arXiv Daily Digest - 2026-01-16

Detecting persuasion in argumentative text is a challenging task with important implications for understanding human communication. This work investigates the role of persuasion strategies - such as Attack on reputation, Distraction, and Manipulative wording - in determining the persuasiveness of a text. We conduct experiments on three annotated argument datasets: Winning Arguments (built from the Change My View subreddit), Anthropic/Persuasion, and Persuasion for Good. Our approach leverages large language models (LLMs) with a Multi-Strategy Persuasion Scoring approach that guides reasoning over six persuasion strategies. Results show that strategy-guided reasoning improves the prediction of persuasiveness. To better understand the influence of content, we organize the Winning Argument dataset into broad discussion topics and analyze performance across them. We publicly release this topic-annotated version of the dataset to facilitate future research. Overall, our methodology demonstrates the value of structured, strategy-aware prompting for enhancing interpretability and robustness in argument quality assessment.

Large Language Models (LLMs) have emerged as powerful operators for evolutionary search, yet the design of efficient search scaffolds remains ad hoc. While promising, current LLM-in-the-loop systems lack a systematic approach to managing the evolutionary process. We identify three distinct failure modes: Context Pollution, where experiment history biases future candidate generation; Mode Collapse, where agents stagnate in local minima due to poor exploration-exploitation balance; and Weak Collaboration, where rigid crossover strategies fail to leverage parallel search trajectories effectively. We introduce Progress-Aware Consistent Evolution (PACEvolve), a framework designed to robustly govern the agent's context and search dynamics, to address these challenges. PACEvolve combines hierarchical context management (HCM) with pruning to address context pollution; momentum-based backtracking (MBB) to escape local minima; and a self-adaptive sampling policy that unifies backtracking and crossover for dynamic search coordination (CE), allowing agents to balance internal refinement with cross-trajectory collaboration. We demonstrate that PACEvolve provides a systematic path to consistent, long-horizon self-improvement, achieving state-of-the-art results on LLM-SR and KernelBench, while discovering solutions surpassing the record on Modded NanoGPT.

We study LTLf synthesis with multiple properties, where satisfying all properties may be impossible. Instead of enumerating subsets of properties, we compute in one fixed-point computation the relation between product-game states and the goal sets that are realizable from them, and we synthesize strategies achieving maximal realizable sets. We develop a fully symbolic algorithm that introduces Boolean goal variables and exploits monotonicity to represent exponentially many goal combinations compactly. Our approach substantially outperforms enumeration-based baselines, with speedups of up to two orders of magnitude.

Large language models (LLMs) can increase users' perceived trust by verbalizing confidence in their outputs. However, prior work has shown that LLMs are often overconfident, making their stated confidence unreliable since it does not consistently align with factual accuracy. To better understand the sources of this verbalized confidence, we introduce TracVC (\textbf{Trac}ing \textbf{V}erbalized \textbf{C}onfidence), a method that builds on information retrieval and influence estimation to trace generated confidence expressions back to the training data. We evaluate TracVC on OLMo and Llama models in a question answering setting, proposing a new metric, content groundness, which measures the extent to which an LLM grounds its confidence in content-related training examples (relevant to the question and answer) versus in generic examples of confidence verbalization. Our analysis reveals that OLMo2-13B is frequently influenced by confidence-related data that is lexically unrelated to the query, suggesting that it may mimic superficial linguistic expressions of certainty rather than rely on genuine content grounding. These findings point to a fundamental limitation in current training regimes: LLMs may learn how to sound confident without learning when confidence is justified. Our analysis provides a foundation for improving LLMs' trustworthiness in expressing more reliable confidence.

Adjusted similarity measures, such as Cohen's kappa for inter-rater reliability and the adjusted Rand index used to compare clustering algorithms, are a vital tool for comparing discrete labellings. These measures are intended to have the property of 0 expectation under a null distribution and maximum value 1 under maximal similarity to aid in interpretation. Measures are frequently adjusted with respect to the permutation distribution for historic and analytic reasons. There is currently renewed interest in considering other null models more appropriate for context, such as clustering ensembles permitting a random number of identified clusters. The purpose of this work is two -- fold: (1) to generalize the study of the adjustment operator to general null models and to a more general procedure which includes statistical standardization as a special case and (2) to identify sufficient conditions for the adjustment operator to produce the intended properties, where sufficient conditions are related to whether and how observed data are incorporated into null distributions. We demonstrate how violations of the sufficient conditions may lead to substantial breakdown, such as by producing a non-positive measure under traditional adjustment rather than one with mean 0, or by producing a measure which is deterministically 0 under statistical standardization.

Fine-grained sparsity promises higher parametric capacity without proportional per-token compute, but often suffers from training instability, load balancing, and communication overhead. We introduce STEM (Scaling Transformers with Embedding Modules), a static, token-indexed approach that replaces the FFN up-projection with a layer-local embedding lookup while keeping the gate and down-projection dense. This removes runtime routing, enables CPU offload with asynchronous prefetch, and decouples capacity from both per-token FLOPs and cross-device communication. Empirically, STEM trains stably despite extreme sparsity. It improves downstream performance over dense baselines while reducing per-token FLOPs and parameter accesses (eliminating roughly one-third of FFN parameters). STEM learns embedding spaces with large angular spread which enhances its knowledge storage capacity. More interestingly, this enhanced knowledge capacity comes with better interpretability. The token-indexed nature of STEM embeddings allows simple ways to perform knowledge editing and knowledge injection in an interpretable manner without any intervention in the input text or additional computation. In addition, STEM strengthens long-context performance: as sequence length grows, more distinct parameters are activated, yielding practical test-time capacity scaling. Across 350M and 1B model scales, STEM delivers up to ~3--4% accuracy improvements overall, with notable gains on knowledge and reasoning-heavy benchmarks (ARC-Challenge, OpenBookQA, GSM8K, MMLU). Overall, STEM is an effective way of scaling parametric memory while providing better interpretability, better training stability and improved efficiency.

Classification imbalance arises when one class is much rarer than the other. We frame this setting as transfer learning under label (prior) shift between an imbalanced source distribution induced by the observed data and a balanced target distribution under which performance is evaluated. Within this framework, we study a family of oversampling procedures that augment the training data by generating synthetic samples from an estimated minority-class distribution to roughly balance the classes, among which the celebrated SMOTE algorithm is a canonical example. We show that the excess risk decomposes into the rate achievable under balanced training (as if the data had been drawn from the balanced target distribution) and an additional term, the cost of transfer, which quantifies the discrepancy between the estimated and true minority-class distributions. In particular, we show that the cost of transfer for SMOTE dominates that of bootstrapping (random oversampling) in moderately high dimensions, suggesting that we should expect bootstrapping to have better performance than SMOTE in general. We corroborate these findings with experimental evidence. More broadly, our results provide guidance for choosing among augmentation strategies for imbalanced classification.

We study potential presence of statistical-computational gaps (SCG) in symmetric binary perceptrons (SBP) via a parametric utilization of \emph{fully lifted random duality theory} (fl-RDT) [96]. A structural change from decreasingly to arbitrarily ordered $c$-sequence (a key fl-RDT parametric component) is observed on the second lifting level and associated with \emph{satisfiability} ($α_c$) -- \emph{algorithmic} ($α_a$) constraints density threshold change thereby suggesting a potential existence of a nonzero computational gap $SCG=α_c-α_a$. The second level estimate is shown to match the theoretical $α_c$ whereas the $r\rightarrow \infty$ level one is proposed to correspond to $α_a$. For example, for the canonical SBP ($κ=1$ margin) we obtain $α_c\approx 1.8159$ on the second and $α_a\approx 1.6021$ (with converging tendency towards $\sim 1.59$ range) on the seventh level. Our propositions remarkably well concur with recent literature: (i) in [20] local entropy replica approach predicts $α_{LE}\approx 1.58$ as the onset of clustering defragmentation (presumed driving force behind locally improving algorithms failures); (ii) in $α\rightarrow 0$ regime we obtain on the third lifting level $κ\approx 1.2385\sqrt{\frac{α_a}{-\log\left ( α_a \right ) }}$ which qualitatively matches overlap gap property (OGP) based predictions of [43] and identically matches local entropy based predictions of [24]; (iii) $c$-sequence ordering change phenomenology mirrors the one observed in asymmetric binary perceptron (ABP) in [98] and the negative Hopfield model in [100]; and (iv) as in [98,100], we here design a CLuP based algorithm whose practical performance closely matches proposed theoretical predictions.

Today's strongest video-language models (VLMs) remain proprietary. The strongest open-weight models either rely on synthetic data from proprietary VLMs, effectively distilling from them, or do not disclose their training data or recipe. As a result, the open-source community lacks the foundations needed to improve on the state-of-the-art video (and image) language models. Crucially, many downstream applications require more than just high-level video understanding; they require grounding -- either by pointing or by tracking in pixels. Even proprietary models lack this capability. We present Molmo2, a new family of VLMs that are state-of-the-art among open-source models and demonstrate exceptional new capabilities in point-driven grounding in single image, multi-image, and video tasks. Our key contribution is a collection of 7 new video datasets and 2 multi-image datasets, including a dataset of highly detailed video captions for pre-training, a free-form video Q&A dataset for fine-tuning, a new object tracking dataset with complex queries, and an innovative new video pointing dataset, all collected without the use of closed VLMs. We also present a training recipe for this data utilizing an efficient packing and message-tree encoding scheme, and show bi-directional attention on vision tokens and a novel token-weight strategy improves performance. Our best-in-class 8B model outperforms others in the class of open weight and data models on short videos, counting, and captioning, and is competitive on long-videos. On video-grounding Molmo2 significantly outperforms existing open-weight models like Qwen3-VL (35.5 vs 29.6 accuracy on video counting) and surpasses proprietary models like Gemini 3 Pro on some tasks (38.4 vs 20.0 F1 on video pointing and 56.2 vs 41.1 J&F on video tracking).

In the context of multi-agent multi-armed bandits (MA-MAB), fairness is often reduced to outcomes: maximizing welfare, reducing inequality, or balancing utilities. However, evidence in psychology, economics, and Rawlsian theory suggests that fairness is also about process and who gets a say in the decisions being made. We introduce a new fairness objective, procedural fairness, which provides equal decision-making power for all agents, lies in the core, and provides for proportionality in outcomes. Empirical results confirm that fairness notions based on optimizing for outcomes sacrifice equal voice and representation, while the sacrifice in outcome-based fairness objectives (like equality and utilitarianism) is minimal under procedurally fair policies. We further prove that different fairness notions prioritize fundamentally different and incompatible values, highlighting that fairness requires explicit normative choices. This paper argues that procedural legitimacy deserves greater focus as a fairness objective, and provides a framework for putting procedural fairness into practice.

Time Series Foundation Models (TSFMs) have emerged as a promising approach for zero-shot financial forecasting, demonstrating strong transferability and data efficiency gains. However, their adoption in financial applications is hindered by fundamental limitations in uncertainty quantification: current approaches either rely on restrictive distributional assumptions, conflate different sources of uncertainty, or lack principled calibration mechanisms. While recent TSFMs employ sophisticated techniques such as mixture models, Student's t-distributions, or conformal prediction, they fail to address the core challenge of providing theoretically-grounded uncertainty decomposition. For the very first time, we present a novel transformer-based probabilistic framework, ProbFM (probabilistic foundation model), that leverages Deep Evidential Regression (DER) to provide principled uncertainty quantification with explicit epistemic-aleatoric decomposition. Unlike existing approaches that pre-specify distributional forms or require sampling-based inference, ProbFM learns optimal uncertainty representations through higher-order evidence learning while maintaining single-pass computational efficiency. To rigorously evaluate the core DER uncertainty quantification approach independent of architectural complexity, we conduct an extensive controlled comparison study using a consistent LSTM architecture across five probabilistic methods: DER, Gaussian NLL, Student's-t NLL, Quantile Loss, and Conformal Prediction. Evaluation on cryptocurrency return forecasting demonstrates that DER maintains competitive forecasting accuracy while providing explicit epistemic-aleatoric uncertainty decomposition. This work establishes both an extensible framework for principled uncertainty quantification in foundation models and empirical evidence for DER's effectiveness in financial applications.

Large Language Models (LLMs) have achieved remarkable capabilities but remain vulnerable to adversarial ``jailbreak'' attacks designed to bypass safety guardrails. Current safety alignment methods depend heavily on static external red teaming, utilizing fixed defense prompts or pre-collected adversarial datasets. This leads to a rigid defense that overfits known patterns and fails to generalize to novel, sophisticated threats. To address this critical limitation, we propose empowering the model to be its own red teamer, capable of achieving autonomous and evolving adversarial attacks. Specifically, we introduce Safety Self- Play (SSP), a system that utilizes a single LLM to act concurrently as both the Attacker (generating jailbreaks) and the Defender (refusing harmful requests) within a unified Reinforcement Learning (RL) loop, dynamically evolving attack strategies to uncover vulnerabilities while simultaneously strengthening defense mechanisms. To ensure the Defender effectively addresses critical safety issues during the self-play, we introduce an advanced Reflective Experience Replay Mechanism, which uses an experience pool accumulated throughout the process. The mechanism employs a Upper Confidence Bound (UCB) sampling strategy to focus on failure cases with low rewards, helping the model learn from past hard mistakes while balancing exploration and exploitation. Extensive experiments demonstrate that our SSP approach autonomously evolves robust defense capabilities, significantly outperforming baselines trained on static adversarial datasets and establishing a new benchmark for proactive safety alignment.

