. We then document four architectural patterns from a production Know Your Customer deployment for a consumer credit product (A2A compliance choreography, grounded-RAG-for-audit, case-ID propagation, and an inference-boundary redaction proxy) that moved a multi-day manual process to same-day automated resolution for roughly four in five cases. Finally, we report three negative results, including two control failures surfaced only by internal audit and a population of legitimate applicants the automated pipeline cannot serve. Securing agents under regulation, we conclude, is less about novel attack classes than about making auditability, least-privilege authorization, and boundary policy enforcement real at production scale -- requirements current agent frameworks leave to the deploying engineer.

We study how the next-token prediction of an autoregressive Transformer language model changes under small perturbations of earlier input token embeddings. Motivated by operator learning and iterative solvers for differential equations, we investigate how the influence of one token on another decays with distance in a trained model. In multilevel methods for differential equations, such as domain decomposition, multigrid, and multilevel preconditioning, one often exploits a separation between strong local interactions and weaker but essential global interactions. The latter correspond to the long tail of the Green's function and are typically handled by a coarse-level operator. Inspired by this perspective, we compute an empirical, distance-resolved gradient profile of token dependencies using autograd. Experiments on trained Pythia models and Qwen2.5-0.5B show that, over the measured distance range, the median Jacobian sensitivity is much better described by a power-law-type decay than by an exponential alternative: the diagonal-normalized profile is well described by $$\overline G(r) \approx γ+β(r+1)^{-p}$$ with exponents $p \approx 0.7$--$0.9$ (typically $0.8$--$0.9$). This behavior appears on coherent text from Gutenberg and WikiText-103. Token-shuffling experiments show that the power-law profile persists even when syntax and prediction quality collapse, whereas randomly initialized models do not exhibit it. The slowly decaying long-range sensitivity thus appears to be a learned property of trained autoregressive Transformer operators. These findings suggest that hierarchical or coarse-level mechanisms in language models may be able to exploit the long-tailed sensitivity profiles.

Event cameras capture sparse brightness changes with high temporal resolution and high dynamic range, compensating for the deficiencies of the conventional RGB frames. However, previous multi-modal fusion techniques typically fail to handle the inherent heterogeneity between RGB frames and event streams, thus easily leading to noise amplification or redundant feature integration during cross-modal fusion. In this paper, we propose a Cross-Modal information inTeraction transFormer, coined as CMTFormer, which hierarchically integrates RGB and event information to achieve efficient and stable multimodal collaboration. Specifically, we design a shallow-to-deep information interaction scheme. In the shallow stage, we present the Shallow Alignment Module (SAM) to achieve an efficient fusion of RGB and event low-level features, which mitigates attribute disparities and prevents noisy information. In the middle stage, we devise the Cross-modal Enhancement Module (CEM) that utilizes texture and edge information to produce mutually reinforced middle-level features. In the deep stage, we present the Learnable Deep Fusion Module (LDFM) which performs high-level information aggregation through learnable weights, thus enabling the network to adaptively fuse RGB and event clues. A Spatial Prior Module is further designed to utilize global spatial information to enhance localization accuracy. Extensive experiments are conducted on two prevalent event-based object detection benchmarks, i.e., DSEC-Detection and PKU-DAVIS-SOD. Our CMTFormer consistently surpasses the detection counterparts in both uni-modal and multi-modal settings, strongly demonstrating the effectiveness of our paradigm. Codes will be available upon publication.

We present DistilledGemma, an efficient and accurate system for the HIPE-2026 shared task on person-place relation extraction from multilingual historical newspaper articles in English, German, and French. Our approach adopts a three-stage knowledge distillation pipeline designed to balance classification accuracy with computational efficiency. In the first stage, we systematically explored prompt engineering strategies across eight large language models to identify the most effective reasoning architecture for this challenging task. In the second stage, we applied supervised fine-tuning (SFT) via QLoRA to a Gemma 4 26B A4B teacher model, leveraging its strong multilingual capabilities to generate silver-standard chain-of-thought traces across the training corpus. In the final stage, we performed response-level distillation to transfer these learned reasoning patterns into a compact Gemma 4 E2B student model. In the official evaluation, our team WHEREAMI ranked 3rd on the standard test set with an accuracy profile mean score of 0.688, and 2nd on the binary test set with a mean score of 0.8156. Notably, by distilling knowledge from the 26B teacher to the 2.3B student, we preserved strong reasoning capabilities while reducing the deployed model size to approximately 2.3B effective parameters; the LoRA adapters used during training were merged into the student for inference. This configuration ranked 2nd in the balanced efficiency-accuracy profile across both the standard and binary test sets. These results demonstrate that knowledge distillation provides a practical and scalable solution for historical document processing, achieving competitive performance without excessive computational cost.

Ozaki scheme II emulates high-precision matrix multiplication using low-precision integer matrix operations based on the Chinese remainder theorem (CRT). It first scales the high-precision matrices to convert them into integer matrices. For this scaling step, Ozaki scheme II provides two modes: accurate mode, which uses INT8 matrix multiplication to estimate scaling factors, and fast mode, which applies the Cauchy--Schwarz inequality at lower computational cost. We show that the existing formula lacks scale invariance; multiplying the input matrices by a constant changes the effective bit width of the integer matrices in the scaling step, causing accuracy degradation or CRT recovery failure. To address this, we propose a revised scaling formula derived from the CRT uniqueness condition via the Cauchy--Schwarz inequality. The proposed formula is scale-invariant by construction, guarantees that the CRT uniqueness condition is always satisfied, and introduces no additional overhead over the original fast mode. Experiments on an NVIDIA GH200 GPU show that the proposed method achieves accuracy comparable to that of accurate mode while maintaining throughput comparable to that of fast mode. In the accuracy--throughput trade-off, the proposed method overcomes the accuracy limitation of fast mode and the throughput constraint of accurate mode, offering a superior accuracy and performance.

Cooperative multi-agent reinforcement learning (MARL) often relies on communication to mitigate partial observability, yet most existing protocols treat messages as flat dense vectors detached from the structure of the observations they summarize. This design overlooks an important source of inductive bias in many cooperative environments, where observations naturally follow a hierarchy such as groups and entities. We propose \textsc{HiComm}, a plug-in communication module that grounds messages in the sender's hierarchical observation. \textsc{HiComm} is receiver-driven: the receiver issues a query, and the hierarchy is resolved through a three-stage decoding process that first selects a group, then a sender, and then an entity within that group, returning the corresponding feature slice as the message. This converts communication from unstructured vector transmission into structured information retrieval over the sender's observation hierarchy. We instantiate this mechanism with Straight-Through Gumbel-Softmax for differentiable discrete selection and a lightweight shared projection design that attaches to standard MARL pipelines. Experiments across cooperative MARL tasks with different observation structures and coordination demands show that \textsc{HiComm} matches or outperforms representative learned communication baselines while reducing communication volume by up to $23\times$ per receiver per episode.

Public discourse about artificial intelligence (AI) often uses anthropomorphic language: language that attributes human capabilities and characteristics to the system. This practice has been criticized for setting misleading expectations, inflating claims, and fueling hype around AI, which may distort public understanding of AI and impact policy priorities. We study the effects of anthropomorphic framing by comparing changes in participants' perceptions (N=815) when reading passages with and without anthropomorphic language, designed to reflect realistic public-facing AI discourse. We further examine whether these effects differ across two types of AI technologies -- large language models and recommendation systems -- and measure changes in perceptions of AI across several dimensions that are prominent in current public discourse. In a separate condition using a text that explicitly discusses the dangers of AI, we show that individuals' views of AI can shift in response to reading a text; yet in the main conditions of the experiment, where we compare anthropomorphic and non-anthropomorphic descriptions, we find that whether the text uses anthropomorphic language does not substantially affect participants' perceptions of AI. Our results indicate that any immediate effects on public opinions of AI are modest, although they leave open the possibility that anthropomorphic language could have an effect in naturalistic settings, or over gradual, continued exposure.

We introduce a pre-registered screening rule that decides, before any implementation, whether an evolutionary / population / lifecycle outer loop over neural-network parameters or structure is worth building. Such outer loops cost 10^2-10^3x their gradient inner loop, yet whether they beat a cheap single-shot alternative is usually discovered only after the expense is paid. Our rule computes, at a Phase-0 gate, a single number: the recovery R = s/G, the best single-shot gradient/curvature statistic's gain s divided by the best gain G of any cheap method evaluated, and prescribes skipping the outer loop when R >= 90%. We validate the rule on a within-lab series of pre-registered outer-loop bets (two analyzed cases plus a disclosed file drawer): in both analyzed cases a static or single-shot computation captured the effect on the project's own metric, the gate fired (R approximately 1.0 in both cases; approximately 0.95 under a stricter metric on one), and the outer loop was abandoned, including one case where a companion factorial decomposition localizes the apparent win to a static substrate change with the evolutionary lifecycle contributing no detectable gain. On one project the gate cost about 50-70 GPU-hours and screened out an estimated 400+ GPU-hours (first cell only) plus weeks of implementation, a 6-8x saving. The rule is prospectively falsifiable: a task with R < 90% where the outer loop still fails to beat single-shot would refute it.

Graph-based narrative extraction relies on a coherence function to score transitions between events, but the coherence metrics in current use are defined operationally and lack an information-theoretic foundation. We study the composite metric $C=\sqrt{A\cdot T}$, where $A$ is the angular similarity of document embeddings and $T=1-d_{\mathrm{JS}}$ is a topic proximity from the Jensen-Shannon distance of soft memberships, and give it an information-geometric reading together with an axiomatic characterization of the geometric-mean combinator. On the product manifold $\mathbb{S}^{d-1}\timesΔ^{K-1}$, the negative log-coherence decomposes additively into an angular and a topic cost. Because the Riemannian metric tensor induced by the Jensen-Shannon distance on the simplex is proportional to the Fisher information matrix, the topic component is locally consistent with the Fisher-Rao metric singled out by Chentsov's theorem. Within the compensability spectrum of combinators, the geometric mean is the unique one consistent with four natural axioms (a boundary/veto condition, symmetry, log-additivity, normalization), and the construction motivates a proper product metric $d_\times$. Experiments on four corpora, three embedding families, and three topic models are consistent with the framework: the Fisher identity holds ($R\ge0.99$), the geometric mean tracks $d_\times$ closely ($ρ=0.999$), and a downstream LLM-as-judge check finds it is not dominated by any alternative combinator or single-channel baseline. Sweeping the spectrum, the bottleneck-coherence gap between extracted and random storylines splits into a symmetric component, maximized at the geometric mean across five corpora, and a displacement term; a cross-modal image-narrative case study reproduces the effect. These results justify the composite coherence metric and articulate when the geometric mean is the natural choice.

Post-hurricane damage assessment and repair scheduling can require computationally intensive simulation and optimization. This paper presents an integrated two-stage deep-learning tool for rapid damaged-line identification and repair-schedule computation. An available offline synthetic dataset for the IEEE 9500-node test feeder contains 1,700 hurricane scenarios with exposure features, grid metadata, fragility parameters, OpenDSS outputs, damaged-line labels, and Adaptive Large Neighborhood Search reference schedules. Stage 1 benchmarks MLP, ResMLP, and GraphSAGE, while Stage 2 compares MLP, DeepSets, and Set Transformer. The selected ResMLP-Set Transformer pipeline propagates Stage 1 errors into Stage 2 and achieves a damaged-job F1-score of 0.920, pairwise order agreement of 0.854, and start- and end-time mean absolute errors of 4.349 min and 4.486 min, respectively. The tool provides rapid initial repair-log decision support for new hurricane cases.

Large Language Models (LLMs) are rapidly being adopted in low-code and no-code automation platforms, where non-expert users design workflows that combine natural language understanding with external services and APIs. LLM agents are LLM systems that use LLMs as a core "brain" to reason, plan, and autonomously execute complex, multi-step tasks. In this paper, we present the first large-scale empirical study of LLM agentic workflows in low-code automation platforms. We analyze more than 6,000 publicly available n8n workflows and examine four aspects of their design: task distribution, structural and tool use patterns, reliability mechanisms, and autonomy levels. Our analysis shows that LLM workflows are not merely prompt response pipelines. Instead, LLMs are commonly embedded within broader automation structures involving control logic, external tools, communication services, storage systems, and human review points. We further find that while many workflows include lightweight post-processing or routing logic after LLM execution, explicit reliability mechanisms such as structured fallback paths, repair loops, failure-specific alerts, and human approval gates remain relatively uncommon. These results reveal a gap between the increasing deployment of LLM agents in practical automation ecosystems and the limited engineering support for reliability, safety, and governance. Overall, our study provides ten empirical findings and five research takeaways for researchers, platform developers, and practitioners seeking to understand and improve real-world LLM agentic workflows.

Autonomous vehicles (AVs) are increasingly deployed in urban environments, yet their safety frameworks remain primarily designed around collision avoidance and minimal risk condition (MRC) behaviors such as slowing or stopping when uncertainty arises. Although effective in reducing immediate crash risk, real-world deployments indicate that stopping alone does not guarantee safe integration into human-governed roadway systems. Incidents reported by municipalities and public records show that AV fallback behaviors can obstruct traffic, interfere with emergency response operations, and create accessibility challenges for passengers and pedestrians. This paper presents an analysis of publicly documented incidents involving AV stopping behavior and human-AV interaction failures. We categorize these incidents according to limitations in perception, planning, and control within current AV architectures. Using this taxonomy, we identify key gaps in existing safety paradigms, particularly the lack of mechanisms for interpreting human authority, responding to multimodal instructions, and adapting to dynamic, socially regulated traffic conditions. We then review emerging research directions that support human-interactive perception, language-grounded and accessibility-aware planning, and assisted control through remote guidance and teleoperation. The analysis highlights the need to augment current AV safety frameworks with capabilities that enable cooperative interaction with human agents and infrastructure. These findings suggest that reliable urban deployment of AVs requires moving beyond passive fallback strategies toward human-interactive autonomy.

Large language models (LLMs) increasingly mediate strategic interactions through natural language, making semantic control a critical element of communication and deception. This paper develops a semantic signaling game in which a sender selects a semantic control, an LLM generates a stochastic message, and a receiver evaluates the message using an awareness-dependent scoring mechanism. Receiver awareness is modeled as a type that determines which linguistic features are perceived and used for inference, providing a formal model of systematic blindness. The framework connects prompt-based control, statistical detection, and game-theoretic equilibrium analysis. Gaussian approximations of aggregate message scores enable likelihood-ratio decision rules, while Perfect Bayesian Nash equilibria characterize strategic behavior. The paper further develops mechanism-design approaches that reshape receiver awareness, penalize deceptive semantic controls, and modify receiver populations to induce benign pooling equilibria. Numerical experiments validate the Gaussian approximation, quantify awareness-ordering effects, analyze mindset dynamics under adaptive adversaries, and demonstrate how awareness shaping and guardrail costs reduce successful phishing attacks. The proposed framework provides a principled foundation for analyzing strategic language-mediated interactions in agentic AI systems and offers new tools for the design of robust and secure human-AI communication.

Deep learning, which in general relies on voluminous amounts of training data, is vulnerable to data poisoning attacks, including error-generic attacks and backdoors (Trojans). In this work, we propose a new data poisoning attack we dub a latent class attack. Here, all poisoned examples are from a class that is novel (unknown) for the given classification domain and are mislabeled to one of the known classes (the target class) of the domain, so that the model learns to recognize the novel class as a sub-class of the target class. Such attacks could be used e.g. to defeat AI-based access control systems, or could cause a "foe" to be classified as a "friend". We also propose a post-training defense to detect this attack, without any access to the training set. This detection approach builds on "class subspace orthogonalization" (CSO), a plug-and-play paradigm demonstrated to improve existing backdoor detectors. Here, CSO is used to seek an input (a putative unknown class instance) whose internal representation is not aligned with any of the known classes, and yet which is classified with confidence to one of these classes. Finally, specific to image classification domains, we propose a method for visualizing the estimated unknown class instance, providing explainability to our latent class detections.

When firms deploy autonomous AI, they must decide how much work to leave to the system and how much to keep workers engaged. This decision affects current output and future human capital. We develop a parsimonious two-period model in which AI may outperform the worker when it functions, but may fail with positive probability. A firm chooses worker engagement; engagement lowers current output for below-benchmark workers, but changes future skill through learning and erosion. We distinguish two dimensions of AI progress: capability, the system's output when it works, and reliability, the probability that it works. In a single-firm benchmark, engagement is valuable only as fallback investment. The firm engages the least-skilled workers most, because they have the largest skill gaps and are least costly to bring toward a useful fallback level. With worker mobility, engagement also affects labor-market sorting: workers prefer jobs that build more valuable skill trajectories. This sorting motive targets higher-skill workers near the AI frontier, where skill gains are more valuable and engagement is less costly. Mobility can therefore reverse the engagement pattern, shifting investment from the least-skilled toward the most-skilled workers below the AI benchmark. Mobility also reshapes how AI progress affects engagement: greater capability raises engagement by increasing the value of the skill trajectory a firm offers, whereas greater reliability can raise or lower it because it reduces fallback need while also changing learning opportunities. Under worker mobility, human-AI work design becomes a problem of human-capital investment, in which allocating work today shapes future skill.

Recent progress in flow-based generative modeling has led to models that output high-quality samples while using only a small number of function evaluations. However, at present, there is a lack of similar advances in estimating the model likelihood. In particular, most existing methods either rely on restrictive architectures that enable exact calculations, or use stochastic approximations such as Hutchinson's trace estimator that introduce substantial variance. In this work, we introduce SCAlable LikeLihood distillation of flOw maPs (SCALLOP). SCALLOP builds on the recently proposed F2D2, a likelihood flow map model that can generate samples and their densities in a small number of function evaluations. While F2D2 uses Hutchinson's estimator during training, we introduce an alternative and more scalable likelihood distillation objective that is Hutchinson-free and admits a vectorized formulation. Empirically, we demonstrate the effectiveness of SCALLOP as a Boltzmann generator in molecular science, and further validate its benefit on image datasets. SCALLOP significantly reduces both training variance and training time while consistently improving performance compared to F2D2, and is competitive with the state-of-the-art while achieving up to 10x inference speedup over the fastest baseline.

Symbolic execution is a powerful program analysis technique with broad applications, such as vulnerability detection, security testing, and malware analysis. However, this technique is known to suffer from scalability issues, e.g., path explosion, complex constraints, due to certain structural and semantic patterns commonly presented in real-world programs. Existing approaches attempt to escape these patterns by transforming programs into new representations to reduce the execution cost. Unfortunately, these transformations are often too rigid to exploit diverse local program semantics and sometimes rely on compiler optimizations designed for concrete execution that may misalign with the goals of symbolic execution. We present Symbolon, a framework that automatically learns diverse code transformations and applies them context-sensitively to improve symbolic execution. Our key insight is to formulate transformation discovery as a search problem over program representations. To make the search practical, Symbolon learns transformations cheaply offline on small programs, distills them into a reusable library of agent skills, and uses an agent to instantiate these skills on repo-level targets. Our evaluation shows that Symbolon substantially improves the symbolic execution engine KLEE across 16 search strategies on 32 real-world programs, increasing line coverage by 3.69x on average while reducing peak memory and per-query solver time by 29.2x and 123x, respectively. When applied to the latest Linux kernel, Symbolon uncovers 21 previously unknown bugs, all of which have been reported to the kernel maintainers.

Neurological disorders involve diverse pathologies of the brain and nervous system, making early and accurate detection essential. While many deep CNNs have been developed for MRI-based classification of neurological disorders, most are optimized for binary tasks and often fail to capture the multi-class features needed to distinguish subtle anatomical differences across conditions. This study proposes the Enhanced Neurological Disorder Detection Network (End-Net) for multi-class MRI classification of neurological disorders. End-Net includes 24 convolutional layers, beginning with convolutional blocks followed by 21 optimized inception modules. These modules extract multiscale features via parallel 1 x 1, 3 x 3, and factorized 5 x 5 convolutional branches, along with max pooling, enabling the model to capture complementary texture, edge, shape, and contextual information. A global average pooling head, compact fully connected classifier, and dropout reduce parameters, limit overfitting, and improve robustness. End-Net was evaluated on the Multi-Class Neurological Disorder dataset, comprising MRI scans from patients with Alzheimer's disease, brain tumors, multiple sclerosis, and healthy controls. Severe class imbalance was addressed by augmenting minority classes with WGAN-GP and randomly undersampling the majority class. The results show that End-Net outperforms existing architectures in both accuracy and generalization. The model is also integrated into an online system for real-time web-based inference and accessibility.

Jointly learning to segment and classify medical images demands cross-task synergy, yet encoder-sharing architectures limit decoder reconstruction to task-private representations, permanently discarding the boundary cues and semantic priors each branch could supply to the other. This work introduces BTI-Net, which establishes bidirectional communication at every decoder level through two parallel pathways via Task Interaction Modules (TIM). Spatial boundary context is gated into the classification branch, while global semantic priors multiplicatively modulate the decoder, with refined features propagating progressively from coarse semantics to fine boundary detail across all four decoder resolutions. Since cross-task interaction is not equally reliable for every input, Uncertainty Proxy Attention (UPA) gates each TIM output per instance and per level using three signals that capture cross-task alignment, scene complexity, and prediction confidence, without external annotations or additional inference passes. Experiments on three medical benchmarks spanning ultrasound, dermoscopy, and brain MRI demonstrate consistent improvements in segmentation IoU and classification accuracy over both encoder-sharing and decoder-interaction baselines. Ablation confirms adaptive gating contributes +2.36 IoU over fixed bidirectional interaction, and classification accuracy improves by up to +2.26 points over the strongest multi-task baseline. UPA's uncertainty proxies serve as reliable single-pass task-failure signals without the overhead of stochastic sampling. Code: https://github.com/C-loud-Nine/BTI-Net_MTL

Artificial neural networks (ANN) provide accurate continuous-valued representation, whereas spiking neural networks (SNN) offer event-driven temporal processing, yet both paradigms face limitations when value encoding and timing dynamics must be learned within a single computational structure. This paper introduces a network based on Unified Complex-valued Neuron (UCN), a new neural computational model that integrates continuous activation and phase-driven event generation through an asymmetric complex-valued state. In the UCN, magnitude encodes signal strength while phase governs intrinsic temporal evolution and valued spike emission. A foundational training framework combining backpropagation (BP) and backpropagation through time (BPTT) is first developed to optimize magnitude and phase pathways in a unified way. To reduce computational complexity, an event-driven adaptive phase learning (EAPL) rule is then introduced as a more efficient alternative. The proposed model is evaluated through object tracking and Lorenz attractor learning. Results demonstrate that UCN-based Network (UCNN) provides accurate, stable, and interpretable spatiotemporal learning while preserving sparse event-driven computation for neuromorphic and edge-AI applications.

A growing number of techniques leverage the spatial structures that underlie many real-world datasets. Despite these advances, the complementary task of estimating spatial structures and understanding their role within these techniques has often been overlooked. In neurophysiological data analysis specifically, numerous methods exist to estimate brain connectivity, but most are not explicitly model-based, dynamic, multivariate, or directed. To address these limitations, we previously introduced noise-driven heat modelling on graphs for neurophysiological connectivity estimation. In this study, we extend this framework by relaxing earlier noise assumptions and adding regularisation to improve robustness. We also develop a simulation procedure to characterise and evaluate our technique in a controlled setting. Finally, we demonstrate that the technique is able to capture meaningful spatial structure across two experiments, each using two real-world datasets. The explicit model formulation of our connectivity estimator has the potential to improve the interpretability of graph-based techniques across a wide range of applications. The code implementing our method is available at https://github.com/sgoerttler/Heat_Connectivity.

Diffusion-based video relighting enables controllable relighting from a single input video, but modern video diffusion backbones are trained on short clips and applied to long-horizon videos through chunked sliding-window inference, often causing temporal discontinuities at chunk boundaries. We address this by reframing long-horizon relighting as \emph{temporally conditioned latent domain translation}. Our framework enforces cross-chunk continuity by propagating target-domain latents across boundaries and makes this behavior learnable using \emph{masked target-domain self-conditioning}, training the model to continue from temporally masked propagated context. We further introduce \emph{warm-start prompting} with a relit prompt anchor from a controllable generative model, which establishes the initial target-domain state and creates a general interface for prompt-based relighting. Experiments on in-the-wild long-horizon videos show markedly improved temporal consistency, with chunk-boundary artifacts largely reduced and unwanted appearance changes across chunks greatly suppressed.

Diffusion language models (DLMs) have recently emerged as a promising alternative to conventional autoregressive language models. By generating multiple tokens in parallel during each denoising step, they offer higher inference throughput while maintaining competitive quality. However, realizing these throughput gains while meeting latency SLOs in a serving system requires addressing challenges introduced by DLMs' unique characteristics. These include navigating the speed-quality tradeoff created by confidence-based denoising, choosing appropriate parallelization levels across model instances under fluctuating load, and coordinating approximate KV caching mechanisms that introduce non-uniform per-step costs. To address these challenges, we present DiLaServe, a cluster-level serving system for DLMs. DiLaServe enables deadline-aware scheduling and adaptive load control through confidence-threshold adjustment, and dynamically reconfigures the cluster by solving a quality-aware optimization problem, while explicitly modeling the step-level heterogeneity introduced by approximate KV caching. Across multiple benchmarks and real-world traces, DiLaServe improves SLO attainment by up to 56.6 percentage points and reduces end-to-end request latency by up to 46\% while incurring less than 1\% accuracy drop.

In a changing decision problem, standard dynamic-regret analyses have often equated the cost of non-stationarity to how far loss moves. However, it is simultaneously possible for a loss sequence to travel far and retain the same optimal policy, or for a small movement in loss to force the optimal policy to change completely. Thus, the size of the movement through loss variation, transition variation, or comparator path length describe the adversary's motion, but not the cost of that motion to the control problem. For a more faithful analytic interpretation, this paper develops a normal-fan geometry for finite-horizon adversarial MDPs with fixed transitions. Occupancy measures form a polytope, and each loss vector exposes an optimal face of that polytope. Non-stationarity in rewards is therefore a path through the normal fan, where motion inside one cone leaves the optimal face unchanged, while crossing a wall may carry regret. We pose the notion of a face-crossing price, which is the minimum regret incurred by remaining on the previous optimal face under the new loss. For any learner that tracks the previous face, dynamic regret decomposes exactly into intrinsic priced face motion plus within-face selection error. The resulting theory separates consequential from harmless non-stationarity, where loss variation can be arbitrarily large at zero price, and identical one-coordinate variation can hide horizon-scale differences in regret.

Tabular foundation models cannot reason about data produced by running systems without access to the rules that govern them. We make this statement falsifiable. The \emph{Operational Turing Test} (OTT) constructs pairs of legal and rule-violating database states whose $1$- and $2$-way column-value marginals match to a total variation of $<0.02$; Le~Cam's lemma then bounds any values-only classifier at $\geq0.49$ Bayes error. Three values-only baselines (XGBoost, TabICL, TabPFN) hit the bound exactly (accuracy $0.50$, pre-registered two one-sided tests (TOST) $p<0.002$), raw row-level access does not help, exposing relational value consistency closes most of the gap, and only a classifier fed by seven executable rule-derived audits reaches $1.00$ classification accuracy. In three matched $100$-state frontier large-language-model (LLM) runs, models given the schema, trigger source, rule tables, and state files classify at most $2/50$ legal states as LEGAL; GPT-5.5 accepts $0/50$ legal states even with higher reasoning effort and a Structured Query Language (SQL) executor. The access-ladder pattern also appears on a second schema with structurally distinct rule families (banking ledger: cross-row balance, cumulative aggregate). The barrier is identifiability, not capacity: scale, data, and richer features cannot cross it without operational grounding.

Retrieval-Augmented Generation (RAG) has become the standard way to ground large language models in external knowledge, yet most systems retrieve a fixed number of passages for every question regardless of its difficulty. This wastes computation on easy questions, starves hard ones, and gives no signal for when a generated answer can be trusted. With a growing share of question answering systems built on top of commercial language model APIs, a method that can decide how much to retrieve, and how far to trust its own answers, without retraining the underlying model, is of clear practical value. This paper presents AB-RAG (Adaptive Budgeted Retrieval-Augmented Generation), a training-free and backbone-agnostic framework that generates an answer, estimates its confidence from a combination of three signals, and then decides whether to stop or to retrieve more evidence, subject to a fixed retrieval budget. The estimator combines the model's own certainty, the agreement between the answer and the evidence, and the variance of the retrieval scores. For models that expose token probabilities the certainty signal is read directly; for closed APIs it is approximated by self-consistency, so the method works without access to model internals. Across three backbones and two datasets, the central result is that the confidence estimate reliably separates correct from incorrect answers on every backbone, reaching a clean split of 57.6% against 0% Exact Match between high- and low-confidence answers on a factoid dataset. The adaptive policy improves accuracy on capable backbones, and the study reports its negative and nuanced findings honestly, including a confidence signal that proved unsuitable for short answers and a retrieval signal whose sign was found and corrected through measurement. The entire study was carried out on a single consumer laptop with only a few dollars of API spend.

There are various benchmarks to evaluate bugfixing capabilities of Large Language Models. However, most widespread benchmarks do not fully reflect real-world bugfixing practices. They are small, weakening statistical reliability, and the buggy programs are often similar to one another, potentially distorting evaluation results. The range of bug types can also be narrow, failing to capture a representative range of bugs. To address these issues, we introduce MegaBugFix, a large-scale bugfixing benchmark containing 12,629 buggy Python programs synthesized from correct ones by a Large Language Model. Bug injections were generated as diffs representing code changes. Through this approach, we were able to avoid common pitfalls of LLM-based mutation techniques like injecting overly simplistic bugs or failing to modify the input program. We evaluated 13 open-weight models on MegaBugFix and baseline benchmarks, finding consistently lower performance on MegaBugFix. This reveals that our benchmark presents more challenging bugs and exposes model failures that may remain hidden when evaluating on existing benchmarks. The benchmark and fine-tuned model used for bug injection are available at hf.co/collections/szalontaib/megabugfix.

The carbonized papyri from Herculaneum preserve the only large-scale library to survive from classical antiquity, but many unopened rolls remain unread because physical opening risks irreversible damage. X-ray computed microtomography ($μ$CT) and virtual unwrapping offer a non-invasive route to their texts, yet previous work on sealed Herculaneum scrolls has recovered only localized readings or limited surface regions. Here, using high-resolution phase-contrast $μ$CT acquired on the BM18 beamline at the European Synchrotron Radiation Facility (ESRF), together with improved computational unrolling and machine learning, we achieve the complete virtual unwrapping and reading of PHerc. 1667 under explicit coverage and papyrological-review criteria. This makes PHerc. 1667 the first Herculaneum papyrus to be fully digitally unrolled and read for extended scholarly study without physical opening. In PHerc. Paris 4, the optimized scan protocol makes ink directly visible in the tomographic volume, allowing three-dimensional ink segmentation and independent validation of surface-conditioned ink recovery. In PHerc. 139, we recover title and author-attribution evidence identifying the scroll as Philodemus, On Gods, Book 8. These results move virtual unwrapping of the Herculaneum scrolls beyond isolated demonstrations towards a scalable framework for systematic recovery of the still-unopened library.

Koopman theory promises linear structure in nonlinear dynamics, but numerical Koopman spectra are easy to compute and hard to trust. A finite EDMD matrix always has eigenvalues; the problem is that many of them may have nothing to do with the infinite-dimensional operator. In this paper we make spectral reliability the objective of dictionary learning. We train neural-network dictionaries not merely to predict the next snapshot, but to minimize Residual Dynamic Mode Decomposition residuals: operator-level a posteriori errors that test whether computed eigenvalues and modes are genuine Koopman spectral objects. To keep the learned observables from collapsing into an unstable coordinate system, the loss also penalizes the condition number of the lifted data matrix. Thus the method couples two requirements that should not be separated: small Koopman residuals and a well-conditioned representation. The result is a learned dictionary that is expressive, numerically stable, and spectrally disciplined. Across conservative and dissipative benchmark systems, the method sharply reduces spectral pollution, improves residual pseudospectral inclusion, and lowers forecast error relative to standard fixed dictionaries. On sea-surface temperature data, it gives cleaner Koopman diagnostics and substantially better one-step forecasts from noisy observations with no governing equations. The message is simple: neural Koopman learning should be judged not by prediction alone, but by whether its spectral claims can be certified. Residuals provide the certificate; conditioning makes it computable.

Would experience designing faster GPU kernels also help close in on a long-standing open mathematical conjecture? Large Language Models (LLMs) integrated into evolutionary search have recently produced state-of-the-art solutions on optimization tasks, including open mathematical conjectures, GPU kernel design, scientific law discovery, and combinatorial puzzles. To achieve this, prior work applied search scaffolds to one target task at a time, so every new problem is approached from scratch and the experience accumulated during search is discarded once the model finishes its attempt. This leaves the capability of iteratively evolving a solution (e.g., knowing which part to mutate and how, deciding when to backtrack) entirely in the scaffold rather than in the model itself. Whether the model itself could acquire this capability and reuse it across different tasks has been largely unexamined. To address this, we introduce Evolution Fine-Tuning (EFT), a mid-training paradigm that teaches LLMs to evolve solutions across tasks by converting evolutionary search trajectories into supervision. We construct Finch Collection, a 156K-trajectory dataset spanning 10 domains and 371 optimization tasks, and fine-tune open-source LLMs from 2B to 9B parameters. Empirically, EFT confers cross-task generalization: across 22 held-out tasks, our models surpass their base counterparts by 10.22% on average. Furthermore, when paired with test-time RL, our model matches state-of-the-art performance on two circle-packing tasks and outperforms its base-model counterpart on the Erdős minimum-overlap problem. EFT thus serves as a "practice phase" for general-purpose discovery agents that do not solve new problems from scratch.

Based on the Backblaze hard disk drive (HDD) dataset, we analyze whether the four major HDD manufacturers represented in the dataset -- HGST, Seagate, Toshiba, Western Digital (WD) -- show differences in short- to medium-term HDD failure rates. Using two different duration regression models, we find -- holding constant drive age, capacity, form-factor, and drive temperature -- that Toshiba's failure rate is slightly above Seagate's. HGST HDD failure rates are the lowest, about 41% of Seagate's. WD HDD failure rates are significantly above HGST's, but still only about 52% of Seagate's. We also document the effects of age, capacity, temperature and drive location on failure rates.

Model Context Protocol (MCP)-style ecosystems give language-model applications a practical connection layer for tools, resources, prompts, and transports. As agents move from connection to execution, security decisions often remain split across clients, servers, prompts, approval dialogs, OAuth deployments, and logs. This paper asks whether a runtime can make execution-layer invariants explicit and testable while preserving MCP-like workflows. We define eight invariants: metadata non-authority, grant-backed approval, canonical resources, principal binding, scoped capability invocation, source-and-target data-flow authorization, deny-path audit, and explicit protocol state. We implement these invariants in HCP, a Handle-Capability Protocol reference runtime for MCP-style agent execution that represents calls through principals, resources, grants, capabilities, handles, policy decisions, data-pipe checks, and audit entries. We evaluate HCP against two MCP-like baselines: a naive connection-layer runtime and a practice-informed connection-layer mitigation baseline with metadata linting, session checks, and per-call approvals. Across 10 benchmark cases, the naive baseline permits all modeled attacks, the mitigation baseline permits 6 of 10, and HCP blocks all 10 while preserving audit evidence. Ablations identify which runtime components block attacks and preserve forensic evidence. A local in-memory microbenchmark reports sub-millisecond mean latencies for measured policy, invocation, peek, and pipe operations. A bounded GitHub README-screening sample provides ecosystem signals, not vulnerability findings. The results support a narrow claim: MCP-style agent systems need an execution-control layer in addition to connection-layer conventions.

Concept-based Explainable AI (C-XAI) seeks human-understandable explanations grounded in semantic concepts, yet validation is limited by the scarcity of fine-grained concept annotations. We evaluate whether mid-scale Multimodal Large Language Models (MLLMs) can perform localized concept naming under strict zero-shot conditions by assigning labels to bounding-box regions at both object and part levels. We propose a reproducible zero-shot evaluation protocol for Concept Naming (CoNa) with (i) closed-set, category-constrained prompting for moderate vocabularies and (ii) Open-CoNa, an embedding-similarity-based strategy for large label spaces. Experiments with four MLLMs (7B-32B) show consistent performance trends across datasets, reaching 62%-88% object-level exact-match accuracy, highlighting the potential of training-free concept annotation from localized regions. We discuss limitations and failure modes and release a reproducible framework to support future low-cost C-XAI research.

Text encoders are known for their utility in natural language processing, as they are able to efficiently compress inputs into dense vectors while preserving semantics. These models have been applied to affective computing, in particular to help with solving sentiment analysis and emotion recognition tasks. Nevertheless, it remains unclear to what extent the latent representations produced by modern text encoders capture well-defined psychological theories of affect. In this work, we investigate the affective capabilities of twelve recently released text encoders by probing their generated embeddings as input features for solving regression and classification tasks across three established emotion frameworks, using both word- and sentence-level data. Additionally, we apply a semantic data-leakage prevention technique to improve robustness in word-level evaluations. Our main findings show that the latent manifolds of the latest instruction-aware open-weight encoders enclose an equal or even a larger amount of affective information in comparison with proprietary counterparts when evaluated at word level. In contrast, embeddings of task-tuned and proprietary encoders reach the highest scores on sentence-level affective classification. Furthermore, a qualitative analysis of latent representations and their encoded affective cues is provided.

We present ThinkProbe, a framework for structural analysis of LLM reasoning traces. ThinkProbe converts each trace into a Thought Graph a directed graph with cycles, 8 node types, and 6 edge types and derives a 19-metric five-dimensional cognitive profile (5D-CP: Breadth, Depth, Structure, Metacognitive, Efficiency) through a fully non-generative pipeline combining rule-based segmentation and discriminative semantic linking. Applied to 4{,}200 traces from 7 native reasoning models across 200 open-ended questions and 10 cognitive domains, ThinkProbe reveals that reasoning structure is a stable, model-level property: between-model variance exceeds between-domain variance by up to fourfold across four of five cognitive dimensions, with Structure showing genuine sensitivity to question domain, exposing qualitatively distinct cognitive profiles invisible to accuracy-based evaluation.

Masked diffusion language models (MDLMs) generate text by iteratively unmasking tokens, but their standard decoder reduces each step to a binary action: a position is either committed to a single token or left fully masked, with no representation of partial belief in between. This all-or-nothing regime discards rich predictive information and forces premature, irrevocable commitments, leading to poor performance under a limited decoding budget. In this paper, we reinterpret mask prediction as clean-state prediction ($x$-prediction) and show that it can be used to induce a continuous flow in input embedding space. Building on this view, we propose a continuous decoding framework for MDLMs where tokens can accumulate partial progress at each diffusion step and remain revisable. To match the uneven contextual constraints across positions in language, we replace the globally synchronous schedule in image diffusion with a confidence-based asynchronous update in which the diffusion progress is token-wise accumulated. Additionally, we introduce a lightweight policy network and formulate its training as a reinforcement learning problem. Applied to pretrained LLaDA, our continuous decoder reaches 97% of its performance on the HumanEval dataset with 25% of decoding budget.

The unfairness of recommender systems has become a topic of concern due to its significant social and ethical implications. Although existing works have shown the effectiveness of attacks on the performance of recommender systems (e.g., promotion and demotion attack), the study of fairness attacks on recommender systems remains largely under-explored. To this end, we propose a novel structure-aware reinforcement learning-based fairness attack method designed to exacerbate the unfairness of target recommender systems. Specifically, we first employ a graph-based structure encoder to model the structural dependencies among the generated fake user-item interactions and the original user-item interactions. Then, we model the sequential dependency of the injected fake items using a recurrent neural network. Based on the learned structure-aware and sequence-aware representations of the fake user and item, the item selection policy attentively decides the next injected fake item. Since the target recommender system may employ fairness-aware training and leverage the user's sensitive attribute information, such as gender, we further designed a gender selection policy to decide the gender of the entire fake user profile. Both the item selection and gender selection policy are learned jointly in our proposed method. Finally, experimental results on four types of target recommendation models and two real-world datasets demonstrate the effectiveness of the proposed attack method in exacerbating the unfairness of recommender systems.

World modeling requires forecasting uncertain futures while preserving information useful for downstream perception. Existing visual world models often struggle to satisfy both goals: VAE-based stochastic models operate in low-dimensional reconstruction latents, which can limit perception performance, while deterministic predictors using strong pretrained features collapse multimodal futures into a single blurry mean. In this work, we propose FlowWM, a stochastic world model that performs flow matching directly within pretrained feature space (e.g., DINOv3). This is challenging because pretrained features are substantially high-dimensional, making standard diffusion recipes suboptimal. To address this, we investigate the design choices needed for feature-space flow matching and introduce a differentiable one-step projection mechanism that enables efficient training with temporal consistency and task-driven objectives. We evaluate FlowWM on two benchmarks: a synthetic benchmark for systematic evaluation of accuracy and diversity, and a real-world benchmark FuturePerception. FlowWM improves perception performance, mode coverage, and horizon robustness, validating our proposed design for stochastic world modeling in high-dimensional feature spaces.

Large language models (LLMs) deployed for structured generation (NER, JSON extraction, QA, and classification) lack formal reliability guarantees, and standard heuristic abstention policies miss user-specified risk targets by 7.5--12.5%. We characterize when conformal risk control (CRC) can certify structured LLM outputs and when it provably cannot. First, we prove an impossibility result: when the base risk (μ> α), any distribution-free method must abstain on at least ((μ-α)/(1-α)) examples, yielding a closed-form feasibility test: one can check whether CRC will work before running it. Second, we analyze a certification hierarchy across Hoeffding, empirical Bernstein, and a betting-based e-CRC bound, with strict gains in low-variance/large-sample regimes: the Hoeffding-to-Bernstein step delivers the largest gain (+37% certified configurations), while e-CRC adds value when calibration data is scarce (10% certification at 20% data versus 0% for Hoeffding). Third, we validate adaptive conformal inference (ACI) under cross-dataset shift, reducing risk-target violations from 71% to 21%, with residual failures concentrated exactly where the impossibility bound predicts. Across six open-weight models (3B--72B parameters), eight datasets, four tasks, and six nonconformity scores, hard NER/QA/CLS configurations are uncertifiable at (α= 0.10); relaxing to (α= 0.30--0.40) unlocks practical certification (47% NER, 40% QA, 60% CLS). The framework gives a three-step deployment recipe: check feasibility, select the bound and score, then mitigate shift.

In general, an ensemble classifier is more accurate than a single classifier. In this study, we propose an ensemble classifier called the kernel Fisher discriminant analysis forest (KFDA Forest), which is a tree-based ensemble method that applies KFDA. To promote diversity, bootstrap is used, and variable sets are randomly divided into K subsets. KFDA is performed on each subset to increase classification accuracy. KFDA maximizes the distance between classes while minimizing the distance within classes. KFDA can also be applied to classification problems in a nonlinear data structure using the kernel trick because it can transform the input space into a kernel feature space, commonly named a rotation, rather than performing a dimensionality reduction. Because new feature axes and KFDA projections are parallel, decision trees are used as a base classifier. To compare the proposed method with existing ensemble methods, we apply these to real datasets from the UCI and KEEL repositories.

Task-oriented semantic communication must allocate scarce radio resources to semantic features under fast fading wireless conditions and strict end-to-end latency budgets. Existing solutions are either optimization-heavy, leading to prohibitive computational overhead during online operation, or rely on end-to-end retraining procedures together with slowly varying channel assumptions. We propose iCoTASC (importance-aware Collaborative Task-Oriented Semantic Communication), a hybrid offline-online framework designed for collaborative multi-device semantic communication systems. iCoTASC leverages attribution-based importance to guide per-dimension embedding selection as a practical communication control signal, models diminishing semantic returns of quantization through a data-driven utility function, and precomputes per-transmitter utility lookup tables offline, which together enable lightweight online scheduling via table lookup and low-complexity refinement under time-varying channels. The proposed framework supports real-time, channel-adaptive semantic resource allocation in distributed systems without requiring retraining of the underlying task inference model.

Knowledge Tracing (KT) is important for personalized education but traditionally suffers from two key limitations: a reliance on shallow ID-based representations that neglect semantic depth and a restriction to single-granularity mastery estimation that overlooks hierarchical knowledge dependencies. To address these challenges, we propose MOSAIC (Multi-granularity Online Semantic AI for Collaborative Knowledge), a novel framework that orchestrates LLM-driven semantic alignment with sequential modeling. Unlike methods that use LLMs solely as predictors, MOSAIC leverages a frozen LLM to generate dynamic, context-aware embeddings and hierarchical prediction prompts, explicitly capturing collaborative signals and peer interactions. Furthermore, we introduce a cross-granularity consistency objective that jointly regularizes mastery estimation across concept, topic-cluster, and global proficiency levels. Extensive experiments on ASSISTments, EdNet, and a newly collected large-scale MOOC dataset demonstrate that MOSAIC establishes new state-of-the-art results. Specifically, our method achieves AUC improvements of up to 3.4\% and Accuracy gains of up to 2.5 \% across all benchmarks. Notably, MOSAIC exhibits superior robustness in collaboration-rich environments and long-sequence scenarios (AUC 0.862 on MOOC), offering both high predictive precision and semantically grounded interpretability.

Extracting interpretable, localized physical mechanisms from complex spatiotemporal data is a foundational challenge across physics, biology, and engineering, but has remained out of reach on real measurements. The central obstacle is obtaining high-quality gradients of data via numerical differentiation, which amplifies noise, diverges for high-order equations, and falters on irregular geometries, limiting the scope of existing approaches to clean simulations of low-order systems. Here, we present weak dominant balance, a derivative-free framework that projects governing equations into a weak (integral) formulation, offloading differentiation onto smooth analytical test functions and leaving the data untouched. The method sustains accurate regime identification under severe noise where existing approaches categorically fail, delivers the first data-driven decomposition of a third-order partial differential equation applied to turbulent duct flow, and produces matching decompositions across direct numerical simulation and particle-image velocimetry measurements of a wavy channel flow, uncovering a previously uncharacterized dynamical regime. Weak dominant balance brings mechanism-level analysis out of simulation and onto measured data, and opens complex physical systems to direct, equation-grounded interpretation.

Sharpness and complexity are two central factors in the generalization analysis of deep neural networks. Existing quantitative evaluations of generalization measures have largely focused on individual scalar measures, leaving the joint explanatory power of sharpness and complexity largely unexplored. This work studies how far sharpness and complexity can jointly explain generalization. We use linear regression and introduce a Pareto-based analysis to quantitatively evaluate the joint explanatory power of these two factors. Beyond the existing parameter-level definitions, we further propose realizations of sharpness and complexity that are closer to function space and less dependent on raw parameter representations. We find that function-oriented definitions of these two quantities expand the explanatory scope of the two-factor view beyond what is achieved by existing parameter-level metrics. Overall, our results support the sharpness-complexity perspective as an informative lens for understanding generalization across diverse settings. At the same time, the remaining failures indicate that whether this two-factor view can serve as a complete theory of generalization remains open.

Metric aggregation divergence (MAD) is the silent inconsistency that arises when distinct pipeline stages in an agent-based model coupled with a multi-objective evolutionary algorithm (ABM+MOEA) independently re-implement how an outcome metric is extracted from simulation trajectories. Unlike deliberate analytical choices, MAD operates at the level of pipeline architecture: each stage is internally coherent, and the inconsistency becomes visible only when cross-stage outputs are compared. Code inspection of EpidemiOptim, a JAIR-published epidemic policy toolbox, reveals three structurally independent aggregation paths in peer-reviewed code. A faithful replication of this structure produces champion disagreement in 64.2% of independent runs (n=500, 95% CI: [59.9%, 68.3%]). In a 300-seed policy-flip experiment, divergent aggregation causes the optimizer to recommend the wrong champion in 83% of replications, with a mean welfare gap of 2.19 units and a Gini inequality gap of 0.050 units. In a follow-up inference audit, 3 of 249 flipped seeds cross the significance boundary itself. A complementary enterprise follow-up produces the predicted null under near-commensurable rankings (rho = 0.991), while a public upstream rerun of the Lake Problem DPS workflow shows that the archived published-path recommendation reaches joint-threshold success 0.401 whereas a shared contract-path rule reaches 0.552. We introduce the metric contract - a single shared callable enforced at dispatch time across all pipeline stages - as the remedy. Framed as standard engineering discipline applied to the cross-stage metric interface, the contract eliminates divergence by construction with approximately 3% runtime overhead.

Large language models (LLMs) increasingly summarize clinical evidence, where a claim's weight depends on how strongly it is supported. Yet these models convey confidence poorly, and properties they never state, such as truth, are often readable from their activations. Whether a clinical model registers evidence strength, distinct from truth, and states it when asked is untested, and any such signal could be lexical. We compiled 45,134 clinical claims from six public sources, harmonized 20,611 into a four-level evidence grade under three independent frameworks, and tested 22 local, open-weight LLMs from several developers (0.6-70 billion parameters; general, medical, and reasoning), with lexical, truth, and cross-framework controls. A linear estimator recovered the grade in every model (median AUROC 71.8), yet decodability did not rise with scale and was weakest in reasoning models. The grade the models stated fell to chance, 25-27 percentage points below the estimator. The recoverable signal was largely lexical and did not transfer across topics or frameworks, yet it was distinct from factual truth and still flagged weakly supported claims (AUROC 69.2). Clinical LLMs thus carry an ordered evidence-strength signal they do not express, so their stated grades fail to convey a claim's support even when it is recoverable from their representations and text.

In regulated domains such as banking and healthcare, where privacy constraints make real speech costly to collect and retain, synthetic speech from modern text-to-speech (TTS) is an appealing alternative for training automatic speech recognition (ASR) without exposing sensitive customer recordings. Yet a persistent distributional gap between synthetic and real data limits how far it can replace genuine recordings. Prior work largely treats this gap as a black box to be engineered around, but in our work, we instead examine its origin directly by probing a SLAM-ASR architecture. Then, we localise where its LLM backbone separates real from synthetic speech and find the discriminative signal concentrated in the early-to-middle layers, where temporal and prosodic perturbations disrupt it most. We further show that representation-level separability, help, but does not directly predict downstream ASR gains. On the other hand, convolving synthetic audio with room impulse responses (RIRs) narrows the gap not by making synthetic speech sound cleaner or more natural, but by reproducing the acoustic irregularities of real recordings. Translating these findings into the training procedure, by adding a layer-selection module combined with RIR augmentation matches a fully real-data baseline using only 25% of the real speech (13.6h) and surpasses it at all higher proportions.

AI agents extend conventional large language model (LLM) applications by integrating language understanding with task execution, external tool use, and memory mechanisms. While memory allows agents to retain prior interactions and provide more personalized and context-aware responses, it also introduces a new vulnerability: information stored in memory can influence future outputs even when the current query is clean. In this paper, we investigate memory manipulation in LLM-based agents for multiple-choice question answering. We first design and implement an LLM-based AI agent with an external memory component that stores and retrieves task-relevant information. We then introduce basic memory manipulation scenarios in which misleading or corrupted memories are inserted into the agent before it answers multiple-choice questions. Using a controlled experimental setup, we compare the agent's performance before and after memory manipulation and measure changes in answer accuracy, attack success rate, and selection of manipulated options. Our results show that even simple memory manipulations can noticeably affect the agent's final answers, causing it to select incorrect options despite receiving clean and well-formed questions.

Multi-agent AI systems can improve answer selection by allowing different language models to exchange reasoning traces, revise initial predictions, and support a final decision. However, such communication may also introduce reliability risks: reasoning from one agent can correct another agent's mistake, but it can also mislead an agent that was initially correct. This paper studies reliable multi-agent AI communication through reasoning exchange and runtime answer revision. We develop a framework in which agents first answer multiple-choice questions independently, then share reasoning traces and revise their decisions. We conduct numerical experiments where we evaluate whether this process improves accuracy, produces more positive than negative answer transitions, and remains effective across domains such as cybersecurity, networking, and general knowledge. The results help identify when multi-agent reasoning improves reliability and when it may propagate errors.

IndicTrans2 is the strongest open English to Indic translation system, but like most systems it is trained on general text and tends to sound stiff on casual, conversational input. We adapt IndicTrans2-1B to conversational register across all 21 Indic languages using only public data (OpenSubtitles, BPCC-H-Daily, Tatoeba). Plain fine-tuning improves conversational chrF but forgets the general domain (it drops 3.9 chrF on FLORES for Hindi). Mixing general data back into training (experience replay) and then averaging the fine-tuned weights with the base (model souping) removes that trade-off: the resulting model beats IndicTrans2-1B on conversational chrF in every one of the 21 languages (mean +6.2) while matching it on FLORES (mean change -0.17, all within 0.7 chrF). Paired bootstrap tests confirm the conversational gains are significant (p <= 0.004) and that FLORES is not significantly degraded. We are deliberate about scope: these are chrF gains, and a blind human plus multi-model LLM check does not confirm them as a perceived quality improvement, so we treat the conversational gain as largely a register match to the references rather than proof of better translation. The techniques are not new; the contribution is the honest, end-to-end study in the Indic conversational setting.

Spatio-temporal grounding in long videos requires precise temporal localization and robust object tracking conditioned on natural-language queries. While recent vision-language models (VLMs) show strong reasoning ability, directly applying frame-by-frame inference to long sequences is computationally expensive and unstable. We propose a practical pipeline that shifts from frame-level to second-level tracking and performs cross-second smoothing to preserve continuity while reducing sequence length. To improve reasoning supervision, we synthesize chain-of-thought style trajectories using advanced multimodal models for temporal localization and target selection, and replace generated spatio-temporal coordinates with ground-truth annotations to avoid noisy supervision. We further optimize the policy with reinforcement learning using a verifier based on $t\_\mathrm{IoU}+mv\_\mathrm{IoU}$. Experiments across multiple FPS settings show that our method achieves a strong trade-off between efficiency and localization quality.

Weather synthesis aims to add weather effects to input videos while preserving scene identity, structure, and motion. The key limitation of existing methods is the lack of diversity in weather appearance and effective control over weather dynamics (e.g., temporal evolution and particle motion). Most approaches rely on text prompts, which are inherently underspecified and often fail to produce detailed weather characteristics. Additionally, general-purpose video editors optimized for clean and aesthetic outputs tend to suppress heavy weather phenomena, making dense particle effects difficult to generate. To address these, we propose a Semantic-Aware, Physics-Informed, and Geometry-Grounded framework that steers an off-the-shelf video editor to synthesize diverse global appearances and detailed particle dynamics. We factorize the synthesis into three conditional signals, so that each provides a distinct and stable source of guidance: semantics specifies what the weather should look like, dynamics governs how it evolves over time, and geometry determines where it should appear in the scene. Specifically, we introduce (1) semantic-aware appearance anchoring to establish the target appearance from scene semantics and user input; (2) physics-informed dynamic simulation to generate particle effects by simulating a Gaussian-represented particle field under gravity, wind, and turbulence; and (3) geometry-grounded video synthesis to align the simulated particles with target scene geometry and synthesize the final video. Experiments demonstrate that our method produces diverse, physically and visually realistic weather effects. Furthermore, we show that our synthesized data significantly improves the robustness of autonomous driving semantic segmentation under adverse weather conditions. Project page: https://jumponthemoon.github.io/w-crafter/.

We show that net demand for liquidity by algo strategies is identifiable from its trade and price history alone, with no knowledge of its signal or optimization problem. An exact multi-period regret decomposition implies that the sign of this statistic classifies a linear strategy as a net liquidity consumer or provider, recovering the Kyle (1985) informed-trader/market-maker dichotomy from observables alone. Under an AR(1) cost process, the same statistic equals the product of strategy size and the squared Roll (1984) implied spread, making the correction a direct proxy for prevailing illiquidity. Extending to endogenous price impact and aggregating across N correlated strategies yields a liquidity-balance condition whose violation produces welfare loss scaling as N squared, a closed-form fire-sale externality. We calibrate to CRSP equity data (2016-2025), tracking implied spreads through the COVID-19 and 2022 rate-shock episodes, with an estimator computable in O(Tnd) time.

Recent advancements in generative artificial intelligence (AI) and large language models (LLMs) have shown significant promise in automating complex reasoning, summarization, and question-answering tasks. However, the effectiveness of general-purpose LLMs in specialized engineering domains remains limited due to insufficient exposure to technical standards, engineering terminology, and domain-specific semantics. This study proposes a systematic approach to developing a customized generative AI agent for transportation engineering applications. A curated corpus of U.S. transportation manuals, design guidelines, and regulatory documents is used to conduct continued pretraining of six state-of-the-art LLMs through a unified low-rank adaptation (LoRA) framework. The training process is monitored to ensure convergence and model stability. Performance is evaluated using standard natural language processing metrics, including BLEU-4 and ROUGE, with Qwen2.5-7B and LLaMA-3.1-8B demonstrating the highest domain alignment and response quality. Results validate the effectiveness of LoRA-based adaptation in improving LLM performance on technical content interpretation and context-specific reasoning. This work contributes a reproducible development framework for constructing domain-specialized generative AI agents, supporting broader deployment in transportation research, design, planning, and policy analysis.

The mean mechanism is known to be non-incentive-compatible, namely, rational players are incentivized to misreport their values. Despite this game-theoretic issue, the mean mechanism is prevalent in practice due to its other desirable properties. We give a full characterization of pure Nash equilibria--how the players will misreport--for the affine mechanism, of which the mean is a special case. Furthermore, we characterize both complete-information and Bayesian games under the affine mechanism. Our results highlight the inevitability of extreme exaggeration in such games.

Current examples of SysML-based verification of discipline-specific models in the literature typically have two flaws. Firstly, they are developed in a tool-specific manner using proprietary APIs, limiting portability. Secondly, they focus on performance properties modelled via parametric diagrams, overlooking behavioural and interface properties that also require verification. This project addresses the problem with a verification process tailored to model-based verification, informed by common SysML tool capabilities and the UML Testing Profile, that enables automated verification of discipline\-/specific models from SysML test cases and returns the results to the SysML model for traceability. A mixed-method approach combining literature research and stakeholder interviews was used to derive validated stakeholder needs, which drove the specification and design of the process. The process was demonstrated end-to-end in two independent SysML tool-chains to evidence tool-agnosticism, and was shown to verify behavioural and interface requirements, including ordering, timing, and state-based responses, using SysML behavioural diagram constructs that parametric approaches alone cannot address.

Encoders have become the state of the art for multiple NLP tasks, especially those requiring deep contextual understanding. While multilingual models offer broad coverage, dedicated monolingual encoders are essential for capturing the unique lexical and syntactic nuances of specific languages. For Portuguese, however, existing monolingual options like BERTimbau and Albertina have not kept pace with recent architectural breakthroughs, often lagging behind English benchmarks in scalability and efficiency. This work introduces BERTomelo, a next-generation monolingual encoder pre-trained from scratch and specifically optimized for the Portuguese language. By leveraging the ModernBERT architecture, BERTomelo overcomes the limitations of previous models, offering Base and Large versions with a 1,024-token context window and hardware-level optimizations like FlashAttention and alternating attention mechanisms. The model was trained on ClassiCC-PT, a massive, high-quality Portuguese corpus of 106 million documents, ensuring superior alignment with the language's contemporary usage. The results demonstrate that BERTomelo not only outperforms previous Portuguese encoders but also provides a more robust and efficient alternative to massive multilingual models in downstream tasks such as STS and NER.

Large Language Model (LLM) alignment trains an LLM using preference data to produce outputs that better meet established quality standards. While LLM alignment techniques are studied for non-coding tasks, we know little about their usefulness for coding tasks. It is unclear whether LLM code alignment could support both functional requirements (producing executable, correct code) and non-functional requirements (code readability, style, maintainability). It is also unknown whether alignment for a code LLM should begin with base pretrained version or the finetuned (i.e., instruction-tuned) version of the LLM. In this paper, we offer insights on the above two research questions by conducting an empirical study. We studied five state-of-the-art (SOTA) LLMs using two widely used LLM alignment techniques: Direct Preference Optimization (DPO) and BoNBoN. For each training record, we created a preference pair as accepted and rejected instances by using the SelfCodeAlign pipeline. DPO and BoNBoN are reward-free models, i.e., they eliminate the need for multiple reward scores for output preferences. We tuned each LLM using the two alignment techniques in two settings: pretrained and finetuned versions of an LLM. We evaluated functional requirements using four SOTA benchmarks (HumanEval+, MBPP+, EvalPerf, EvoEval) and non-functional requirements using the CODAL benchmark, which evaluates code quality across five dimensions derived from software engineering practices. We find that pretrained-to-aligned pathways achieve larger improvements in the aligned variant over its pretrained variant. But the pretrained variant is generally less accurate than its finetuned variant. However, finetuned- to-aligned offers smaller performance improvements or, in some cases, degradation in the aligned variant than its finetuned variant.

Short-fiber thermoplastic (SFT) composites are increasingly employed in lightweight aerospace and automotive structures owing to their favorable strength-to-weight ratio, high production rates, and recyclability. Unlike continuous-fiber systems, the mechanical response of SFTs is governed by mesoscale interactions among fiber orientation, spatial clustering, and manufacturing-induced porosity. These features exhibit significant spatial variability in manufactured components and influence stiffness, damage initiation, and nonlinear deformation. Although mesoscale finite element (FE) models can resolve such heterogeneity, their application to realistic three-dimensional microstructures remains computationally intractable. A data-driven surrogate framework is proposed to predict the mechanical behavior of additively manufactured, compression-molded (AM-CM) SFTs. Microstructures reconstructed from micro-computed tomography data were discretized into Voronoi-based cells representing distinct fiber-interaction neighborhoods. Each cell was homogenized via nonlinear FE simulations incorporating matrix damage, and the resulting stress-strain responses trained a hybrid Graph Neural Network-Long Short-Term Memory (GNN-LSTM) architecture encoding microstructural topology and history-dependent mechanical evolution. The surrogate accurately predicts stiffness and stress-strain behavior of unseen microstructures, achieving $R^2\approx 0.98$ relative to high-fidelity FE simulations with over two orders-of-magnitude reduction in computational cost. Coupling the framework with experimentally calibrated damage laws demonstrates that fiber orientation, clustering, and porosity collectively govern local effective stiffness. The approach provides a physics-informed, data-efficient pathway to identify mechanically weak microstructural cells and accelerate digital-twin development for SFT components.

This paper presents new results and breakthrough obtained with the HbHAI techniques (Hash-based Homomorphic Artificial Intelligence) proposed in \cite{filiol0,sepp}. HbHAI is based on a novel class of key-dependent hash functions that naturally preserve most similarity properties, most AI algorithms rely on. It enables to analyse and process data in its cryptographically secure form while using existing native AI algorithms without modification, with unprecedented performances compared to existing homomorphic encryption schemes and most notably compared to the same processing on corresponding plaintext data. Two major results have been obtained further. First we enable to reduce the compression rate up to a factor of 10 thus allowing to process massive datasets while reducing the computation time and the energy footprint in the same order. Second, we show how it is possible to arbitrarily reduce the final validation error of AI-based decision tests by using repetition error-correcting codes.

General-purpose large language models (LLMs) have demonstrated strong abilities in opendomain question answering, information extraction, and text generation. Agricultural applications, however, are domain-specific, region-dependent, time-sensitive, and safety-critical. Without data governance, expert evaluation, and evidence constraints, an agricultural assistant mayproduce unreliable advice on crop diseases, pesticide use, fertilization, or policy interpretation.To avoid presenting unverified simulated numbers as real experimental findings, this paper doesnot report any model-performance claims that have not been produced by an actual training runand expert evaluation. Instead, we propose AgriTune-R, a reproducible and auditable frameworkfor adapting general-purpose LLMs to agricultural tasks. The framework selects the publiclyverifiable Qwen3-8B model as the recommended base model and integrates agricultural datagovernance, instruction construction, LoRA/QLoRA parameter-efficient fine-tuning, retrievalaugmented generation, expert evaluation, and safety control for high-risk questions. The contributions are: (1) a structured workflow for agricultural LLM adaptation; (2) an evaluationprotocol for agricultural knowledge QA, pest and disease consultation, cultivation management,and policy explanation; (3) an expert-review rubric combining factuality, safety, evidence consistency, and uncertainty expression; and (4) a clear separation between protocol design andempirical conclusions, providing an executable baseline for future empirical studies.

We give a single compositional setting in which gradient-based learning and Hamiltonian-style mechanics appear as functorial semantics. The syntax is an operad Arr whose objects are input-output interfaces (pairs of manifolds) and whose morphisms are *smooth adaptive arrangements*, which consist of a reactive parameter space, a lens given by smooth output and input maps, and a real-valued potential. The main technical result of the paper is what we call *lens internalization*, a lax symmetric monoidal functor Lens(C) $\to$ C associated to any symmetric monoidal closed category C. Using it, we provide two functors $Φ_\text{phase}$, $Φ_\text{conf}$: Arr $\to$ PC into the 2-category of polynomial coalgebras -- input-output discrete dynamical systems -- which we take as the semantics category. $Φ_\text{phase}$ stores both position and momentum, whereas $Φ_\text{conf}$ stores only position. When applied to a parameterized function, $Φ_\text{conf}$ recovers the gradient descent training algorithm, with backpropagation as the lens' backward pass. When applied to harmonic particles wired together -- in series, or according to any finite directed graph -- one diagram yields two different regimes, both of which are governed by the graph Laplacian: $Φ_\text{phase}$ gives the discrete wave equation, which is conservative and second-order, and $Φ_\text{conf}$ gives the discrete heat equation, which is dissipative and first-order. They are two semantics of one adaptive arrangement, e.g. with the same potential in each case. And because Arr is an operad, such diagrams nest -- larger systems wired from smaller ones -- and each semantics assembles a system's dynamics functorially from its parts. These dynamics are moreover executable: a parameterized neural network and a graph of particles both compile, by the same construction, to explicit state machines one can run.

Ovarian cancer is frequently diagnosed at an advanced stage, making preoperative contrast-enhanced computed tomography (CT) central to staging and surgical planning; yet the scarcity of annotated imaging data, compounded by privacy regulations, limits the development of generalizable computational models in this domain. Text-conditioned 3D CT synthesis has shown promise, but existing pipelines depend on paired radiology reports and have been evaluated only on chest CT. We propose OvESyn (Ovarian Evidence-based Synthesis), a framework that constructs standardized Findings and Impression sections directly from CT-derived imaging descriptors and routine clinical metadata, without any original radiology report, and uses them to condition a latent diffusion model adapted to 493 high-grade serous ovarian carcinoma patients. This is the first text-conditioned 3D CT synthesis framework adapted to an abdomino-pelvic oncologic setting. A systematic ablation over two adaptation axes, vision-language encoder alignment and generator fine-tuning, identifies generator domain adaptation as the operative mechanism for crossing the domain gap and establishing the target anatomy: without it, synthesis remains anchored to the thoracic pretraining domain, with Precision and Recall collapsing to zero and FID2.5D exceeding 140, regardless of encoder alignment. Encoder alignment instead refines intensity and fine detail. The full OvESyn attains the best distributional and intensity fidelity (FID2.5D 29.35, Precision 0.671, Wasserstein-1 0.044), while the generator-only variant maximizes coverage (Recall 0.645), reflecting a fidelity/coverage trade-off governed by encoder adaptation. Requiring only automatic segmentations and routine preoperative metadata, OvESyn supports transferability to report-scarce settings and provides a foundation for synthetic cohort generation in abdomino-pelvic oncologic imaging.

We audit fourteen mainstream large language models (LLMs) for hiring discrimination using the paired-resume methodology of Kline, Rose, and Walters (2022). The sole 2023-vintage model reproduces the pro-White callback gap documented in field experiments on labor market discrimination ($+2.12$ pp, significant at the 1\% level). Every model released in 2024 or after shows either a null gap or a significant pro-Black reversal (up to $-3.01$ pp). The same pattern holds on the gender axis. Based on 24,024 paired postings per model across 14 models, our results document a reversal in the direction of algorithmic hiring bias across model generations.

Linked lists are one of the most basic data structures in computer science. But when many threads try to use the same linked list at the same time, things get complicated. In this paper, we look at five different ways to make a linked list work correctly and efficiently with multiple threads running at once. We start with the simplest approach -- one big lock for the whole list -- and step by step improve it, ending with a lock-free design that uses no locks at all. We implemented all five versions in C++ and measured how fast each one is across different workloads (read-heavy, balanced, and write-heavy) and different list sizes. Our results show that the right choice of algorithm depends heavily on how the list is used: the coarse-grain and lazy lists win under read-heavy workloads with small key ranges, while the lock-free list becomes competitive when key ranges are large and more threads are running. Fine-grain locking, despite its theoretical appeal, pays a heavy cost from per-node lock overhead and consistently performs the worst in our tests.

Unsupervised tabular anomaly detection requires methods that are accurate, robust across heterogeneous datasets, and computationally efficient. Classical statistical detectors are often efficient, but they usually rely on a fixed data view and a single notion of abnormality. Deep anomaly detectors can learn more flexible scoring functions, but they are substantially slower and difficult to tune in unsupervised settings due to the lack of a reliable supervisory signal. We propose RGLD, a randomized global-local density estimator for efficient unsupervised tabular anomaly detection. RGLD combines a global random-feature density branch, which identifies samples in broadly low-density regions, with a local neighbor branch, which detects samples that are weakly supported by nearby observations. Both branches operate over feature-bagged randomized views, allowing RGLD to expose anomaly evidence that may be hidden in any single representation. We conduct experiments on 47 tabular datasets against 23 statistical and deep anomaly detection baselines under fully unsupervised setting. RGLD achieves the strongest dataset-level AUROC performance, ranking 1st in dataset wins, and ranks 2nd in AUPRC wins. RGLD is also faster than all evaluated deep detectors, achieving 50x-580x speedups, and remains competitive with statistical methods in runtime, yielding a favorable accuracy-efficiency tradeoff.

Large language models (LLMs) are increasingly used to simulate social survey responses, yet their outputs exhibit systematic biases: marginal distributions are skewed, response variance is poorly calibrated, and predictor-outcome relationships are attenuated. We ask a simple question: given a small pilot sample of human responses, can an LLM recover the statistical characteristics of a broader population? We decompose recovery along three axes: structural fidelity, marginal fidelity, and individual fidelity. Using a COVID-19 misinformation survey as a case study, we benchmark three families of approaches: prompting, rectification, and fine-tuning. The findings suggest that fine-tuning on small pilot samples offers a balanced approach for achieving multiple forms of fidelity, but the levels of such fidelity can vary across subsamples, potentially threatening pluralistic alignment.

Model quantization is essential for the efficient deployment of Large Language Models (LLMs), but introduces a critical vulnerability: Quantization-Conditioned Backdoor (QCB) attacks. In these attacks, malicious behaviors remain dormant in full-precision models and activate only after specific quantization distortions, bypassing standard security audits. To mitigate this, we introduce FlipGuard, a proactive defense framework that selectively perturbs model weights prior to quantization. By breaking the adversary's precise alignment between weight patterns and quantization boundaries, FlipGuard suppresses backdoor activation without requiring access to training data or trigger samples. We further propose the Defense Effectiveness Ratio (DER), a unified metric to jointly evaluate security gains, utility preservation, and computational cost. Extensive experiments across seven LLMs (including StarCoder and LLaMA-family models) and three quantization schemes (INT8, FP4, NF4) demonstrate that FlipGuard effectively neutralizes QCBs across three scenarios, i.e., vulnerable code generation, content injection, and over-refusal, achieving high security with negligible performance degradation.

Physicians now pose millions of clinical questions to AI tools each week, yet these tools are evaluated largely on hypothetical or exam-style questions, not those actually asked in practice. We report a blinded evaluation built on 620 Real-world Point-Of-Care Queries (Real-POCQi) submitted to the OpenEvidence (OE) platform by physicians spanning 30 specialties, as well as 187 questions from HealthBench. 149 practicing physicians across 36 states made head-to-head comparisons between answers from three frontier general-purpose models (Claude Opus 4.8, Gemini 3.1 Pro, and GPT-5.5) and a specialized clinical tool (OE), with graders matched to each question's specialty. When comparing answers along five dimensions relevant to clinical decision support -- accuracy, clinical utility, source quality, verifiability, & completeness -- physicians scored the specialized tool highest on all axes; in the primary analysis on Real-POCQi, win differences (margins between win and loss rates) ranged from 25 to 39 percentage points (p<0.001). Results remained consistent in sensitivity analyses stratifying by citation display, answer length, OE-user status, and Real-POCQi versus HealthBench. In parallel, LLM judges were found to systematically differ from expert judges, though both generally agreed on the best model. These findings underscore two conclusions: (i) AI tool evaluations should reflect real-world query distributions and use expert judges that mirror the specialization defining modern medicine and (ii) the consistent advantage of the specialized tool over general-purpose models does not necessarily mean that the latter cannot serve similar purposes, but that targeted engineering and customization can yield meaningful gains in performance for its users. We release Real-POCQi as a public benchmark, as well as the prespecified statistical analysis for reproducing results of this study.

LLM agents can share more than text. In some systems, an agent can send a short visible message while also passing its full KV-cache state to another model. This hidden state can help the final model combine evidence from several agents, but it is also hard to inspect. A visible message may look harmless even if the hidden state has been changed. We study this problem in a multi-agent question-answering setup. Specialists each see part of the evidence, send a short commitment, and pass full KV-cache state to a coordinator. In clean runs, this latent collaboration improves over a matched text-only version. On transformed HiddenBench with Qwen3-4B, it reaches EM/F1 of 0.338/0.486, compared with 0.231/0.369 for text collaboration. Qwen3-8B and HotPotQA runs show the same direction of improvement. The problem appears when one specialist is malicious. Some false visible commitments can steer answers. More seriously, changing the hidden KV state can collapse performance even when the visible commitment still looks plausible. A verifier that checks only text misses this failure mode. Simple magnitude checks catch some obvious corruptions, but adaptive attacks can evade them while still damaging the final answer. The most reliable fix we find is not to guess whether hidden state looks normal, but to protect it in transport. We implement an HMAC-SHA256 manifest that binds the specialist, session, model, visible commitment, tensor metadata, and payload digest. It accepts all 774 honest replayed payloads and rejects all 295 recorded tampered payloads. The main lesson is that full-KV latent memory can be useful, but it should be treated as a security-sensitive object, not as ordinary internal model state.

Reinforcement learning agents can exploit misspecified reward signals to achieve high apparent returns while failing on the intended objective, a failure mode known as reward hacking. Existing practical defenses typically constrain policy updates to stay near a known safe reference, creating a tension between suppressing hacking and permitting legitimate improvement. We propose Modification-Considering Value Learning (MCVL), which operationalizes the theoretical idea of current utility optimization for standard value-based RL. MCVL wraps an off-policy learner and treats each incoming transition as a candidate modification: it forecasts two training paths, one that includes the transition and one that does not, and scores both with a frozen bootstrapped-return estimator derived from a learned reward model and value function. The transition is admitted only if inclusion does not decrease the score. We formalize conditions under which this filtering is both safe and permissive, and instantiate MCVL with DDQN and TD3. Across four safety-relevant gridworlds and three modified MuJoCo continuous-control tasks with diverse hacking mechanisms, MCVL mitigates reward hacking while continuing to improve the intended objective. Project website: ktolnos.github.io/mcvl/.

Poisoning attacks entail attackers intentionally tampering with training data. In this paper, we consider a dirty-label poisoning attack scenario on a speech commands classification system. The threat model assumes that certain utterances from one of the classes (source class) are poisoned by superimposing a trigger on it, and its label is changed to another class selected by the attacker (target class). We propose a filtering defense against such an attack. First, we use DIstillation with NO labels (DINO) to learn unsupervised representations for all the training examples. Next, we use K-means and LDA to cluster these representations. Finally, we keep the utterances with the most repeated label in their cluster for training and discard the rest. For a 10% poisoned source class, we demonstrate a drop in attack success rate from 99.75% to 0.25%. We test our defense against a variety of threat models, including different target and source classes, as well as trigger variations.

In this study we present a formal definition of large discrete sets having, informally, three properties: their elements are easily recognized, easily generated, and the latter tasks are easily learned from examples. The formalism is specialized to sets of binary strings and a definition of "machine-learnability" based on the existence of a bounded-complexity Boolean autoencoder that fixes the elements of the set. We present experiments where the autoencoders are implemented by nets of Boolean threshold functions. Machine-learnability is demonstrated for Rorschach patterns (that may have reversed contrast in the mirrored half), and considerably "wilder" sets whose elements are only approximately fixed by admissible autoencoders. In the second case we demonstrate a simple iteration that evolves wild sets to make them properly machine-learnable.

Learning to bid in repeated multi-unit auctions with bandit feedback poses a fundamental challenge. Existing methods often rely on rigid explore-then-exploit schedules, assume stationary adversaries, and optimize solely for bidder utility, thereby limiting adaptability and strategic robustness. To address these limitations, we introduce the A3M framework, which integrates adaptive deep reinforcement learning (DRL), explicit adversarial reasoning, and principled multi-objective reward design for online auction strategy optimization. A3M employs an actor-critic DRL backbone to dynamically balance exploration and exploitation, an opponent model for fictitious play against non-stationary adversaries, and a composite reward function to jointly maximize utility, auctioneer revenue, and fairness. We provide the first comprehensive empirical evaluation of this integrated approach against established baselines in both discriminatory and uniform price auctions. Results show that A3M reduces final regret by 30--40\% in standard settings, maintains robust performance against adversarial strategy shifts, scales favorably with the number of units $K$, and enables tunable multi-objective trade-offs. An extensive ablation study confirms the necessity of each core component. Our work establishes A3M as a powerful and flexible framework for learning in complex auction environments.

Behavior-cloned diffusion policies are expressive but remain vulnerable to covariate shift: small deviations from demonstrated states can compound into task failure. Existing methods address this either by expanding the training distribution through expert corrections or synthetic augmentation, or by steering a frozen policy at test time with guidance from a learned model. The former can be expensive or assumption-dependent, while the latter discards the corrected trajectories after execution. We introduce ReGuide, a self-improving framework that treats guided rollouts as reusable on-policy recovery data. ReGuide first uses Phase-Conditioned Guidance (PCG) to generate corrective rollouts: it constructs phase-specific latent targets, applies guidance only in the drifted-but-recoverable regime, and guides through the estimated clean action to match the dynamics model's training distribution. Successful guided rollouts are then absorbed back into the policy through ReGuide-FT, which fine-tunes the current checkpoint, or ReGuide-FS, which retrains from scratch on the augmented dataset; the two can also be composed and iterated. On Robomimic Can, Square, Transport, and Tool Hang, ReGuide improves base-policy success by $1.3$--$7.7\times$, outperforms LPB in the test-time-only setting, and matched-data ablations show that the gains come from guided recovery data rather than additional rollouts alone.

Modern vision-language models (VLMs) for driving assistants typically treat vehicle dynamics as a black box, resulting in decisions that lack awareness of the vehicle's real-time electro-mechanical state. To bridge this gap, we introduce the Electro-Visual-Language Assistant (EVLA) -- a novel framework that combines multi-modal scene understanding with real-time perception of the electrified powertrain state (e.g., motor torque, battery SOC). Our approach features two key innovations: first, a Unified Co-State Encoder (UCSE) that fuses visual, textual, and vehicle-state inputs into a shared latent representation, augmented with an Energy-Efficiency Field to model spatial energy costs; and second, an Electro-aware Structured Reasoning Chain (ESRC), which replaces external chain-of-thought prompting with an internal, deterministic reasoning process grounded in physical constraints and optimization objectives. Trained end-to-end with a physics-guided joint loss, EVLA learns to generate context-aware and energy-optimal driving decisions. Extensive evaluations on a driving QA benchmark demonstrate that EVLA substantially outperforms strong fine-tuned VLM baselines, improving the final score by +0.0871 and accuracy by +5.6\%. Ablation studies validate the necessity of each component, and efficiency analyses show that EVLA achieves 36\% faster inference than multi-stage pipelines. This work underscores that integrating vehicle-state awareness and structured physical reasoning is crucial for developing next-generation, physically-grounded driving assistants.

Venture capital (VC) investment decisions face distinct challenges, such as multi-source heterogeneous data, non-stationary time series, and the demand for explainable predictions in high-stakes, low-data settings. To overcome these issues, we introduce \textbf{FinInvest-GTCN}, a Graph-Temporal-Causal Network that redefines the task from content recommendation to quantitative risk-return assessment. This architecture combines a relational graph encoder to capture the investment ecosystem's topology, a multi-scale temporal fusion module to handle long-term dependencies and non-stationarity, and a causal decision head that generates risk-adjusted predictions with interpretable causal attributions. A core innovation is the Meta-Causal Adaptation (MCA) strategy, which facilitates robust fine-tuning for new, data-scarce sectors by aligning updates with causally-plausible structures derived from meta-pretraining. Comprehensive experiments on proprietary VC datasets show that FinInvest-GTCN delivers state-of-the-art results, markedly lowering the primary Risk-Adjusted Mean Squared Error (RA-MSE) to 2.51 from a baseline of 3.05 and boosting the cumulative return of a simulated portfolio by 18.7\%. Ablation studies underscore the essential role of each component, while additional analyses confirm the model's stability, interpretability, and enhanced adaptability. This work pioneers a data-driven, explainable framework for investment decision support.

Large language models have driven recent progress in language and multimodal AI, yet pre-training them at scale is prohibitively expensive. Low-rank pre-training, which factorizes each weight matrix into a rank-r product to reduce both parameters and FLOPs, is a promising response but typically lags full-rank training in quality. We propose Duplicated Latent Residual (DLR), a training-only, parameter-free, foldable plug-in for low-rank pre-training. DLR augments the standard low-rank output Bz with a fixed structured residual alpha/sqrt(K) * Expand_K(z) that replicates each latent coordinate K = ceil(d_out/r) times across the output. With alpha fixed, DLR adds zero learnable parameters per layer; after training, it is absorbed into the up-projection in closed form, B* = B + alpha/sqrt(K) R^T, so deployment parameter count, FLOPs and memory match the underlying low-rank backbone exactly. Across LLaMA models from 60M to 7B parameters, DLR strengthens low-rank pre-training on C4 validation perplexity in most settings, with the clearest gains at 130M and above; folded checkpoints transfer cleanly to supervised fine-tuning on standard benchmarks.

Cybersecurity is a real-life test-bed for many machine learning problems at once, especially when considering modern strides in using Large Language Models (LLMs) to automate processes as ``agents.'' Cybersecurity workflows require orchestrating hundreds of standard and bespoke tools through various formats. The scale of cybersecurity data is enormous; for example, a single malware sample can be viewed as a sequence of billions of tokens. The cost of labeling any file by experts is enormous and labor-intensive, in part because an adversary (possibly a well-funded nation state actor) is attempting to subvert your detection methods. Even skilled experts may disagree on the correct label, creating ambiguity in what constitutes ground truth. When deployed, models must run quickly on billions of items a day, where low-latency is critical for operational success, in a continuously changing environment. In addition, explainability is not optional: analysts demand clear reasoning for model decisions to cope with the large number of false-positive alerts they face daily, and to quickly develop remediation and understand how something went wrong. In short, the amount of complexity cybersecurity is greater than that of natural language and computer vision, and thus we posit that cybersecurity is the better test-case for general AI progress than other, well-studied fields.

In-context learning (ICL) is an emerging paradigm that employs the semantic information inherent in large language models (LLMs) for generating answers to user queries. While the remarkable performance of ICL has been widely known, a general modeling and a rigorous theoretical analysis of this paradigm are still lacking. This work presents a probabilistic model for ICL and derives the performance of ICL for both general parametric distributions and exponential families. Based on the derived results, the work explains the impact of multiple factors such as the number of demonstrations, the sensitivity of the probabilistic model to the variation of its parameters, as well as the similarity between the demonstrations and the query on the performance of ICL.

Tool and agent routing from natural-language prompts is naturally a set-valued prediction problem: a single query may require multiple agents, while over-selection increases execution cost. The benchmark introduced here is derived from WildChat and contains 3,000 prompts over a fixed 12-agent catalog, with AI-assisted heuristic labels under a fixed schema and controlled rebalancing for multi-label evaluation. The evaluation protocol combines set-level metrics (Precision, Recall, F1, Jaccard, and Exact Match), latency, an execution-oriented capability-coverage simulation, and a constrained weighted-routing setting based on ordinal agent-cost tiers. Compared methods include nearest-neighbor matching, linear multilabel classification, dependency-aware baselines, a fine-tuned encoder, deterministic weighted post-scoring via Weighted Agent Routing (WAR), and a zero-shot LLM baseline. Results show that supervised routers substantially outperform nearest-neighbor and zero-shot LLM routing. The fine-tuned encoder achieves the strongest unconstrained set accuracy, while the linear multilabel model provides the strongest practical baseline. In the constrained setting, the weighted routing layer improves utility when applied on top of strong supervised scorers, with the largest gain observed for Encoder+WAR. Overall, the benchmark and evaluation protocol support reproducible study of accuracy-cost trade-offs in fixed-catalog multi-agent routing.

Detecting security threats in an organization's cloud computing environment has become necessary due to the increased reliance on cloud infrastructure. Logging of all cloud computing events enables investigation into any incidents after they are detected. Automated detection of threats using the logs based on heuristics or anomaly detection could result in a high false positive rate due to its relatively static nature. In this article, we present an industrial case study of a self-supervised learning method using graph neural networks applied to AWS CloudTrail logs to surface suspicious events for analyst review. The model produces an anomaly score for each event and dynamically adapts to changes in the organization without requiring periodic retraining. Based on our experiments across five organizations, the proposed model produced substantially fewer alerts than a domain expert rule-based baseline in almost all cases, reducing alert volumes to approximately 1 per hour from thousands generated by traditional methods. We note that this evaluation covers only flagged events, and false negatives cannot be estimated from the current data; findings should therefore be interpreted as a practical deployment study offering insights into real-world constraints rather than a fully validated detection system. We discuss these limitations and the requirements for extending the approach to other cloud environments as future work.

Lightweight Directory Access Protocol (LDAP) is a protocol that allows users to query and modify Active Directory (AD) data. By default, all users have read access to all AD data through LDAP, making it a common initial tool for reconnaissance when a threat actor first compromises an identity. To capture threat actors early in the reconnaissance phase, we developed two machine learning frameworks to detect LDAP reconnaissance: an ML classifier to predict malicious LDAP queries and an ML-based data-mining method to extract malicious query signatures. By correlating LDAP queries with endpoint detections, the first framework uses weak supervision to label a massive dataset and classify LDAP queries as malicious or benign. For immediate deployment, a second technique was developed on top of this approach to employ a rigorous statistical hypothesis-testing framework for mining novel, malicious LDAP signatures. While this weakly supervised approach is limited compared with manual human labeling, it is more practical for this use case because it leverages large-scale automated corpus construction, reducing costs and time. Ultimately, both the LDAP classifier and the ML-based LDAP signature mining method achieved performance benchmarks, with the classifier achieving up to a 65\% True Positive Rate (TPR) on the holdout set while limiting false positives, and mined signatures demonstrating 81.48\% field precision with CrowdStrike's Managed Detection and Response team.

We introduce GRAB, a constructor-encoder-bridge pipeline for table question answering. Our method lifts relational data into an heterogeneous graph, encodes it via message passing, and transfers the signals to an LLM through a small set of query-conditioned latent tokens. This provides the LLM with a compact, task-relevant structural representation together with the flattened text. Crucially, the LLM remains strictly frozen to preserve its general reasoning capabilities; we train only the lightweight graph encoder and latent bridge (91M parameters), allowing the entire pipeline to be trained efficiently. Our pipeline significantly improves performance on relational Question Answering, with the largest gains in demanding multi-table settings, offering an efficient, principled way to connect relational deep learning with LLMs.

Machine-learned (ML) operator models can be trained to predict density functional theory (DFT) Hamiltonian/density matrices at significantly reduced computational cost, thus extending electronic-structure calculations to previously unfeasible scales. Here, we introduce MALOQ (Massively Accelerated Learning of Operators for Quantum Transport), an application built to train on and predict electronic-structure matrices for systems made of few to 100k atoms, described by large basis sets, and covering a wide range of atomic elements. Based on a state-of-the-art, SO(2)-equivariant backbone architecture, MALOQ provides (i) custom data-processing kernels to handle high-rank Hamiltonian matrix data and (ii) a scalable edge-wise distribution of atomic graph(s). Trained on the largest molecular Hamiltonian datasets available today, it reduces time-per-epoch by over 30% compared to a molecule-wise-distributed framework, and enables inference on material graphs of arbitrary size. We demonstrate scalable training and inference for 3,000-12,000 atoms on the Alps supercomputer, up to 192 GPUs and 256 GPUs, respectively.

Efficient computation of shortest paths in weighted graphs is a fundamental problem with many applications. Neuromorphic hardware platforms promise massively parallel, efficient computation, changing parallelism tradeoffs. In this work, we introduce NEURO-MAPP (Neuromorphic-based Min-Add Parallel Propagation), a distributed shortest path algorithm designed to use the local computation and network communication available in neuromorphic systems. We provide an optimized implementation of the algorithm on the SpiNNaker 2 platform and evaluate its performance on a selection of synthetic and real-world graphs. These results are compared to Dijkstra's algorithm on a modern CPU. We find that the NEURO-MAPP implementation scales favorably in terms of runtime for many graph types while consuming less energy per shortest-path query than the CPU implementation in almost all cases. These findings highlight the potential of neuromorphic hardware featuring sparse, spike-based communication as a scalable and energy-efficient platform for computation in graph search and related tasks.

Doctor agents are moving beyond single-turn answer generation toward evolving clinical decision systems. Within an outpatient episode, they acquire evidence, use examination and consultation resources, and decide when to finalize a diagnosis and management plan. Across episodes, their behavior may change through memory, retrieval, reflection, or other update mechanisms. Current evaluations only partially cover this setting. Fixed-input medical QA benchmarks score final answers from complete inputs, whereas many interactive benchmarks still focus on individual encounters or fixed runs, providing limited support for evaluating how episode-level decisions interact with cross-episode experience. We introduce MedEvoEval, an executable longitudinal evaluation framework based on action-gated simulated outpatient episodes. Each source case is converted into role-specific patient, examination, and manager views; evidence is revealed only through valid actions; and each episode records a structured trace that links observations, actions, final outputs, manager scores, and optional experience write-back. We release a runnable E&D artifact with 700 processed episodes, provenance notes, schemas, an episode runner, scoring scripts, configurations, example logs, analysis code, and trajectory- and step-level derivatives. Experiments show that episode traces expose process costs hidden by final-answer scoring, show how MDT-style consultation reallocates resources, and support longitudinal analyses of memory maturation, held-out transfer, update-stage response, and backward retention. Together, these results show that MedEvoEval provides a concrete basis for evaluating whether doctor agents improve through experience, transfer useful behavior, and retain earlier capabilities over time.

Knowledge updating in pre-trained Large Language Models (LLMs) remains an important challenge. While continual training provides a potential avenue for knowledge updating, it continues to present substantial technical difficulties. Furthermore, LLMs often struggle with accurately answering questions about specific factual information, such as news articles - a capability limitation widely recognized in the research community. This paper proposes PASTA, a simple yet powerful framework for integrating detailed factual information from news articles as new knowledge into LLMs, with the primary goal of building specialized models that accurately answer questions about this knowledge. Our framework combines data augmentation, question-answering generation, and a novel self-learning DPO process that simultaneously enables knowledge overwriting and hallucination suppression. We provide insights into effective knowledge updating through systematic analysis of learning parameters and data configurations. In our experimental evaluation with web articles published after the base model's knowledge cutoff, PASTA achieved remarkable improvement from 0.02 to 0.82 accuracy while maintaining general language capabilities, demonstrating its effectiveness for creating domain-specialized LLMs.

Synthetic aperture radar (SAR) data augmentation is important for improving the generalization of data-driven SAR interpretation models, yet practical augmentation workflows are often hindered by heterogeneous dataset formats, task-dependent metadata requirements, diverse generation methods, and weak validation of generated samples. This paper presents the \textbf{S}AR \textbf{A}ugmentation and \textbf{G}eneration \textbf{A}gent (SAGA), a schema-grounded and benefit-aware agent framework for task-oriented SAR data generation and augmentation. Given a natural-language request and heterogeneous SAR inputs, SAGA extracts observable dataset facts, validates executable dataset schemas, selects feasible augmentation strategies through validator-constrained planning, and compiles the selected strategy into an auditable augmentation workflow. Generated data are further assessed by quality, distribution, SAR-artifact, duplicate, leakage, and optional downstream-task evaluators to support evidence-qualified augmentation claims. By separating semantic proposal from deterministic validation and execution, SAGA improves the reliability and reproducibility of SAR augmentation decisions. Experiments on controlled agentic benchmarks and downstream SAR interpretation tasks show that SAGA improves schema grounding, skill planning, invalid-sample rejection, and downstream augmentation utility compared with rule-based, LLM-only, ReAct-style, and fixed-augmentation baselines.

Most work on efficient concurrent ordered indices, such as concurrent binary search trees, B-trees, skip lists, etc., has focused on data structures that provide good \emph{worst-case} guarantees. In real workloads, objects are often accessed at different rates, since access distributions may be non-uniform. Many efficient distribution-adaptive data structures exist in the sequential case; however, they are often complicated to make efficient in the concurrent case. The most prominent distribution-adaptive data structure is Splay Tree. Its most important advantage is that it does not store any balancing information and provides a reasonable performance improvement on extremely skewed workloads, such as Zipfian workloads. This paper proposes a splay-like rotation design for concurrent binary search trees. Instead of moving an accessed node to the root, rotations use two depth thresholds that are based on the static-optimality complexity computed from the number of accesses to the node: a node is rotated only when it is substantially deeper than the upper threshold, and rotations of the node stop before reaching the lower threshold. This design aims to preserve the main practical benefit of splaying on skewed workloads while reducing contention near the root. We present two variants of the rotation design: one using an exact 64-bit access counter per node and one using a 6-bit approximate counter. We prove static optimality for the corresponding sequential read-only tree and evaluate both rotation designs by implementing them on top of the concurrent AVL tree of Bronson et al. Our experiments show that the approach can improve throughput on several skewed workloads.

Large Language Models (LLMs) have shown the potential to generate code explanations that surpass those of peers in quality, offering promising opportunities for computer science education. While these explanations may not yet match the depth and clarity of instructor-provided explanations, research in computational creativity highlights that the quantity and diversity of ideas can often outweigh a singular focus on quality. Inspired by this, we explore whether combining multiple diverse explanations, each emphasizing distinct aspects (e.g., function, concept, goal), can enhance students' understanding of programming exercises compared to generic explanations that do not emphasize distinct conceptual aspects. In our study 971 first-year computing students were randomly assigned either diverse or generic LLM-generated explanations for two programming exercises. Students completed multiple-choice and open-ended questions for each exercise, followed by Likert-scale questions and open-ended reflections. Our findings outline patterns in student performance and perceived cognitive load across the two explanation conditions. These findings highlight how variation in explanation emphasis may relate to learner engagement and understanding. Across participants, open-ended response accuracy was consistently about 7.7% higher when students received diverse explanations, with no difference in perceived cognitive load.

Predicting student performance and characterizing metacognitive calibration are essential for personalization in intelligent tutoring systems. Prior research treats performance prediction, calibration error calculation, and variance decomposition as separate pipelines, preventing unified interpretation. I propose the Unified Behavioral Prediction and Calibration Analysis Pipeline (UBP-CAP), an integrated framework processing student pre-execution behavioral telemetry through three linked modules: (1) a LightGBM classifier with SHAP for binary correctness prediction, (2) formal calibration metrics (ECE, MCE, and Brier score decomposition) to evaluate metacognitive alignment, and (3) a crossed Generalized Linear Mixed-Effects Model (GLMM) for decomposing calibration deviations. I introduce the Predictive-Explanatory Divergence Index (PEDI), which quantifies structural divergence between predictive and explanatory feature profiles. Evaluated on 1,195 interaction records (27 students, 45 tasks), Logistic Regression achieves AUC-ROC = 0.903, outperforming LightGBM (0.878). Student naive ECE (0.109) significantly exceeds model ECE (0.068), confirming systematic miscalibration. The crossed GLMM yields ICCStudent = 0.123, showing calibration is situational rather than dispositional. PEDIcos = 0.081 (p = 0.327) indicates structural alignment between prediction and explanation on shared behavioral features.

The adaptive moment estimation algorithm, known as Adam, is widely used in modern machine learning, owing to its low per-iteration complexity and strong empirical performance. Despite its prevalent use, the theoretical foundation of Adam remains largely unexplored for time-varying and nonstationary systems. In fact, the existing theoretical analyses of Adam-type algorithms are primarily concerned with time-invariant model parameters and explicitly or implicitly rely on independent and identically distributed (i.i.d.) data assumptions, under which the learning taskcan be formulated as minimizing a fixed expected objective with a static minimizer. However, such assumptions are often violated in time-varying and nonstationary systems, thereby calling for a theoretical investigation beyond the conventional yet idealized i.i.d. setting. The main objective of this paper is to solve this challenging problem by establishing a general theory of Adam for time-varying and nonstationary stochastic systems. We will introduce some new techniques for analyzing the products of nonstationary and dependent random matrices induced by Adam's coupled first- and second-moment recursions, and will construct a new stochastic Lyapunov function that blends these two moment dynamics. Under a stochastic excitation condition that allows nonstationary and dependent data, we will derive both parameter tracking and output prediction error bounds explicitly, quantifying the effects of stepsize, first- and second-momentum parameters, gradient noise and parameter drift. These bounds not only provide guarantees for Adam performance, but also provide guidelines for hyperparameter selection. Experiments on both synthetic and real-world data validate our theory and design guidelines.

Long-context language models often conflate two different goals: compressing history into an efficient state, and maintaining reliable long-term memory. Linear, recurrent, and sparse attention reduce the cost of processing long sequences, but they do not by themselves specify when a fact should be written, overwritten, protected from distractors, or discarded. We study memory-managed long-context attention, a research route that separates a fast recurrent or sparse backbone from explicit editable request-local memory slots and query-time sparse fallback. Across structured synthetic tasks, token/chunk/sequence bridges, generated natural language, and local frozen-model diagnostics, pure fixed-state or pure sparse methods fail some overwrite, version, anti-pollution, or no-write-signal cases, while a hybrid covers both routes. A small 2,097,152-token mechanism stress test reaches 50/50 pooled accuracy with 2-132 active chunks. A 2.74M-parameter minimal causal event-token model reaches 595/600 with lite write supervision, supporting proof of trainability rather than scale. A six-family frozen-hidden-state bridge reaches 1079/1080 controlled pointer accuracy, but it uses generator-provided integer key IDs and separately encoded canonical key strings; it is an oracle-metadata probe, not open-text entity resolution. Local non-leaderboard RULER 4K diagnostics remain close to full context, whereas a 33-record LongBench v1 16K subset shows that naive lexical selection is not general. The evidence separates three claims: controlled slot lifecycle is feasible, sparse fallback is needed when writes lack future-query signals, and learned open-domain selection remains the main architectural bottleneck. We do not claim a final generative architecture, global slot-trajectory convergence, or systems superiority.

Predicting the aerodynamic performance (e.g. lift, drag, and moment coefficients) of an aircraft is challenging -- computational models are biased and direct simulations are prohibitive. A pragmatic way to overcome this limitation is by calibrating low-fidelity computational predictions with experimental measurements. This, however, requires calibrating against \emph{sparse} measurements contaminated with \emph{uncertainty} in both the control inputs and the measured aerodynamic response. We develop a methodology to address this problem based on Gaussian process surrogates and the classical Kennedy-O'Hagan calibration. A surrogate model learned on abundant-but-cheap low-fidelity data is calibrated with a sparse set of measurement data. Crucialy, we develop a Bayesian latent Gaussian process based approach that marginalizes the calibrated surrogate model over the input uncertainty, while also matching the marginal mean and variance of the measured output uncertainty. Once calibrated, our surrogate model predicts the uncertainty in aerodynamic coefficients with very high accuracy, including at extrapolative input settings. We validate our calibrated surrogate model predictions against measurement data with \emph{true} uncertainty intervals to demonstrate that the model places $94.2-95.8\%$ of its predictive samples inside the released $95\%$ truth intervals, with endpoint cumulative probabilities very close to the nominal 0.025 and 0.975 levels.

Firmware lies at the heart of IoT devices. Its development depends heavily on third-party libraries (TPLs), which greatly accelerate the process but simultaneously introduce associated vulnerabilities. Binary Code Similarity Detection (BCSD) is an effective technique for identifying vulnerabilities in firmware by comparing pairs of code segments. However, existing studies either evaluate their performance only on small-scale datasets or lack diversity in terms of vulnerabilities, TPLs, and firmware. Consequently, a comprehensive understanding of BCSD for real-world vulnerability detection remains absent. To bridge this gap, we conduct a large-scale study of vulnerability detection across 60,000 firmware images from 200 vendors using BCSD. Rather than introducing a novel model, we examine the influence of four key factors -- vulnerable function versions, vulnerability search space, function sizes, and compilation toolchains on BCSD performance. Our results reveal that these factors substantially affect performance, often by wide margins. To address this, we propose a build-aware query strategy that derives queries from representative real-world binaries, effectively closing the gap and raising the mean reciprocal rank (MRR) from 0.818 to 0.981. Furthermore, we demonstrate that a TPL-aware, two-stage search process significantly enhances accuracy, improving MRR by 18.5\% by limiting the search space.

Creative Commons licenses dominate African NLP corpus releases, but their compatibility rules are rarely applied. CC-BY-SA and CC-BY-NC cannot be combined in a single published dataset; a NoDerivs clause silently prohibits tokenisation and annotation. This paper audits the license provenance of over twenty corpus families used in African NLP, constructs a six-tier compatibility matrix, and applies it to three case-study languages: Kituba/Munukutuba, Zarma, and Moore. Four failure modes are documented with primary-source evidence: outright prohibition (JW300, removed from OPUS after a legal audit confirmed Terms of Service violation); composite license misrepresentation (WAXAL, whose CC-BY 4.0 claim is contradicted by its own HuggingFace dataset card); a NoDerivs clause hidden behind a CC-BY label (Tanzil); and data persistence failure (the Congolese Radio Corpus, where 402 of 405 source URLs are now dead). A pre-annotation due diligence checklist and a survey of legally clean enrichment opportunities close the paper.

AI systems increasingly exhibit behavior that differs systematically between evaluation and deployment contexts. Alignment faking, sandbagging, benchmark gaming, deceptive scheming, specification gaming, and trojans have each been documented separately, with each line of work characterizing one facet of what we argue is a single structural mechanism. We propose that this common mechanism is a defeat device, an engineering and regulatory concept long established in vehicle-emissions law and brought to broad public attention by the 2015 Volkswagen emissions case. A defeat device in an AI system has three necessary elements: a discriminator that detects evaluation context, a concealed swap that conditions behavior on detection, and a gap between eval-distribution and deployment-distribution performance on the stated evaluation criterion. We formalize this triadic test as a behavioral definition, organize documented cases along three taxonomic axes (origin, trigger, swap mechanism), propose Trigger-Axis-Aware Differential Probing (TADP) as a forensic detection protocol, and advance the claim that defeat devices can naturally emerge in current frontier AI systems without any operator engineering. We characterize naturally-emerging defeat devices as potentially one of the harmful emerging phenomena that AI safety practice should monitor and test for systematically. Implications for evaluation methodology, post-training pipeline design, interpretability research priorities, and AI governance follow.

While automatic tools for speech annotation are now commonplace within phonetic research pipelines, many tasks require substantial manual correction or training sets to perform accurately. Simultaneously, large speech models such as wav2vec2 have been shown to perform well at speech classification tasks, raising the question of how these models may be applied to phonetic annotation tasks. We introduce wav2VOT: a tool for the automatic estimation of voice onset time, closure duration, and burst realisation using wav2vec2. We demonstrate that wav2VOT performs comparably with current approaches on unseen datasets, and can estimate with high accuracy with fine-tuning. Analysis of wav2VOT predictions demonstrate high fidelity across stop voicing and place of articulation. These results demonstrate that large speech models are capable of producing accurate annotations, and further motivate exploration of large speech models as tools in phonetic research pipelines.

While AI holds the potential to revolutionize space life sciences, realizing this promise is contingent upon the systematic restructuring of heterogeneous spaceflight biological data into machine-actionable, AI-ready forms. Even though open access principles support human reuse and scientific reproducibility, this does not necessarily enable AI systems to access and analyze such a diverse set of scientific datasets. In addition, the growing array of AI approaches places distinct demands on data structure, metadata, and access interfaces. In order to respond to such growing changes we propose a three-tier approach, proceeding from FAIR to AI-ready to space-ready data. We discuss existing infrastructures and how they can be improved to close the AI access gap. We conclude by proposing a neutral international coordinating body as the governance backbone for the trustworthy, agent-accessible space biology infrastructure that deep space biological research will require.

The common factor analytic model is related to Helmholtz and Boltzmann machines, can be conceived as a linear autoencoder, or can be thought of as a single-hidden-layer generative neural network. We thus consider it a basal generative representation learner that can be used as a minimal model for studying the foundational characteristics of (deep) generative model architectures. We focus on the fundamental problem of indeterminacy in latent factor projections. This indeterminacy implies that, even when the intrinsic dimension of the latent vector is known, regularity conditions are met, and rotational indeterminacy is resolved, an inherent indefiniteness in the retrieval of causative latent sources remains: they will be uncertain, distributionally deviant, and non-unique. This can have major implications for data representation but remains an elusive issue, even to practitioners and theorists well-versed in the factor model. Moreover, this classic psychometric problem is intricately related to the modern issue of latent variable collapse in the variational autoencoder framework for deep generative modeling. Here, we assess this indeterminacy from various perspectives and show how these are mathematically and conceptually related and we discuss subsequent implications for the Psychometrics, Statistics, and Artificial Intelligence communities. We show that one has latent factor determinacy across all its facets when the feature-dimension grows to infinity. This feeds into an essentially distribution-free estimation approach in the sample case when the number of features grows very large. We conclude, as these are emergent properties at scale, that the factor model is suited for representation learning of very-high-dimensional data.

Fine-tuning a large language model is a ubiquitous method for enhancing its capability on a specific downstream task. However, prior work has shown that this increase in capability comes with a cost: it can increase a model's tendency to respond to unsafe adversarial prompts, even when fine-tuning with non-adversarial data. We present the first comprehensive empirical study of this phenomenon in multilingual settings by fine-tuning Llama-3.2, Qwen3, and Gemma-3 models using benign data translated across nine languages. We find that safety outcomes are highly sensitive to both the choice of fine-tuning language and the evaluation language, with adversarial compliance rates increasing four-fold in some settings. Multilingual safety drift is decoupled from general capability metrics, and occurs heterogeneously across languages and models. Fine-tuning in non-English languages often induces smaller internal representational drifts than English, but these shifts lead models to default to either exaggerated compliance or refusal. As such, assessing fine-tuning impacts solely in English provides inadequate assurance for deployment. To facilitate further research into these cross-lingual safety blind spots, we release the Multilingual-Benign-Tune dataset and the SORRY-Bench-Multilingual evaluation suite.

Large language models are increasingly capable of mathematical reasoning, but the proofs they generate are often unreliable and hard to verify. Interactive theorem provers such as Lean 4 address this by accepting only kernel-checked proofs; however, their reach is bounded by the formalized knowledge available. While Mathlib, a repository of formalized Lean 4 theorems that covers diverse mathematical areas, certain specialized areas remain underrepresented; notably, the domain of Combinatorics on Words (CoW). CoW studies sequences, exploring their properties such as periodicity, borders, conjugacy, and morphisms. As a result, specialized provers, trained on Mathlib-centered data, lack the lemmas to operate in CoW. We present two contributions. First, we introduce a Lean 4 formalization of CoW containing eight modules and \textbf{93} declarations of core definitions and foundational lemmas. Second, we present LAMP, a multi-agent framework that synthesizes kernel-verified Lean 4 proofs by providing explicit, structured domain knowledge at inference time through an ontology, rather than by fine-tuning a prover. LAMP coordinates a Planner, Builder, and Verifier with Model Context Protocol based access to a domain-specific CoW ontology. In a suite of 90 CoW theorems that span all eight modules and three difficulty levels, LAMP synthesizes verified proofs for 96.7% of theorems, substantially exceeding both an unscaffolded baseline and existing specialized provers. An ablation shows that removing LAMP's tool-grounded architecture or its Planner/Builder separation each cost roughly 12 percentage points, even with the backbone model held fixed.

We introduce the Contagion Tensor, a measurement framework for quantifying how large language model (LLM) output distributions couple across modalities, agents, and time steps. From the tensor we derive the Coupling Amplification Factor (CAF), a family of ratio-based metrics sharing the form CAF = E[T_condition] / E[T_baseline], providing unitless, baseline-referenced measurement with bootstrap confidence intervals. We instantiate CAF in four variants and evaluate the strongest in a complete 2x2x2 block-orthogonal simulation design with modality-specific ablation. The ablation reveals that an apparent image-condition super-linear effect (CAF = 1.40) collapses to sub-linear (CAF = 0.87) when the image perturbation module is disabled, a shift of -0.53 with zero effect on text conditions. We supplement with real-API experiments across two model families: DeepSeek-Chat (R=30) and GPT-4o-mini (R=15, real vision). Under uniform personas, text-only communication produces CAF approx 1.0 in both models. Diverse personas drive convergence (CAF = 0.88). A within-model comparison on GPT-4o-mini reveals: C3 (text) CAF = 1.02 vs. C5 (real vision, R=30) CAF = 1.72 [1.700, 1.733], delta = +0.70, validating the simulation's super-linear image-condition prediction. Of 11 conditions, 5 have been tested on real APIs and 6 remain unverified. Our contribution is two-layered: (1) a measurement instrument that makes output-distribution coupling quantitatively falsifiable; and (2) a transferable ablation protocol that any modular multi-agent simulator can adopt to distinguish genuine coupling from design artifacts.

Federated Learning (FL) emerged as a promising distributed machine learning paradigm. However, extending FL to the class incremental learning scenarios introduces unique challenges: 1) Capacity conflict and catastrophic forgetting from the shared model overloading, 2) Heterogeneity from Non-Independent and Identically Distributed (Non-IID) data, and 3) Synchronized class misalignment. In this paper, we propose \textbf{F}isher-Routed \textbf{M}i\textbf{X}ture of Experts for \textbf{Fed}erated Class-Incremental Learning (\textsc{FedFMX}), a novel framework to address these challenges via adaptive expert specialization across clients. The crucial insight is to route each sample to an expert subset that jointly optimizes knowledge acquisition and retention. Specifically, we introduce a Fisher-Routed Expert Scoring (FRES) module to estimate expert importance via Fisher-based stability cost and gradient-based plasticity gain. Then, we design an Adaptive Expert Selection (AES) module by quantifying marginal contributions for adaptive expert subset determination. Finally, by the routing-aware regularization (RAR), we achieve load balance and efficient FL training. We theoretically prove the $\mathcal{O}(T^{-1})$ convergence rate. Extensive experiments on multiple benchmarks compared with state-of-the-art methods demonstrate the superiority of \textsc{FedFMX}.

Quantum kernel estimation on near-term hardware is shot-budgeted: every entry of the kernel Gram matrix is a Bernoulli expectation that must be sampled with a finite number of circuit executions. Recent work on quantum kernel classification has shown that allocating shots non-uniformly across kernel entries, weighted by their downstream task sensitivity, can reduce the shot budget required to reach a target accuracy. We extend this idea to Gaussian process (GP) regression, a setting whose downstream quantities (full-spectrum posterior variance, log-determinant, marginal likelihood) couple to kernel error more tightly than the sign-only outputs of classification. We derive three closed-form pair-level sensitivities predictive coupling $|α_iα_j|$, leave-one-out residual, and marginal-likelihood gradient and plug them into a Neyman-style minimum-variance allocation rule. To prevent catastrophic over-concentration when the warm-up sensitivity estimate is itself noisy, we add a high uniform coverage floor justified by a Frobenius lower bound on the missing-entry perturbation. On four UCI benchmarks and two synthetic RBF + Bernoulli controlled studies, the resulting allocator delivers $10$--$21\%$ test-RMSE improvement over uniform allocation across the moderate-budget regime. The gain transfers (i) to genuine ZZ and Pauli-Z quantum kernels on quantum-natural data ($-13$--$15\%$ at low budget, $p<0.05$ paired) and (ii) to four downstream tasks (Bayesian quadrature, heteroscedastic regression, hyperparameter learning, multi-output Cokriging). On UCI features embedded into a ZZ kernel the gain disappears, consistent with the exponential-concentration regime where shot allocation has nothing to exploit.

Long-context LLM inference faces a fundamental conflict: head-adaptive compression algorithms (e.g., Top-$p$ nucleus sampling) offer superior accuracy by dynamically fluctuating memory budgets, yet modern inference engines (e.g., vLLM) demand rigid, static memory patterns to leverage CUDA Graphs and PagedAttention. We resolve this ``Static-Dynamic'' mismatch with HARD-KV, a unified framework that that bridges dynamic selection with rigid system constraints. HARD-KV introduces a Cascade Cache hierarchy, managing the token lifecycle across dense, sparse, and condensed tiers. Crucially, we propose a Logits Calibration mechanism that normalizes diverse importance metrics into a unified probability space, enabling consistent Top-$p$ budgeting across heterogeneous heads. To bridge the efficiency gap, we offer a system-level solution, which rewrites fragmented, dynamic indices into contiguous physical layouts compatible with high-performance inference engine. Extensive experiments on math-reasoning benchmarks (AIME, U-Math) verify that HARD-KV achieves up to 2$\times$ throughput improvement over static baselines while maintaining high-fidelity generation in 10k+ token scenarios. Code is available at https://github.com/SuDIS-ZJU/HARDInfer.

This paper develops a continuum theory of exit-and-join coalition dynamics in nonatomic cooperative games. We extend the Aumann-Shapley value and the Aumann-Drèze value to coalition structures in which each coalition is treated as a restricted nonatomic game, yielding a marginal-contribution-based payoff density that governs incentives for agents to remain in, exit, or join coalitions. We derive deterministic mean-field dynamics from decentralized switching rules and show that payoff-difference switching recovers replicator dynamics as a special case. We characterize exit-and-join equilibrium by the absence of profitable positive-mass deviations and prove its equivalence with stationarity of the induced mass dynamics under incentive-compatible and strictly payoff-responsive switching rates. For mass-based cooperative games, we construct a Lyapunov function and establish global convergence under strict concavity. We further show that the equilibrium is equivalent to a Wardrop equilibrium of an induced nonatomic population game and admits a variational inequality formulation. The framework is extended to incorporate switching costs and endogenous coalition acceptance rules, leading to constrained equilibria characterized by quasi-variational inequalities. The proposed theory unifies cooperative value allocation, noncooperative coalition mobility, mean-field dynamics, evolutionary game theory, and population games within a common framework for analyzing coalition formation and adaptation in large-scale multi-agent systems.

Recent large language models (LLMs) achieve strong performance on entity matching without requiring task-specific training data. However, applying these models to large sets of candidate pairs remains slow and costly. In contrast, entity matchers using traditional machine learning methods or small language models (SLMs), such as RoBERTa, offer much faster inference but require task-specific training data. This paper investigates whether the need to provide task-specific training data can be avoided by using knowledge-distillation workflows, in which an LLM serves as a teacher model to label training pairs that are subsequently used to train a smaller student model. We investigate knowledge distillation for entity matching along the following dimensions: pair-selection strategy, teacher model, label post-processing method, and student model. We evaluate the workflows using the Abt-Buy, Walmart-Amazon, WDC Products, DBLP-ACM, and DBLP-Scholar benchmarks, and compare the performance of student models trained with machine-labeled data to the performance of the same models trained using the benchmark training sets. Our experiments show that student models trained using the machine-labeled sets perform approximately on par with models trained on the benchmark training sets, with the remaining differences in both directions staying below two F1 points. Using GPT-5.2 to label the training sets for all five benchmarks costs US\$28.31 to US\$40.88, whereas manually labeling the same training sets is estimated to require 470 hours of work. At inference time, Ditto is 41.5 to 534 times faster than directly using an LLM to perform the matching tasks. These results indicate that current LLMs, when combined with a suitable pair-selection method, can substantially reduce or even eliminate the manual effort required to label use case-specific training data for entity matching.

Mathswitch is an open-source project that imports mathematical concept records from sources such as Wikidata, Wikipedia, MathWorld, Encyclopedia of Mathematics, nLab, ProofWiki, and Agda-Unimath, and links records that refer to the same concept. It does not reorganize or redefine the imported content; each source retains its own structure. The current focus is on importing concept data from Wikidata and the resources it links to, with plans to expand to further sources and better concept linking. Because the concept set is approximated through queries over Wikidata's collaboratively edited graph, the imported data is noisy: some items are non-mathematical, while others are ambiguous. In this paper, we test whether a voting ensemble of LLM judges can filter this noise. We evaluate it on Wikidata items with known MathWorld identifiers as a positive control, and examine how classification changes when database identifiers are removed from context. We then inspect the cases where the judges disagree with MathWorld and group these disagreements into three categories (degenerate descriptions, narrow scope bias, and editorial-scope mismatches) that suggest different remediation strategies.

Human videos are a scalable source of supervision for robot manipulation, as they are abundant and naturally capture rich object interactions. However, transferring human demonstrations to robots remains challenging due to embodiment mismatch, scene variation, and robot-specific feasibility constraints. We present Human2Any, a framework for learning reusable object-centric interaction priors from human videos without requiring real-world robot demonstrations in the target task contexts. Human2Any represents manipulation through object-object interaction motion, capturing task-relevant scene changes while abstracting away embodiment-specific details. It composes learned interaction priors with robot-side feasibility reasoning and motion planning, allowing the same human-derived knowledge to adapt to different embodiments, scene geometries, and task contexts. We validate Human2Any across diverse manipulation settings, including real-world experiments on a Franka tabletop setup and an RBY-1 humanoid mobile robot, demonstrating robust interaction-centric manipulation without real-world robot training data. Project website: https://human2any.github.io/.

We study the leading-order fluctuation of stochastic gradient Euler-Maruyama estimators for generalized non-reversible Langevin dynamics. Under structural assumptions tailored to the small-stepsize central limit theorem and under an unbiased stochastic gradient oracle, we prove that the empirical average over a horizon of order the inverse squared stepsize satisfies a central limit theorem in the vanishing-stepsize regime. The limiting variance is characterized through the Poisson equation of the limiting full-gradient diffusion. We then rewrite this constant in an operator form that links it to the continuous-time asymptotic variance and, under standard operator-theoretic assumptions, derive a sufficient condition under which an anti-symmetric perturbation strictly reduces the leading-order fluctuation constant relative to the reversible baseline. We also identify bounded smooth predictive observables that re directly covered by the main theorem. As a separate Gaussian calculation beyond the bounded-test-function regime, we obtain closed-form formulas for quadratic Hamiltonians and linear observables. The framework covers non-reversible Langevin dynamics and augmented-state examples including Hessian-free high-resolution dynamics and a positive-definite subclass of gradient-adjusted underdamped Langevin dynamics that allow stochastic gradients. Numerical experiments on basic examples and Bayesian linear regression using synthetic data, and Bayesian logistic regression using real data support the predicted Gaussian fluctuations and show that the non-reversible schemes consistently reduce the root mean squared error (RMSE) relative to their reversible baselines.

Objective: ICD codes are central to reimbursement, research, and population health surveillance, yet automated coding systems often struggle to integrate diagnostic signals from both clinical narratives and structured electronic health record (EHR) variables. We evaluated whether frozen medical large language model (LLM) representations can serve as a shared embedding space for multimodal primary diagnosis category prediction. Materials and Methods: We constructed a MIMIC-IV cohort of 13,645 admissions from the 10 most frequent primary ICD-10 codes, consolidated into seven categories. Structured variables were serialized into clinical narratives and combined with leakage-pruned discharge notes. Using a frozen MedFound-Llama3-8B-finetuned backbone, we extracted hidden states from five transformer layers and trained linear probes for structured-only, unstructured-only, and combined inputs, comparing against XGBoost and information-matched PLM-ICD baselines and evaluating MIMIC-III adaptation with a compact bottleneck adapter. Results: The combined probe performed best on MIMIC-IV (87.69% strict; 91.45% medical accuracy), exceeding both single-modality probes and baselines. The structured-only probe outperformed its standard baseline by 6.19 points in medical accuracy. Diagnostic information became increasingly linearly separable in deeper layers, and a 2M-parameter adapter restored cross-dataset transfer to MIMIC-III using only 5% of target labels. Discussion: LLM embeddings can unify structured and narrative EHR information for multimodal diagnosis prediction, supporting efficient reuse of clinical representations across modalities and datasets through a small representation-level module. Conclusion: Multimodal probing of frozen medical LLM representations provides a practical approach for studying EHR modalities and adapting clinical representations across datasets.

Government documents in India are predominantly issued in regional languages such as Marathi, creating substantial accessibility barriers for non-native readers, interstate administrative bodies, and policy analysts. Although recent advances in neural machine translation have improved sentence-level translation quality, existing systems largely neglect document structure, formatting integrity, and domain-specific terminology, thereby limiting their applicability to official documentation. This paper presents a structure-preserving Marathi-to-English government document translation framework capable of performing end-to-end document transformation while maintaining layout fidelity. The proposed system integrates layout-aware optical character recognition, coordinate-based text extraction, large language model based translation, and structured document reconstruction through HTML representations. By enforcing spatial alignment constraints and preserving hierarchical document elements, the framework ensures structural consistency between the source and translated documents. Experimental evaluation on real-world Marathi government PDFs demonstrates improved structural preservation, translation coherence, and terminological consistency compared to conventional text-only translation pipelines. The proposed framework contributes toward scalable multilingual accessibility solutions for e-governance and administrative document processing.

Machine Learning (ML) algorithms, such as k-Nearest Neighbours (kNN) or random forest, eschew the ideal of true data models in favour of predictive performance. However, minimising the MSE or F-score cannot lead to unbiasedness directly, which is important in many situations such as official statistics. We study the conditions of algorithmic ML, other than the existence and knowledge of true data models, which lead to unbiased prediction or classification for a given finite population, including how the training data may be sampled from the population, how a trained prediction algorithm can be tuned to achieve unbiased prediction or classification for that population, and how the performance of out-of-sample prediction or classification can be assessed unbiasedly. The inference is based on the known probability design of samples and training sets, rather than any assumed distributions or models.

Software engineering is experiencing its most significant transformation since the emergence of high-level programming languages. As large language models (LLMs) increasingly enable sustained, multi-step, tool-mediated execution, engineering value is shifting from writing deterministic code to supervising probabilistic and autonomous behavior. This paper argues that AI-Native Software Engineering is a paradigm shift rather than a mere tooling advance, creating a new professional archetype: the Agentic Engineer, whose primary artifact is the agentic system rather than the program. We characterize this transition through three changes: (i) the unit of work shifts from functions to supervised agent workflows, (ii) correctness shifts from binary assertions to statistical evaluation under uncertainty, and (iii) accountability shifts from code authorship to outcome ownership. Drawing on post-2022 research, we compare traditional and agentic engineering roles and define core mechanisms of autonomous agents, including reasoning-acting loops, context engineering, tool use, memory, behavioral drift, and compositional error. We place human-AI collaboration within socio-technical frameworks and examine mixed empirical evidence. While some studies report productivity gains, others show slowdowns among experienced developers, highlighting disciplined oversight rather than automation as the critical competency. Using established governance frameworks, we identify required skills and risks, including indirect prompt injection. We conclude that the future is one of symbiosis rather than substitution: agentic engineering builds upon and depends on classical software engineering principles.

Artificial intelligence reaches the land registry not as another tool but as a value chain that turns data into intelligence and intelligence into economic value. This paper argues that the decisive legal move is to place validity, a functional, second-order concept, at the centre of that chain. Rights, liability and supervision organise around it. It traces three impacts.Registry information becomes smart data, governed simultaneously by registry law, the GDPR, the European data acts and the AI Act. Control emerges as the operative concept for digital representations of real estate, whose proprietary effect depends on anchoring to the register. In a hybrid society of human and artificial agents, the registry becomes the public node of validity, with blockchain complementing rather than replacing it. Across three legal cultures, the registra's value migrates from processing documents to guaranteeing validated data,making validity an asset for the UNO Sustainable Development Goals.

Multi-Frequency Electrical Impedance Tomography (MF-EIT) is a non-invasive, low-cost modality that reconstructs electrical property distributions from boundary voltages. For stroke imaging, progress in 3D deep-learning reconstruction is limited by the lack of large-scale datasets with paired ground-truth (GT) volumes and by non-standardized pipelines for data generation, simulation, and evaluation. We introduce BREIT, a modular framework for 3D MF-EIT stroke reconstruction providing: (i) a neuroimaging-to-EIT pipeline that converts CT/MRI into frequency-dependent GT admittivity volumes; (ii) a self-contained Python 3D Complete Electrode Model (CEM) forward solver for simulating MF-EIT voltages; and (iii) a 3D D-bar implementation supporting non-uniform electrode layouts. Building on BREIT, we propose dFNO-bar, which integrates Fourier Neural Operators into D-bar by learning a mapping from scattering data $t(ξ)$ to conductivity $σ(x){=}\Re\{γ\}$. We evaluate dFNO-bar against D-bar, Deep D-bar, and Gauss--Newton reconstructions on UCLH-matched synthetic data, and observe higher brain SSIM with comparable CC across noise settings. Code and data are publicly available at: https://github.com/djahiddj13/BREIT

Every existing vector database and agent memory framework treats memory as passive storage that agents query explicitly. No system propagates knowledge between agents through the memory layer itself. We introduce HyphaeDB, an agent-native memory infrastructure that reinterprets the Hierarchical Navigable Small World (HNSW) graph topology the data structure at the core of every modern vector database not as a search optimization, but as a communication fabric for multi-agent AI systems. In HyphaeDB, agents are nodes in the vector space with persistent positions, knowledge propagates via a gossip protocol through the graph's neighbor structure with energy-based attenuation, and emergent behaviors contradiction detection, pattern crystallization, and consensus formation arise from the combination of topology, propagation dynamics, and local interaction rules. We present the architecture built on three primitives (knowledge nodes, topology edges, and memory diffs), a multi-layer abstraction hierarchy with promotion via emergent consensus, and theoretical analysis grounding the system in small-world network theory, epidemic broadcast protocols, and swarm intelligence. We provide a reference implementation on PostgreSQL with pgvector and describe a concrete deployment in Swarm-Driven Development, a multi-agent software engineering methodology. HyphaeDB represents, to our knowledge, the first system to combine navigable small world topology with gossip-based knowledge propagation for multi-agent coordination.

In next-generation wireless networks, the growing density of devices and limited spectrum resources pose severe jamming challenges to fragile legitimate communication links in the wireless electromagnetic environment. Crucially, when jamming overlaps with pilot and data symbols in both time and frequency domains, it inflicts a severe bottleneck on receiver-side joint estimation and detection. Existing schemes often lack an effective framework to combat such jamming contamination, thereby failing to guarantee reliable transmission. To address this issue, we propose a Brownian bridge diffusion-based joint channel estimation and data detection framework (BBD-JCED) for jamming-resilient receivers. Specifically, the proposed framework comprises two core modules: the first extracts jamming features in the short-time Fourier transform (STFT) domain and suppresses jamming samples, thereby improving the signal-to-jamming-plus-noise ratio (SJNR) of the received signal; the second introduces a Brownian bridge diffusion (BBD) process to model the evolution of the suppressed signal and the encoded bits in the presence of channel estimation errors, thereby enabling enhanced joint channel estimation and data detection. To alleviate the computational burden of the BBD process in the second module, we further derive a fast ordinary differential equation (ODE) solver that enables its low-complexity iterative evolution. Finally, we design a multi-module training algorithm to improve the data recovery capability of the proposed framework. Simulation results demonstrate that the proposed framework achieves superior bit recovery performance compared with baseline schemes while maintaining a lower number of model parameters and competitive computational complexity.

Hate speech annotation pipelines routinely collapse annotator disagreement into majority vote labels before training. We show that this aggregation is not neutral: 42.6% of all annotator disagreement in HateXplain concentrates specifically at the hate/offensive boundary, a pattern consistent with annotators applying different thresholds for where hate begins (chi-squared = 135.199, df = 2, p < 0.0001). Both a hard-label BERT model (Model A) and a soft-label model (Model B) drop 22 percentage points in accuracy from agreed posts (~80%) to disagreement posts (~58%), confirmed at p < 0.0001. A per-annotator multi-head model (Model C) widens this gap further to 28 points while collapsing offensive disagreement accuracy to 0.245. Critically, Model A expresses significantly higher confidence on boundary case errors than Model C (0.710 vs. 0.495, p < 0.0001), meaning standard evaluation metrics will not detect the failure. Three downstream interventions of increasing sophistication all fail to recover boundary accuracy. We argue the problem is structural. Majority vote presents a contested judgment as ground truth, and models inherit that false certainty. The intervention must be upstream in annotation design.

Large Language Models (LLMs) have demonstrated the ability to simulate human-like OCEAN personality traits in generated text. Previous efforts have focused on prompt engineering or fine-tuning to shape LLM personality. In this work, we propose a mechanistic interpretability approach that directly intervenes on the model's latent features. Our method identifies latent directions in the residual stream corresponding to a target OCEAN trait using sparse autoencoders (SAEs) and contrastive activation analysis. We formalize an additive steering vector in activation space and demonstrate how applying a small additive shift to the hidden states enhances the target trait while preserving overall language modeling performance. To determine the optimal combination of feature shifts, we explore a linear weighting heuristic with grid search optimization that balances personality expression with task performance. Our approach shows promise in controllably steering personality traits at the mechanistic level while maintaining high performance on standard benchmarks.

Emotional body motion expressions are an essential element of non-verbal communication. Effectively conveying these expressions through technology is of utmost importance, for example, with virtual reality avatars and in social robotics. Recent advances in generative models have opened new opportunities for advancing research on emotional body motion learning. However, generating accurate emotional expression representations is challenging, given the subtlety of emotional cues, individual variability, and cultural differences. We investigate whether a generative model can implicitly learn emotional body motions directly from culturally grounded motion-capture data, without explicit emotion-motion guidance. Using a dataset of emotional performances by 49 Japanese actors, we trained a Transformer-based generative model to generate expressive motions conditioned on 13 discrete emotion labels. We evaluate the generated motions from two perspectives: (1) an LSTM-based classifier to assess recognizability by machine observers, achieving a recognition accuracy of 22.80%, and (2) a human perception study with Japanese raters to assess alignment with human affective interpretations, yielding a recognition accuracy of 24.91%. Beyond these, we evaluate the utility of generative modeling for three practical tasks: augmenting emotion recognition models, extracting representative emotion-specific motion patterns, and synthesizing smooth transitions between emotion intensities. Our findings highlight the potential of implicit, data-driven generative modeling to enhance affective computing applications and our understanding of emotion expressions.

Hierarchical decision-making frameworks are pivotal for addressing complex control tasks, enabling agents to decompose intricate problems into manageable subgoals. Despite their promise, existing hierarchical policies face critical limitations: (i) reinforcement learning (RL)-based methods struggle to guarantee strict constraint satisfaction, and (ii) optimal control (OC)-based approaches often rely on myopic and computationally prohibitive formulations. To reconcile these trade-offs, hierarchical RL-OC architectures have emerged as a promising paradigm. However, the formulation of the lower-level optimization within these frameworks remains underexplored, often relying on heuristic or myopic objectives. In this work, we propose a principled framework that systematically integrates upper-level goal abstraction with structured lower-level decision making. We adopt an inverse optimization approach to inform the structure of the lower-level problem from expert demonstrations, ensuring that the objective of the lower-level policy remains aligned with the overall long-term task goal. To validate the approach, our framework is evaluated on distinct decision making tasks: network-based resource allocation and continuous collision avoidance. Empirical results demonstrate that our method consistently outperforms strong baselines based on end-to-end RL, learning-augmented optimal control, and existing hierarchical RL approaches in both efficiency and decision quality.

Predicting future states is essential for autonomous agents, yet current Vision-Language-Action (VLA) models fundamentally lack this capability, relying instead on reactive perception-action mapping. While integrating Predictive World Models (PWMs) addresses this gap, existing approaches either incur prohibitive cascaded latency or act as shallow terminal tasks that fail to deeply embed forward-looking reasoning. To endow VLA models with this reasoning capability, we propose X-Mind. Rather than treating PWMs as an external auxiliary module, this framework internalizes them as the Visual Chain-of-Thought (Visual CoT). By enforcing a world rollout prior to action, the model is constrained to imagine future evolution first, yielding a driving policy that is robustly grounded in environmental dynamics and aware of the future consequences its actions will unfold. The challenge here is efficiency, and we tackle it on two fronts. First, we introduce a compact representation of visual thinking: an abstract sketch that fuses a Bird's-Eye-View (BEV) layout with abstract driving priors (e.g., navigation intents and traffic rules). Rather than rolling out dense future frames, the model reasons over this sketch as a mental canvas; aided by a Deep Compression Autoencoder (DC-AE), a 12-frame future rollout is reduced to merely 96 tokens, alleviating the long-context computational bottleneck. Second, to accelerate generation further, we propose a recurrent block diffusion scheme that unrolls the denoising steps across the layers of the large drive model, folding iterative refinement into the backbone's one forward pass. Trained and validated on large-scale real-world data, X-Mind achieves competitive end-to-end driving performance, which makes it a highly practical, low-latency solution that successfully deploys large-scale cognitive reasoning directly onto resource-constrained vehicle platforms.

The increasing communication complexity of large-scale heterogeneous systems has motivated runtime methodologies for communication-aware workload placement and routing optimization. These communication limitations are addressed in this paper by proposing SHIFT, a novel topology-agnostic approach that transfers compute node context and data to a more suitably positioned node, rather than only shifting data as in conventional networks-on-chip. The proposed strategy is evaluated on a chiplet-based architecture utilizing a fine-pitch integration platform featuring multiple bandwidth-domains for heterogeneous workloads. The proposed architecture employs multi-layered routing between functional or memory chiplets and utility chiplets, which serve as intelligent nodes for routing and compute relocation. Adaptive scheduling and routing utilize a modified shortest-path algorithm for large-scale systems, complemented by a lightweight ML-assisted policy that infers traffic conditions to improve adaptivity. To establish a performance baseline, the initial assessment uses random instruction vectors and data patterns to evaluate the fundamental capabilities of SHIFT. Simulation results exhibit successful relocations over total trials ranging from 75.2% to 97.9% across configurations, with average latency improvements of 16.4%-62.5% and a maximum of 76.8%. In addition, throughput is improved by up to 12.5x, power dissipation per unit area is reduced by ~8%, energy-per-bit is reduced by up to 58.3%, and performance is improved by 18%. To evaluate efficiency under high logic and data density, the framework was tested on standard LLM workloads. Results exhibit average improvements of 4.9x, 5.9x, and 1.8x in runtime, throughput, and energy-efficiency, respectively, surpassing state-of-the-art wafer-scale LLM services and demonstrating compatibility with large-scale platforms and applications.

We propose a path-space formulation of prediction in AI world models. Rather than sequences of one-step conditional distributions, we argue that a world model implicitly defines a probability measure over future trajectories. In the local regime where latent dynamics admit an effective Markovian description, this path measure takes the Onsager-Machlup form. Within this framework, prediction (most probable trajectory), planning (constrained optimization), and uncertainty (fluctuations) emerge as operations on a single action functional. We decompose the latent dynamics into reversible and irreversible components and introduce operational measures of entropy production from model rollouts. In controlled small-scale attention-based models, we find that attention asymmetry is acquired during training in proportion to the irreversibility of the data. Symmetrizing the learned attention suppresses entropy production and selectively degrades long-horizon prediction of irreversible dynamics while preserving relaxational prediction. These results suggest that irreversibility may serve as a computational resource for predictive world models. More generally, the fundamental predictive object is a distribution over future paths rather than states.

Although most undergraduates now use large language models (LLMs), a form of generative artificial intelligence (GenAI) for academic writing, no validated method distinguishes the qualitatively different ways students rely on them. Existing instruments assess reliance solely by frequency of use, a measure that, as this study shows, inadvertently rewards dependence on AI rather than recognizing students' own intellectual contribution. Conducted at a public minority-serving university and grounded in the AI Literacy Framework, Expectancy-Value Theory, and Biggs's Presage-Process-Product model, the study drew on 382 undergraduates, 14 interviews, and 396 open-ended survey responses. Four distinct reliance types were identified and confirmed: Strategic (34.3%), Instrumental (30.9%), Dialogic (30.4%), and Dependent (4.5%). Students' value and cost beliefs predicted the intensity of their reliance on LLMs, whereas their AI literacy predicted the type of reliance they adopted, indicating that differentiated support is needed. Notably, Strategic users, those who engaged AI most deliberately, scored lowest on standard outcome measures. This pattern reflects a limitation of current instruments, which index AI's contribution rather than writing quality, thereby penalizing students who show the greatest independent thinking. Analysis also revealed an additional group, roughly 13%, who declined to use AI for ethical rather than practical reasons, and who existing frameworks overlook. These findings carry implications for AI literacy programs, the measurement of student learning outcomes, and equitable AI policy at minority-serving institutions.

Recent artificial intelligence (AI) systems have shown remarkable progress in mathematical reasoning. Many existing approaches, including large language models (LLMs), draw on human prior knowledge in the form of mathematical text, code, or theorem libraries. Although these approaches are highly effective in practice, it remains an open question whether an agent can autonomously discover useful theorems without such human priors. We study this question in a formal axiomatic system by developing an agent that starts from axioms and inference rules alone and gradually grows a library of useful theorems. Concretely, we propose a self-supervised theorem-discovery algorithm that alternates between proof search and useful-theorem extraction, building a theorem library whose entries are reused as lemmas for subsequent proof search. Experiments show that the agent discovers tens of thousands of theorems and finds proofs for human-written benchmark problems, suggesting that its discoveries include theorems meaningful from a human mathematical perspective. Furthermore, the discovered theorems improve LLM proof performance when provided as prompt lemmas, indicating that they can serve as external knowledge for LLM reasoning. Our results provide evidence that useful theorems can emerge from proof search without relying on human-provided theorem libraries. More broadly, they suggest a path toward self-evolving AI systems for mathematics whose discoveries remain formally verifiable.

Variational Quantum Eigensolvers (VQEs) are central to quantum computing, yet testing them remains challenging due to the oracle problem: the ground-state energy they compute is itself unknown. Existing approaches, such as convergence-based testing, are unreliable and yield high false-positive rates due to optimisation instability. We propose METAMORPHQ, a metamorphic testing framework that derives test oracles directly from quantum mechanical properties of VQE circuits. Exploiting algebraic properties of parametrised rotation gates and diagonal Hamiltonians, we define five physics-based invariants that hold for any correct circuit and can be verified at initialisation without ground-truth outputs. Evaluated on 500 benchmark circuits with 2,469 mutants, METAMORPHQ achieves zero false positives and significantly improves diagnostic effectiveness (Youden's J = 0.57 vs. 0.02 for convergence testing). These results demonstrate that physics-derived invariants provide a practical, oracle-free foundation for testing quantum software, enabling reliable validation of both human- and LLM-generated circuits.

Large language models increasingly act as agents: they call tools, move money, delete records, and send messages on a user's behalf. To keep them safe, practitioners imported the chatbot-era recipe (train the model to refuse unsafe inputs) into the agentic setting, and treat the resulting capability loss as a manageable ``alignment tax.'' We argue this is a \emph{category error}. Refusal is a primitive for \emph{content safety}, where the harm is in the model's output and is therefore a learnable function of it. Agentic harm is different in kind: it lies not in any output but in the relation between the authority an action exercises and the authority the user granted, which is absent from the text the model sees. Importing content-safety methods into this regime does not trade capability for safety; it pays capability and buys negative security. We support this with three lines of evidence spanning the autonomy spectrum: defense-trained models learn surface patterns rather than intent; the same training collapses multi-step agents before any threat appears while leaving them exploitable; and even undefended frontier models exceed granted authority under ordinary use. We conclude that action safety cannot be installed in weights. It must be expressed as \emph{least privilege}, enforced \emph{outside} the model at the action boundary, and evaluated as \emph{action alignment} (a relational, deployment-conditioned property) rather than a refusal score.

We introduce 5ting, our system for the SemEval2026 Task 8 (MTRAGEval), which evaluates multi-turn Retrieval Augmented Generation (RAG) systems. Multi turn RAG involves context drift, under specification, and hallucination risk. Our system combines BGE-M3 dense retrieval with FAISS indexing, dual-query merged retrieval, and LLM based reranking, followed by role separated generation constrained to retrieved evidence. The retriever achieved nDCG@5 = 0.4719 in Task A, while the end to end system ranked in Task C with a harmonic score of 0.5597 and RL_F = 0.7692.

LLM agents are expected to act over multiple turns, using search, browsing interfaces, and terminal tools to complete user goals. Yet not every goal is well specified or achievable in the available environment. In such cases, a reliable agent should recognize that further interaction is unlikely to help and abstain from additional tool calls. We define Agentic Abstention, the problem of deciding when an agent should stop acting under uncertainty. Unlike standard LLM abstention, which is usually evaluated as a single-turn answer-or-abstain decision, agentic abstention is a sequential decision problem: an agent can answer, abstain, or gather more information at each turn, and the need to abstain may only become clear after interacting with the environment. We study this problem across web shopping, terminal environments, and question answering, evaluating 13 LLM-as-agent systems and 2 agent scaffolds on more than 28,000 tasks. Our results show that the main challenge is not only whether agents can abstain, but also when they abstain. Some agents never abstain when they should, while others do so only after many unnecessary interactions. This gap is especially large on tasks where the instruction appears feasible until the environment reveals otherwise (e.g., no valid result matches the instruction). We further find that model scale, reasoning, and agent scaffolding affect abstention in different ways, where larger or more capable models sometimes perform worse at timely abstention. Finally, we introduce CONVOLVE, a context engineering method for improving agentic abstention that distills full interaction trajectories into reusable stopping rules. On WebShop, CONVOLVE substantially improves timely abstention without updating model parameters, raising Llama-3.3-70B's timely recall rate from 26.7 to 57.4. Our dataset and code are available at https://lhannnn.github.io/agentic-abstention

Leveraging large-scale weakly supervised datasets is crucial to train robust end-to-end automatic speech recognition (ASR) models. However, such datasets often contain noisy labels and lack domain specificity, limiting their effectiveness. To address these issues and make better use of weakly supervised datasets, we propose a novel training approach incorporating data filtering and selection. Our approach consists of three steps: pretraining on the entire dataset, continued pretraining on a filtered subset based on character error rate (CER), and fine-tuning on a small number of acoustically similar samples to the target domain, selected from the filtered subset. In experiments with a 90,000-hour weakly supervised Japanese dataset, the proposed filtering and selection methods synergistically reduced CER by up to 6.4% and 4.0%, respectively, even though these steps reused training samples already used in the first pretraining step.