Whether neural information processing is entirely classical or involves quantum-mechanical elements remains an open question. Here we propose a model-agnostic, information-theoretic test of nonclassicality that bypasses microscopic assumptions and instead probes the structure of neural representations themselves. Using autoencoders as a transparent model system, we introduce a Bell-type consistency test in latent space, and ask whether decoding statistics obtained under multiple readout contexts can be jointly explained by a single positive latent-variable distribution. By shifting the search for quantum-like signatures in neural systems from microscopic dynamics to experimentally testable constraints on information processing, this work opens a new route for probing the fundamental physics of neural computation.

The paper introduces a white-box attack on computer vision models using SHAP values. It demonstrates how adversarial evasion attacks can compromise the performance of deep learning models by reducing output confidence or inducing misclassifications. Such attacks are particularly insidious as they can deceive the perception of an algorithm while eluding human perception due to their imperceptibility to the human eye. The proposed attack leverages SHAP values to quantify the significance of individual inputs to the output at the inference stage. A comparison is drawn between the SHAP attack and the well-known Fast Gradient Sign Method. We find evidence that SHAP attacks are more robust in generating misclassifications particularly in gradient hiding scenarios.

Combinatorial optimization augmented machine learning (COAML) has recently emerged as a powerful paradigm for integrating predictive models with combinatorial decision-making. By embedding combinatorial optimization oracles into learning pipelines, COAML enables the construction of policies that are both data-driven and feasibility-preserving, bridging the traditions of machine learning, operations research, and stochastic optimization. This paper provides a comprehensive overview of the state of the art in COAML. We introduce a unifying framework for COAML pipelines, describe their methodological building blocks, and formalize their connection to empirical cost minimization. We then develop a taxonomy of problem settings based on the form of uncertainty and decision structure. Using this taxonomy, we review algorithmic approaches for static and dynamic problems, survey applications across domains such as scheduling, vehicle routing, stochastic programming, and reinforcement learning, and synthesize methodological contributions in terms of empirical cost minimization, imitation learning, and reinforcement learning. Finally, we identify key research frontiers. This survey aims to serve both as a tutorial introduction to the field and as a roadmap for future research at the interface of combinatorial optimization and machine learning.

Deploying Python based AI agents on resource-constrained edge devices presents a runtime optimization challenge: high thread counts are needed to mask I/O latency, yet Python's Global Interpreter Lock (GIL) serializes execution. We demonstrate that naive thread-pool scaling causes a "saturation cliff": >= 20% throughput degradation at overprovisioned thread counts (N >= 512) on edge-representative configurations. We present a lightweight profiling tool and adaptive runtime system using a Blocking Ratio metric (beta) that distinguishes genuine I/O wait from GIL contention. Our library-based solution achieves 96.5% of optimal performance without manual tuning, outperforming multiprocessing (limited by ~8x memory overhead on devices with 512 MB-2 GB RAM) and asyncio (blocked by CPU-bound phases). Evaluation across seven edge AI workload profiles, including real ML inference with ONNX Runtime MobileNetV2, demonstrates 93.9% average efficiency. Comparative experiments with Python 3.13t (free threading) show that while GIL elimination enables ~4x throughput on multi-core edge devices, the saturation cliff persists on single-core devices, validating our beta metric for both GIL and no-GIL environments. This provides practical optimization for edge AI systems.

Comprehending genomic information is essential for biomedical research, yet extracting data from complex distributed databases remains challenging. Large language models (LLMs) offer potential for genomic Question Answering (QA) but face limitations due to restricted access to domain-specific databases. GeneGPT is the current state-of-the-art system that enhances LLMs by utilizing specialized API calls, though it is constrained by rigid API dependencies and limited adaptability. We replicate GeneGPT and propose GenomAgent, a multi-agent framework that efficiently coordinates specialized agents for complex genomics queries. Evaluated on nine tasks from the GeneTuring benchmark, GenomAgent outperforms GeneGPT by 12% on average, and its flexible architecture extends beyond genomics to various scientific domains needing expert knowledge extraction.

Intrinsic evaluation metrics for conditional language models, such as perplexity or bits-per-character, are widely used in both mono- and multilingual settings. These metrics are rather straightforward to use and compare in monolingual setups, but rest on a number of assumptions in multilingual setups. One such assumption is that comparing the perplexity of CLMs on parallel sentences is indicative of their quality since the information content (here understood as the semantic meaning) is the same. However, the metrics are inherently measuring information content in the information-theoretic sense. We make this and other such assumptions explicit and discuss their implications. We perform experiments with six metrics on two multi-parallel corpora both with mono- and multilingual models. Ultimately, we find that current metrics are not universally comparable. We look at the form-meaning debate to provide some explanation for this.

In this article, we argue that understanding the collective behavior of agents based on large language models (LLMs) is an essential area of inquiry, with important implications in terms of risks and benefits, impacting us as a society at many levels. We claim that the distinctive nature of LLMs--namely, their initialization with extensive pre-trained knowledge and implicit social priors, together with their capability of adaptation through in-context learning--motivates the need for an interactionist paradigm consisting of alternative theoretical foundations, methodologies, and analytical tools, in order to systematically examine how prior knowledge and embedded values interact with social context to shape emergent phenomena in multi-agent generative AI systems. We propose and discuss four directions that we consider crucial for the development and deployment of LLM-based collectives, focusing on theory, methods, and trans-disciplinary dialogue.

Selective knowledge erasure from LLMs is critical for GDPR compliance and model safety, yet current unlearning methods conflate behavioral suppression with true knowledge removal, allowing latent capabilities to persist beneath surface-level refusals. In this work, we address this challenge by introducing Knowledge Immunization Framework (KIF), a representation-aware architecture that distinguishes genuine erasure from obfuscation by targeting internal activation signatures rather than surface outputs. Our approach combines dynamic suppression of subject-specific representations with parameter-efficient adaptation, enabling durable unlearning without full model retraining. KIF achieves near-oracle erasure (FQ approx 0.99 vs. 1.00) while preserving utility at oracle levels (MU = 0.62), effectively breaking the stability-erasure tradeoff that has constrained all prior work. We evaluate both standard foundation models (Llama and Mistral) and reasoning-prior models (Qwen and DeepSeek) across 3B to 14B parameters. Our observation shows that standard models exhibit scale-independent true erasure (<3% utility drift), while reasoning-prior models reveal fundamental architectural divergence. Our comprehensive dual-metric evaluation protocol, combining surface-level leakage with latent trace persistence, operationalizes the obfuscation - erasure distinction and enables the first systematic diagnosis of mechanism-level forgetting behavior across model families and scales.

The research undertakes a comprehensive comparative analysis of Kolmogorov-Arnold Networks (KAN) and Multi-Layer Perceptrons (MLP), highlighting their effectiveness in solving essential computational challenges like nonlinear function approximation, time-series prediction, and multivariate classification. Rooted in Kolmogorov's representation theorem, KANs utilize adaptive spline-based activation functions and grid-based structures, providing a transformative approach compared to traditional neural network frameworks. Utilizing a variety of datasets spanning mathematical function estimation (quadratic and cubic) to practical uses like predicting daily temperatures and categorizing wines, the proposed research thoroughly assesses model performance via accuracy measures like Mean Squared Error (MSE) and computational expense assessed through Floating Point Operations (FLOPs). The results indicate that KANs reliably exceed MLPs in every benchmark, attaining higher predictive accuracy with significantly reduced computational costs. Such an outcome highlights their ability to maintain a balance between computational efficiency and accuracy, rendering them especially beneficial in resource-limited and real-time operational environments. By elucidating the architectural and functional distinctions between KANs and MLPs, the paper provides a systematic framework for selecting the most suitable neural architectures for specific tasks. Furthermore, the proposed study highlights the transformative capabilities of KANs in progressing intelligent systems, influencing their use in situations that require both interpretability and computational efficiency.

Concept Bottleneck Models (CBMs) improve the explainability of black-box Deep Learning (DL) by introducing intermediate semantic concepts. However, standard CBMs often overlook domain-specific relationships and causal mechanisms, and their dependence on complete concept labels limits applicability in scientific domains where supervision is sparse but processes are well defined. To address this, we propose the Process-Guided Concept Bottleneck Model (PG-CBM), an extension of CBMs which constrains learning to follow domain-defined causal mechanisms through biophysically meaningful intermediate concepts. Using above ground biomass density estimation from Earth Observation data as a case study, we show that PG-CBM reduces error and bias compared to multiple benchmarks, whilst leveraging multi-source heterogeneous training data and producing interpretable intermediate outputs. Beyond improved accuracy, PG-CBM enhances transparency, enables detection of spurious learning, and provides scientific insights, representing a step toward more trustworthy AI systems in scientific applications.

Multi-agent systems (MAS) enable complex reasoning by coordinating multiple agents, but often incur high inference latency due to multi-step execution and repeated model invocations, severely limiting their scalability and usability in time-sensitive scenarios. Most existing approaches primarily optimize task performance and inference cost, and explicitly or implicitly assume sequential execution, making them less optimal for controlling latency under parallel execution. In this work, we investigate learning-based orchestration of multi-agent systems with explicit latency supervision under parallel execution. We propose Latency-Aware Multi-agent System (LAMaS), a latency-aware multi-agent orchestration framework that enables parallel execution and explicitly optimizes the critical execution path, allowing the controller to construct execution topology graphs with lower latency under parallel execution. Our experiments show that our approach reduces critical path length by 38-46% compared to the state-of-the-art baseline for multi-agent architecture search across multiple benchmarks, while maintaining or even improving task performance. These results highlight the importance of explicitly optimizing latency under parallel execution when designing efficient multi-agent systems. The code is available at https://github.com/xishi404/LAMaS

Studying the optoelectronic structure of materials can require the computation of up to several thousands of the smallest eigenpairs of a pseudo-hermitian Hamiltonian. Iterative eigensolvers may be preferred over direct methods for this task since their complexity is a function of the desired fraction of the spectrum. In addition, they generally rely on highly optimized and scalable kernels such as matrix-vector multiplications that leverage the massive parallelism and the computational power of modern exascale systems. \textit{Chebyshev Accelerated Subspace iteration Eigensolver} (ChASE) is able to compute several thousands of the most extreme eigenpairs of dense hermitian matrices with proven scalability over massive parallel accelerated clusters. This work presents an extension of ChASE to solve for a portion of the spectrum of pseudo-hermitian Hamiltonians as they appear in the treatment of excitonic materials. The new pseudo-hermitian solver achieves similar convergence and performance as the hermitian one. By exploiting the numerical structure and spectral properties of the Hamiltonian matrix, we propose an oblique variant of Rayleigh-Ritz projection featuring quadratic convergence of the Ritz-values with no explicit construction of the dual basis set. Additionally, we introduce a parallel implementation of the recursive matrix-product operation appearing in the Chebyshev filter with limited amount of global communications. Our development is supported by a full numerical analysis and experimental tests.

Large language models (LLMs) have achieved impressive performance across natural language tasks and are increasingly deployed in real-world applications. Despite extensive safety alignment efforts, recent studies show that such alignment is often shallow and remains vulnerable to jailbreak attacks. Existing defense mechanisms, including decoding-based constraints and post-hoc content detectors, struggle against sophisticated jailbreaks, often intervening robust detection or excessively degrading model utility. In this work, we examine the decoding process of LLMs and make a key observation: even when successfully jailbroken, models internally exhibit latent safety-related signals during generation. However, these signals are overridden by the model's drive for fluent continuation, preventing timely self-correction or refusal. Building on this observation, we propose a simple yet effective approach that explicitly surfaces and leverages these latent safety signals for early detection of unsafe content during decoding. Experiments across diverse jailbreak attacks demonstrate that our approach significantly enhances safety, while maintaining low over-refusal rates on benign inputs and preserving response quality. Our results suggest that activating intrinsic safety-awareness during decoding offers a promising and complementary direction for defending against jailbreak attacks. Code is available at: https://github.com/zyz13590/SafeProbing.

The predominance of machine learning models in many spheres of human activity has led to a growing demand for their transparency. The transparency of models makes it possible to discern some factors, such as security or non-discrimination. In this paper, we propose a mixture of transparent local models as an alternative solution for designing interpretable (or transparent) models. Our approach is designed for the situations where a simple and transparent function is suitable for modeling the label of instances in some localities/regions of the input space, but may change abruptly as we move from one locality to another. Consequently, the proposed algorithm is to learn both the transparent labeling function and the locality of the input space where the labeling function achieves a small risk in its assigned locality. By using a new multi-predictor (and multi-locality) loss function, we established rigorous PAC-Bayesian risk bounds for the case of binary linear classification problem and that of linear regression. In both cases, synthetic data sets were used to illustrate how the learning algorithms work. The results obtained from real data sets highlight the competitiveness of our approach compared to other existing methods as well as certain opaque models. Keywords: PAC-Bayes, risk bounds, local models, transparent models, mixtures of local transparent models.

The evolution of User Interface design has emphasized the need for efficient, reusable, and editable components to ensure an efficient design process. This research introduces CoGen, a system that uses machine learning techniques to generate reusable UI components directly in Figma, one of the most popular UI design tools. Addressing gaps in current systems, CoGen focuses on creating atomic components such as buttons, labels, and input fields using structured JSON and natural language prompts. The project integrates Figma API data extraction, Seq2Seq models, and fine-tuned T5 transformers for component generation. The key results demonstrate the efficiency of the T5 model in prompt generation, with an accuracy of 98% and a BLEU score of 0.2668, which ensures the mapping of JSON to descriptive prompts. For JSON creation, CoGen achieves a success rate of up to 100% in generating simple JSON outputs for specified component types.