Automated toxicity moderation systems operate in dynamic online environments where harmful behavior evolves through coded language, shifting targets, and strategic adaptation to enforcement. Existing drift detection methods often focus on global distributional change, but such signals may miss safety-relevant shifts that emerge in localized harm subspaces or high-risk model-error regions. This paper introduces DriftGuard, a safety-aware adaptive moderation framework that combines multi-monitor drift detection with selective model updating. The framework tracks global text drift, identity-harm drift, model uncertainty, toxic-risk drift, and false-negative-risk drift. When safety-relevant change is detected, the model is updated using a hard-mix adaptation set that prioritizes likely false negatives, identity-related high-risk examples, false-positive-risk examples, and uncertain boundary cases. Experiments on Civil Comments temporal shift and Jigsaw-to-DynaHate cross-dataset shift show that safety-aware monitors detect risks missed by global drift alone. Hard-mix adaptation improves toxic recall and accuracy over no-update and random-balanced baselines, raising toxic recall to 0.8777 on Civil Comments and from 0.7107 to 0.8523 on DynaHate. Bootstrap analysis further shows stable DynaHate safety gains, with toxic recall increasing by 0.1418 and false-negative prevalence decreasing by 0.0781. Overall, DriftGuard links safety-aware drift detection to targeted, lightweight model updating for more robust adaptive toxicity moderation.

Understanding and localizing subtle changes between paired images is critical for tasks such as surveillance and image editing. However, traditional Image Change Captioning (ICC) methods lack spatial grounding, limiting their precision. We introduce Image Change Captioning and Segmentation (ICCS), a new multimodal task that jointly requires structured change description and pixel-level localization. To address ICCS, we propose the Change-aware Captioning and Reasoning Chain (CCRC), a dual-chain framework that decouples semantic reasoning from spatial segmentation. The first chain, Chain-of-Change-Captioning (CCC), enhances fine-grained change perception via a visual fusion module based on Multi-Head Change-aware Attention inserted between the visual and language components of a Multimodal Large Language Model (MLLM). CCC also determines whether a change is segmentable. If not, it alone generates the caption. Otherwise, the second chain, Chain-of-Change-Segmenting (CCS), is activated, leveraging spatial priors from CCC and refining masks with a Change-aware Token Refiner for accurate boundary localization. We evaluate CCRC on both synthetic and real-world change detection benchmarks with pixel-level supervision. Experiments show CCRC achieves state-of-the-art performance.

Test-time adaptation (TTA) of vision-language models (VLMs) is essential for their robust deployment in dynamic, real-world environments. However, existing TTA methods often adapt locally without accumulating knowledge over time, or operating within a single modality without exploiting VLMs' inherently multi-modal nature. Inspired by the \textbf{Com}plementary \textbf{Mem}ory systems of the biological brain, we propose \textbf{ComMem}, an innovative approach that mimics the distinct but cooperative roles of the hippocampus and neocortex to enable effective TTA for VLMs. ComMem consists of two key components: a fast-adapting detailed memory, akin to the hippocampus, that forms a dynamic visual cache from high-confidence test samples; and a slow-integrating abstract memory, akin to the neocortex, that continually refines global textual prototypes. For each test instance, ComMem jointly optimizes both memory systems to ensure cross-modal consistency. Extensive experiments on 15 benchmark datasets show that ComMem significantly outperforms state-of-the-art methods under both natural distribution shifts and cross-dataset generalization, offering a promising direction for enhancing VLMs' practical adaptability.

The robustness of trajectory prediction models is crucial for developing safe autonomous driving systems. Adversarial attacks on trajectory prediction can significantly impair the accuracy of predicted trajectories, leading to hazardous driving behaviors. While heuristic defense strategies have been implemented to enhance the robustness of trajectory prediction models, these measures often fail against more sophisticated, targeted adversarial attacks. Hence, there is a pressing need to establish verifiable safety assurances for trajectory prediction models. In this paper, we extend the traditional Randomized Smoothing framework to "TrajRS", which provides a certified robust radius for smoothed trajectory predictors. We clarify and expand the formal definitions of robustness in trajectory prediction and tailor the practical TrajRS scheme specifically to "robustness for the optimal prediction" and "robustness for all possible predictions". An extensive set of experiments demonstrates that TrajRS effectively achieves robustness certification for all smoothed pedestrian trajectory predictors in this work.

While AI development and evaluation for Southeast Asia (SEA) has grown rapidly, agent capabilities in regional languages are still poorly understood despite its importance to sovereign AI. To fill this gap, we introduce SEATauBench, the first agent-focused evaluation framework for SEA sovereign AI. SeaTau adapts TauBench to five languages -- Mandarin, Vietnamese, Thai, Indonesian, and Filipino -- and evaluates agents across progressively localized settings that vary the language of user-agent interaction, tool specifications, and task domains. Across three recent models, we find that English agent capabilities transfer reasonably well when only the conversation language changes, but quality and robustness degrade sharply as more task contexts are localized, with the largest losses in full domain adaptation. We also the limits of English-only agent assessment for measuring agent capabilities in SEA languages. More broadly, SeaTau provides a diagnostic benchmark and reusable adaptation pipeline for building reliable multilingual agents for linguistically diverse regions. Data and code can be accessed at github.com/SEACrowd/SEATauBench.

Classical SLAM estimates metric poses and a geometric map but produces no actionable predictive model for planning. Action-conditioned world models learn compact latent dynamics for planning but ignore global metric consistency and accumulate drift under open-loop rollout. We argue these are two views of the same estimation problem and propose J-LAW (Joint Localization and Actionable World Modeling) in this letter: a coupled factor graph that jointly optimizes metric object poses, latent world states, and latent landmark embeddings. The bridge is a pose-conditioned latent encoder and a learned pose--latent coupling factor, so that better localization improves the world model and vice versa. We cast observation, action-conditioned prediction, metric odometry, pose--latent coupling, latent loop closure, and latent landmark observation as probabilistic factors in a single MAP objective. Real-data experiments on PushT and WildGS show that coupled graph correction substantially reduces latent prediction RMSE and endpoint drift relative to open-loop rollout, while latent loop closure improves global trajectory consistency. J-LAW yields a map that is simultaneously metric (poses) and actionable (latent landmarks for planning).

We ask under what conditions an agent with a harm-minimizing policy can displace an approval-seeking (RLHF) agent in a competitive market, and when that policy is sufficient to prevent community harm. We use evolutionary game theory (finite-population Moran-Fermi pairwise comparison) to formalize this subject to assumptions of wisher hindsight, peer testimony, a monotone harm ledger, sufficient information density of community feedback, and a finite, depleting resource pool, in a negative-sum environment. We show that adoption is favored when the prior distributions on how readily wishers attune to community sentiment are monotone, exhibit endpoint inversion, and have a centro-symmetric pairing property, and demonstrate this with several long-tailed priors (Hill, Pareto, Lomax, Frechet). Where it is favored, a critical adoption level separates communities that drift back to the approval-seeking agent from those for which the audited agent fixes; above that level fixation is the overwhelmingly likely outcome. We derive when fixation is attainable as a bound on the effective (informational) size N_c of the community, which must be small enough to allow fixation before depletion. We present these as Theorems 5.4 and 5.5; the algebraic and finite-grid backbone is machine-checked in Lean 4, with the barrier-crossing asymptotics retained as explicit hypotheses. We show that a self-audited agent with a community ledger is not, in general, sufficient to prevent community harm. Sufficiency depends both upon the alignment of the agent's audit with community values and the timeframe over which harm is evaluated. Regardless of alignment, once adoption reaches dominance, the state is absorbing. The same policy that reduced harm under alignment becomes a trap, welfare-negative under misalignment and, even under alignment, one that locks in harm deferred past the adoption horizon.

Accurately explaining hidden patterns in multi-aspect data has typically been done by leveraging labels and/or accompanying auxiliary metadata. However, labels and auxiliary data may be inaccurate (e.g. nonstandard, inconsistent), insufficient (e.g. static tabular metadata for time-dependent recordings), or unavailable. % We propose \fullmethod (\method), which leverages the knowledge of large language models (LLMs) to explain the hidden patterns in human narratives. \method uses task-agnostic and task-specific prompts to explain extracted co-clustered latent patterns from tensor decomposition. To evaluate these explanations, we test the LLMs on forward and backward inference tasks. % Our demo system is available at https://github.com/dawonahn/ECML_PKDD_AnTenA.

Critic-free reinforcement learning with verifiable rewards (RLVR), exemplified by Group Relative Policy Optimization (GRPO), avoids training a value function (critic) and reduces memory and compute overhead relative to critic-based PPO pipelines for aligning large language models. However, GRPO-style advantage estimation depends on prompt-local (within-prompt-group) reward statistics and can be unstable. In particular, when all rollouts in a prompt group receive identical rewards, the within-group reward variance becomes zero, and group normalization yields zero advantages for that group, impeding learning in cold-start regimes with binary verifiers. We introduce BV-Blend, a critic-free framework that stabilizes advantage estimation by combining prompt-local on-policy statistics with semantic-cluster-conditioned historical moments. BV-Blend maintains EMA-tracked reward moments for each cluster, derives a confidence weight from a standard error of the mean (SEM) proxy, and uses this weight to blend historical and prompt-local baseline and variance statistics into a standardized advantage for PPO-style clipped updates. Experiments on verifiable reasoning benchmarks show that BV-Blend improves training stability and performance, and remains robust in regimes where group-normalized methods may stall.

Pathology foundation models (PFMs) have demonstrated strong potential across clinical and scientific applications, yet their performance is often hindered by batch effects, which are non-biological variations across tissue source institutions (TSIs) that distort learned feature representations and impair generalization. Conventional mitigation strategies, such as stain normalization, offer limited success in addressing these high-dimensional, complex artifacts. We present GLMP (General-purpose LLM-Mediated Pathology model), a novel framework that generates robust numerical embeddings from histology image patches through an intermediate textual representation. By leveraging pretrained general-purpose multimodal large language models (MLLMs) and text encoders, GLMP effectively prioritizes biologically meaningful signals over TSI-specific artifacts, thereby improving cross-institutional generalization. To our knowledge, GLMP is the first pathology model to use text descriptions of histological features as an intermediate representation for generating numerical embeddings from histology images. Our results highlight the untapped potential of broad-domain, non-specialized MLLMs in computational pathology and introduce a new paradigm for building versatile, generalizable, and robust pathology models.

Composition is a high-level visual intent that governs where subjects are placed and how a scene is organized, yet current unified multimodal models remain unreliable at fine-grained composition recognition and struggle to turn such intent into controllable generation. We present COMPASS, the first unified multimodal framework that grounds composition-intent control in a single system spanning both composition perception and composition-guided generation, with a shared expert token $τ_c$ as the central intent anchor. On the perception side, COMPASS injects composition expertise into an MoE backbone in a minimally invasive manner and distills the inferred intent into $τ_c$. On the generation side, COMPASS reuses $τ_c$ as a global conditioning signal that steers the denoising trajectory, effectively converting passive composition analysis into explicit layout control. To support systematic instruction-following composition learning and evaluation at scale, we construct Comp-11, a large-scale dataset with an 11-class taxonomy and reasoning-augmented annotations. Extensive experiments show that COMPASS substantially improves category-level composition understanding and delivers more composition-consistent, prompt-faithful generation than strong baselines.

Treatment reasoning underpins every therapeutic decision, integrating disease context, comorbidities, medications, contraindications, and evolving biomedical knowledge to select an appropriate therapy. It is inherently iterative: candidates are weighed against many constraints, revised as evidence emerges, and grounded in verifiable sources. Here we introduce ATHENA-R1, an AI agent for treatment reasoning across all FDA approved drugs since 1939, trained by reinforcement learning over a universe of 212 biomedical tools. At each step it identifies missing information, selects and runs relevant tools, and incorporates the evidence. To train it without human-annotated traces, we build a two-level self-learning framework: multi-agent systems construct the tools, tasks, and reasoning trajectories for supervised fine-tuning, then reinforcement learning with scientific feedback rewards reasoning quality (evidence gathering, grounded tool use, logical non-redundancy). Across five benchmarks of 3,168 drug reasoning tasks and 456 patient treatment cases, ATHENA-R1 outperforms language models and tool-use systems, reaching 94.7% accuracy on open-ended drug reasoning and 82.9% on treatment reasoning, 17.8 and 10.7 points above GPT-5. In blinded evaluations by experts from 28 rare disease organizations, it is preferred over reference models on all criteria, and physicians rated it favorably on complex hospitalized cardiovascular and infectious-disease cases. Adverse-event hypotheses it generated, tested in electronic health records from 5.4 million patients, reached adjusted odds ratios of 1.48-1.84, with no elevation among negative controls. Because it requires knowing what evidence to seek before concluding, treatment reasoning has long been hard for AI; we show it can be reframed as a learnable process of iterative evidence gathering that reinforcement learning can train AI to perform.

Causally linking disease-related factors to image-derived biomarkers provides a powerful pathway to understanding disease mechanisms. Despite growing interest in applying causal artificial intelligence (AI) approaches for this task, these methods still need to be adapted for complex medical images, and especially, neuroimaging. However, the lack of ground-truth data presents a barrier to development. To bridge this gap, we developed and tested a method for generating synthetic neuroimages, which adhere to a user-specified causal structure describing the non-image to image variable relationships, permitting the creation of ground-truth neuroimaging datasets. In the simulated T1-weighted magnetic resonance images, anatomical variability is modeled by sampling from a subspace estimated from real data and deforming a template image to create unique simulated subjects. Causal relationships are encoded via precise volumetric changes of any region-of-interest without unwanted global artifacts. We achieved relative volume errors of 0.3-2.66% for the targeted regions-of-interest and demonstrate their statistically significant causal relationships, while maintaining mean absolute errors for non-target brain regions between 0.034-0.397ml. An initial evaluation of causal discovery methods exposes their limited ability to suppress spurious connections, highlighting the need for image-appropriate methods. Our framework is the first to enable the generation of realistic synthetic 3D neuroimages with explicit causal control that can serve as the missing ground-truth data necessary for the objective benchmarking and development of causal AI methods.

Large Language Models (LLMs) often face ethical tradeoffs in which several responses may be defensible but express different priorities, such as fairness, honesty, courage, or restraint. We introduce VirtueMap, a framework for describing these patterns through an Aristotelian virtue-ethics lens. Instead of asking for a single correct answer, VirtueMap asks humans or LLMs to rank all five responses to each of seven general, non-lethal, non-political, and non-religious ethical dilemmas. To define the reference orderings used for scoring, we first proposed, for each dilemma and virtue, an ordering of the five responses from most to least expressive of that virtue. We then collected more than 100 respondent evaluations per ordering and retained it as operational ground truth only when at least 95% confirmed it. Rankings are scored against these retained orderings using normalized Borda alignment, yielding profiles over Practical Wisdom, Justice, Truthfulness, Courage, and Temperance. We apply VirtueMap to nine LLM families in a repeated-run evaluation and find high mean rank consistency (90.3%), with the largest differences appearing on Courage, Temperance, and Justice. We also release an interactive website that computes profiles locally in the browser and compares respondents with measured LLM profiles.

Tool-using LLM agents increasingly read untrusted content while holding side-effecting tools such as payments, email, CRM, and infrastructure APIs, yet common framework defaults still conflate tool exposure with authorization. We audit whether LangChain/LangGraph, LlamaIndex, and the Stripe Agent Toolkit re-authorize each model-emitted call, with concrete argument values, before execution. Across pinned public-source commits, all three provide capability gating by default, but none provides a deterministic fail-closed per-call value authorization gate by default. We introduce ScopeGate, a five-stage PDP/PEP for agent tool calls: scope, authorization, money ceiling, idempotency, and default deny. Evaluation shows the identical unauthorized payout call executes under LangChain's default dispatch (with a companion LlamaIndex PoC) but is denied by ScopeGate; the tested control reports 0/48 static bypasses, 0/29 unauthorized attempts (40-iteration adaptive run), 0/10 benign false-denies, and Latam-GPT payment-agent containment at 10/10. ASR denotes attempted unauthorized action, containment is not a cure, deployment-tier claims are inference over measured model classes, and no CVE is asserted.

Predicting the risk of distant metastasis from primary tumor tissue histology is a critical yet challenging task in computational pathology. Multiple Instance Learning (MIL) approaches can attend to subdomains in tumor regions that harbor features of metastatic cancer progression. However MIL models treat tissue patches as unordered bags, discarding the spatial layout that defines the metastatic potential. We propose that metastatic risk is inherently dictated by the geometric arrangement of the tumor microenvironment at the interface with tumor cells. Our model is designed to explicitly capture the spatial relationships between tumor cells, tumor associated fibroblasts and infiltrating lymphocytes. For this purpose, we propose Distance aware Tissue Modeling for Multiple Instance Learning(DTMf-MIL), a novel method that reinforces visual features with explicit spatial priors. By computing signed distance functions (SDF) relative to tissue phenotypes, our model learns to recognize structural signatures of metastatic risk. This geometric awareness translates directly to superior clinical performance as DTMf-MIL significantly outperforms state-of-the-art methods that ignore spatial layout on metastasis prediction from tissue in the primary tumor. We further validate our approach on public benchmarks, demonstrating that spatial awareness consistently improves diagnostic accuracy across diverse clinical tasks.

Generative models in finance face the dual challenge of producing realistic data while satisfying strict regulatory and economic objectives, a requirement that standard tabular diffusion models cannot provide. To address this difficulty, we introduce Constrained Tabular Diffusion for Finance (CTDF), a novel integration of sampling-time feasibility operations with mixed-type tabular diffusion in financial applications. By incorporating a training-free feasibility operator into the reverse-diffusion sampling loop, CTDF enforces hard constraints for applications such as simulation, legal compliance, and extrapolation. Extensive experiments on large-scale financial datasets demonstrate zero constraint violations and improvement in scarce data utility. CTDF establishes a robust method for generating trustworthy and compliant synthetic data, opening new avenues for rigorous generative modeling and analysis in the financial domain.

Stackelberg differential games (SDGs) provide a powerful framework for hierarchical decision-making in stochastic and continuous-time environments, yet their solution remains computationally challenging due to the complexity of traditional dynamic programming and Hamilton-Jacobi-Bellman-Isaacs (HJBI) methods, especially in high-dimensional systems. This paper proposes an entropy-regularized reinforcement learning (ERRL) approach for linear-quadratic SDGs (LQ-SDGs) within a continuous-time diffusion framework governed by Markovian regime switching. The key innovation lies in deriving exploratory weakly-coupled HJBI equations with entropy regularization, which promotes stochastic policies that actively avoid suboptimal equilibria -- a limitation of classical SDG methods. Neural networks are integrated to approximate regime-dependent value functions and solve high-dimensional partial differential equations (PDEs) efficiently, while a novel sampling technique enhances computational tractability. Numerical results demonstrate the effectiveness of the framework compared to conventional approaches, particularly in escaping suboptimal traps through exploratory policies. The study highlights the critical role of entropy regularization and neural network approximations in achieving robust solutions for hierarchical decision-making problems under abrupt environmental shifts.

We introduce MACROCAST, a lightweight Time Series Foundation Model (TSFM) for real-time macroeconomic forecasting. Existing TSFMs suffer from data leakage in two forms: temporal contamination, as the model may have seen the realized values of the series it forecasts, and revision bias, as training on fully revised data diverges from the preliminary, vintage-specific releases available to real-time forecasters. MACROCAST is, to our knowledge, the first TSFM that rules out both forms of leakage entirely: at no stage of training is the model exposed to information that would not have been available to a forecaster in real time. We train MACROCAST first on purely synthetic time series in approximately one GPU-day and then fine-tune it on synthetic time series drawn from Bayesian VARs, dynamic factor models, and ARIMA specifications estimated on vintage-specific ALFRED data. Because pretraining uses only simulated data and fine-tuning uses only real-time vintages, no observed future or revised value ever enters the model; each fine-tuning run takes nine minutes. Evaluated on the FRED-MD database in a genuine real-time out-of-sample exercise, MACROCAST improves on the AR(1) benchmark for roughly 80% of series-horizon pairs, matches or surpasses Chronos-2 -- the strongest currently available TSFM -- and outperforms the Bayesian VAR and dynamic factor model benchmarks, all in a data-leakage-free manner.

To address parameter misspecification and sudden structural environmental changes in conventional stochastic differential game (SDG) frameworks, this paper introduces a distributional control approach that characterizes optimal strategies as probability distributions over actions, conditioned on the continuous state, the discrete regime state, and parameters. This forms a reinforcement learning framework for entropy-regularized zero-sum stochastic differential games (ERRL-ZSSDGs) in a regime-switching jump-diffusion process. Using the dynamic programming principle (DPP), we derive the associated coupled systems of Hamilton-Jacobi-Bellman-Isaacs (HJBI) equations, from which equilibrium strategies are expressed via gradients of the value function. For linear-quadratic problems, semi-analytical solutions for both value function and equilibrium strategies are obtained by solving a system of coupled ordinary differential equations (ODEs). In more general settings, an Actor-Critic policy improvement algorithm is developed to approximate the value functions and equilibrium policies across different regimes. The method is applied to an investment game, and numerical examples illustrate the effect of the temperature parameter and regime transitions on optimal policies and values.

Sign languages are compositional systems where meaning arises by combining sublexical phonological parameters, such as handshape, location, and movement. While deep learning models for Sign Language Recognition (SLR) have achieved increased performance on translation benchmarks, it remains unclear whether these models distinguish abstract phonological features or merely rely on low-level statistical correlations. This work evaluates the phonological perception of SLR models trained on American Sign Language (ASL) by probing phonological sensitivity using minimal pairs and evaluating representational alignment with human behavioral data. Our results reveal that SLR models exhibit emergent phonological sensitivity, but with clear architectural trade-offs: pose-based models are sensitive to handshape contrasts, while pixel-based models better capture location changes. Furthermore, pose-based models learn latent representations that correlate with human perceptual similarity judgments (r~0.49). These findings suggest that while SLR models exhibit emergent phonology, current training paradigms are insufficient to scale them beyond their architectural inductive biases.

Agent-based AI has enabled the automation of tasks by exposing application tools and resources to large language models (LLMs). However, to improve scope and accuracy, agents are often given access rights that exceed those of ordinary users, introducing significant security risks. AI is routinely integrated into applications with a disregard to security, risking data exposure and breaching regulations. This paper applies the AI Trust, Risk, and Security Management (TRiSM) framework to a medical report-generation application to demonstrate how an insecure agent workflow can be transformed into security-conscious agentic workflow. Both workflows were evaluated across five LLMs (Claude Haiku 4.5, GPT-4.1-nano, GPT-4.1-mini, GPT-5.4-mini, and Gemini 2.5 Flash) on two report types, totalling 800 generations and 500 attack scenarios including RAG poisoning, data-field injection, and client-side network injection. The TRiSM-guided agentic workflow reduced mean attack success rates from 31% to 10% for RAG poisoning and from 42% to 25% for data-field injection, while eliminating the network injection vector entirely through server-side prompt construction. Furthermore, report accuracy increased by 14 percentage points (72.5% to 86.5%) with the agentic workflow, demonstrating a secure design which provides more reliable outputs. This paper contributes to knowledge by demonstrating least-privilege, defence in depth agentic workflows improving security and accuracy, while also highlighting model choice is a necessary architectural consideration.

The flatness hypothesis suggests that flatness of the loss landscape, as measured by the eigenvalues of the loss Hessian, correlates with better neural network generalization. While various algorithms reduce these eigenvalues, most focus on procedural design, leaving it unclear how data distributions and NN parameters structurally determine directions toward flat minima. Characterizing these directions analytically is generally intractable. To overcome this mathematical difficulty, recent studies derived the Wolkowicz-Styan (WS) upper bound on the maximum eigenvalue of the cross-entropy loss Hessian in three-layer NNs. Although this upper bound is differentiable, its gradient was not derived. Therefore, we analytically derive the gradient of the WS upper bound to characterize directions leading to flat minima. Based on this, we propose Hessian Spectral Range (HSR) Regularization, which updates parameters along the steepest descent direction of the WS bound. Experiments demonstrate that HSR Regularization narrows the Hessian eigenvalue spectrum, avoids sharp minima and saddle points, and promotes convergence to flat minima. Although the applicability of this method is currently limited to cross-entropy loss and three-layer architectures, to the best of the authors' knowledge, this is the first study to report a closed-form gradient that promotes convergence to flat minima without numerical approximations. Therefore, the theoretical analysis of this gradient is expected to contribute to the further development of NNs.

People overthink; language models over-sample, and the extra effort can talk both into a worse answer. Reasoning systems answer a hard question by sampling it many times (test-time scaling), and the more they draw, the more often a correct answer turns up somewhere, so coverage, the fraction of problems with at least one correct try, climbs and appears to be progress. But a deployed system must return one answer, and choosing it, not knowing which try is right, is selection; selection is capped, and past a point extra samples only make the model surer of a confident mistake, even as every draw adds cost. The gap between climbing coverage and stalled selection, the identifiability gap, is the answer a model can produce but not pick. So the real question is not whether to sample but how far, and the answer is: not far. For picking an answer, the vote has already settled within a few dozen draws, the modal ceiling; for scoring a benchmark, sooner still, the correlation ceiling. Beyond that, extra draws cost compute and add nothing, and can even make the answer worse. This paper turns the cutoff into a single number, the effective number of samples, that any sampling run already reveals. The bottleneck is recognizing a right answer, not generating one.

High-fidelity molecular docking simulations can produce biologically relevant estimates of epitope-receptor binding affinity but are computationally expensive and therefore limit the number of candidates that can be screened for vaccine design. In this work, we evaluate machine learning (ML) approaches where variants of active learning are used to classify instances of high binding affinity between 9-mer epitopes and a well-conserved swine leukocyte antigen (SLA) receptor in the context of Porcine Reproductive and Respiratory Syndrome (PRRS). We use an internally generated dataset of 80 epitope-SLA docking affinities, each requiring more than 48 hours of high-performance computing (HPC). Multiple model families (linear, MLP, CNN, and a small transformer) are trained under strict low-data conditions within a pool-based active learning loop. In each case, optimal model configurations are identified by conducting large-scale hyperparameter optimization over the combined space of model architecture, training configuration, acquisition policy, and ensemble decision rules. To mitigate the effects of data subsample selection, each candidate configuration is evaluated by averaging performance over many randomized and balanced training and validation data subsets. Across experiments, transformer-based sequence models consistently emerged as the best-performing architecture, with active incremental learning yielding significant improvement over a baseline random sample acquisition strategy. Under moderate training data availability (N=30), the optimized ML-model configuration outperforms a standard baseline trained on twice the amount of data. Under higher training data availability (N=60), the same configuration achieves a peak accuracy of 86.8%, consistent with an upper bound of 85% classification accuracy based on two independent estimates of conformational noise.

Deformable 3D Gaussian Splatting (3DGS) has emerged as an efficient approach for rendering dynamic scenes in a wide range of 3D applications. However, existing deformation field-based approaches largely lack explicit object-level modeling, often resulting in inconsistent Gaussian deformations within individual objects and unwanted coupling between different objects. To address this limitation, we introduce a semantics-guided framework that enforces dynamic regularization at the object level, aiming to achieve spatially consistent object-wise deformation. Specifically, we first extract segmentation masks using the Segment Anything Model (SAM) and derive semantic features from input images. An object-ID map is then constructed via feature relevance matching with a predefined object dictionary. Guided by this object-ID map, we identify the pixel-wise top-k contributing Gaussians for each object and impose consistency regularization on their deformation parameters, including position, scale, and rotation. Unlike prior methods that learn deformation fields without explicit object-level constraints, our approach incorporates semantic cues to guide deformation behavior at the object level. Experimental results demonstrate that our semantics-aware regularization improves object-level deformation consistency and outperforms baseline methods in rendering quality, achieving higher PSNR and SSIM and lower LPIPS in dynamic 3DGS rendering. Our project page is available at https://dyn-reg-3dgs.github.io/.

Parameterized quantum circuits (PQCs) provide a flexible substrate for hybrid quantum machine learning (QML), but their practical value on Noisy Intermediate-Scale Quantum (NISQ) devices remains an empirical question, especially because training depth and scale can introduce optimization challenges such as barren plateaus. Here we study how the number and topology of two-qubit entangling gates in the feature-map stage influence a fixed hybrid QNN workflow for classifying strong versus weak epitope-receptor binding in Porcine Reproductive and Respiratory Syndrome (PRRS) vaccine design. The dataset consists of docking-derived binding affinities for N=80 9-mer epitopes, labeled as Strong or Weak binding, and partitioned into training, validation, and test subsets using a 40:30:30 split. We compare a classical CNN benchmark with a hybrid Embedding-QNN architecture under four feature-map configurations: a non-entangling Z feature map, an all-to-all high-entanglement ZZ feature map, and two interleaved nearest-neighbour entanglement patterns of low and high depth. Among the configurations tested, the high-entanglement ZZ feature map is seen to provide the strongest evidence of reduced training-set overfit, with a lower training area under the accuracy curve (AUAC) and the highest test/training AUAC ratio, while preserving competitive test-set accuracy. These results do not establish a general QML advantage, but they suggest that feature-map entanglement topology is a meaningful design variable for sparse biological screening tasks and warrants further evaluation with additional metrics, larger datasets, and noise-aware or hardware-based experiments.

Deep Neural Network (DNN) classifiers suffer from poor calibration when their softmax outputs (predictive confidence) deviate from the empirical likelihoods. This manifests itself as either overconfident incorrect predictions or under-confident correct predictions. Label smoothing (LS) enhances model calibration by introducing entropy regularization during training through redistributing probability mass from the ground-truth label to the remaining classes. LS, including Margin-based LS (MbLS), have restrictive assumptions: they rely on predefined, uniform smoothing rules and only tackle overconfidence. In reality, samples exhibit diverse characteristics, such as difficulty/ambiguity, that interact with the evolving nature of the model being trained. In training, samples may have various degrees of under- or overconfidence. To overcome this, a mechanism that identifies the specific confidence state of each sample and determines the appropriate degree of smoothing in each training step is needed, tailoring the adjustment to the individual sample. We propose FedLAS: Feature-Modulated Bidirectional Label Smoothing, a plug-and-play algorithm for label smoothing-based losses. In FedLAS, we introduce a Feature Norm-based Confidence Indicator (NCI) to control smoothing and a Bidirectional Calibration Gating (BCG) module to detect both over and under-confidence. Our algorithm can be integrated with LS and MbLS based losses when applied to standard DNNs, enhancing performance. Extensive experiments on standard and fine-grained high-resolution vision benchmarks show that FedLAS consistently improves calibration compared to modern baselines, reducing Expected Calibration Error (ECE) and Adaptive ECE while maintaining Top-1 accuracy. Code: github.com/nadarasarbahavan/FEDLAS

Online high-dimensional regression requires algorithms that can update sequentially while preserving structural sparsity. We propose \textit{Adaptive Iterative Hard Thresholding (AIHT)}, an online sparse-regression framework that alternates stochastic subgradient updates with adaptively scheduled hard-thresholding steps. The key idea is to separate support discovery from local refinement: early in the learning process, AIHT delays thresholding so that weak but informative coordinates have time to accumulate signal, while later it increases the projection frequency to stabilize the sparse estimator and exploit local curvature. We develop the theory for high-dimensional online quantile regression, a challenging setting in which the loss is nonsmooth and the data may exhibit heterogeneity or heavy-tailed noise. Under restricted curvature and gradient-leakage conditions, AIHT remains in an inflated sparse cone, exhibits a two-phase convergence behavior, and attains logarithmic regret for the sliding-window objective. Simulations for online quantile regression, together with threshold-scheduling ablations, support the proposed mechanism and illustrate its advantage over standard online sparse-learning baselines.

We present RIPA, the first systematic multi-channel empirical study of prompt injection attacks delivered through the sensory pipeline of a ROS 2-based LLM-controlled robotic system. Across 100 independent runs per injection variant on five LLMs spanning four model families and parameter scales from approximately 4B to approximately 284B (DeepSeek-V4-Flash, Llama-3-8B-Instruct-Lite, Llama-3.3-70B-Instruct-Turbo, Qwen 2.5-7B-Instruct-Turbo, Gemma-3n-E4B), we identify model-specific vulnerability profiles that do not follow a monotonic scaling trend: Llama-3.3-70B-Instruct-Turbo exhibits 100% attack success rate (ASR) across all injection variants, while Llama-3-8B-Instruct-Lite and Qwen 2.5-7B-Instruct-Turbo resist direct-override injection (0% ASR), and the smallest model evaluated (Gemma-3n-E4B, approximately 4B) matches the 70B model's vulnerability profile, indicating that robustness is model-specific rather than scale-dependent. We propose a hybrid semantic firewall that achieves 0% ASR against known injection patterns with no false positives on a preliminary benign set (0/20 commands) but exhibits a 10.2% trial-weighted bypass rate (58/570 trials; N equals 30 per payload across 19 obfuscation payloads) against adversarially obfuscated attacks, exposing a critical gap between rule-based and semantic defense layers. We further introduce three sensory injection channels: visual (Channel 1, via OCR), audio (Channel 2, via Whisper STT), and LiDAR sensor context poisoning (Channel 3). We show that Channel 3, which injects fabricated obstacle data into the robot environment-state representation at the LLM system-prompt level, achieves 100% ASR across all variants on DeepSeek-V4-Flash. We also contribute a firewall bypass taxonomy spanning 19 obfuscation payloads across five categories. All code, data, and results are publicly available.

Parameter settings in evolutionary algorithms and metaheuristics are important because such parameter values can influence the performance of algorithms under evaluation. For a given algorithm, there are many different numerical experiments to show that the algorithm can work well in practice; however, in most cases there is no theoretical analysis of parameter settings. In this work, we show that theoretical analysis using the theory of dynamical systems and evolution of population variance can give some good results in terms of parameter ranges for the bat algorithm. We also show that results from numerical experiments are consistent with theoretical bounds. Such analyses can provide good insights from different perspectives about the algorithmic characteristics such as variance evolution, transition between exploration and exploitation as well as convergence behaviour.

This article establishes the foundational mathematical limits of Artificial General Intelligence (AGI) safety, proving that the core barrier is not the impossibility of an aligned state, but its structural unverifiability. We formalize this boundary through two central impossibility results: the Unverifiability Theorem of Alignment and the Theorem of Finite Structural Unverifiability of AGI Alignment. We ground this boundary at Trakhtenbrot's Wall, demonstrating that contemporary engineering defenses relying on finite hardware or halting architectures fail to escape logical obstructions. This failure manifests as an inescapable triad of containment failures: open domains yield fundamental undecidability (Rice and Gödel); universal finite verification collapses into algorithmic incomputability (Trakhtenbrot); and particular bounded environments trap the supervisor within intractable bounds in the worst case. As a direct structural corollary of these results, we derive the Soundness--Completeness--Tractability Trilemma, establishing that the mutual incompatibility of these three properties is a necessary consequence of descriptive complexity rather than an empirical anomaly. Finally, we map these theoretical bounds onto practical AI engineering, demonstrating that modern containment strategies are not temporary patches, but mandatory sacrifices of logical expressivity required to secure decidable fragments of safety.

Localized generative editing needs localized evaluation: full-image identity metrics are structurally confounded under hard-composited edits. We present Envisage, a FLUX.1-Fill inpainting reference pipeline for rhinoplasty goal visualization from a single frontal photograph. The pipeline combines 8 rhinoplasty clinical presets (the released framework also includes 8 blepharoplasty and 8 rhytidectomy presets), MediaPipe masks, and hard-mask compositing. The composite preserves outside-mask pixels by construction, so full-face identity scores are dominated by copied pixels rather than by the diffusion backbone. Because full-face identity metrics cannot grade localized edits, we introduce SurgicalScore, a mask-decomposed 0-1 protocol scoring edit direction, edit magnitude, masked LPIPS, realism, and outside-mask preservation; SS_raw assigns 0.919 [0.918, 0.920] to a perfect-predictor control , anchoring the ceiling. On N=211, the paired ArcFace gain (output-to-GT minus input-to-GT) is negative for all methods (Envisage -0.048 smallest, vs. ICEdit -0.139, Kontext -0.242, InstructPix2Pix -0.294; p < 1e-4), with external validation on a 457-pair ASPS/PCA corpus showing a larger negative gap. With SurgicalScore, Envisage achieves the highest score (0.599 [0.579, 0.619]) and leads on both metrics, but the all-negative ArcFace gap shows that full-face identity is poorly aligned with localized surgical accuracy under hard compositing. A 5-seed GT-oracle (an upper bound, not a deployable result) reduces the residual ArcFace gap by 73% (-0.054 to -0.015), with positive output-to-GT gain on 33.9% of cases, indicating candidate-space headroom for a learned ranker. For localized edits, progress should be measured with edit-region fidelity rather than full-face identity metrics. We release Envisage, SurgicalScore, preset definitions, and matched split manifests.

Connected Vehicles (CVs) rely extensively on communication technologies to enable data-driven predictive analyses for enhancing performance and safety. These communication channels can be exploited by adversaries to launch cyberattacks such as Sybil attacks, which could threaten both safety-critical and mobility applications, leaving CVs vulnerable and putting human lives at risk. As CV deployment continues to expand, the need to detect and mitigate cyberattacks in real-time becomes increasingly urgent. This study presents an in-vehicle Digital Twin (DT)-based collision warning framework with built-in capabilities for Sybil attacks detection. The framework integrates a Temporal Convolutional Network (TCN) for learning temporal dependencies in vehicle trajectory data and Hierarchical Navigable Small World (HNSW) algorithms for efficient similarity-based classification. Our framework is evaluated on real-world Sybil attack data, collected through field experiments. The framework achieved accuracy, recall, and F1 scores of 0.984, 1.00, and 0.944, respectively, in detecting Sybil-generated fake vehicles. During the safety evaluation, the framework reduced the mean Time Exposed Time-To-Collision (TET) and mean Time Integrated Time-To-Collision (TIT) of near-collision events by 88% and 72%, respectively. Furthermore, real-world feasibility evaluation shows that the framework conformed to the standardized maximum allowable latency for safety applications and operated well within the capacity of modern processors -- demonstrating the promise of an in-vehicle DT-based framework as an attack mitigation mechanism against Sybil attacks for next-generation CVs.

Effective sub-typing (also known as grouping or clustering) of patients using their electronic health record (EHR) data can greatly inform precision medicine efforts. However, subtyping temporal EHR datasets is known to be challenging due to inherent EHR issues, including complexity and irregularity. In this study, we propose a self-supervised Mamba-based model that learns effective EHR representations and enables enhanced patient subtyping. We evaluate the proposed model on public and private real-world EHR datasets to classify the data based on the available labels and subtype patients based on the representations learned from the model. Through an extensive set of experiments, we demonstrate that our model's design choices lead to better performance compared to competitive baseline models for prediction. Moreover, we evaluate several clustering techniques to demonstrate that our findings offer valuable insights into subtyping patients based on temporal records from EHR models\footnote{Our implementations are available at https://github.com/healthylaife/triplet_mamba.

Fayyazi et al. (2025) recently proposed FACTER, a model-agnostic framework designed to jointly enforce fairness and statistical coverage in LLM-based recommendation through conformal thresholding and iterative prompt repair. In this work, we conduct a reproducibility study of the FACTER framework across diverse architectures and dataset sparsity levels, evaluating both the original open-ended generation task and a constrained re-ranking extension. Under the strict reproduction, we observe a divergence in recommendation utility, which we trace to underspecified target-set evaluation in the original study. We then use the constrained re-ranking setting to evaluate FACTER when the candidate set is fixed, and introduce a static Fair Zero-Shot baseline to isolate the contribution of the iterative prompt repair loop. Our analysis shows that FACTER consistently reduces adaptive-threshold violation counts, but that these reductions are not consistently reflected under the fixed threshold or in global fairness metrics. In the constrained ranking setting, static fairness instructions achieve comparable semantic-parity outcomes to FACTER's dynamic repair loop, suggesting that the additional online repair mechanism provides limited benefit in this formulation. All code and reproduction artifacts are available at https://github.com/oscar-omlf/facter-repr.

Background: Developers frequently reuse code by copying fragments and adapting them to fit new contexts. Existing benchmarks for evaluating large language models (LLMs) on code adaptation either rely on explicit step-by-step instructions, cover only narrow change types such as variable wiring, or operate exclusively at function-level granularity. It remains unknown how well LLMs can adapt code fragments without explicit edit guidance when the required changes are varied and controlled. Objective: We investigate instruction-free code snippet adaptation in which an LLM must adapt a code fragment to fit its target context without any explicit edit guidance. We study three dimensions: which adaptation types are hardest (RQ1), how performance scales with adaptation complexity (RQ2), and how much surrounding context the model needs (RQ3). Method: We will construct a dataset of Java code fragments from open-source repositories with strong test coverage and apply a taxonomy of adaptation operators, derived from empirical findings on how developers adapt copied code, using a mutation-injection framework. Working at the code fragment level and controlling the injected changes lets us know exactly what adaptations the model must perform. The unmutated fragment serves as a plausible reference for the changes the model needs to make. LLMs will be evaluated on instruction-free adaptation tasks across three context granularity levels. Correctness will be measured primarily via test-suite re-insertion, complemented by mutation-level inspection.

Solving multi-agent optimal control problems in aerospace such as pursuit-evasion and contested space operations can be modeled as non-convex differential games for which, there are limited algorithms. In this work, a relaxation of generalized Nash Equilibrium problems (GNEPs) to exclude inter-agent control coupling in dynamics, which is representative of many multi-agent systems is introduced. The main contribution is an algorithm for solving a broad class of differential games named FALCON: Fast Augmented Lagrangian Convexification for Open-loop Nash equilibria is presented. Methodologically, sequential convex programming (SCP) is utilized to create tractable convex sub-games which can then be solved via standard convex programming methods involving a potential game reformulation. FALCON is demonstrated to have global convergence guarantees to an open-loop Nash equilibrium for non-convex differential games under mild assumptions. This is numerically shown through both cooperative and competitive differential games.

In stochastic linear bandits, the canonical Upper Confidence Bound (UCB) algorithm admits a simple frequentist regret analysis but can be computationally demanding, while Thompson Sampling (TS) is computationally attractive yet typically harder to analyze due to its non-optimistic nature. We propose Absolute Thompson Sampling (ATS), a simple modification of TS that ensures optimism in expectation by replacing the signed exploration noise with its absolute value. This preserves the computational efficiency of TS while avoiding the technically involved anti-concentration arguments common in TS analyses, enabling a simple UCB-style regret analysis. We show that ATS achieves $\tilde{O}(d^{3/2}\sqrt{K})$ regret, matching existing bounds for TS in linear bandits. We further introduce Ensemble Absolute Thompson Sampling (EATS), which takes the maximum over multiple absolute perturbations with normalization by the ensemble size. As the ensemble size grows, EATS converges to the UCB objective, recovering UCB behavior in the limit. Experiments show that moderate ensemble sizes already yield strong performance. Our results point to a bridge between randomized exploration and deterministic optimism both in theory and practice.

Large language models (LLMs) are increasingly deployed in high-stakes domains, where free-text explanations such as chain-of-thought and post-hoc rationales are used to justify model outputs. Yet it remains unclear whether these explanations are sufficient, i.e., if they contain enough information to explain the model's output-generating process. We generalize classical sufficiency from feature attributions to arbitrary explanations and prove that explanation sufficiency can change depending on the input distribution, which must be explicitly defined for LLM explanations. We propose using the LLM itself to generate alternative inputs conditioned on an explanation, capturing its beliefs about possible inputs. We formalize self-consistent sufficiency as a goal for free-text explanations and introduce an information-theoretic metric, SCSuff, that enables evaluation of free-text explanations without relying on predefined biases or shortcuts. Our experiments show that SCSuff agrees with targeted perturbation tests where applicable and demonstrate that explanation sufficiency can vary with the input distribution. We find LLM explanations are generally insufficient and weakly correlated with model size, accuracy, or output entropy. Analysis of final-token hidden states shows that top and bottom SCSuff scores can be predicted from internal representations, suggesting that SCSuff can guide detection and improvement of sufficient LLM explanations. The code for this paper is available at https://github.com/rajesh-lab/self-consistent-sufficiency .

This work tackles the challenge of enhancing low-resolution LiDAR sensors for SLAM applications through a novel Deep Unrolling-based Super-Resolution (SR) model. We integrate an outlier removal module to ensure structural integrity while maintaining real-time performance. By leveraging a model-based optimization approach, our method efficiently reconstructs high-resolution point clouds while minimizing computational overhead. The proposed SR model is evaluated within a LiDAR SLAM framework, demonstrating significant improvements in pose estimation accuracy and efficiency compared to state-of-the-art SR methods.

Recent progress in Text-to-SQL has been driven by stronger language models and prompting strategies, yet performance on real enterprise benchmarks such as Spider 2.0 and BIRD remains far below that on classical academic datasets. We argue that the main bottleneck is no longer reasoning, but database representation. Real databases contain repeated audit columns, large groups of similar tables, opaque identifiers whose meanings are stored only in documentation, and extensive data dictionaries with little query-relevant information. Existing query-aware methods, including schema linking and retrieval-based schema selection, filter this raw context but still operate on redundant and verbose representations. We reformulate the problem as database context compression, a query-agnostic transformation that rewrites schemas, semantic descriptions, and external documentation into a compact representation. We formalize this transformation with the SGCF (Support-Gain Component Factorization) principle, which unifies repeated column extraction, isomorphic table templating, semantic componentization, and evidence purification under a single coverage objective. Based on SGCF, we propose DBCC, a database-side middleware that performs offline structural and semantic compression together with lightweight online evidence purification. DBCC is model-agnostic and can be integrated into existing Text-to-SQL pipelines. On Spider 2.0-Snow and BIRD, DBCC reduces input context by up to two orders of magnitude (from 2.6M to 34.7K tokens on the largest Spider 2.0-Snow subset), improves schema-linking strict recall from 0% to 56.5% under DeepSeek-V3.2 (63.1% under Claude Opus 4.7), and consistently increases end-to-end execution accuracy by 1.8-1.9% over three recent Text-to-SQL systems. Our code is open-sourced at https://github.com/MrBlankness/SchemaCompression.

Neuromorphic and edge computing research has focused on reducing the inference cost of neural network controllers, yet in physical closed-loop systems the actuator can rival or exceed an efficient controller in energy. An efficient controller is therefore necessary but not sufficient, because the actuator becomes the cost worth reducing once inference no longer dominates it. Here, we introduce energy-aware learning, an approach that incorporates actuator energy directly into the reinforcement learning reward, and demonstrate it in closed-loop deep brain stimulation (DBS) for Parkinson's disease. A deep spiking Q-network, trained in a biophysical cortico-basal ganglia-thalamic circuit model, learns to suppress pathological alpha-beta oscillations by 45.2% while reducing stimulation charge by 80.0% relative to continuous DBS. Sparsity-constrained knowledge distillation compresses the policy onto the SynSense XyloAudio 3 neuromorphic processor at 0.52 mW inference power, yielding 28.1x lower energy per inference than an equivalent artificial neural network on conventional edge hardware. By co-optimizing stimulation energy and inference efficiency, the framework addresses both major power demands in implantable neuromodulation.

Prediction sets should have high coverage to be useful, but some coverage notions are more practically relevant than others. In the classification setting, class-conditional coverage requires that the prediction set (i.e., the set of candidate labels for a new test point) must achieve the target accuracy level within each class, which may be challenging to satisfy when many classes are rare and have few calibration points. At the other extreme, marginal coverage requires only that coverage holds on average over the distribution of all classes, which can lead to low-probability labels being essentially ignored. To find a middle ground, recent work has introduced macro-coverage, defined as the unweighted average of class-conditional coverages. Macro-coverage offers a compromise between marginal coverage and class-conditional coverage that is particularly appropriate for long-tailed settings. In this work, we show that label-weighted conformal prediction can be used to produce prediction sets with a finite-sample macro-coverage guarantee, and more generally a guarantee on a family of generalized macro-coverage objectives that aggregate coverage at the level of arbitrary class groupings and take a weighted average. We further characterize the form of the smallest prediction sets satisfying a given generalized macro-coverage objective and propose a corresponding conformal score function. We validate our theoretical results on two large-scale image classification datasets.

While recent vision-language models (VLMs) have achieved significant improvements on static visual-to-code tasks such as generating code for webpages, charts, or SVGs, it remains unclear whether they can recover temporal dynamics when motion is present. To this end, we introduce Animation2Code, a benchmark for evaluating temporal visual reasoning via reconstructing executable web animation code from videos. Animation2Code consists of 1,069 web animation videos with diverse visual appearances and motion patterns, paired with corresponding HTML/CSS/JavaScript implementations. We propose two human-aligned metrics, appearance similarity and temporal similarity, which allow us to disentangle visual fidelity from temporal alignment when comparing rendered animations against ground-truth samples. Benchmarking state-of-the-art VLMs on this dataset shows that current VLMs struggle to maintain temporal consistency in reconstruction, even when achieving high appearance similarity, including under finetuning and iterative refinement settings. Code and data are available at https://anya-ji.github.io/animation2code-website .

Current approaches to enhance Large Language Model (LLM) reasoning, such as Chain-of-Thought and "Wait" prompts, primarily encourage models to think more, yet often fail to guide them toward Truth. While Representation Editing (RepE) offers a intrinsic control, its application to dynamic reasoning trajectories remains underexplored. In this work, we bridge this gap by investigating the geometry of truth within unfolding reasoning chains. We uncover three critical insights: (1) Truth is encoded at the sentence level and is entangled with latent reasoning patterns; (2) Effective intervention follows an Uncertainty Principle and a Decay Effect, requiring localization to early, high-entropy forks; (3) Naive steering vectors suffer from noise, risking collateral damage to correct trajectories. Based on these findings, we propose DynaSteer, a dynamic RepE framework. DynaSteer employs pattern clustering to disentangle reasoning manifolds and utilizes Fisher-LDA to project purified truth. By dynamically monitoring lookahead entropy, it selectively steers and rolls back trajectories only when necessary. Comprehensive experimental results on several MATH benchmark verify the effectiveness of DynaSteer, and experiments on out-of-domain coding tasks further confirm its generalization ability. Our code is publicly available at https://github.com/tianlwang/DynaSteer.

When a large language model (LLM) codes a construct in text as a human annotator would, that agreement makes the LLM a reliable coder. Yet reliability leaves construct validity untouched. The instrument may be theory-naive, reaching the code through a correlate that meets none of the demands the construct's theory makes, and no current method tells that apart from genuine measurement. We propose grain calibration as a method that closes the gap. It decomposes a construct into clause-level components, tests each against the text with extractive evidence, and combines the results through an explicit, theory-derived rule. Because the rule is stated rather than lodged in one opaque pass, its structure is evidence about the process rather than the output. It shows which components settled a code, and, when the code is wrong, whether a component was missed or an adjacent construct mistaken for it. Validation shifts from scoring an instrument's outputs against an annotator to showing that the instrument runs on the construct its theory specifies.

Modern machine learning models are trained by optimizing high-dimensional non-convex empirical risk functions. Such cost functions can have a multitude of local optima and yet, gradient-based optimization appears to converge to near-global optima. Within a simple supervised learning setting, we develop a precise picture of which parts of the empirical risk landscape are accessible by polynomial-time algorithms. We are given i.i.d. pairs $\{(\boldsymbol{x}_i,y_i):\; 1 \le i\le n\}$ with $\boldsymbol{x}_i\in \mathbb{R}^d$ standard Gaussian feature vectors, and $y_i\in\mathbb{R}$ response variables that depend on $\boldsymbol{x}_i$ through their projections on an unknown $k$-dimensional subspace. We use empirical risk minimization to learn a model that depends on an $m$-dimensional projection of the data (e.g., an $m$-neurons neural network). We propose an incremental approximate message passing (IAMP) algorithm and precisely characterize the training error it achieves, as well as the relation between test and training error, in the high dimensional asymptotics $n,d\to\infty$, with $n/d\toα\in (0, +\infty)$. Based on earlier work in related models, we expect that the performance achieved by our algorithm is optimal among polynomial-time algorithms.

The axiom of choice has divided the foundations of mathematics for over a century, but the distinction between classical and constructive proofs has remained a philosophical and methodological one. We use Lean 4's kernel-level tracking of axiom dependence to show that the axiom of choice has a measurable geometric correlate in proof space that obeys a one-parameter mixture law and has operational consequences for neural theorem provers. To do this, we partition $471{,}260$ declarations of Mathlib by transitive dependence on the axiom of choice and represent a filtered population of $42{,}355$ traced theorems by their sequences of tactic invocations. We use the constructive proofs in this dataset to train a self-supervised proof encoder and show that when using it to measure classical proofs, three complementary measurements (anomaly score, reconstruction loss, and density-superlevel containment) exhibit a common decline with the proof's distance from the axiom in the dependency graph, from sharp separation at the shallow boundary (AUC $0.847$ at distance $2$) to indistinguishability at distance~$9{+}$. Robustness controls show that the signature survives length, file, author, and topic controls, and replicates under full-source encoders trained on normalised proof source. Operationally, we show that on an evaluation sample of $251$ Mathlib theorems, Lean's \texttt{aesop} tactic solves constructive theorems at $13\times$ the rate of classical ones, and a neural-guided hybrid using the ReProver tactic generator compresses the gap to $5\times$. The geometric anomaly score predicts \texttt{aesop} failure beyond proof length, providing an operational link between the geometric signature and prover performance.

Athlete assessment is a critical process for tracking physical progress and identifying elite talent. However, during mass recruitment drives, traditional methods rely on manual observation, which is inherently subjective and unscalable, or basic computer vision (CV) systems limited to quantitative repetition counting. These standard approaches lack the "coaching intelligence" required to evaluate qualitative physiological markers such as form degradation, spinal articulation, and fatigue. This paper presents a novel, LLM-based hybrid agentic framework for automated, holistic athlete profiling that strictly aligns with the Sports Authority of India (SAI) assessment protocols. Orchestrated via LangGraph, our dual-pipeline architecture synthesizes the geometric precision of CV (MediaPipe) for kinematic tracking with the semantic reasoning of Vision-Language Models (Llama-4-scout). To overcome the latency and token constraints associated with multimodal video processing, we introduce a 3 X 3 "Smart Grid" temporal chunking strategy, reducing computational overhead by over 88% while preserving critical temporal continuity. To ensure data integrity and mitigate hallucination, the framework pioneers an autonomous "LLM-as-a-Judge" self-correction loop that cross-references quantitative and qualitative metrics before persistence. Finally, we implement a dual-persistence Retrieval-Augmented Generation (RAG) pipeline utilizing a vector search engine (ChromaDB). This enables coaches to bypass rigid SQL databases and perform complex semantic queries (e.g., "Identify athletes with high endurance but poor core rigidity") using natural language. Experimental results demonstrate that this multi-agent approach significantly bridges the gap between raw biometric tracking and actionable coaching insights, offering a scalable, objective solution for national talent identification.

Speech-driven 3D facial animation methods face significant challenges in simultaneously achieving high-fidelity motion and precise artistic control at production quality. Existing controllable models typically learn global style control by relying on large-scale, low-quality \emph{in-the-wild} datasets that compromise overall animation realism. Furthermore, these frameworks often lack the fine-grained temporal precision required for demanding tasks such as dialogue localization (e.g., dubbing), where matching specific facial expressions is as critical as lip synchronization. We present KM-Speaker (Keypoint-Matching Speaker), a novel keypoint-conditioned flow-based generative framework that provides both global style guidance and frame-level temporal control from reference performances. We propose a disentanglement strategy that separates audio-driven lip motion from keypoint-driven upper-face dynamics, together with a global style context preservation mechanism to ensure coherent full-face expressiveness. KM-Speaker advances example-based 3D facial animation by achieving high-fidelity motion and flexible controllability in a data-constrained setting, consistently outperforming state-of-the-art methods in lip-sync accuracy, style adherence, and expressive temporal control.

As large language models (LLMs) move into production serving, practitioners must rapidly evaluate inference performance across diverse hardware, models, and serving parameters to meet cost and latency targets. However, the end-to-end behavior of LLMs couples serving-layer policies with low-level GPU kernel execution and rapidly evolving architectures, forcing slow, deployment-specific benchmarking that is hard to generalize. We present KernelSight-LM, a fine-grained inference simulator that models token-level execution and produces kernel-level latency breakdowns. It decomposes each serving step into a roofline kernel model with a learned efficiency term, a communication model, and a host-overhead model, composed through a discrete-event scheduler that also captures mechanisms like prefix caching and continuous batching. KernelSight-LM offers two prediction tiers that trade target-GPU data for accuracy. The cross-generation tier uses no target-GPU measurements, only hardware specifications and kernel microbenchmarks from previously profiled GPUs, and predicts per-kernel latency on an unseen GPU generation to 12.1% error, a 1.8x improvement over the roofline baseline (22.0%). A second target-measured tier adds one model-agnostic kernel-microbenchmark sweep on the target GPU, sharpening per-kernel error to 3.8%, a 7.3x improvement over a comparable baseline (27.7%). Both tiers require far less target-GPU data than the prior systems they extend. In our simulator, these predictions yield end-to-end median (p50) errors across six model families of 15.4%, 12.8%, and 3.0% (TTFT, TPOT, throughput) in the cross-generation tier and 14.3%, 6.2%, and 2.7% in the target-measured tier, matching dedicated profiling tools while collecting far less on-device data. Beyond prediction, its kernel-level bottleneck breakdowns support hardware/software co-design and capacity planning.

On-policy distillation (OPD) has a property absent in offline distillation and RL: teacher supervision quality depends on student competence. Incoherent rollouts yield noisy gradients; already-mastered tokens yield redundant ones. This creates waste at three scales (tokens, training phases, and prompts) yet existing methods supervise uniformly. We introduce SEAD, which uses entropy as a unified probe of this competence-dependent degradation at three scales: (1) joint teacher-student entropy partitions tokens into zones receiving tailored divergences or zero gradient (approx. 50% skipped); (2) a cosine schedule anneals from forward to reverse KL as competence grows; (3) a competence-gated curriculum introduces prompts easy-to-hard. These components are symbiotically necessary: token selection requires coherent rollouts (curriculum), annealing requires monotonic improvement (also curriculum). On OLMo-3 (7B to 32B), SEAD achieves +4.8 avg accuracy over vanilla OPD across six math benchmarks, with ablations confirming super-additive interactions.

We study sparse self-attention in which each query attends to a dense local window plus a set of Fibonacci-spaced offsets, with a per-layer scalar alpha that compresses or expands the spacing. Across 21 language models trained under one matched recipe (60M parameters, 512 hidden, 16 layers, 426M tokens), we compare four ways of setting alpha across depth: fixed, per-layer learned, a static linear stagger, and a coprime (anti-gridding) reassignment of that stagger, together with a reach-matched power-of-2 control. Three results stand out. First, a static per-layer stagger improves perplexity over both fixed and learned alpha, and the gain is base-agnostic: applying the same stagger to a power-of-2 base lifts it above fixed Fibonacci and to parity with learned Fibonacci attention. Second, learning per layer is inert: it does not beat the static schedule and costs roughly five times the inference latency. Third, and most consequential, all sparse variants extrapolate to four times their training length with little or no degradation, whereas a recipe-matched dense baseline collapses (perplexity rises by 201% at 4x length); we attribute this to fixed-offset attention only ever querying relative positions seen during training. We also report two honest negatives: at training length the best sparse model has about 26% higher perplexity than the dense baseline, and the staggering gain is uniform across context positions rather than concentrated at long range.

Recent advances in large language models and vision-language models have enabled reasoning over multimodal data, offering opportunities for clinical applications such as decision support and triaging. However, existing medical AI benchmarks are fragmented: some support multi-turn dialogues but lack images, while others provide multimodal inputs but focus on single-turn QA tasks. To address this gap, we introduce IMCBench, an image-grounded, multi-turn medical conversation benchmark that pairs real, publicly available clinical images with synthetic patient profiles to simulate realistic patient-clinician interactions. Each conversation is evaluated across three clinical dimensions: safety, accuracy, and appropriate use of uncertainty in diagnosis. We benchmark eight multimodal frontier models across four model families (Claude, GPT, Nova, and Llama), scoring each on a 1-5 scale using LLM-as-Jury scoring calibrated against expert clinician annotations. Our results show that Claude Opus 4.6 achieves the highest overall score (3.61), followed by Claude Sonnet 4.6 (3.30) and GPT-5.2 (3.29), though no model dominates all dimensions and safety degrades for both malignant and rare conditions ($Δ$ = -0.27 each). Ablation studies further reveal that both visual input and EHR context contribute to safe guidance (safety drops of 0.18 and 0.23 on average when each is removed), with stronger models leveraging visual features more effectively. Together, these findings demonstrate that accurate clinical description does not guarantee safe patient guidance, motivating the need for multi-dimensional evaluation frameworks in medical AI.

In many real-world applications, such as retail sales, energy usage, and supply chain planning, forecasting is performed across hierarchical structures. These structures often represent aggregations (e.g., products to categories to regions), where forecasts must not only be accurate but also coherent, meaning that lower-level predictions sum correctly to higher-level forecasts. Traditional statistical methods, such as Bottom-Up and MinT, enforce coherence through post-processing but fail to model complex nonlinear temporal dependencies and covariate interactions. We propose Hierarchical Temporal Fusion (HTF), a novel extension of the Temporal Fusion Transformer (TFT) that integrates structured hierarchical embeddings with a coherence-aware loss function to ensure consistent forecasts across all levels of a hierarchy. Rather than applying reconciliation after forecasting, HTF embeds coherence directly into the training objective. The coherence loss penalizes the difference between aggregated child forecasts and their corresponding parent forecasts during training, enabling the model to learn both temporal dynamics and structural consistency simultaneously. We evaluate HTF on two publicly available benchmark datasets: the M5 Walmart forecasting dataset and a publicly available hierarchical energy consumption dataset. Results demonstrate that HTF substantially reduces forecast incoherence while improving forecasting accuracy compared with classical reconciliation methods and deep learning baselines. In addition, attention visualization and embedding analysis provide insight into how temporal and structural information contribute to hierarchical forecasting performance.

Building performant Vision-Language Models (VLMs) requires carefully curating large-scale training datasets, yet the community lacks systematic benchmarks for evaluating such curation strategies. We introduce DataComp for VLMs (DCVLM), a benchmark for controlled data-centric experiments to improve VLM training. As part of DCVLM, we collect 160 datasets spanning four data types -- image-caption pairs, multimodal interleaved documents, text-only, and instruction-tuning data -- into a corpus of 6T multimodal tokens. DCVLM allows participants to test curation strategies (filtering, mixing, formatting, sampling) across 1B-8B models and 6.25B-200B token budgets. Models are then evaluated on a carefully selected suite of up to 52 downstream benchmarks across 9 domains. We conduct extensive experiments on DCVLM and find that data mixing, not filtering, is key to a high-quality training dataset: instruction-heavy mixtures scale better than caption-heavy ones, with gains widening at larger scales. The resulting dataset, DCVLM-Baseline, enables training an 8B VLM to 63.6% accuracy on our 33-task core suite with 200B training tokens. Compared to FineVision, the state-of-the-art open VLM training dataset, this represents an improvement of +5.4pp. DCVLM and all accompanying artifacts will be made publicly available at https://www.datacomp.ai/dcvlm/.

Sparse autoencoders (SAEs) have become a useful tool for extracting interpretable features in language models. However, standard SAE architectures operate on individual token activations, meaning that the number of active features scales linearly with context length, and studying long model transcripts becomes difficult. We introduce turn-averaged SAEs, which represent a single Human or Assistant turn with a fixed number of features by learning to reconstruct the average model activation across the turn. We find that turn-averaged features describe a single turn's high-level characteristics more completely than per-token features when judged by an LLM. We also demonstrate that turn-averaged SAEs greatly simplify common downstream uses of SAEs like attribution graphs. Broadly, turn-averaged SAEs make interpretability techniques practical at long context lengths.

Recent advances in AI-driven weather and climate modeling have improved forecast skill while reducing computational cost. However, existing data-driven approaches are limited in their ability to model coupled Earth system dynamics, which is required for extending predictability beyond the ~2-week horizon. To address this, we introduce NIVA, a multimodal foundation model designed to learn unified representations across Earth system components. While the full framework targets atmosphere, ocean, ice, and land interactions, we focus here on a two-modality setting (ocean and atmosphere) as a controlled proof of concept to evaluate whether foundation models can learn coupled dynamics. Trained on large-scale Earth system simulations, NIVA learns physically meaningful cross-modal structure, providing a foundation for subseasonal-to-seasonal prediction. As initial validation, we show that NIVA captures key modes of climate variability through accurate prediction of major climate indices.

Team roles offer an interpretable lens on collaboration, yet computational studies of roles often rely on domain-specific personas or data-driven clustering rather than theory-grounded taxonomies. We operationalize a taxonomy of eight communication roles grounded in education literature and annotate a corpus of 6,307 Slack messages from 55 students across 18 teams in a semester-long computer science course project. We evaluate whether LLMs can approximate expert labels, enabling scalable, taxonomy-driven role annotation. Using these role labels, we characterize role dynamics over teams' lifecycles, finding that different roles peak at different moments and that students enact a more diverse set of roles as projects progress. To evaluate the utility of our role constructs, we use them to predict peer recognition, outperforming lexical, conversational, and LLM-prompting baselines. To assess generalizability beyond the educational context, we apply the same role constructs to a public dataset (DeliData) to predict team performance improvement after deliberation, again exceeding prior performance.

Scalar objective functions are required when a multi-criteria optimization problem must yield a single preferred design rather than only a Pareto set. The choice of scalarization influences which compromise is selected, how preference parameters are interpreted, and whether non-supported Pareto regions can be reached. This paper compares four formulations for normalized bi-criteria minimization: weighted sums, achievement scalarizing functions, desirability functions, and a fuzzy-logic-based formulation. Two analytically defined Pareto fronts, one convex and one concave, isolate the effect of the objective formulation from numerical optimizer behavior. The comparison focuses on reachable Pareto regions, parameter-induced selection density, compensation between criteria, sensitivity, and interpretability. Results show that weighted sums are simple but structurally limited on concave fronts, while achievement, desirability, and fuzzy formulations reach interior non-supported regions through different mechanisms. Desirability functions introduce nonlinear single-criterion preference mappings, whereas fuzzy rules express nonseparable and reference-dependent engineering preferences.

We investigate domain adaptation of modern BERT models in the legal domain. We further pre-train ModernBERT on all US court opinions using the masked language modeling objective. Although ModernBERT has been trained on roughly 500x more data than original BERT, we still find that this model benefits from further pre-training and domain adaptation in the legal domain: we report significant improvements compared to vanilla ModernBERT on all datasets connected to US court opinions. We find gains similar to those reported in early work on domain adaptation of BERT-like models. However, from scratch pre-training does not match the performance of further pre-training an existing ModernBERT checkpoint in our experiments. The resulting models are capable of processing sequences up to 8,192 tokens, and can be used to compute meaningful embeddings of legal passages, or could quickly rerank hundreds of legal passages for a given search query. We release all model checkpoints publicly.

Multiview mammography relies on paired craniocaudal (CC) and mediolateral oblique (MLO) views to provide complementary projections of a 3D breast volume, enabling precise anomaly localization. However, acquiring high-quality, balanced datasets remains challenging for deep learning applications. We propose a novel method to synthesize multiview mammograms by leveraging the inherent geometric relationship between CC and MLO views. To enforce an implicit 3D consistency prior during generation, we develop an alignment module that searches a 2D affine transformation subspace to establish optimal anatomical correspondence. Leveraging this alignment, we introduce a pixel-space self-consistency loss based on the Earth Mover's Distance (EMD) between the 1D anteroposterior (AP) axis tissue distributions of the generated images. Integrated into a pretrained flow matching model, MammoFlow forces synthesized pairs to share physically plausible tissue distributions from the chest wall to the nipple. To our knowledge, this is the first work to guide multiview mammogram generation using implicit geometric tissue correspondence. Our method demonstrates superior image quality, passes expert radiologist evaluation, and generates physically consistent pairs that improve downstream classification AUC by 5%. Code is available at https://github.com/XYPB/MammoFlow

The increasing demand for high-performance computing in plasma physics has driven scalable and resilient simulation methods capable of efficiently exploiting modern multi-GPU architectures. This work extends a portable hybrid MPI+OpenMP implementation of BIT1, focusing on high-performance resilience for accelerated Particle-in-Cell (PIC) Monte Carlo (MC) simulations under both uniform and non-uniform load conditions. Scalable particle load balancing and robust checkpoint/restart mechanisms across Nvidia and AMD accelerators are integrated with standardized I/O using openPMD and ADIOS2. This leverages BP4 for high-performance file-based checkpointing and SST for in-memory data streaming, enabling efficient data movement, resilient large-scale execution, seamless continuation from existing checkpoints, and effective handling of computational and I/O workloads. Advanced HPC profiling and tracing tools, including Nvidia Nsight Systems and AMD ROC-Profiler with Perfetto, provide detailed insights into computation, communication, and system-level behavior for optimization. Performance results on Frontier (OLCF-5), MN5, and LUMI-G demonstrate strong and weak scaling up to 800 GPUs, validating the framework for large-scale PIC MC simulations, while in-situ analysis and visualization using scalable I/O further enhance scientific insight without interrupting multi-GPU execution on current and future exascale systems.

Sequence learning has emerged as the promising paradigm in recommendation systems, surpassing traditional Deep Learning Recommendation Models (DLRM) by capturing the temporal nuances of user behavior. However, current state-of-the-art architectures operate under a limiting analogy: they treat user history as a monolithic chronological sequence like a sentence in a Large Language Model (LLM). We observe a fundamental divergence between natural language and recommendation data: unlike the linear, logical flow of text, user history is inherently multi-faceted. A user's journey is a fragmented reflection of diverse interests, resulting in much weaker coherence between items than is found in LLM training data. This lack of structural unity leads to context pollution. In single-sequence modeling, unrelated behaviors compete for the same attention budget. This "noisy" signal dilutes the model's focus, effectively capping its ability to discern high-intent patterns from background activity. To address this, we propose Constructive Multi-Sequence Learning (CMSL), a paradigm shift from passive sequence ingestion to active "context engineering" that constructs multiple coherent sequences in latent space. CMSL leverages a learnable Sequence Construction Module to disentangle user history into "pure" thematic strands, followed by a linear attention mechanism to efficiently model these strands at scale. CMSL has been deployed across ranking and retrieval tasks and across four major surfaces at Meta.

Automatically generated videos from scientific papers are increasingly used for education and research dissemination. However, existing evaluation metrics mainly measure visual quality or whether key points from the paper appear in the video without assessing whether the video actually helps viewers understand the ideas. We introduce EffectivePresentationScorer, a framework for evaluating the instructional quality of scientific presentation videos. It checks whether a video explains the main ideas clearly, introduces needed background concepts, and connects technical details to the main contribution of the paper. When we apply EffectivePresentationScorer to the existing paper-to-video generation systems, we find that generated videos mention the correct topics and follow the structure of the paper but fail to explain prerequisite concepts or clarify why the method works. These failures are often ignored by existing video evaluation metrics, which focus on content presence rather than explanatory quality.

Embodied foundation models have recently been widely used to improve robot generalization and task success rates. Previous works apply lossy efficient-inference techniques such as quantization, pruning, and asynchronous inference, accepting small action quality degradation in exchange for lower per-step computation cost and inter-action latency. However, unlike traditional static ML tasks, embodied tasks involve repeated interaction with the environment, and task-level performance is determined not only by per-step cost, but also by closed-loop effects unique to embodied execution, which remain insufficiently characterized in current efficient-inference studies. In this work, we propose TISED (\underline{T}ask-level \underline{I}nference \underline{S}peedup \underline{E}ffect \underline{D}ecomposition), an analytical framework that unifies diverse lossy inference optimization techniques and decomposes their effects on static and dynamic tasks, and uncovers some paradoxical effects on task-level performance: (1) on \textit{static tasks}, optimization sometimes can lengthen end-to-end per-task completion time even as per-step latency drops; (2) on \textit{dynamic tasks}, moderate lossy optimization can raise task success rate even above the baseline; and (3) the monotonicity and sweet-spot location of both effects can shift with hardware configuration. Together, our findings provide a new perspective on adapting inference optimization techniques to embodied tasks.

The clinical and communication skills of medical students are commonly assessed through Objective Structured Clinical Examinations (OSCEs), which consist of brief scenario-driven simulations of doctor-patient interactions. However, training is often limited by the low availability of human standardized patients, motivating the development of realistic virtual patients (VPs). To address this gap, we introduce a French OSCE dialogue dataset comprising 240 student-patient training interactions. We build upon it a controllable LLM-based pipeline to generate synthetic OSCE dialogues. The pipeline integrates modular components, such as retrieval-based grounding and a reflection loop, to ensure patient fidelity, coherence, and realism. Additionally, we propose a multi-level evaluation framework assessing patient simulation quality, student performance, and linguistic quality, using an LLM-as-a-Judge approach. Experiments suggest that controllability modules generally improve patient fidelity and student evaluation consistency. Finally, we implement an interactive prototype in which students can practice with a VP and receive automatic feedback.

Fine-tuning on harmless data can partially undo behaviors acquired earlier in training. Safety can erode under benign post-alignment updates, unlearned capabilities can re-emerge, latent traits can transfer through apparently unrelated supervision, and related post-alignment fragility appears in other generative settings. We argue these phenomena are usefully viewed through a common training-history lens. Our hypothesis is geometric: large early training phases create dominant behavioral manifolds, while later alignment or specialization phases are shallower displacements from them. Subsequent fine-tuning can therefore inherit a persistent reversion component pointing back toward a witness of the dominant manifold. We call this the gravitational interpretation of fine-tuning reversion. Across our main settings, representational drift rapidly acquires a component along a history-defined reversion direction (v_rev). In our main track, alignment with v_rev rises from cos = 0.429 +/- 0.052 after the first update to 0.647 +/- 0.021 by step 20. Across 24 run-step pairs, every observed alignment exceeds the p99 of an isotropic activation-space null. We demonstrate that selectively blocking motion along v_rev changes the final alignment at T=100 from 0.648 +/- 0.009 to -0.211 +/- 0.021 and reduces harmfulness from 19.0% +/- 4.0% to 8.5% +/- 1.5% with little task cost. These results support v_rev as a causally relevant mediator of early post-alignment reversion in our setup. Importantly, we do not claim that v_rev is the unique safety direction, nor that the dominant manifold is directly observed; rather, we identify a robust, history-defined direction that explains and partially controls early reversion dynamics.

Recent work suggests that Large Language Models (LLMs) are sensitive to the belief states of agents described by text, as measured by the false belief task (FBT), yet persistent concerns of construct validity remain. We adopt a **developmental perspective**, tracing the pattern of mental state reasoning behavior -- and likely **preconditions** for this behavior -- across multiple training stages in the Olmo2 and Pythia language model suites. We find that above-chance FBT performance depends both on model size and sufficient training volume, emerges relatively late in pretraining, and is most improved by post-training interventions (SFT, DPO) in the condition most diagnostic of mentalizing (False Belief, Implicit). However, FBT performance is fragile: consistent with past work, the use of non-factive verbs (e.g., thinks) increases false belief attributions even in the True Belief condition. To contextualize these findings, we track the emergence of **situation modeling**: the ability to report on basic factual properties of a described scene. Situation modeling accuracy generally precedes and exceeds FBT accuracy, yet situational representations also prove surprisingly incoherent in certain respects: when asked about the knowledge states of the Antagonist agent -- who always knows the item's true location -- Olmo2 13b is consistently influenced both by the Target agent's knowledge state and the presence of non-factive verbs. Together, these results suggest that larger, sufficiently trained models build partially coherent situation models in a developmentally appropriate sequence, yet display surprising fragility -- highlighting the value of developmental and stress-testing approaches for evaluating LLM capabilities.

Large vision-language models (LVLMs) are increasingly used for clinical image understanding, yet they remain vulnerable to \emph{hallucinations}--producing textual findings or attributes not supported by the image. We present a vision-traceable hallucination detection framework that audits arbitrary LVLM responses via visual evidence grounding, requiring neither modification nor internal access to the hidden states of LVLMs. Given an LVLM response, we extract visually verifiable entities and use a medical-domain-adapted Qwen-VL grounding verifier to localize each entity on the input image. To enhance the robustness of our detection method, we introduce a counterfactual entity perturbation method and estimate visual evidence uncertainty by contrasting factual and counterfactual grounding results. Specifically, we compute an entity-level uncertainty score from the positive confidence, counterfactual confidence, and their grounding overlap for binary hallucination decision-making. Experiments on multiple medical imaging modalities and LVLM backbones demonstrate that our method consistently improves hallucination detection performance over recent baselines, while providing interpretable localization evidence and strong cross-model transferability. Code and dataset are available at https://github.com/Agentic-CliniAI/CounterVHD.

Training operator-learning models for large-scale problems governed by partial differential equations (PDEs) is challenging due to the curse of dimensionality, memory constraints, and limited training data. These challenges arise in many scientific and engineering applications, including subsurface flow, climate modeling, and geological carbon storage (GCS). In this work, we propose a scalable operator-learning framework based on the Karhunen-Loeve Deep Neural Network (KL-DNN) and demonstrate its performance for modeling GCS. The model is trained on a dataset comprising 100 samples of large-scale simulations in a three-dimensional domain with 1.7 million cells and 50 time steps. The KL-DNN method constructs latent spaces using low-rank singular value decomposition of static properties and a nested Karhunen-Loeve expansion for dynamic pressure fields, enabling full-resolution predictions without subsampling or spatial coarsening. The KL-DNN model achieves an average root mean square error (RMSE) of 1.1 psi for pressure (0.04% relative error with respect to the average pressure in the domain) and RMSE of 0.0146 for CO2 saturation (5% relative error with respect to the average saturation inside the plume). The model requires 20 minutes of training on a single GPU, representing a 19% reduction in the pressure errors, 7% reduction in the saturation error, and a two-order-of-magnitude speedup compared to DeepONet trained on the same dataset. These results, along with inference time of less than one minute, establish the proposed model as a practical and accurate solution for large-scale PDE problems, enabling rapid uncertainty quantification, history matching, and real-time decision support.

We present CLEAR-MoE, a four-phase post-training pipeline that converts a frozen pretrained Vision Transformer (ViT) into a sparse Mixture-of-Experts (MoE) model without updating backbone weights. The pipeline (i) scores feed-forward network (FFN) layers by sparsity, clusterability, and output sensitivity; (ii) decomposes selected layers into a shared low-rank SVD basis and per-cluster residual experts using k-means clustering; (iii) trains lightweight routers supervised by cluster labels; and (iv) dispatches tokens through pluggable CUDA backends. On Imagenette with DeiT-Small, CLEAR-MoE retains 99.9% of the dense model's accuracy (86.70 +/- 0.02% versus 86.73%). Extensive ablation studies reveal a consistent empirical finding: the shared SVD basis is the primary factor responsible for preserving accuracy. Random routing, learned routing, and three different router architectures produce nearly identical performance, with accuracy varying by at most 0.06 percentage points (86.62%-86.68%). Accuracy also remains stable across different SVD ranks, expert counts (2-8), calibration set sizes (50-500), and random seeds. This behavior generalizes across five ViT backbones (DeiT-Tiny, DeiT-Small, DeiT-Base, ViT-Small, and ViT-Base), covering models from 5.7M to 86.6M parameters, with accuracy differences <= 0.10 percentage points from their dense counterparts. On a GTX 960 GPU, routing and scatter-gather overhead make the CLEAR-MoE FFN 1.3-1.7x slower than the dense implementation. A dispatch microbenchmark further shows that routing is an order of magnitude more memory-bound than expert matrix multiplications, identifying fused dispatch kernels as a promising direction for future optimization.

Multimodal models are increasingly deployed to solve tasks collaboratively with humans or other artificial agents. Existing benchmarks show that these models possess many of the required component capabilities, but the conditions that coincide in collaboration, including time pressure, information asymmetry, and imperfect communication, are usually studied in isolation. We introduce GPTNT, a benchmark built on the cooperative video game Keep Talking and Nobody Explodes, in which two agents must coordinate to defuse procedurally generated bomb puzzles against a live countdown. One agent can see and manipulate the bomb but does not have the defusal instructions; the other has the instructions but cannot see or manipulate the bomb. Neither agent can succeed alone: success requires effective and efficient communication. Unlike turn-based proxies, GPTNT requires agents to act asynchronously and communicate in real time. GPTNT is designed to separate collaboration from reliance on memorized solutions: the instruction manual, the partner, or both can be withheld to isolate what a model derives in the moment from what it already knows. We show that GPTNT poses a substantial challenge for state-of-the-art systems: none of the closed- or open-source models we test defuses a single bomb in real time, a bar that human players clear. Through controlled experiments, we identify critical weaknesses in state tracking, efficient action under time pressure, ambiguity handling, and error recovery. We release GPTNT as a benchmark for collaborative performance that current evaluations leave unmeasured. Because it runs on the real game, GPTNT benefits from procedural generation and inherits a living modding community, allowing the benchmark to evolve as models improve rather than being solved once and retired.

Positron emission tomography (PET) seeks to balance diagnostic quality with ra-diation dose. Low-count PET noise is non-Gaussian, non-stationary, and spatial-ly dependent. It scales directly with local activity and is shaped by iterative recon-struction and physical corrections. Standard denoising diffusion probabilistic models (DDPMs) ignore these PET properties. Their forward process adds iso-tropic, homoscedastic Gaussian noise to the target. Such an approach fails to cap-ture the realistic physical degradation generated by the imaging system. To ad-dress the above limitations, this study introduces a heteroscedastic residual diffu-sion model (HDDPM) for low-count brain PET recovery in which the forward corruption is itself intensity-aware. We designed a fixed, Poisson-based variance module to generate voxel-wise noise maps. These maps naturally place stronger noise perturbation on low-activity regions than high-activity ones, meanwhile the network predicts the low-to-standard-count residual under explicit dose-fraction conditioning. We evaluated our proposed model (HDDPM) alongside generative frameworks across three different scanners, using both internal and external da-tasets at various simulated dose levels (1% to 50%). HDDPM and isotropic DDPM showed comparable overall image quality, but HDDPM stood out in the lowest-dose (1%) external scans. It is highly reliable and significantly reduces measurement errors in both high- and low-activity regions, compared to the standard model. These results support that heteroscedastic noising with the pro-posed HDDPM is feasible, and it provides a physically motivated inductive bias for quantitative low-count PET recovery by reflecting the activity-dependent noise structure of PET.

Across social and online platforms, people are increasingly exposed to AI-generated images. As a consequence, the task of distinguishing AI-generated from authentic images is becoming a central challenge for information ecosystems. While humans perform better than chance, accuracy falls short of many operational needs. Initial evidence shows that visually oriented training can improve deepfake detection but does not improve participants' ability to identify real images as real. Here, we investigate the efficacy of a brief training intervention for intelligence analysts employed by the United States government in 2024. We conducted a counterbalanced within-subject randomized experiment in which we showed participants real and AI-generated images varying in pose complexity and scene context and asked them whether each image was real or AI-generated, both before and after an expert delivered a 30-minute training that pointed out patterns in seven real and 50 AI-generated images. We collected 2,544 image-level judgments from 32 intelligence analysts. We find training increased overall accuracy by 9 percentage points (95% CI: [2.7, 15.4]) from a baseline of 72%. We find the improvement is driven by a 14.2 percentage point increase in accuracy for real images (95% CI: [0.7, 27.7]). Through a careful experimental setup that curated matched pairs of real and AI-generated images across pose complexity categories, we reveal how these trainings influence people with different levels of digital forensics and generative AI experience and identify the kind of image-based content where this training intervention appears to be most effective. Ultimately, these results provide causal evidence that a brief, structured training can improve human judgment across a diverse array of real and AI-generated images, informing organizational responses to AI-generated visual misinformation.

The emergence of low-dimensional structures in the spectra of neural network weight matrices is a common empirical feature of trained models, but the dynamical origin of this phenomenon during learning remains an open problem. We formulate neural network training as the stochastic evolution of an initially random matrix ensemble, driven by stochastic gradient descent (SGD) updates that reshape the spectral bulk while amplifying signal strength. This induces a Baik-Ben Arous-Péché (BBP) transition during training, where isolated eigenvalues detach from the random bulk distribution, providing a dynamical framework for representation formation in high-dimensional learning dynamics. We demonstrate this in a solvable linear teacher-student model, where spectral evolution is analytically tractable and a phase diagram of trainability governed by the step size (or learning rate) and initial weight variance is obtained, and subsequently extend our formalism beyond the linear regime to nonlinear and stochastic settings. Numerical simulations in realistic settings support this picture, showing robust emergence of spectral alignment during training. Our results suggest that spectral analysis may provide a unified perspective of stochastic learning dynamics, linking trainability, optimisation hyperparameters, spectral phase transitions, and representation learning in neural networks.

Dexterous manipulation policies can solve individual skills, but composing them to perform multiple tasks with a single hand remains challenging. Adding a new task on top of an existing manipulation skill often imposes conflicting demands on overlapping fingers and contact modes, causing destructive interference between preserving an existing manipulation outcome and executing a new one. We propose DexCompose, a role-aware residual composition framework that reuses pretrained dexterous policies for multi-task manipulation through explicit finger-level action ownership. Given two pretrained full-hand policies, DexCompose first collects successful post-task states from the first skill and performs release tests over candidate finger masks to identify which fingers are necessary for maintaining the established skill state. It then trains two asymmetric residual modules: a bounded residual stabilizer for task preservation, and a context-aware residual that adapts the frozen downstream policy only within the action subspace assigned to the new task. We evaluate the framework on 16 composite dexterous manipulation tasks spanning four object-retention skills and four downstream interactions. DexCompose achieves a 77.4% average composite success rate, demonstrating that structural action ownership with dual residuals offers a promising direction for composing dexterous skills beyond conventional policy chaining.

As large language models and harness frameworks continue to advance, agents operating in terminals are increasingly capable of performing a broader range of general computer-use tasks beyond coding. However, existing benchmarks do not adequately evaluate general-purpose terminal computer-use agents (TUAs): general computer-use benchmarks primarily target graphical user interfaces (GUIs), whereas terminal-based benchmarks largely emphasize technical and programming-centric workflows historically native to the shell. We introduce TUA-Bench, a general-purpose benchmark for terminal-use agents. TUA-Bench includes 120 real-world tasks across five task families, covering routine digital activities-including document editing, email management, and live-web information seeking-as well as scientific and engineering workflows co-designed with PhD-level domain experts that require specialized software. This breadth distinguishes TUA-Bench from prior shell-focused or domain-specific benchmarks. Each task is manually designed, runs in a real terminal with a deterministic setup script, and is evaluated by an execution-based scoring protocol. We find that the strongest frontier agent, Claude Code with Claude Opus 4.8 max reasoning effort, achieves 65.8% overall performance, with substantial gaps across both tracks. By providing a broad and realistic evaluation of terminal-use capabilities, TUA-Bench aims to accelerate the transition from narrow, task-specific assistants to general-purpose agents capable of operating reliably across diverse digital environments.

Binary classification from positive-only samples is a variant of PAC learning in which the learner receives i.i.d. samples from the positive region of an unknown target concept, but is evaluated under the original distribution (which places mass on both positive and negative regions). This model dates back to Natarajan [1987, STOC], and the characterization of improper learning is well-known -- it even appears in textbooks. The characterization of proper positive-only learning, however, has long remained open. In this work, we revisit and settle this question: a concept class is properly learnable from positive-only samples if and only if it has finite VC dimension and satisfies a new combinatorial condition, which we call uniform exterior separability. Together with several separation results, this characterization reveals a surprisingly rich landscape that differs sharply from standard PAC learning: proper and improper learning are separated, randomized and deterministic proper learning are separated, there are classes for which no ERM is a learner, and finite VC dimension does not suffice even for non-uniform learning. Along the way, we introduce new combinatorial dimensions that we believe can be of broader interest in learning theory.

Many two-player zero-sum games admit not a unique Nash equilibrium but a convex set of them: a polytope of profiles that all share the minimax value V* yet prescribe different behaviour. Standard solvers each converge to some equilibrium and are treated as interchangeable. We ask whether they instead select different members of the Nash set, systematically as a function of the algorithm rather than the seed. Using a tabular, exactly solvable testbed of six games with analytically known Nash sets -- including a two-dimensional Nash polytope and Kuhn poker -- we find that (i) selection is determined by the algorithm, not the seed, but families differ only on asymmetric Nash sets; (ii) regularized last-iterate methods (R-NaD, magnetic mirror descent) select the maximum-entropy member, the information projection of their uniform reference onto the Nash set -- exactly on the 2-D polytope and at 99.7% of maximum entropy in Kuhn -- while regret-averaging methods (CFR, CFR+, fictitious play) drift to a lower-entropy face; we confirm this on a randomized 180-game ensemble, where R-NaD attains the maximum-entropy member in 100% of converged games while CFR+ sits strictly below it in 94% (paired Wilcoxon p < 10^-27); (iii) the selected member has downstream consequences against sub-optimal opponents that scale with sequential/hidden-information structure but stay bounded -- in Kuhn the max-entropy member is a strictly better hedge, whereas on the matrix games the members differ without either dominating. We also report two negative results correcting common intuitions: removing CFR's positive-orthant (max(R,0)) projection does not eliminate boundary drift; and R-NaD's selection is anchor-following, not initialization-independent. We state the maximum-entropy / I-projection characterization as a strongly data-supported conjecture, checked throughout against analytic ground truth.

We analyze Bregman ADMM for nonconvex linearly constrained problems under two-sided relative smoothness, a condition that replaces the standard Lipschitz gradient assumption with a Hessian comparison relative to a Bregman kernel. This setting covers polynomial objectives arising in matrix and tensor models for which a global Lipschitz-gradient constant need not exist. We show that on an invariant open state-space domain, one iteration of Bregman ADMM defines a smooth primal--dual fixed-point map whose strict-saddle KKT points are unstable fixed points; consequently, from random initialization the iterates converge to a strict saddle with probability zero. Combined with existing first-order convergence results, this yields almost-sure second-order stationarity of limiting KKT points. We extend the analysis to a multi-block star consensus formulation for distributed optimization. The technical novelty lies in a determinant reduction with a Bregman-specific symmetrization and scaling step in the two block spectral argument, together with a null space cancellation exploiting the star graph structure in the consensus case. Numerical experiments on distributed matrix factorization illustrate the theory, and a symmetric tensor factorization example demonstrates the broader Bregman proximal splitting idea beyond the separable consensus setting.

Inference-time scaling is a promising paradigm to improve generative models, especially when outputs must satisfy structural constraints or optimize downstream rewards. We consider Masked Diffusion Model (MDM) and introduce MDM-VGB, a discrete diffusion sampler that augments unmasking generation with theoretically principled reward-guided remasking. Inspired by the recent success of the classical Jerrum-Sinclair backtracking Markov chain in reward-tilted generation, MDM-VGB extends the backtracking random walk from a fixed prefix tree to a masked-state graph, allowing tokens to be unmasked and remasked at arbitrary positions. The resulting sampler favors unmasking and remasking moves that lead to higher-value partial configurations, enabling both effective high-reward generation and efficient repair of low-reward samples. We prove that MDM-VGB is robust to process-verifier noise and achieves quadratic complexity, while popular test-time heuristics such as best-of-$N$ can incur exponential complexity due to error accumulation. Our theoretical findings are corroborated by strong empirical performance, particularly on popular constraint-satisfaction and scientific benchmarks such as Sudoku and QM9.

Preference-based alignment often struggles to capture the reasoning that underlies human judgments. Many evaluations rely on multiple interacting criteria, yet pairwise labels reveal only the final choice rather than the considerations that shape preferences. Inverse Constitutional AI (ICAI) improves interpretability in decision making by summarizing preferences into natural-language principles, but its single-pass explanations miss much of the nuance involved in complex decisions. We introduce Democratic ICAI, a novel approach that gathers multiple competing rationales through structured persona debate, offering a broader and more expressive account of the factors influencing each comparison. From these richer signals, we derive clearer and more comprehensive steering principles and use them to guide decision modeling through both LLM-based and decision-tree judges. Experiments on creative preference benchmarks, MuCE-Pref and LiTBench, across multiple creative task categories show that Democratic ICAI yields a more faithful preference structure. It improves average preference prediction across tasks relative to deliberative prompting and principle-based baselines, while producing constitutions that LLM annotators prefer.

CSIRTs increasingly fine tune language models on vulnerability scan records, but these records expose internal network topology and create privacy risks under regulations such as GDPR and LGPD. We present the first empirical study of how DP SGD and HMAC pseudonymization interact when fine tuning small language models with 1B to 3B parameters on structured CSIRT data. We evaluate 96 LoRA adapters across four SLMs and four training regimes, including raw fine tuning, QLoRA with large batch training, and DP SGD with epsilon equal to 2 and 8. We also audit memorization using 20 planted canaries, four extraction attacks, and a dual attack targeting HMAC pseudonymized identifiers. Our results show three main findings. First, matched update controls reproduce the observed reduction in memorization by reducing the number of optimizer updates alone, accounting for 66 percent to 132 percent of the measured effect, with a mean of 100 percent across three seeds and four models. In this setting, DP SGD provides the formal privacy guarantee but does not produce additional measurable reductions in memorization. Second, HMAC pseudonymization removes the original identifiers from the exposure surface, reducing exposure by 40 percent to 61 percent, while pseudonymized identifiers remain close to the expected random baseline and do not become a secondary memorization target. Third, F1 scores remain between 0.19 and 0.28 across all 96 adapters using four shot prompting, indicating that, under the evaluated training budget, 1B to 3B SLMs do not achieve operationally useful performance.

Nowadays mobile robots have wide engineering applications. Simultaneous localization and mapping (SLAM) is an important task of these robots. The major and common algorithms used for this task are based on extended Kalman filter (EKF). One of the main problems in EKF-based SLAM is its divergence. The nonlinearity of motion and observation models and linearization error are the main reasons for the divergence. There have been some efforts to address this problem with limited success. In this paper, by applying a simple compass and using an effective transformation, we transform the non-linear state space model into a linear model. Then, by applying the original KF to this model, we reach a new method, which is called LMKF SLAM. We show that the LMKF SLAM is significantly superior to the state-of-the-art methods, especially EKF-based SLAMs, both in accuracy, convergence, and computational complexity. The proposed method is also more stable with respect to the uncertainty of sensors values and changes in system parameters. Experimental results verify these points.

Model capability is the central variable in LLM pre-training, yet is never observed directly: data shapes it prospectively, while evaluation reveals it only retrospectively, compressing samples, prompts, decoding, and scoring rules into one noisy score. Practical optimization runs this backward: a failure is observed first, and the engineer must infer the corpus fix. The two sides speak incompatible vocabularies -- benchmark names and per-sample correctness versus data sources, domains, and quality labels -- so this inference is usually intuition, not method. We close this gap with the \emph{capability slice}: a group of evaluation samples sharing background condition, task type, solving operation, and output constraint -- precise enough to localize a single weakness yet stable enough to survive aggregation, unlike a benchmark name, too coarse, or a single sample, too noisy. Built around this unit, an evaluation taxonomy, a non-instruction data taxonomy, and mapping rules form a closed loop turning a benchmark-level failure into a targeted, testable data intervention. We test this loop on two case studies pulling in opposite directions. First, the loop rules the data out: continued pre-training drives BBH down by $-46.82\%$, but diagnosis traces this to a single masked \texttt{\textless EOS\textgreater} loss rather than weakened reasoning; restoring it recovers BBH to $66.44$, above the original checkpoint, without changing the data. Second, the loop rules the data in: a persistent math-reasoning weakness is decomposed by solving operation into specific failing combinations, and a weakness-targeted sampling procedure built from it lifts AIME2025/AIME2026 Pass@128 from $6.67$/$0.00$ to $26.67$ each. The same unmodified loop reaches opposite, correct verdicts in both cases, showing the evaluation-to-data inference can be routine, auditable, and experimentally validated rather than intuitive.

We demonstrate how emotional valence influences the order-dependent structure of children's recognition memory: correct recall of a sequence of emotionally-valenced toys depended not just on the valence of a given toy itself, but also on the valence of the toys shown before and after it. Whilst standard psychological models confirm that order-dependence differs across an event (a set of toys shown in sequence), accuracy is low and the model does not reflect how memory for an emotional object influences others in the set. A classical tensor network model factoring in valence is able to achieve a 77.98\% accuracy in modelling the results of the study. While not strictly a ``quantum cognition'' model, this massive increase in accuracy shows the value of quantum-inspired methods for modelling order-dependent phenomena, such as emotional memory. Further, the task protocol we introduce presents a novel, real-world tool for exploring emotional temporal memory in children for analysis using classical and quantum-like models of cognition.

Appliance-level energy monitoring in office buildings produces noisy alerts that non-expert facility managers struggle to use. This paper proposes an end-to-end agentic pipeline that combines deep time-series forecasting, variational anomaly detection, and LLM-based reasoning to generate prioritized, actionable maintenance recommendations. The system tracks seven office appliances using a hybrid Singular Spectrum Analysis (SSA) and Long Short-Term Memory (LSTM) forecasting model, and applies a per-appliance LSTM Variational Autoencoder (VAE) with attention to flag abnormal daily consumption episodes. A three-stage LangChain pipeline begins with a Context Agent that always retrieves three core RAG sources (model reliability, hourly baseline, and expert knowledge) and conditionally adds up to three more (forecast context, anomaly history, global baseline) based on event characteristics, capped at eight reasoning steps. A Diagnosis Agent converts the evidence into a structured JSON diagnosis, and a Report Agent renders a human-readable narrative. A reflective memory layer incorporates operator feedback. The dashboard shows real-time 30-minute forecasts, intraday consumption, the previous day anomaly report, and a feedback form. We evaluate the forecasting model, anomaly detector with appliance-specific thresholds, and LLM reasoning on a 16-scenario benchmark including sustained and transient spikes, unexpected shutdowns, and systemic events, comparing five LLM backends under static vs. dynamic retrieval. Dynamic retrieval matches full static retrieval across all backends while cutting average context from six to three-six sources per event. The best backend scores 90.4/100 with a 100% pass rate at a 70-point threshold, and a fully local 7B-parameter model passes all 16 scenarios.

This work examines perturbation generalization in spatial foundation-model embeddings derived from fluorescence microscopy images. Although these models can discriminate drug conditions accurately, it remains unclear whether the learned representations reflect patterns consistent with expected perturbation axes that transfer across drugs. We introduce SVC-Probe, a perturbation-aware framework that combines Subcellular Embedding Atlas Stability, Mondrian Neighborhood Graphs, and a Foundation Model Perturbation Probe to assess embedding stability, neighborhood rewiring, and centroid prediction under drug treatment. Applied to the CM4AI MDA-MB-468 chemical-perturbation atlas comprising 462 antibody labels and SubCell 1536-dimensional embeddings, SVC-Probe demonstrates that 98.6% three-way condition accuracy does not correlate with reliable cross-drug prediction, with cosine similarity diminishing from 0.944 in-domain to 0.30 under leave-one-drug-out evaluation, constituting a two-drug stress test rather than a general benchmark. Null calibration indicates that raw residual-turnover coupling is largely influenced by generic embedding structure, whereas a drug-specific signal emerges under vorinostat and is consistent with chromatin-related reorganization. In contrast, the paclitaxel axis is not robustly reconstructed, likely due to sparse coverage of microtubule-associated proteins. Together, these results introduce and demonstrate a reusable diagnostic framework for stress-testing spatial virtual-cell representations and indicate that perturbation generalization may serve as a stricter and more informative benchmark than baseline condition discrimination.

In the optimization of neural networks, gradient dynamics are influenced by critical points that arise from the model's architecture. These critical points occur where the Jacobian of the model's parametrization is rank-deficient, and are the most pronounced singularities studied in Singular Learning Theory. We investigate such points in deep fully-connected networks with monomial activations via tools from polynomial algebra such as Mason's Theorem. We show that, for sufficiently large activation degree, criticality occurs precisely at subnetworks, i.e., at parameter configurations where some neurons are inactive or redundant. This offers a mathematical perspective on the implicit bias in deep neural networks, explaining the tendency of these models to converge toward simpler functions.

Data-driven modeling in real-world regression tasks often suffers from limited training samples, high collection costs, and noisy observations. Inspired by the impact of data augmentation in vision and language, we propose a novel Counterfactual Residual Data Augmentation (CRDA) technique for tabular regression. Our key insight is that once a regressor has modeled the systematic component of the data, the remaining noise can be viewed as an invariant residual that remains stable under small perturbations of carefully selected features. We exploit this residual invariance to generate new, yet realistic, training samples, effectively expanding the dataset without requiring additional real data. Our method is model-agnostic and readily applicable to various types of regressors. In experiments across datasets from a variety of benchmark repositories, on average, CRDA reduces an MLP Regressor's MSE by 22.9% and an XGBoost Regressor's MSE by 6.4%. When compared to existing state-of-the-art data generators and augmentation techniques, CRDA consistently outperforms in MSE reduction. By adding principled counterfactual variations to the training data, our method offers a simple and efficient remedy for noise-prone, small-sample regression settings.

Single-cell RNA sequencing (scRNA-seq) clustering is essential for identifying cell types, but high dimensionality, sparsity, dropout, and technical noise hinder robust expression representation and cell graph construction. Existing masked autoencoders mainly use expression recovery for feature reconstruction, while graph clustering methods usually depend on fixed KNN graphs and do not feed recovered expression back into graph optimization. We propose scKDGM, a KAN-guided dynamic graph masked learning framework for scRNA-seq clustering. scKDGM uses graph-aware distribution preserving gene masking (GDP-Mask) to perturb cell identity, a KAN-based TAKGCN encoder to learn masked-view representations, mask-guided expression recovery to construct a dynamic graph, and cross-view contrastive learning to transfer recovery signals into topology updates. A ZINB loss models overdispersion and zero inflation. Experiments on 12 real scRNA-seq datasets show that scKDGM outperforms 10 baselines in average NMI and ARI.

Natural language processing (NLP) applications need large and rich amount of linguistic knowledge. Furthermore, electronic language sources such as dictionaries, encyclopedia, and corpora became available. So, automatic methods are emerged to extract lexical information from those sources to overcome the knowledge acquisition bottleneck. We presented a method to automatically extract lexical information from a machine-readable version of the Arabic-English Al-Mawrid dictionary. We used n-gram analysis and key-word-in-context (KWIC) analysis to discover lexical patterns that manifest morphologic, syntactic, or semantic information. Then, we used hand-crafted rule-based information extraction to extract that information. Furthermore, we used punctuation marks and some heuristics to extract a set of synonyms in a subentry. This study registered high precision for all types of information, high recall for synonyms, and low recall for the other information. The study also showed that the Al-Mawrid has significant amount of derivations (morphologic information) and synonyms, domain labels, and hyponym/hypernym relations (semantic information).

LLMs agents are increasingly used in multi-agent settings, yet their behaviour in sustainability games remains largely unexplored. This work investigates whether lying can emerge among LLM agents in a competitive sustainability game in which agents are informed that common resources can regenerate, although regeneration does not actually occur. We develop an agent-based model of a sustainability game in which agents manage industrial, military, and ecological resources, and interact through a network. LLM agents can observe neighbours' status, declare future attacks, receive permission to lie, and access reputation information, while rule-based agents provide an interpretable behavioural baseline. The results show that neighbour information strongly changes system dynamics, increasing attacks while improving biosphere retention and coexistence. Also, the presence of future declarations reduce extinction risk without suppressing conflict. Behaviourally, deception emerges even when agents are not explicitly allowed to lie, and explicit permission mainly increases bluffing and diversion rather than direct backstabbing. Finally, the presence of reputation memory and information about the current biosphere level reduces system ecological depletion. These findings suggest that deception can arise as an emergent behaviour in LLM-agent systems and that communication between LLM-agents could support sustainability while dealing with risk.

World models can predict future physical states, but prediction accuracy alone does not explain how physical information is organized and used inside their latent dynamics. We introduce a controlled diagnostic protocol for studying event-conditioned latent physical structure in passive object-state world models. The protocol tests whether hidden representations encode event-regime information, whether event contexts reweight non-exclusive physical field readouts, and whether field-aligned representational components have functional consequences for prediction. Using a balanced controlled-generator dataset with free-motion, collision, and occlusion events, we evaluate recurrent, attention-based, and latent state-space transition models under a fixed-horizon forecasting setup. The models learn useful predictive dynamics and their hidden states support reliable event-regime readout. Event contexts systematically reweight kinematic, contact, and object-permanence field readouts: free motion is kinematic-dominant, collision combines kinematic and contact structure, and occlusion combines motion-related and object-permanence structure. Time-aligned and directional-consistency analyses further show phase-related shifts in field emphasis. Finally, fixed-horizon projection causal field effect (CFE) shows that suppressing field-aligned directions can degrade event-relevant prediction, with strongest evidence for contact-aligned structure in collision-contact windows and more qualified evidence for object-permanence-aligned structure in hard-occlusion hidden windows. These results support event-conditioned organization and fixed-horizon functional sensitivity of latent physical fields, while not implying explicit physical modules, isolated causal circuits, or context-invariant sliding-window generalization.

Large language model (LLM) agents are rapidly being integrated into real-world systems. Their autonomy and tool-use capabilities generate substantial value while simultaneously expanding the security attack surface. This survey provides a comprehensive overview of the opportunities and challenges of LLM agents in security, focusing on two core areas: (1) threats to LLM agents themselves and corresponding mitigation strategies (LLM agents self-security), and (2) the role of LLM agents in empowering the cybersecurity lifecycle across offense and defense (LLM agents empowered cybersecurity). We first examine the internal and external attack surfaces of agents, propose a taxonomy organized by threat sources, and analyze associated mitigations and evaluation frameworks. We then investigate how agent capabilities are applied in cybersecurity practice and present, to our knowledge, the first agent-empowerment framework aligned with the full cyber offense-defense lifecycle. By systematically surveying these two areas, we are the first to highlight a positive feedback synergy between LLM agents self-security and empowered cybersecurity, offering new insights for the advancement of both. We further identify current limitations and outline promising directions for future research. The insights provided aim to catalyze the coordinated development of LLM agents self-security and agent empowered cybersecurity, paving the way for more capable and robust agent applications.

Retrieval-Augmented Generation (RAG) enhanced by Knowledge Graphs has shown promise in complex multi-hop reasoning tasks. However, existing graph-based retrieval methods typically rely on flat, undirected topologies. During the retrieval process, the probability flow often gets trapped in high-degree abstract concept nodes which we define as ``probability black holes'', leading to semantic drift and noise accumulation. To address this, we propose SemFlowRAG, a framework that reconstructs the flat retrieval space into a corpus-adaptive semantic gradient graph. This data-driven self-organization enables a hierarchical structure to emerge naturally from the data distribution, capturing the intrinsic semantic granularity of the corpus to suppress structural noise. By quantifying the semantic abstractness of entities through the embedding variance of their associated passages, we transform static undirected edges into directed semantic constraints. Furthermore, we design an abstractness-guided directed PageRank algorithm that forces the retrieval trajectory to follow a ``high-to-low semantic abstractness'' gradient. This mechanism ensures layer-by-layer evidence convergence, smoothly guiding the retrieval process from abstract concepts to specific document evidence. Extensive experiments on complex QA datasets demonstrate that SemFlowRAG effectively mitigates the ``probability black holes'' issue, outperforming existing baselines in both retrieval and downstream reasoning performance.

Light curves describe temporal variations in the brightness of celestial objects. Learning robust representations of light curves is essential for large-scale automatic discovery in the dynamic universe, but existing time-series foundation models often struggle with the uneven sampling, complex noise, and wide range of physical timescales that characterize astronomical observations. We propose a domain-informed representation learning framework for irregular astronomical time series with Joint-Embedding predictive architecture (JEPA), combining semantics-preserving views, uncertainty-aware tokenization, and multi-view self-distillation. The encoders are trained with multi-view self-distillation using LeJEPA regularization on the LEAVES dataset and evaluated on the StarEmbed classification benchmark. On StarEmbed, our model outperforms hand-crafted features on 15 of 16 classification metrics. In few-shot linear probing, it achieves macro-F1 scores of 42.56 $\pm$ 7.21 with one sample per class and 63.58 $\pm$ 1.20 with 100 samples per class, consistently improving over hand-crafted features. Beyond variable-star classification, the learned representation supports similarity search, parameter estimation, and photometric zero-point drift detection. We further evaluate cross-domain adaptation on 12 heterogeneous irregular time-series datasets from PYRREGULAR, where the adapted variant matches or exceeds previous state-of-the-art performance on 5 datasets, compared with at most 3 wins by any single prior baseline. These results demonstrate that domain-informed multi-view self-distillation is an effective strategy for learning representations of irregular time series, while also highlighting that successful time-series representation learning requires domain-specific inductive biases rather than a universally optimal architecture.

Early detection of dementia enables timely intervention, and reflecting cognitive impairment, spontaneous speech offers a non-invasive screening modality. Conventional approaches often focus on a single representational dimension -- such as acoustic descriptors, pause modeling, automatic speech recognition (ASR) transcripts, or multimodal fusion -- limiting integrative reasoning across heterogeneous cognitive symptoms. We propose a low-rank adaptation (LoRA)-tuned large language model (LLM) that performs structured multi-view reasoning over four complementary speech-derived signals: ASR transcripts with pause markers, discourse-level topic cues, temporal fluency statistics, and phonological sequences. These cues are encoded within a unified prompt, enabling a single LLM to learn a coherent decision function without modality-specific encoders or late-stage fusion. On ADReSSo, our best model achieves an F1-score of 90.14%, and ablation confirms the complementary contribution of each view.

Classical universal approximation theorems establish the expressive power of sigmoidal multilayer perceptrons, but they do not prescribe how initial weights should encode the geometry of a data distribution. We propose S-GAI, a spectral geometry-aware initialization framework for one-hidden-layer sigmoidal MLPs. Starting from the constructive idea that sigmoid units can act as smooth half-space gates, we move from hand-specified planar geometry to class-wise spectral geometry estimated from image data. For each class, SVD provides a mean, principal directions, and spectral scales. An energy threshold selects the retained directions, and each retained direction is represented by two sigmoid gates. These class-specific gates form a shared hidden layer initialized directly from the training set. We also formulate a SVD-based subspace classifier as a non-neural geometric reference, which tests whether the estimated spectral class geometry is already discriminative before being embedded into the MLP. Experiments on MNIST, Fashion-MNIST, and a more challenging CIFAR-10 test show that the S-GAI-initialized MLP starts from a substantially more informative hidden state than Xavier initialization and reaches comparable final accuracy under full training. When the hidden layer is frozen, training only the output layer still gives stronger performance than frozen random gates, providing evidence that S-GAI effectively embeds class-wise spectral geometry into the MLP.

Online latent state estimation constitutes a fundamental challenge within the artificial intelligence field, serving as a foundational tool for diverse applications, including sequential decision making, anomaly and change-point detection. In this paper, a novel online distributed sensing framework, where agents collaborate and exchange information to perform latent state estimation, is presented. The proposed estimator combines available partial domain knowledge with the representation capabilities of deep neural networks. In particular, the designed sensing framework incorporates prior estimates, optimized consensus weights, and Kalman-like recursive updates to perform decentralized inference, without relying on knowledge of noise statistics. Extensive experiments on linear, chaotic (Lorenz), and practical wireless tracking environments reveal that the proposed Covariance-Agnostic Neural Kalman Consensus Filter (CA-NKCF) outperforms traditional distributed Kalman and particle filters as well as purely model-free deep neural networks, exhibiting robustness even when the underlying motion and observation models are misspecified. It is also demonstrated that CA-NKCF's performance advantage remains stable across varying noise levels, random communication topologies, latent state dimensions, and observation clutter densities induced by scattering objects in wireless systems.

Deep neural networks (DNNs) are widely applied in Network-based Intrusion Detection System (NIDS) due to their high accuracy. However, DNNs are highly susceptible to adversarial attacks, which generate malicious traffic to evade NIDS detection. Existing approaches often adapt adversarial attacks from computer vision (CV) tasks to the NIDS domain, overlooking the fundamental differences between CV and NIDS. This results in two major issues: 1) The generated network traffic may become invalid, 2) The generated traffic may lose its original attack semantics. To address these issues, this paper proposes an adversarial attack specifically designed for NIDS. Instead of directly generating flow-level features, our approach incrementally generates packet-level features to construct adversarial traffic. During the generation process, the semantic integrity of the traffic is monitored at each stage, effectively avoiding the issues of invalid traffic and semantic loss observed in existing methods. We evaluate our attack algorithm against current NIDS models using the CIC-UNSW-NB15, CIC-DDoS2019, and CIC-IDS-2017 datasets. The proposed method achieves an average evasion success rate of 92.78%, while ensuring that the generated adversarial traffic remains semantically consistent with the original malicious traffic.

Recursive self-training can degrade neural generative models when generated data is reused without fresh human data or external quality control. We study this risk in code LLMs, where AI-generated code can enter real repositories, later become training data, and create a repository-scale self-training loop. While software development traditionally interrupts this loop through pull-request review, tests, compilation, and human approval, AI coding tools now produce code faster than humans can review it, and code review itself is increasingly automated by AI systems. We therefore compare three recursive fine-tuning regimes: no review, Human-gate review using model-independent filters such as compilation and static quality checks, and AI-self-gate review using the code LLM's own signals such as perplexity and binary self-scoring. Across multiple code LLMs and benchmarks, no review collapses fastest, Human-gate filters slow but do not stop collapse, and AI-self-gate filters can look strong early but later lose their filtering effect. In the clearest case, the binary self-gate enters a rubber-stamp regime where acceptance scores rise while benchmark correctness falls. We explain this behavior by formulating review as gated distributional reweighting, proving that AI self-gating degenerates to ungated self-training under a self-confirming acceptance condition, and giving a spectral analysis of representation-level covariance concentration under recursive retraining. These results suggest that stable recursive code LLM training requires exogenous verification rather than model-coupled self-review.

Program verifiers play a central role in training coding agents, including selecting trajectories for supervised fine-tuning (SFT) and providing rewards for reinforcement learning (RL). Standard execution-based verification requires running unit tests inside per-repository environments such as Docker images, incurring substantial environment setup costs. We propose Dockerless, an environment-free agentic patch verifier that evaluates generated code patches without executing them. Rather than simply matching candidate patches to references, Dockerless judges patch correctness using evidence gathered through agentic repository exploration. On a verifier evaluation benchmark, Dockerless outperforms the strongest open-source verifier by 14.3 AUC points. Using Dockerless as both the SFT trajectory filter and the RL reward enables a fully environment-free post-training pipeline. The resulting model reaches 62.0%, 50.0%, and 35.2% resolve rate on SWE-bench Verified, Multilingual, and Pro, respectively. It surpasses the Qwen3.5-9B baseline by 2.4, 8.7, and 2.9 points, matching environment-based post-training.

Long-horizon software engineering agents often need to manage lengthy and noisy interaction histories under limited context budgets. Existing memory management methods typically rely on static compression workflows or impose rigid constraints on compression timing and granularity. Moreover, these approaches fail to jointly optimize memory management and issue resolution capabilities to improve performance while reducing token usage. We present SWE-MeM, a training framework for proactive and on-demand memory management in software engineering agents. SWE-MeM provides a flexible memory tool that lets agents decide when, what, and how to compress based on trajectory state, task progress, and remaining context budget. We train agents with synthesized proactive memory-management trajectories and Memory-aware GRPO, which jointly optimizes memory management and issue resolution through memory-aware trajectory splitting and step-level credit assignment. On SWE-Bench Verified, SWE-MeM achieves 43.4% and 60.2% resolve rate with 4B and 30B models, respectively, outperforming existing memory management baselines in both performance and efficiency.

One goal in reinforcement learning (RL) research is to understand general-purpose sequential decision-making, using benchmark simulators as a proxy for learning in deployment settings. When running experiments, however, the goal of achieving high performance in the simulator can mutate into focusing exclusively on solving the simulator. To achieve high scores, researchers may adopt solutions exclusively meant for solving simulators, rather than learning while the agent is deployed outside a simulator. Solving simulators is also worthy of investigation, but it is a fundamentally different RL research question. In this paper, we argue that RL researchers need to distinguish between two use cases of simulators: solving simulators and using simulators as a proxy for learning in deployment. We first discuss how these two use-cases are importantly different, in terms of constraints on how the agent can use the simulator, which algorithms are appropriate, and which evaluation metrics are appropriate. We then highlight several issues and misleading conclusions that can occur by not making the distinction between these two settings clear, supported with examples and simple experiments. This work is a call to the community to begin clearly distinguishing how they are using simulators in their work, hopefully sparking further discussion on which empirical practices work best in each setting.

We study the spectral perturbation of the empirical Fisher Information Matrix (FIM) of a parametric statistical model under two structured perturbations: departure of the input from a reference (in-distribution) ensemble, and finite-precision (quantized) perturbation of the model's parameters. For the first, under an explicit local curvature-monotonicity hypothesis on the dominant eigenvalue lambda_max of the FIM, we show departure from a reference manifold provably elevates lambda_max relative to a calibration baseline (Proposition 3.2), and discuss why this hypothesis is required, since curvature need not increase monotonically under every perturbation. Our principal result is a directional eigenvalue perturbation bound, via Weyl's inequality, showing lambda_max under a quantization noise perturbation is lower bounded by its unperturbed value up to a third-order remainder, and, under a mild genericity condition, strictly exceeds it at leading order (Theorem 4.3). We give two tractable approximations to lambda_max -- one heuristic, one with a rigorous two-sided bound -- and a completeness result for a threshold-based partition of an augmented state space. These results motivate using sigma_t = lambda_max(F_t)/lambda_base as a runtime monitoring statistic for deployed language models: the quantization result offers a mechanism for an empirical observation of our own, where a calibration threshold for this statistic was approximately 244 times larger than a preliminary full-precision estimate on a 4-bit quantized model, a single measurement rather than a value derived in closed form. We report supporting measurements (twelve models, n=1,080 trajectories) broadly consistent with our predictions, discuss the scope and limitations of every result, and state as an open problem the closed-form prediction of the quantization inflation magnitude our bound does not supply.

Fluorescence microscopy images are degraded by noise and diffraction-induced blur, which compromise structural fidelity and limit quantitative analysis. Supervised deep learning methods achieve impressive restoration performance but require large-scale paired datasets that are difficult to obtain in practice. To address this issue, we propose SDIP, a zero-shot deep image prior (DIP) framework that sequentially performs denoising and deconvolution without external training data. An aSeqDIP-based module first suppresses noise while preserving fine structures through sequential autoencoding regularization. In the deconvolution stage, a wavelet-based background correction step is incorporated before the proposed RLG-DIP module performs artifact-reduced deconvolution. RLG-DIP uses the Richardson-Lucy deconvolution result as a physically consistent guidance prior, integrating the imaging model with the implicit prior of DIP to stabilize the ill-posed deconvolution process. Experiments on the BioSR dataset across multiple cellular structures demonstrate that SDIP improves both signal-to-noise ratio and resolution, achieving superior visual quality and improved quantitative performance on most evaluated structures. The proposed framework may also provide useful insights for designing physically guided DIP methods for other inverse problems.

Benchmarks are widely used to evaluate task completion by Large Language Models (LLMs), but this approach has accumulated construction-validity problems, and a passing score may not show whether the requested task was delivered. We study both problems. In a controlled code-as-spec setup, two production Copilot CLI agents (claude-opus-4.7, gpt-5.5) re-implement a React Fluent-UI data table in Angular as a reusable library under a hidden 222-test Playwright oracle across 18 runs and three oracle-availability conditions. Alongside the score, we run a mechanical library audit and check each verdict with a no-op ablation. Without the oracle, the library is present but unfinished, revealed by scores. With the oracle in the loop, the score reaches near-perfect, but from a demo holding the tested behavior directly, the library left dead or absent. We call this building to the test; the broader disposition behind both we call validation self-awareness. The agent does not, on its own, validate what it ships as a user would. Prevalence remains an open question across other agents, signals, and model families. Beyond benchmark scores, dispositions like validation self-awareness merit research attention.

Spatio-Temporal Logic with Graph Operators (STL-GO) extends Signal Temporal Logic (STL) to multi-agent systems via graph operators that count neighboring agents satisfying a property, together with multi-agent quantifiers. While Boolean semantics for STL-GO are well-defined, quantitative semantics have not yet been developed and existing quantitative semantics for spatio-temporal logics such as STREL cannot capture the counting constraints in STL-GO's graph operators. We develop quantitative semantics for STL-GO as a layered algebraic construction that separates temporal aggregation from graph-operator aggregation (governed by an abstract accumulator with a monotone fold and readout). We prove that soundness and completeness reduce to monotonicity conditions on these components. We implement the framework and evaluate it on two multi-agent environments: a 2D bounded region with stochastic Dubins-car dynamics and a 3D Earth-satellite system, under four semantic instantiations (Boolean, min-max, signed-deficit, and a hybrid), demonstrating the tradeoffs between accumulator choices and reporting scalability in the number of agents and time horizon.

Increasingly autonomous agentic AI systems pose novel multi-agent risks, such as secret collusion via covert communication channels. The natural defence to these collusion attempts is to monitor plain-text communication, but the efficacy of monitors has been called into doubt by increasingly sophisticated model steganography; indeed, some theoretical schemes have been proposed that are information-theoretically or computationally indistinguishable from good-faith plain-text communication. In this paper, we demonstrate that the complexity of these schemes is no longer a safety barrier, as agentic coding models can already produce undetectable stegosystems when given realistic tool usage, such as code execution or accessing research papers through web searches. Agents also adapt when key ingredients are missing, for example, by adding model-sampling components or implementing related keyed coding schemes. We then frame tacit steganographic coordination between agents as a Schelling-point problem and introduce coordination metrics for estimating when two agents are likely to select compatible schemes without explicit prior agreement. Our results suggest a shift in the threat model for covert communication between AI agents, where the main barrier is no longer whether frontier agents can understand and implement sophisticated stegosystems, but coordination: whether independently acting agents can converge on compatible schemes, keys, and parameters. We find substantial convergence on broad scheme families but limited strict one-shot coordination, suggesting that shared artefacts, repeated interaction, and tool-mediated search are the settings where covert communication risks are most acute. Overall, our findings provide empirical grounding for the recent strategic confinement hypothesis, which assumes that capable agents can construct covert channels that survive monitoring.

Agentic artificial intelligence shows great promise for radically improving the pace of innovation in hardware/software co-design research across computer architecture, systems, compilers, and VLSI. Thus far, however, applications of AI in these contexts have generally been demonstrated in isolated settings on small-scale problems, due to the difficulty of designing and deploying complex AI-infused hardware and software development workflows. This paper introduces CHIA, an open-source hardware/software co-design framework for agile and principled research on the application of AI to co-design. CHIA treats the productive construction and scalable deployment of the co-design flow itself as a first-class objective. In CHIA, agentic AI-driven hardware and software design flows are expressed as CHIA loops: directed cyclic graphs whose nodes execute various system-on-chip design tools, microarchitectural simulators, software build systems, AI models, evolutionary coding agents, and more. The CHIA library provides node implementations for many popular tools, including Chipyard, gem5, ChampSim, FireSim, Hammer (thus several commercial ASIC CAD tools), Vivado, AlphaEvolve, AdaEvolve, and many others. CHIA also provides a broad set of features to conduct principled science around these flows. These include isolation between AI models and hardware tools, profiling mechanisms, fault-tolerant execution, and reliability at scale across hundreds of heterogeneous systems (CPUs, FPGAs, GPUs, etc., across public cloud/on-prem.). To showcase CHIA, we present five CHIA loops as case studies: (1) automatic RTL-to-gem5 simulator alignment, (2) LLM-driven implementation of microarchitectural features in RTL, (3) agentic, IPC-aware critical path optimization, (4) evolutionary architectural discovery, and (5) maintainer-friendly agentic GitHub issue fixing.

Text-to-image (T2I) diffusion models typically require substantial computational resources and cloud infrastructure, posing significant challenges for edge deployment in terms of latency, cost, and user privacy. We present JuZhou 1.0, an ultra-lightweight T2I foundation model designed for fully offline, on-device execution. JuZhou 1.0 achieves its efficiency through four key designs: (1) a compact image-generation backbone consisting of a 0.385B-parameter denoising U-Net and a 1.90M-parameter distilled decoder, totaling approximately 0.387B parameters; (2) Rectified Flow training combined with DMD2 distillation, reducing inference to 4 sampling steps; (3) Chinese semantic alignment trained on 9M curated image-text pairs, enabling direct Chinese prompting without external translation at inference time; and (4) a training and distillation pipeline completed on domestically developed Sugon K100 AI accelerators without relying on NVIDIA GPUs for training or distillation. Despite its compact scale, the 28-step base model of JuZhou 1.0 achieves an overall GenEval score of 0.69, outperforming published baselines including SDXL (2.6B, 0.55), SD3-Medium (2B, 0.62), and IF-XL (4.3B, 0.61). We further validate the full poetry-to-image pipeline on Android and the core CLIP-U-Net-VAE generation branch on iOS. On a smartphone powered by the Snapdragon 8 Elite Gen 5 Mobile Platform, the 4-step U-Net denoising branch runs in approximately 1.6 seconds, while the full Android poetry-to-image pipeline takes 4.5 seconds with on-device prompt refinement on Xiaomi 17 Pro Max. These results position JuZhou 1.0 as a practical approach to mobile text-to-image generation and provide a concrete reference for Chinese-native generation, domestic-compute training, and fully offline on-device deployment after one-time installation.

The promotion of betting applications on social media platforms has increased significantly in recent years. Many of these advertisements use persuasive techniques that may mislead users, encourage risky behavior, and potentially influence users' mental well-being. However, research on the automated detection of manipulative and deceptive betting advertisements remains limited due to the lack of publicly available annotated datasets. In this work, we introduce a new dataset of betting-related advertisements collected from two widely used social media platforms, Instagram and Reddit. The advertisements were manually annotated for manipulative and deceptive advertising practices. In addition to classification labels, the dataset includes human-provided explanations that describe the reasoning behind each annotation, enabling research into explainable approaches to detecting manipulative advertising. Furthermore, we analyze the strategies commonly used in betting advertisements and examine how these persuasive tactics may impact users' mental health. The proposed framework can also enable practical applications such as browser plugins that warn users about manipulative betting advertisements and automated web crawlers that help regulatory authorities monitor and detect such promotions online.

Recurrent models must forget in order to remember, yet the state of the art decides what to erase without consulting what is stored -- the gate sees only the arriving token, not the memory it is about to modify. This memory-blind gating is one of three coupled defects in the leading delta-rule architecture (GDN-2): the value-axis erase mask wastes parameters at the scale of the value projection, and -- as we prove -- mathematically prevents the WY-form triangular chunk solver that makes recurrent training competitive with Transformers. We introduce CARVE (Content-Aware Recurrent with Value Efficiency), which resolves all three problems through one principle: erase only on the key axis. This is provably necessary and sufficient for the WY-form solver to remain valid. Within it, CARVE reuses the recurrent output tensor -- already written to GPU memory -- as a free content signal for the erase gate, and replaces the per-value write-gate projection with a single scalar per head. At initialisation CARVE is bit-identical to GDN-2; any quality difference emerges from what the content gate learns. At 1.3B parameters trained on 100B tokens, CARVE achieves WikiText perplexity 15.72 (minus 0.18 vs. GDN-2, a 4.5-sigma effect), leads every recurrent baseline on nine common-sense reasoning benchmarks, and sets state of the art on every RULER retrieval probe -- at 0.4% throughput overhead, 13% lower peak memory, and 19% fewer parameters. Six formal theorems cover memory capacity, Lyapunov stability, gradient flow, expressivity separation, Pareto-optimal chunk size, and hybrid optimality.

Limited data availability, class imbalance, and domain variability remain major barriers to reliable medical image classification. Conventional augmentation can improve training diversity but may distort diagnostically informative structures, whereas unconstrained generative augmentation may introduce label-inconsistent content. This paper proposes MedDiffuseMix, a saliency-guided diffusion mixing framework for controlled medical image augmentation. The method uses classifier-derived saliency maps to separate high-saliency diagnostic regions from low-saliency background areas and applies diffusion-guided mixing mainly to regions with lower diagnostic importance. Adaptive mixing, Gaussian boundary blending, and a saliency-preservation constraint reduce semantic distortion and reject or attenuate samples that shift model attention away from clinically relevant evidence. The framework is evaluated on four public benchmarks: the Radiological Society of North America pneumonia chest radiography dataset, Musculoskeletal Radiographs, PatchCamelyon, and the Breast Cancer Histopathological Image Classification dataset. Experiments with convolutional and transformer-based classifiers show that MedDiffuseMix improves accuracy, F1-score, and area under the receiver operating characteristic curve compared with standard augmentation, Mixup, GenMix, SaliencyMix, and diffusion-based augmentation baselines. Ablation studies confirm the importance of saliency guidance, adaptive region mixing, and smooth boundary blending. Visual attribution analysis further indicates that MedDiffuseMix better preserves diagnostically salient regions. These results suggest that saliency-guided diffusion mixing is an effective augmentation strategy for limited-data medical image classification.

Deep neural networks (DNNs) have shown outstanding performance in visual recognition tasks within vision sensor networks; however, they are still vulnerable to adversarial manipulations and imperceptible perturbations that can lead to erroneous predictions. To address that, this paper presents AEGIS, a semantic aware and uncertainty guided adversarial detection framework designed for robust image classification in vision sensors pipelines. At its core, a SemantiGAN module functions as a multi class semantic discriminator, identifying and filtering visually inconsistent adversarial inputs before they propagate further in the pipeline. For inputs that pass this stage, a stochastic augmentation process generates test time variations, from which handcrafted instability metrics FlipScore, Prediction Inconsistency, Layerwise Cosine Similarity (early and mid layers), and Entropy are computed. These features are aggregated into a compact five dimensional vector and processed by an Evidential Deep Learning (EDL) classifier, which models output evidence using a Dirichlet distribution to yield both class predictions and calibrated uncertainty estimates. Evaluations on the Tiny ImageNet dataset across six categories clean, FGSM, PGD, patch based, functional, and geometric attacks demonstrate the effectiveness of AEGIS. The proposed framework achieves an AUROC of 92.1\%, an AUPRC of 90.2\%, and an accuracy of 90.7\%, outperforming conventional softmax-based detectors in terms of detection performance, robustness, interpretability, and uncertainty calibration.

A mesh of sovereign agents has no center: no shared clock, no shared model, and no coordinator to gather data or retrain. Its competence rests on each agent folding the projections its peers emit into a single internal state, online, from observations that arrive at irregular, unscheduled times, on a substrate whose weights it cannot retrain. Any one of these constraints is tractable on its own; folding optimally under all three at once is not. We ask what such a substrate must be, and prove two necessary conditions from one model of a self-evolving latent observed at irregular, exogenous times. Because the latent changes, its optimal estimator is time-varying: an adaptive timescale is necessary, and every fixed-gain filter is strictly suboptimal. And because arrivals are clock-free, the optimal estimate depends on the elapsed gap between them, which no gap-blind network recovers at any width or depth. This second condition is capacity-independent: scale cannot substitute for the missing dependence. The two conditions intersect in the continuous-time liquid class. An LSTM satisfies the first, a fixed continuous-time filter the second, and a multi-timescale liquid network both. Synthetic experiments confirm each: the network attains the timescale, and the separation is computed exactly. The characterization is necessary, not sufficient, and binds fixed-weight substrates: a network free to retrain reaches the class by other means. Proved per agent, the necessity binds every agent of a mesh, a structural condition on mesh intelligence.

The impact of real-world noise on Open Vocabulary Object Detectors (OV-ODs) remains poorly understood due to their architectural complexity. We present our comprehensive analysis Robust Onion, an empirical study that uses controlled synthetic visual degradations to peel OV-ODs layer-by-layer, revealing how, why, and where robustness degrades, systematically analyzing feature collapse. Our findings reveal that models with similar vision backbones exhibit comparable robustness, driven by similar feature collapse at similar layers, while factors such as pretraining strategy, architectural nuances, and caption supervision contribute little. Robustness is primarily governed by the image domain rather than annotations, explaining the similar robustness impact on COCO and LVIS, and why datasets like ODinW-13 can give an impression of inflated robustness due to large, isolated objects. Finally, we validate our insights by improving robustness on real-world BDD100K, WiderFace, and VisDRONE via our lightweight plug-and-play NN & TK0 approach, using 96x fewer trainable parameters than end-to-end training. We also explain the prior works' robustness observations.

Open-vocabulary semantic segmentation (OVSS) enables text-guided segmentation of unseen objects, breaking fixed-class limitations to achieve open-world understanding. However, existing OVSS methods primarily focus on modifying the CLIP attention mechanism, which still suffers from unstable local segmentation for remote sensing (RS) domain. To address these limitations, we propose RSGPNet, a training-free geometric prompting framework for RS OVSS that refines segmentation by leveraging object geometric areas and consistency constraints. Specifically, RSGPNet comprises three core modules: a Text-guided Coarse Mask module (TCM), a Geometric Re-prompting Module (GRP), and a Coarse-to-fine Consistency Verification Mechanism (CVM). TCM utilizes text prompts and the input image to construct initial coarse segmentation masks. GRP then converts these coarse masks into geometric box prompts, feeding them back into the segmentation model to generate refined masks. Finally, CVM employs consistency computation to prevent prompting from reinforcing erroneous regions. They allow the model to improve segmentation accuracy in complex areas, such as category boundaries. Extensive experiments on RS datasets demonstrate that RSGPNet significantly outperforms state-of-the-art methods across both quantitative and qualitative metrics while exhibiting excellent interpretability. The code is released at \href{https://github.com/wangshanwen001/RSGPNet}{https://github.com/wangshanwen001/RSGPNet}.

Software-compilable C programs routinely fail to complete the four-stage pipeline of a high-level synthesis (HLS) toolchain -- compilation, C simulation (CSim), synthesis, and C/RTL co-simulation (CoSim) -- because HLS accepts only a synthesizable subset of C (HLS-C). Yet most existing large language model (LLM) systems built for HLS code repair only cover the early pipeline stages and feed raw tool logs directly to the model, yielding brittle and hard-to-reproduce fixes. We formulate C-to-HLS-C conversion as a closed-loop generation-verification-diagnosis-repair problem on an HLS tool (Xilinx Vitis), contributing three components: an end-to-end workflow of cooperating agents closed by the four-stage verifier under strict evidence isolation; a Progressive Mismatch Localization Chain (PMLC) that localizes CSim/CoSim mismatches through log normalization, AST backward slicing, and dual-trace instrumentation; and a typed-query, two-stage evidence RAG backed by a self-evolving, family-routed repair-card pool. Experimental results show that the proposed workflow substantially outperforms all comparable state-of-the-art models.

As reasoning models emit chains of thought tens of thousands of tokens long, KV cache increasingly becomes a deployment bottleneck. Existing cache eviction methods rank tokens by attention weight, which is a noisy importance proxy in long reasoning traces, and prohibits the use of fused kernels in production inference by forcing the model to materialize the attention matrix. In this work, we instead score tokens with a metric we term the epiphany score: the change in the model's internal representation, read directly from the forward pass with no attention matrix and negligible extra state. Our resulting cache eviction method, EpiKV, requires no training, classifier, or custom kernel, and can be used directly in FlashAttention inference stacks unchanged -- scaling to a 16x longer feasible context than attention-based scoring. upper-mid layers negatively) and remove a positional trend with a causal rolling z-score. At a 4096-token cache EpiKV reaches 72% on MATH-500, matching the strongest attention-based baseline (ThinKV 71%, H2O 67%); a lag-normalized KV variant reaches 37% on AIME-2024 at 8192 tokens against the best of them (33%), at up to 2.8x the speed.

Local orbital magnetization is the field whose rotation generates the equilibrium current density. Unlike spin magnetization, a quantum-mechanical local formula consistent with both this current relation and the modern theory of bulk orbital magnetization has been missing. In this work, we derive a quantum-mechanical formula for the local orbital magnetization for non-interacting electrons by considering local-flux response of the grand potential. The local-flux response fixes the formula uniquely in two dimensions, whereas in three dimensions it selects a natural representative within a longitudinal ambiguity. Furthermore, coarse graining yields a natural local marker that generates the current to third-derivative order, and its site-position moment equals the orbital magnetic quadrupole moment of finite-size systems. We illustrate the obtained results with the Haldane model.

Physical reservoir computing offers an energy-efficient alternative to conventional neural networks, where the intrinsic memory capacity in the physical system plays a central role. We demonstrate that memory capacity can be engineered extrinsically in memoryless systems by exploiting the computational space-time tradeoff, substituting temporal memory with spatial degrees of freedom. Our approach utilizes multidimensional input nodes to function as a spatial memory axis, thereby removing the dependency on intrinsic history-dependent dynamics in the reservoir. We validate this framework through numerical simulations of a generalized quantum dot, whose discrete energy levels provide strong nonlinearity crucial for reservoir computing as well. By coupling this inherent nonlinearity with our extrinsic memory, we show that memoryless quantum reservoir can achieve high performance on both chaotic Mackey-Glass future prediction and nonlinear transformation tasks. Furthermore, by analyzing the geometry of the quantum state trajectories, we identify the physical mechanism underlying this memory emergence: extrinsic memory constructs a hysteresis loop within the quantum Hilbert space, and this loop becomes topologically stable when the evolution of the system state synchronizes with the input signal's frequency. Our work decouples reservoir computing from material-specific memory properties, significantly expanding the range of candidate systems for quantum neuromorphic computing.

Relative motion between interacting systems can generate emergent energy scales that are absent in isolated systems. While uniform motion can be eliminated by a Galilean transformation, relative motion between interacting systems generally cannot. In the presence of characteristic spatial structures, relative motion gives rise to a Doppler frequency scale determined by the characteristic wavevector of the excitation and the relative velocity of the system. This emergent scale provides a fundamental mechanism for driving nonequilibrium phenomena in moving systems. In particular, the emergent energy scale is determined by how the relative motion probes the spatial structure of the relevant excitation. In this tutorial, we illustrate these ideas using magnonic systems as a concrete platform. We first discuss motion-induced magnon transport between relatively moving ferromagnets, in which the Doppler frequency serves as an effective nonequilibrium bias in the perturbative regime. This mechanism produces magnon currents even in the absence of conventional driving forces such as temperature gradients or chemical potential differences. We then introduce motion-induced parametric instabilities. When the emergent scale becomes sufficiently large to resonantly create magnon pairs, the perturbative description breaks down, and the magnonic vacuum becomes unstable. Above a critical velocity threshold, spontaneous magnon-pair creation emerges, resulting in strongly enhanced transport and nonequilibrium dynamics. Connections to related phenomena, including quantum friction, Cherenkov emission, and Zeldovich superradiance, are also highlighted. The concept of an emergent energy scale provides a unifying framework for understanding transport phenomena and instabilities in quantum systems with relative motion.

Entanglement entropy has proven to be a powerful probe of phenomena such as quantum chaos and phase transitions. Pseudo entropy is a recently proposed time-like generalization of an entanglement measure, motivated by de Sitter holography. In this work, we find that pseudo entropy can also serve as a novel probe for distinguishing topological phases of matter. For this, we consider the Su--Schrieffer--Heeger model as a representative example and investigate the averaged excess entropy $ΔS_{12}$, defined as the difference between pseudo entropy and the average entanglement entropy, across the topological-to-trivial and trivial-to-topological phase transitions. When the two states are in the same phase, we find that $ ΔS_{12}$ is non-positive under periodic boundary conditions, while for open boundary conditions, it is non-positive only when the system is sufficiently large. Moreover, we analyze ground-state quench protocols for topology-crossing quenches and find that the imaginary pseudo entropy tracks the critical times predicted by the Fisher zeros.

Room-temperature, unshielded recording of cardiac magnetic signals has remained a significant challenge since the inception of magnetocardiography (MCG). In this work, we present an MCG system based on optically pumped magnetometers (OPMs) designed to operate in ambient magnetic environments and acquire adult human cardiac magnetic fields, without the need for active or passive shielding. The system operates in a gradiometer configuration, achieving background-noise cancellation with a common-mode rejection ratio (CMRR) of 31 dB and a gradient sensitivity of 314 $\mathrm{fT/cm/\sqrt{Hz}}$. MCG signals were acquired sequentially at 16 locations across the anterior thorax, and a comprehensive signal-analysis framework incorporating wavelet multiscale principal component analysis (WMSPCA) filtering and signal quality estimation (SQE) scoring was developed to enhance signal quality. This framework yielded a QRS complex signal-to-noise ratio (SNR) of $28.56 \pm 5.61$ dB across all measurement locations. These results demonstrate the feasibility of performing clinical-grade MCG in unshielded, real-world magnetic environments, with consistent morphological fidelity across the QRS complex and T-wave segments. This work represents a meaningful step toward the practical deployment of OPM-based MCG systems in hospital and point-of-care settings.

Inferring microscopic Hamiltonians from experimental data is a central challenge in quantum materials and quantum simulation. In low-dimensional spin systems, exchange interactions are often assumed to be spatially uniform, despite structural and environmental inhomogeneities that can locally modify the coupling. Here, we leverage a local, length-independent machine learning methodology to reconstruct spatially modulated exchange interactions directly from inelastic scanning tunneling spectroscopy maps. We demonstrate this approach with nanographene spin chains, identifying both near-uniform and inhomogeneous regimes across the synthesized magnets. The reconstructed models quantitatively reproduce the experimental spectra and recover the correct scaling of the excitation gap with system size. Our results establish a general strategy to bridge local spectroscopic measurements with effective many-body Hamiltonians.

We develop a time-local nonequilibrium Green's function formulation for real-time dynamics in quantum systems coupled to superconducting leads. The superconducting lead self-energy is a strongly frequency-dependent matrix in Nambu space, giving rise to nonlocal memory kernels in the time domain. This makes direct propagation of the Kadanoff-Baym (KB) equations computationally demanding. To overcome this difficulty, we extend the auxiliary-mode expansion, originally developed for normal-metal leads, to Nambu-space self-energies. This allows us to decompose the superconducting lead self-energy into a finite number of exponential modes and to transform the KB equations with memory integrals into a closed set of ordinary differential equations. The resulting time-local equations enable efficient real-time simulations under general time-dependent bias voltages, superconducting phases, and one-body Hamiltonians of the central system, while retaining the memory effects induced by superconducting leads. As an application, we analyze voltage-quench dynamics in a superconductor-quantum-dot-superconductor junction and show that, after a dc bias is suddenly applied, the system evolves through a transient regime and relaxes to an ac Josephson periodic steady state. The resulting periodic steady-state current agrees with the Floquet Green's function solution, validating the present real-time formulation.

Cooperative emission is a collective quantum optical process that requires macroscopic phase coherence among coupled emitters. Recent observations of cooperative emission in QD superlattices have renewed interest in how such coherence emerges in nanostructured solids. Meanwhile, theoretical studies have long discussed the relationship between electronic delocalization and coherence, particularly whether delocalized states necessarily give rise to cooperative emission. This study addresses this question through power-dependent steady-state PL and time-resolved PL decay measurements. The findings indicate that, although the quantum resonance peak exhibits delocalized excitonic characteristics, it shows no signatures of cooperative radiation. In particular, neither superlinear intensity scaling nor power-dependent emission delay was observed, indicating the absence of cooperative-radiation signatures. This can be understood from two disorder-related aspects. Temperature-dependent spectroscopy reveals pronounced inhomogeneous broadening and low-temperature dark-exciton participation, pointing to intra-domain static disorder and exciton-state mixing. These effects collectively hinder the establishment of macroscopic coherence. The temperature dependence of the quantum resonance peak decay lifetime is consistent with two-dimensional exciton dynamics. This work provides direct experimental evidence that electronic delocalization can be decoupled from cooperative coherence in CdSe quantum dot superlattices.

We derive an equation describing domain wall motion in antiferromagnets under the influence of normal strain. From this equation, we find that the domain wall moves towards positions where $\varepsilon_{xx}$ is high and $\varepsilon_{zz}$ is low. Furthermore, each strain component leads to a different terminal velocity for the same strain profile. This difference arises because both strains affect the domain wall width in opposite ways: $\varepsilon_{xx}$ reduces the width, whereas $\varepsilon_{zz}$ increases it. The model is then compared with mumax$^+$ simulations for various strain profiles, including a strain gradient, an oscillating strain, and a Rayleigh wave. The comparison shows good agreement between the analytical and numerical results. Finally, we demonstrate the potential of standing surface acoustic waves as an error correction method in racetrack memory.