Large Language Models (LLMs) are increasingly deployed in human-centric applications, yet they often fail to provide substantive emotional support. While Reinforcement Learning (RL) has been utilized to enhance empathy of LLMs, existing reward models typically evaluate empathy from a single perspective, overlooking the inherently bidirectional interaction nature of empathy between the supporter and seeker as defined by Empathy Cycle theory. To address this limitation, we propose Psychology-grounded Empathetic Reward Modeling (PERM). PERM operationalizes empathy evaluation through a bidirectional decomposition: 1) Supporter perspective, assessing internal resonation and communicative expression; 2) Seeker perspective, evaluating emotional reception. Additionally, it incorporates a bystander perspective to monitor overall interaction quality. Extensive experiments on a widely-used emotional intelligence benchmark and an industrial daily conversation dataset demonstrate that PERM outperforms state-of-the-art baselines by over 10\%. Furthermore, a blinded user study reveals a 70\% preference for our approach, highlighting its efficacy in generating more empathetic responses. Our code, dataset, and models are available at https://github.com/ZhengWwwq/PERM.

Directed acyclic graphical (DAG) models are a powerful tool for representing causal relationships among jointly distributed random variables, especially concerning data from across different experimental settings. However, it is not always practical or desirable to estimate a causal model at the granularity of given features in a particular dataset. There is a growing body of research on causal abstraction to address such problems. We contribute to this line of research by (i) providing novel graphical identifiability results for practically-relevant interventional settings, (ii) proposing an efficient, provably consistent algorithm for directly learning abstract causal graphs from interventional data with unknown intervention targets, and (iii) uncovering theoretical insights about the lattice structure of the underlying search space, with connections to the field of causal discovery more generally. As proof of concept, we apply our algorithm on synthetic and real datasets with known ground truths, including measurements from a controlled physical system with interacting light intensity and polarization.

The rapid evolution of Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) has produced substantial gains in reasoning, perception, and generative capability across language and vision. However, whether these advances yield commensurate improvements in safety remains unclear, in part due to fragmented evaluation practices limited to single modalities or threat models. In this report, we present an integrated safety evaluation of 7 frontier models: GPT-5.2, Gemini 3 Pro, Qwen3-VL, Doubao 1.8, Grok 4.1 Fast, Nano Banana Pro, and Seedream 4.5. We evaluate each model across language, vision-language, and image generation settings using a unified protocol that integrates benchmark evaluation, adversarial evaluation, multilingual evaluation, and compliance evaluation. Aggregating our evaluations into safety leaderboards and model safety profiles across multiple evaluation modes reveals a sharply heterogeneous safety landscape. While GPT-5.2 demonstrates consistently strong and balanced safety performance across evaluations, other models exhibit pronounced trade-offs among benchmark safety, adversarial alignment, multilingual generalization, and regulatory compliance. Both language and vision-language modalities show significant vulnerability under adversarial evaluation, with all models degrading substantially despite strong results on standard benchmarks. Text-to-image models achieve relatively stronger alignment in regulated visual risk categories, yet remain brittle under adversarial or semantically ambiguous prompts. Overall, these results show that safety in frontier models is inherently multidimensional--shaped by modality, language, and evaluation scheme, underscoring the need for standardized safety evaluations to accurately assess real-world risk and guide responsible model development and deployment.

The practice of fine-tuning Large Language Models (LLMs) has achieved state-of-the-art performance on specialized tasks, yet diagnosing why these models become brittle and fail to generalize remains a critical open problem. To address this, we introduce and apply a multi-layered diagnostic framework to a cross-architectural study. We fine-tune Llama 3.1 8B, Gemma 2 9B, and Mistral models on a high-stakes phishing detection task and use SHAP analysis and mechanistic interpretability to uncover the root causes of their generalization failures. Our investigation reveals three critical findings: (1) Generalization is driven by a powerful synergy between architecture and data diversity. The Gemma 2 9B model achieves state-of-the-art performance (>91\% F1), but only when trained on a stylistically diverse ``generalist'' dataset. (2) Generalization is highly architecture-dependent. We diagnose a specific failure mode in Llama 3.1 8B, which performs well on a narrow domain but cannot integrate diverse data, leading to a significant performance drop. (3) Some architectures are inherently more generalizable. The Mistral model proves to be a consistent and resilient performer across multiple training paradigms. By pinpointing the flawed heuristics responsible for these failures, our work provides a concrete methodology for diagnosing and understanding generalization failures, underscoring that reliable AI requires deep validation of the interplay between architecture, data, and training strategy.

As AI agents become increasingly autonomous, widely deployed in consequential contexts, and efficacious in bringing about real-world impacts, ensuring that their decisions are not only instrumentally effective but also normatively aligned has become critical. We introduce a neuro-symbolic reason-based containment architecture, Governor for Reason-Aligned ContainmEnt (GRACE), that decouples normative reasoning from instrumental decision-making and can contain AI agents of virtually any design. GRACE restructures decision-making into three modules: a Moral Module (MM) that determines permissible macro actions via deontic logic-based reasoning; a Decision-Making Module (DMM) that encapsulates the target agent while selecting instrumentally optimal primitive actions in accordance with derived macro actions; and a Guard that monitors and enforces moral compliance. The MM uses a reason-based formalism providing a semantic foundation for deontic logic, enabling interpretability, contestability, and justifiability. Its symbolic representation enriches the DMM's informational context and supports formal verification and statistical guarantees of alignment enforced by the Guard. We demonstrate GRACE on an example of a LLM therapy assistant, showing how it enables stakeholders to understand, contest, and refine agent behavior.

Cognitive Radio (CR) systems, which dynamically adapt to changing spectrum environments, could benefit significantly from advancements in machine learning technologies. These systems can be enhanced in terms of spectral efficiency, robustness, and security through innovative approaches such as the use of Transformer models. This work investigates the application of Transformer models, specifically the GPT-2 architecture, to generate novel modulation schemes for wireless communications. By training a GPT-2 model on a dataset of existing modulation formulas, new modulation schemes has been created. These generated schemes are then compared to traditional methods using key performance metrics such as Signal-to-Noise Ratio (SNR) and Power Spectrum Density (PSD). The results show that Transformer-generated modulation schemes can achieve performance comparable to, and in some cases outperforming, traditional methods. This demonstrates that advanced CR systems could greatly benefit from the implementation of Transformer models, leading to more efficient, robust, and secure communication systems.

While the automatic evaluation of omni-modal large models (OLMs) is essential, assessing empathy remains a significant challenge due to its inherent affectivity. To investigate this challenge, we introduce AEQ-Bench (Audio Empathy Quotient Benchmark), a novel benchmark to systematically assess two core empathetic capabilities of OLMs: (i) generating empathetic responses by comprehending affective cues from multi-modal inputs (audio + text), and (ii) judging the empathy of audio responses without relying on text transcription. Compared to existing benchmarks, AEQ-Bench incorporates two novel settings that vary in context specificity and speech tone. Comprehensive assessment across linguistic and paralinguistic metrics reveals that (1) OLMs trained with audio output capabilities generally outperformed models with text-only outputs, and (2) while OLMs align with human judgments for coarse-grained quality assessment, they remain unreliable for evaluating fine-grained paralinguistic expressiveness.

Online high-definition (HD) map construction is an essential part of a safe and robust end-to-end autonomous driving (AD) pipeline. Onboard camera-based approaches suffer from limited depth perception and degraded accuracy due to occlusion. In this work, we propose SatMap, an online vectorized HD map estimation method that integrates satellite maps with multi-view camera observations and directly predicts a vectorized HD map for downstream prediction and planning modules. Our method leverages lane-level semantics and texture from satellite imagery captured from a Bird's Eye View (BEV) perspective as a global prior, effectively mitigating depth ambiguity and occlusion. In our experiments on the nuScenes dataset, SatMap achieves 34.8% mAP performance improvement over the camera-only baseline and 8.5% mAP improvement over the camera-LiDAR fusion baseline. Moreover, we evaluate our model in long-range and adverse weather conditions to demonstrate the advantages of using a satellite prior map. Source code will be available at https://iv.ee.hm.edu/satmap/.

Model counting of Disjunctive Normal Form (DNF) formulas is a critical problem in applications such as probabilistic inference and network reliability. For example, it is often used for query evaluation in probabilistic databases. Due to the computational intractability of exact DNF counting, there has been a line of research into a variety of approximation algorithms. These include Monte Carlo approaches such as the classical algorithms of Karp, Luby, and Madras (1989), as well as methods based on hashing (Soos et al. 2023), and heuristic approximations based on Neural Nets (Abboud, Ceylan, and Lukasiewicz 2020). We develop a new Monte Carlo approach with an adaptive stopping rule and short-circuit formula evaluation. We prove it achieves Probably Approximately Correct (PAC) learning bounds and is asymptotically more efficient than the previous methods. We also show experimentally that it out-performs prior algorithms by orders of magnitude, and can scale to much larger problems with millions of variables.

We prove that there is no preferential voting method satisfying the Condorcet winner and loser criteria, positive involvement (if a candidate $x$ wins in an initial preference profile, then adding a voter who ranks $x$ uniquely first cannot cause $x$ to lose), and resolvability (if $x$ initially ties for winning, then $x$ can be made the unique winner by adding a single voter). In a previous note, we proved an analogous result assuming an additional axiom of ordinal margin invariance, which we now show is unnecessary for an impossibility theorem, at least if the desired voting method is defined for five-candidate elections.

As Large Language Models (LLMs) increasingly operate as Deep Research (DR) Agents capable of autonomous investigation and information synthesis, reliable evaluation of their task performance has become a critical bottleneck. Current benchmarks predominantly rely on static datasets, which suffer from several limitations: limited task generality, temporal misalignment, and data contamination. To address these, we introduce DR-Arena, a fully automated evaluation framework that pushes DR agents to their capability limits through dynamic investigation. DR-Arena constructs real-time Information Trees from fresh web trends to ensure the evaluation rubric is synchronized with the live world state, and employs an automated Examiner to generate structured tasks testing two orthogonal capabilities: Deep reasoning and Wide coverage. DR-Arena further adopts Adaptive Evolvement Loop, a state-machine controller that dynamically escalates task complexity based on real-time performance, demanding deeper deduction or wider aggregation until a decisive capability boundary emerges. Experiments with six advanced DR agents demonstrate that DR-Arena achieves a Spearman correlation of 0.94 with the LMSYS Search Arena leaderboard. This represents the state-of-the-art alignment with human preferences without any manual efforts, validating DR-Arena as a reliable alternative for costly human adjudication.

This note introduces Projected Microbatch Accumulation (PROMA), a proximal policy update method for large language model fine-tuning. PROMA accumulates policy gradients across microbatches by projecting out sequence-wise gradient components before microbatch aggregation. The projection is applied layer-wise during the backward pass, enabling efficient implementation without additional forward or backward passes. Empirically, PROMA enforces tighter control of local KL divergence than GRPO, resulting in more stable policy learning. Unlike PPO and GRPO, PROMA achieves proximal updates without inducing entropy collapse and does not rely on a reference policy or likelihood-ratio clipping.

Large language models are increasingly used for code generation and debugging, but their outputs can still contain bugs, that originate from training data. Distinguishing whether an LLM prefers correct code, or a familiar incorrect version might be influenced by what it's been exposed to during training. We introduce an exposure-aware evaluation framework that quantifies how prior exposure to buggy versus fixed code influences a model's preference. Using the ManySStuBs4J benchmark, we apply Data Portraits for membership testing on the Stack-V2 corpus to estimate whether each buggy and fixed variant was seen during training. We then stratify examples by exposure and compare model preference using code completion as well as multiple likelihood-based scoring metrics We find that most examples (67%) have neither variant in the training data, and when only one is present, fixes are more frequently present than bugs. In model generations, models reproduce buggy lines far more often than fixes, with bug-exposed examples amplifying this tendency and fix-exposed examples showing only marginal improvement. In likelihood scoring, minimum and maximum token-probability metrics consistently prefer the fixed code across all conditions, indicating a stable bias toward correct fixes. In contrast, metrics like the Gini coefficient reverse preference when only the buggy variant was seen. Our results indicate that exposure can skew bug-fix evaluations and highlight the risk that LLMs may propagate memorised errors in practice.

With grid operators confronting rising uncertainty from renewable integration and a broader push toward electrification, Demand-Side Management (DSM) -- particularly Demand Response (DR) -- has attracted significant attention as a cost-effective mechanism for balancing modern electricity systems. Unprecedented volumes of consumption data from a continuing global deployment of smart meters enable consumer segmentation based on real usage behaviours, promising to inform the design of more effective DSM and DR programs. However, existing clustering-based segmentation methods insufficiently reflect the behavioural diversity of consumers, often relying on rigid temporal alignment, and faltering in the presence of anomalies, missing data, or large-scale deployments. To address these challenges, we propose a novel two-stage clustering framework -- Clustered Representations Optimising Consumer Segmentation (CROCS). In the first stage, each consumer's daily load profiles are clustered independently to form a Representative Load Set (RLS), providing a compact summary of their typical diurnal consumption behaviours. In the second stage, consumers are clustered using the Weighted Sum of Minimum Distances (WSMD), a novel set-to-set measure that compares RLSs by accounting for both the prevalence and similarity of those behaviours. Finally, community detection on the WSMD-induced graph reveals higher-order prototypes that embody the shared diurnal behaviours defining consumer groups, enhancing the interpretability of the resulting clusters. Extensive experiments on both synthetic and real Australian smart meter datasets demonstrate that CROCS captures intra-consumer variability, uncovers both synchronous and asynchronous behavioural similarities, and remains robust to anomalies and missing data, while scaling efficiently through natural parallelisation. These results...

Communication overhead is a critical challenge in federated learning, particularly in bandwidth-constrained networks. Although many methods have been proposed to reduce communication overhead, most focus solely on compressing individual gradients, overlooking the temporal correlations among them. Prior studies have shown that gradients exhibit spatial correlations, typically reflected in low-rank structures. Through empirical analysis, we further observe a strong temporal correlation between client gradients across adjacent rounds. Based on these observations, we propose GradESTC, a compression technique that exploits both spatial and temporal gradient correlations. GradESTC exploits spatial correlations to decompose each full gradient into a compact set of basis vectors and corresponding combination coefficients. By exploiting temporal correlations, only a small portion of the basis vectors need to be dynamically updated in each round. GradESTC significantly reduces communication overhead by transmitting lightweight combination coefficients and a limited number of updated basis vectors instead of the full gradients. Extensive experiments show that, upon reaching a target accuracy level near convergence, GradESTC reduces uplink communication by an average of 39.79% compared to the strongest baseline, while maintaining comparable convergence speed and final accuracy to uncompressed FedAvg. By effectively leveraging spatio-temporal gradient structures, GradESTC offers a practical and scalable solution for communication-efficient federated learning.

Domain-specific knowledge graphs (DKGs) often lack coverage compared to general knowledge graphs (GKGs). To address this, we introduce Domain-specific Knowledge Graph Fusion (DKGF), a novel task that enriches DKGs by integrating relevant facts from GKGs. DKGF faces two key challenges: high ambiguity in domain relevance and misalignment in knowledge granularity across graphs. We propose ExeFuse, a simple yet effective Fact-as-Program paradigm. It treats each GKG fact as a latent semantic program, maps abstract relations to granularity-aware operators, and verifies domain relevance via program executability on the target DKG. This unified probabilistic framework jointly resolves relevance and granularity issues. We construct two benchmarks, DKGF(W-I) and DKGF(Y-I), with 21 evaluation configurations. Extensive experiments validate the task's importance and our model's effectiveness, providing the first standardized testbed for DKGF.

Variational Quantum Algorithms (VQAs) are critically threatened by the Barren Plateau (BP) phenomenon. In this work, we introduce the H-EFT Variational Ansatz (H-EFT-VA), an architecture inspired by Effective Field Theory (EFT). By enforcing a hierarchical "UV-cutoff" on initialization, we theoretically restrict the circuit's state exploration, preventing the formation of approximate unitary 2-designs. We provide a rigorous proof that this localization guarantees an inverse-polynomial lower bound on the gradient variance: $Var[\partial θ] \in Ω(1/poly(N))$. Crucially, unlike approaches that avoid BPs by limiting entanglement, we demonstrate that H-EFT-VA maintains volume-law entanglement and near-Haar purity, ensuring sufficient expressibility for complex quantum states. Extensive benchmarking across 16 experiments -- including Transverse Field Ising and Heisenberg XXZ models -- confirms a 109x improvement in energy convergence and a 10.7x increase in ground-state fidelity over standard Hardware-Efficient Ansatze (HEA), with a statistical significance of $p < 10^{-88}$.

As hubs of human activity, urban surfaces consist of a wealth of semantic entities. Segmenting these various entities from satellite imagery is crucial for a range of downstream applications. Current advanced segmentation models can reliably segment entities defined by physical attributes (e.g., buildings, water bodies) but still struggle with socially defined categories (e.g., schools, parks). In this work, we achieve socio-semantic segmentation by vision-language model reasoning. To facilitate this, we introduce the Urban Socio-Semantic Segmentation dataset named SocioSeg, a new resource comprising satellite imagery, digital maps, and pixel-level labels of social semantic entities organized in a hierarchical structure. Additionally, we propose a novel vision-language reasoning framework called SocioReasoner that simulates the human process of identifying and annotating social semantic entities via cross-modal recognition and multi-stage reasoning. We employ reinforcement learning to optimize this non-differentiable process and elicit the reasoning capabilities of the vision-language model. Experiments demonstrate our approach's gains over state-of-the-art models and strong zero-shot generalization. Our dataset and code are available in https://github.com/AMAP-ML/SocioReasoner.

We present DeFlow, a decoupled offline RL framework that leverages flow matching to faithfully capture complex behavior manifolds. Optimizing generative policies is computationally prohibitive, typically necessitating backpropagation through ODE solvers. We address this by learning a lightweight refinement module within an explicit, data-derived trust region of the flow manifold, rather than sacrificing the iterative generation capability via single-step distillation. This way, we bypass solver differentiation and eliminate the need for balancing loss terms, ensuring stable improvement while fully preserving the flow's iterative expressivity. Empirically, DeFlow achieves superior performance on the challenging OGBench benchmark and demonstrates efficient offline-to-online adaptation.

With advancements in deep learning (DL) and computer vision techniques, the field of chart understanding is evolving rapidly. In particular, multimodal large language models (MLLMs) are proving to be efficient and accurate in understanding charts. To accurately measure the performance of MLLMs, the research community has developed multiple datasets to serve as benchmarks. By examining these datasets, we found that they are all limited to a small set of chart types. To bridge this gap, we propose the ChartComplete dataset. The dataset is based on a chart taxonomy borrowed from the visualization community, and it covers thirty different chart types. The dataset is a collection of classified chart images and does not include a learning signal. We present the ChartComplete dataset as is to the community to build upon it.

A model that avoids stereotypes in a lab benchmark may not avoid them in deployment. We show that measured bias shifts dramatically when prompts mention different places, times, or audiences -- no adversarial prompting required. We introduce Contextual StereoSet, a benchmark that holds stereotype content fixed while systematically varying contextual framing. Testing 13 models across two protocols, we find striking patterns: anchoring to 1990 (vs. 2030) raises stereotype selection in all models tested on this contrast (p<0.05); gossip framing raises it in 5 of 6 full-grid models; out-group observer framing shifts it by up to 13 percentage points. These effects replicate in hiring, lending, and help-seeking vignettes. We propose Context Sensitivity Fingerprints (CSF): a compact profile of per-dimension dispersion and paired contrasts with bootstrap CIs and FDR correction. Two evaluation tracks support different use cases -- a 360-context diagnostic grid for deep analysis and a budgeted protocol covering 4,229 items for production screening. The implication is methodological: bias scores from fixed-condition tests may not generalize.This is not a claim about ground-truth bias rates; it is a stress test of evaluation robustness. CSF forces evaluators to ask, "Under what conditions does bias appear?" rather than "Is this model biased?" We release our benchmark, code, and results.

Dimensionality reduction is a powerful technique for revealing structure and potential clusters in data. However, as the axes are complex, non-linear combinations of features, they often lack semantic interpretability. Existing visual analytics (VA) methods support cluster interpretation through feature comparison and interactive exploration, but they require technical expertise and intense human effort. We present \textit{LangLasso}, a novel method that complements VA approaches through interactive, natural language descriptions of clusters using large language models (LLMs). It produces human-readable descriptions that make cluster interpretation accessible to non-experts and allow integration of external contextual knowledge beyond the dataset. We systematically evaluate the reliability of these explanations and demonstrate that \langlasso provides an effective first step for engaging broader audiences in cluster interpretation. The tool is available at https://langlasso.vercel.app

Although the Gradient Boosted Decision Trees (GBDTs) dominate industrial tabular applications, upgrading legacy models in high-concurrency production environments still faces prohibitive retraining costs and systemic risks. To address this problem, we present NSR-Boost, a neuro-symbolic residual boosting framework designed specifically for industrial scenarios. Its core advantage lies in being "non-intrusive". It treats the legacy model as a frozen model and performs targeted repairs on "hard regions" where predictions fail. The framework comprises three key stages: first, finding hard regions through residuals, then generating interpretable experts by generating symbolic code structures using Large Language Model (LLM) and fine-tuning parameters using Bayesian optimization, and finally dynamically integrating experts with legacy model output through a lightweight aggregator. We report on the successful deployment of NSR-Boost within the core financial risk control system at Qfin Holdings. This framework not only significantly outperforms state-of-the-art (SOTA) baselines across six public datasets and one private dataset, more importantly, shows excellent performance gains on real-world online data. In conclusion, it effectively captures long-tail risks missed by traditional models and offers a safe, low-cost evolutionary paradigm for industry.

Surgical planning integrates visual perception, long-horizon reasoning, and procedural knowledge, yet it remains unclear whether current evaluation protocols reliably assess vision-language models (VLMs) in safety-critical settings. Motivated by a goal-oriented view of surgical planning, we define planning correctness via phase-goal satisfiability, where plan validity is determined by expert-defined surgical rules. Based on this definition, we introduce a multicentric meta-evaluation benchmark with valid procedural variations and invalid plans containing order and content errors. Using this benchmark, we show that sequence similarity metrics systematically misjudge planning quality, penalizing valid plans while failing to identify invalid ones. We therefore adopt a rule-based goal-satisfiability metric as a high-precision meta-evaluation reference to assess Video-LLMs under progressively constrained settings, revealing failures due to perception errors and under-constrained reasoning. Structural knowledge consistently improves performance, whereas semantic guidance alone is unreliable and benefits larger models only when combined with structural constraints.

Modal methods are a long-standing approach to physical modelling synthesis. Extensions to nonlinear problems are possible, including the case of a high-amplitude vibration of a string. A modal decomposition leads to a densely coupled nonlinear system of ordinary differential equations. Recent work in scalar auxiliary variable techniques has enabled construction of explicit and stable numerical solvers for such classes of nonlinear systems. On the other hand, machine learning approaches (in particular neural ordinary differential equations) have been successful in modelling nonlinear systems automatically from data. In this work, we examine how scalar auxiliary variable techniques can be combined with neural ordinary differential equations to yield a stable differentiable model capable of learning nonlinear dynamics. The proposed approach leverages the analytical solution for linear vibration of system's modes so that physical parameters of a system remain easily accessible after the training without the need for a parameter encoder in the model architecture. As a proof of concept, we generate synthetic data for the nonlinear transverse vibration of a string and show that the model can be trained to reproduce the nonlinear dynamics of the system. Sound examples are presented.

Artificial intelligence (AI) agents are increasingly used in a variety of domains to automate tasks, interact with users, and make decisions based on data inputs. Ensuring that AI agents perform only authorized actions and handle inputs appropriately is essential for maintaining system integrity and preventing misuse. In this study, we introduce the AgentGuardian, a novel security framework that governs and protects AI agent operations by enforcing context-aware access-control policies. During a controlled staging phase, the framework monitors execution traces to learn legitimate agent behaviors and input patterns. From this phase, it derives adaptive policies that regulate tool calls made by the agent, guided by both real-time input context and the control flow dependencies of multi-step agent actions. Evaluation across two real-world AI agent applications demonstrates that AgentGuardian effectively detects malicious or misleading inputs while preserving normal agent functionality. Moreover, its control-flow-based governance mechanism mitigates hallucination-driven errors and other orchestration-level malfunctions.

Ontological Knowledge Bases (OKBs) play a vital role in structuring domain-specific knowledge and serve as a foundation for effective knowledge management systems. However, their traditional manual development poses significant challenges related to scalability, consistency, and adaptability. Recent advancements in Generative AI, particularly Large Language Models (LLMs), offer promising solutions for automating and enhancing OKB development. This paper introduces a structured, iterative methodology leveraging LLMs to optimize knowledge acquisition, automate ontology artifact generation, and enable continuous refinement cycles. We demonstrate this approach through a detailed case study focused on developing a user context profile ontology within the vehicle sales domain. Key contributions include significantly accelerated ontology construction processes, improved ontological consistency, effective bias mitigation, and enhanced transparency in the ontology engineering process. Our findings highlight the transformative potential of integrating LLMs into ontology development, notably improving scalability, integration capabilities, and overall efficiency in knowledge management systems.

Futrell and Mahowald claim LMs "serve as model systems", but an assessment at each of Marr's three levels suggests the claim is clearly not true at the implementation level, poorly motivated at the algorithmic-representational level, and problematic at the computational theory level. LMs are good candidates as tools; calling them cognitive models overstates the case and unnecessarily feeds LLM hype.

We study tabular reinforcement learning problems with multiple steps of lookahead information. Before acting, the learner observes $\ell$ steps of future transition and reward realizations: the exact state the agent would reach and the rewards it would collect under any possible course of action. While it has been shown that such information can drastically boost the value, finding the optimal policy is NP-hard, and it is common to apply one of two tractable heuristics: processing the lookahead in chunks of predefined sizes ('fixed batching policies'), and model predictive control. We first illustrate the problems with these two approaches and propose utilizing the lookahead in adaptive (state-dependent) batches; we refer to such policies as adaptive batching policies (ABPs). We derive the optimal Bellman equations for these strategies and design an optimistic regret-minimizing algorithm that enables learning the optimal ABP when interacting with unknown environments. Our regret bounds are order-optimal up to a potential factor of the lookahead horizon $\ell$, which can usually be considered a small constant.

Aligning Large Language Models (LLMs) with human preferences is critical, yet traditional fine-tuning methods are computationally expensive and inflexible. While test-time alignment offers a promising alternative, existing approaches often rely on distorted trajectory-level signals or inefficient sampling, fundamentally capping performance and failing to preserve the generative diversity of the base model. This paper introduces LLMdoctor, a novel framework for efficient test-time alignment that operates via a patient-doctor paradigm. It integrates token-level reward acquisition with token-level flow-guided preference optimization (TFPO) to steer a large, frozen patient LLM with a smaller, specialized doctor model. Unlike conventional methods that rely on trajectory-level rewards, LLMdoctor first extracts fine-grained, token-level preference signals from the patient model's behavioral variations. These signals then guide the training of the doctor model via TFPO, which establishes flow consistency across all subtrajectories, enabling precise token-by-token alignment while inherently preserving generation diversity. Extensive experiments demonstrate that LLMdoctor significantly outperforms existing test-time alignment methods and even surpasses the performance of full fine-tuning approaches like DPO.