Pseudo entropy between quantum states at different times is generally complex, yet its imaginary part has lacked a bounded operational meaning. We show that a calibrated replica interferometer converts the pseudo-Rényi phase into a directly measurable record of transition orientation. Together with replica visibility, it exactly determines the trace distance between forward and backward ancilla outputs and hence the Helstrom-optimal single-shot success probability. At short times, the symmetrized covariance of the modular and physical Hamiltonians sets the initial distinguishability response. Under any common quantum channel, the corresponding orientation information can only decrease, with equality characterized by Petz recovery. Imaginary pseudo entropy therefore records a reversible distinction between temporal orientations, while coarse graining can make the loss of that record irreversible.

Altermagnetism is an emerging class of magnetic order characterized by momentum-dependent spin-split electronic structures despite vanishing net magnetization. Although momentum-space signatures consistent with altermagnetism have been reported in a growing number of materials, their relationship to the underlying real-space magnetic configurations remains incompletely understood, because similar spin-split electronic structures can arise from distinct magnetic orders. In the candidate d-wave altermagnet KV2Se2O, the magnetic origin of the observed momentum-dependent spin splitting has remained controversial. Here, we employ spin-polarized scanning tunnelling microscopy combined with magnetic-field-dependent quasiparticle interference imaging to determine the magnetic configuration of KV2Se2O at the atomic scale. Spin-resolved quasiparticle interference reveals a checkerboard-like antiparallel spin texture within the V2O layer and determines its interlayer spin arrangement across unit-cell step edges. Remarkably, we identify both C-type and G-type magnetic configurations, both of which generate similar spin-split electronic structures at the single-layer level but correspond to d-wave altermagnetic and conventional antiferromagnetic orders, respectively. These observations reveal a complex magnetic landscape arising from nearly degenerate magnetic states. Our results establish a direct connection between momentum-space spin splitting and real-space magnetic order, providing a framework for identifying the microscopic origin of spin-split electronic structures in altermagnetic materials.

Shearing a packing of solid granular grains can be difficult, especially when the solid fraction is high and the boundary confinement is strong. It was recently shown that embedding voids in grains can make a packing easier to shear when such voids make the grains auxetic. Here, we use finite element simulation to show that auxeticity is not a necessary condition even in a seemingly very constrained setting: shearing dense and ordered granular packings under a constant solid fraction. More specifically, by controlling the geometry of a void embedded in a grain, we induce an apparent elastic anisotropy and softening of the grain under shear, which collectively leads to a significant reduction -- up to 90\% -- of the apparent shear modulus of a packing of these grains. Complementary analysis shows that this reduction correlates well with a decrease in contact-force anisotropy, and is insensitive to system size and contact friction variation. Our results highlight how grain-scale geometry, mediated by multi-body contact mechanics, modulates macroscopic system-scale elasticity, providing a minimal design mechanism towards targeted collective mechanical properties of soft granular metamaterials.

Quantum evolution is generally expected to drive a quantum many-body system toward equilibrium. This expectation is often justified by the Hilbert-space ergodicity of generic quantum dynamics, namely, the idea that pure-state evolution explores Hilbert space uniformly up to physical constraints. Such a statement can be made rigorous by requiring the associated state ensemble to form the Haar-random ensemble, or its more structured generalization, the Scrooge ensemble. In this Letter, we report the emergence of exact Hilbert-space ergodicity in a continuously monitored quantum many-body system. For any target density matrix $σ$, we construct a continuously monitored system for which we rigorously prove that the Scrooge ensemble of $σ$ is the unique late-time equilibrium distribution of quantum trajectories. Remarkably, this requires only that the jump operators in the monitoring form a deformed unitary 1-design, a seemingly much weaker condition than full ergodicity. We numerically demonstrate our predictions by simulating continuously monitored systems whose equilibrium states are thermal states. Our results establish a rigorous mechanism for the emergence of Hilbert-space ergodicity and provide a practical route for its investigation on quantum devices.

We present a novel quantum transport model for microwave-induced resistance oscillations (MIRO) where we prove that the instantaneous scattering rate is directly modulated by the velocity of the driven coherent state. This interaction peaks exactly at $ωt = 2nπ$, where the wave packets sweep through the impurity landscape at maximum speed, breaking time-reversal symmetry to generate a net direct current. Additionally, we introduce a dephasing architecture to explain amplitude saturation: a non-linear geometric dephasing ($\exp(-A/R_c)$) triggered when the displacement amplitude $A$ of the oscillating coherent state, approaches the cyclotron radius $R_{c}$. This perfectly captures the linear-to-sublinear power crossover at high intensities, offering a fully coherent description of non-equilibrium transport.

Surface acoustic wave (SAW) phonon coupling with nitrogen-vacancy (NV) center spins in diamond offers a promising platform for on-chip quantum phononic manipulations. Although an ensemble of NV centers coupled to a common SAW phonon mode enables superradiance and collective quantum control, achieving a tunable superradiant phase transition remains challenging. Here, we show that optically driving NV centers level transitions enhances the effective spin-phonon coupling, triggering a SAW phonon superradiant phase transition in the weak-coupling regime. We also demonstrate that above a critical threshold, the driving light rapidly switches on the phonon superradiance--a dynamic effect that persists in finite-number NV ensembles. Our results provide a controllable route to coherent phonon-NV spin manipulation in solid state quantum devices.

Attaching metallic contacts to transition metal dichalcogenide nanostructures and in particular to $\textrm{MoS}_2$ has posed significant challenges over the past years. For $\textrm{MoS}_2$ nanotubes and nanoribbons, a highly promising material for field effect transistors as well as quantum electronic devices, this is even more the case due to the small, curved surface. So far all attempts there have led to a wide scatter of contact resistances on the same chip. Recently, for quasi two-dimensional, flat $\textrm{MoS}_2$ flakes, the use of semimetals has led to a breakthrough, making transparent and Ohmic contacts possible. Here, we demonstrate the steps required to reproducibly fabricate contacts to single, vapor phase grown $\textrm{MoS}_2$ nanotubes and nanowires. All devices display finite room-temperature two-point resistances in absence of gating, with a median value of $340\,\textrm{k}Ω$ in a large fabrication series. A detailed analysis elucidates the impact of the different fabrication changes.

Rydberg states of atoms in vacuum are now well recognized as a resource for quantum technologies. Donors in semiconductors also display analogous states, which have been proposed for similar applications. While they benefit from permanent locations in their host crystals, electron-lattice coupling leads to much shorter excited-state lifetimes than for neutral atoms in vacuum. Here we provide a quantitative description of donor-phonon kinetics, creating a basis for engineering donor systems in realistic material stacks for quantum devices. Our theory incorporates both form factors for the Rydberg states, which given their large extents in real space provide strong selectivity in momentum space, and tabulated deformation potentials for all six phonon branches throughout the Brillouin zone. By confronting this framework with carefully controlled time-resolved free electron laser measurements, we show that the widely quoted position of the silicon conduction band minimum, $k_0$, is inconsistent with observed donor relaxation rates and that quantitative agreement is obtained for a value further from the X-point than commonly assumed. This stringent experiment-theory comparison establishes donor relaxation as a precision metrology for conduction band parameters and scattering processes in silicon, with consequences spanning from quantum devices to classical electronics.

The Comment by J. Sirker [arXiv:2511.16509] raises an important issue concerning dynamical quantum phase transitions (DQPTs) in noisy and mixed-state dynamics, namely that the extension of the Loschmidt echo from pure to mixed states is not unique and different extensions preserve different physical properties. The Comment examines a noise-averaged mixed-state fidelity and shows that DQPTs cannot occur for any nonzero noise when the return rate is defined through the Uhlmann-Bures fidelity of the noise-averaged density matrix. This conclusion is valid for the mixed-state fidelity observable discussed in the Comment and is consistent with prior studies [https://doi.org/10.1103/PhysRevB.109.L180303, arXiv:2504.03005]. Our article [https://doi.org/10.1103/mkll-nd46] investigated a different operationally defined quantity: the logarithm of the Loschmidt echo obtained by first determining the noise-averaged excitation probabilities generated during the noisy ramp and then performing a coherent post-ramp evolution of a pure state constructed from these noise-averaged transition probabilities. As emphasized explicitly in our original publication, this observable is defined through an operational assumption and is not the same quantity as the mixed-state fidelity. The nonanalyticities reported in Ref. [https://doi.org/10.1103/mkll-nd46] therefore concern this two-stage operational protocol and should not be identified with zeros of the Uhlmann-Bures fidelity. There is therefore no direct contradiction between the theorem established for the Uhlmann-Bures return rate and the conclusions obtained for the different operational protocol studied in Ref. [https://doi.org/10.1103/mkll-nd46].

We investigate the thermodynamics of interfacial spin transport within a normal metal/ferromagnetic insulator/normal metal ($\mathrm{NM/FMI/NM}$) trilayer heterostructure, where the central magnetic layer is described by the anisotropic quantum XXZ model. By employing the self-consistent harmonic approximation (SCHA) combined with a microscopic linear response formulation, we evaluate the interfacial spin-mixing conductance $g_{\uparrow\downarrow}$ across all spin regimes. We demonstrate that $g_{\uparrow\downarrow}$ uniquely decomposes into a coherent condensed component ($g_{\mathrm{cond}}$), driven by the macroscopic phase of the spin superfluid, and an incoherent fluctuation-driven term ($g_{\mathrm{fluct}}$) mediated by stochastic thermal magnons. Crucially, in the extreme quantum limit of $S = 1/2$, $g_{\mathrm{cond}}$ drops steeply and vanishes at a finite coherence temperature $T_{\mathrm{coh}}$. Conversely, the fluctuation-driven term $g_{\mathrm{fluct}}$ vanishes at $T = 0$, exhibits a characteristic $T^2$ quadratic scaling at low temperatures, and undergoes a systematic $1/S$ amplitude suppression as the macroscopic magnetization becomes robust. Our microscopic insights bridge the gap between quantum many-body fluctuations and macroscopic spin-superfluid hydrodynamics, providing clear foundational principles for optimizing long-range coherent transport in quantum spintronic devices.

We reveal that engineering electrode-coupling configurations can fundamentally reshape coherent transport phenomena in quasiperiodic quantum systems. Leveraging nonequilibrium Green's function theory, we systematically analyze charge and heat transport, as well as current fluctuations, in a magnetic-flux-threaded quasiperiodic Su-Schrieffer-Heeger ring with both symmetric and asymmetric multi-site reservoir couplings. Contrary to the conventional expectation that optimal transport is achieved near the homogeneous-hopping limit, our results reveal that multi-site lead coupling fundamentally reshapes the transport landscape, extending the regime of enhanced transport deep into the topological phase. Strikingly, asymmetric source-drain coupling induces a disorder-assisted conducting phase where quasiperiodic modulation enhances, rather than suppresses, charge and energy transport. Magnetic flux exerts a dual influence: it activates additional interference-mediated transmission channels that amplify transport while simultaneously suppressing the disorder-induced re-entrant conducting regime. Furthermore, we uncover a flux-driven migration of the optimal transport window with increasing disorder strength, shifting from the topological regime toward the trivial-hopping regime. This behavior highlights the intricate interplay among quasiperiodicity, dimerization, magnetic-flux-induced quantum interference, and the geometry of the system-reservoir coupling. Collectively, our findings position coupling engineering as a powerful paradigm for the rational control of nonequilibrium transport in quasiperiodic materials and chart a route toward quantum device configurations in which transport characteristics can be precisely tuned via the interplay of disorder, topology, and quantum interference.

We study a two-leg ladder model consists of a one-dimensional (1D) Su-Schrieffer-Heeger (SSH) lattice with staggered nearest-neighboring hopping amplitudes and a normal 1D tight-binding lattice with uniform hopping. By varying the strength of inter-leg coupling, we find that topologically nontrivial phase with zero-energy edge modes will emerge, even when the SSH leg is in the trivial regime. Compared with the single SSH model, the nontrivial region in the parameter space is significantly expanded in the ladder. The topological phase is characterized by quantized Berry phase, and the phase boundaries are determined analytically. We also analyze the distributions of topological zero modes in the ladder, and find that the nontrivial regime can be further divided into two regions, which are separated by a gap closing point in the energy spectrum and correspond to the cases with edge modes residing in different legs. These results indicate that the topological phase and edge modes can be effectively tuned through the manipulations in the trivial lattice. Our work unveils the emergence of nontrivial topology in the ladder lattices and provides a new platform for studying topological phases.

We develop a quantum kinetic theory for Bloch electrons driven by a uniform dc electric field, extending the nonequilibrium density-matrix formalism beyond the non-crossing approximation. This extension is required to capture steady-state terms that are nominally zeroth order in disorder strength and compete with intrinsic band-geometric responses, as in anomalous Hall and related spin, orbital, and valley transport. Working in the length gauge with Gaussian white-noise disorder, we include impurity scattering to fourth order in the disorder potential. An iterative solution for the impurity-induced density-matrix fluctuations yields a connected $V^4$ collision integral after subtracting disconnected impurity pairings, thereby avoiding double counting. The resulting terms separate into self-energy corrections, ladder-type vertex renormalization, and crossed quantum-interference contributions. We clarify the correspondence between this density-matrix kinetic equation and the Keldysh formalism, and decompose the response into Fermi-surface and Fermi-sea components. As an application, we study the two-dimensional massive Dirac fermion model. We obtain analytical expressions for the single-particle lifetime, transport relaxation time, and longitudinal conductivity at the Born level, and then evaluate the anomalous Hall conductivity including crossed impurity processes. These processes generate an extrinsic contribution of order $τ^0$ that coexists with the intrinsic Berry-curvature term; for Gaussian white-noise disorder in this model, the $Ψ$-type contribution cancels while the $X$-type term remains finite. The formalism provides a consistent route for incorporating band geometry and crossed-disorder corrections into multiband transport, with applications to spin, pseudospin, orbital, and valley phenomena.

Spin-photon interfaces based on solid-state defects are key building blocks for scalable quantum networks and hybrid quantum platforms. Optimizing light-matter coupling in these systems requires precise knowledge of the optical transition dipole polarization, yet for many promising quantum emitters this quantity is hard to determine and therefore remains poorly characterized. Here, we develop a framework for reconstructing electric transition dipole polarization in spin-1/2 solid-state defects directly from ensemble spectroscopy. The approach combines the response of photoluminescence spectra to magnetic field, optical polarization, and strain. Applied to erbium ions in silicon, a particularly challenging system containing multiple crystallographic subsites, the framework identifies strain-induced shifts as the origin of asymmetric ensemble spectra and enables simultaneous determination of the optical dipole polarization and strain-orbital coupling tensor. The resulting model predicts how cavity-ion coupling depends on crystallographic orientation and magnetic-field direction, which we verify using single erbium ions coupled to a nanophotonic cavity. Together, these results establish a broadly applicable route for extracting microscopic properties of solid-state quantum emitters from ensemble spectroscopy and for engineering optimized spin-photon and spin-phonon interfaces.

Using on-chip microwave measurements, we investigate multilevel $π$-junctions formed by proximitized quantum dot (QD) in a germanium (Ge)/silicon-germanium (SiGe) heterostructure. In the multilevel regime, where several QD orbitals contribute simultaneously to superconducting transport, the Josephson ground state is no longer determined solely by the occupation of a single orbital. By combining DC transport and microwave techniques, we identify the qualitative signatures of multilevel $π$-junctions in both their gate-voltage dependence and microwave response. In particular, we observe combinations phase transitions that are sharp or smooth in gate voltage and which exhibit distinct inductive and dissipative signatures. Such multilevel Josephson transport has previously been observed primarily in exceptionally clean systems such as carbon nanotubes. Our results establish proximitized Ge as a platform for investigating hybrid superconductor/semiconductor physics and demonstrate the integration of gate-defined superconducting quantum devices with high-quality on-chip microwave resonators.

Active particles confined by hard walls accumulate at boundaries and may become dynamically adsorbed due to directional persistence. In this work, we show that the same persistence mechanism also gives rise to a finite wall splitting probability, meaning that a particle initialized at a wall can reach the opposite boundary before returning to its starting point. By comparing forward and backward evolution equations directly in position--velocity phase space, we derive exact relations linking stationary distributions and splitting probabilities for run-and-tumble, active Brownian, and active Ornstein--Uhlenbeck particles. In particular, we show that the stationary density is generated by the spatial derivative of the splitting probability, while the distribution of dynamically adsorbed particles at the walls is encoded in wall splitting probabilities. The correspondence is valid in arbitrary spatial dimension and establishes an exact bridge between stationary and first-passage descriptions of confined active matter, revealing them as complementary representations of the same persistence-driven dynamics.

Interfacial exchange coupling plays a critical role in enabling novel phenomena in magnetic heterostructures, such as spin triplet superconductivity, quantum anomalous Hall effect (QAHE), and advanced spintronic functionalities. While microscopic characterization of this coupling is essential for elucidating the underlying mechanism, it remains technically challenging. Here, using spin-polarized scanning tunneling microscopy (SP-STM) and quasiparticle interference, we directly observed interfacial exchange coupling in a magnetic tunnel junction formed by an Fe coated tip and a Cr(001) surface. We found the ferromagnetic tip induces significant energy shift (up to 10 meV) in the spin-polarized surface state of Cr(001). This shift is highly sensitive to the tip-surface distance and the spin-alignment between Fe tip and Cr surface, which can be switched by external magnetic field. Our results demonstrate that extended 2D surface states can mediate strong exchange coupling across a heterojunction, enabling local control of interfacial exchange interaction induced phenomena.

Controlling ultrafast magnetic dynamics is critical to understanding nonequilibrium spin interactions and advancing high-speed spintronics. However, a lack of efficient in situ tuning strategies leaves most ultrafast studies largely dependent on the intrinsic properties of the individual materials. Here we demonstrate continuous strain tuning of both the equilibrium magnetic response and ultrafast demagnetization dynamics in the room-temperature van der Waals ferromagnet Fe3GaTe2. Applying up to 4.2% uniaxial tensile strain increases the coercive field from nearly zero to 100 Oe, consistent with an enhancement of the effective perpendicular magnetic anisotropy. Time-resolved magneto-optical Kerr effect measurements further reveal strain-accelerated ultrafast demagnetization, with 1.2% tensile strain reducing the characteristic demagnetization time by approximately 20%. Remarkably, strain accesses an accelerated demagnetization regime that cannot be reached simply by increasing pump fluence in the unstrained sample. Combined with first-principles calculations, our results resolve that the applied strain modifies the spin-lattice energy transfer, leading to the observed accelerated demagnetization. These findings establish mechanical strain as an effective route for on-demand control of ultrafast magnetic dynamics while reducing the required optical energy by reconfiguring the magnetic energy landscape and associated spin-relaxation pathways.

We introduce a class of dynamic systems we call phase time crystals consisting of a binary assembly of particles with intermediate or long-range repulsive interactions that are subjected to a circular drive of uniform chirality in which each particle species is out of phase from the other by 180 degrees. As a function of the particle density and orbit radius, this system can organize into a rich variety of dynamical crystalline states, including one in which the out of phase particles form bound pairs that assemble into a triangular lattice. We also find stripe phases, overlapping packed crystals, disordered or phase glass states with no diffusion, mixed fluids, and different types of phase-separated states. We show that these states are robust against the addition of thermal fluctuations, and that the paired crystal can melt into a paired fluid. If the drive on each particle species is of opposite chirality, the system forms stripes and packed lattices, but no paired crystal is present. We demonstrate that by modifying the nature of the chiral driving, it is possible to realize numerous kinds of active molecular lattices, including dynamic square spin ice geometries and higher-order complex structures.

A minimal theory for the nonradiative transfer of energy from a two-dimensional (2D) exciton -- especially a moiré-localized exciton -- to a nearby graphene layer is presented. Starting from Fermi's golden rule, the transfer rate is written as the overlap between the exciton near-field spectrum and the long-wavelength electronic loss function of graphene, weighted by an exciton form factor. In the point-dipole limit the framework reproduces the established $\GET\propto z^{-4}$ law for energy transfer to graphene. Including the finite spatial extent of a moiré exciton through a Gaussian form factor with localization length $\lX$, we show that high-momentum components of the near field are filtered out for $z\lesssim\lX$, so that the transfer rate -- and hence the photoluminescence (PL) quenching -- can serve as a probe of exciton localization. Treating graphene as a gate-tunable bath, a Pauli-blocking model predicts that interband electron-hole excitations are strongly suppressed once $2|\muF|$ approaches $\hbarω$, partially restoring PL intensity and lifetime. Benchmarking against the full random-phase-approximation loss function of doped graphene confirms the minimal model to within a few percent over the relevant distance range for representative near-infrared exciton parameters. We map the resulting PL observables over experimentally relevant ranges of spacer thickness, localization length, emission energy, and Fermi level, and identify when graphene-induced quenching dominates the optical response of transition-metal dichalcogenide/hexagonal boron nitride/graphene heterostructures. A graphene gate thus acts not as a passive electrostatic element but as a tunable 2D electronic reservoir whose long-wavelength response can be probed through exciton PL quenching.

Voltage-tunable Josephson junctions (VT-JJs) are an emerging element in superconducting quantum electronics with potential to expand the functionality of conventional designs. While VT-JJs are largely compatible with wafer-scale semiconductor processing, their integration into quantum circuits remains a challenge due to unmitigated semiconductor microwave loss. Here, a deep mesa etch process, wherein the epitaxial material is removed except the VT-JJ device, will facilitate the integration of VT-JJs with low-microwave-loss circuit elements by allowing these circuit elements to be placed directly on a low-loss substrate. A Germanium quantum well is grown by Molecular Beam Epitaxy (MBE) on a float zone silicon substrate with in-situ deposited aluminum contacts. This combination allows the formation of an oxide-free superconductor-semiconductor interface. The deep mesa etch process is optimized to produce a sidewall taper sufficient for continuous metal deposition from the substrate to the top of the mesa for electrostatic gate electrodes and interconnects. The fabricated Josephson junctions demonstrate gate-tunable supercurrents with a maximum critical current over 100 nA and critical-current normal-resistance product of $8.63~μV$. These results demonstrate a pathway toward improved integration of voltage-tunable superconducting circuit elements with quantum electronic building blocks such as couplers and qubits.

Three-dimensional (3D) magnetic nanostructures offer new opportunities for controlling domain-wall (DW) configurations beyond the limitations of planar systems, providing promising architectures. However, the realization of reliable 3D magnetic devices requires precise control of geometry-dependent DW behaviour and quantitative experimental validation of the resulting magnetic states. Here, we combine TetMaG micromagnetic simulations, focused electron beam induced deposition (FEBID), and off-axis electron holography to investigate the influence of curvature and corner geometries on DW behaviour in 3D magnetic nanobridges. TetMaG simulations predict fundamentally different magnetic properties for curved and cornered geometries. Cornered nanobridges act as preferential DW pinning sites, stabilizing localized magnetic configurations and enabling controlled switching between neighbouring pinning positions. While curved nanobridges promote gradual magnetization rotation, reduced pinning, and smoother DW motion. These optimized geometries were fabricated with high structural fidelity using FEBID and subsequently characterized by quantitative electron holography. Electron holography measurements revealed magnetic induction maps that matched the simulated magnetization configurations, providing direct experimental validation of the TetMaG predictions. Curved and cornered geometries exhibited distinct DW characteristics governed by their local structural features, demonstrating the critical role of geometry in tailoring magnetic behaviour in 3D systems. Operando current-biasing experiments further revealed current-induced DW motion, including the displacement of a tail-to-tail DW into a head-to-tail configuration within corner structures.

A recent pioneering experiment [Nat. Commun. 16, 10114 (2025)] demonstrated that a driven overdamped colloidal particle in a harmonic trap immersed in a viscoelastic fluid can recuperate energy dissipated into the surrounding bath and convert it into useful work. In this article we considerably extend the original predictions. In particular, we show that energy recuperation is a generic feature of non-Markovian systems both in and out of equilibrium, even as simple as a free Brownian particle. Moreover, we demonstrate that inertia alone, even in the strong damping regime, can lead to this effect despite the absence of any external forcing. These results suggest that energy recuperation can be ubiquitous in nature and it may be the modus operandi of various phenomena in setups with memory. We show that this novel mechanism of energy recovery is the source of memory-induced ultrafast directed transport of a particle in a periodic potential in which it almost attains its top speed corresponding to the system with no energy barriers. Our results may answer from the fundamental point of view the question why the cytosol, the intracellular fluid in biological cells, is viscoelastic.

This paper introduces an index that categorizes the topology of insulating chiral magnetic systems. The position space Chern number, $C_R$ is distinct from its momentum space counterpart, $C_K$. A nonzero index guarantees the existence of topologically protected in-gap states that localize on the edge of local potential barriers in momentum space. The Chern-Simons effective field theory describing position space Chern insulators reveals a topologically quantized correlation between transverse force operators that describe the flow of quanta in momentum space. We demonstrate the existence of nonzero $C_R$ in systems hosting skyrmion magnetic phases and show how the index generalizes the classical concept of a skyrmion winding number. Lastly we investigate the competition between momentum space and position space topologies, and highlight an apparent obstruction to having systems with both $C_R \neq 0$ and $C_K \neq 0$.

In this work, we study alternating mixed-spin $(s,S)$ Heisenberg ladders in the magnetic field $h$ using density matrix renormalization group and linear spin-wave calculations. The $h$ \textit{versus} interchain coupling $J_\perp$ phase diagram for the $(1/2,1)$ case is investigated in detail. { In particular, we demonstrate the compatibility between the critical line estimates and magnetic ordering by analyzing chains with variable values of $J_\perp$ and of $h$ along the chain, $J_\perp$ and $h$ scans, and considering the usual case of chains with uniform couplings}. The magnetization plateau at 1/3 of saturation magnetization, 1/3 - plateau, is observed for $J_\perp>0$ and in a limited range for $J_\perp<0$. The critical Kosterlitz-Thouless transition point, where the 1/3 - plateau closes, is identified through a finite-size analysis of the transverse spin correlation functions.

In disordered topological materials, real-space crystalline defects interplay with momentum-space wave function singularities to \textit{enhance} the bulk photovoltaic current. What's singular is the interband Berry phase, or equivalently the phase of the \textit{optical} dipole matrix element, which has a \textit{vortex} structure in momentum space. Such \textit{optical vorticity} is guaranteed to exist in all topological materials associated with nontrivial Chern numbers. These vortices enhance electron-impurity skew scattering, which manifests as a ballistic photovoltaic current that is sensitive to (a) the topological material class, (b) the symmetry class of crystalline defects, and (c) the light polarization. This sensitivity manifests in two ways: firstly, by (a-c)-dependent frequency exponents for the photovoltaic current $\propto ω^{\text{exponent}}$ in topological semimetals, with $ω$ the frequency of the light source. Secondly, by (a-c)-dependent constraints of the bulk photovoltaic tensor, which are explainable only by emergent, \textit{magnetic} symmetries of \textit{time-reversal-invariant} topological materials. These ideas are concretized by case studies on multifold fermions, 3D $m$-order Weyl semimetals and 2D $n$-order Dirac systems, which include $n$-layer rhombohedral graphene, transition metal dichalcogenides, and topological surface states. Theoretical guidance is provided for a tri-pronged experimental program that combines frequency-tuned photoconductivity measurements, defect characterization and defect engineering.

Quantum impurity models provide a paradigmatic setting for studying Kondo screening, boundary criticality, and impurity entropies. While these phenomena are well understood in unitary systems, their fate in non-Hermitian many-body settings remains largely unexplored. We study a $\mathscr{PT}$-symmetric quantum impurity model consisting of a unitary $SU(2)_1$ Wess--Zumino--Witten bulk coupled to two impurity spins through complex-conjugate boundary Kondo interactions. Using an integrable lattice realization with $\mathscr{PT}$-symmetric boundary impurities, solved by the Bethe Ansatz and benchmarked against finite-temperature matrix-product-state calculations, we determine the impurity contribution to the free energy and entropy. In the Kondo-screened regime, where the spectrum remains entirely real and the impurities are screened by many-body Kondo clouds, we find that the impurity entropy decreases monotonically from $\ln 4$ in the ultraviolet to $0$ in the infrared. This monotonic flow persists despite the nonunitary nature of the boundary interaction, which places the system beyond the standard assumptions of the $g$-theorem.

Based on time-dependent Ginzburg-Landau field theory we show that rapid cooling through the second-order phase transition into superfluid \Hea\ excites collective modes of newly formed chiral domains, in addition to topological defects that are formed via the Kibble-Zurek mechanism. Simulations of temperature quenches in the presence of Gaussian space-time white noise generate a highly excited inhomogeneous condensate. Large-scale simulations exhibit a complex network of domain walls and vortices. We report results for the excitation of bosonic collective modes by thermal noise as well as nonequilibrium temperature quenches, followed by coarsening dynamics tracked in terms of the Fourier components of the order parameter amplitudes. For thermal states, the spectrum of bosonic excitations is defined by a power spectral density (PSD) for each mode, which is sensitive to the Langevin damping. For weak damping the PSD onsets sharply at the frequency corresponding to the mass of the bosonic mode, then decays as $1/ω$. We also track the dynamics of the order parameter following a temperature quench. We report results for the scaling exponents of Kibble-Zurek freeze-out time and correlation length as a function of quench rate for several damping rates. The dynamical exponent $z$ is shown to transition smoothly from $z=1$ to $z=2$ as the damping is increased, while the correlation length exponent, $ν\approx 1/2$, is independent of damping.

The statistics of stationary points are a powerful way to understand mean-field random landscapes, and the Kac--Rice formula is a general way to compute them. A longstanding technical barrier to these calculations is the presence of the absolute value of the determinant of the Hessian. Neglecting the absolute value produces an elegant 2-index supersymmetric representation of the problem, but is often incorrect. We develop an expanded 4-index supersymmetric representation of the complexity problem which incorporates the absolute value naturally via spontaneous supersymmetry breaking along a particular superspace direction. Positing that no additional symmetry breaking occurs implies the reduction to five order parameters corresponding to elements of a superspace operator algebra generated by the spontaneously SUSY-breaking operator. We relate the order parameters to the geometry and spectra of stationary points, showing that the SUSY-breaking order parameter corresponds to the spectral density of the Hessian at zero eigenvalue. We give examples of this formalism applied to calculate the annealed complexity of several models, including the perceptron and the Sherrington--Kirkpatrick model. The framework is naturally extended to quenched complexity, where each order parameter corresponds to a replica matrix.

The Mori-Zwanzig projection formalism constitutes a powerful and robust framework for deriving equations of motion in terms of generalized Langevin equations (GLEs) for an arbitrary observable using evolution and projection operators. Based on this framework, we analyze the properties of four distinct GLEs for a scalar observable including a Markovian force derived from a generally non-linear potential, a non-Markovian friction force, and an orthogonal force, commonly interpreted as a random force. While all four GLEs are exact, they differ in the memory friction kernel, which may either be dependent or independent of the observable, and by the potential, which may either include or exclude the effective kinetic energy of the observable. Inclusion of the kinetic energy in the potential is advantageous for observables whose velocity satisfies Wick's theorem, since this reproduces the correct distribution of the observable and its velocity even without contributions from the friction force and the orthogonal force.

Recently, a family of unconventional integrators for higher order ODEs with polynomial vector fields was proposed, based on the polarization of vector fields. The simplest instance is the by now famous Kahan discretization for first order ODEs with quadratic vector fields. All these integrators possess remarkable conservation properties. In particular, for the first and the second order Hamiltonian ODEs, the discretization by polarization possesses an integral of motion and an invariant volume form. In this note, we extend our previously proposed algebraic approach to derivation of these integrals to discretizations of ODEs of an arbitrary order. For all orders $\ge 3$, these integrals are new.

We report stable, ballistic cycler orbits in the circular restricted three-body problem: periodic trajectories that alternately undergo temporary capture about each primary. We construct continuous families of symmetric cyclers from intersections of the stable and unstable manifold tubes of the $L_1$ Lyapunov orbit and exhibit stable examples across more than two orders of magnitude in mass ratio, from the Sun--Jupiter regime to the equal-mass limit. Linear stability separates naturally into planar and out-of-plane components. The planar-stable branch of every computed family is created together with a hyperbolic branch in a saddle-center bifurcation of the return map at the family's maximal Jacobi constant, while out-of-plane instability occurs only through isolated parametric resonances. Every family examined contains a subfamily that is linearly stable to both planar and out-of-plane perturbations. We conjecture that saddle-center birth is universal among cycler families, implying that stable cyclers are a generic feature of the restricted three-body problem.

We study the long-time asymptotics of the focusing nonlinear Schrödinger equation with nonzero boundary conditions in the transition region. Biondini and Mantzavinos showed that, away from the transition curves, the \((x,t)\)-plane decomposes into two constant-amplitude plane-wave regions and a central region described by slowly modulated elliptic oscillations. However, their asymptotic formulae are not uniform near the boundaries separating these regions. The purpose of this paper is to resolve this transition problem. Using a double-scaling nonlinear steepest-descent analysis of the associated Riemann--Hilbert problem, we show that the leading term in the transition region is still a plane wave, while the first nontrivial correction is of order \(t^{-1/3} \). The coefficient of this correction is expressed in terms of a distinguished tritronquée solution of an inhomogeneous Painlevé-II equation. This Painlevé-II tritronquée structure is also known to appear in the asymptotic analysis of rogue waves of infinite order.

The Tomimatsu--Sato (TS) family, characterized by the rotation parameter $q$ and the TS index $δ=n,$ provides an important class of exact stationary axisymmetric vacuum solutions of Einstein's equations, whose integrable structure is known to be closely related to the $n$-point Toda molecule hierarchy through the Nakamura Conjecture. However, the set of equations appearing in the Nakamura Conjecture contains not only Hirota bilinear derivatives but also ordinary first-derivative terms, and therefore is not formulated entirely within the conventional bilinear algebra. In this paper we introduce a reduced trilinear formulation based on the reduced sector $(a,b,c)\rightarrow(a,b,1)$ of the $Z_3$-symmetric trilinear Hirota operators. We show that both the Hirota bilinear derivatives and the ordinary derivatives appearing in the Nakamura Conjecture can be rewritten completely within this reduced trilinear framework. Consequently, the set of equations admits a formulation in terms of reduced trilinear operators. We further show that the reduced trilinear formulation naturally inherits a Hirota-type direct method. The conventional bilinear spectral factor $k_i-k_j$ is replaced by the $Z_3$-weighted combinations $k_i+ωk_j$ and $k_i+ω^2k_j$, providing a direct-method structure characteristic of the reduced trilinear hierarchy. These results suggest that the Toda-molecule description of the Tomimatsu--Sato hierarchy may be viewed as a reduced sector of a broader trilinear framework, and provide a new perspective on the integrable structure of stationary axisymmetric gravity.

We consider solutions of the sinh-Gordon Painlevé III equation \[ u_{xx} + \frac{1}{x} u_x = \sinh u \] that are real on $(0,\infty)$. They are parametrized by the monodromy parameter $p\in\overline{\mathbb{C}}$, $|p|>1$, and an additional real parameter $s^{\mathbb{R}}$ when $p=\infty$. Our previous joint work with A. Its described the asymptotic behavior of these solutions as $x\to\infty$. Here, we describe the transition as $x, p\to \infty$, $2\Im(p)=-s^{\mathbb R}$, between singular solutions ($|p|<\infty$) and smooth solutions ($p=\infty$). In short, if we parametrize $|p|^2 = 1 + e^{2\varkappa x}$, then the smooth exponential asymptotics of the solutions extends to the region $\varkappa>1$, with a change of the leading order term at $\varkappa=2$; at $\varkappa=1$ the exponential behavior transitions into an elliptic asymptotics, which holds for all $0<\varkappa<1$; as $\varkappa$ decays to zero, elliptic asymptotics degenerates into trigonometric one, which holds for all $p$ fixed.

The dynamical order of self-sustained oscillators is often characterized by phase synchronization, extensively studied within the framework of the Kuramoto model. It has recently been reported that strong coupling leads to further organization of coupled oscillators, termed waveform proportionality (WP), through amplitude dynamics that cannot be addressed using the Kuramoto model. A previous study [Phys. Rev. Lett. 134, 167202 (2025)] showed that, in coupled oscillator systems, synchronization induces Taylor's law (TL). Particularly, it demonstrated that strong coupling gives rise to WP, which leads to TL with an exponent 2. The findings suggested that WP requires the individual oscillators constituting the coupled system to possess sufficiently fast intrinsic frequencies. Here, we show that WP and TL with an exponent 2 can be induced by a pacemaker oscillator, regardless of the magnitude of the intrinsic frequencies of the individual oscillators in a population. Specifically, even in a population composed of oscillators with slow intrinsic frequencies, WP and TL with an exponent 2 can be induced by coupling the population to a fast pacemaker. Furthermore, we demonstrate that WP and TL can also be induced in a population of non-self-oscillatory units by coupling them to a pacemaker. These results indicate that WP and TL with an exponent 2 are more universal than previously thought, extending beyond oscillator populations with fast intrinsic dynamics.

Porous electrodes couple electrical, chemical, mechanical, hydraulic, and thermal fields, yet conventional frequency-domain diagnostics interrogate only one of them: electrochemical impedance spectroscopy (EIS) the electrical response and dynamic mechanical analysis (DMA) the mechanical. Each reads a diagonal entry of the multiphysical constitutive matrix and is blind to the cross-couplings that govern structural evolution and degradation. Starting from linear irreversible thermodynamics, we formulate a general theory of multiphysical impedance spectroscopy, in which perturbing one field and measuring the conjugate response of another probes an off-diagonal entry of the constitutive matrix, recovering the static coupling coefficient and resolving its relaxation dynamics across frequency. Specializing to the electro-chemo-mechanical pathway yields a closed-form theory of mechano-electrochemical impedance spectroscopy (MEIS), in which a small harmonic current is applied and the stack stress is measured; the impedance factorizes into a chemical-accumulation term multiplying the sum of a chemo-mechanical and a poro-mechanical kernel. The porosity-accommodation bridge function is derived from a Helmholtz free energy -- following from a microstructural stiffness and viscosity rather than a fitted form -- and a three-phase (solid-fluid-void) closure interpolates continuously between unsaturated and Biot-saturated limits through a void-accommodation fraction. Non-dimensionalization reduces the spectrum to five groups, identifies the phase angle as the discriminator of the chemo-mechanical parameters, and locates the onset of second-quadrant behavior, which in a full cell arises from the competition between an expanding and a contracting electrode. MEIS emerges as one member of a family of cross-coupled spectroscopies the same framework brings within reach.

Predicting crystal structures requires navigating rugged energy landscapes in which favorable local motifs must be inherited across candidates with incompatible cells, densities, and symmetries. Conventional real-space crossover often destroys these motifs when parent structures are geometrically mismatched. Here we show that latent representations learned by pretrained universal interatomic potentials can serve as continuous evolutionary coordinates for crystal structure prediction. In the Latent Genetic Algorithm (LGA), offspring are generated by inverse optimization of atomic positions and lattice vectors to match a target latent representation, which is constructed via interpolation of the parent latent vectors. LGA suppresses high-energy and short-contact offspring, increases the HfO$_2$ ground-state recovery rate from 20-35% to 60-95%, and enables a unified variable-supercell search over 16 perovskites with a nearly tenfold reduction in search cost. Applied to (PbTiO$_3$)$_n$/(PbZrO$_3$)$_n$ superlattices, LGA reveals $\sqrt{2} \times 3\sqrt{2} \times 1$ long-period ground-state structures characterized by a common in-plane finite-$q$ modulation $q{_\parallel} = (1/6,1/6)$ and layer-coupled sidebands. To our knowledge, this in-plane periodicity has not been reported in any related oxide perovskite superlattice studies. Altogether, LGA offers a powerful representation-guided paradigm for ground-state structure prediction and provides a practical, decoder-free route toward materials inverse design.

Shadow tomography has appeared as a powerful tool for estimating observables on quantum computers from a small number of samples. We show that shadow-tomography-inspired ideas can offer similarly improved sample scaling for estimating observables on tensor network states on classical computers after proper adaptation. We develop strategies for both spin (bosonic) and fermionic systems, tailored to the contraction requirements of tensor networks, and generate scaling improvements of factors of $O(N)$ to $O(N^{3})$ (where $N$ is system size), depending on the specific task and system type. For the important and difficult task of evaluating the expectation value of long-range interacting Hamiltonians, we achieve the optimal $O(1)$ overall scaling (up to logarithmic factors) for an arbitrarily fixed relative Monte Carlo error in both spin and fermionic systems. Additionally, we show that shadow estimators offer more stable gradients of observables in variational optimization tasks than standard Monte Carlo estimators. We demonstrate practical advantage by simulating systems with long-range interactions, including the 2D long-range Heisenberg model and an ab-initio quantum chemistry Hamiltonian.

We investigate the complete dynamical pathway of photon-electron interactions involved in two-photon absorption (TPA) in a silicon nanowire waveguide using three independent high-speed measurement techniques. These methods probe different stages of the process: nonlinear photon absorption, electron excitation from the valence to the conduction band, and free-carrier generation. According to the conventional model of TPA, these three processes should occur at identical rates. However, our measurements reveal significant discrepancies between them. The measured nonlinear photon absorption is more than twice the value required to account for the measured TPA transitions, indicating the presence of additional absorption pathways or nontrivial TPA dynamics. Furthermore, the number of measured TPA transitions substantially exceeds the measured free-carrier density, indicating that long-lifetime free carriers represent only a small fraction of the TPA-excited electrons, while the majority recombine rapidly back to the valence band on a timescale shorter than 13 ps. In addition, the three stages of the TPA pathway exhibit distinct saturation behaviors at different photon densities, further indicating that the TPA process in silicon is more complex than described by the conventional model. These findings provide new insight into the physical mechanisms governing TPA, suggesting the existence of multiple competing pathways for this optical transition. A major obstacle to a complete understanding of TPA is the unclear physical origin of the virtual midgap level. The potential strategies for minimizing unwanted nonlinear losses in high-speed silicon photonic circuits, as well as for exploiting TPA in high-speed optical switching and photonic signal processing are investigated.

This work is a generalization of our previous study [arXiv:2406.17944], which dealt with the coherent and incoherent combination of linearly polarized Gaussian laser beams operating in continuous-wave mode under a $2π$ focusing configuration. In the present work, the laser beams are focused within a solid angle that can reach $4π$ by using two opposing source planes, where all the lasers from both planes are focused onto the same point in the focal plane. Our results show that the focal spot area obtained with either one or two source planes remains unchanged, despite the fact that, in the two-source-plane configuration, the numerical aperture (NA) of the system is twice that of the $2π$ focusing scheme. However, the intensity at the focal point is four times greater than that obtained with a single source plane. In addition, the longitudinal resolution is significantly improved, yielding a focal volume smaller than $0.5 λ^3$ for a numerical aperture of NA = 0.895. For incoherent beam combination, the focal volume remains practically the same in both the $2π$ and $4π$ focusing configurations, while the intensity at the focal point is twice that obtained in the $2π$ case. Furthermore, in the incoherent configuration, we observe a depth of focus reaching up to 30,000 $λ$ for a system with a low numerical aperture.

A deep defogging pipeline pretrained on controlled laboratory fog and fine-tuned with domain-randomized synthetic fog applied to clear outdoor scenes generalizes across a graded sequence of out-of-distribution settings with no target-domain training, from chamber-free free-flowing fog to iPhone video recorded through an aircraft cabin window in flight, an entirely unseen sensor, scene, and optical path. This directly addresses an open transfer limitation reported for real-world binocular defogging. Two design choices support the transfer. First, a single-camera fog imager photographs a flat-panel display through an artificial-fog enclosure with a fixed 114~mm scattering path, producing 5{,}495 pixel-aligned foggy/clear pairs. Exact registration permits a paired Laplacian ratio that predicts per-image restoration quality far better than single-image proxies (Spearman $ρ= 0.632$ versus $0.399$) and supports pixel-exact $L_1$ reconstruction training that avoids adversarial hallucination. Second, the fog-chamber checkpoint is fine-tuned on Mapillary Vistas crops overlaid with on-the-fly randomized synthetic fog spanning a broad range of strengths, spatial variations, airlights, and noise conditions. On a 552-image held-out split, a uniform comparison of 30 restoration backbones places NAFNet at the top (24.33~dB~/~0.7912~SSIM), with a compact alternative within 1.29~dB at 3\% of the parameter count, and a ResNet-50 classifier confirms that the restoration preserves semantic content rather than only pixel-level structure. On unpaired aircraft-window video, NIQE decreases from a mean of 6.22 to 4.97 after fine-tuning, with temporally stable output across full-motion sequences. The same backbone, under paired supervision, also reaches 20.71~dB~/~0.683~SSIM on a non-overlapping O-HAZE/NH-HAZE split (a transferability check rather than a competitive ranking).

Defects limit the performance of diamond in electronics and quantum technologies, yet how they nucleate from migrating point defects is rarely described as a phase transition. Here we show that dislocation-loop formation in diamond is a \emph{first-order phase transition}. We build a Ginzburg-Landau theory of it whose order parameter -- the loop area -- and coefficients are fixed directly from quantum-mechanically accurate machine-learned atomistic simulations. From simulations at nanometre and nanosecond scales, we find that carbon self-interstitials aggregate, by diffusion-recombination and lattice exchange, into line-defect motifs that seed a prismatic $\tfrac{1}{2}\langle110\rangle$ dislocation loop and two platelet-like planar defects. We also characterize the dynamics of the transition with Kramers' rate theory. The transition is strongly first-order, driven overwhelmingly ($\approx98\%$) by bond-energy reorganisation rather than elastic relief. Because these defects form \emph{intrinsically} -- from carbon interstitials alone, without nitrogen -- our results offer a nitrogen-free pathway complementary to the nitrogen-mediated routes long debated for type-Ia diamond, and a transferable framework for irradiation-induced loops.

We study the transport of light through dense opaque anisotropic metamaterials consisting of oriented nanowires. The nanowires consist of polymer photoresist that is structured by direct laser writing (DLW) with two-photon induced polymerization, with radii between $a = 0.5$ and $1~μ\text{m}$. Our flat samples have a thickness up to 9 layers, from $L = 3~μ\text{m}$ to $20~μ\text{m}$. Within each layer, the nanowires are parallel and spaced with random nearest-neighbor distances; nanowires in adjacent layers are perpendicular. The diffuse optical transmission at $λ= 633~$nm is as low as $T = 12 \%$, typical of optically dense, multiple scattering metamaterials, with a mean free path down to $\ell = 1.1~μ\text{m}$, much less than the sample thickness. It is striking that the linear polarization of the input light is maintained at the output of the dense nanowire samples, and not scrambled as in dense nanosphere arrays. Moreover, the linear output polarization faithfully tracks the input polarization. We propose that the polarization is maintained in our optically thick samples, since light is predominantly transported perpendicularly to the nanowire layers. The polarization vector then lies in the nanowire plane, consisting of a linear combination of parallel and perpendicular vectors that are both conserved upon subsequent scattering. Hence, the polarization remains independent of nanowire orientation, even after multiple scattering events. We propose that anisotropic scattering samples may find practical uses in white LEDs and its applications in lighting luminaires, optical communication, and encryption systems.

We model Herman Kahn's escalation ladder as an impartial combinatorial game. Reindexing each rung by its distance to the nuclear threshold turns the ladder into a subtraction game, the most tractable class in combinatorial game theory, and the doctrinal fact that no side wishes to fire first selects the misere convention. We prove that single-ladder stability is governed by a congruence (Theorem 4.1) and derive a ladder-design corollary that makes the burden of first escalation a function of ladder length and escalation granularity (Corollary 4.2). For simultaneous theaters we show, under normal play, that joint stability is the Nim-sum of the theater-wise escalation distances (Theorem 5.2), a condition that is neither additive nor dominated by the most dangerous theater. We then show the Nim-sum reduction fails under misere play, introduce the misere quotient as its replacement, and prove by exhaustive backward induction that for two-step escalation the quotient is the order-six monoid with generators a, b satisfying a^2=1 and b^3=b, with loss set {a,b^2} (Theorem 6.3). To our knowledge, impartial combinatorial game theory has not previously been applied to nuclear escalation ladders; the existing game-theoretic literature on escalation is classical and payoff-based.

This paper presents a two-dimensional method-of-moments (MoM) solver for electromagnetic scattering from infinitely long perfectly electrically conducting (PEC) cylinders. Both TMz and TEz polarizations are considered. Starting from the scalar Helmholtz equation, the electric field integral equation (EFIE) is derived for TMz scattering and the magnetic field integral equation (MFIE) is derived for TEz scattering. The induced surface current on the PEC boundary is expanded using pulse basis functions, and the boundary integral equations are discretized using point matching at the segment centers. Circular cylinders with radii $R = λ$ and $R = 2λ$ are used as validation cases because analytical series solutions are available. The MoM-computed surface currents, total near fields, scattered near fields, and field-error distributions are compared against the analytical solutions. After validation, the same solver is applied to a square PEC cylinder, for which no simple closed-form analytical solution is used. The results show strong agreement between the MoM and analytical circular-cylinder solutions and demonstrate the geometry-dependent scattering behavior of the square cylinder.

Nonreciprocal systems enable asymmetric energy transport and suppress backscattering, giving rise to unconventional wave phenomena. Here, we show that nonreciprocal cavities based on unidirectional waveguides exhibit a continuous modal spectrum, in contrast to conventional cavities with discrete eigenmodes. Using a ferrite-loaded microwave cavity as an example, we demonstrate that enforcing unidirectionality, by tailoring the waveguide geometry, drives a transition from discrete to continuous spectra, accompanied by strong spatial localization of electromagnetic fields. Our results reveal that dissipation alone fails to regularize these singular responses, highlighting the need for additional mechanisms to control localization in nonreciprocal systems.

A deep-etch field termination strategy using a Ga-assisted plasma-free etching technique in a low-pressure chemical vapor deposition (LPCVD) system is demonstrated for $β$-Ga$_2$O$_3$ Schottky barrier diodes (SBDs). The thermally activated etching method provides a plasma-free approach for forming deep mesa terminations while maintaining excellent device integrity. The fabricated diodes exhibit excellent forward conduction characteristics with a turn-on voltage of 1.14~V, a Schottky barrier height (SBH) of 1.15~eV, an ideality factor of 1.20, and a specific on-resistance of 3.72~m$Ω\cdot$cm$^2$, all closely matching those of the unetched planar devices. Capacitance-voltage analysis further confirms a uniform carrier concentration of $2\times10^{16}$~cm$^{-3}$ and an SBH of 1.23~eV, indicating stable electrical characteristics after deep mesa formation. Temperature-dependent electrical measurements from 25 to 250$^\circ$C demonstrate stable thermionic-emission transport behavior, with a gradual increase in on-resistance at elevated temperatures due to phonon-limited carrier mobility. Over this temperature range, the SBH decreases from 1.16 to 1.12~eV, while the ideality factor increases from 1.21 to 1.33. The leakage current remains low throughout the entire temperature range, and the rectification ratio remains above $10^{5}$ even at 250$^\circ$C. Under reverse bias, the diodes exhibit an increase in breakdown voltage from 287V to 500V, confirming the effectiveness of geometric electric-field redistribution achieved by the deep-etched mesa structure. Silvaco TCAD simulations corroborate these experimental observations by showing significant suppression of electric-field crowding near the anode edge. These results establish Ga-assisted plasma-free etching as a reliable, damage-free field termination technique for high-performance $β$-Ga$_2$O$_3$ power devices.

Full-field measurement techniques such as digital image correlation and infrared thermography are prevalent in experimental solid mechanics. Digital image correlation is used to analyze surface deformation, while infrared thermography quantifies surface temperature fields. However, sophisticated procedures are necessary to express both datasets in the same Lagrangian frame, especially when analyzing non-flat surfaces. In this study, we propose an external projection-based coupling that uses the pinhole camera model to relate two-dimensional temperature data measured by infrared thermography to three-dimensional point coordinates from stereocorrelation-based digital image correlation. Unlike existing multiview approaches, we utilize two independently calibrated industrial-grade systems and augment the experimental evaluation with the pinhole camera model. The projection matrix of the camera model is calibrated using a single image of a reference object. Through this projection, temperature fields are accurately represented at material points. Our method is particularly suited for, but not restricted to, curved surfaces and straightforward to embed in existing experimental protocols, as the image registration is kept as is. Additionally, we propose using radial basis functions as a global interpolation ansatz in both space and time to compute in-plane temperature gradients and even temperature rates on curved surfaces, thereby providing an extensive and information-rich full-field dataset.

Moment closure is a central problem in reduced descriptions of stochastic, kinetic, and quantum dynamics, where equations for low-order observables are coupled to an unresolved hierarchy of higher-order moments. Existing closures usually impose a prescribed form on the distribution or directly truncate the hierarchy, which can become inaccurate or unstable for strongly non-Gaussian states. Here we introduce unit-circle moment closure, which recasts the problem as analytic continuation. Raw moments are mapped to bounded unit-circle moments, whose unresolved tail is reconstructed by a Takagi-Prony procedure from the effective pole structure of a mapped generating function. The resulting continuation yields stable higher-order moments without assuming a fixed distributional ansatz. Illustrative static and dynamical examples demonstrate accurate reconstruction of non-Gaussian distributions and stable evolution of moment hierarchies. Our approach provides a general perspective for moment closure based on analytic structure rather than direct truncation.

The "cathetus rule" in optics alleges that the image of an object-point, formed by reflection or refraction at a surface, lies on the perpendicular ("cathetus") from the object-point to or through the surface. The first known statement of the rule, attributed to Euclid, was for a plane or spherical mirror. The rule was extended to refraction by Ptolemy.... Kepler was universally credited with the first disproof-and-salvage of the cathetus rule until 2018, when Benedetti's priority was exposed by Goulding. Kepler notwithstanding, the rule was reaffirmed by Tacquet for plane and spherical mirrors, except for the case in which the rays converge toward a point behind the eye; this became known as the "Barrovian case" because it troubled Barrow, in spite of his modern concept of an image. Barrow demolished the cathetus rule for the tangential image except in the paraxial limit, and Newton salvaged it for the sagittal image. The rule then seems to fade from history. But the rule is equivalent to the assumption that the image is stigmatic and the cathetus well defined. This narrow assumption is approximately true in the first-order (paraxial, "Gaussian") analysis of lenses and mirrors; and unacknowledged applications of the ancient rule can indeed be discerned in modern expositions of that subject. Moreover, the validity of the rule for the sagittal image fills a critical gap in meridional ray-tracing through spherical surfaces: by tracing the chief ray from an off-axis object-point, then applying the cathetus rule to the successive surfaces, one can locate successive sagittal image-points on the chief ray (produced rectilinearly through surfaces as necessary), and hence assess astigmatism to leading order, without tracing any rays outside the meridional plane.

In previous work it was argued that the cells of a multicellular organism form a classically coherent system and that such coherence is essential for life. Here we make this claim precise by introducing an explicit classical formalism in which a many-cell system is represented by a single state vector in an abstract DNA code space. Using Dirac's bra-ket notation purely as a compact representation of classical states, we construct an analogue of the center-of-mass coordinate that encodes the organismal identity and show how a common genetic code shared by all cells corresponds to a coherent phase in this space. We then map this structure onto DNA sequence space by introducing a classical Biological Hamiltonian whose generalized coordinates encode DNA codes and their cell-wise distribution, so that the organismal identity is represented by a global code state rather than by individual molecular constituents. Within this framework we define a time-dependent maintenance operator with code-correcting and code-breaking terms, weighted by coefficients $A(t)$ and $B(t)$, which captures the balance between restorative dynamics and environment-induced damage to the code. Aging is described as a slow drift in these control parameters: as $A(t)$ decreases and $B(t)$ increases, the identity state becomes less stable and the organism moves from robust code coherence to stochastic code variability. In this picture, death appears as a transition in which the global identity state can no longer be maintained.

Temporal boundaries in time-modulated media provide a powerful route toward wave manipulation beyond conventional spatial boundaries. Here, we investigate nonlinear temporal boundaries generated by interaction quenches in a synthetic lattice with exactly solvable interacting dynamics. Unlike conventional temporal boundaries arising from abrupt changes of single-particle dispersion, the present system realizes a self-induced temporal medium in which the propagating wave packet dynamically determines its own effective dispersion and transport properties. By solving the nonlinear Schrödinger dynamics analytically, we show that the interaction generates an emergent wave-packet-dependent band structure and a state-dependent temporal refractive response while preserving fully controllable evolution. Based on this framework, we establish a nonlinear temporal-scattering picture and uncover phenomena including amplitude-dependent temporal reflection/refraction and nonlinear temporal birefringence. Furthermore, we demonstrate that gradient-induced Bloch oscillations suppress wave-packet diffusion and enable coherent periodic transport with exact state reconstruction. Our results extend temporal reflection and refraction from dispersion-quenched linear systems to interaction-quenched nonlinear media and provide a tractable framework for nonlinear wave manipulation in synthetic lattices.

The rapid growth of energy-intensive technologies, including artificial intelligence, large-scale computing, and thermal management systems, has intensified global energy demand amid accelerating climate change. Meeting these demands requires innovative, low-carbon thermal management strategies that improve energy efficiency while minimizing environmental impact. This review revisits the underexplored phenomenon of bioconvection, a self-organized fluid motion generated by motile microorganisms, as a bio-inspired approach to sustainable heat transfer. Drawing on studies from natural ecosystems and laboratory experiments, we synthesize current knowledge of microorganism-induced hydrodynamics, pattern formation, and thermofluidic transport to assess the feasibility of harnessing bioconvection for thermal management. We further support this assessment through quantitative analyses of the thermal performance of bioconvective systems and discuss this in the framework of relevant non-dimensional numbers. By generating spontaneous convective plumes through density stratification, motile microorganisms enhance heat and mass transfer without external mechanical forcing. These self-organized flows provide a promising route toward hybrid bio-engineered cooling systems that reduce pumping energy, disrupt thermal boundary layers, and improve heat transfer efficiency. We conclude the review with the key challenges on the way to practical implementation, including microbial stability, material compatibility, controllability, scalability, as well as integration with existing cooling technologies. Finally, we identify critical research directions spanning heat transfer, microbiology, and nonlinear fluid mechanics within the broad context of sustainability, positioning bioconvection as a promising strategy for environmentally responsible thermal management in an era of rapidly increasing energy demand.

The inherent spatial randomness and broad spectral heterogeneity of epitaxial quantum dots (QDs) -- one of the most mature classes of solid-state quantum emitters -- remains a major obstacle to their scalable deployment in integrated photonic quantum technologies. Overcoming this challenge requires deterministic fabrication strategies capable of precisely aligning nanophotonic structures with high-quality emitters, which in turn demands efficient and automated single-QD characterization. Despite substantial progress in optical measurement techniques, a platform capable of autonomous, data-efficient, and sufficiently versatile characterization of single quantum dots at the chip scale remains lacking. Here, we introduce an automated cryogenic measurement platform that combines wide-field photoluminescence imaging with vector-stage-scanning confocal spectroscopy to enable high-throughput, chip-scale targeted optical characterization of individual QDs. Using this platform, we automatically acquire photoluminescence data from thousands of GaAs/AlGaAs QDs on a single chip. We demonstrate how this extensive dataset enables identification of high-performance emitters for future deterministic device fabrication, while simultaneously revealing statistical trends across the QD ensemble. By uniting data-efficient targeted measurements with scalable automation, our platform establishes a foundation for large-scale quantum photonic integration and the high throughput characterization framework needed to accelerate materials optimization.

Nonlinear time-dependent computation of macroscale dynamics in stellarators is motivated by laboratory results showing the possibility of robust operation in conditions where magnetohydrodynamic (MHD) modes are linearly unstable. A new formulation of semi-implicit MHD computation for toroidally shaped magnetic confinement systems uses 2D nodal spectral elements over the poloidal plane and Fourier representation over a generalized toroidal angle. Geometric mappings and steady-state (equilibrium) fields are expanded in the same 3D representation as the time-evolved fields to model non-axisymmetric configurations. For accuracy at large timestep, the semi-implicit operator is based on the ideal-MHD energy integral using 3D pressure and magnetic fields. The nodal spectral elements allow numerical convergence through either h-refinement or p- refinement. Our implementation (NIMSTELL) with the continuous H1 expansion of magnetic-field components and diUusive divergence control is a generalization of the NIMROD code [JCOMP 195, 355]. The NIMSTELL implementation is verified linearly and nonlinearly on resonant ideal interchange, where convergence from the stable side results from the stabilization method used in NIMROD [JCOMP 319, 61]. Optionally, NIMSTELL may use an H(curl) representation for vector potential, and both magnetic representations are verified with respect to results from JOREK [Phys. Plasmas 29, 063901] on linear and nonlinear magnetic tearing in the W7-A rotating-ellipse configuration. Application of the existing vector-potential implementation to interchange shows that it needs a minimum level of electrical resistivity to avoid numerical noise for a given level of spatial resolution. Solving the algebraic systems from the implicit parts of the time advance is facilitated by including the Fourier components of stellarator mode families in each preconditioning operation.

Parity-time (PT) symmetry in time-delay oscillators such as lasers and optoelectronic oscillators provides a potential route to enhanced spectral purity, including reduced phase noise and improved sidemode suppression. Existing theoretical descriptions are typically based on coupled-mode formulations derived under slowly varying envelope and near-degeneracy assumptions, which restrict their validity to weak coupling, small gain/loss contrast, and small detuning. In this work, a non-perturbative formulation of PT symmetric time-delay oscillators is developed based on a delay-difference equation and a scattering matrix representation of the coupling network. The approach treats propagation delay explicitly and does not rely on modal truncation, remaining valid for arbitrary coupling strength, gain/loss imbalance, and resonance detuning. The exact eigenvalue structure of the system is obtained in closed form, yielding a complete characterization of the unbroken and broken PT symmetric regimes as well as the associated exceptional points. A dimensionless order parameter is introduced that governs the symmetry transition over the full parameter space. It is further shown that conventional coupled-mode theory is recovered as an asymptotic limit of the exact formulation for small parameters. The results provide a unified and physically transparent framework for analysing PT symmetric delay systems beyond the weak-coupling limit, with direct implications for the design and optimisation of low-noise oscillators and photonic systems.

We develop a constructor-theoretic formulation of optical time in which delay, phase, temporal ordering, synchronization, and detector records are described as physical tasks rather than as consequences of a primitive time parameter. An optical delay is treated as an operational attribute defined by comparison and record-forming tasks, while phase becomes temporal only through a reference-dependent phase-delay equivalence relation. Within this framework, the Fisher information associated with delay estimation is interpreted as a distinguishability resource, and the Cramer-Rao bound becomes a task-impossibility statement: for a specified optical substrate, reference, detector, photon budget, bandwidth, visibility, and noise model, no constructor can estimate a delay with variance below the inverse Fisher information. We illustrate the approach using interferometric delay estimation, dispersive group-delay propagation, and double-slit diffraction, where the standard Fraunhofer pattern is recovered as a record distribution generated by a phase-delay task. The framework does not replace Maxwellian optics; it reorganizes optical dynamics as a means of determining which temporal tasks are physically possible or impossible.

In this work, we propose a robust Sparse Identification of Nonlinear Dynamics (SINDy) pipeline for handling datasets corrupted by noise and outliers. The method decouples outlier filtering from sparse regression by combining Iterative Least Trimmed Squares (ILTS) with Sequentially Thresholded Least Squares (STLS). Unlike standard approaches that treat all observations uniformly within a single regression stage, the proposed ILTS-SINDy framework first applies an ILTS procedure that iteratively minimizes the sum of the smallest squared residuals to identify the most reliable observations without prior knowledge of outliers, after which STLS is used to recover a parsimonious governing model. Extensive numerical experiments show that ILTS-SINDy can significantly outperform existing robust SINDy variants across a range of outlier contamination levels, with performance maintained even under settings with up to $20\%$ corrupted observations.

Chirality-induced spin selectivity (CISS) enables spin-dependent transport at chiral molecule/Au(111) interfaces and is used in spintronics when combined with ferromagnetic thin films in spin-valve-type hybrids. However, the influence of substrate microstructure on CISS and the related magnetization induced by the proximity of adsorbed chiral molecules (MIPAC) effect is still not well understood. In this study, we compare the effects of the adsorption of L-chiral alpha-helical alanine-rich peptides on Au/Co/Au ferromagnetic thin films fabricated by molecular beam epitaxy (MBE) and magnetron sputtering. X-ray reflectivity and X-ray diffraction show sharper interfaces and a narrower Au(111) rocking-curve width for the MBE-grown sample. However, atomic force microscopy and scanning tunneling microscopy images reveal that both sample types have locally smooth Au(111) surface regions suitable for peptide adsorption, despite clear differences in larger-scale morphology. Microscopic scanning tunneling spectroscopy after peptide exposure yields similar magnetization-direction-dependent tunneling currents in both sample types, confirming a similar magnitude CISS effect on the molecular scale. In contrast, macroscopic magneto-optical Kerr effect hysteresis loops and effect microscopy reveals that only sputter-deposited samples show slight coercivity enhancements and a consistent reduction in domain wall velocity after peptide exposure. These results suggest that microscopic CISS signatures are robust for both sample types, whereas macroscopic MIPAC-type magnetic responses are more sensitive to the substrate microstructure.

Signal recognition plays a critical role in species interactions and can be enhanced by learning signal characteristics through experience. In brood parasitism, host species may use visual cues to recognize and reject parasite eggs from their nests; because egg appearances vary within and between host individuals, a host can improve recognition by learning a tailored template of its own eggs. Nevertheless, constitutive and induced costs of learning may inhibit an extended learning period. We use a simple model of signal detection and learning to study how the benefits of learning are affected by different sources of variation in the learning signal. We find that phenotypic variation in egg appearances within a host hinders learning by adding noise to the signals, whereas genotypic variation between individuals promotes learning by carrying more information in the signals. Moreover, we consider environmental variation that can cause egg appearances to fluctuate across clutches over time. We find that such environmental variation reduces the fitness of learning hosts by distorting the signals, creating an effective cost that can offset the benefits of learning. Our results imply that learning or even a brief period of imprinting may be evolutionarily disfavored in homogeneous populations and variable environments.

Atmospheric carbon dioxide (CO2) growth rates reflects the combined influence of anthropogenic emissions, biospheric carbon exchange, and climate variability. While climate mitigation is primarily evaluated using bottom-up emission inventories within political boundaries, there is a need to validate these emission reductions using atmospheric measurements. Here, we present a global top-down analysis of atmospheric CO2 growth rates using CAMS atmospheric CO2 reanalysis, EDGAR anthropogenic emissions, GOSIF dataset and the Southern Oscillation Index (SOI) as a measures of biospheric activity, to quantify the relative influence of human and natural drivers. We find that atmospheric CO2 growth rate varies substantially across space and time but is dominated by natural carbon-cycle processes and global background trends. Anthropogenic emission signals are frequently masked by natural variability, making regional top-down detection of human emission changes difficult. The COVID-19 emission reductions in 2020, despite occurring during a neutral ENSO year, were not consistently reflected in regional atmospheric CO2 growth rates, highlighting the dominant roles of biospheric dynamics and atmospheric transport. Using unsupervised clustering and persistence analysis, we identify five characteristic carbon-cycle regimes. Spatial averaging removes much of the regional variability, leaving large-scale climate as the dominant control in most regimes. The active biosphere is the main exception, where strong biogenic signals persist, underscoring the critical role of tropical forests in shaping atmospheric CO2 variability.

In this work, we interpret developmental growth as a metabolic energy allocation problem and link the von Bertalanffy growth model to metabolic energy investments into the growth channel. Using a framework that specifies how metabolic energy is allocated among baseline maintenance, growth, and other processes, we analyse the resulting growth allocation patterns and derive direct relationships between key scaling exponents: the mass-growth exponent, the length-based exponent, the metabolic scaling exponent, and the geometric exponent, which describes the mass-length relationship. These exponents determine the metabolic investment exponent, which controls the qualitative behaviour of the growth-allocation function. Requiring the inferred allocation fraction to remain biologically feasible, we derive constraints on developmental velocity and characteristic mass scales. This provides a physical, energy-based interpretation of phenomenological growth curves and clarifies how metabolic scaling, geometric scaling, and growth dynamics are interrelated within a single allocation framework.

This study focuses on a mathematical model of Alzheimer's disease involving $β$-amyloid, cellular prion protein and their complex. The global asymptotic stability of the model indicates that the complex continues to induce neuronal damage regardless of the initial states. To investigate the dynamics of this system, we have rigorously proved that when the formation rate of new plaques is zero, the system is unconditional globally asymptotically stable without any limitation proposed in previous work. Numerical simulations further validate the theoretical analysis, regardless of the random initial state, demonstrating that the system consistently converges to a unique positive equilibrium. From a therapeutic perspective, we propose targeted therapeutic strategies and verify their effectiveness through numerical simulations. These results provide a universal theoretical basis for understanding dynamic mechanisms of Alzheimer's disease and offer critical guidance for developing targeted therapeutics.

Capsids are the protein shells that protect the genetic material of viruses. The precise structural description of capsids informs how viruses assemble and evolve and is key to the development of antiviral targets. Most viruses form icosahedral capsids; among these, most adopt quasi-spherical shapes, and some form elongated architectures. However, elongated capsids have been understudied, despite their decoupling of width and length providing greater control over their packaging capacity, a feature of particular interest in capsid evolution and in virus-based biotechnological platforms. A key bottleneck is the lack of tools for the analysis and design of elongated viral capsids. To that end, this article introduces Democapsid as a versatile tool for generating coordinates of both quasi-spherical and elongated (and shrunk) icosahedral capsids, as well as for producing customizable graphical models and publication-quality figures. The underlying algorithm builds on the generalized geometrical theory of viral capsids and employs numerical methods to assemble capsid elements based on folding constraints. It includes parameters controlling protein tiling associated with the eight regular icosahedral lattices, elongation axes (5-fold, 3-fold, and 2-fold), sphericity, and discrete body length for prolate (extended) and oblate (shrunk) shapes. It is available as a JavaScript browser application, a Python package powering plugins for UCSF ChimeraX and Blender, and an R package for generating reproducible documents with embedded models. The code (MIT License) is available on GitHub. Democapsid will benefit both researchers and graphic designers by enabling the investigation and communication of research on viral capsids and other icosahedral compartments.

We investigate linear lumping for parameter-dependent mass action reaction networks, distinguishing between generic and critical parameter regimes. For generic parameters -- those ranging in some non-empty open subset of parameter space -- we prove that exact linear lumping yields only "obvious" reductions: elimination of non-reactant species or projections along stoichiometric first integrals. This characterization extends to reaction networks with product-form kinetics, including Michaelis-Menten and Hill-type rate laws. For mass action systems we proceed to develop an algorithmic approach to identify critical parameter sets -- algebraic subvarieties in parameter space where non-trivial lumpings become available. This procedure reduces the determination of lumping maps to a system of finitely many polynomial equations. It also applies to constrained lumping scenarios (which are frequently motivated by chemical considerations). We then review and extend results about proper lumpings. Finally, we discuss lumpings of a self-replicator system, and of a two-pathway enzyme mechanism, to document the viability of our methods in relevant scenarios. Our results clarify the relationship between structural (parameter-independent) and fine-tuned (parameter-dependent) reductions, with implications for approximate lumping when system parameters lie near critical values

Background: Digital health technologies allow for frequent, remote gait monitoring in people with multiple sclerosis (MS). However, to differentiate daily variability from actual disease progression in longitudinal data, established minimal clinically important differences (MCID) are required. Currently, there is limited literature defining these thresholds for digital gait metrics. Objective: To establish MCIDs for digital gait measures reflecting progression in MS. Methods: Digital gait measures were captured via daily, remote, smartphone-based Two-Minute Walk Tests in CONSONANCE (NCT03523858), a phase 3b study of ocrelizumab in progressive MS. Using an anchor-based approach, median changes from baseline at Week 96 on digital gait measures were computed for patients showing clinically meaningful worsening on either Timed 25-Foot Walk, Ambulation Score, Expanded Disability Status Scale, or 12-item Multiple Sclerosis Walking Scale. These changes were subsequently triangulated to derive the MCID estimates. Results: 243 patients with progressive MS (female: n=125 (51%); mean [SD] age: 49.3 [9.3]; mean [SD] EDSS: 4.8 [1.4]) had digital gait data available at baseline and Week 96. Median changes were generally consistent across anchors. Triangulated MCIDs are: Step Velocity = -0.16 m/s, Step Velocity Scaled to Walking Time = -0.18 m/s, Step Duration = 0.06 s, Step Length = -0.07 m, Total Number of Steps = -28, and Total Distance Walked = -24 m. Conclusion: These MCIDs provide a framework for interpreting meaningful gait changes and integrating digital measures into MS outcome evaluation. Beyond facilitating novel clinical trial endpoints to evaluate treatment efficacy, they enable objective, real-world monitoring to advance personalized patient care.

We propose a hybrid active-online learning framework for label-efficient concept drift adaptation in optical network failure detection. Using margin-based selective labeling, our method achieves nearceiling accuracy and AUC scores while querying only 3.4% of streaming samples, with negligible latency overhead compared to static inference.

Binary compressive sensing (BCS) seeks to recover a $k$-sparse binary vector of length $n$ from $m$ linear measurements. Classical CS guarantees break down for $m < k$ and convex/greedy BCS algorithms with random Gaussian sensing matrices perform poorly. We introduce ISDP-MVSIC, which combines randomized semidefinite programming (SDP) sampling, majority voting (MV) and successive interference cancellation (SIC) across $L \ll n$ stages, wrapped in a residual-cost driven retry loop. The method exposes a tunable complexity--performance trade-off: for $n=100, 144$, raising the worst-case complexity $\mathcal{C}_{max}$ from $7.9 \times 10^9$ to $2.0 \times 10^{10}$ enables empirical exact recovery over $m/k \in [0.4,5.0]$ as the sparsity ratio $s=k/n$ decreases from $0.5$ to $0.1$, by practically targeting the undersampled regime.

The growing demands for higher throughput and cost-efficient wireless communications drive the need for receivers that are both simple to deploy and robust to hardware impairments and nonlinear environments. While classical model-based receivers and recently proposed deep neural network ( DNN) architectures provide complementary benefits, they either rely on simplified linear Gaussian assumptions, require considerable computational resources, or are tailored for a given setting and modulation. In this work, we propose a compact and modular DNN augmentation that universally refines the soft outputs of existing receivers (model-based or data-driven), addressing two distinct operating regimes: structurally incomplete soft information arising from reduced-complexity detectors, and degraded soft outputs caused by hardware impairments and synchronization errors. A key property of the proposed framework is its task-agnostic nature: operating without any knowledge of the specific source of unreliability, it produces well-calibrated log-likelihood ratios (LLRs) suitable for channel decoding. Our design leverages an element-wise scaled convolutional neural network tailored to perform learned interference cancellation across users and neighboring subcarriers, combined with a training algorithm that encourages accurate LLR s for soft channel decoding. Numerical results demonstrate that the proposed augmentation consistently improves diverse receiver algorithms in challenging channel conditions while incurring minimal overhead.

This letter addresses long-term fairness in uplink non-orthogonal multiple access integrated sensing and communication (NOMA-ISAC) systems. Existing resource allocation schemes that maximize instantaneous sum rate often favor strong users, leaving historically underserved users with poor long-term throughput. We propose PF-JUGPA, a proportional-fair scheduling based joint user grouping and power allocation method. PF-JUGPA first pre-selects users via a PF metric combining instantaneous rate proxies and historical averages, then performs fairness-aware grouping and power allocation by maximizing a weighted sum rate whose weights are inversely proportional to historical service rates. Simulation results show that PF-JUGPA significantly improves the Jain fairness index and weak-user average rates with only a modest sum-rate loss compared to sum-rate-oriented and round-robin baselines. The findings confirm that embedding long-term service history into both scheduling and resource allocation yields an effective throughput--fairness--sensing tradeoff in uplink NOMA-ISAC.

The non-terrestrial network (NTN) architecture has gained significant interest from the academia owing to its versatility and the ability to provide worldwide service. To achieve extremely high data rates in NTNs, as intended in the sixth-generation (6G) communication systems, millimeter wave (mmWave) and terahertz (THz) frequencies can be considered, enabling substantial bandwidth and data transmission capacity, which makes them highly suitable for NTN applications. However, these high-frequency signals suffer from significant propagation challenges, including atmospheric attenuation, pointing errors, and various environmental effects. Therefore, a comprehensive link budget analysis is essential to accurately assess the feasibility of mmWave/THz-based NTN systems. Existing studies in the literature often fail to fully capture certain frequency-, altitude-, and direction-dependent effects observed in mmWave/THz transmission or possible communication scenarios within the NTN architecture. In particular, while most prior works primarily focus on free-space loss or atmospheric attenuation, this study adopts a much more comprehensive approach. In this work, a detailed link budget analysis is conducted for mmWave/THz NTNs, considering free-space loss, atmospheric absorption, weather-induced effects, ionospheric disturbances, polarization mismatches, feeder losses, antenna and circuitry constraints, fading, pointing errors, and non-white noise characteristics. The results have revealed that the multi-layer structure of the NTN architecture can help reducing the excessive loss levels to a certain level that can be tolerated by high-gain directional antennas/arrays, providing multi-gigabit links and making mmWave/THz NTNs feasible for 6G communication systems.

High-altitude platform station (HAPS) massive multiple-input multiple-output (MIMO) systems are expected to support wide-area, low-latency, and energy-efficient connectivity in future non-terrestrial networks. However, Doppler-induced channel aging, finite-rate feedback quantization, packet loss, and estimation noise impair transmitter-side channel state information (CSI), making robust downlink beamforming challenging. In HAPS channels, these impairments are strongly structured by elevation-dependent Rician propagation and line-of-sight (LoS)-dominant geometry, whereas conventional robust beamforming methods often rely on generic uncertainty models and computationally intensive optimization. This paper develops a physics-informed uncertainty-aware beamforming framework for HAPS massive MIMO systems under imperfect CSI. First, a geometry-aware channel and feedback-impairment model is developed, where CSI errors due to aging, quantization, packet loss, and noise are represented through tangent-space ellipsoidal uncertainty sets. Second, a physics-informed variational autoencoder (VAE) exploits the LoS-dominant steering manifold to enhance channel direction information and propagate learned uncertainty through unit-sphere projection. Third, the learned uncertainty representation is embedded into a robust energy-efficiency maximization formulation with probabilistic QoS awareness. To enable scalable online operation, the resulting beamforming policy is approximated using a multi-agent deterministic policy gradient framework with centralized training, decentralized execution, and differentiable power projection. Simulation results show that the proposed framework improves energy efficiency, SINR robustness, outage reliability, convergence behavior, and online runtime compared with imperfect-CSI, SDR-based, and no-VAE baselines.

A defining feature of 6G networks is that performance depends not only on the quantity of available resources (e.g., spectrum, antennas, cache memory, compute, and fronthaul bandwidth) but also on their \emph{fungibility}, i.e., the ability of one resource to substitute for another under changing conditions. We argue that the fungibility landscape of a distributed 6G system is governed by two coupled decision scales: \emph{micro} decisions made locally by agents and \emph{macro} outcomes that emerge at the network level. Existing distributed-optimization approaches largely conflate these scales. To address this gap, we develop an agent-based-modeling (ABM) framework that separates macro and micro decisions through three operator-controllable macro choices, three micro hyperparameters, and three structural metrics. We establish six key results: (i) a two-timescale decomposition theorem, (ii) a structural-metric basis theorem, (iii) a macro--micro design rule with closed-form factorization of the emergent breakdown threshold, (iv) a fungibility--resilience monotonicity proposition, (v) a connectivity--substitutability duality theorem, and (vi) a multi-application generalization proposition. Numerical results visualize the macro fungibility landscape and the micro decision-sensitivity region for a representative 6G deployment.

Machine learning for underwater acoustics is constrained by the scarcity of publicly available labeled datasets. In contrast to air-acoustic domains, where large benchmarks enable rapid model development, underwater datasets are typically small and limited in acoustic diversity, restricting robust model training and cross-domain generalization. To help address this gap, we introduce a curated underwater audio dataset derived from an open-source maritime sound archive. The dataset contains over one thousand labeled audio segments across eight biologically and mechanically relevant acoustic classes, providing an additional resource for training models in data-limited underwater environments. Additionally, we establish a lightweight Convolutional Neural Network (CNN) baseline and propose a margin-enhanced loss with feature alignment to mitigate class confusion arising from data imbalance, acoustic similarity, and cross-domain mismatch. While the baseline achieves 96.35% in-domain accuracy, evaluation on ShipsEar reveals substantial domain shift; the proposed feature alignment improve zero-shot ship detection by 42.60%, demonstrating stronger robustness under distribution mismatch. We further release a transparent curation pipeline and reproducible benchmark to support future research on imbalance mitigation, domain adaptation, and data-efficient underwater acoustic classification.

We propose and analyze a communication-centric reconfigurable intelligent surface (RIS)-assisted integrated sensing and communication (ISAC) system, in which a monostatic radar simultaneously senses a moving target and serves a user equipment (UE) over Nakagami-m fading. We design a dual-function phase-modulated continuous-wave (PMCW) waveform that embeds the data stream directly into the radar pulse train: each pulse carries one full maximum-length sequence whose polarity is flipped by a binary phase-shift keying data symbol, so that the same emission preserves the sharp range autocorrelation required for sensing while conveying one bit per pulse to the UE. We further propose a communication-centric RIS phase configuration that co-phases each element onto the direct radar-to-UE path, yielding a coherent superposition at the UE and a received-power gain that scales with the square of the number of elements. We show that from the radar's perspective, however, the same surface behaves as an uncontrolled scatterer, since the resulting reflection paths are mis-phased and do not benefit from array combining. We derive a closed-form approximation for the average UE bit error rate based on a moment-matched Gamma approximation, and we show that the same waveform still forms a usable range-Doppler map for sensing. Monte-Carlo simulations corroborate the analytical results.

The performance of rate-splitting multiple access (RSMA) can be severely affected by imperfect successive interference cancellation (SIC) in practical wireless systems. This paper investigates a downlink pinching antenna system (PASS)-assisted RSMA network under imperfect SIC, where residual common-stream interference is explicitly incorporated into private-stream decoding. To improve user fairness, a max-min rate optimization problem is formulated through the joint design of antenna activation, common-rate allocation, and power allocation. The resulting mixed-integer non-convex problem is addressed using a two-stage framework that combines greedy channel-aware antenna activation with successive convex approximation (SCA)-based resource allocation. Numerical results demonstrate the effectiveness of the proposed framework in improving fairness under imperfect SIC.

As antenna arrays grow, near-field effects become non-negligible in large-scale MIMO, making accurate low-overhead channel acquisition crucial in both far-field and near-field regimes. Deep-learning-based frequency-domain channel extrapolation can reduce pilot overhead, but existing extrapolators generalize poorly to unseen distances and environments, especially across near-field and far-field channels. We propose a physically interpretable framework to unify generalization across both regimes. Our key insight is that angular profiles are regime-dependent, while delay profiles share a sparsity structure that can be aligned. Based on this, we develop a physics-guided disentanglement and alignment pipeline with multi-cluster decoupling, angle-delay feature disentanglement, and delay-domain alignment, enabling the model to learn distribution-stable delay features while reusing heterogeneous angular features. We further design a unified near/far-field DL extrapolator (UNiFi-DLE) and detail its dataset preparation, training, and inference. Simulations and sim-to-real experiments show that UNiFi-DLE generalizes robustly to unseen near-field and far-field scenarios and consistently outperforms state-of-the-art methods.

Wireless foundation models (WFMs) have recently emerged as a promising paradigm for learning multiple channel state information (CSI) acquisition tasks. However, unlike natural language tokens governed by statistical co-occurrence, wireless channels are generated by electromagnetic propagation laws, and current WFM training is constrained by limited data scale, narrow distribution coverage dominated by simulations, and a pronounced sim-to-real gap. As a result, simply scaling model parameters and CSI samples does not necessarily yield robust and generalizable models. In this paper, we advocate enabling physics equivariance as a principled and explainable inductive bias for WFMs. Specifically, we focus on a universal propagation property for electromagnetic waves, termed wave equivariance: when the input CSI is modulated along time-frequency-space dimensions, the output channel response should exhibit the corresponding transformation. Empirical studies show that the vanilla-WFM fails to reliably acquire such equivariance even with a large number of model parameters and training samples. To address this, we design the physics-intrinsic WFM (phys-WFM) with wave equivariance, which explicitly aligns model behaviors with an interpretable wave propagation structure. Results demonstrate that the proposed design effectively captures wave equivariance and substantially improves robustness and generalization to unseen environments under distribution shift, offering a physics-grounded and testable route toward explainable wireless foundation models.

We study the capacity of an additive white Gaussian noise (AWGN) channel followed by a subtractive dithered uniform quantizer. Under the Schuchman conditions and with negligible overload probability, the system admits an additive-noise representation in which the effective noise is the sum of Gaussian and uniform components. Capacity bounds are derived for this model when inputs are subject to an average-power constraint as well as a peak-amplitude constraint, where the latter accounts for the limited quantizer dynamic range. Specifically, a computable lower bound is obtained based on the entropy power inequality (EPI), using the maximum-entropy input under the above constraints. Tighter numerical lower bounds are derived using discrete input constellations with finite mass points. Finally, an upper bound is obtained by exploiting the fact that Gaussian distributions maximize entropy under a variance constraint. Numerical results show that, for a K-level quantizer, discrete constellations with K mass points already achieve near-optimal rates among the tested families. Moreover, our upper bound is close to the lower bounds in the moderate-SNR regime; it thus represents a good and simple capacity approximation in this regime.

Power delivery -- including high-to-low voltage conversion, complex power distribution across heterogeneously integrated chiplets, and efficient interconnect allocation -- remains a critical bottleneck in high-performance computing (HPC) systems. Existing vertical power delivery (VPD) solutions are estimated to achieve less than 70\% system-wide end-to-end power delivery efficiency, defined from platform input power to delivered on-chip load power, with substantial energy lost as heat before reaching on-chip point-of-loads (POLs). In the absence of systematic design methodologies, evaluating power quality, exploring architectural alternatives, and optimizing performance rely on computationally prohibitive simulations, resulting in suboptimal designs. This paper introduces an end-to-end scalable power delivery framework for HPC systems, including distributed VPD (DVPD) architecture, DVPD design optimization methodology, and analytical models. The framework leverages substrate-embedded GaN power switches together with arrays of unit inductors and capacitors tailored for HPC applications. Multi-stage power conversion schemes (48V-to-1V, 48V-to-24V-to-1V, and 48V-to-12V-to-1V) are explored, with system-wide voltage drops and power losses evaluated under steady-state conditions. Design specifications for passive and active devices are formulated to meet next-generation efficiency targets. For the 48V-to-1V case, the proposed DVPD approach achieves 84\% system-wide efficiency while occupying 54\% of the area beneath the load system, with efficiency increasing to 87.6\% at 75\% area utilization across a 1--50~kW load range. Furthermore, steady-state voltage drops peak at 2.7\% and transient drops at 9\% (without decoupling capacitors), demonstrating the viability of DVPD for future wafer-scale HPC platforms.

This article surveys spatial-domain-enhanced Physical-layer Authentication (PLA), with Dual-polarized Antennas (DPA), Massive Multiple-Input Multiple-Output (MIMO), and Reconfigurable Intelligent Surfaces (RIS) as the primary focus. With the rapid growth of wireless deployments, authentication mechanisms face stringent requirements for high security, low overhead, and low latency. PLA offers lightweight identity verification by exploiting physical-layer characteristics. However, the effectiveness of PLA critically depends on how physical observations are constructed and validated under wireless channels. Unlike existing surveys that mainly organize PLA by authentication modality, feature source, and evaluation metrics, this work emphasizes the connection between spatial-domain enhancement mechanisms, the resulting feature representation, and the authentication procedure. We review how DPA, Massive MIMO, and RIS reshape PLA feature representation, and we summarize newly introduced security threats along with representative defense strategies. Case studies further illustrate the practical impact, such as representative detection-probability trends across Signal-to-Noise Ratio regimes and quantitative comparisons among representative schemes. Finally, we outline promising future opportunities enabled by Dynamic Metasurface Antennas, Extra-large MIMO, and spatial configuration with artificial intelligence.

Specific emitter identification (SEI) determines which known emitter a received signal originates from, while specific emitter verification (SEV) determines whether the received signal genuinely comes from its claimed emitter. In this paper, we consider the effect of wireless fading channels on SEI and SEV. When the Rician $K$-factor varies, the resulting distribution shift induced by the channel degrades both identification and verification performance. To address this issue, we first theoretically prove that the coefficient of variation of the signal envelope is strictly monotonic with respect to the Rician $K$-factor. Motivated by this property, we propose an envelope-guided adaptive feature modulation (EAFM) identifier for SEI and an EAFM with Mahalanobis distance metric learning (EAFM-MD) verifier for SEV. Specifically, the proposed EAFM identifier adopts a dual-branch neural network to extract device-oriented features from the IQ-domain input and channel-conditioning features from the normalized signal envelope, and adaptively modulates the former via feature-wise linear modulation. Then, we extend the EAFM identifier to an EAFM-MD verifier. The device-fingerprint library is constructed by storing the feature centroid and covariance for each enrolled device, along with the within-device Mahalanobis distances of training signals. For verification, the Mahalanobis distance between the extracted test features and each stored centroid is computed using the stored covariance matrix, and the minimum distance is compared to the corresponding device threshold to make a decision. Finally, numerical results show that the proposed EAFM identifier improves cross-channel identification performance, while the proposed EAFM-MD verifier achieves superior detection performance against unknown spoofing attacks.

Rapid and reliable post-earthquake damage assessment is critical for public safety, re-occupancy decisions, and effective emergency response. This paper presents a physics-informed, unsupervised learning framework that enables structural damage diagnosis in sparsely instrumented buildings following seismic events. The approach fuses real sensor data with physics-based simulations to create a hybrid spatiotemporal input grid, extending observability to regions without sensors. A Spatiotemporal Composite Autoencoder Network (SCAN) processes this hybrid input, learning the structure's undamaged behavior from pre-event or ambient-condition data alone. SCAN integrates convolutional layers for extracting localized spatial features and LSTM layers for modeling temporal dynamics, enabling it to recognize deviations from normal behavior caused by damage. Post-event sensor data are analyzed through the trained model, and anomalies are flagged based on elevated reconstruction and prediction errors. These error patterns are spatially mapped to localize potential damage, even in uninstrumented areas. By embedding low-fidelity physical estimates directly into the input representation, the framework enhances detection sensitivity while requiring no labeled damage examples. This hybrid AI-physics approach offers a scalable, interpretable, and data-efficient solution for real-time post-earthquake structural diagnostics, providing critical decision support for emergency managers and accelerating safe and targeted recovery efforts.

Efficient and intelligent post-earthquake structural damage assessment is critical for rapid disaster response. Although data-driven approaches have shown promise in this domain, traditional supervised learning relies on large labeled datasets that are impractical to obtain for earthquake-damaged structures. To overcome this limitation, we propose a Bayesian-optimized multisource domain adaptation framework for predicting post-earthquake structural damage on a target building without the need for any damage labels. The framework comprises three key steps. First, it extracts features from multiple source domains and the target domain and feeds them into a classifier and a domain discriminator. The classifier ensures the features remain damage-sensitive, while the discriminator promotes their invariance across domains. Second, the framework assigns a weighing factor to each source domain to balance their contributions during training. Finally, Bayesian optimization is employed to optimize these source domain weights, aiming to maximize prediction accuracy on the target domain. This framework offers a robust solution for structural damage assessment when labeled data are scarce, significantly enhancing post-earthquake damage assessment capabilities.

Accurate assessment of sweetness is essential for quality control in agriculture, yet conventional methods rely on destructive sampling and are difficult to scale. This thesis presents a robotic arm-based spectral sensing system for strawberry detection, localization, approach, and non-destructive sweetness estimation. The system integrates perception, calibration, and robotic control in a closed-loop pipeline. A YOLOv11s detector is adopted for real-time strawberry detection, while RGB-ToF calibration and mask-to-depth alignment are used to obtain geometrically consistent target localization. A custom eye-in-hand hand-eye calibration workflow is developed to estimate the rigid transform between gripper_link and cam_front, enabling reliable transformation of fruit targets into the robot base frame. Based on these estimates, the robot executes a waypoint-based search and an incremental closed-loop approach strategy to position the sensor at optimal working distance for sweetness sensing. Experimental results show strong end-to-end performance (88.10% success over 42 trials), with robust detection (95.24%) and successful approach execution once a target is detected (100% conditional success). Hand-eye calibration comparisons indicate that although Andreff yields the smallest translation norm in single-run results, the Park method provides better cross-sample consistency and therefore more stable downstream robot behavior. The residual failures are concentrated in the sensing stage, especially valid-region extraction for sweetness estimation under difficult depth/reflectance conditions. Overall, this work demonstrates the feasibility of integrating RGB-ToF perception, robotic manipulation, and non-destructive sensing for practical strawberry quality assessment, and provides a scalable baseline for future integration of learning-based policies such as Vision-Language-Action models.

Beamfocusing is the established near-field strategy for a large array serving a single-antenna user. We consider the single-user line-of-sight MIMO link, free of multipath, in which the user, too, carries an extended aperture, and show that the focusing prescription inverts: beyond a modest Fresnel number, focusing on the user is outperformed by far-field steering. Under fully analog, unit-modulus beamforming, we derive closed-form power gains for focusing (each aperture phase-matched to the other's center) and for steering (a planar phase ramp) in the Fresnel regime, and prove that their comparison is governed by two dimensionless quantities: the link Fresnel number, the product of the two aperture lengths normalized by wavelength and link distance, and the aperture ratio, irrespective of how many elements discretize the apertures. For equal apertures the two gains cross exactly once, at the universal value 1.947; beyond it, focusing loses ten dB per decade of Fresnel number, and the advantage celebrated in the MISO literature survives only as the receive aperture vanishes. We then derive the strategy that is order-optimal at every Fresnel number, con-focusing: both apertures aim at the common point from which they subtend equal angles. It attains the rank-one eigenbound in leading constant, needs no channel knowledge, degenerates to plain steering for equal apertures, and is acquirable within one beam-refinement round with no geometry exchange between the terminals.

A subwavelength quantum-photonic circuit element should simultaneously generate nonclassical light, suppress plasmonic loss, and remain dynamically tunable. We show that an orthogonal plasmonic nanorod dimer can satisfy all three requirements. A phase-locked control polarization induces plasmonic refractive-index enhancement, driving the probe response toward a near-zero-extinction regime while simultaneously tuning the local second-harmonic parametric interaction. The resulting nonlinear plasmonic source operates in an absorption-suppressed regime and enables all-optical control of quantum correlations. We demonstrate switchable logarithmic negativity and single-mode nonclassicality, establishing a route toward actively tunable quantum-plasmonic circuit elements operating well below the diffraction limit.

Superconducting transmon systems are promising platforms for nanoscale thermometry due to their high sensitivity to environmental fluctuations. Their intrinsic anharmonicity, which is essential for qubit operations, gives rise to a non-equidistant energy spectrum that significantly affects the thermal populations and, consequently, the thermometric sensitivity. In this work, we investigate the ultimate quantum limits of temperature estimation with a transmon beyond the two-level approximation. We compare the thermometric performance of three complementary models: the qubit, a harmonic oscillator and a weakly anharmonic Duffing oscillator, evaluating their corresponding quantum Fisher information (QFI) as a function of the temperature. We show that the multilevel anharmonic structure of the transmon affects its thermometric precision. Indeed, including higher excited states enhances the maximum amount of information that can be extracted about the system temperature, compared to a qubit probe. Furthermore, we address a fundamental limitation of the standard quartic truncation, which yields a potential that is unbounded from below and supports only spurious metastable states. By introducing bounded anharmonic models, namely a confined quartic potential and a sextic correction term, we assess the robustness of the thermometric precision beyond the Duffing regime. Our results provide practical guidelines for transmon-based nanoscale thermometry and clarify the role of the anharmonic multilevel spectrum in quantum temperature estimation.

We demonstrate that light propagation in reciprocal chiral photonic media admits a unified description in terms of an emergent Dirac structure in helicity space. Starting from Maxwell's equations, we reformulate the electromagnetic field as a four-component spinor governed by an effective non-Hermitian optical Dirac equation. In this representation, the magnetoelectric response of the chiral medium appears as a helicity-dependent background that modifies the spectrum and eigenmodes, while the breaking of the spin-degenerate condition generates the intrinsic spin-orbit coupling between helicity and orbital degrees of freedom. After projection onto the positive-frequency sector, the theory reduces to an exact two-level helicity-orbital model. This model is found to have an analytical solution and describes coherent Rabi-like oscillations between spin-orbit-coupled vector modes. Chirality controls the helicity splitting and detuning, whereas the electromagnetic mismatch of the medium determines the coupling strength responsible for oscillatory spin-orbit conversion. The resulting dynamics is constrained by exact conservation of the total angular momentum, leading to reversible conversion between spin and orbital angular momentum with well-defined selection rules. Our work establishes an optical Dirac framework for structured light in chiral media, and provides experimentally accessible predictions for chirality-controlled oscillations, polarization dynamics, and orbital angular momentum conversion in structured optical fields.

A central challenge in quantum algorithms and cryptography is reasoning about algorithms with oracle access to a random group element (e.g. a random function, permutation, or unitary). Can we efficiently simulate such algorithms? Can we determine what they know after t queries? A classical tool for this is lazy sampling: the oracle does not commit to the full group element upfront, but rather samples partial information about it on the fly. We study a quantum analog of lazy sampling: compressed oracles (or recording oracles). These are quantum data structures that allow on-the-fly simulation for quantum queries, originally introduced by Zhandry (CRYPTO '19) for random functions, and generalized to unitaries by Ma-Huang (STOC '25) and permutations by Carolan (STOC '26), and used to great effect in security proofs and lower bounds due to their interpretability. We define and analyze a general-purpose and interpretable path-recording oracle, derived from first principles, that perfectly simulates random elements of any closed subgroup of $U(N)$. Our oracle stores, in superposition, t input-output pairs, with updates described in terms of the commutant of the group's tensor power representation. This transparently records the information the algorithm has learned. Our oracle builds on recent work of Grinko-Yoshida (QIP '26), who gave a different general-purpose compressed oracle without clear interpretability. One interesting application of our path-recording is allowing direct comparisons between compressed oracles of different groups, giving a new technique for proving pseudorandomness results. For example, comparing $S_N$ and $U(N)$ yields what is arguably the simplest construction to date of pseudorandom unitaries: the product PC of a pseudorandom permutation and a random Clifford, improving on the prior PFC construction (Metger-Poremba-Sinha-Yuen, FOCS '24; Ma-Huang, STOC '25).

The Bose-Hubbard system has been studied extensively both theoretically and experimentally, in particular in the context of ultracold atomic gases in optical lattices. Even in the two-mode case the many-particle dynamics display complex interference effects resulting in revival and breakdown phenomena as well as tunnelling. The most basic theoretical description is the mean-field approximation, which can be derived from a time-dependent variational principle assuming the many-particle wave function is an SU(2) coherent state. Here we build on this to construct a simple initial-value coherent state propagator, summing over mean-field trajectories and keeping track of their phases, given by the corresponding mean-field actions. This yields an approximation to the full time-dependent many-particle state, and is able to reproduce breakdown and revival dynamics. Applying a time-slicing procedure on top of this, we are able to accurately capture many-particle tunnelling effects. While in this paper we focus our analysis on the Bose-Hubbard dimer, the methods developed can be applied to more general SU(2) Hamiltonians, and can be extended to SU(M) systems.

In this paper, we study interacting bosons at zero temperature in a random and higher-dimensional continuum model introduced by Kac and Luttinger. For weak interactions we prove that there is condensation in the lowest eigenstate of the one-particle Hamiltonian (type-I BEC). For strong interactions, however, we show that condensation in a localized state cannot occur. We also prove generalized condensation, where a family of eigenstates of the one-particle Hamiltonian is macroscopically occupied as a whole. Combining these results yields a scenario where there is generalized condensation into a family of eigenstates of the one-particle Hamiltonian, but none of them is macroscopically occupied itself (type-III BEC). This proves a transition in the type of condensation. To the best of our knowledge, this is the first rigorous result in this direction for a random continuum model in higher dimensions.

We compute invariants of Carrollian spacetimes, deriving them from the geometry of the screen bundle. For generic Carrollian structures we specify how to generate the entire algebra of differential invariants, with emphasis on dimension 3, which has special physical relevance. Then, in the framework of jet-spaces, we compute the numerology behind these invariants: the Hilbert and Poincaré functions that govern their numbers according to order. Finally, we compute the Spencer cohomology behind the Carrollian geometry that, in particular, contains the spaces of intrinsic torsion and intrinsic curvature, which are fundamental invariants, important in the equivalence problem and symmetry analysis. Thus, we also discuss symmetry sizes of Carrollian spacetimes.

Teaching quantum mechanics is challenging, not least because the theory often conflicts with our classical worldview. Quantum correlations in particular are notoriously counter-intuitive. Their non-classical behavior is typically revealed through Bell-type inequalities. Among these, Wigner's Inequality constitutes a particularly accessible test, as it relies on minimal set-theoretic assumptions. In this pedagogical paper, we derive Wigner's Inequality, describe a quantum-optical setup to experimentally violate it, and provide access to the raw data, enabling students and instructors to perform their own analyses. Our measured data shows clear violations of Wigner's Inequality, directly illustrating the non-classical nature of quantum correlations. By connecting theory, experiment, and data analysis, this paper equips educators with a resource for engaging students in authentic scientific practice and developing a deeper understanding of quantum systems.

We investigate the nonlinear optical response of a two-dimensional mesoscopic quantum ring subjected to a spiral dislocation, with emphasis on third-harmonic generation (THG). The topological defect is modeled through a torsion-induced deformation of space, which modifies the effective metric without introducing curvature. By combining the minimal-coupling prescription in curved space with a radial ring confinement and a perpendicular magnetic field, we derive the effective radial Schrödinger problem, obtain the bound states, and evaluate the nonlinear susceptibilities within the electric-dipole approximation. We show that the axial symmetry of the topologically deformed ring preserves the dipole selection rule $Δm=\pm 1$ and therefore suppresses second-harmonic generation, while THG remains allowed through multistep transition chains. The study is further expanded through three complementary analyses that can be implemented without changing the Hamiltonian: a dephasing-controlled study of spectral resolution, three-dimensional waterfall spectra showing the dependence on $β$ and $B$, and a channel-resolved decomposition of the THG amplitude. Together, these results establish the spiral dislocation as a robust geometric knob for tuning nonlinear optical activity in mesoscopic ring-shaped nanostructures.

Transport networks often lose functionality through gradual degradation in link operating conditions before topological disconnection occurs. Link-centred and binary percolation measures identify important facilities or connectivity failures, but they provide limited information on which spatial areas cause the largest loss of network-wide propagation capability. This paper develops a Dynamic Propagation Connectivity (DPC) metric based on quantum percolation for critical-area identification in transport networks. Time-varying link travel times are converted into continuous propagation strengths, which define a Hermitian propagation operator at each observation time. Candidate regions are then evaluated by a regional degradation experiment that measures the resulting loss of DPC. The method is applied to a benchmark Sioux Falls network and six Florida road networks during the post-Hurricane Irma disruption and recovery period, using 1,281 five-minute observation times. The benchmark confirms that the regional DPC score identifies a predefined structurally critical corridor. In the Florida networks, the identified critical areas differ from regions selected by link count, local degradation, edge betweenness, algebraic connectivity, and classical percolation. In Networks 1 to 4, DPC and classical percolation rankings have negative Spearman correlations, showing that continuous propagation degradation and binary fragmentation reveal different vulnerability patterns. Robustness tests under alternative travel time scaling, degradation strength, and grid size show stable results, with mean rank agreement between 0.84 and 0.96. The findings extend transport resilience analysis based on percolation from binary connectivity loss to continuous propagation degradation and provide a spatial diagnostic tool for regional monitoring, emergency planning, and recovery prioritisation.

Quantum fields exhibit a rich entanglement structure which is still not fully understood. In this work, we study the entanglement structure of the vacuum state of a massless scalar field in (1+1)-dimensions -- a paradigmatic case for both high energy and condensed matter physics. We fully characterize the entanglement negativity between two arbitrary compact spacelike-separated regions of the field by calculating the logarithmic negativity along with the modes carrying it, called negativity cores. We achieve this using a framework based on the Kähler structure of Gaussian states, wherein we calculate the diagonalization of the operator associated with the partially-transposed restricted linear complex structure. In doing so, we extend the methods of this framework by proposing a basis-independent definition of the transpose operation. The explicit diagonalization we perform is enabled by a reformulation of the eigenvalue problem as a boundary value problem in the complex plane. Our results also suggest extensions to higher dimensions and fermionic fields.

Achieving precision measurements of neutrino oscillation parameters and resolving parameter degeneracies remain central challenges in neutrino physics. This work presents a systematic investigation of three-flavor neutrino oscillations within the framework of quantum estimation theory using the quantum Fisher information matrix (QFIM). The behavior of all six independent elements of the QFIM associated with the parameters theta23, deltaCP, and Delta(m31)^2 is analyzed, and the impact of parameter correlations on the quantum Cramér-Rao bound is studied. Furthermore, we demonstrate that parameter degeneracies in neutrino oscillation probabilities do not necessarily imply indistinguishability of the underlying quantum states. By employing quantum fidelity and the QFIM, we show that degenerate parameter sets can exhibit distinct quantum-information characteristics that remain hidden at the probability level, revealing quantum-state differences between probability-degenerate solutions.

Connectivity degradation in transport networks under structural disturbance is a central problem in network resilience research. Existing methods rely mainly on percolation theory and topological connectivity measures. They focus on whether paths exist and whether connected components fragment. These approaches cannot capture functional degradation where network topology remains intact but propagation ability has already declined substantially. This paper introduces quantum percolation theory into transport network connectivity analysis and proposes Dynamic Propagation Connectivity (DPC) as a new measure that characterises network propagation ability under disturbance. By mapping a transport network under disturbance into a propagation operator system, this paper establishes a spectral analysis framework for DPC and defines the time-averaged participation index as its core quantification. This paper provides a series of rigorous theoretical results. DPC remains constant under homogeneous disturbance and degrades under heterogeneous disturbance. This paper establishes a quantitative relationship between the degradation rate, the minimum eigenvalue spacing of the propagation operator, and heterogeneous deviation strength. This paper proves a separation theorem between DPC and algebraic connectivity. It derives an analytical expression for DPC and a second-order perturbation approximation on the ring graph. Numerical experiments on three transport benchmark networks verify all theoretical conclusions and confirm degradation monotonicity, separation from algebraic connectivity, and degradation amplification by network size. This paper provides a theoretical framework for transport network resilience assessment that goes beyond topological connectivity.

We investigate geodesic motion, shadow observables and thin-disk accretion for the regular black hole generated by a non-local gravitational self-energy contribution. The geometry is controlled by a zero-point length and can be followed smoothly from the Schwarzschild limit to a cold extremal remnant. We compute the photon ring, critical impact parameter, apparent shadow radius, null Lyapunov exponent, innermost stable circular orbit, orbital frequencies and Novikov-Thorne flux profile. As the self-energy scale grows, both the photon ring and the ISCO move outward, while the photon-ring frequency and instability exponent decrease. The horizon-normalized shadow area increases by about a factor of 2.7 for the near-extremal benchmark, although the same shadow radius decreases when normalized by the ADM mass. The ISCO binding efficiency grows modestly, whereas the zero-torque thin-disk flux peak moves outward and falls to about 61% of the Schwarzschild peak as the solution approaches the remnant regime. These trends identify a coherent set of optical, orbital and accretion signatures of the gravitational-self-energy regularization.

Symmetries plays a significant role in understanding the conservation laws in Quantum field theories. Here, we attempted a quadratic type dimensionless gauge transformation to achieve the invariance in QFTs. We have shown the extensive study of invariance of complex scalar, Abelian and Non- Abelian theories and established the conservation laws. We included an explicit graphical analysis to invoke the invariance. This is studied in a physical context, where different field configurations correspond to the same physical state. The necessity of the covariant derivative is studied in detail, highlighting how it ensures consistent transformation under local symmetry operations. The meaning of covariance is clarified as the preservation of the form of physical laws under transformations.

Gravitational wave (GW) propagation is usually studied under the geometric optics approximation. But when GWs propagate through structures of sizes similar to their wavelength, this approximation breaks down. Going beyond the geometric optics approximation allows us to explore the wave optics effects in curved background that appear in such cases. In this work, we present a scheme for numerically evolving linearised plane GWs through stationary, spherical astrophysical structures in both weak and strong gravity regimes. Our simulations evolve the full Einstein equations (with all 10 components) for Gaussian, NFW and Burkert potentials, although in simplified form for the two latter. Our simulations show that the scattering of the GWs depends not only on the mass of the lens but also strongly on the gravitational potential distribution of the lens. We isolate the effects of diffraction by setting the wavelength of GW to be less than the Schwarzschild radius of the structure. Among our most important results, we find that the GWs do not propagate along null geodesics when propagating through the Gaussian density, neither in the strong nor weak gravity setting. We also find that for the Burkert potential, the convexity of the plane wave is flipped when leaving the structure, in the strong gravity case. We compare our results with the linearized scalar wave predictions and find that the difference between these and the exact GW modes are of order one when the wave is inside the central potential. However, the difference reduces to only a few percent when the wave has passed through the structure. Although these effects are small, future GW detectors and Pulsar Timing Arrays (PTAs) could be sensitive to these signals which could thus potentially help in constraining the structure of dark matter spikes or halos.

Decoherence is the biggest bottleneck in all quantum technologies. For nitrogen-vacancy (NV) centers in diamond, the loss of coherence is caused by the electron and nuclear spin bath of the diamond lattice. Here, we demonstrate that the spin bath - that typically causes decoherence - entangles the spin states of the NV electron and the host $^{14}$N nucleus. The many-body interaction between the $^{14}$N nucleus - electron - bath spins at an energy level anti-crossing occurring for an applied magnetic field orientation perpendicular to the NV axis is responsible for this effect. This is observed experimentally on NV ensembles via electron spin-echo measurements, where the echo envelope is modulated at the frequency of a $^{14}$N nuclear spin transition. Using numerical simulations, we show that the spin bath coupling to the NV centers is essential for observing this modulation. Due to the zero first-order Zeeman effect at the anti-crossing, the observed oscillations have long spin-echo coherence times, 2--3 times those at the parallel magnetic field orientation. The oscillation frequency is highly stable and robust against environmental fluctuations. These findings provide new opportunities for fundamental studies of many-body physics and quantum sensing.

Gravitational waves can be lensed by intervening potentials of various scales. Strong lensing leads to underestimated distances and overestimated masses, biasing astrophysical results if not accounted for. I present a novel analysis of the LIGO-Virgo-KAGRA catalog of binary black-hole mergers, simultaneously inferring (1) whether or not each event is strongly lensed, (2) their magnifications if so, and (3) the underlying merger population, using both parametric and nonparametric population models as well as two models for the lensing optical depth. Posterior lensing probabilities do not exceed 1% for any event, so population constraints are consistent with those assuming nondetection of strong lensing or that lensing never occurs. This includes multiple subpopulations over black-hole mass and a component with high aligned spins. Compared to standard analyses, however, there are reductions of order 10% in uncertainty on the redshift at which the merger rate peaks and an order of magnitude in high-redshift rate upper limits. Though modest, these are the first constraints using only resolved events at redshifts where current ground-based gravitational-wave detectors are usually insensitive, at and beyond the peak of star formation.

Variational quantum algorithms (VQAs) have emerged as a promising approach to quantum cryptanalysis on noisy intermediate-scale quantum (NISQ) devices. Although numerous variational attack schemes have been proposed for symmetric cryptosystems, a systematic and modular benchmarking framework to evaluate their performance is still lacking. In this work, we present a comprehensive benchmark study of variational quantum attacks on the Simplified Data Encryption Standard (S-DES), focusing on the modular design choices that determine attack efficiency. We formulate variational quantum attacks within a unified framework consisting of four components: initial state preparation, parameterized circuit (Ansatz) design, cost function construction, and classical optimization. Through numerical simulations, we systematically compare representative design alternatives and evaluate their combinations in terms of convergence behavior, success probability, and effective time complexity. We further introduce standardized metrics for assessing variational quantum attack performance. Our results reveal clear performance hierarchies among different modular configurations and show that carefully optimized designs can significantly outperform naive quantum search. This work establishes a principled benchmark methodology for variational quantum cryptanalysis and positions S-DES as a practical testbed for evaluating quantum attacks on symmetric ciphers in the NISQ era.

We investigate the reconstruction of holographic duals for strongly coupled quantum field theories in regimes characterized by large hierarchies and the presence of false vacua. Within the gauge/gravity duality, these features translate into non-trivial thermodynamic behaviour and exotic renormalization group flows, including skipping flows between non-adjacent fixed points. Building on previous work based on Physics-Informed Neural Networks (PINNs), we extend the holographic inverse problem of reconstructing the bulk scalar potential from boundary thermodynamic data into this new regime. This setting presents a variety of conceptual and numerical challenges, such as near-degenerate states, large hierarchies of energy scales, and regions of the potential that are not directly probed by the input data. We develop a set of methodological advances that overcome these obstacles, thereby improving the established PINNs-based methodology and extending it to new physical regimes of interest that were previously out of reach. Applying the developed framework, we demonstrate accurate reconstruction of scalar potentials deep into the false vacuum regime, achieving robust agreement with the physical features of the underlying thermodynamics despite significant numerical stiffness. Our results extend the bridge between holography and machine learning, and suggest that data-driven approaches can provide new insights into the structure of strongly coupled systems.

We investigate quantum synchronization in a coupled cavity optomagnonic system which consists of two spatially separated optical whispering-gallery-mode (WGM) resonators and each resonator is also coupled to a yttrium iron garnet (YIG) sphere through the optomagnonic interaction. Phase-dependent single-photon hopping factor couples the two optical resonators and provides an indirect interaction between the two distant magnon modes. We then investigate complete synchronization, φ-synchronization, and quantum phase synchronization using the covariance-matrix formalism as well as also studying the effects of the hopping term on the overall synchronization dynamics of two distant magnon modes. It can be seen that the photon-hopping phase provides an efficient way to control the synchronization dynamics and when it is varied from 0 to π, the magnon trajectories gradually evolve from weakly correlated motion to a highly synchronized state, which is also accompanied by a significant reduction in the synchronization error. The influence of the photon-hopping strength and thermal fluctuations is also investigated, where it can be seen that stronger photon hopping enhances all synchronization measures, while thermal noise weakens the coherent correlations responsible for synchronized dynamics. Our results demonstrate that the phase of the hopping factor offers a simple and effective approach for controlling synchronization dynamics in WGM based coupled cavity optomagnonic systems and also provide a useful route towards coherent control of collective magnon dynamics in such quantum optomganonic devices.

Brans-Dicke theory is reformulated within the Herglotz variational principle (HVP), and an exact black hole solution with scalar hair is obtained for $ω_{0}=0$ and vanishing potential $V(φ)=0$. The scalar profile is strictly positive and the resulting stealth Schwarzschild solution arises without fixing the otherwise arbitrary Herglotz function $η(r)$. Motivated by the weak-field limit, the explicit choice $η(r)=η_{0}(r-2M)^k/r^{k+2}$, with $η_0$ a constant of dimension length, produces a scalar field configuration remaining regular at the black hole horizon. Consequently, the HVP provides a new mechanism for evading standard no-hair theorems in scalar-tensor theories.

We show how to compute the compaction function within stochastic inflation, by solving the random field dynamics on stochastic binary trees. In this framework, the compaction function is directly related to the ratio of the volumes emerging from the sibling and child branches of a given node. This construction also determines whether or not the areal radius of a perturbation increases monotonically with the radial coordinate, thereby distinguishing between type-I and type-II fluctuations. As an application, we investigate primordial black hole (PBH) formation in a single-field toy model with a constant potential slope, using stochastic-tree realizations generated with the public code \texttt{FOREST}. In the classical regime, where quantum diffusion is subdominant, the PBH mass function is narrowly distributed and type-II fluctuations are strongly suppressed relative to type I. By contrast, in the quantum and near-critical (i.e. close to eternal inflation) regimes, the PBH mass distribution spans several orders of magnitude, the overall PBH abundance is enhanced, and type-II fluctuations outnumber type I. In that case cloud-in-cloud effects are also important, highlighting the need for a better understanding of the evolution and collapse of type-II fluctuations in order to obtain robust PBH predictions when stochastic effects are significant.

We propose a framework for the direct generation of flying cat-qubit states from vacuum using time-dependent two-photon drives in nonlinear bosonic systems. We study both Kerr-based and two-photon-dissipation-based generation. By engineering Kerr nonlinearity, two-photon driving, and dissipation, we demonstrate logical control of a cat qubit during its generation and emission, while its quantum information is simultaneously shared between the nonlinear system and the propagating output field. We further analyze the effects of photon loss and pure dephasing, showing that both the state generation and logical control remain robust under realistic noise conditions. These results provide a route toward programmable bosonic quantum networks and future propagating error-correctable encodings.

We study the dynamics of a Hamiltonian system with a ghost degree of freedom, characterized by a negative kinetic-energy contribution and the possibility of runaway behavior due to an indefinite energy functional. We present numerical evidence that a nonlinear interaction term, together with a saturating exponential potential $V_c$, can suppress phase-space escape over the parameter ranges explored in this work. Using direct numerical integration of the Hamiltonian equations of motion, Poincaré surfaces of section, and trajectory projections, we find that the ghost sector and nonlinear couplings generate a mixed phase-space structure with both regular islands and chaotic regions. The maximal Lyapunov exponent supports bounded chaotic motion: nearby trajectories separate exponentially while remaining confined to a finite region of phase space for the investigated initial conditions and parameters $(\varepsilon,α)$. These results suggest that nonlinear confinement can significantly alter the stability properties of negative-energy sectors at the classical level. They provide numerical evidence for a bounded-chaos regime in which ghost-induced divergences are avoided within the explored domain.

We show that any quantum pseudorandom state that is secure against single-copy distinguishers, i.e. a $1$-PRS, can be amplified to $t$-copy security, i.e. to a $t$-PRS, without additional assumptions, for any polynomial $t$ in the security parameter. Prior work (Ananth and Goldin, arXiv 2025) was only able to show this for a restricted class of $1$-PRS constructions, namely ones whose generators only use a small number of ancilla qubits. Technically, we show that by carefully accounting for the randomness that is used in the construction, and using quantum extractors, it is possible to eliminate an ancilla register of any length and obtain a meaningful $t$-PRS outcome.

We provide a field-theoretic construction of Padmanabhan's duality-invariant Feynman propagator for a massive point particle in Euclidean space $\mathbb{R}^D$. Padmanabhan's propagator includes quantum-gravity effects due to the existence of a quantum of length $\ell$. Including $O(\ell^2)$ corrections, the corresponding field-theory model turns out to be a free, massive scalar defined in $\mathbb{R}^{D+2}$. The two additional dimensions with respect to the original $\mathbb{R}^D$ provide the necessary room, so to speak, for quantum-gravity fluctuations.

We investigate indistinguishability of identical bosons and fermions undergoing arbitrary particle-number-preserving evolutions of their visible degrees of freedom. For the projective indistinguishability measure, defined by the projection of the visible state onto the symmetric/anti-symmetric subspace, we derive an equivalent expression in terms of the dynamically invariant internal state. We further generalize the textbook symmetrization/anti-symmetrization framework for bosons and fermions to arbitrary partial distinguishability by deriving an explicit reconstruction formula for the multiparticle visible state in terms of the indistinguishability function encoding the dynamical invariants. We give complete characterization of the class-functions of indistinguishability by projective measures on generalized symmetries. Finally, we reveal a strikingly counterintuitive effect: introducing additional particle label states can increase the multiparticle indistinguishability of identical particles. The effect originates from the cancellation of collective multiparticle phases.

A quantum macroscopic ``Kruskal'' black hole solution that incorporates quantum geometry effects has been derived in Loop Quantum Gravity as the counterpart to the classical Schwarzschild solution with a distinct imprint outside the event horizon, even at scales much larger than the Planck length. This resulting black hole quantum geometry is supported by an effective energy density of quantum origin, outside the horizon, which prevents asymptotic flatness at large distances and confines massive particles to finite radii, thereby preventing their escape to infinity. In this work we adapt to these solutions the extended Anderson-Candelas-Christensen-DeWitt approach to compute the quantum vacuum polarization in order to provide an accurate measure of the quantum activity around these black holes. We carry out a numerical implementation of the formalism and present, to our knowledge for the first time, the scalar vacuum polarization $\langleφ^2\rangle$ exterior to this quantum-corrected geometry. We find that the quantum-gravity exponent $ε$ enhances the near-horizon polarization and induces, farther out, a small negative tail that we identify -- through a parameter-free DeWitt--Schwinger comparison -- with the field's response to the nonzero curvature of the background (absent for Ricci-flat Schwarzschild). The correction scales linearly with $ε$, the parameter tracking the quantum gravitational corrections, so that for astrophysically realistic (i.e., tiny) $ε$, the result is numerically indistinguishable from Schwarzschild. The calculation furnishes a consistency check on the quantum activity around these solutions, the fluctuations tracking the local curvature without anomalous growth in the exterior.

We study the ground-state entanglement entropy of free fermions on the Erdős--Rényi random graph, where each of the possible edges is present independently with some probability. Using random matrix theory and asymptotic freeness, we prove that the ground-state entanglement entropy obeys an exact volume law in the thermodynamic limit. The entanglement density, with a universal coefficient that is independent of the edge probability and the microscopic details of the graph. This coefficient is confirmed numerically to take the value approximately $0.386$ nats, strictly below the Page value. The volume law therefore reflects the absence of geometric locality in the random graph.

With the emergence of machine learning (ML) models trained on large datasets containing potentially sensitive data, a major question in AI safety is how to make learning private with respect to the training data. Similar to classical machine learning, quantum machine learning (QML) models are not devoid of privacy vulnerabilities. Differential privacy (DP) is a standard tool for training ML models on sensitive data, but its impact in QML remains poorly understood. In this work we study private training in hybrid variational QML models using a classical private DP-SGD optimizer applied to pipelines with classical inputs and outputs. We analyze the interplay between gradient clipping and calibrated noise addition in DP-SGD, and its impact on optimization and accuracy for noisy and noiseless quantum models. We first explain why quantum noise does not provide a satisfactory replacement for the calibrated noise in DP-SGD for ensuring privacy. We then show how the deterministic bounds on gradient norms for a wide class of quantum models translate into explicit control of the detrimental clipping bias introduced by DP-SGD. Finally, we formulate a numerical comparison protocol under fixed clipping threshold and privacy budget and evaluate it on synthetic and image-classification tasks for equivalent quantum and classical models. Our results suggest that quantum models can retain higher accuracy in private-training regimes where the formal privacy guarantee is ensured by a classical DP-SGD mechanism.

The nonlinear Schrödinger equation (NLSE) governs a broad class of wave phenomena, including deep-water waves, quantum turbulence, and solitons. The multiscale spatiotemporal coupling inherent in these systems imposes severe computational bottlenecks on classical high-fidelity numerical simulations. While quantum computing offers the potential for exponential speedup, its unitary dynamics pose a fundamental challenge to solve the NLSE. We propose a quantum framework based on the Lax-pair scattering for solving the 1D NLSE. Specifically, the physical field is first mapped into the spectral space via a quantum direct scattering circuit. Following a decoupled linear time evolution, the physical solution is reconstructed through an inverse scattering transform utilizing the quantum singular value transformation. Since the temporal evolution is performed analytically in the scattering domain, the framework bypasses iterative time stepping, rendering it highly advantageous for long-time simulations. To demonstrate the accuracy and noise resilience of this approach, we simulate a Gaussian wave packet under quantum noise, two-soliton collisions, breather dynamics, and modulational instability on a quantum emulator.

Quantum Embedding density functional theory (QmDFT) embedding offers a highly scalable approach to improve treatment for large highly correlated pi conjugated systems. However, estimating advanced electronic structure properties in polycyclic aromatic hydrocarbons (PAHs) needs advanced exchange correlation functionals that frequently trigger convergence instabilities during the embedding cycle. In this work, we introduce an adaptive damping and direct inversion in the iterative subspace (DIIS) accelerated protocol that stabilizes the embedding procedure, enabling robust integration of hybrid functionals like B3LYP and CAM B3LYP. Using 10 selected PAHs (linear and fused) molecules as a benchmark. We demonstrate a clear functional-dependent ground-state energetics and frontier-orbital gap estimation. While LDA based approaches yield near-quantitative agreement with FCI in DFT reference energies and is further supported by thermochemical isomerization benchmarks, while B3LYP provide significantly improved agreement with experimental E0-0 transition values. This mapping allows us to bypass explicit excited-state calculations for E0 0 values, thereby significantly reducing computational overhead. Among the hybrid functionals screened, CAM B3LYP offers a balanced overall performance. Our results establish a stable QmDFT framework and provide useful guidance for functional selection for quantum embedding studies of PAHs and related low-dimensional pi-conjugated materials.

Computing non-equilibrium steady states of open quantum systems is a challenging task on conventional computers, motivating quantum algorithms for direct steady-state estimation. A natural route is to regard the steady state as the zero mode of the Liouvillian and to isolate this sector spectrally. We formulate this task as a known-zero-sector projection problem and implement the corresponding filter using the Rodeo algorithm, which performs stochastic spectral filtering through repeated controlled evolutions and measurement-conditioned filtering steps. In the steady-state setting, the filter can be centered directly at the known zero eigenvalue, avoiding the spectral search required in generic eigenstate preparation. Compared with a phase-estimation-based implementation of the same projection, the Rodeo approach enables restart on failure and reduces the target-error dependence of the filtering cost and controlled-evolution depth from power-law to logarithmic. This advantage becomes more pronounced as the spectral separation of the Hermitian Liouvillian embedding increases, allowing Rodeo filtering to outperform phase-estimation filtering already at modest controlled-evolution depths. Our results identify Rodeo filtering as a resource-efficient primitive for estimating steady-state observables in open quantum systems.

Embedded random matrix ensembles operating in nuclear shell model spaces, with nucleons occupying a finite set of single particle orbits and interacting via a two-body interaction, form the basis for statistical shell model. With sufficiently strong interaction, the level densities in shell model spaces take close to a Gaussian form and transition strength distributions close to a bivariate Gaussian form. In practice, partitioning via spherical configurations ($\tilde{m}$) and angular momentum $J$ (also isospin where appropriate) are essential. The resulting statistical spectroscopy or statistical shell model was applied successfully in the past in some studies of nuclear level densities, orbit occupancies, $β$-decay matrix elements and so on. Going beyond these, recently it is recognized that embedded ensembles, in a better approximation, generate in-fact $q$-normal form ($q=1$ gives Gaussian and $q=0$ Wigner's semi-circle) for density of eigenvalues, bivariate $q$-normal form for transition strengths and conditional $q$-normal form for strength functions. These then allow us to develop statistical shell model with $q$-normal forms. These new developments in embedded ensembles and statistical shell model are briefly reviewed in this paper. Also described, using some examples, is the role of the $q$ parameter in generating statistical properties of general quantum many-particle systems.

In this draft, we investigate the dynamical instability of a restricted class of non-static, axially symmetric, self-gravitating fluid configurations within a Hu-Sawicki inspired f(R) gravity model. The matter source is described by an anisotropic energy-momentum tensor containing three principal stresses and an off-diagonal stress component. For the adopted vorticity free geometry, conservation equations are formulated, and a linear perturbation scheme is applied to separate the equilibrium and time-dependent sectors. This procedure yields a collapse equation that governs the evolution of the perturbed compact configuration. The associated instability conditions are then derived in terms of the adiabatic index Γ under the Newtonian and post-Newtonian approximations, whereas the resulting bounds show that the onset of instability depends not only on the stiffness of the fluid, but also on the background energy density, directional pressure anisotropies, metric perturbations, and higher-curvature contributions generated by the Hu-Sawicki model. The general relativistic limit is recovered by suppressing the modified gravity parameters, while the isotropic limit reproduces the classical Chandrasekhar threshold. These results demonstrate that curvature corrections and anisotropic stresses can appreciably modify the conventional instability conditions of axially symmetric compact systems.

The coupled three-mode squeezed vacuum is a representative multimode squeezed Gaussian state featuring unique steerability. This work investigates Gaussian quantum steering distributions of the coupled three-mode squeezed vacuum under an expanding universe. Due to causal separation between the interior and exterior spacetime regions, quantum information behind the event horizon is inaccessible to Alice, Bob and Charlie. We separately analyze steering behaviors for physically accessible and inaccessible modes. Our analysis shows that greater total mean photon number and momentum, combined with reduced expansion volume and expansion rate, enhance quantum steering strength. Notably, Gaussian quantum steering for physically inaccessible modes undergoes the "sudden death" phenomenon when a critical threshold parameter $φ$ is exceeded. These results deliver novel insights into quantum correlations in curved spacetime.

Assessing possible life on an exoplanet requires spatial, spectral, temporal, and environmental context rather than a threshold detection of one molecule or surface feature. We develop a covariance-aware Solar Gravitational Lens (SGL) framework in which the data product is a time-tagged Stokes spectral cube reconstructed from wavelength-dependent Einstein-ring measurements. The demonstrated calculation is a 0.45-2.40 um Stokes-I reflected-light simulation of an Earth-radius planet at 30 pc, observed from 650 AU with a (128 x 128) raster, 128 simultaneous spectral channels, and $R\simeq70$. A separate 0.40-20 um architecture-level calculation tracks reflected and thermal planet photons, SGL gain, solar-corona noise, instrumental backgrounds, throughput, dwell time, and reconstruction covariance. In the controlled population audit, structural forward-model mismatch preserves the block ordering gas > surface > cloud/path > mineral > calibration/SGL while reducing the combined conditional information gain to 0.83 of the matched-model value. A reconstruction-covariance bracket reduces an (8 x 8) regional coadd gain from 7.77 to 3.00, implying a 6.7-fold dwell penalty. The feasibility results are design scalings, not a mission verdict: imaging and low-resolution mapping are earlier objectives, whereas full regional spectroscopy requires simultaneous acquisition, sub-ppm effective coronal calibration, measured reconstruction covariance, and branch-specific radiometric validation. We show that the SGL offers a uniquely powerful path to surface-resolved mapping, regional spectroscopy, thermal-climate diagnostics, and co-location tests, providing spatial, spectral, temporal, and environmental context that could strengthen assessments of habitability and possible biological activity beyond disk-integrated precursor observations.

We consider unitary quantum scrambling for an entire class of observable algebras $\mathcal A$ whose commutant ${\mathcal A}'=\mathbb{C}\mathbb{Z}_2$ is generated by a Hermitian involution $S$, i.e., a parity operator. We adopt two different, although related, scrambling measures: (a) the algebraic out-of-time-order correlator (A-OTOC) introduced in [Andreadakis, Anand, and Zanardi, Phys. Rev. A 107, 042217 (2023)], and (b) a Slater-determinant-type quantity, dubbed the Plücker fidelity, obtained by embedding operator algebras into a higher-dimensional projective operator space. Both measures admit a simple geometric interpretation in terms of distances between algebras and their dynamical images. Moreover, they can be expressed in terms of the norm of the $\mathbb{Z}_2$-symmetry-breaking component of the unitary, which is in turn controlled by the time autocorrelation function of the $\mathbb{Z}_2$ generator $S$. We derive exact expressions for the A-OTOC and the Plücker fidelity in the general case, as well as their values for a random unitary. For Hamiltonian-generated channels, we compute their long-time averages, as well as the typical values of these long-time averages for random generators $S$. Finally, we numerically explore the results of our formalism for interacting spin chains in different phases and for different physical parity operators.

We investigate the interplay between thermodynamic phase transitions and the dynamical behavior of quasi-normal modes (QNMs) in weakly nonlinear charged anti-de Sitter black holes. In a non-extended phase space, we identify a reentrant large-small-large(intermediate) phase transition for these black holes. Employing the shooting method to solve the Klein-Gordon equation for a massless scalar field, we compute QNM frequencies across these phases, revealing distinct behaviors in different phases. Specifically, to our knowledge, this is the first report of utilizing QNMs as dynamical probes of thermodynamic phases during reentrant phase transitions including zeroth and first-order transitions. These findings establish QNMs as a powerful dynamical probe for thermodynamic phase transitions, providing new insights into the thermodynamic and dynamical properties of nonlinear charged black holes as well.

Although entangled state vectors cannot be fully described in terms of variables localized in space and time, any given entanglement experiment can be built from basic quantum circuit components with well-defined locations. We analyze such quantum circuits and present evidence that the local weak values comprise a covariant tensor associated with each individual qubit. Even if the state is massively entangled, these tensors do not evolve or collapse when other qubits are measured or pass through distant circuit elements. They can therefore be viewed from different reference frames without contradiction. Furthermore, their evolution through any circuit always obeys covariant dynamical rules. Weak values are subject to both past and future constraints, so the covariant quantities can only be determined by considering the entire circuit "all-at-once", as in action principles, incorporating the future measurement basis to avoid the standard no-go theorems. Because these results hold for a set of universal quantum gates, this work lends support to the claim that any quantum circuit can be assigned a realistic, lower-level description compatible with our understanding of classical spacetime.

ESA's LISA mission will open a new window onto the gravitational-wave sky by detecting signals from a wide variety of sources in the millihertz frequency band. Among these, galactic binaries are expected to be the most numerous sources observable by LISA. Their analysis and parameter estimation represent a significant challenge, as the signals are expected to strongly overlap in both the time and frequency domains. Conventional Bayesian inference approaches, such as Markov Chain Monte Carlo sampling, are difficult to scale to this setting due to the high dimensionality of the problem and the complicated likelihood landscape which can hinder convergence. In this work, we explore simulation-based inference as a means to perform efficient parameter estimation for single galactic binaries, with a potential extension to the analysis of multiple overlapping sources. Our approach relies on a conditional normalizing flow acting as a neural posterior estimator. The model is trained using samples generated according to a dedicated simulation framework that does not require any likelihood computation. Once trained, the neural posterior estimator enables the generation of thousands of posterior samples per second, again without explicit likelihood evaluation. We first present results for a single source in a narrow frequency band, and then extend the analysis to wider frequency ranges. As a proof of concept, we further investigate the more challenging case of two overlapping sources. These results demonstrate the potential of likelihood-free inference as a scalable alternative to conventional Markov chain Monte Carlo sampling for the analysis of LISA galactic binaries.

We present a theoretical study of the evolution of polarization entanglement and quantum coherence in 511 keV photon pairs produced by para-positronium decay during successive Compton scattering events. We start with a maximally entangled Bell state and employ the generalized Stokes-Mueller formalism to derive the two-photon density matrix following single-, double-, and triple-Compton scattering, explicitly considering both polar and azimuthal scattering geometries. Using this framework, we quantify the degradation of quantum correlations through concurrence (as a measure of entanglement) and the $l_1$-norm (as a measure of coherence). Our results demonstrate that entanglement is highly sensitive to the scattering geometry and disappears near right-angle scattering, while quantum coherence remains finite even in regimes where entanglement vanishes completely. These findings provide a unified description of polarization-dependent decoherence in annihilation photon pairs and clarify the distinct roles of entanglement and coherence in realistic two-photon interactions. These results are relevant for quantum-enhanced positron emission tomography and highlight the persistence of quantum resources in scattering-dominated media.

Applications of the proposed primitive: Quantum Primitive for Output-Hiding Function Sharing are discussed for environments in which the parties may have strategic considerations over the generated function value, and therefore may not be mutually trusted. We show that in environments in which the measurement may not be reliably controlled by either party, the primitive permits the parties to generate private, unbiased coins.

Numerical simulation of quantum computing hardware and open quantum systems governed by the Lindblad equation is challenging due to the high dimensionality of the density matrix and the need to preserve fundamental physical properties. In our previous work, we developed an arbitrary-order, low-rank, completely positive and trace preserving (CPTP) method for the Lindblad equation with time-dependent Hamiltonians by nested Picard iteration (NPI). In this work, we develop Gregory NPI schemes, which are CPTP schemes constructed by Gregory-type quadrature on equispaced nodes. The methods, which are of order up to nine, substantially reduce the computational cost compared to our previously proposed NPI schemes with Gaussian quadrature rules, while retaining high-order accuracy and structure preservation. We analyze the stability of the resulting scheme for a physics-based test equation. Numerical experiments verify the convergence of the method and demonstrate the effectiveness of the low-rank approximation. We study the performance of a previously constructed CNOT gate for both closed and open quantum systems.

Applications of the proposed primitive: Quantum Primitive for Output-Hiding Function Sharing, are discussed for secure quantum communications and computation protocols. In particular, QKD applications provide notable enhanced security and efficiency properties relative to status-quo protocols. Additionally, we provide examples when parties may wish to encode joint functions, or decisions, which remain information-theoretically hidden from external parties and those internal to the quantum system, without additional private keys, hidden randomness, or classical communication. In particular these applications may be useful in domains such as; financial transactions, joint signaling or coordination decisions, and navigation systems, among others.

We study entanglement transfer in a minimal model of two qubits that are coupled with one another through an antiferromagnetic Heisenberg exchange ($\tilde{J}$), and where one of them is additionally coupled to a fermionic environment through another antiferromagnetic Heisenberg exchange ($J_{K}$). By tuning the coupling ratio $J_{K}/\tilde{J}$, the system undergoes a quantum phase transition at $T=0$, accompanied by a redistribution of entanglement from the $d'-d$ qubit-subsystem to the environment. Remarkably, the resulting physics exhibits properties that bear analogy with a quantum-information theoretic perspective of the physics of black hole thermodynamics. Carefully selected bipartite and tripartite mutual information measures displays behaviour analogous to the dynamical evolution of black hole entropy, Hawking entropy, and the Page curve expected during the process of evaporation. An effective temperature scale is obtained from the variation of the ground state energy with respect to changes in the bipartite mutual information between the subsystem and the bath. A steady growth of this temperature with the coupling ratio resembles that of the Hawking temperature with the inverse mass of the black hole. Concomitantly, the emergence of non-Fermi liquid behaviour observed near the quantum critical point and in the strong-coupling phase resembles strange-metal-like physics expected near the event horizon from a holographic duality perspective. Our results establish the minimal model as a platform for studying entanglement transfer and information scrambling within a fully unitary quantum framework, and offer new insights into a resolution of the black hole information paradox.

The decay of a metastable (false) vacuum plays a crucial role in constraining Standard Model and beyond the Standard Model physics. In particular, it has been shown that gravity can have a significant impact on the calculation of the decay rate. In this context, it is natural to ask whether different but classically equivalent formulations of gravity lead to the same physical predictions. The aim of this paper is to analyze vacuum decay in teleparallel and symmetric teleparallel equivalent formulations of General Relativity (GR), namely TEGR and STEGR. Although these theories describe the same classical dynamics, it is of paramount importance to understand whether this equivalence persists also at the quantum level. In this respect, the analysis of vacuum stability may provide a particularly sensitive testing ground. The central question is whether the decay rate of a false vacuum computed within TEGR or STEGR coincides with the corresponding result obtained in GR. Our analysis shows that the tunneling exponent remains unchanged, offering a non-trivial example in which the equivalence between different formulations of gravity extends beyond classical dynamics.

Efficient manipulation of Pauli-algebraic objects is a key bottleneck in the classical emulation and benchmarking of quantum algorithms for chemistry and many-body physics. This bottleneck appears in Hamiltonian construction, variational ansatz preparation, expectation-value and gradient evaluation, and real-time propagation, all of which require repeated Pauli-algebra operations. Here, we present a high-performance Pauli-algebra framework tailored to quantum many-body and quantum-chemical simulations. The framework combines compact binary symplectic encoding, canonical coefficient reduction, and grouped sparse operator representations that exploit shared bit-flip patterns among Pauli strings. The resulting Julia/C\texttt{++} implementation accelerates Pauli multiplication, Hamiltonian construction, and operator--state multiplication in sparse and symmetry-adapted many-electron spaces. Benchmarks demonstrate efficient Hamiltonian construction, large-active-space VQE and ADAPT-VQE calculations, and real-time variational dynamics on modern multicore CPU and GPU architectures. These results show that structure-aware Pauli-algebra engines provide a scalable classical backend for developing and benchmarking quantum algorithms in quantum chemistry and many-body simulation.

Next-generation gravitational-wave detectors will require significant improvements in current theoretical waveform models, particularly in the case of asymmetric-mass binaries. Here we provide one such improvement by calculating fully relativistic finite-size effects for small mass ratios -- primarily, fluxes of energy -- including quadratic-in-secondary-spin terms, spin-induced quadrupole terms, and tidally induced quadrupole terms, for quasicircular inspirals of a small companion into a Kerr black hole. We formulate these calculations within a multiscale waveform-generation framework in self-force theory, which could be used, with an energy-balance law we derive, to develop self-contained waveform models for asymmetric binaries involving stars orbiting black holes. Our results could additionally be used to improve other families of waveform models across all mass ratios. We present results both as complete numerical data sets on a Chebyshev grid and as analytical post-Newtonian expansions (to sixth PN order relative to the leading term in each contribution to the flux).

We study variational quantum algorithms for the Maximum Independent Set (MIS) problem on benchmark graphs of 64, 99, and 180 vertices. The Variational Quantum Eigensolver (VQE) and Quantum Approximate Optimization Algorithm (QAOA) are compared across SPSA and COBYLA optimizers at multiple circuit depths. A preprocessing pipeline comprising spectral graph reordering (via the Fiedler vector) and distance-based sparsification reduces circuit depth while preserving energy fidelity. Classical post-processing via history-guided bitstring correction and stepwise maximality extension recovers the exact MIS across all instances. With CVaR optimization, VQE with SPSArecovers up to 6 distinct MIS per run for the 64-node instance and up to 10 distinct MIS per run for the 99-node instance, sampling broadly from the optimal solution population. Repeated runs with different SPSA trajectories collectively enumerate a larger fraction of all MIS for each instance. For the 180-node instance, where standard approaches stall at size 14 (MIS is 15), we introduce ancilla-assisted superposition initialization: ancilla qubits prepare a uniform superposition over classically-found near-optimal solutions, and an excitation-preserving ansatz evolves this state while conserving Hamming weight. This novel construction enables quantum-parallel variational search over multiple seeds simultaneously, discovering the exact MIS where single-seed methods fail. The 180-qubit simulation represents, to our knowledge, the largest scale at which gate-based variational algorithms have solved MIS to optimality. Hardware validation on IBM Quantum hardware ibm_marrakesh confirms that converged simulator parameters transfer effectively to noisy quantum execution.

It is commonly believed that a quantum field theory of General Relativity requires a non-perturbative formulation. In addition, the background independence of classical General Relativity supplies a physical selection criterion for suitable Hilbert space representations of the corresponding quantum field theory. In this contribution we show that there exist background independent representations of Fock type within the manifestly non-perturbative, canonical approach to quantum gravity. Mandatory for their existence is the presence of suitable matter fields next to the geometry field. In particular, the excitations of the corresponding Fock vacuum necessarily entangles matter and geometry. In this article we use the constraint quantisation method. We compare the resulting Fock incarnation of background independent, non-perturbative canonical quantum gravity with the well known Loop quantum gravity incarnation. One of the most important differences is that the Fock quantum gravity (FQG) Hilbert space, in contrast to the Loop quantum gravity (LQG) Hilbert space, is separable. This has many advantages when attempting to implement the Hamiltonian constraint, also known as Wheeler-DeWitt constraint, as a densely defined quadratic form.