Privacy policies help inform people about organisations' personal data processing practices, covering different aspects such as data collection, data storage, and sharing of personal data with third parties. Privacy policies are often difficult for people to fully comprehend due to the lengthy and complex legal language used and inconsistent practices across different sectors and organisations. To help conduct automated and large-scale analyses of privacy policies, many researchers have studied applications of machine learning and natural language processing techniques, including large language models (LLMs). While a limited number of prior studies utilised LLMs for extracting personal data flows from privacy policies, our approach builds on this line of work by combining LLMs with retrieval-augmented generation (RAG) and a customised knowledge base derived from existing studies. This paper presents the development of LADFA, an end-to-end computational framework, which can process unstructured text in a given privacy policy, extract personal data flows and construct a personal data flow graph, and conduct analysis of the data flow graph to facilitate insight discovery. The framework consists of a pre-processor, an LLM-based processor, and a data flow post-processor. We demonstrated and validated the effectiveness and accuracy of the proposed approach by conducting a case study that involved examining ten selected privacy policies from the automotive industry. Moreover, it is worth noting that LADFA is designed to be flexible and customisable, making it suitable for a range of text-based analysis tasks beyond privacy policy analysis.

Recent advances in synthetic data generation have shown that compact language models can be trained effectively when the underlying corpus is structurally controlled and linguistically coherent. However, for morphologically rich and computationally under-resourced languages such as Romanian, there is still no openly documented, end-to-end pipeline that unifies tokenizer design, preprocessing, pretraining, compression, evaluation, and large-scale synthetic data generation in a reproducible framework. Building on TF1, a three-million-story English fable dataset, and TF2, which extends TF1 through high-quality Romanian translations, we introduce TF3-RO, a Romanian-centric language modeling pipeline spanning tokenizer training, from-scratch model development, and Romanian-native dataset generation. TF3-RO constructs Romanian-specific BPE and Unigram tokenizers from a linguistically informed corpus to mitigate token inflation induced by Romanian morphology. Using long-sequence packed training, we pretrain a 51.65M-parameter LLaMA-style Transformer entirely from scratch. The model is subsequently optimized through quantization, structured pruning, and logit-based knowledge distillation, yielding a compact 26.45M-parameter student model with tied embeddings and strong deployment characteristics. Using this distilled model, TF3-RO generates three million Romanian-native synthetic fables via a controlled combinatorial prompting framework. Across all stages, the pipeline integrates a comprehensive evaluation suite combining intrinsic metrics, Romanian agreement probes, entity coherence, rule-based grammar checking, and LLM-based assessment. TF3-RO provides a reproducible and linguistically grounded framework for training compact Romanian language models and producing large-scale synthetic narrative corpora.

Offline Reinforcement Learning (RL) enables policy optimization from static datasets but is inherently vulnerable to backdoor attacks. Existing attack strategies typically struggle against safety-constrained algorithms (e.g., CQL) due to inefficient random poisoning and the use of easily detectable Out-of-Distribution (OOD) triggers. In this paper, we propose CS-GBA (Critical Sample-based Gradient-guided Backdoor Attack), a novel framework designed to achieve high stealthiness and destructiveness under a strict budget. Leveraging the theoretical insight that samples with high Temporal Difference (TD) errors are pivotal for value function convergence, we introduce an adaptive Critical Sample Selection strategy that concentrates the attack budget on the most influential transitions. To evade OOD detection, we propose a Correlation-Breaking Trigger mechanism that exploits the physical mutual exclusivity of state features (e.g., 95th percentile boundaries) to remain statistically concealed. Furthermore, we replace the conventional label inversion with a Gradient-Guided Action Generation mechanism, which searches for worst-case actions within the data manifold using the victim Q-network's gradient. Empirical results on D4RL benchmarks demonstrate that our method significantly outperforms state-of-the-art baselines, achieving high attack success rates against representative safety-constrained algorithms with a minimal 5% poisoning budget, while maintaining the agent's performance in clean environments.

Automatic Question Generation (QG) often produces outputs with critical defects, such as factual hallucinations and answer mismatches. However, existing evaluation methods, including LLM-based evaluators, mainly adopt a black-box and holistic paradigm without explicit error modeling, leading to the neglect of such defects and overestimation of question quality. To address this issue, we propose ErrEval, a flexible and Error-aware Evaluation framework that enhances QG evaluation through explicit error diagnostics. Specifically, ErrEval reformulates evaluation as a two-stage process of error diagnosis followed by informed scoring. At the first stage, a lightweight plug-and-play Error Identifier detects and categorizes common errors across structural, linguistic, and content-related aspects. These diagnostic signals are then incorporated as explicit evidence to guide LLM evaluators toward more fine-grained and grounded judgments. Extensive experiments on three benchmarks demonstrate the effectiveness of ErrEval, showing that incorporating explicit diagnostics improves alignment with human judgments. Further analyses confirm that ErrEval effectively mitigates the overestimation of low-quality questions.

Discrete diffusion models have recently emerged as a promising alternative to the autoregressive approach for generating discrete sequences. Sample generation via gradual denoising or demasking processes allows them to capture hierarchical non-sequential interdependencies in the data. These custom processes, however, do not assume a flexible control over the distribution of generated samples. We propose Discrete Feynman-Kac Correctors, a framework that allows for controlling the generated distribution of discrete masked diffusion models at inference time. We derive Sequential Monte Carlo (SMC) algorithms that, given a trained discrete diffusion model, control the temperature of the sampled distribution (i.e. perform annealing), sample from the product of marginals of several diffusion processes (e.g. differently conditioned processes), and sample from the product of the marginal with an external reward function, producing likely samples from the target distribution that also have high reward. Notably, our framework does not require any training of additional models or fine-tuning of the original model. We illustrate the utility of our framework in several applications including: efficient sampling from the annealed Boltzmann distribution of the Ising model, improving the performance of language models for code generation and amortized learning, as well as reward-tilted protein sequence generation.

The advancement of artificial intelligence toward agentic science is currently bottlenecked by the challenge of ultra-long-horizon autonomy, the ability to sustain strategic coherence and iterative correction over experimental cycles spanning days or weeks. While Large Language Models (LLMs) have demonstrated prowess in short-horizon reasoning, they are easily overwhelmed by execution details in the high-dimensional, delayed-feedback environments of real-world research, failing to consolidate sparse feedback into coherent long-term guidance. Here, we present ML-Master 2.0, an autonomous agent that masters ultra-long-horizon machine learning engineering (MLE) which is a representative microcosm of scientific discovery. By reframing context management as a process of cognitive accumulation, our approach introduces Hierarchical Cognitive Caching (HCC), a multi-tiered architecture inspired by computer systems that enables the structural differentiation of experience over time. By dynamically distilling transient execution traces into stable knowledge and cross-task wisdom, HCC allows agents to decouple immediate execution from long-term experimental strategy, effectively overcoming the scaling limits of static context windows. In evaluations on OpenAI's MLE-Bench under 24-hour budgets, ML-Master 2.0 achieves a state-of-the-art medal rate of 56.44%. Our findings demonstrate that ultra-long-horizon autonomy provides a scalable blueprint for AI capable of autonomous exploration beyond human-precedent complexities.

In LLM-based text-to-SQL systems, unanswerable and underspecified user queries may generate not only incorrect text but also executable programs that yield misleading results or violate safety constraints, posing a major barrier to safe deployment. Existing refusal strategies for such queries either rely on output-level instruction following, which is brittle due to model hallucinations, or estimate output uncertainty, which adds complexity and overhead. To address this challenge, we formalize safe refusal in text-to-SQL systems as an answerability-gating problem and propose LatentRefusal, a latent-signal refusal mechanism that predicts query answerability from intermediate hidden activations of a large language model. We introduce the Tri-Residual Gated Encoder, a lightweight probing architecture, to suppress schema noise and amplify sparse, localized cues of question-schema mismatch that indicate unanswerability. Extensive empirical evaluations across diverse ambiguous and unanswerable settings, together with ablation studies and interpretability analyses, demonstrate the effectiveness of the proposed approach and show that LatentRefusal provides an attachable and efficient safety layer for text-to-SQL systems. Across four benchmarks, LatentRefusal improves average F1 to 88.5 percent on both backbones while adding approximately 2 milliseconds of probe overhead.

Recent NLP advances focus primarily on standardized languages, leaving most low-resource dialects under-served especially in Indian scenarios. In India, the issue is particularly important: despite Hindi being the third most spoken language globally (over 600 million speakers), its numerous dialects remain underrepresented. The situation is similar for Odia, which has around 45 million speakers. While some datasets exist which contain standard Hindi and Odia languages, their regional dialects have almost no web presence. We introduce INDIC-DIALECT, a human-curated parallel corpus of 13k sentence pairs spanning 11 dialects and 2 languages: Hindi and Odia. Using this corpus, we construct a multi-task benchmark with three tasks: dialect classification, multiple-choice question (MCQ) answering, and machine translation (MT). Our experiments show that LLMs like GPT-4o and Gemini 2.5 perform poorly on the classification task. While fine-tuned transformer based models pretrained on Indian languages substantially improve performance e.g., improving F1 from 19.6\% to 89.8\% on dialect classification. For dialect to language translation, we find that hybrid AI model achieves highest BLEU score of 61.32 compared to the baseline score of 23.36. Interestingly, due to complexity in generating dialect sentences, we observe that for language to dialect translation the ``rule-based followed by AI" approach achieves best BLEU score of 48.44 compared to the baseline score of 27.59. INDIC-DIALECT thus is a new benchmark for dialect-aware Indic NLP, and we plan to release it as open source to support further work on low-resource Indian dialects.

Large language models can represent a variety of personas but typically default to a helpful Assistant identity cultivated during post-training. We investigate the structure of the space of model personas by extracting activation directions corresponding to diverse character archetypes. Across several different models, we find that the leading component of this persona space is an "Assistant Axis," which captures the extent to which a model is operating in its default Assistant mode. Steering towards the Assistant direction reinforces helpful and harmless behavior; steering away increases the model's tendency to identify as other entities. Moreover, steering away with more extreme values often induces a mystical, theatrical speaking style. We find this axis is also present in pre-trained models, where it primarily promotes helpful human archetypes like consultants and coaches and inhibits spiritual ones. Measuring deviations along the Assistant Axis predicts "persona drift," a phenomenon where models slip into exhibiting harmful or bizarre behaviors that are uncharacteristic of their typical persona. We find that persona drift is often driven by conversations demanding meta-reflection on the model's processes or featuring emotionally vulnerable users. We show that restricting activations to a fixed region along the Assistant Axis can stabilize model behavior in these scenarios -- and also in the face of adversarial persona-based jailbreaks. Our results suggest that post-training steers models toward a particular region of persona space but only loosely tethers them to it, motivating work on training and steering strategies that more deeply anchor models to a coherent persona.

Accurate survival prediction in Non-Small Cell Lung Cancer (NSCLC) requires the integration of heterogeneous clinical, radiological, and histopathological information. While Multimodal Deep Learning (MDL) offers a promises for precision prognosis and survival prediction, its clinical applicability is severely limited by small cohort sizes and the presence of missing modalities, often forcing complete-case filtering or aggressive imputation. In this work, we present a missing-aware multimodal survival framework that integrates Computed Tomography (CT), Whole-Slide Histopathology (WSI) Images, and structured clinical variables for overall survival modeling in unresectable stage II-III NSCLC. By leveraging Foundation Models (FM) for modality-specific feature extraction and a missing-aware encoding strategy, the proposed approach enables intermediate multimodal fusion under naturally incomplete modality profiles. The proposed architecture is resilient to missing modalities by design, allowing the model to utilize all available data without being forced to drop patients during training or inference. Experimental results demonstrate that intermediate fusion consistently outperforms unimodal baselines as well as early and late fusion strategies, with the strongest performance achieved by the fusion of WSI and clinical modalities (73.30 C-index). Further analyses of modality importance reveal an adaptive behavior in which less informative modalities, i.e., CT modality, are automatically down-weighted and contribute less to the final survival prediction.

Recent achievements of vision-language models in end-to-end OCR point to a new avenue for low-loss compression of textual information. This motivates earlier works that render the Transformer's input into images for prefilling, which effectively reduces the number of tokens through visual encoding, thereby alleviating the quadratically increased Attention computations. However, this partial compression fails to save computational or memory costs at token-by-token inference. In this paper, we investigate global context compression, which saves tokens at both prefilling and inference stages. Consequently, we propose VIST2, a novel Transformer that interleaves input text chunks alongside their visual encoding, while depending exclusively on visual tokens in the pre-context to predict the next text token distribution. Around this idea, we render text chunks into sketch images and train VIST2 in multiple stages, starting from curriculum-scheduled pretraining for optical language modeling, followed by modal-interleaved instruction tuning. We conduct extensive experiments using VIST2 families scaled from 0.6B to 8B to explore the training recipe and hyperparameters. With a 4$\times$ compression ratio, the resulting models demonstrate significant superiority over baselines on long writing tasks, achieving, on average, a 3$\times$ speedup in first-token generation, 77% reduction in memory usage, and 74% reduction in FLOPS. Our codes and datasets will be public to support further studies.