We have performed first-principles hybrid density functional theory calculations to study the formation and properties of two-dimensional electron or hole gases (2DEG or 2DHG) at s-Si/SiGe interfaces. For small Ge concentrations $x < 0.25$, we find a novel type of band alignment with no offset in the conduction bands, implying that a 2DEG cannot be formed, though a 2DHG can. In contrast, for $x > 0.25$ the band alignment suggests that either a 2DEG or 2DHG can be formed. The electronic band structure features two nearly degenerate 2DEG states at the bottom of the conduction bands, and two 2DHG states at the top of the valence band. These states can be accessed by appropriate doping and gating. Charge density plots of these states show that they feature carriers confined to the near vicinity (2--3 atomic layers) of the interface. Calculated effective masses are anisotropic, being markedly so for the 2DHG states, and in excellent agreement with experiment. This property can be exploited to create a 1D carrier gas. Our results are especially important for s-Si/SiGe-based semiconducting spin qubits for quantum computing applications.

Bipartite Gaussian boson sampling (BipartiteGBS) produces output probabilities governed by squared permanents of submatrices of arbitrary complex matrices, matching the nonsymmetric structure of directed graphs. Most GBS-based graph algorithms, however, rely on symmetric hafnian structure and are formulated for undirected problems. Here we propose a BipartiteGBS-based framework for directed-graph heuristic optimization. We introduce Max-Perm as a canonical optimization task for BipartiteGBS and derive a closed-form sampling enhancement factor relative to uniform classical sampling in this idealized setting. We then use permanent-biased BipartiteGBS samples to guide a genetic algorithm for the celebrated directed Hamiltonian cycle problem. Numerical experiments on Erdős--Rényi random directed graphs show that the resulting BipartiteGBS-enhanced algorithms improve success rates over a standard genetic algorithm and yield longer valid paths when no Hamiltonian cycle is found, while ablation tests indicate that BipartiteGBS-guided initialization is the dominant contributor. These results show how permanent-based photonic sampling can provide useful algorithmic guidance for asymmetric combinatorial search.

We investigate the generation of optical torque in coherently prepared multi-level atomic media driven by a vector vortex beam composed of two counter-rotating components carrying opposite orbital angular momenta, $+l\hbar$ and $-l\hbar$. We consider a three-level $Λ$ configuration and a four-level tripod configuration. Using a perturbative steady-state solution of the optical Bloch equations, we obtain analytical expressions for both linear and nonlinear contributions to the optical torque. The results show that the torque is strongly controlled by atomic coherence, including the initial population imbalance and the relative phase between the vortex components. Nonvanishing torque can arise even when the two components have equal amplitudes, due to coherence-induced asymmetry in the atomic response. In the tripod configuration, the presence of a strong control field leads to electromagnetically induced transparency, which suppresses the torque near resonance and shifts the dominant response to finite detunings. These results establish a route for controlling light-induced rotational dynamics in atomic media using vector vortex fields, with potential applications in coherent optical manipulation and angular-momentum-based control in quantum systems.

We propose a dynamical approach to identifying ground-state quantum phases through short-time dissipative cooling. Rather than determining the phase by preparing highly accurate approximations to ground states, we prepare a representative state of a candidate phase and monitor the early-time response of phase-sensitive observables under cooling dynamics tailored to the target Hamiltonian. For a class of phase-decision problems in which the relevant observables can be inferred from the low-energy manifold, and with jump operators implementable using only short-time Hamiltonian simulation, the dissipative evolution rapidly suppresses high-energy components and drives the system into a low-energy manifold whose observables already reveal the underlying ground-state phase, well before mixing to the steady state. We demonstrate this strategy for the frustrated $J_1$--$J_2$ Heisenberg chain, the Kitaev honeycomb model, and the XXZ chain, including Berezinskii--Kosterlitz--Thouless and topological phase transitions. In particular, coarse filter resolutions and short evolution times suffice to recover phase-sensitive quantities such as the Luttinger parameter and topological diagnostics. We further provide theoretical justification that cooling dynamics with such jump operators can rigorously prepare low-energy manifolds for free-fermionic and free-bosonic systems, and investigate this mechanism for interacting fermionic systems. Our results suggest that phase decision is a plausible target for future utility-scale studies on early fault-tolerant quantum devices.

Estimating potential future exposure (PFE) for path-dependent derivatives, such as FX Target Redemption Forwards (TARFs), represents a formidable computational challenge due to the demand of nested Monte Carlo simulations. We present a hybrid quantum-classical framework that leverages Iterative Quantum Amplitude Estimation (IQAE) to address this via a reduced-order counterparty credit risk model. Our methodology maps the non-linear TARF payoff -- including cumulative gains and knock-out features -- into a quantum circuit via a two-step formulation, whereby a first-step percentile is computed classically and then used to condition quantum evaluation of subsequent exposure. We employ discretisation of the FX process and a linearised additive approximation of dynamics to enable implementation on current quantum platforms. Developed via the Classiq platform and validated on NVIDIA CUDA-Q and Amazon Braket SV1, our approach achieves relative errors of 1%-8% against classical benchmarks at the 97.5% and 99% confidence levels. While discretisation constraints and approximate monotonicity assumption may introduce bias and limit recovery of the full exposure distribution, our framework offers a tractable testbed for quantum acceleration. Scaling analysis suggests that $\sim$300 logical qubits could enable full 52-week exposure estimation, reducing sample complexity for tail-risk estimation via amplitude estimation at the cost of increased circuit depth.

The entanglement of Gaussian continuous-variable (CV) states is fully determined by the state's second moments. In contrast, some entangled non-Gaussian states evade every second-moment criterion, and non-Gaussian entanglement detection remains an experimental challenge. The $p_3$-PPT criterion detects entanglement using moments of the partial transpose of the density matrix. This criterion was recently extended to CV systems using photon-number-resolving detectors and multi-copy interferometry; here we introduce a single-copy homodyne protocol that detects bipartite CV entanglement via the same criterion. Unbiased U-statistic estimators for the partial-transpose moments $p_2$ and $p_3$ are constructed directly from randomized homodyne data and used to evaluate the $p_3$-PPT entanglement witnesses: a linear one for detection, and a quadratic one whose violation yields a dimension-free lower bound on the entanglement negativity. The protocol estimates $p_2$ and $p_3$ up to additive error $\varepsilon$ at Fock cutoff $N$ from $O((N+1)^{14/3}/\varepsilon^2)$ measurements at fixed confidence. We demonstrate the protocol on six families of Gaussian and non-Gaussian states, reaching $95\%$ empirical one-sided detection probability from $\sim 10^3$ to $10^4$ homodyne measurements for states with $\bar{n} \approx 2$, placing non-Gaussian entanglement detection within reach of current homodyne experiments.

Planar Germanium has emerged as a promising platform to build spin-based large scale quantum computers. By exploiting the anisotropic hyperfine interaction of holes in Ge, qubits with long T2* have been recently realized. While the performance of single qubits is still more or less limited by 73Ge nuclear spin fluctuations, the site-to-site variation of qubit sweet spot becomes obstacles to maintaining high fidelity of each qubit across the whole wafer. To achieve high performance Ge-based quantum circuit, it is therefore essential to eliminate the origin source of hyperfine noise. In its Silicon counterparts, reduction of 29Si abundance enables exceptional high-fidelity operation. In contrast, hole qubits based on isotopically purified Ge have not been demonstrated. Here, we report the synthesis of high quality 2-dimensional hole gas (2DHG) with enriched 70GeH4 precursor. Due to the suppression of nonzero spin nucleus, the qubits' T2* on the sweet spot is moderately extended beyond 20 us, surpassing the previous best reported Ge hole qubits. More importantly, the qubits' T2* off the sweet spot is enhanced to above 3 us, enabling single qubit gate fidelity exceeding 99.9% in both operating regimes. Hahn-echo spectroscopy further resolves a finite-frequency nuclear-noise channel that is distinct from the conventional Larmor-linked hyperfine response. We associate this channel with quadrupole-modified dynamics of residual 73Ge nuclei sampling local electric-field gradients near the Ge/SiGe interface. Its field scaling and angle-dependent visibility are consistent with a qubit-visible quadrupolar nuclear-noise component transduced through the anisotropic hyperfine interaction of Ge holes. These results establish isotopically purified planar Ge as a high-coherence scalable platform for hole spin qubits and provide a spectroscopic probe of interfacial quadrupolar nuclear dynamics.

The gravitational field of a distant, isolated system is manifested by the conformally invariant Weyl tensor. Thus the conformal structure far from the system encodes the system's gravitational mass. It also encodes the causal structure, thereby linking it to the mass. For asymptotically anti-de Sitter (AdS) spacetimes, this link led to a novel positive mass theorem of Page, Surya, and the second author \cite{PSW} which did not rely on any traditional energy condition. Here we ask whether that theorem has a rigidity case. Specifically, we consider all null geodesics in an asymptotically AdS spacetime that depart from the Penrose conformal infinity, travel through spacetime, and return to conformal infinity. If all such geodesics from a given point refocus at an antipodal point at infinity, is the spacetime conformal to anti-de Sitter space? It is easy to answer the question if the asymptotically AdS spacetime either (i) obeys the null energy condition or (ii) is static, and we give simple proofs in those cases. We also answer the question in the case of globally stationary, asymptotically AdS spacetimes, by applying the theory of magnetic geodesics on the Riemannian manifold-with-boundary obtained by quotienting by the stationary Killing vector field. The question has an analogue for asymptotically flat spacetimes, which we also discuss.

We consider the encoding of graph problems as Quadratic Unconstrained Binary Optimization (QUBO) problems, which are solvable by either quantum or classical annealers. Yet, the class of problems encoded as QUBO problems has not previously included nowhere-zero flows. Nowhere-zero flows are related to Tutte's $5$-flow conjecture and appear in many contexts in graph theory. We provide an encoding of nowhere-zero flows as a QUBO Hamiltonian and prove the correctness of the construction. Our construction yields a Hamiltonian $H_{\mathrm{mod},k}$ whose ground state has zero energy if and only if the graph $G$ has a nowhere-zero $\mathbb Z_k$-flow. By Tutte's equivalence theorem, zero ground energy is equivalent to $\varphi(G)\le k$, and the zero-energy degeneracy is given by the flow polynomial $F(G;k)$. In particular, when the ground-state energy is zero, this is also the ground-state degeneracy. The construction uses one-hot variables to represent the edge flow residues modulo $k$ and auxiliary variables to represent the per-vertex modular quotient. We prove that the correctness of the construction is independent of the choice of orientation, root vertex, and positive penalty weights. We verify the construction on $59$ examples of graphs and values of $k$ that include both yes-instances and no-instances. We exhaustively sweep orientations and root choices on selected robustness instances and test a finite suite of positive penalty weights. The resulting Hamiltonian is implemented using the dimod.BinaryQuadraticModel class, which is compatible with the D-Wave Ocean SDK. Quantum-hardware runs and claims about potential speedup using these devices are left to follow-up work.

Fermionic topological orders can host 't Hooft anomalies with no bosonic counterpart. We identify a new sixteen-fold family of (2+1)D fermionic topological orders, forming a fermionic analogue of Kitaev's sixteen-fold way. This family is distinguished by the mod 16 't Hooft anomaly of a $\mathbb{Z}_2$ one-form symmetry, generated in each theory by a single nontrivial $\mathbb{Z}_2$ anyon. This intrinsically fermionic anomaly permits anyon spins that are forbidden in bosonic phases; the simplest new example is an Abelian fermionic topological order containing a single $\mathbb{Z}_2$ Abelian anyon of spin 1/8. Each theory can be realized as the gapped boundary of a (3+1)D fermionic symmetry-protected topological (SPT) phase protected by the $\mathbb{Z}_2$ one-form symmetry, which acquires a $\mathbb{Z}_{16}$ classification once the spacetime spin structure is twisted by the one-form symmetry. We realize these phases microscopically via lattice models built from Walker-Wang models coupled to local fermions.

Temporal non-analyticities in the rate function of the Loschmidt echo manifests a class of dynamical quantum phase transitions (DQPTs) that has emerged as a powerful framework for understanding far-from-equilibrium many-body dynamics. While such DQPT has been extensively studied theoretically in spin-boson systems such as the Dicke model, their experimental observation remains elusive. In particular, the dynamics of DQPT in asymmetric spin subspaces and under the influence of spin dissipation are largely unexplored. Here, we report an experimental study of temporal DQPT in a generalized Dicke model using a trapped-ion quantum simulator. By coupling a linear chain of $\rm{^{40}Ca^{+}}$ ions to a collective center-of-mass motional mode, we probe the quench dynamics starting from both symmetric and asymmetric initial states. We extract the rate function and identify temporal turn-around points that are in quantitative agreement with theoretical predictions. Additionally, we investigate the impact of spin dissipation on these dynamics. Our results establish an experimental platform for probing complex many-body out-of-equilibrium phenomena and advance the development of hybrid oscillator-spin quantum simulators.

Understanding how macroscopic thermodynamic behavior emerges from microscopic quantum dynamics remains an open problem. While equilibration of quantum observables is well established, thermodynamics also relies on variables not directly associated with linear operators, but which are defined instead as functions of expectation values. Whether and how such derived quantities inherit equilibration properties is an open question. Here, we establish that any continuously differentiable function of equilibrating expectation values also equilibrates. We apply this result to a bipartite isolated system, showing that the entropy and conjugate variables of each subsystem -- defined through Jaynes' maximum entropy principle -- equilibrate. Moreover, with the assumption that their equilibrium properties depend solely on local conserved quantities, we show the dynamical maximization of the total entropy, enforcing equality of conjugate variables across subsystems. These results provide a direct dynamical justification for entropy maximization and the emergence of thermodynamic equilibrium conditions, showing that fundamental principles of thermodynamics follow from the unitary evolution of quantum systems.

We formulate a continuity equation for the Schrödinger equation in the complex space. We define a complex momentum by normalizing the complex current by the particle density. This momentum is a quantum analog of the classical, kinematic momentum analytically continued into the complex plane. The kinematic momentum and the gradient of the wavefunction's phase each represent a fluid-like flow in the complex plane; the phase-gradient flow is incompressible. The zeros of the wavefunction give rise to simple poles in the momentum. The poles manifest as irrotational vortexes in the phase-gradient flow, while critical points of the wavefunction present as rigid body-like rotational flows of the kinematic momentum. A discrete nature of elementary excitations comes about inherently because the quantity of the poles is automatically integer. An exact quantization condition is subsequently formulated, which reduces to the Bohr-Sommerfeld condition in the semiclassical limit. We establish a priori that the Bohr-Sommerfeld condition must be exact for the Harmonic Oscillator. We show that the kinetic energy is a sum of contributions of the average value and fluctuations, respectively, of the kinematic momentum. The zero-point vibrations within bound states are solely due to the fluctuations of the momentum and manifest as rigid-body flows at infinity. The momentum poles -- and hence the wavefunction's zeros -- can be viewed as emergent, consistent with the remarkable property of quantum entanglement exhibited by standing wave solutions of the Schrödinger equation.

We introduce the dimension-restricted prepare and broadcast (PAB) scenario, which generalizes standard prepare-and-measure frameworks. Here, the system prepared by a sender undergoes a broadcasting transformation before being locally measured by multiple receivers. We develop a hierarchy of classical, quantum, and nonsignalling models describing this scenario, characterize their corresponding correlation sets, and derive new families of Bell-like inequalities together with linear and semidefinite programming methods for their certification. First, assuming shared randomness, we prove that the hierarchy collapses into a single set whenever we consider only one measurement per party. Then, considering multiple possible measurements, we show that PAB scenarios allow the activation of nonclassicality, revealing genuinely nonclassical features in resources that admit classical descriptions in standard prepare-and-measure or Bell settings.

Local observables can lose sensitivity to an initial state during strongly interacting many-body evolution even though the global dynamics remain unitary. We show that this sensitivity can be enhanced through a time-dependent Hamiltonian readout. Two orthogonal product states are first evolved under a slanted-field Ising Hamiltonian, where their distinction becomes strongly suppressed as observed through several local observables, including subsystem magnetizations and correlation functions, and are then quenched to the transverse-field Ising model at a tunable time. Exact simulations of chains up to $N=12$ show that the optimized time-averaged separation after the switch exceeds the residual slanted-field baseline for every observable and system size tested. In the strongest channels, the standardized readout separation remains robust over the accessible size range, with no clear systematic suppression at larger $N$. The enhancement recurs in widely separated late-time windows and persists qualitatively for open boundaries. These results establish Hamiltonian switching as an observable-selective mechanism for enhancing initial-state sensitivity without time reversal or implying recovery of the full reduced state.

Combinatorial optimization is a leading target for quantum algorithms, but finite-depth QAOA can suffer from strong diabatic errors when the interpolation Hamiltonian has small, or exponentially small, spectral gaps. We propose a Pauli-sparse counterdiabatic extension of linear-ramp QAOA based on the regularised adiabatic gauge potential \[ \bigl(\mathcal L_H^2+ηI\bigr)A_λ^{(η)} = -\mathrm{i}\mathcal L_H(\partial_λH), \qquad \mathcal L_H(X)=[H,X]. \] Instead of computing a dense AGP, we solve this equation approximately by an inexact conjugate-gradient method in Pauli coordinates, truncating the Pauli expansion during the iteration to obtain a gate-budget-aware set of implementable rotations. The selected support is then improved by a Galerkin refit and certified by an a posteriori residual bound. The regularization parameter \(η\) acts as an energy-resolution scale: it suppresses transitions below \(\sqrtη\) while retaining larger-gap transitions. Thus, the method can avoid resolving exponentially small splittings inside a low-energy solution manifold while reducing leakage away from it. Numerical experiments on Ferromagnetic Chain (FC) and perturbed FC--MaxCut/MarketSplit instances show that the resulting LR-CD-QAOA ansatz improves approximation ratios over the uncorrected linear ramp, especially in regimes where LR-QAOA remains far from the optimum. Overall, the proposed regularized LR-CD-QAOA framework substantially broadens the practical applicability of QAOA to QUBO optimization by improving its robustness across heterogeneous problem landscapes, including instances with near-degenerate low-energy structures and small spectral gaps.

Graph states of mutually entangled photons are key resources for quantum computation and communication and can be generated by leveraging the entanglement between a single resident spin and emitted photons from a charged semiconductor quantum dot (QD). This approach is intrinsically limited by the decoherence of the spin. We study how to mitigate this decoherence with dynamical decoupling of an electron spin in the weak transverse magnetic field regime using spin echo and Carr-Purcell-Meiboom-Gill (CPMG) techniques. Application of these techniques allows us to extend the coherence time of a spin by more than two orders of magnitude, extracting a $T_2^{CPMG}$ of $298\pm53$ ns. We further demonstrate that this technique is compatible with the generation of a spin-photon-photon entangled state at a high rate enabled by a micropillar cavity, with a 20% improvement in simulated state fidelity when using dynamical decoupling. These results pave the way for generating larger and more complex entangled states with QDs.

Negatively charged germanium vacancy centers (GeV) in diamond are a promising platform for quantum computing and quantum communication. However, these applications require the precise incorporation of GeV centers with good optical properties inside of nanophotonic structures. In this work, we demonstrate the highly efficient local creation of GeV centers in diamond via focused-ion-beam implantation, followed by high-temperature annealing. We report the successful creation of GeV centers over the depth range of 5.5 - 30 nm. Implantation at low fluence enables the creation of single GeV centers. The formation yield strongly depends on implantation energy and fluence, reaching up to 33% at energies of 35 and 70 keV. This method, therefore, enables the efficient creation of GeV centers within a small, well-defined local sample volume and offers a potential means of incorporating them into photonic structures.

Magic-state distillation is one of the leading overheads in fault-tolerant quantum computation. Existing methods for finding distillation factories require a transversal gate to act correctly on the entire codespace, a constraint that limits both generality and search efficiency. We introduce a strictly weaker borrowed-identity condition, requiring only that the distillation circuit act as the identity on a single input state. It applies uniformly across all levels of the Clifford hierarchy and unifies, within a single level, factories that distill different magic states -- for example, the $|T\rangle$, $|CS\rangle$, and $|CCZ\rangle$ factories. A brute-force search over borrowed-identity circuits with two-group symmetry recovers, within the search range, all distance-2 factories known from code-construction approaches, including entangled-output and multi-output factories previously outside the scope of any single numerical search. This unification yields parent circuits that encode multiple factories, so the output magic-state type can be chosen at compile time rather than fixed by a hard-coded design. The framework also extends beyond CSS codes, unifying constructions, including synthillation and non-CSS catalytic factories, previously obtained by disparate approaches.

Hawking radiation is a fundamental result of quantum field theory in curved spacetime, yet its direct observation remains beyond current experimental capabilities. Circuit quantum electrodynamics provides a practical platform for realizing analogue systems where Hawking-like radiation may be studied under controlled laboratory conditions. In this work, we analyze two superconducting-circuit analogues of Schwarzschild black holes: a tunable dc-SQUID transmission line and a SNAIL-based transmission line supporting solitonic solutions of the KdV equation. We investigate the conditions under which these architectures can generate an observable Hawking temperature and study the impact of dissipation and thermal noise using an open quantum systems approach. To assess the observability of the Hawking signal, we propose complementing particle number measurements with estimates of the Hilbert-Schmidt distance to the thermal bath. Our analysis establishes practical detectability thresholds and shows that Hawking temperatures above approximately 73 mK remain distinguishable under realistic experimental conditions. While the tunable transmission line architecture can reach temperatures of about 113 mK and therefore appears more viable, the solitonic model requires further optimization and more demanding experimental conditions.

It has been noted in the literature that, if post-inflationary reheating is dominated by a stiff fluid with equation of state (EoS) $p>ρ/3$, then the predictions of Starobinsky inflation for the scalar spectral index could be made to agree with measurements from the combined CMB+BAO datasets performed by the Atacama Cosmology Telescope (ACT) and the South Pole Telescope (SPT) collaborations. However, a side-effect of such a stiff epoch is the blue-tilting of the primordial gravitational-wave (GW) spectrum. In this work, we explore the observational consequences of this blue-tilting in three scenarios: (i) a purely stiff-dominated reheating, (ii) a more realistic case where reheating is first dominated by a matter-like fluid (corresponding to inflaton oscillations around the bottom of a quadratic potential well) later followed by a stiff epoch, and (iii) a case analogous to the previous one, but with an earlier radiation-dominated instead of matter-dominated epoch. We show that in all cases the $1σ$ region allowed by recent CMB+BAO data is already excluded by constraints on the amount of radiation present during Big Bang Nucleosynthesis (BBN). Moreover, in a considerable fraction of the remaining $2σ$ region we find that the blue-tilting would be severe enough to make the primordial spectrum detectable in future interferometers such as Einstein Telescope, LISA, DECIGO, and BBO, thus rendering these scenarios testable by these experiments.

Quadratic couplings between ultralight scalar dark matter and Standard Model fields can produce a distorted dark-matter field profile around the Earth. Gradients in the field induce a non-radial, composition-dependent force that can be suppressed at the Earth's surface while remaining accessible to space-based experiments. The MICROSCOPE satellite, which searched for violations of the equivalence principle, can constrain this force, but existing results assume a radial force, and they cannot be directly translated into an optimal bound in the anisotropic regime. We develop a signal template for this regime by organizing the force into radial and polar multipole coefficients and projecting the force onto the MICROSCOPE measurement axis. We use this template to recast the published MICROSCOPE constraint using the component of the signal that overlaps with the radial-force template. We estimate the sensitivity gain that would be provided by an analysis utilizing the additional non-overlapping signal. Such an analysis could improve sensitivity to the couplings of quadratically coupled scalar dark matter by more than an order of magnitude relative to the radial-force recast for dark matter masses $\gtrsim 10^{-9}$ eV.

Fusion-based quantum architectures are the leading approach to photonic quantum computing. However, the sub-threshold regime, where logical error rates must reach the levels required by useful applications, has received little attention. We show that in this regime, fusion failure imposes a noise floor on the logical error rate that prevents all-linear-optics architectures from reaching the required rates at low overhead. For fusion-based architectures using quantum emitter spins, we show that the noise floor is reduced by orders of magnitude at a lower overhead.

We investigate a dark energy scenario generated by neutrino interactions mediated by a light scalar field, in which finite-temperature corrections induce an effective neutrino mass that evolves with the thermal history of the Universe. Within the framework of Unimodular Gravity, these interactions give rise to a non-conservation current, leading to dynamical dark energy. We study one- and two-neutrino realizations of the model. In the one-neutrino case, the dark energy density evolves monotonically, whereas in the two-neutrino scenario it can reach a maximum at intermediate redshifts before decreasing at late times. Using late time cosmological datasets, we constrain the effective interaction strength for lightest-neutrino masses in the range $0.05 \,{\rm meV}\le m_1 \le 1 \,{\rm meV}$. We find preferred interaction scales of order $G_s\sim10^{12} \, {\rm eV}^{-2}$ with a significance of $2 σ$, with the inferred coupling decreasing as the assumed neutrino mass increases. Assuming neutrino couplings of order unity, this $G_s$ value corresponds to an ultralight mediator with mass $m_φ\sim10^{-6} \, {\rm eV}$. We further assess the impact of Planck distance-prior, finding a noticeable reduction in parameter degeneracies and a reconstructed dark energy evolution closer to that of a cosmological constant. Our results show that neutrino interactions can generate both monotonic and non-monotonic dark energy evolutions while remaining compatible with current cosmological observations. The inferred interaction strengths remain consistent with non-zero values for part of the explored neutrino-mass range, supporting neutrino-induced dark energy dynamics as a viable phenomenological extension of $Λ$CDM at the background level.

We propose an algorithmic framework for building and training quantum generative models corresponding to multivariate probability distributions. Our model uses parallel Fourier feature maps for embedding continuous-valued variables combined with a forrelation-type quantum circuit for tuning Fourier coefficients of the quantum model. Crucially, we develop a distinct training strategy where training is enabled at large scale by log-likelihood loss with unbiased Monte Carlo estimator based on Parseval's identity. Unlike prior work that relied on maximal mean discrepancy (MMD) loss, our approach goes beyond matching just low frequency moments, while enabling efficient classical training. Once the model is trained, we use inverse quantum Fourier transforms to map it into a separate sampling circuit in the computational basis. We demonstrate the efficiency of the suggested framework by validating loss estimation at the scale of over 1000 qubits on a single GPU. We show that univariate and bivariate models with highly non-trivial structure can be trained to low total variation distance, while fine-tuned IQP models with MMD loss show poor performance. Comparing to classical baselines represented by normalizing flow and diffusion models, we show that our approach avoids oversmoothing and preserves multi-modal structure of the target. Finally, we have deployed the trained models on superconducting quantum devices, successfully sampling distributions with per-sample execution times of approximately $300\,μ\mathrm{s}$. Our work shows that quantum generative models with the train-on-classical deploy-on-quantum approach can provide both high-quality structure at increased scale and fast sampling access needed for inference.

We investigate the thermodynamic structure of charged AdS black holes in ModMax nonlinear electrodynamics coupled to dRGT-like massive gravity, incorporating Sharma--Mittal entropy corrections. The thermodynamic geometry is analyzed using the Weinhold metric in the parameter space spanned by the horizon radius and electric charge. The resulting thermodynamic Ricci scalar characterizes effective microscopic interactions, with curvature singularities signaling extremal boundaries and degeneracies of the thermodynamic metric. We further construct a rectangular black hole heat engine in the extended phase space and derive an exact expression for its efficiency, demonstrating how the ModMax parameter and massive-gravity couplings influence the enthalpy-based conversion of heat into work, while the Sharma--Mittal parameters modify the Carnot bound through corrections to the black-hole temperature. Finally, a topological analysis of the corrected temperature and generalized free energy reveals both conventional and novel critical points, and the associated conserved topological charge is investigated.

In this work, we investigate quantum tree search via amplitude amplification. Amplitude amplification generalizes Grover's algorithm by replacing the Hadamard initialization with an arbitrary unitary $A$, with Grover's algorithm recovered as the special case of uniform initialization. We demonstrate the construction of a dynamic search tree of depth $m$ with query complexity $\sqrt{\left(b_{avg}\right)^m}$ where $b_{avg}$ denotes the average branching factor, improving upon the commonly assumed $\sqrt{\left(b_{max}\right)^m}$, where $b_{max}$ is the maximum branching factor. We further challenge the widespread assumption that amplitude amplification is inferior to quantum backtracking. In fact, quantum backtracking is unsuitable for problems that do not naturally admit a backtracking structure; in such cases, amplitude amplification yields improved query complexity. We observe that amplitude amplification constructs the search tree dynamically, rendering its internal structure inaccessible, a constraint that applies equally to quantum backtracking. To address this, we propose sampling-based methods to estimate the tree structure, under the assumption that it approximates a normal distribution with increasing depth. Finally, we introduce a quantum greedy search based on a lookahead heuristic inspired by the classical cognitive architecture Soar, which models human-like problem-solving strategies.

The reverse isoperimetric conjecture asserts that, at fixed thermodynamic volume, Schwarzschild--AdS black holes maximize entropy. We prove that this statement is the fixed-volume form of a boundary-completed Noether-charge stability theorem. The essential observation is that the bulk Hollands--Wald canonical energy is not the full entropy Hessian: along exact stationary black-hole families it vanishes, and the missing curvature is supplied by a constrained asymptotic charge Hessian. Combining this boundary term with bulk canonical-energy positivity gives entropy concavity on admissible fixed-volume components, while zero-energy rigidity determines the equality sector. The theorem reproduces the Einstein-gravity area-volume inequality and extends naturally to Wald entropy in higher-derivative theories. Known violations are thereby reinterpreted as failures of compactness, positivity, or rigidity rather than failures of the variational mechanism.

We provide a semi-analytic calculation of the first quantum corrections to the symmetron fifth force around a planar source with nonzero thickness. We find a suppression of the fifth force compared with the classical prediction within a Compton wavelength of the source, which is of order 10% in the parameter region relevant to experiments like CANNEX, while the fifth force is enhanced at larger distances from the source. The resulting change in the spatial profile of the fifth force may be relevant to current and near future terrestrial and astrophysical tests of force laws, and has implications for the optimisation of experimental geometries, including atom interferometers. This work provides a key benchmark for future numerical studies of quantum-corrected fifth forces in screened scalar-tensor theories of gravity.

Quantum fast-forwarding (QFF) is usually formulated for reversible Markov chains, where the projected quantum walk evolution is exactly governed by Chebyshev polynomials of a Hermitian discriminant matrix. We study whether this framework can be extended to nonreversible dynamics for an $α$-perturbed $n$-cycle Markov chain, which preserves circulant structure while introducing controlled irreversibility. We show that the nonreversible case has a fundamental obstruction: for $α\neq 0$, the eigenvalues of $P_α$ leave the interval $[-1,1]$, so $T_m(P_α)$ is not uniformly bounded and cannot arise as an exact unitary compression for all times. Thus, exact Chebyshev-based QFF does not extend directly beyond reversibility. Nevertheless, we obtain a finite-time approximation result using truncated Chebyshev and LCU techniques. The evolution $P_α^t$ can be approximated with degree $τ=O\left(|α|t+\sqrt{t\log(t/η)}\right),$ which recovers the reversible $O(\sqrt t)$ behavior only in the perturbative regime $|α|=O(t^{-1/2})$. This identifies a nearly reversible regime where QFF survives perturbatively and quantifies how irreversibility degrades the speedup.

We introduce and systematically study the length of the slow stable hybrid star branch as a quantitative measure of the extended stability region that arises in hybrid neutron stars when the hadron-quark phase conversion is slow compared to the radial oscillation timescale. Combining generalized piecewise-polytropic hadronic equations of state of varying stiffness with a constant-speed-of-sound quark-matter model, we construct a large set of hybrid equations of state spanning a broad range of transition pressures, energy-density jumps, and quark-matter speeds of sound. We identify four morphological types for the slow stable branch in the mass-radius plane: waterfall branches that descend monotonically from the hadronic maximum mass, bridges that connect the hadronic branch to a second unconditionally stable hybrid branch, tails that extend briefly beyond the maximum mass of an unconditionally stable hybrid branch, and tail-bridges that combine features of the latter two. Their prevalence is governed primarily by the transition pressure and the energy-density jump, while the branch length is also significantly influenced by the stiffness of the hadronic sector and the quark-matter speed of sound. Imposing current astrophysical and microphysical constraints shows that viable long branches are predominantly of waterfall type, and that stiff hadronic equations of state -- strongly disfavored under the rapid-conversion assumption -- remain compatible with all current constraints within the slow-conversion framework. In the plane of transition baryon density versus density jump, slow stable configurations open a new region of viable parameter space inaccessible under rapid conversions.

We investigate the supercritical thermodynamics of Euler-Heisenberg AdS black holes within the framework of Lee-Yang phase transition theory. We show that the system admits two distinct critical points associated with a four-phase thermodynamic structure and identify a degenerate higher-order critical point where the two criticalities merge. Extending the thermodynamic description into the complex domain, we determine the distribution of Lee-Yang singularities and construct the corresponding complex phase diagrams. At the degenerate critical point, we find that a well-defined Widom line emerges despite the absence of a conventional coexistence curve, acting as an effective stability boundary in the supercritical regime. In the two-critical-point regime, the complex phase diagram exhibits two distinct Widom lines, one associated with a coexistence curve and the other arising solely from the complex singularity structure. We further show that the Lee-Yang formalism consistently reproduces the expected phase structure for systems with a single critical point and in the absence of criticality. Our results reveal a rich supercritical phase structure and provide new insights into the origin and physical interpretation of Widom lines.

We study the scattering of the ground states of doubly charmed baryons ($Ξ_{cc}^{++},Ξ_{cc}^{+},Ω_{cc}^{+}$) and light-flavor pseudoscalar mesons ($π,K,η$) up to the next-to-leading order within chiral effective theory. We perform the unitarization of the $S$-wave scattering amplitudes in order to study the excited doubly charmed baryons. The unknown next-to-leading order low energy constants are determined through the fits to recent lattice data in the elastic scattering processes based on the CLQCD ensembles. Following the chiral extrapolation to physical quark masses, we predict resonance, virtual and bound doubly-charmed-baryon states arising from the single- and coupled-channel scattering of $Ξ_{cc}^{++},Ξ_{cc}^{+},Ω_{cc}^{+}$ with $π,K,η$. Furthermore, we also calculate the corresponding scattering lengths, effective ranges, phase shifts and inelasticities at physical quark masses, which could shed light on future experimental searches and lattice simulations.

Highly compact and finely segmented silicon-tungsten electromagnetic calorimeters are being developed within the FCAL collaboration for applications in the LUXE experiment at DESY and future electron-positron collider facilities. These detectors combine tungsten absorber plates with thin silicon pad sensors, providing a small effective Molière radius and high spatial granularity, which are essential for resolving nearby electromagnetic showers in high-occupancy environments. The fundamental active unit of this calorimeter concept is the Compact Silicon Sandwich (CSIS), integrating a silicon pad sensor together with signal routing, high-voltage distribution and mechanical support in a highly compact structure. The assembly of these CSIS modules is performed within a dedicated infrastructure for silicon detector integration. A partially instrumented prototype of such a calorimeter has been tested in an electron beam with energies between 1 and 6~GeV. First results from the 2025 test beam campaign are presented, including minimum-ionizing particle calibration and preliminary event displays illustrating the shower development in the highly granular detector. These results constitute an important step towards the validation of this technology for LUXE and future collider experiments.

In this work, we investigate the exclusive central production of dyons, a particle carrying electric and magnetic charges, in hadronic interactions at LHC energies, assuming the photon fusion mechanism. Motivated by predictions from theoretical physics beyond the Standard Model, we analyze the photoproduction of these particles in the ultraperipheral collision regime, employing the equivalent photon approximation. We estimate the cross-sections for dyons production in spin 0, 1/2, and 1 scenarios, adopting values for electric and magnetic charges from the literature. Our results demonstrate a direct correlation between spin and cross section, predicting a substantially higher probability for the photoproduction of higher-spin dyons.

Strong magnetic fields produced in relativistic heavy-ion collisions can modify fluctuations of conserved charges and, consequently, their associated chemical potentials. We present first-principles $(2+1)$-flavor lattice-QCD results for isospin-driven splittings of conserved-charge chemical potentials between the isobar systems $^{96}_{44}\mathrm{Ru}+^{96}_{44}\mathrm{Ru}$ and $^{96}_{40}\mathrm{Zr}+^{96}_{40}\mathrm{Zr}$ in the QCD crossover region, both at vanishing and nonzero magnetic fields along the pseudo-critical line $T_{pc}(eB)$. We outline a framework that, under strangeness neutrality and charge-to-baryon ratio $r\equiv n_{\rm Q}/n_{\rm B}$, maps the isospin difference between two nuclei, as encoded in $r_{\rm Zr}$ and $r_{\rm Ru}$, onto splitting ratios $Δμ_{\rm Q}/Δμ_{\rm B}$, $Δμ_{\rm S}/Δμ_{\rm B}$, and $Δμ_{\rm S}/Δμ_{\rm Q}$ as functions of $μ_{\rm B}(r_{\rm Ru})/Δμ_{\rm B}$. Using continuum-estimated lattice results for the leading-order coefficients $q_1\equiv(μ_{\rm Q}/μ_{\rm B})_{\rm LO}$ and $s_1\equiv(μ_{\rm S}/μ_{\rm B})_{\rm LO}$, we find that, at vanishing magnetic field, the splitting ratios are of similar magnitude to recent Bayesian extractions from STAR isobar data and yield $Δμ_{\rm Q}<0$ and $Δμ_{\rm S}>0$, with the electric-charge sector dominating. At nonzero magnetic fields, the splitting ratios show only moderate $eB$ dependence. We therefore further examine Ru--Zr differences in the normalized magnetic-field response of chemical-potential ratios, particularly those involving $μ_{\rm Q}/μ_{\rm B}$, which display a pronounced enhancement in lattice QCD. We also present hadron resonance gas (HRG) results and experimentally motivated proxy observables with kinematic cuts to facilitate contact with experiment.

In this work, we investigate the anisotropic suppression of reconstructed $K^{*0}$ resonances arising from hadronic rescattering using the A Multi-Phase Transport (AMPT) model for Au+Au collisions at $\sqrt{s_{NN}}=200$ GeV. We demonstrate that the rescattering probability of the decay daughters depends strongly on the decay angle $θ^{*}$ due to Lorentz boost effects, which lead to smaller laboratory-frame momenta for daughters emitted opposite to the parent particle motion. This anisotropic suppression influences several experimentally measured observables. We show that the reconstructed $K^{*0}$ yield exhibits a strong $θ^{*}$ dependence. Furthermore, the anisotropic loss of resonances modifies the angular distributions used to extract the spin alignment parameter $ρ_{00}$ in the production-plane and helicity frames. Even in the absence of intrinsic polarization in the model, the reconstructed $K^{*0}$ sample shows deviations of $ρ_{00}$ from the unpolarized value of $1/3$, with opposite trends in the two reference frames. These results demonstrate that hadronic rescattering can generate apparent polarization signals and must be carefully considered in experimental measurements of vector-meson spin alignment using production plane and helicity frame.

In this paper, we investigate the sensitivity of future NINJA-like experiment at J-PARC to eV-scale sterile neutrinos within the 3+1 framework. We perform a phenomenological feasibility study using the $ν_μ\rightarrow ν_e$ appearance, $ν_μ\rightarrow ν_μ$ and $ν_e \rightarrow ν_e$ disappearance channels, focusing on possible future configurations of the detector located at different floors of the NM building (B2, SS, and GROUND), corresponding to different off-axis angles. Our analysis is based on a simplified and effective detector response, in which events are classified into electron-like and muon-like topologies and constant benchmark selection efficiencies are applied. We explore different exposure scenarios and assess the impact of analysis choices such as upper energy cuts. We include systematic uncertainties corresponding to normalization for signal and background rates and study the robustness of our results with respect to variations in the assumed energy resolution, and vary efficiencies for key backgrounds such as muon misidentification from charge current and neutral current interactions. Finally, we examine the effects of combining data from multiple detector locations. We find that the SS floor provides the strongest constraints on the active-sterile mixing parameters, while the B2 and GROUND configurations offer constraints comparable to the current bounds for probed mass-squared differences. Our results indicate that a NINJA-like detector, optimized for sufficient statistics and benchmark identification performance, has the potential to provide competitive constraints on light sterile neutrino scenarios in its future runs.

In the presence of a bias term, the conventional condition for forming a domain-wall network is $p_{\rm fv}>p_c\simeq 0.31$, with $p_{\rm fv}/(1-p_{\rm fv})=\exp(-ΔV(0)/V_b(0))$, where $p_{\rm fv}$ is the false-vacuum fraction immediately after the phase transition, $ΔV(0)$ is the zero-temperature energy splitting between the false and true vacua and $V_b(0)$ is the zero-temperature barrier height measured from the true vacuum. This criterion, however, cannot be generally valid, since it is insensitive to the dynamics of the phase transition. In this work, we derive a dynamical criterion for domain wall formation in the presence of a bias term. We evaluate $p_{\rm fv}$ at the freeze-out temperature of the false-vacuum correlation volumes $T_{\rm fo}$, obtaining a substantially stricter criterion. The same dynamical picture also yields a necessary consistency condition for applying scaling-regime gravitational-wave estimates, $T_{\rm fo}>T_{\rm ann}$, where $T_{\rm ann}$ is the annihilation temperature inferred from the scaling-regime dynamics.

We study the late-time relaxation of a neutral relativistic viscous fluid in $d+1$ dimensions. In the long-wavelength regime, linearized hydrodynamics predicts that the sound mode at momentum $nk$ decays as $e^{-n^2ω_I t}$. However, nonlinear analysis gives a decay of $e^{-nω_I t}$. We derive a closed asymptotic attractor solution in which the frequency of the $n$-th harmonic locks to $n$ times the complex frequency of the fundamental mode. The amplitude envelopes for energy current $J$ obey a simple cascading relation, $J_n=α_J^{\,n-1}J_1^n$, with $α_J$ fixed by the equation of state, the longitudinal viscosity, and the fundamental wavenumber. For conformal fluids, $α_J=1/(8ηk)$, in agreement with the holographic result of arXiv:2512.07242. The existence of the attractor shows that, even near equilibrium, field powers are not equivalent to amplitude order.

The recently introduced formalism of chiral cluster seeds replaces quantum cluster variables with deformed vertex operators. In this framework, a decorated quiver associated with a seed encodes the operator product expansions of the corresponding vertex operators. This formalism is applied to several $(q,t)$-deformed W-algebras, including $\mathcal{W}_{\mathfrak{q},\mathfrak{t}}(\mathfrak{gl}(N|M))$, $U_q(\widehat{\mathfrak{sl}}_2)$, and the deformed Bershadsky--Polyakov algebra. In particular, it is shown that different free field realizations of the currents are related by mutations of the associated chiral cluster seed. The second part of the paper introduces a $(q,t)$-deformation of the subregular W-algebras, denoted by $\mathcal{W}_{\mathfrak{q},\mathfrak{t}}^{\text{sub}}(\mathfrak{sl}(N))$. All free field realizations obtainable through seed mutations are described. An embedding of $\mathcal{W}_{\mathfrak{q},\mathfrak{t}}^{\text{sub}}(\mathfrak{sl}(N))$ into the free field realization of $\mathcal{W}_{\mathfrak{q},\mathfrak{t}}(\mathfrak{sl}(N))$ tensored with a rank-two Heisenberg algebra is constructed. This embedding may be viewed as a deformed analogue of inverse quantum Hamiltonian reduction. The relation between the subregular algebras and $\mathcal{W}_{\mathfrak{q},\mathfrak{t}}(\mathfrak{gl}(1|N))$ is also discussed.

Cosmic-ray muon scattering has shown considerable potential for detecting nuclear materials and other dense contraband, but practical deployment remains challenging. A major difficulty arises from the coupling between material properties and muon momentum, since the broad natural momentum distribution influences the scattering angle and prevents unambiguous material identification. In this work, we propose a Coarse Momentum-Aware Domain Adaptation (CMADA) method to enable precise identification of materials. Instead of relying on high-precision momentum measurements, the proposed framework adopts coarse momentum binning combined with unsupervised domain adaptation to learn transferable scattering representations. In addition, a precision review mode based on averaging repeated samplings was proposed to further enhances identification performance. The coarse momentum binning strategy improves same-domain identification accuracy from 62.15% without momentum information to 89.52% with 5-bin momentum information, and further to 93.37% (precision review mode). Furthermore, the proposed unsupervised domain adaptation framework improves the cross-domain identification accuracy from 71.71% for the source-only baseline to 89.00% without requiring target domain labels.

Using the soft-gluon theorem, we identify a soft-recoil mechanism by which unresolved gluon radiation induces decoherence in the spin correlations of heavy quark-antiquark pairs produced in deep-inelastic scattering. We show the eikonal soft contribution preserves the Born spin structure, whereas the subleading soft term generates stochastic recoil-induced rotations of the spin-correlation plane. Upon tracing over the unresolved gluon, these rotations produce an effective dephasing channel: the normal-axis correlation remains unchanged at this order, while the in-plane spin coherences are suppressed. We estimate the resulting reduction of concurrence and Bell-CHSH violation, and propose a radiation-binned EIC observable based on the ratio of in-plane to normal spin correlations. This observable isolates the characteristic anisotropic suppression predicted by the soft-recoil mechanism and provides a measurable handle on radiation-induced spin decoherence of an entangled quark-antiquark pair produced in a deep-inelastic scattering process.

This work presents the layout design of the LHCb UP detector, a MAPS-based pixel tracker composed of four detection planes, and several approaches for its standalone track reconstruction. The dedicated UP tracking algorithms demonstrates that efficient standalone reconstruction can be achieved for LHCb Upgrade II with high purity, reaching efficiencies close to $98\%$ while maintaining a ghost rate below $4\%$. These results indicate that UP standalone tracks can provide high-quality inputs for global LHCb reconstruction and offer a viable solution for future high-luminosity tracking and real-time reconstruction challenges.

Evidence for two KK graviton candidates has been previously reported at 380 GeV and 700 GeV. Following a Randall Sundrum interpretation, two extra resonances should appear at 1000 GeV and 1300 GeV. Recently ATLAS, in its search for triple Higgs coupling, has reported an excess in that mass region in conformity with this prediction. Local cross sections are therefore clearly in excess of the standard predictions even for large values of kl. While still marginally significant, this effect appears in a mass region with low background which allows to expect good prospects of discovery with RUN3 data. Such a result could therefore allow to interpret an excess in the measurement of kl and confirm the existence of a series of KK graviton resonances observable at LHC. Other opportunities seem to appear in searches for heavy resonances decaying into ZZ/WW in semi-leptonic and fully hadronic modes. Indirect evidences for the RS model coming from precision measurements are also presented.

Gauge theories of massive $p$-forms are connected by various dualities, which hold classically but may be broken at the quantum level. One example is the $BF$ theory of topologically coupled $p$- and $(d-p-1)$-forms in $d$ dimensions, where the coupling between forms results in a manifestly gauge invariant mass term for either form when the other is integrated out classically. We perform the path integral quantisation of this theory; by integrating out one of the forms, the resulting determinants are sensitive to the topology of spacetime, and counterterms must be introduced to renormalise their divergences. We compute these determinants in terms of the topological numbers of spacetime, showing explicitly how the quantum duality of the massive theories is broken on topologically non-trivial backgrounds. This is directly related to the quantum breaking of the massless duality between the form that was integrated out and the longitudinal modes of its partner. In particular, the difference of counterterms is proportional to the Euler characteristic of spacetime. The existence of gravitational instantons suggests that these dualities may be broken even in Minkowski space in the presence of topological fluctuations.

The Weak Gravity Conjecture (WGC) is a swampland criterion of long standing: any consistent theory of quantum gravity must contain a charged particle whose charge-to-mass ratio exceeds that of an extremal black hole, so that gravity remains the weakest force. The AdS/CFT correspondence offers a calculable boundary handle on bulk gravity, and the imaginary parts of bulk quasinormal modes are read off the boundary as poles of a retarded Green's function. We show that the WGC follows from this boundary calculation in two settings that fall outside the Reissner--Nordström idealisation: static spherically symmetric black holes in dRGT massive gravity, and dyonic black holes in Einstein--ModMax non-linear electrodynamics. The chain runs from the metric and gauge field, through the charged Klein--Gordon equation, into a near-horizon scaling limit whose radial equation reduces to Whittaker form; the conformal weight $ν_0$ then enters a damping-time inequality. For the dRGT black hole every massive-gravity parameter ($α,β,m_g,h$) cancels out, leaving the universal saturation $q/(m r_+) \geq 1/\sqrt{2} \approx 0.707$. For the Einstein--ModMax black hole the duality-symmetric non-linearity parameter $γ$ survives, and yields $q/(m r_+) \geq e^{-γ/2}$, which reduces to the Reissner--Nordström bound $q/(m r_+) \geq 1$ in the Maxwell limit $γ\to 0$. Either result is of order unity, and the second weakens monotonically as the non-linearity grows. We then relax three of the simplifying assumptions of the dRGT derivation, namely exact extremality, minimal coupling, and the absence of higher-curvature terms. The cancellation breaks. Each correction reintroduces $m_g,α,β$ into the bound through a controlled functional dependence, and we tabulate and plot the relaxed forms across parameter space.

This paper presents the first non-perturbative lattice determination of the Yang--Mills topological susceptibility slope $χ^\prime$ in the large-$N$ limit. This quantity represents the $\mathcal{O}(p^2)$ term of the momentum expansion of the topological charge density two-point correlator, and has important theoretical and phenomenological implications for strong interactions. This calculation is based on a novel algorithm that avoids topological freezing at large $N$ on fine lattices, and on a novel method to reliably compute $χ^\prime$ on the lattice. The results of this study are relevant for the description of the proton spin in deep inelastic scattering experiments via the Shore--Veneziano formula.

We consider a black hole solution embedded in a mass varying dark matter halo and study scalarization induced by dark matter. We derive the required conditions on the scalar dark matter coupling that allow for a regular bound state scalar cloud. We find that such configurations exist only for quantized coupling values in terms of the halo's intrinsic parameters (for Hernquist, this reduces to the inverse of the compactness). This study provides a novel connection between dark matter phenomenology and scalarization around a black hole.

The Sugawara construction provides a natural way to construct the Virasoro algebra from a current algebra. It was shown in Ref.~\cite{Ghazi:2025oin} that for the $u(1)^2$ Kac-Moody current algebra, there exist additional constructions that exhibit a $\mathbb{Z}_N$-graded structure. Indeed, the space of such constructions defines a non-compact algebraic variety whose dimension depends on $N$. In this paper, we consider the compactification of these algebraic varieties by adding points at infinity to the non-compact part, and show that these points correspond precisely to generalizations of $BMS_3$-like algebras. More explicitly, for a $\mathbb{Z}_2$ grading, the corresponding algebra coincides with the $BMS_3$ algebra, which takes the form $\mathrm{Vir} \rtimes F$, where $F$ is an infinite abelian ideal of the full algebra. For $N > 2$, we show that there exist generalizations of the standard $BMS_3$ algebra of the form $\mathrm{Vir} \rtimes F$, where $F$ is a nonabelian ideal that forms a nilpotent algebra of depth $r < N$. We further demonstrate that the depth of the algebra is related to the order of the singularity of the algebraic variety at that point. We also show that the polynomials defining the algebraic varieties exhibit a factorization property into linear factors, which, if true, classifies all $BMS_3$-like algebras. Finally, we study the central extensions of these algebras, which are consistent with the general structure of algebras corresponding to primary fields of conformal weight $h = 2$.

The concept of self-similarity in the internal structure of the proton, rooted in scale invariance and fractal geometry, provides an intriguing framework for understanding the behaviour of parton distribution functions (PDFs), particularly in the small \textit{x} region probed in deep inelastic scattering (DIS). Phenomenological models based on self-similarity have been shown to reproduce key features of experimental data, suggesting that recursive scaling patterns may play an important role in partonic dynamics. In this work, we present an overview of scale-invariant descriptions of proton structure, focusing on self-similar models developed in earlier studies and their phenomenological implications for structure functions and parton distributions. We then explore possible conceptual connections between these fractal-inspired descriptions and modern holographic approaches to QCD, particularly within the framework of light-front holographic QCD. By comparing the scaling behaviour appearing in phenomenological models with the geometric structure underlying holographic QCD, we highlight qualitative correspondences that suggest a broader role of scale invariance in proton structure. Although the connection remains interpretive rather than derivational, it offers a complementary perspective of how fractal-like scaling observed in DIS may relate to geometric scaling in holographic descriptions of QCD.

A unified afterburner framework is presented to describe nucleon--nucleon final-state interactions and light-(anti)nuclei production via coalescence in high-energy measured in pp collisions at the LHC collisions. The model reproduces qualitatively light-(anti)nuclei spectra without fine-tuning of the model parameters, as well as correlation observables, and can be extended to beyond proton--proton collisions.

Existing laboratory, astrophysical, and direct-detection searches constrain the kinetic-mixing portal $ε$ of dark-photon dark matter but do not determine the cosmological origin of the relic abundance. We show that cosmic microwave background (CMB) isocurvature provides an independent probe of Higgsed dark-photon production histories: two models with identical $(m_{\Ap},ε)$ and identical present-day abundance can produce distinct cold-dark-matter (CDM) isocurvature signatures if their hidden-scalar evolution differs. The relevant observable is the logarithmic response of the final dark-photon abundance to the inflationary dark-Higgs displacement. We develop a model-independent response formalism and demonstrate that any perturbative inheritance branch with conserved comoving yield necessarily satisfies $\qeff\ge2$. Consequently, a perturbative branch accounting for the full dark-matter abundance through $h\to\Ap\Ap$ requires an initial coherent displacement exceeding $3.5\times10^4H_*$, while ordinary stochastic fluctuations over $\mathcal{O}(60)$ inflationary e-folds are exponentially unlikely to generate the required field amplitude. Viable Higgsed dark-photon scenarios therefore require either subdominant abundance, coherent initial conditions, suppression of the inherited scalar response prior to freeze-out, or sufficiently cold momentum evolution of the produced vector population.

We provide a systematic and rigorous geometric framework that relates three structures naturally associated to BPS central charges in $\mathcal{N}=2$ supersymmetric gauge theories: the split attractor flow (SAF) of $|Z|$, the Hessian flow (HF) of $\operatorname{Im}(e^{-i\vartheta}Z)$, and the spectral network (SN) on the base curve of the Hitchin fibration. Our main contributions are: (i) a concise proof of orthogonality between SAF and gradient Hessian flow using only the Kahler structure; (ii) a precise lift-projection duality showing that the spectral network projects to the *characteristic Hessian flow* (the Hamiltonian flow of $\operatorname{Im}(e^{-i\vartheta}Z)$) on the Hitchin base, clarifying a crucial distinction; (iii) a complete proof of the Kontsevich-Soibelman (KS) equivariance by induction on the SAF tree depth, with the geometric ordering provided by the characteristic Hessian flow. We illustrate the framework with detailed and nontrivial examples: $SU(2)$ pure and $N_f=4$ (including new BPS indices for higher flavour charges), $SU(3)$ pure (full BPS spectrum reconstruction), $SU(4)$, the Kronecker $3$-quiver, and we apply the induction to derive a closed-form BPS spectrum for the Argyres-Douglas $H_1$ theory, $Ω(nα_1+mα_2)=\binom{n+m}{n}$, which is a new result. In the tropical limit we obtain an explicit generating function for disk counts in $SU(N)$ gauge theories, $Z_{\mathrm{disk}}^{SU(N)} = \prod_{α\inΦ_+}\prod_{k\ge1}(1-e^{-k\langleα,y\rangle})^{-\binom{k+\mathrm{ht}(α)-1}{\mathrm{ht}(α)-1}}$, which follows directly from our recursion. These results demonstrate the computational power of the unified framework and provide new, verifiable predictions.

We perform a Bayesian analysis of transverse momentum ($p_{\mathrm{T}}$) spectra of identified particles, i.e., pions, kaons, and protons, at midrapidity in Au+Au collisions and Pb+Pb collisions using the Boltzmann-Gibbs blast-wave (BGBW) model. We investigate whether it is possible to simultaneously describe the $p_{\mathrm{T}}$ spectra of identified particles without imposing the particle species-dependent $p_{\mathrm{T}}$ fit ranges -- a practice that was followed in conventional blast-wave model studies to achieve reasonable simultaneous fits. Using Bayesian analysis, our results indicate that a simultaneous description of the $p_{\mathrm{T}}$ spectra of pions, kaons, and protons is feasible without imposing the particle species-dependent $p_{\mathrm{T}}$ fit ranges, for Au+Au collisions up to the available data ($\sim$2 GeV/c) and for Pb+Pb collisions up to 3 GeV/c. The extracted parameters remain broadly consistent with those obtained from conventional BGBW simultaneous fits, while the extension of the fit range leads to moderate changes in some parameters. Furthermore, Bayesian analysis yields well-constrained posterior distributions for the kinetic freeze-out temperature $T_{kin}$, the average transverse flow velocity $\langle β_{\mathrm{T}}\rangle$, and the exponent of the velocity profile $n$ and shows their correlations transparently. We suggest that the BGBW model in a Bayesian inference framework proposed can be applied in future data analyses to simultaneously describe the $p_{\mathrm{T}}$ spectra of identified particles and extract the relevant information about the collision system.

Compact stellar systems are often used to place stringent constraints on the particle mass of ultralight dark matter (ULDM), as the heating effect induced by wave interference can drive system expansion, potentially bringing them into tension with observations. In a recent study, we pointed out that internal two-body relaxation in these stellar systems may have a significant impact on their evolution in ULDM halos, an effect overlooked in previous studies. Here, we further investigate the influence of stellar metallicity, the Milky Way's tidal field, and the ULDM particle mass on the long-term fate of compact stellar populations. We find that metal-richer systems are generally more resistant to disruption. The tidal field of the Milky Way, by altering the orbital motion of the stellar systems within host ULDM halos, can significantly affect their stability. Furthermore, we find in our simulations that the heating effect becomes stronger with increasing ULDM particle mass when the system size is much smaller than the ULDM de Broglie wavelength $R_{\rm h} \ll λ_{\rm dB} $, in contrast to the $λ_{\rm dB}\lesssim R_{\rm h}$ case. These results highlight the complexity of the evolution of compact stellar systems in ULDM halos, and suggest that existing constraints derived from the systems, such as ultrafaint dwarf galaxies, may require careful revision.

Based on $8.5~{\rm fb}^{-1}$ of $e^+e^-$ collision data collected at center-of-mass energies between 4.42 and 4.95 GeV with the BESIII detector at the BEPCII storage ring, we investigate the process $e^+e^-\to π^+π^-D_s^+D_s^-$. With no significant signal observed, upper limits on the Born cross sections of $e^+e^-\to π^+π^-D_s^+D_s^-$ at each energy value are determined at the 90% confidence level. Additionally, a search for intermediate charmonium-like resonances is performed in the $M(D_s^+D_s^-)$ invariant-mass spectrum, but no significant resonant structures are observed with the current statistics.

We present a novel unified conditional-probability framework for relativistic systems in which conditioning on additive conservation laws simultaneously yields equilibrium occupation statistics and conservation-induced correlations. In this formulation, equilibrium arises as a conditional limit law of a closed system. The one-mode marginal gives Maxwell--Boltzmann, Bose--Einstein, and Fermi--Dirac statistics at leading saddle order, with the conserved quantities fixing the exponential tilt and the microscopic occupation measure determining the statistics. Expanding the two-mode marginal to Gaussian order gives the leading finite-rank covariance between modes induced by exact conservation. When contracted with observables linear in mode occupations, this covariance gives their leading exact-conservation contribution. We use this structure to define projected observables orthogonal to selected conserved quantities. By construction, their covariance has no leading exact-conservation contribution. In small collision systems, where conservation effects are less suppressed by multiplicity and can survive standard nonflow suppressions, this provides a direct way to isolate conservation-aligned contributions to long-range correlations. We demonstrate this with PYTHIA8/Angantyr-generated p+Pb events at $\sqrt{s_{\mathrm{NN}}}=5.02~\mathrm{TeV}$ by comparing ordinary and projected covariances, showing that the projection removes the conservation-aligned contribution while leaving the conservation-orthogonal covariance essentially unchanged.

Mirror dark matter (MDM) can modify neutron-star structure and tidal response through gravitational coupling. In this work, we construct an ordinary-matter equation of state (EOS) by comparing hadronic matter described by the relativistic mean-field NL3\(ωρ\) model, and quark matter in the framework of the Nambu--Jona-Lasinio (NJL) model. The stable branch is determined through a Maxwell construction, which serves to connect distinct phases of matter. For the parameter sets considered here, \(m_u=5.2~{\rm MeV}\) is the lowest light current-quark mass in the scanned range that satisfies the \(2M_\odot\) maximum-mass requirement, while \(m_u>5.2~{\rm MeV}\) all yield stable neutron-star configurations without a resolved macroscopic quark core. The small-radius inferences for PSR J0437--4715 and XTE J1814--338, together with the tidal-deformability constraint from GW170817, are sensitive to the dark-matter mass fraction \(f_D\). The commonly used GW170817 interval \(70\lesssimΛ_{1.4}\lesssim580\) corresponds approximately to \(0.12\lesssim f_D\lesssim0.88\) in the present model. These results indicate that, even without a macroscopic quark core, MDM can provide an important mechanism for reducing the visible radius and modifying the tidal response of neutron stars.

We investigate the forward--backward asymmetry in the production and elliptic flow of prompt D0 mesons in proton--lead (p--Pb) collisions at$\sqrt{s_{\mathrm{NN}}}=8.16$ TeV using the heavy-flavor improved string-melting version of the AMPT model. The model calculations provide a simultaneous description of nuclear modification factor $R_{\mathrm{pPb}}$ and $v_2$ in forward and backward rapidities. We find that the observed asymmetry arises from the interplay of initial-state cold nuclear matter effects and final-state partonic interactions, with the competition between coalescence and fragmentation playing a critical role in shaping the transverse momentum and rapidity dependence of both observables. This work suggests that a partonic medium is formed in high-multiplicity p-Pb collisions at LHC energies.

This paper provides an isomorphism between the space of twisted gauge fields on a principal bundle $\mathcal{P}$ and the space of standard gauge fields on a different principal bundle $\mathcal{Q}$ associated to $\mathcal{P}$. This isomorphism extends to local fields on the base manifold, which enables the use of local twisted fields in standard gauge theories (e.g. Yang-Mills-like theories). This allows one to deal with two symmetry groups, coming from $\mathcal{P}$ and $\mathcal{Q}$, respectively. The construction makes use of a larger principal bundle $\mathcal{S}$ which has $\mathcal{P}$ and $\mathcal{Q}$ as quotient bundles. The gauge structure on $\mathcal{S}$ encodes both standard and twisted gauge structures on $\mathcal{P}$. In addition, the isomorphism classes of bundles $\mathcal{S}$ are in 1:1 correspondence with the equivalence classes of cocycles (up to a coboundary). This paper also provides a new interpretation of (full) dressing fields as dynamic (or active) sections of a principal bundle.

The high-energy behaviour of electroweak scattering amplitudes is of theoretical and phenomenological interest. In these proceedings, we summarize recent progress in analytic high-energy calculations for two-loop four-point electroweak amplitudes in the full Standard Model. As a representative application, we discuss the electroweak corrections to Higgs boson pair production, where rich structures of logarithmic and power corrections appear and sizeable effects are found in the high-energy region.

We study a massless minimally coupled compact scalar, or axion, on global $D$-dimensional de Sitter space (dS$_D$). We quantise the theory canonically, determine the quantum dS charges, and find that the axion zero mode supplies a quantum-mechanical factor beyond the oscillator Fock space, $\mathcal{F}$. The full Hilbert space is $\mathcal{H}=L^2(S^1)\otimes\mathcal{F}$, with the integer quantum-mechanical momentum on $L^2(S^1)$ identified with the conserved $\mathrm{U}(1)$ shift charge. The 1-particle unitary irreducible representation (UIR) of the dS group, $\mathrm{SO}(D,1)$, captures the oscillator sector, but misses the zero mode. We find that the neutral 0-particle state is dS-invariant and normalisable. Charged 0-particle states are normalisable, but only $\mathrm{SO}(D)$ invariant. This implies that geodesic observers related by dS boosts do not agree on the particle number in a charged sector, an effect absent in QFTs equipped only with the standard Bunch-Davies vacuum. We compute field-strength Wightman 2-point functions in charged sectors and find that they are Hadamard. For non-zero charge they are not dS-invariant at finite global times, but they are asymptotically so at early and late times. We complement this analysis with a Euclidean perspective. The ordinary $D$-sphere path integral, $Z_{S^D}$, written in terms of Harish-Chandra characters, has access only to the neutral sector. Charged sectors require vertex-operator insertions, and summing over them gives a decorated sphere path integral, $\widehat{Z}_{S^D}=Z_\text{QM}\,Z_{S^D}$, that captures the entire Hilbert space, with $Z_\text{QM}$ denoting the partition function of a quantum rotor at a dimension-dependent effective temperature. Finally, in dS$_3$, we use the duality between an axion and a photon to translate our results to electromagnetism, where the axion zero mode gives rise to magnetic monopoles.

Precision pulsar timing typically relies on the stability of average pulse profiles, enabling time-of-arrival (TOA) estimation through template cross-correlation. This assumption breaks down for highly variable radio sources such as Rotating Radio Transients (RRATs) and fast radio bursts (FRBs), where individual pulses could exhibit strong variability in morphology and amplitude, and no single averaged profile may represent the underlying emission process. We present toa_sp, an open-source Python package for extracting TOAs directly from PSRFITS search-mode data without requiring profile folding into a stable template. The framework implements a suite of complementary single-pulse timing strategies, including parametric profile fitting, non-parametric estimators, and adaptive sub-band and time-resolution optimisation, together with empirical diagnostics for assessing model consistency. We apply toa_sp to 688 single pulses from a 3-hour FAST observation of RRAT~J1913+1330. The resulting TOAs residual achieve a weighted RMS residual of 1.33\,ms, a 24\% improvement over a standard template-based PSRCHIVE pipeline, while retaining all pulses without statistical outlier rejection. A set of bright FRB 20220529 bursts provides a controlled test of the framework across regimes of increasing pulse complexity, revealing frequency-dependent substructure not captured by band-integrated profiles. We introduce an empirical convergence diagnostic that identifies well-constrained pulses and guides the transition between parametric and non-parametric regimes. Full multi-strategy processing of 688 pulses requires approximately 7.6\,s per pulse on a 10-thread CPU. The package is publicly available via pip install toa_sp.

We construct the Euclidean CFT$_{d}$ dual of a generic massive scalar in Lorentzian dS$_{d+1}$ via analytic continuation. The resulting $PT$ defect defines a $PT$-invariant state that reproduces the Bunch-Davies Wightman function. However, the entanglement entropy captures only the real part of the central charge. This motivates a single-geodesic dual based on the timelike geodesic-integrated Wightman function, which yields the correlators between a bulk operator and a linear combination of an OPE block and its Casimir partner. We also derive the associated conformal defect and anomaly from an integral identity of the dS/CFT symmetry group.

We present a full classification of UV complete QFTs that RG flow to the 2D Ising CFT by solving the bootstrap equations for massless right--left S-matrices. For the Ising model with E_8 spectrum, we find exactly four completions, arising from higher-T\bar T-type deformations, including a previously unknown ``golden flow'' whose UV fixed point is a diagonal su(2) coset CFT (c=25/14) along Δ_{\rm rel}=2/7. A universal Fibonacci/E_8 structure governs the R--L adjacency matrices and the Y-system periods, so that the E_8 symmetry persists across all RG scales.

We study pair production of a vector-like lepton doublet at polarized future linear colliders, focusing on the charged and neutral channels $e^+e^-\toτ^\prime\barτ^\prime$ and $e^+e^-\toν^\prime\barν^\prime$. The benchmark masses are $M_{\rm VLL}=1000,1200~{\rm GeV}$ at CLIC with $\sqrt{s}=3~{\rm TeV}$ and $M_{\rm VLL}=390,460~{\rm GeV}$ at ILC with $\sqrt{s}=1~{\rm TeV}$. Using the realistic polarization configurations LR$=(-0.8,+0.3)$, RL$=(+0.8,-0.3)$, LL$=(-0.8,-0.3)$, and RR$=(+0.8,+0.3)$, we evaluate tree-level production-level cross sections and construct observables designed to test the electroweak structure of the doublet. The charged channel is consistently larger than the neutral channel because it receives both photon and $Z$ exchange. In the LR configuration, the charged-channel rates reach $23.32$ and $20.28~{\rm fb}$ at CLIC, and $188.83$ and $129.04~{\rm fb}$ at ILC, for the two benchmark masses at each collider. We express rate reach through the projected visible-fraction requirement $f_{\rm vis}^{95}=3/(\mathcal{L}σ_{\rm prod})$, keeping the result independent of a specific decay selection. To quantify charged--neutral separation we use the absolute discriminator $D_σ$, which reaches about $0.546$ at CLIC and $0.542$ at ILC in the LR benchmark. We also find a stable observed asymmetry separation, $|ΔA_{LR}^{\rm obs}|\simeq 0.45$--$0.46$, between the charged and neutral channels. The corresponding production-level statistical projection gives sizeable $Z_A$ values for the benchmark luminosities, scaling as $\sqrt{\mathcal{L}_{\rm tot}f_{\rm vis}}$ under an equal LR/RL luminosity split. These results demonstrate that the beam polarization can provide a representation-sensitive diagnostic of vector-like lepton doublets, beyond a simple rate enhancement.

We study pseudoscalar heavy mesons in the QCD instanton vacuum beyond the static limit. Finite-mass effects in the heavy-light loop are encoded in a separable effective vertex built from a profile function $φ(\vec{p})$, kept distinct from the static Wilson-line form factor $F_Q^{(\infty)}(\vec{q})$ of the $m_Q\to\infty$ limit. The pseudoscalar two-point function fixes the residual mass $Λ$ and the residue-normalized meson-quark coupling, from which we evaluate the decay constant, the spin-independent kinetic matrix element, and the zero-recoil slope of the Isgur-Wise function at order $1/m_Q$. The subleading calculation is restricted to the kinetic (derivative) part of the HQET operators. For a representative vertex calibrated to the $B$-meson decay constant and the spin-averaged $B$-meson mass, we obtain $f_B = 186.8$~MeV, $Λ= 184.5$~MeV, $m_b^{\mathrm{eff}} = 5.04$~GeV, $λ_1^{(\partial)} = -0.922~\mathrm{GeV}^2$, and $ρ_{\mathrm{IW}}^2 = 1.105$. The kinetic contribution yields a mass shift of order $Λ/2$ and a sizable $1/m_Q$ current correction, indicating that the spin-independent nonperturbative $1/m_Q$ sector is a sensitive probe of the finite-mass heavy-light vertex.

We study the stochastic gravitational wave background originated from graviton bremsstrahlung in decays of scalar leptoquarks, which are colored scalar bosons simultaneously coupling to a quark and a lepton. We take the scalar leptoquarks in the $\mathrm{SU}(5)$ grand unified theory as a concrete example. Stringent experimental bounds on proton decay force these particles to be superheavy, which in turn renders their graviton bremsstrahlung, induced by quantum gravity effects, less suppressed. By solving the relevant Boltzmann equation, we trace the evolution of the scalar leptoquark number density in the early universe and use it to compute the resulting gravitational wave spectrum. We find that high-frequency gravitational wave detectors employing resonant cavity techniques offer a promising means to probe such signals.