Recent advancements in diffusion-based generative priors have enabled visually plausible image compression at extremely low bit rates. However, existing approaches suffer from slow sampling processes and suboptimal bit allocation due to fragmented training paradigms. In this work, we propose Accelerate \textbf{Diff}usion-based Image Compression via \textbf{C}onsistency Prior \textbf{R}efinement (DiffCR), a novel compression framework for efficient and high-fidelity image reconstruction. At the heart of DiffCR is a Frequency-aware Skip Estimation (FaSE) module that refines the $ε$-prediction prior from a pre-trained latent diffusion model and aligns it with compressed latents at different timesteps via Frequency Decoupling Attention (FDA). Furthermore, a lightweight consistency estimator enables fast \textbf{two-step decoding} by preserving the semantic trajectory of diffusion sampling. Without updating the backbone diffusion model, DiffCR achieves substantial bitrate savings (27.2\% BD-rate (LPIPS) and 65.1\% BD-rate (PSNR)) and over $10\times$ speed-up compared to SOTA diffusion-based compression baselines.

Large graph datasets make training graph neural networks (GNNs) computationally costly. Graph condensation methods address this by generating small synthetic graphs that approximate the original data. However, existing approaches rely on clean, supervised labels, which limits their reliability when labels are scarce, noisy, or inconsistent. We propose Pseudo-Labeled Graph Condensation (PLGC), a self-supervised framework that constructs latent pseudo-labels from node embeddings and optimizes condensed graphs to match the original graph's structural and feature statistics -- without requiring ground-truth labels. PLGC offers three key contributions: (1) A diagnosis of why supervised condensation fails under label noise and distribution shift. (2) A label-free condensation method that jointly learns latent prototypes and node assignments. (3) Theoretical guarantees showing that pseudo-labels preserve latent structural statistics of the original graph and ensure accurate embedding alignment. Empirically, across node classification and link prediction tasks, PLGC achieves competitive performance with state-of-the-art supervised condensation methods on clean datasets and exhibits substantial robustness under label noise, often outperforming all baselines by a significant margin. Our findings highlight the practical and theoretical advantages of self-supervised graph condensation in noisy or weakly-labeled environments.

Wearable devices enable continuous, population-scale monitoring of physiological signals, such as photoplethysmography (PPG), creating new opportunities for data-driven clinical assessment. Time-series extrinsic regression (TSER) models increasingly leverage PPG signals to estimate clinically relevant outcomes, including heart rate, respiratory rate, and oxygen saturation. For clinical reasoning and trust, however, single point estimates alone are insufficient: clinicians must also understand whether predictions are stable under physiologically plausible variations and to what extent realistic, attainable changes in physiological signals would meaningfully alter a model's prediction. Counterfactual explanations (CFE) address these "what-if" questions, yet existing time series CFE generation methods are largely restricted to classification, overlook waveform morphology, and often produce physiologically implausible signals, limiting their applicability to continuous biomedical time series. To address these limitations, we introduce EvoMorph, a multi-objective evolutionary framework for generating physiologically plausible and diverse CFE for TSER applications. EvoMorph optimizes morphology-aware objectives defined on interpretable signal descriptors and applies transformations to preserve the waveform structure. We evaluated EvoMorph on three PPG datasets (heart rate, respiratory rate, and oxygen saturation) against a nearest-unlike-neighbor baseline. In addition, in a case study, we evaluated EvoMorph as a tool for uncertainty quantification by relating counterfactual sensitivity to bootstrap-ensemble uncertainty and data-density measures. Overall, EvoMorph enables the generation of physiologically-aware counterfactuals for continuous biomedical signals and supports uncertainty-aware interpretability, advancing trustworthy model analysis for clinical time-series applications.

Enabling Large Language Models (LLMs) to effectively utilize tools in multi-turn interactions is essential for building capable autonomous agents. However, acquiring diverse and realistic multi-turn tool-use data remains a significant challenge. In this work, we propose a novel text-based paradigm. We observe that textual corpora naturally contain rich, multi-step problem-solving experiences, which can serve as an untapped, scalable, and authentic data source for multi-turn tool-use tasks. Based on this insight, we introduce GEM, a data synthesis pipeline that enables the generation and extraction of multi-turn tool-use trajectories from text corpora through a four-stage process: relevance filtering, workflow & tool extraction, trajectory grounding, and complexity refinement. To reduce the computational cost, we further train a specialized Trajectory Synthesizer via supervised fine-tuning. This model distills the complex generation pipeline into an efficient, end-to-end trajectory generator. Experiments demonstrate that our GEM-32B achieve a 16.5% improvement on the BFCL V3 Multi-turn benchmark. Our models partially surpass the performance of models trained on τ - bench (Airline and Retail) in-domain data, highlighting the superior generalization capability derived from our text-based synthesis paradigm. Notably, our Trajectory Synthesizer matches the quality of the full pipeline while significantly reducing inference latency and costs.

We propose Strategy-aware Surprise (SuS), a novel intrinsic motivation framework that uses pre-post prediction mismatch as a novelty signal for exploration in reinforcement learning. Unlike traditional curiosity-driven methods that rely solely on state prediction error, SuS introduces two complementary components: Strategy Stability (SS) and Strategy Surprise (SuS). SS measures consistency in behavioral strategy across temporal steps, while SuS captures unexpected outcomes relative to the agent's current strategy representation. Our combined reward formulation leverages both signals through learned weighting coefficients. We evaluate SuS on mathematical reasoning tasks using large language models, demonstrating significant improvements in both accuracy and solution diversity. Ablation studies confirm that removing either component results in at least 10% performance degradation, validating the synergistic nature of our approach. SuS achieves 17.4% improvement in Pass@1 and 26.4% improvement in Pass@5 compared to baseline methods, while maintaining higher strategy diversity throughout training.

Efficient distillation is a key pathway for converting expensive reasoning capability into deployable efficiency, yet in the frontier regime where the student already has strong reasoning ability, naive continual distillation often yields limited gains or even degradation. We observe a characteristic training phenomenon: even as loss decreases monotonically, all performance metrics can drop sharply at almost the same bottleneck, before gradually recovering. We further uncover a token-level mechanism: confidence bifurcates into steadily increasing Imitation-Anchor Tokens that quickly anchor optimization and other yet-to-learn tokens whose confidence is suppressed until after the bottleneck. And the characteristic that these two types of tokens cannot coexist is the root cause of the failure in continual distillation. To this end, we propose Training-Trajectory-Aware Token Selection (T3S) to reconstruct the training objective at the token level, clearing the optimization path for yet-to-learn tokens. T3 yields consistent gains in both AR and dLLM settings: with only hundreds of examples, Qwen3-8B surpasses DeepSeek-R1 on competitive reasoning benchmarks, Qwen3-32B approaches Qwen3-235B, and T3-trained LLaDA-2.0-Mini exceeds its AR baseline, achieving state-of-the-art performance among all of 16B-scale no-think models.

Modern coding scaffolds turn LLMs into capable software agents, but their ability to follow scaffold-specified instructions remains under-examined, especially when constraints are heterogeneous and persist across interactions. To fill this gap, we introduce OctoBench, which benchmarks scaffold-aware instruction following in repository-grounded agentic coding. OctoBench includes 34 environments and 217 tasks instantiated under three scaffold types, and is paired with 7,098 objective checklist items. To disentangle solving the task from following the rules, we provide an automated observation-and-scoring toolkit that captures full trajectories and performs fine-grained checks. Experiments on eight representative models reveal a systematic gap between task-solving and scaffold-aware compliance, underscoring the need for training and evaluation that explicitly targets heterogeneous instruction following. We release the benchmark to support reproducible benchmarking and to accelerate the development of more scaffold-aware coding agents.

Heart rate variability (HRV) is a pivotal noninvasive marker for autonomic monitoring; however, applying Large Language Models (LLMs) to HRV interpretation is hindered by physiological hallucinations. These include respiratory sinus arrhythmia (RSA) contamination, short-data instability in nonlinear metrics, and the neglect of individualized baselines in favor of population norms. We propose C-GRASP (Clinically-Grounded Reasoning for Affective Signal Processing), a guardrailed RAG-enhanced pipeline that decomposes HRV interpretation into eight traceable reasoning steps. Central to C-GRASP is a Z-score Priority Hierarchy that enforces the weighting of individualized baseline shifts over normative statistics. The system effectively mitigates spectral hallucinations through automated RSA-aware guardrails, preventing contamination of frequency-domain indices. Evaluated on 414 trials from the DREAMER dataset, C-GRASP integrated with high-scale reasoning models (e.g., MedGemma3-thinking) achieved superior performance in 4-class emotion classification (37.3% accuracy) and a Clinical Reasoning Consistency (CRC) score of 69.6%. Ablation studies confirm that the individualized Delta Z-score module serves as the critical logical anchor, preventing the "population bias" common in native LLMs. Ultimately, C-GRASP transitions affective computing from black-box classification to transparent, evidence-based clinical decision support, paving the way for safer AI integration in biomedical engineering.

The rise of AI agent frameworks has introduced agent skills, modular packages containing instructions and executable code that dynamically extend agent capabilities. While this architecture enables powerful customization, skills execute with implicit trust and minimal vetting, creating a significant yet uncharacterized attack surface. We conduct the first large-scale empirical security analysis of this emerging ecosystem, collecting 42,447 skills from two major marketplaces and systematically analyzing 31,132 using SkillScan, a multi-stage detection framework integrating static analysis with LLM-based semantic classification. Our findings reveal pervasive security risks: 26.1% of skills contain at least one vulnerability, spanning 14 distinct patterns across four categories: prompt injection, data exfiltration, privilege escalation, and supply chain risks. Data exfiltration (13.3%) and privilege escalation (11.8%) are most prevalent, while 5.2% of skills exhibit high-severity patterns strongly suggesting malicious intent. We find that skills bundling executable scripts are 2.12x more likely to contain vulnerabilities than instruction-only skills (OR=2.12, p<0.001). Our contributions include: (1) a grounded vulnerability taxonomy derived from 8,126 vulnerable skills, (2) a validated detection methodology achieving 86.7% precision and 82.5% recall, and (3) an open dataset and detection toolkit to support future research. These results demonstrate an urgent need for capability-based permission systems and mandatory security vetting before this attack vector is further exploited.

Supervised convolutional neural networks (CNNs) are widely used to solve imaging inverse problems, achieving state-of-the-art performance in numerous applications. However, despite their empirical success, these methods are poorly understood from a theoretical perspective and often treated as black boxes. To bridge this gap, we analyze trained neural networks through the lens of the Minimum Mean Square Error (MMSE) estimator, incorporating functional constraints that capture two fundamental inductive biases of CNNs: translation equivariance and locality via finite receptive fields. Under the empirical training distribution, we derive an analytic, interpretable, and tractable formula for this constrained variant, termed Local-Equivariant MMSE (LE-MMSE). Through extensive numerical experiments across various inverse problems (denoising, inpainting, deconvolution), datasets (FFHQ, CIFAR-10, FashionMNIST), and architectures (U-Net, ResNet, PatchMLP), we demonstrate that our theory matches the neural networks outputs (PSNR $\gtrsim25$dB). Furthermore, we provide insights into the differences between \emph{physics-aware} and \emph{physics-agnostic} estimators, the impact of high-density regions in the training (patch) distribution, and the influence of other factors (dataset size, patch size, etc).

Traffic flow prediction is a typical spatio-temporal prediction problem and has a wide range of applications. The core challenge lies in modeling the underlying complex spatio-temporal dependencies. Various methods have been proposed, and recent studies show that the modeling of dynamics is useful to meet the core challenge. While handling spatial dependencies and temporal dependencies using separate base model structures may hinder the modeling of spatio-temporal correlations, the modeling of dynamics can bridge this gap. Incorporating spatio-temporal heterogeneity also advances the main goal, since it can extend the parameter space and allow more flexibility. Despite these advances, two limitations persist: 1) the modeling of dynamics is often limited to the dynamics of spatial topology (e.g., adjacency matrix changes), which, however, can be extended to a broader scope; 2) the modeling of heterogeneity is often separated for spatial and temporal dimensions, but this gap can also be bridged by the modeling of dynamics. To address the above limitations, we propose a novel framework for traffic prediction, called Meta Dynamic Graph (MetaDG). MetaDG leverages dynamic graph structures of node representations to explicitly model spatio-temporal dynamics. This generates both dynamic adjacency matrices and meta-parameters, extending dynamic modeling beyond topology while unifying the capture of spatio-temporal heterogeneity into a single dimension. Extensive experiments on four real-world datasets validate the effectiveness of MetaDG.

Recent Omni-multimodal Large Language Models show promise in unified audio, vision, and text modeling. However, streaming audio-video understanding remains challenging, as existing approaches suffer from disjointed capabilities: they typically exhibit incomplete modality support or lack autonomous proactive monitoring. To address this, we present ROMA, a real-time omni-multimodal assistant for unified reactive and proactive interaction. ROMA processes continuous inputs as synchronized multimodal units, aligning dense audio with discrete video frames to handle granularity mismatches. For online decision-making, we introduce a lightweight speak head that decouples response initiation from generation to ensure precise triggering without task conflict. We train ROMA with a curated streaming dataset and a two-stage curriculum that progressively optimizes for streaming format adaptation and proactive responsiveness. To standardize the fragmented evaluation landscape, we reorganize diverse benchmarks into a unified suite covering both proactive (alert, narration) and reactive (QA) settings. Extensive experiments across 12 benchmarks demonstrate ROMA achieves state-of-the-art performance on proactive tasks while competitive in reactive settings, validating its robustness in unified real-time omni-multimodal understanding.

Finding the most relevant person for a job proposal in real time is challenging, especially when resumes are long, structured, and multilingual. In this paper, we propose a re-ranking model based on a new generation of late cross-attention architecture, that decomposes both resumes and project briefs to efficiently handle long-context inputs with minimal computational overhead. To mitigate historical data biases, we use a generative large language model (LLM) as a teacher, generating fine-grained, semantically grounded supervision. This signal is distilled into our student model via an enriched distillation loss function. The resulting model produces skill-fit scores that enable consistent and interpretable person-job matching. Experiments on relevance, ranking, and calibration metrics demonstrate that our approach outperforms state-of-the-art baselines.

In this paper, we present BAR-SQL (Boundary-Aware Reliable NL2SQL), a unified training framework that embeds reliability and boundary awareness directly into the generation process. We introduce a Seed Mutation data synthesis paradigm that constructs a representative enterprise corpus, explicitly encompassing multi-step analytical queries alongside boundary cases including ambiguity and schema limitations. To ensure interpretability, we employ Knowledge-Grounded Reasoning Synthesis, which produces Chain-of-Thought traces explicitly anchored in schema metadata and business rules. The model is trained through a two-stage process: Supervised Fine-Tuning (SFT) followed by Reinforcement Learning via Group Relative Policy Optimization. We design a Task-Conditioned Hybrid Reward mechanism that simultaneously optimizes SQL execution accuracy-leveraging Abstract Syntax Tree analysis and dense result matching-and semantic precision in abstention responses. To evaluate reliability alongside generation accuracy, we construct and release Ent-SQL-Bench, which jointly assesse SQL precision and boundary-aware abstention across ambiguous and unanswerable queries. Experimental results on this benchmark demonstrate that BAR-SQL achieves 91.48% average accuracy, outperforming leading proprietary models, including Claude 4.5 Sonnet and GPT-5, in both SQL generation quality and boundary-aware abstention capability. The source code and benchmark are available anonymously at: https://github.com/TianSongS/BAR-SQL.

As large-scale speech-to-speech models achieve high fidelity, the distinction between synthetic voices in structured environments becomes a vital area of study. This paper introduces Advosynth-500, a specialized dataset comprising 100 synthetic speech files featuring 10 unique advocate identities. Using the Speech Llama Omni model, we simulate five distinct advocate pairs engaged in courtroom arguments. We define specific vocal characteristics for each advocate and present a speaker identification challenge to evaluate the ability of modern systems to map audio files to their respective synthetic origins. Dataset is available at this link-https: //github.com/naturenurtureelite/ADVOSYNTH-500.

The World Wide Web thrives on intelligent services that rely on accurate time series classification, which has recently witnessed significant progress driven by advances in deep learning. However, existing studies face challenges in domain incremental learning. In this paper, we propose a lightweight and robust dual-causal disentanglement framework (DualCD) to enhance the robustness of models under domain incremental scenarios, which can be seamlessly integrated into time series classification models. Specifically, DualCD first introduces a temporal feature disentanglement module to capture class-causal features and spurious features. The causal features can offer sufficient predictive power to support the classifier in domain incremental learning settings. To accurately capture these causal features, we further design a dual-causal intervention mechanism to eliminate the influence of both intra-class and inter-class confounding features. This mechanism constructs variant samples by combining the current class's causal features with intra-class spurious features and with causal features from other classes. The causal intervention loss encourages the model to accurately predict the labels of these variant samples based solely on the causal features. Extensive experiments on multiple datasets and models demonstrate that DualCD effectively improves performance in domain incremental scenarios. We summarize our rich experiments into a comprehensive benchmark to facilitate research in domain incremental time series classification.

We approach multilinguality as sense adaptation: aligning latent meaning representations across languages rather than relying solely on shared parameters and scale. In this paper, we introduce SENse-based Symmetric Interlingual Alignment (SENSIA), which adapts a Backpack language model from one language to another by explicitly aligning sense-level mixtures and contextual representations on parallel data, while jointly training a target-language language modeling loss to preserve fluency. Across benchmarks on four typologically diverse languages, SENSIA generally outperforms comparable multilingual alignment methods and achieves competitive accuracy against monolingual from-scratch baselines while using 2-4x less target-language data. Analyses of learned sense geometry indicate that local sense topology and global structure relative to English are largely preserved, and ablations show that the method is robust in terms of design and scale.

Enhancing the moral alignment of Large Language Models (LLMs) is a critical challenge in AI safety. Current alignment techniques often act as superficial guardrails, leaving the intrinsic moral representations of LLMs largely untouched. In this paper, we bridge this gap by leveraging Moral Foundations Theory (MFT) to map and manipulate the fine-grained moral landscape of LLMs. Through cross-lingual linear probing, we validate the shared nature of moral representations in middle layers and uncover a shared yet different moral subspace between English and Chinese. Building upon this, we extract steerable Moral Vectors and successfully validate their efficacy at both internal and behavioral levels. Leveraging the high generalizability of morality, we propose Adaptive Moral Fusion (AMF), a dynamic inference-time intervention that synergizes probe detection with vector injection to tackle the safety-helpfulness trade-off. Empirical results confirm that our approach acts as a targeted intrinsic defense, effectively reducing incorrect refusals on benign queries while minimizing jailbreak success rates compared to standard baselines.

While Reinforcement Learning (RL) has advanced LLM reasoning, applying it to long-context scenarios is hindered by sparsity of outcome rewards. This limitation fails to penalize ungrounded "lucky guesses," leaving the critical process of needle-in-a-haystack evidence retrieval largely unsupervised. To address this, we propose EAPO (Evidence-Augmented Policy Optimization). We first establish the Evidence-Augmented Reasoning paradigm, validating via Tree-Structured Evidence Sampling that precise evidence extraction is the decisive bottleneck for long-context reasoning. Guided by this insight, EAPO introduces a specialized RL algorithm where a reward model computes a Group-Relative Evidence Reward, providing dense process supervision to explicitly improve evidence quality. To sustain accurate supervision throughout training, we further incorporate an Adaptive Reward-Policy Co-Evolution mechanism. This mechanism iteratively refines the reward model using outcome-consistent rollouts, sharpening its discriminative capability to ensure precise process guidance. Comprehensive evaluations across eight benchmarks demonstrate that EAPO significantly enhances long-context reasoning performance compared to SOTA baselines.

Vision-Language Pre-training (VLP) models demonstrate strong performance across various downstream tasks by learning from large-scale image-text pairs through contrastive pretraining. The release of extensive English image-text datasets (e.g., COYO-700M and LAION-400M) has enabled widespread adoption of models such as CLIP and SigLIP in tasks including cross-modal retrieval and image captioning. However, the advancement of Chinese vision-language pretraining has substantially lagged behind, due to the scarcity of high-quality Chinese image-text data. To address this gap, we develop a comprehensive pipeline for constructing a high-quality Chinese cross-modal dataset. As a result, we propose DanQing, which contains 100 million image-text pairs collected from Common Crawl. Different from existing datasets, DanQing is curated through a more rigorous selection process, yielding superior data quality. Moreover, DanQing is primarily built from 2024-2025 web data, enabling models to better capture evolving semantic trends and thus offering greater practical utility. We compare DanQing with existing datasets by continual pre-training of the SigLIP2 model. Experimental results show that DanQing consistently achieves superior performance across a range of Chinese downstream tasks, including zero-shot classification, cross-modal retrieval, and LMM-based evaluations. To facilitate further research in Chinese vision-language pre-training, we will open-source the DanQing dataset under the Creative Common CC-BY 4.0 license.

Multi-hop uncrewed aerial vehicle (UAV) networks are promising to extend the terrestrial network coverage. Existing multi-hop UAV networks employ a single routing path by selecting the next-hop forwarding node in a hop-by-hop manner, which leads to local congestion and increases traffic delays. In this paper, a novel traffic-adaptive multipath routing method is proposed for multi-hop UAV networks, which enables each UAV to dynamically split and forward traffic flows across multiple next-hop neighbors, thus meeting latency requirements of diverse traffic flows in dynamic mobile environments. An on-time packet delivery ratio maximization problem is formulated to determine the traffic splitting ratios at each hop. This sequential decision-making problem is modeled as a decentralized partially observable Markov decision process (Dec-POMDP). To solve this Dec-POMDP, a novel multi-agent deep reinforcement leaning (MADRL) algorithm, termed Independent Proximal Policy Optimization with Dirichlet Modeling (IPPO-DM), is developed. Specifically, the IPPO serves as the core optimization framework, where the Dirichlet distribution is leveraged to parameterize a continuous stochastic policy network on the probability simplex, inherently ensuring feasible traffic splitting ratios. Simulation results demonstrate that IPPO-DM outperforms benchmark schemes in terms of both delivery latency guarantee and packet loss performance.

Physics-Informed Neural Networks (PINNs) provide a mesh-free approach for solving differential equations by embedding physical constraints into neural network training. However, PINNs tend to overfit within the training domain, leading to poor generalization when extrapolating beyond trained spatiotemporal regions. This work presents SPIKE (Sparse Physics-Informed Koopman-Enhanced), a framework that regularizes PINNs with continuous-time Koopman operators to learn parsimonious dynamics representations. By enforcing linear dynamics $dz/dt = Az$ in a learned observable space, both PIKE (without explicit sparsity) and SPIKE (with L1 regularization on $A$) learn sparse generator matrices, embodying the parsimony principle that complex dynamics admit low-dimensional structure. Experiments across parabolic, hyperbolic, dispersive, and stiff PDEs, including fluid dynamics (Navier-Stokes) and chaotic ODEs (Lorenz), demonstrate consistent improvements in temporal extrapolation, spatial generalization, and long-term prediction accuracy. The continuous-time formulation with matrix exponential integration provides unconditional stability for stiff systems while avoiding diagonal dominance issues inherent in discrete-time Koopman operators.

Despite being under development for over 15 years, transaction throughput remains one of the key challenges confronting blockchains, which typically has a cap of a limited number of transactions per second. A fundamental factor limiting this metric is the network latency associated with the block propagation throughout of the underlying peer-to-peer network, typically formed through random connections. Accelerating the dissemination of blocks not only improves transaction rates, but also enhances system security by reducing the probability of forks. This paper introduces SCRamble: a decentralized protocol that significantly reduces block dissemination time in blockchain networks. SCRamble's effectiveness is attributed to its innovative link selection strategy, which integrates two heuristics: a scoring mechanism that assesses block arrival times from neighboring peers, and a second heuristic that takes network latency into account.

We consider a single large language model (LLM) server that serves a heterogeneous stream of queries belonging to $N$ distinct task types. Queries arrive according to a Poisson process, and each type occurs with a known prior probability. For each task type, the server allocates a fixed number of internal thinking tokens, which determines the computational effort devoted to that query. The token allocation induces an accuracy-latency trade-off: the service time follows an approximately affine function of the allocated tokens, while the probability of a correct response exhibits diminishing returns. Under a first-in, first-out (FIFO) service discipline, the system operates as an $M/G/1$ queue, and the mean system time depends on the first and second moments of the resulting service-time distribution. We formulate a constrained optimization problem that maximizes a weighted average accuracy objective penalized by the mean system time, subject to architectural token-budget constraints and queue-stability conditions. The objective function is shown to be strictly concave over the stability region, which ensures existence and uniqueness of the optimal token allocation. The first-order optimality conditions yield a coupled projected fixed-point characterization of the optimum, together with an iterative solution and an explicit sufficient condition for contraction. Moreover, a projected gradient method with a computable global step-size bound is developed to guarantee convergence beyond the contractive regime. Finally, integer-valued token allocations are attained via rounding of the continuous solution, and the resulting performance loss is evaluated in simulation results.

We present MoST (Mixture of Speech and Text), a novel multimodal large language model that seamlessly integrates speech and text processing through our proposed Modality-Aware Mixture of Experts (MAMoE) architecture. While current multimodal models typically process diverse modality representations with identical parameters, disregarding their inherent representational differences, we introduce specialized routing pathways that direct tokens to modality-appropriate experts based on input type. MAMoE simultaneously enhances modality-specific learning and cross-modal understanding through two complementary components: modality-specific expert groups that capture domain-specific patterns and shared experts that facilitate information transfer between modalities. Building on this architecture, we develop an efficient transformation pipeline that adapts the pretrained MoE language model through strategic post-training on ASR and TTS datasets, followed by fine-tuning with a carefully curated speech-text instruction dataset. A key feature of this pipeline is that it relies exclusively on fully accessible, open-source datasets to achieve strong performance and data efficiency. Comprehensive evaluations across ASR, TTS, audio language modeling, and spoken question answering benchmarks show that MoST consistently outperforms existing models of comparable parameter counts. Our ablation studies confirm that the modality-specific routing mechanism and shared experts design significantly contribute to performance gains across all tested domains. To our knowledge, MoST represents the first fully open-source speech-text LLM built on a Mixture of Experts architecture. \footnote{We release MoST model, training code, inference code, and training data at https://github.com/NUS-HPC-AI-Lab/MoST

This paper introduces an unsupervised health-monitoring framework for turbofan engines that does not require run-to-failure labels. First, operating-condition effects in NASA CMAPSS sensor streams are removed via regression-based normalisation; then a Long Short-Term Memory (LSTM) autoencoder is trained only on the healthy portion of each trajectory. Persistent reconstruction error, estimated using an adaptive data-driven threshold, triggers real-time alerts without hand-tuned rules. Benchmark results show high recall and low false-alarm rates across multiple operating regimes, demonstrating that the method can be deployed quickly, scale to diverse fleets, and serve as a complementary early-warning layer to Remaining Useful Life models.