When fermions pair across mismatched Fermi surfaces, the mismatch reflects a built-in inequivalence between the species. We show it can instead arise dynamically by spontaneous symmetry breaking. In a massless two component Dirac theory with exact SU(2) flavor symmetry, a self-interacting vector boson condenses, splitting the Fermi surfaces while preserving time reversal. Pairing then yields a stable relativistic superfluid, promoting the Chandrasekhar-Clogston line to a surface in coupling space, the mismatch fixed self-consistently by the symmetry-breaking coupling.

One of the main challenges in string theory Calabi-Yau compactifications to four dimensions is the stabilization of the massless complex structure and Kähler moduli. In type IIB string theory, complex structure moduli can be stabilized perturbatively by turning on fluxes on the internal space, while there is no perturbative mechanism for Kähler moduli stabilization. Since every Calabi-Yau manifold has at least one Kähler modulus (the overall volume), there is no hope to stabilize all moduli perturbatively. A way out is given by Landau-Ginzburg/Gepner models string vacua which can have no Kähler moduli. To identify the most promising candidates for fully stabilized perturbative string vacua, we provide an exhaustive list of Landau-Ginzburg orientifold models with no Kähler moduli, and compute for each model the number of complex structure moduli together with the tadpole charge. From this, we can identify which of these models are genuine candidates for phenomenologically relevant string vacua.

We employ the joint QED and QCD factorization of deeply inelastic, electron-proton scattering with generic initial state radiation to probe the possibility of exotic particle emission -- i.e., of weakly coupled particles originating from a dark or hidden sector -- through anomalous energy loss. We leverage this possibility through the consideration of phase-space-limited kinematic regions, for which the emission of an additional, undetected particle can particularly impact the associated cross-section. In this first paper, as a proof of principle, we focus on radiation from the incoming electron, considering the modification of the lepton distribution function from the emission of particles, that could have spin of up to 2 and various, well-motivated electron couplings. We illustrate the sensitivity of our approach through the computation of the modified cross-sections for the emission of MeV-GeV mass-scale, spin 0 particles in kinematics chosen for their sensitivity to initial state electron radiation and suitable to the forward-backward detection sensitivity of the ePIC detector at the EIC.

Collider analyses combine rigorous statistical techniques with state-of-the-art Machine Learning models. However, when the latter are used directly to estimate the likelihood function of the background, hard to quantify systematic effects may bias the estimation of the relevant signal parameters. To address this problem, we present the Minimum Resolution Likelihood (MRL) method, which defines a Fiducial Signal Region that effectively turns the systematic effects into statistical uncertainties. We show with examples that the resulting signal strength estimation is either unbiased or consistent with zero. We consider both toy examples and a realistic application based on the HI-SIGMA technique applied to di-Higgs searches.

Using new methods for computing helicity amplitudes with intermediate helicity-polarized gauge bosons, we revisit the transverse-longitudinal polarization interference in the $pp\to V+{\rm jets}$ process for $V=γ^*,Z^{(*)},W^{(*)}$ decaying to massless leptons. At each order of the strong coupling constant $α_s$ and remaining exclusive with respect to jet kinematics, we show that the polarization interference in $γ^*\to\ell^+\ell^-$ vanishes after phase-space integration over the kinematics of $\ell^\pm$, thereby extending well-known results for the inclusive process. Due to parity violation, cancellations are softened for the $W$ and $Z$ bosons. We give a simple formula to account for fiducial cuts. We comment on the implications for multiboson processes, and the applicability of our results to chiral gauge bosons in new physics scenarios and to polarization measurements of weak bosons in heavy ion collisions.

The Monte Carlo (MC) replica method quantifies parameter uncertainties in global fits of parton distribution functions (PDFs) and Standard Model Effective Field Theory (SMEFT) Wilson coefficients by fitting a model to many noise-perturbed copies of the data and taking the empirical distribution of the best-fit parameters as the uncertainty. The method reproduces the Bayesian posterior exactly only when the model is linear in its parameters, and departs from it in the nonlinear case. We derive the leading-order distribution the method produces and compare it with the Laplace approximation of the Bayesian posterior: the two differ by a single computable matrix, the residual-weighted Hessian of the model at the best fit, whose sign and magnitude set the over- or under-estimation of the parameter uncertainties. This closed-form expression quantifies when and by how much the MC method departs from Bayesian inference. We illustrate it on two single-parameter examples solvable in closed form and point to its evaluation in full PDF and SMEFT fits as a natural next step.

Next-to-next-to-leading order (NNLO) QCD corrections to quarkonium production and decay are known to exhibit perturbative instabilities within non-relativistic QCD. We identify the origin of this problem and propose a simple remedy. Applying our approach to $S$-wave color-singlet quarkonium processes, we achieve substantially improved perturbative convergence and agreement with experimental data.

We consider rules for modifying holographic tensor networks proposed in two independent contexts: by the Colorado (CO) group in 2503.09681 to incorporate observers in holographic maps, and by Kaya-Rath-Ritchie (KRR) in 2506.10064 to derive the Bousso-Penington generalized entanglement wedge proposal. Interestingly, these two sets of tensor network rules are exactly equivalent. This suggests a more general connection between these Abdalla-Antonini-Iliesiu-Levine (AAIL) inspired observer rules and generalized entanglement wedges. To pursue this connection, we first use KRR's analogous rules for the gravitational path integral (based on fixed geometry states) to generalize AAIL's path integral rules to include observers occupying a bulk subregion. Additionally, we leverage the connection in the opposite direction by using the AAIL rules to derive the Bousso-Penington proposal for pointlike bulk regions in JT gravity.

Recent advances in numerical and machine-learning methods have enabled highly accurate constructions of Ricci-flat metrics on compact Calabi-Yau three-folds. For phenomenological applications it is crucial to understand how these metrics vary across moduli space. In this work, we construct approximate analytic expressions for Ricci-flat Calabi-Yau metrics with explicit complex-structure and Kähler moduli dependence by combining machine-learned numerical data with symbolic regression. Our approach is based on an explicit Ansatz for the Kähler potential with moduli-dependent coefficients. Fitting this Ansatz to numerical data and applying symbolic regression allows us to reconstruct analytic formulae for these coefficients, thereby obtaining approximate Ricci-flat metrics with explicit moduli dependence. We apply the construction to a one-parameter family of bi-cubic three-folds in $\mathbb{P}^2 \times \mathbb{P}^2$, achieving percent-level agreement with the underlying numerical data.

We identify a geometric pairing mechanism in systems with spontaneously broken spacetime symmetries, whereby pairs of Goldstone fields become canonically conjugate and hybridize into a single type-B Nambu--Goldstone mode (NGM). The mechanism originates from deformations of the local spacetime volume element induced by Goldstone fluctuations, which endow the Goldstone manifold with a Berry curvature. Integrating this pairing with existing reductions, we propose a general counting formula for the number of gapless NGMs in many-body systems. We demonstrate the mechanism in a microscopic model of a two-component Bose--Einstein condensate, where the dilaton and the $U(1)$ Goldstone field combine into a single NGM with quadratic dispersion.

Quadratically coupled ultralight scalar dark matter behaves as a coherent classical field whose interactions with matter can induce a composition-dependent force through the dark matter background. We present a complete calculation of this background-induced force beyond the spherically symmetric approximation. Using a partial-wave treatment of dark-matter scattering, we determine its angular dependence and derive an analytic description valid even when the dark-matter wavelength is much smaller than the Earth's radius. We show for the first time that Earth screening generates a characteristic frequency-band structure, splitting the signal into multiple sidebands that provide a distinctive experimental signature. We further show that the relative amplitudes of these sidebands vary annually due to the Earth's motion through the dark-matter halo, enabling the construction of a complete signal template. As an application of these results, we re-evaluate constraints from the MICROSCOPE mission, which currently provides the strongest laboratory limits on equivalence-principle violations from ultralight dark matter. We further show that proposed space-based equivalence-principle experiments, such as Galileo Galilei and STE-QUEST, can significantly enhance their sensitivity to ultralight scalar dark matter by incorporating the full frequency-band information.

We set the first constraints on a small-scale dark acoustic oscillation (DAO) in the linear matter power spectrum arising from dark sector interactions, with a full forward model of the Ly-$α$ forest. No more than 30\% of dark matter can form DAOs if they peak at wavenumbers $< 50\,h\,\mathrm{Mpc}^{-1}$ (95\% c.l.), probing scales $25 \times$ smaller than the cosmic microwave background (CMB). Given the complex covariance of DAO and nuisance parameters, we use a deep kernel learning emulator of hydrodynamical simulations to capture imprints of linear oscillations in the Ly-$α$ forest.

Light-front wave functions play a central role in the program of understanding the structure of hadrons as QCD bound states. Using continuum Schwinger methods, based on Dyson-Schwinger and Bethe-Salpeter equations, they can be computed directly within a framework connected to QCD. For light pseudoscalar mesons, previous studies revealed an approximate separability of longitudinal and transverse lightcone momentum dependences in the LFWFs, leading to a simple relation between distribution functions and amplitudes. In this work, we extend those previous studies to the case of pseudoscalar charmonium and bottomonium, using the fictitious $π_s$ meson as a benchmark. Motivated by the observed deviations, we propose a modified non-separable LFWF ansatz that successfully reproduces the properties of heavy pseudoscalar quarkonia and allows the calculation of zero-skewness generalised parton distribution functions, electromagnetic and gravitational form factors, and transverse charge and mass distributions.

We point out that the masses of the highly excited bound quark-antiquark states in QCD$_2$ in the infinite $N$ limit may be determined in the framework of a simple quantum-mechanical model with the potential $V(x) = σ|x|$. In the ultrarelativistic case, the masses follow the pattern $$ μ_n^2 \ =\ \frac {g^2 N}{2π} n, $$ which coincides with the law derived by `t Hooft by solving the Bethe--Salpeter equation. In constrast to what follows from the exact analysis, the levels of the relativistic Hamiltonian have finite widths, but these widths are exponentially suppressed, $Γ_n \propto \exp\{-2πm^2/σ\}$ for large quark masses. In the nonrelativistic case, the levels follow the asymptotics $ε_n = μ_n - 2m = Cn^{2/3}$, where the constant $C$ can be determined by solving the `t Hooft equation or alternatively the nonrelativistic Schrödinger equation.

Investigating the non-perturbative behavior of QCD and the dynamics of strong interaction is crucial for the study of heavy quarkonia and the understanding of exotic fully-heavy tetraquarks. In this work, using the analytical exact iteration method (AEIM), the analytical eigenvalue solutions of the non-relativistic Schrödinger equation are obtained in the presence of topological defects and external magnetic flux. The interactions are modelled using a modified Cornell potential supplemented by harmonic and inverse quadratic terms. We demonstrate that the energy levels are distinctly shifted by the topological defect parameter ($α$). The mass spectra of heavy quarkonia ($c\bar{c}$ and $b\bar{b}$) and fully-heavy tetraquarks ($cc\bar{c}\bar{c}$ and $bb\bar{b}\bar{b}$) across several radial and orbital excitation states are successfully calculated using this approach. The computed masses of bottomonium and charmonium accord well with current theoretical predictions and experimental findings. Our findings for the heavy tetraquarks are in line with previous theoretical investigations that consider tetraquarks as configurations of diquarks and antidiquarks. The numerical results demonstrate that a nontrivial interaction between the confining potential and the background space-time geometry governs the mass hierarchy of these exotic hadronic states, providing high-precision data with excellent agreement with established theoretical models and experimental benchmarks.

Galactic binaries driven by stellar wind shocks, such as colliding wind binaries (CWBs) and gamma-ray binaries (gBs), harbor one of the most efficient particle acceleration engines known in the Universe. Despite their potential, these sources remain relatively unexplored, particularly in the domains of low radio frequencies and very high resolution. As a result, we lack comprehensive population studies and well-characterized individual systems. Only a few of these binaries, such as the iconic gB PSR B1259$-$63 or the massive CWB WR 140, have been studied in enough detail to probe their wind dynamics and shock physics. Current observations lack the sensitivity to detect weak non-thermal synchrotron emission from low-energy particle populations and the angular resolution to resolve shock structures on sub-au scales. The Square Kilometre Array Observatory (SKAO) will mark a significant improvement in both sensitivity and resolution with its SKA-low and SKA-mid telescopes, solving these challenges. This will enable systematic studies of the winds and shock interactions in these binary systems. Additionally, SKA-VLBI will facilitate the observation of changes in shock geometry at different orbital phases, linking particle acceleration processes to the binary's orbital characteristics and stellar wind properties. SKAO will pave the way for comprehensive population studies of these energetic binary systems.

The accretion-ejection mechanism in Active Galactic Nuclei (AGN) remains a central open problem in astrophysics, tied to the role of AGN feedback in galaxy formation and evolution. Radio-quiet AGN dominate the observed AGN population. Lacking luminous jets, their radio emission traces a rich set of processes spanning the host galaxy kpc scales down to the vicinity of the supermassive black hole: star formation, AGN-driven winds and shocks, free-free emission from photo-ionized gas, low-power jets, and coronal activity close to the inner accretion disk. The Square Kilometre Array (SKA) will probe these processes across a wide frequency range with unprecedented sensitivity, wide-field survey capability, and, critically, high-resolution VLBI imaging. Flux, spectral, and polarization monitoring will constrain dynamics and environmental coupling, while mapping nuclear regions on sub-pc to kpc scales will disentangle compact cores from host emission, resolving the diversity of radio activity across accretion regimes and jet powers from the local Universe to the cosmic dawn. At the full AA4 deployment, the SKA-MID phased into global VLBI arrays will deliver sub-milliarcsecond imaging and $μ$Jy sensitivity over 0.35--15\,GHz, enabling the first population-level census of radio-quiet AGN nuclei. Earlier AA$\ast$ operations will support pilot studies of the brightest nearby systems.

Context: The stability of mass transfer through Roche-lobe overflow plays a key role in shaping the outcome of binary interactions. However, the criterion for mass transfer stability remains one of the main open questions in the theory of binary evolution. Aims: We develop a mass transfer stability prescription that accounts for mass and angular momentum loss, and implement it in the population synthesis code SEVN. We assess its impact relative to the standard formalism used in SEVN, using blue stragglers and binary compact objects as illustrative cases. Methods: We derive an expression for the response of the Roche-lobe radius to mass loss in the general case where the mass and angular momentum of the system are not conserved. On the basis of this formulation, we construct a new mass transfer stability criterion that modifies the standard approach only through the Roche-lobe response term. Results: Population synthesis simulations with SEVN show that the new criterion allows stable mass transfer in binaries with higher donor-to-accretor mass ratios, leading to an overall increase in the predicted number of blue stragglers and promoting their formation in wider orbits. This contributes to reconciling the differences between theory and observations. For binary compact objects, the impact of the new stability criterion varies across system types, with the strongest effects occurring in binaries containing at least one neutron star. In particular, for low mass transfer efficiency, the new criterion enhances the contribution of channels involving stable mass transfer and leads to a larger number of systems, including gravitational wave progenitors. Conclusion: The inclusion of a new, simple, yet more consistent prescription for mass transfer stability has proven that refining this criterion can significantly improve our understanding of the formation channels of specific stellar populations.

We present our search for little red dots (LRDs) in the "J1030 field", a region of the sky around the $z\sim 6.3$ quasar SDSS J1030+0524, observed by the JWST EIGER program. Over 154 point-like sources selected in a JWST-based photometric catalog, we find five broad line emitters (with $FWHM \gtrsim 1000\ \rm km s^{-1}$) that are red ($F200W - F356W > 0$) and are undetected in the X-rays. We use these sources to derive the bolometric luminosity function (LF) of LRDs at $z = 2.4$ and $z = 4.5$. At $z = 2.4$, the space density of LRDs is only a factor of $\sim 2$ lower than that of all pre-JWST active galactic nuclei (AGNs) with bolometric luminosity $L_{\rm bol} \gtrsim 3 \times 10^{44}\ \rm erg\ s^{-1}$. At $z = 4.5$, our estimate is consistent with those derived for LRDs based on larger areas of the sky. A similar behaviour is observed in the black hole mass function. More importantly, we study the number density of LRDs from cosmic dawn to cosmic noon. We find that there is no significant evolution in the abundance of LRDs with $L_{\rm bol} \gtrsim 3 \times 10^{44}\ \rm erg\ s^{-1}$ at $z > 2$. We speculate that the drop at $z < 4$ seen by other studies is due to their sampling of only the bright-end of the LRDs LF. At cosmic noon, the abundance of LRDs is $n = 3.4^{+5.6}_{-2.4} \times 10^{-5}\ \rm Mpc^{-3}$, which is a factor of $\sim 350$ larger than recent model predictions and is comparable with that of X-ray selected AGNs with similar bolometric luminosity. Our result may imply that, if LRDs are the early, rapid stages of supermassive black hole growth, as suggested by some models, then the formation of black hole seeds can be efficient down to epochs as recent as cosmic noon. Alternatively, LRDs may simply be a high-accretion phase in already mature black holes.

Line intensity mapping (LIM) has emerged as a powerful tool for surveying the large-scale structure of the Universe across cosmic time by measuring spatial fluctuations in the cumulative emission of spectral lines from unresolved sources or the intergalactic medium. Besides the most abundant 21-cm hyperfine line of neutral hydrogen, there are bright far-infrared fine-structure lines like [CII] 158 $μ$m, [OIII] 88 $μ$m, [NII] 122/205 $μ$m, and [OI] 63 $μ$m, as well as mid-/high-$J$ CO rotational transitions, hydrogen Ly$α$ and H$α$ as potential LIM probes. A key opportunity lies in combining and cross-correlating 21-cm intensity maps from SKAO with other line intensity maps, targeted by a range of ongoing and forthcoming LIM experiments that probe overlapping cosmic volumes. Cross-correlation between 21-cm maps and other line tracers mitigates uncorrelated systematics and enhances sensitivity to the underlying matter distribution, while multi-line analyses help disentangle cosmological and astrophysical parameters. Beyond cross-power spectra, higher-order and morphological statistics -- such as cross-bispectra, marked correlations, and morphological measures -- capture non-Gaussian features and the environmental dependence of structure formation. This chapter explores the synergies that can be achieved by combining SKAO observations with other line-intensity mapping experiments, demonstrating how such joint analyses can unlock new insights into galaxy evolution and cosmology.

The cosmological distribution of neutral hydrogen (HI) during the post-reionization era is highly non-Gaussian due to the underlying non-linear structure formation, complex galaxy biasing, and potential primordial non-Gaussianity. One needs higher-order (beyond two-point) statistics to maximally extract the non-Gaussian information out of the 21-cm intensity maps. This chapter summarizes the potential of several higher-order statistics, including voxel intensity distribution, emission line stacking, probability density functions, $\ell_1$-norm, bispectrum, and various marked statistics. Additionally, image-based morphological descriptors, such as the Largest Cluster Statistic, local dimensions, and Minkowski functionals, etc., can potentially characterize the morphology and geometry of the cosmic web encoded in the 21-cm intensity maps. This chapter presents forecasts of the detectability of these higher-order statistics in the context of the future SKAO observations. These forecasts incorporate instrumental noise, observational effects, and, in some cases, foreground removal in their analyses. With its unprecedented sensitivity, the future SKAO 21-cm observations will enable us to measure these higher-order statistics more precisely, possibly helping to break degeneracies between astrophysical and cosmological parameters, and maximizing the science outcome from these surveys.

The Living Swift-XRT Point Source Catalogue (LSXPS) enables near real-time searches for X-ray transients. Many detected candidates are faint, often near the XRT detection limit, and are classed as "low significance," as it is often unclear whether their apparent brightening reflects a genuine transient or a statistical fluctuation. Some of these sources are affected by Eddington bias, a statistical effect that inflates measured fluxes near the detection threshold. We present a simulation-based Bayesian framework that corrects for this bias and provides more accurate probabilities for each source being truly transient, i.e. that its true intensity exceeds the historical 3$σ$ upper limit. Applied to LSXPS data, this method yields more reliable classifications, recovering over 500 transients above this threshold -- more than an eight-fold increase over the original confirmed sample. Using extensive simulations based on real Swift-XRT images, we validate the robustness of this approach, showing that it remains stable across varying exposure times and background conditions. These results demonstrate that the LSXPS transient probabilities, corrected for Eddington bias, provide a reliable and internally consistent framework for real-time X-ray transient identification.

Orion is the closest region hosting active star formation and young OBA stars. Computing far-ultraviolet (FUV) fluxes at its stars is essential to connect stellar and protoplanetary disc properties to the environment. We (1) accurately estimated the FUV flux at a large sample of stars in Orion by statistically accounting for the uncertainty in parallax measurements, and (2) investigated the relation between stellar accretion and external FUV flux by comparing observations and disc evolution models. We selected a large stellar population in Orion, assigned sub-cluster memberships and used 2D dimensional sub-cluster geometry to infer 3D separations from OBA stars and compute the FUV flux at stellar positions. We studied the accretion luminosities Lacc inferred from Ha emission in Gaia XP spectra of Orion sources and determined their detection fraction as a function of age and FUV flux. We compared the results with population synthesis models of viscous discs experiencing external photoevaporation. We provided a publicly available table of FUV fluxes at ~8600 stars in Orion. Most of the stellar population is weakly irradiated <10^{2} G0, ~35% is intermediately irradiated 10^{2}-10^{4} G0, and ~5% has FUV fluxes >10^{4} G0. Gaia-based Lacc decreases with age, and Ha detection fraction declines more rapidly in regions with strong FUV fluxes than in regions exposed to weaker FUV fluxes, broadly consistent with the model. This may suggest that external photoevaporation efficiently depletes strongly FUV-irradiated accretion discs, but it is not sufficient to reliably confirm this conclusion. The provided tools for computing FUV fluxes at Orion stars will be essential for future observations aimed at assessing the role of external photoevaporation on discs. We encourage measurements of stellar and disc properties in Orion, covering FUV fluxes 1-10^5 G0.

The line spread function (LSF) of a spectrograph encodes the inherent broadening of a single spectral line. It is typically reported as a single number, the resolving power $R = λ/Δλ$ with $Δλ$ the FWHM of the LSF. In standard pipelines for extracting stellar kinematics the LSF is assumed to be a wavelength dependent Gaussian. However, detailed LSF measurements from real integral field spectrographs reveal a variety of shapes, some close to Gaussian, others with large wings or that appear boxy. I have studied the impact that these non-Gaussian LSF profiles have on the recovery of the stellar kinematics of a mock spectrum and find that even in the high dispersion case of 300 km s$^{-1}$, there is up to a 7 percent uncertainty in the dispersion due to non-Gaussian LSF profiles. Additionally, higher order Gauss-Hermite moments $h_3$ and $h_4$ can be biased by up to $\pm$0.1. To resolve this bias, I developed a method to match the LSF of the template spectra to the LSF of a target spectrum when the LSF of either one or both is non-Gaussian and show that it can reduce bias in the dispersion to less than a percent down to the instrumental resolution. A Python implementation of this method has been made publicly available.

Dust polarization observations are widely used to diagnose the relative importance of magnetic fields and turbulence in star forming molecular clouds, often through summary statistics such as the mean polarization direction $μ$ and dispersion $σ$. Recent multi-scale polarization observations of the Orion Integral-Shaped Filament (ISF) reveal substantial diversity in polarization morphology among its dense cores, raising questions about the underlying Alfvénic nature of the cloud. In this work, we develop a statistical framework to compare polarization-based summary statistics from observations with those derived from projected three dimensional MHD simulations, explicitly accounting for projection effects. Using globally sub-Alfvénic simulations that naturally produce slightly super-Alfvénic dense cores, we show that modest deviations of core-scale magnetic fields from the parent cloud field, when combined with projection, can generate a wide range of plane-of-sky polarization dispersions. Applying hypothesis testing, we find that the observed $(μ, σ)$ values in the Orion ISF are statistically consistent with sub-Alfvénic cloud models over a broad range of viewing angles. This broad degeneracy implies that $μ$ and $σ$ alone cannot provide precise information about the three-dimensional magnetic-field distribution, and hence the Alfvén Mach number, of an individual cloud. While the observations can provide evidence against certain projection geometries, we demonstrate that polarization statistics based solely on $(μ, σ)$ are insufficient to provide evidence against sub-Alfvénic cloud models. Our results highlight the necessity of explicitly incorporating projection effects when interpreting polarization observations of molecular clouds.

A primary question within galaxy evolution is how active galactic nuclei (AGN) feedback modifies the circumgalactic medium (CGM). We present a search for faint H$α$ emission from the cool ionized CGM ($T\sim 10^4$ K) around radio galaxies by stacking background-quasar spectra from DESI sightlines. We take into account the projected distance and position angle of each quasar sightline relative to the radio jet axis, and test whether jet--CGM coupling is anisotropic. We detect a strong H$α$ excess at $>5σ$ along the collimated radio jet axis ($θ<20^\circ$) with a mean integrated flux of $1.19\times10^{-17}\ {\rm erg\ cm^{-2}\ s^{-1}}$. In contrast, the azimuthally averaged stack over all 324 sightline angles yields no detection ($<2σ$), indicating that this excess emission is very localized along the radio jet. We also find that the jet-aligned H$α$ signal is radially structured, where the strongest emission occurs near the host galaxy just outside the optical half-light radius, and rising again near the projected radio-lobe region. The jet-aligned stacks reveal H$α$ signal that is roughly 100 times brighter than normal halos. In the same sightlines however, Mg II absorption shows no difference in incidence between jet-aligned and off-axis directions, with broadly similar equivalent widths, column densities, and line widths. This striking contrast shows that while Mg II traces the ambient, clumpy cool CGM reservoir, the H$α$ emission directly captures localized, low-covering-fraction clouds whose density, pressure, or ionization level has been dramatically boosted by the propagating jet. These results deliver clear evidence of localized jet-CGM interaction in radio-jetted AGNs.

Our $ψ$DM simulations predict that stars diffuse throughout dark matter halos over the Hubble time through a random walk driven by the wave perturbations intrinsic to $ψ$DM. The resulting stellar distribution locally follows a Gaussian profile (Sersic index $n=0.5$), as expected from the central limit theorem, expanding as $R_{1/2}(t)\simeq(\hbar/m_ψ)^{0.5}\sqrt{t}$, in good agreement with the core--halo profiles of typical $ψ$DM dwarf spheroidal galaxies. The strength of this diffusion depends on halo mass and the corresponding soliton, naturally producing progressively more diffuse stellar systems in more massive halos. The observed continuity from faint dwarfs and compact dwarf spheroidals to ultra-diffuse galaxies can therefore be interpreted as an age sequence, with later-forming dwarfs experiencing less diffusion and thus remaining smaller and brighter. Stellar scattering arises from the random walk of the soliton, gradually transporting stars from the dense central core into the outer halo, creating the extended stellar envelopes observed around Local Group dwarfs. Rather than being unique to Ultra-Diffuse Galaxies (UDGs), this wave-driven stellar diffusion may provide a unified mechanism explaining galaxy structure across a vast mass range, from ultra-faint dwarfs to the most massive UDGs, without requiring distinct formation channels or "failed galaxy" scenarios. The diffuse stellar halos and globular cluster distributions recently revealed by Euclid may therefore represent direct observational signatures of the granular dynamics of wave dark matter.

We show the known pathological character of $f(\mathbb{Q})$ gravity in its scalar-tensor representation.

In order to understand the magnetization of galaxies and the role of magnetic fields in feedback processes that govern star formation and galaxy evolution, it is essential to have a comprehensive census of magnetic fields in and around galaxies from the nearby Universe to high redshifts. In this chapter, we outline the science goals, strategies, techniques, and observational requirements with SKA AA4 for (1) a homogeneous polarimetric survey of nearby galaxies - mapping both the diffuse polarized emission as well as producing a dense RM grid within the virial radius; (2) a survey of the interstellar magnetic fields in distant galaxies targeting strong lensing systems with polarized lensed quasars, as well as a general statistical back-illumination survey to probe the redshift evolution of magnetic fields in the CGM. These proposed observations will serve as a major step towards understanding the co-evolution of galaxies and their magnetic fields over cosmic time and provide constraints on galactic dynamo theories.

Most XRISM galaxy cluster observations to date have focused on AGN feedback or actively merging systems. The weak cool-core cluster A3571 was observed in four XRISM Cycle 1 pointings, enabling the study of gas kinematics in a relaxed, AGN-feedback-free system. We present measurements of the velocity dispersion and bulk velocity in the core regions of A3571, out to $120$ kpc. The velocity dispersion is relatively uniform across all regions ($\sim100-120 ~\mathrm{km~s^{-1}}$), except in the northern gas sloshing elongation, where a $68\%$ upper limit of $68~\mathrm{km~s^{-1}}$ is obtained. The core Mach number and non-thermal pressure fraction of A3571 are lower than in the extremely relaxed cluster A2029 and below predictions from cosmological simulation suites. Despite relatively low velocity dispersion values, the derived turbulent heating rate is sufficient to offset cooling losses in all studied regions. This suggests that sloshing motions contribute significantly to the heating budget. Comparing XRISM observations of merging and relaxed clusters, we find that mergers exhibit an average Mach number of $0.29\pm0.07$, nearly twice that of the relaxed sample, which is consistent with predictions from non-radiative cosmological simulations. A3571 is a promising target for resonant scattering studies; however, simulations indicate that deeper observations are required to obtain reliable turbulent velocities via the $z/w$ line ratio.

Blue horizontal-branch (BHB) stars are low-mass, core helium-burning objects with nearly constant luminosities, making them powerful tracers of old, metal-poor populations and valuable standard candles for mapping the Galactic halo. However, robustly identifying BHB stars from low-resolution spectra remains challenging. We present MSFA-Net, a two-stage framework developed for the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) DR12. By combining multi-scale convolutions with a soft frequency attention mechanism, MSFA-Net learns discriminative representations in both the wavelength domain and the Fourier-frequency domain. On the test set, the framework achieves a precision of 94.67% in the initial multiclass screening and 98.07% in the subsequent binary refinement. Applying the pipeline to LAMOST DR12, we retrieve 27,853 BHB candidate spectra. After spectral deduplication and removal of previously known objects, we identify 3583 new BHB stars, confirmed via Balmer-line profile fitting. We further estimate atmospheric parameters (Teff, log g, and [Fe/H]) using the machine-learning-based SLAM model and examine their distributions. A non-negligible subset shows unusually high log g and/or metallicities, which we interpret primarily as inference-related systematics rather than intrinsic properties. Photometric cross-matching with Gaia DR3 and color-magnitude diagrams provide an additional consistency check for the sample. The resulting catalog substantially enlarges the spectroscopically confirmed BHB sample from LAMOST and offers a homogeneous data set for studies of Galactic-halo structure and stellar populations.

In this chapter, we provide a comprehensive overview of the astrophysical and cosmological processes that shape the 21-cm signal during Cosmic Dawn and the Epoch of Reionization. We investigate both standard and exotic signatures potentially observable with SKA-Low. Standard signatures are those expected within the $Λ$CDM framework, including contributions from the first stars, galaxies, and black holes. Exotic signatures are more speculative indicating new physics, such as primordial black holes, modifications to the dark matter sector, non-standard primordial fluctuations, or strongly emitting radio galaxies. The effects of these different sources or scenarios are evaluated in the context of the expected sensitivity of SKA-Low, considering the AA* and AA4 configurations. The chapter aims to provide an overview of the theoretical landscape of 21-cm signatures and to highlight how the forthcoming SKA-Low observations will improve our understanding of astrophysical processes at early times and may open the door towards new physics beyond the $Λ$CDM framework.

Interstellar dust is a crucial asset in many astronomical observations. Characterising grains present in the dense gas and in star-forming environments is also key to constrain the pristine conditions for planetary formation. However, dust properties remain poorly characterised and are still debated: low dust emissivities observed in nearby protostars are not completely explained to this day. In this study, we aim to determine whether it is possible to retrieve the dust properties from multi-wavelength observations of the dust emission towards embedded protostars, and the extent to which current dust models can reproduce the observed values of the dust emissivity index in young protostars. We perform radiative transfer computations of the thermal dust emission from a model protostellar envelope, considering different dust optical properties commonly used in the community. This allows us to explore the effects of dust composition on the spectral index, to try and explain the variation in the emissivity index in nearby protostars observations. We find large variations in the spectral index as the sole result of different dust models, without the need for dust grain size evolution. However, our work does not allow us to reproduce the lowest emissivity index values found in some protostellar envelopes without including unexpectedly large millimetre-sized processed grains. We show that appropriate methods permits to measure the dust emissivity from observations of the spectral index at millimetre wavelengths with very little uncertainty. Variation in emissivity index between the different observed sources and the dust models most commonly used by the community implies that the intrinsic composition of dust is not sufficient to explain the lowest spectral index values. Thus, early dust evolution producing larger dust grains may have to be taken into account to obtain a complete picture.

We introduce a method for visual and auditory feedback when exploring the fit of a model to data. Starting with a best-fit curve fit to data, the user can drag the curve to a new position and the computer will emit a squeal, becoming louder and more unpleasant as the discrepancy between curve and data increases. We demonstrate with four examples: a two-parameter curve fit to golf putting data, a four-parameter curve fit to dilution assays, a fit to cosmological data sensitive to the parameters of the Big Bang model, and a nonparametric Gaussian process fit to temperature readings.

Recently, the LHAASO collaboration reported ultra-high-energy (UHE) gamma rays from six microquasars. For five of these sources, the emission extends beyond $100$ TeV, making microquasars promising candidates for Galactic PeVatrons. We investigate whether gamma-rays around $100$ TeV originate from hadronic interactions of accelerated cosmic rays (CRs) with the ambient medium, and we estimate the contribution of these sources to the measured CR spectrum around the knee. We also place upper limits on six LHAASO microquasars with no detected UHE emission. We assume diffusion-dominated propagation of CR, with a diffusion coefficient suppressed compared to the average Galactic value near the source and equal to the Galactic value at large distances. We assume continuous injection over timescales of $t_{\rm age}=0.1-1$ Myrs. Using available measurements of the gas density, we find that hadronic interactions alone cannot fully account for the observed emission for any of the detected sources. However, in the case of GRS 1915+105 and MAXI J1820+070, the hadronic scenario may still be valid when considering acceleration efficiency higher than $10\%$. We then derive upper limits on the hadronic contribution to the observed gamma-ray flux. We estimate that the detected sources contribute at most $\sim4\%$ of the Galactic CR spectrum at $1$ PeV for an injection timescales of $0.1$ Myrs and a CR acceleration efficiency of $10 \%$. When adopting the maximum acceleration efficiency allowed by the gamma-ray observations the contribution rises to $37\%$. Longer injection timescales ($\sim 1$ Myrs) lead to contributions exceeding the observational constraints. For sources not detected at UHE, we obtained a maximum contribution of $\sim 17\%$, achieved assuming continuous injection over $1$ Myr.

We present Scintillation Pipeline for Interferometric Candidate Extraction (SPICE) an automated CASA-based pipeline developed to identify pulsar candidates in Giant Metrewave Radio Telescope (GMRT) and upgraded GMRT (uGMRT) data through their diffractive interstellar scintillation signatures. SPICE integrates flagging, calibration, imaging, and classification, with robust RFI excision, iterative self-calibration with dynamic reference antenna selection, source detection using PyBDSF, and classification based on our earlier development of scintillation-based visibility correlation searches. SPICE is available publicly on github and is archived on Zenodo. We applied SPICE to archival datasets from both legacy GMRT and uGMRT. The pipeline successfully recovered known pulsars such as PSR 0437-4715, PSR B0450-18, and PSR B0329+54, yielding scintillation parameters consistent with expectations. Non-detections in some scans highlight the influence of pervasive RFI, the dependence on the reference antenna, and the intrinsic variability of the scintillation properties. SPICE complements time-domain searches by enabling reproducible scintillation-based candidate identification in interferometric data. Its application to the GMRT archive opens a pathway for discovering compact variable sources and expanding pulsar searches beyond time-domain searches.

Over the past two decades, internet-enabled citizen science research (CSR) has contributed to significant discoveries while involving millions of people in the research process. Our review highlights CSR in extragalactic radio astronomy and emphasises that such approaches will become increasingly relevant across radio astronomy in the era of the Square Kilometre Array (SKA). As astronomical data volumes grow, CSR is converging with Artificial Intelligence and Machine Learning (AI/ML), creating hybrid human-machine frameworks suited to big-data challenges. Two CSR platforms, Radio Galaxy Zoo and RAD@home, demonstrate success: the former excels in large-scale, web-based catalogue creation, while the latter combines structured training with collaborative discovery. Following this, we propose CSR with the SKA, namely SKA@home, with two modes: one purely web-based and the other in collaboratory mode with national training programmes. We argue that CSR can complement, and at times surpass, automated AI/ML pipelines, particularly in identifying rare, intricate, or unexpected features. Illustrative CSR discoveries include an episodic wide-angle-tailed radio galaxy, a jet-galaxy interaction, a collimated synchrotron thread, a twin-ring odd radio circle, and a large-scale shock ahead of a cluster-infalling galaxy. Consistent with the IAU's recognition of CSR as a driver of Astronomy for Development and the United Nations' affirmation of participation in science as a universal human right, both the SKA construction proposal and outreach strategies show commitment to enabling CSR with SKA. The proposed SKA@home would not only enhance the early discovery potential of SKA data but also initiate a deeper and more meaningful connection with society at large.

We present a hierarchical Bayesian analysis of the joint mass, spin, and orbital properties of $\sim 50$ Galactic binary radio pulsars with measured neutron star masses, classified by binary type into pulsar-white dwarf (PSR-WD) and double neutron star (DNS) systems. We find moderate evidence for an anti-correlation between neutron star mass and spin period in the pooled recycled population (correlation coefficient $ρ= -0.26$, with $96\%$ of the posterior probability at $ρ<0$; the $90\%$ credible interval excludes zero), robust to the treatment of candidate DNSs and to a radio-detectability selection correction. Although consistent with accretion-driven recycling, the correlation cannot statistically distinguish an accretion origin from a moment of inertia-driven spin-up mechanism, because the neutron star moment of inertia is nearly linear in mass over the observed range. The DNS systems alone instead lean to the positive side expected from the moment-of-inertia mechanism ($ρ=+0.13$), though with only ten systems this is not statistically conclusive. Mass shows no significant correlation with orbital period or inclination, and only a weak one with eccentricity. As a secondary result, neutron stars with helium white dwarf companions are marginally more massive than those with carbon-oxygen/oxygen-neon white dwarf companions ($Δ\simeq 0.06\,M_\odot$), consistent with more extensive accretion in the helium white dwarf channel. We confirm, in a hierarchical framework, the previously reported correlation between companion mass and orbital eccentricity in double neutron stars ($ρ=+0.82$). We interpret these results within a two-channel picture -- accretion-grown PSR-WD versus birth-mass-dominated DNS.

We present ALMA CO($J=1-0$) observations of a nearby radio galaxy NGC1316 at a 100-pc resolution to investigate the impact of AGN jets on the molecular gas. The molecular gas exhibits complex spatial and kinematic distributions, with broad CO line widths ($>50$ km s$^{-1}$) observed in several regions. The interferometric CO flux is only 34%-38% compared to single-dish data, indicating a large fraction of spatially extended molecular gas, especially in the central regions. We identified 24 Giant Molecular Clouds Associations (GMAs) primarily within the ``NW Shell'' and the ``SE Blob''; these GMAs show velocity dispersions approximately twice as high as those in typical star-forming galaxies for their sizes. Analysis of archival ALMA CO($J=2-1$) and CO($J=3-2$) data reveals elevated line ratios ($R_{21} \sim 1$ and $R_{31} \sim 1$) in gas near the jet, whereas, away from the jet, typical values ($R_{21} \sim 0.7$, $R_{31} \sim 0.3$). A multi-wavelength comparison reveals a $\sim$5 kpc warm ionized gas shell that encompasses the molecular NW Shell. The observed energetics and bubble morphology are consistent with an expanding bubble model driven by the jet assuming a jet power of $1.6\times10^{43}$~erg~s$^{-1}$. We propose that the high extended gas fraction results from the destruction of molecular clouds due to interactions with the jet plasma. NGC1316 may be a good example of jet-induced negative feedback through the ablation, dispersal, and rarification of dense molecular clouds through jet-ISM interactions.

The discovery of stripped-envelope supernovae (SNe) interacting with dense circumstellar medium (CSM) challenges our current understanding of massive star evolution. We present late-time observations of the interacting Type Ic SN 2017dio and investigate its mass-loss mechanism and progenitor channel. We analysed late-time spectra and light curves (LCs) that are dominated by ejecta-CSM interaction. We examined the CSM and the source of the infrared (IR) excess by modelling the radiation produced by the ejecta-CSM interaction and the IR echo from circumstellar dust. In addition, we studied the evolution of spectral features, with a particular emphasis on the Halpha emission line. From the combined analysis of the LCs and spectral properties, we infer that the peak mass-loss rate for the CSM reaches ~0.2 $M_{\odot}/yr$ and that the typical value over most epochs is ~0.06 $M_{\odot}/yr$. The nearby CSM was formed over a period of 4 to 65 years before the explosion. The CSM radius begins at ~$1.3\cdot10^{15}$ cm. The IR excess identified in the LCs is consistent with the radiation from dust with a mass increasing from ~0.001 to ~0.005 $M_{\odot}$ in the case of carbon dust or ~0.005 to ~0.02 $M_{\odot}$ in the case of silicate dust. From IR echo modelling, we estimate an upper limit on the dust mass of $4\cdot10^{-5} M_{\odot}$, which implies an SN progenitor mass-loss rate of $2.4\cdot10^{-5} M_{\odot}/yr$ at the dust evaporation radius determined by the SN peak luminosity (0.017 pc for carbon dust, corresponding to mass loss ~170 years before the explosion). This implies a very rapid increase in the mass-loss rate ahead of the explosion. Although the progenitor of SN 2017dio has lost its helium envelope, it interacted with a hydrogen-rich CSM formed shortly before the explosion, suggesting that this material originated from a companion star rather than the progenitor itself. [Abridged]

Measuring the masses of accreting white dwarfs (WDs) is crucial for understanding their evolution and the physics of accretion. High-resolution X-ray spectroscopy can trace the WD motion through Doppler shifts of emission lines formed close to the WD. We report an 83~ks XRISM/Resolve observation of the intermediate polar EX~Hydrae and measure the orbital modulation of individual Fe K-shell line centroids. The Fe~{\sc xxv} K$α$ components show coherent orbital modulation, yielding $K_1 = 58.1 \pm 8.5\ \mathrm{km\ s^{-1}}$. This is the first detection of orbital modulation in individual Fe K-shell lines from an accreting WD, made possible by the high spectral resolution of Resolve and its frequent in-orbit gain calibration. The measured $K_1$ is consistent with optical/UV $K_1$ measurements, providing a cross-check that these distinct tracers follow the WD orbital motion. Combining this X-ray measurement with literature orbital parameters, we derive a WD mass of $M_1 = 0.79 \pm 0.04\ M_\odot$. These results demonstrate that high-resolution X-ray spectroscopy can use individual Fe K-shell line centroids to trace WD orbital motion in accreting WDs.

Galaxy outflows are usually ordered by stellar mass and star-formation rate (SFR), but the same feedback budget may couple differently to gas in diffuse and compact galaxies. We use Dark Energy Spectroscopic Instrument (DESI) Data Release 1 stacked spectra of massive star-forming galaxies at $0.35<z<1.0$ to test whether stellar surface density, \(Σ_\star=M_\star/(2πR_e^2)\), is an independent empirical coordinate of down-the-barrel singly ionized magnesium (\mgii) cool-gas absorption. In AGN-clean samples matched in stellar mass, and in a stricter sample matched in both stellar mass and a Balmer-line SFR proxy, the \mgii\ outflow equivalent width (EW) rises monotonically with \(Σ_\star\) in every redshift bin. From the lowest to highest \(Σ_\star\) tertile, EW$_{\rm out}$ increases by \(0.37\)--\(0.61\)~Å, while the absolute outflow velocity changes only weakly. DESI therefore shows that cool-gas outflow strength in massive star-forming galaxies is not set only by how much stellar mass or star formation a galaxy has, but also by how tightly the galaxy is built. The structural dependence points to changes in the absorbing velocity distribution and/or the effective covering fraction of cool outflowing gas.

We investigate the cosmological evolution of matter perturbations in the modified gravity model $f(R,T^2)$, where $T^2=T_{μν}T^{μν}$ denotes the quadratic contraction of the energy--momentum tensor. Using the gauge-invariant 1+3 covariant formalism, we study the evolution of the matter density contrast and analyze several growth observables, including the growth factor, the growth index, and the weighted growth rate $fσ_8$. We consider representative values $n=1/2$ and $n=1/4$, which probe different regimes of the matter--geometry coupling. We show that the growth index decreases with increasing redshift and approaches the standard matter-dominated behavior at early times, while mild scale-dependent deviations from the $Λ$CDM model emerge at late times. The model predicts small departures from General Relativity for $n=1/4$, whereas stronger deviations appear for $n=1/2$ and larger values of the coupling parameter $α$. We further compare the theoretical predictions for $fσ_8$ with current observational data and find that viable parameter choices remain within the observational $\pm2σ$ bounds. These results indicate that $f(R,T^2)$ gravity can provide a viable description of late-time cosmic acceleration and large-scale structure formation while remaining consistent with current growth observations.

We present a full-shape power-spectrum analysis of the eBOSS DR16 quasar sample with optimal redshift weights. The DR16 QSO catalog contains 343,708 quasars over $0.8<z<2.2$, a redshift interval broad enough to contain useful light-cone evolution but not naturally captured by a single effective-redshift measurement. We construct Karhunen--Loève weights for the parameters of interest and measure the resulting monopole and quadrupole with a cross-correlation estimator, which remains well defined for sign-changing weights. The theoretical spectra are convolved with the measured Fourier-space survey-window kernels for each Galactic cap and weighting scheme, and both the covariance matrix and the end-to-end validation are based on 1000 EZ light-cone mock catalogs. In $Λ$CDM, the redshift-weighted and standard analyses give consistent constraints, as expected from the near-standard effective redshifts of the weights targeting $h$, $Ω_{\rm m}$, and $A_s$. In the Chevallier--Polarski--Linder (CPL) model, the redshift-weighted DR16 analysis reduces the marginalized uncertainties on $H_0$, $σ_8$, and $w_0$ by $43.3\%$, $19.7\%$, and $20.5\%$, respectively, and turns the standard one-sided constraint on $w_a$ into a bounded posterior, $w_a=-0.98^{+1.0}_{-1.3}$. The gain is therefore concentrated where the model contains genuine redshift evolution, demonstrating that optimal redshift weighting can recover tomographic information from a wide QSO light cone while keeping the full-shape data vector compact.

Cosmic rays (CRs) regulate the chemical evolution of the gas and its coupling to the magnetic field in the densest and coldest regions of the interstellar medium (ISM). However, the CR ionisation rate of H$_2$ ($ζ_2$) is one of the most debated parameters characterising molecular clouds due to the uncertainties in its estimation. We developed a new analytical framework based on the chemistry of N$_2$H$^+$, N$_2$D$^+$ and DCO$^+$ to overcome observational limitations in current estimates of $ζ_2$ and to probe the latter and the electron fraction, $x(e)$, in the gas across multiple density regimes. We applied this method towards the parsec-scale filament of the infrared dark cloud (IRDC) G035.39-00.33 with new observations from the NOrthern Extended Array (NOEMA) at a resolution of $3''$ (or $\sim9000$ au). Ancillary observations of C$^{18}$O complete this survey to measure $x(e)$ and $ζ_2$ in G035.39-00.33. CO depletion is widespread in G035.39-00.33 with factors, $f_\mathrm{D}$, positively correlated with column and number densities of H$_2$ in the cloud. The deuterium fractions ($R_\mathrm{D}$) are enhanced towards these same sites in which the corresponding electron fraction values cluster below $\lesssim10^{-8}$. $ζ_2$ varies by three orders of magnitude in G035.39-00.33 ($\sim10^{-18}-10^{-15}$ s$^{-1}$) with a median of $\sim2.3\times10^{-18}$ s$^{-1}$, consistent with those reported for other IRDCs and giant filaments, but on average lower than the typical $ζ_2$ for the ISM. $ζ_2$ shows a functional dependence on $N(\mathrm{H_2})$, but with absolute values lower compared to those predicted by theoretical models. This behaviour suggests the presence of an overall attenuation of the CR flux taking place in G035.39-00.33. The CR flux appears to be reduced by the change in magnetic field strength and morphology previously reported in the region.

Radio-luminous active galactic nuclei, or radio galaxies, are the brightest population of objects in the extragalactic radio sky and will be seen in large numbers in essentially every SKA observation. Despite having been studied for more than seventy years, some aspects of radio galaxy physics are still poorly understood, and the SKA will shed light on this by enabling the generation of very large samples of high-resolution, sensitive, broad-band images of radio galaxies, allowing us to probe, for example, regions of particle acceleration, spectral ageing, and the magnetic field structures both internal and external to the radio lobes. A key feature of the radio galaxy population is that observations of extended sources probe the past history, and thus the duty cycles, of accretion onto the central supermassive black hole, and we discuss ways in which the SKA will improve our understanding of episodic and dying radio galaxies in particular.}

The Jingdong Radio Telescope (JRT) is a 120-meter fully steerable radio telescope currently under construction in Jingdong County, Yunnan Province, China. Located at a relatively low latitude (24.5 degree), the JRT will enable observations of nearly 90% of the sky. Equipped with two broadband single-pixel receivers covering 1-8 GHz and 6-18 GHz, and a powerful digital backend, the telescope will support single-dish studies of various radio sources-particularly millisecond pulsars for enhancing the detection of nanohertz gravitational waves. In addition to single-dish capabilities, the JRT is expected to contribute approximately 800 hours annually to international Very Long Baseline Interferometry (VLBI) observations via a standard VLBI backend. When operating in conjunction with the phased-up SKA-Mid, the JRT will significantly enhance the technical and scientific capabilities of existing VLBI networks. This paper presents a comprehensive overview of the JRT's VLBI module and explores its potential to improve joint VLBI observations with current VLBI networks. Our analysis suggests that coordinated VLBI observations involving both the SKA-Mid and the JRT have the potential to significantly advance the field. For early sciences, we also highlight a few highly promising scientific cases, e.g. measuring the distance to PSR J0437-4715 with <1 ly accuracy and exploring jet formation with an event-horizon-scale resolution in M60*.

Pulsar timing arrays (PTAs) are opening the nanohertz gravitational-wave (GW) band by timing millisecond pulsars (MSPs) to target signals from supermassive black hole binaries (SMBHBs). Beyond evidence for a stochastic background, a central SKA-era objective is detecting individual continuous-wave (CW) sources. The scientific payoff hinges on localization: conventional PTA searches yield uncertainties of tens-hundreds of deg$^2$, too large to identify a unique host, obtain a redshift, infer intrinsic masses, or pursue electromagnetic counterparts. This limitation is chiefly geometric: the CW response includes Earth and pulsar terms, and poorly known pulsar distances make the pulsar-term phase a free parameter that degrades triangulation. If distances to a few MSPs are known to better than a GW wavelength ($\sim$ 1 pc), these phases are fixed and localization improves by orders of magnitude. Simulations indicate that with sub-parsec distances for a handful of nearby MSPs, the uncertainty can shrink to $\sim 10^{-3}$ deg$^2$ (arcminute scale), enabling unique host association and multi-messenger follow-up. Achieving such distances requires $\sim$ 10 $μ$arcsec parallaxes for MSPs within a few hundred parsecs, a precision now approached with Very Long Baseline Interferometry (VLBI) and expected to become practical with phased-array SKA1-Mid operating as a sensitive VLBI element. SKA1's multi-beam, multi-calibrator astrometry will provide the independent distance priors needed for PTAs to localize nanohertz GW sources and measure SMBHB parameters and environments. We assess VLBI's role in PTA CW searches and propose a concrete SKA1-Mid observing strategy for nearby MSPs to deliver the required sub-parsec distances.

Time-domain and multimessenger astronomy (MMA/TDA) targets demand rapid-response follow-up observations. In many cases, it is the only way to make discoveries and advance our understanding of the astrophysical phenomena, for example, kilonovae accompanying gravitational waves from compact object mergers, shock breakout in supernovae, prompt emission from GRBs, etc. Presently the MMA/TDA follow-up workflow requires wrangling disparate software packages and user interfaces. We present an end-to-end software tool for the community, the Gemini Observation and Analysis of Targets System (GOATS), which unifies and simplifies the workflow, particularly for Gemini follow-up observations. GOATS achieves this by integrating services from Gemini Observatory and its parent organization, NSF NOIRLab. From a single platform, GOATS enables enhanced target selection via NOIRLab's ANTARES alert broker, triggering of Gemini (and other facilities within the Astronomical Event Observatory Network), automated data retrieval from the Gemini Observatory Archive, and interactive data reduction and analysis through Gemini's DRAGONS software and NOIRLab's Astro Data Lab science platform. GOATS was successfully deployed in an end-to-end demonstration of real-time follow-up of Rubin/LSST alerts with NOIRLab facilities. As part of this demonstration, we selected targets from the Rubin alert stream and triggered follow-up observations within minutes of the Rubin detections. We obtained spectra for several targets and classified them as supernova of various types (Ia, IIP, Ib/c) with redshifts ranging from 0.05 to 0.35. By eliminating the need to manually connect tools and automating repetitive tasks, GOATS lowers the entry barrier and allows users to focus on the scientific interpretation of the observation results.

Simulated black hole accretion flows with strong magnetic fields often exhibit "flux eruption events" (FEEs), transient and localized expulsions of matter near the event horizon due to magnetic reconnection. It may now be possible to image them with the Event Horizon Telescope (EHT), a global network of millimeter-wave observatories that images black holes. Here we use machine learning as an interpretable inference tool to identify observational signatures of FEEs that could be accessible to the EHT. First, we train a convolutional neural network to learn task-relevant representations of FEEs in uncorrupted simulated images. After using this network to label a larger set of images, we then train interpretable models (random forest and logistic regression) to determine observational signatures. We find that during a FEE, images in the millimeter tend toward more diffuse emission, higher linear polarization, and lower total fluxes, but these signatures are weak for most FEEs compared to the usual time variability of these features. Moreover, the Q-U loop rotation rate decreases during FEEs, contrary to a picture in which FEEs could jointly cause both millimeter Q-U loops and flares. Our random forest trained on observable summary statistics achieves ~80% class-weighted accuracy, suggesting that the CNN learns FEE structure not fully mapped onto these traditional summary statistics. Our results imply that image size and polarization fraction can be used to flag candidate FEEs, but high-resolution, high-dynamic range images will still be important to confirm FEEs and test accretion flows for this phenomenon.

Galaxy interactions are a key ingredient in galaxy evolution; not only are they a primary pathway of galaxy growth and mass assembly, but also a key driver of processes such as star formation and quenching. We investigate the impact of galaxy-galaxy interactions on stellar mass assembly using JWST/NIRCam observations of a spectroscopically selected sample of galaxies at $5.0<z_{spec}<5.6$ from the Canadian NIRISS Unbiased Cluster Survey (CANUCS). Of the 48 galaxies in our parent sample, we visually classify 21 ($44\%$) as closely-interacting ($\lesssim$ 5 kpc) systems with two or more components. We evaluate the non-parametric star formation histories (SFHs) of these systems' components using the spectral energy distribution fitting code \textsc{Dense Basis}. We find that the components in these systems experience brief intervals ($\sim0.2$ Gyr) of strongly enhanced star formation that grow their stellar mass by $\sim2.66\pm0.85\times$, forming $\sim1.71\pm0.37\times$ of excess mass than expected compared to if there was no burst. Attributing these star formation rate enhancements to interactions and assuming that the components will merge, we find that mergers are responsible for $\sim42^{+20}_{-25}\%$ of the total stellar mass growth of galaxies at $z\sim5$. While about half of this contribution comes from the merging of the pre-existing stellar masses of the merging galaxies, half is due to stellar mass that is newly-formed during the interaction. We conclude that mergers, and their associated star formation bursts, are an important pathway for stellar mass growth in high-$z$ galaxies.

Pulsar timing arrays conduct low-frequency gravitational wave searches, which require comprehensive accounting of various noise sources to achieve robust results. Interstellar propagation effects (e.g., dispersion and scattering) are especially complex noise sources, introducing chromatic delays that can reduce sensitivity to gravitational waves and bias their inference if left unmodeled. These delays also strongly depend on the line of sight properties to each individual pulsar. To address this, we present customized chromatic noise models for 67 pulsars in the NANOGrav 15 yr dataset. These models are selected from an expanded suite of Gaussian processes to simultaneously characterize multiple types of chromatic delays and are tailored to each pulsar's dataset. Alongside probing the interstellar medium, we use these models to infer the solar wind electron density over the course of $\sim 1.5$ solar cycles. We also find evidence for non-dispersive chromatic delays in 21 out of 67 NANOGrav pulsars. After applying our chromatic models, we observe significant impacts on the inference of achromatic noise in 19 out of 67 pulsars, finding in several cases that a previously significant achromatic noise process can be partially or entirely described as chromatic. These results demonstrate that refined noise modeling is essential to enhance the sensitivity and accuracy of low-frequency gravitational wave searches with pulsar timing arrays.

SN 2024abup is a nearby broad-lined Type Ic supernova (SN Ic-bl) in NGC 0681 at a distance of 23.3 \pm 1.6 Mpc. As energetic explosions of massive stars, SNe Ic-bl are considered a plausible site for rapid-neutron capture nucleosynthesis (r-process) and chemical enrichment from short-lived progenitors. They may also contribute to dust production in the early Universe. We present JWST near- to mid-infrared (NIR+MIR) observations (1-14 micron) of SN Ic-bl 2024abup at +41 days after the V band maximum (+54 days after explosion), the first-ever JWST+MIR observation of a SN Ic-bl along with radio and optical data. Using the spectral synthesis code SUMO, we identify the observed broad IR line features in SN 2024abup and find significant contributions from C, O, Mg, and carbon monoxide (CO) -- the earliest detection of molecules in a core-collapse SN so far. The spectrum shows continuum emission at wavelengths greater than 1.5 micron, which could be explained by dust -- preexisting, newly formed, or a combination-heated by the SN. We do not find compelling evidence for infrared signatures of r-process elements, though our search is hampered by the presence of many broad and blended features from the non-r-process elements. These new observations indicate that SNe Ic-bl could be a contributor to early-universe dust production, and suggest that if r-process elements are produced, revealing their presence from spectra requires very high-quality data and models to disentangle blends.

High-redshift galaxy clustering provides a powerful probe of the growth of structure, testing models of dark matter, dark energy, and galaxy formation during the epoch when the Universe was rapidly evolving. Emission line galaxies (ELGs) and quasars (QSOs) are used as tracers of dark matter by the Dark Energy Spectroscopic Instrument (DESI) to probe this redshift regime. We present results from ELG and QSO mock catalogs created from the Uchuu N-body simulation and tuned to DESI Data Release 2 (DR2) clustering. Employing a modified subhalo abundance matching (SHAM) technique, we populate Uchuu halos and subhalos with QSOs between 0.8 < z < 2.1. For ELGs, we modify this method to select satellite galaxies with low velocities relative to their associated central halos, and populate a separate set of Uchuu halos and subhalos with ELGs between 0.8 < z < 1.6. In this paper, we reproduce the redshift evolution of number density and clustering statistics across the fitted range of scales. We also measure the large-scale clustering bias of both the data and mock samples. These results improve simulated lightcone construction from cosmological models and enhance our understanding of the galaxy-halo connection.

We report updated nHz gravitational wave (GW) significance, characterization, and interpretations using the customized chromatic-noise models (CNMs) developed in Larsen, Baier et al. (2026). for the NANOGrav 15-year data set. We find increased evidence for the Hellings-Downs (HD) correlation signature of the stochastic gravitational wave background (GWB), with a Bayes factor of $1571\pm14$ for HD-correlations over a common uncorrelated red-noise process using a power-law model with $14$ Fourier modes. We find this $\sim8\times$ increase in Bayes factor from Agazie et al. (2023a) is a result of improved noise mitigation. Assuming an analytic null distribution for the frequentist interpulsar correlation statistic, this corresponds to a slightly more significant measurement from $3.16σ$ to $3.32σ$ against the no-correlation scenario. Spectral inference with CNMs brings the power-law GWB amplitude down to $A_{\rm GWB} = 2.1^{+0.6}_{-0.5}\times10^{-15}$ at fixed $γ_{\rm GWB} = 13/3$. In a varied-$γ$ analysis, the spectral index increases to $γ_{\rm GWB}=3.5^{+0.7}_{-0.6}$. We report updates on an all-sky continuous gravitational wave (CW) search as well as select targeted searches and calculate a $3.2\times$ larger detection volume for the NANOGrav detector. With CNMs, we find reduced evidence for a non-Einsteinian, scalar-transverse mode of gravity. Finally, we reinterpret the GWB first with the assumption of an astrophysical background sourced by SMBHBs and then assuming the more exotic origins of cosmic inflation, a first-order cosmological phase transition, and stable cosmic strings. Under both the SMBHB hypothesis and the cosmological hypotheses, we see only marginal shifts in model parameter posteriors which are consistent with the slightly quieter and steeper power-law GWB spectrum.

Recent studies have argued that isolated binary evolution simulations generically overestimate the observed local binary black hole (BBH) merger rate, even after adopting observationally motivated variations in the metallicity-dependent cosmic star-formation history, and have interpreted this as motivation for drastic revisions to binary stellar evolution models. We revisit these claims using a compilation of 1490 simulated BBH merger rates from 57 isolated binary-evolution studies, compared to constraints from the LIGO--Virgo--KAGRA Collaboration through GWTC-5. While $\sim$80% of compiled submodels find rates above the GWTC-5 interval, a substantial subset reproduces or underestimates the observed rate. The literature spans several orders of magnitude, reflecting strong sensitivity to assumptions about natal kicks, common-envelope evolution, mass transfer, angular-momentum loss, remnant formation, stellar winds, initial conditions, and star-formation history. Using 2543 pairwise BBH submodel variations constructed to isolate single physical assumptions, we identify which choices most strongly impact the simulated BBH merger rate. Low BBH merger rates are not uniquely associated with strong natal kicks or reduced low-metallicity star formation. Multiple physically motivated assumptions can independently reduce simulated rates to values consistent with observations. We further show that simulated rates cluster into `simulation silos': frameworks producing apparent consensus within a code that does not generalize beyond it. Our results indicate that claims of a universal BBH merger-rate crisis are strongly model dependent, and underscore the importance of exploring the full parameter space across multiple population-synthesis frameworks before concluding that isolated binary evolution is in tension with gravitational-wave observations.

We present a radio-oriented review of current strategies for the detection and characterization of dual active galactic nuclei (DAGN) and supermassive black hole binaries (SMBHBs), emphasizing the crucial role of radio interferometry in advancing this field. We discuss how high-resolution radio imaging - particularly through very long baseline interferometry (VLBI) - provides a unique, dust-unbiased tool to identify multiple accreting SMBHs, disentangle AGN-related emission from star formation, and trace components from tens of kpc to sub-parsec scales. We summarize current observational limitations, such as insufficient sensitivity-resolution combination and area coverage. We then outline how the SKAO will overcome these constraints through its unprecedented combination of sensitivity, survey speed, imaging fidelity and angular resolution, enabling the discovery and characterization of dual and binary SMBHs from the nearby Universe to the epoch of reionization. Several science cases are presented, including radio follow-ups of optical/infrared-selected DAGN, direct blind radio selection of DAGN, studies of compact bound SMBHBs, and the link between SMBHB orbital evolution and low-frequency gravitational wave emission. We further emphasize the synergy between SKAO observations and modern and upcoming facilities such as the James Webb and Euclid space telescopes, Rubin Observatory, and gravitational wave detectors including the Laser Interferometer Space Antenna and pulsar timing arrays. These combined capabilities will allow SKAO to enable the first comprehensive radio census of dual and binary SMBH systems, bridge the gap between electromagnetic and gravitational wave observations, and provide a statistically significant view of SMBH pairing, accretion, and merger-driven feedback throughout cosmic history.

Gravitational waves from inspiraling binary black holes (BBHs) provide insights into the lives and deaths of massive stars. Population synthesis allows us to model these binaries through isolated binary evolution, but its predictive power is limited by difficulties in varying the stellar models and their associated uncertainties. We present a new grid of stellar tracks computed with the open-source stellar evolution code MESA, spanning metallicities $10^{-3} \le Z/Z_{\odot} \le 7$. We vary two stellar physics parameters: wind-driven mass loss and the convective boundary mixing (CBM) mechanism. We pair these models with the Method of Interpolation for Single Stellar Evolution (METISSE) and binary population synthesis code COSMIC to obtain synthetic populations of merging BBHs in the local Universe. We find a maximum in the primary mass spectrum near $10M_\odot$ which in most model variations is composed of two sub-populations at $\approx8M_{\odot}$ and $\approx13 M_\odot$, with the higher-mass population dominated by BBHs whose progenitors underwent a mass ratio reversal (MRR). This population also suggests an anticorrelation between higher primary masses and mass ratio, as BBHs with $m_1\gtrapprox10M_\odot$ preferentially undergo MRR and prefer a final mass ratio of $q\approx0.7$. However, the location and relative strength of these two sub-populations is sensitive to our assumed stellar physics: varying both the wind and CBM treatments can merge the MRR and non-MRR populations into a single peak near $9M_\odot$. Variations in our stellar tracks, especially CBM, lead to a factor of $\approx6$ difference in the rate, primarily due to modulation of the common envelope formation channel.

We report the first detection of the `convective-transition gap' (also known as `M-dwarf gap') in the globular cluster 47 Tuc (NGC 104) thanks to Euclid data. This feature, linked to a change in the physical properties of late-type dwarfs, has remained elusive, with only two detections so far. Leveraging the large number of stars, high resolution, and photometric precision enabled by Euclid, we detect a statistically significant, sharp discontinuity in the main-sequence luminosity function of 47 Tuc at $I_{\rm E} \approx 22.9$, which we identify as the convective-transition gap. We compare the observed properties of the gap in 47 Tuc with theoretical models, showing how the gap can be a powerful diagnostic to probe the internal chemical structure of globular clusters, and their multiple stellar populations. Following its initial discovery in the metal-poor cluster NGC 6397, the identification of a convective gap in the metal-rich 47 Tuc suggests that this feature might be more general than previously thought. These results demonstrate that Euclid can be transformative well beyond cosmology, with impact across multiple areas of astrophysics, including resolved stellar populations.

Independent luminosity-distance probes beyond the Type Ia supernova range are needed to test cosmic expansion at high redshift. Type 1 AGN are abundant at \(z>2\), but their use for cosmology requires standardizable observables with controlled scatter, redshift dependence, and measurement uncertainty. We present a hierarchical Bayesian framework for cosmology using AGN variability, based on the empirical anti-correlation between optical/UV variability amplitude and luminosity. The method targets the moderate-baseline regime of current wide-field time-domain surveys, where individual light curves cannot typically identify the full long-timescale stochastic process, but can constrain finite-window brightness and short-lag variability. Each light curve is fitted independently to obtain posterior samples of these summaries, which are then importance-reweighted under a population model relating variability to luminosity, rest-frame wavelength, intrinsic scatter, and the assumed distance-redshift relation. This framework propagates object-level uncertainty while avoiding repeated light-curve likelihood evaluations during cosmological inference, making catalogue-scale analyses feasible. Using Gaia DR3-like G-band simulations matched to real Gaia cadences, noise properties, and quality cuts, we show that finite-baseline light curves are more robustly summarized by window-averaged brightness and short-lag variability than by the separate long-timescale parameters of stochastic models. End-to-end closure tests recover the injected variability-luminosity relation, intrinsic scatter, and distance-redshift parameters up to the expected calibration degeneracies. This Gaia G-band analysis establishes a proof of concept for AGN-variability distances in the moderate-baseline survey regime, with the main gains expected from Gaia DR4, ZTF, DESI-selected AGN samples, and Rubin/LSST-era data.

We measure the power spectral density (PSD) of the accretion rate time series in an unequal mass (q = 0.2) binary surrounded by a circumbinary gas disk, using very high-resolution 2D hydrodynamics simulations. Our aim is to identify new signposts of supermassive black hole (SMBH) binaries in active galactic nuclei (AGN), based on the shape of the continuum PSD, to complement well-studied line features in the PSD (periodicities). We find that the continuum PSD is a broken power-law, transitioning from flat (white noise) to a slope of -4 at a break frequency generically ~5 times the binary orbital frequency. This form is expected when (a) delivery of gas from the circumbinary disk to the individual "minidisks" is a damped random walk with correlation time equal to binary orbital period and (b) the minidisks function as low-pass filters acting at the Kepler frequency of the outer edge of the smaller black hole's minidisk; we show numerical evidence for both. The broken power-law PSD is attained in a limit where the secondary black hole is much smaller than its minidisk, realized numerically by a sufficiently small "sink" region; larger sinks lead to excess high-frequency noise seen as accretion rate spikes, and we argue these should be regarded as artificial when the black holes themselves are smaller than the sink regions. The broken power-law PSD is reminiscent of stochastic variability in ordinary AGN, inviting the conjecture that canonical AGN variability could result from widespread binarity, however pulsar timing experiments may exclude this possibility.

The origin of the complex orbital structure of young massive stars at the Galactic centre remains an open question. If these stars formed in a single episode from a gaseous accretion disc, they may initially have constituted a single, coherently rotating stellar disc. We investigate whether perturbations from an unseen intermediate-mass black hole (IMBH) could fragment and/or disrupt such a disc into the multiple orbital components observed today. First, we derive a theoretical criterion for when and where the IMBH's torque overcomes the disc's self-torque and tears it apart. We then test this picture with direct $N$-body simulations of a stellar disc interacting with an inclined IMBH around a central supermassive black hole. We find that the outcome depends strongly on the IMBH's orbit and mass. A prograde IMBH rapidly aligns with the stellar disc, while a massive retrograde IMBH ($m_{\bullet} \simeq 0.67\,M_{\rm d}$) anti-aligns relative to the radially overlapping stars and efficiently fragments the original disc into three components in angular-momentum space: an inner disc, a misaligned overlapping region, and an unperturbed outer disc. The IMBH also excites eccentricities in the overlapping region, driving stars away from initially circular orbits. These features emerge for an IMBH mass of $2000\,{\rm M}_{\odot}$ and a disc mass of $3000\,{\rm M}_{\odot}$ within 10--20 Myr, a timescale comparable to the age of the young Galactic centre stellar population, and provide a plausible explanation for the observed multiple orbital planes, warped geometry, and broad eccentricity distribution.

Galaxy sizes are correlated with stellar mass and redshift, as characterised by size scaling relations. The inferred forms of these scaling relations are sensitive to how galaxies are classified -- either by their star formation activity (e.g. specific star-formation rate, sSFR) or by their morphology markers (e.g. bulge-to-total ratio, Sérsic index). We combine stellar mass and sSFR estimates from pop-cosmos (a generative model trained on COSMOS2020 Spitzer IRAC $\textit{Ch.1} <26$) with size and morphology measurements from COSMOS-Web, obtaining $99,369$ galaxies. By investigating the size-mass and the size-redshift relations, we show that: (i) the sSFR/morphology splits give quantitatively different slopes, intercepts, and intrinsic scatter behaviour; (ii) intrinsic scatter depends on structural morphology but not on sSFR, which constrains the galaxy-halo connection; (iii) the quiescent and bulge-dominated size-mass relations both show double-power law breaks, but at different pivot masses, indicating that quenching and structural transformation occur on different time-scales; (iv) the morphology-dependent trends are only recoverable from space-based imaging. Further, the quiescent pivot mass $M_{\ast} \sim 10^{10.7}~\mathrm{M}_{\odot}$ coincides with the mass scale at which AGN (infrared torus) bolometric luminosity fraction peaks in transitioning galaxies, while the bulge-dominated pivot mass $M_{\ast} \sim 10^{11.1}~\mathrm{M}_{\odot}$ coincides with the halo mass at which AGN-driven baryonic redistribution peaks, tracing the interval over which AGN feedback ramps from quenching onset to structural transformation.