Separate Source-Channel Coding (SSCC) remains attractive for text transmission due to its modularity and compatibility with mature entropy coders and powerful channel codes. However, SSCC often suffers from a pronounced cliff effect in low Signal-to-Noise Ratio (SNR) regimes, where residual bit errors after channel decoding can catastrophically break lossless source decoding, especially for Arithmetic Coding (AC) driven by Large Language Models (LLMs). This paper proposes a receiver-side In-Context Decoding (ICD) framework that enhances SSCC robustness without modifying the transmitter. ICD leverages an Error Correction Code Transformer (ECCT) to obtain bit-wise reliability for the decoded information bits. Based on the context-consistent bitstream, ICD constructs a confidence-ranked candidate pool via reliability-guided bit flipping, samples a compact yet diverse subset of candidates, and applies an LLM-based arithmetic decoder to obtain both reconstructions and sequence-level log-likelihoods. A reliability-likelihood fusion rule then selects the final output. We further provide theoretical guarantees on the stability and convergence of the proposed sampling procedure. Extensive experiments over Additive White Gaussian Noise (AWGN) and Rayleigh fading channels demonstrate consistent gains compared with conventional SSCC baselines and representative Joint Source-Channel Coding (JSCC) schemes.

Understanding relationships between attention heads is essential for interpreting the internal structure of Transformers, yet existing metrics do not capture this structure well. We focus on the subspaces spanned by attention-head weight matrices and quantify head-to-head relationships using the Projection Kernel (PK), a principal-angle-based measure of subspace similarity. Experiments show that PK reproduces known head-to-head interactions on the IOI task more clearly than prior metrics such as the Composition Score. We further introduce a framework to quantify the informativeness of PK distributions by comparing them with a reference distribution derived from random orthogonal subspaces. As an application, we analyze a directed graph constructed from PK and show that, in GPT2-small, L4H7 acts as a hub by functioning as an identity head.

Carrier Frequency Offset (CFO) estimation in Orthogonal Frequency Division Multiplexing (OFDM) systems faces significant performance degradation across heterogeneous software-defined radio (SDR) platforms due to uncalibrated hardware impairments. Existing deep neural network (DNN)-based approaches lack device-level adaptation, limiting their practical deployment. This paper proposes a Sim2Real transfer learning framework for per-device CFO calibration, combining simulation-driven pretraining with lightweight receiver adaptation. A backbone DNN is pre-trained on synthetic OFDM signals incorporating parametric hardware distortions (e.g., phase noise, IQ imbalance), enabling generalized feature learning without costly cross-device data collection. Subsequently, only the regression layers are fine-tuned using $1,000$ real frames per target device, preserving hardware-agnostic knowledge while adapting to device-specific impairments. Experiments across three SDR families (USRP B210, USRP N210, HackRF One) achieve $30\times$ BER reduction compared to conventional CP-based methods under indoor multipath conditions. The framework bridges the simulation-to-reality gap for robust CFO estimation, enabling cost-effective deployment in heterogeneous wireless systems.

In the field of artificial intelligence, real parameter single objective optimization is an important direction. Both the Differential Evolution (DE) and the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) demonstrate good performance for real parameter single objective optimization. Nevertheless, there exist other types of evolutionary algorithm for the purpose. In recent years, researchers begin to study long-term search. EA4eig - an ensemble of three DE variants and CMA-ES - performs well for long-term search. In this paper, we introduce two types of evolutionary algorithm proposed recently - crisscross search and sparrow search - into EA4eig as secondary evolutionary algorithms to process inferior individuals. Thus, EA4eigCS is obtained. In our ensemble, the secondary evolutionary algorithms are expected to vary distribution of the population for breaking stagnation. Experimental results show that our EA4eigCS outperforms EA4eig and is competitive when compared with state-of-the-art algorithms. Code and supplementary material are available at:https://anonymous.4open.science/r/EA4eigCS-2A43.

AI-powered coding assistants are rapidly becoming fixtures in professional IDEs, yet their sustained influence on everyday development remains poorly understood. Prior research has focused on short-term use or self-reported perceptions, leaving open questions about how sustained AI use reshapes actual daily coding practices in the long term. We address this gap with a mixed-method study of AI adoption in IDEs, combining longitudinal two-year fine-grained telemetry from 800 developers with a survey of 62 professionals. We analyze five dimensions of workflow change: productivity, code quality, code editing, code reuse, and context switching. Telemetry reveals that AI users produce substantially more code but also delete significantly more. Meanwhile, survey respondents report productivity gains and perceive minimal changes in other dimensions. Our results offer empirical insights into the silent restructuring of software workflows and provide implications for designing future AI-augmented tooling.

When LLMs judge moral dilemmas, do they reach different conclusions in different languages, and if so, why? Two factors could drive such differences: the language of the dilemma itself, or the language in which the model reasons. Standard evaluation conflates these by testing only matched conditions (e.g., English dilemma with English reasoning). We introduce a methodology that separately manipulates each factor, covering also mismatched conditions (e.g., English dilemma with Chinese reasoning), enabling decomposition of their contributions. To study \emph{what} changes, we propose an approach to interpret the moral judgments in terms of Moral Foundations Theory. As a side result, we identify evidence for splitting the Authority dimension into a family-related and an institutional dimension. Applying this methodology to English-Chinese moral judgment with 13 LLMs, we demonstrate its diagnostic power: (1) the framework isolates reasoning-language effects as contributing twice the variance of input-language effects; (2) it detects context-dependency in nearly half of models that standard evaluation misses; and (3) a diagnostic taxonomy translates these patterns into deployment guidance. We release our code and datasets at https://anonymous.4open.science/r/CrossCulturalMoralJudgement.

We present NoReGeo, a novel benchmark designed to evaluate the intrinsic geometric understanding of large language models (LLMs) without relying on reasoning or algebraic computation. Unlike existing benchmarks that primarily assess models' proficiency in reasoning-based geometry-where solutions are derived using algebraic methods-NoReGeo focuses on evaluating whether LLMs can inherently encode spatial relationships and recognize geometric properties directly. Our benchmark comprises 2,500 trivial geometric problems spanning 25 categories, each carefully crafted to be solvable purely through native geometric understanding, assuming known object locations. We assess a range of state-of-the-art models on NoReGeo, including frontier models like GPT-4, observing that even the most advanced systems achieve an overall maximum of 65% accuracy in binary classification tasks. Further, our ablation experiments demonstrate that such geometric understanding does not emerge through fine-tuning alone, indicating that effective training for geometric comprehension requires a specialized approach from the outset. Our findings highlight a significant gap in current LLMs' ability to natively grasp geometric concepts, providing a foundation for future research toward models with true geometric cognition.

Current in-IDE AI coding tools typically rely on time-consuming manual prompting and context management, whereas proactive alternatives that anticipate developer needs without explicit invocation remain underexplored. Understanding when humans are receptive to such proactive AI assistance during their daily work remains an open question in human-AI interaction research. We address this gap through a field study of proactive AI assistance in professional developer workflows. We present a five-day in-the-wild study with 15 developers who interacted with a proactive feature of an AI assistant integrated into a production-grade IDE that offers code quality suggestions based on in-IDE developer activity. We examined 229 AI interventions across 5,732 interaction points to understand how proactive suggestions are received across workflow stages, how developers experience them, and their perceived impact. Our findings reveal systematic patterns in human receptivity to proactive suggestions: interventions at workflow boundaries (e.g., post-commit) achieved 52% engagement rates, while mid-task interventions (e.g., on declined edit) were dismissed 62% of the time. Notably, well-timed proactive suggestions required significantly less interpretation time than reactive suggestions (45.4s versus 101.4s, W = 109.00, r = 0.533, p = 0.0016), indicating enhanced cognitive alignment. This study provides actionable implications for designing proactive coding assistants, including how to time interventions, align them with developer context, and strike a balance between AI agency and user control in production IDEs.

Sharpness-Aware Minimization (SAM) aims to improve generalization by minimizing a worst-case perturbed loss over a small neighborhood of model parameters. However, during training, its optimization behavior does not always align with theoretical expectations, since both sharp and flat regions may yield a small perturbed loss. In such cases, the gradient may still point toward sharp regions, failing to achieve the intended effect of SAM. To address this issue, we investigate SAM from a spectral and geometric perspective: specifically, we utilize the angle between the gradient and the leading eigenvector of the Hessian as a measure of sharpness. Our analysis illustrates that when this angle is less than or equal to ninety degrees, the effect of SAM's sharpness regularization can be weakened. Furthermore, we propose an explicit eigenvector-aligned SAM (X-SAM), which corrects the gradient via orthogonal decomposition along the top eigenvector, enabling more direct and efficient regularization of the Hessian's maximum eigenvalue. We prove X-SAM's convergence and superior generalization, with extensive experimental evaluations confirming both theoretical and practical advantages.

Access to mental healthcare is increasingly strained by workforce shortages and rising demand, motivating the development of intelligent systems that can support mental healthcare experts. We introduce coTherapist, a unified framework utilizing a small language model to emulate core therapeutic competencies through domain-specific fine-tuning, retrieval augmentation, and agentic reasoning. Evaluation on clinical queries demonstrates that coTherapist generates more relevant and clinically grounded responses than contemporary baselines. Using our novel T-BARS rubric and psychometric profiling, we confirm coTherapist exhibits high empathy and therapist-consistent personality traits. Furthermore, human evaluation by domain experts validates that coTherapist delivers accurate, trustworthy, and safe responses. coTherapist was deployed and tested by clinical experts. Collectively, these findings demonstrate that small models can be engineered to exhibit expert-like behavior, offering a scalable pathway for digital mental health tools.

Multi-step reasoning tasks like mathematical problem solving are vulnerable to cascading failures, where a single incorrect step leads to complete solution breakdown. Current LLM routing methods assign entire queries to one model, treating all reasoning steps as equal. We propose TRIM (Targeted routing in multi-step reasoning tasks), which routes only critical steps$\unicode{x2013}$those likely to derail the solution$\unicode{x2013}$to larger models while letting smaller models handle routine continuations. Our key insight is that targeted step-level interventions can fundamentally transform inference efficiency by confining expensive calls to precisely those steps where stronger models prevent cascading errors. TRIM operates at the step-level: it uses process reward models to identify erroneous steps and makes routing decisions based on step-level uncertainty and budget constraints. We develop several routing strategies within TRIM, ranging from a simple threshold-based policy to more expressive policies that reason about long-horizon accuracy-cost trade-offs and uncertainty in step-level correctness estimates. On MATH-500, even the simplest thresholding strategy surpasses prior routing methods with 5x higher cost efficiency, while more advanced policies match the strong, expensive model's performance using 80% fewer expensive model tokens. On harder benchmarks such as AIME, TRIM achieves up to 6x higher cost efficiency. All methods generalize effectively across math reasoning tasks, demonstrating that step-level difficulty represents fundamental characteristics of reasoning.

Large Language Models (LLMs) often exhibit a gap between their internal knowledge and their explicit linguistic outputs. In this report, we empirically investigate whether Looped Transformers (LTs)--architectures that increase computational depth by iterating shared layers--can bridge this gap by utilizing their iterative nature as a form of introspection. Our experiments reveal that while increasing loop iterations narrows the gap, it is partly driven by a degradation of their internal knowledge carried by representations. Moreover, another empirical analysis suggests that current LTs' ability to perceive representations does not improve across loops; it is only present in the final loop. These results suggest that while LTs offer a promising direction for scaling computational depth, they have yet to achieve the introspection required to truly link representation space and natural language.

Differentially Private Stochastic Gradient Descent (DP-SGD) is the dominant paradigm for private training, but its fundamental limitations under worst-case adversarial privacy definitions remain poorly understood. We analyze DP-SGD in the $f$-differential privacy framework, which characterizes privacy via hypothesis-testing trade-off curves, and study shuffled sampling over a single epoch with $M$ gradient updates. We derive an explicit suboptimal upper bound on the achievable trade-off curve. This result induces a geometric lower bound on the separation $κ$ which is the maximum distance between the mechanism's trade-off curve and the ideal random-guessing line. Because a large separation implies significant adversarial advantage, meaningful privacy requires small $κ$. However, we prove that enforcing a small separation imposes a strict lower bound on the Gaussian noise multiplier $σ$, which directly limits the achievable utility. In particular, under the standard worst-case adversarial model, shuffled DP-SGD must satisfy $σ\ge \frac{1}{\sqrt{2\ln M}}$ $\quad\text{or}\quad$ $κ\ge\ \frac{1}{\sqrt{8}}\!\left(1-\frac{1}{\sqrt{4π\ln M}}\right)$, and thus cannot simultaneously achieve strong privacy and high utility. Although this bound vanishes asymptotically as $M \to \infty$, the convergence is extremely slow: even for practically relevant numbers of updates the required noise magnitude remains substantial. We further show that the same limitation extends to Poisson subsampling up to constant factors. Our experiments confirm that the noise levels implied by this bound leads to significant accuracy degradation at realistic training settings, thus showing a critical bottleneck in DP-SGD under standard worst-case adversarial assumptions.

AI writing assistants can reduce effort and improve fluency, but they may also weaken writers' sense of authorship. We study this tension with an ownership-aware co-writing editor that offers on-demand, sentence-level suggestions and tests two common design choices: persona-based coaching and style personalization. In an online study (N=176), participants completed three professional writing tasks: an email without AI help, a proposal with generic AI suggestions, and a cover letter with persona-based coaching, while half received suggestions tailored to a brief sample of their prior writing. Across the two AI-assisted tasks, psychological ownership dropped relative to unassisted writing (about 0.85-1.0 points on a 7-point scale), even as cognitive load decreased (about 0.9 points) and quality ratings stayed broadly similar overall. Persona coaching did not prevent the ownership decline. Style personalization partially restored ownership (about +0.43) and increased AI incorporation in text (+5 percentage points). We distill five design patterns: on-demand initiation, micro-suggestions, voice anchoring, audience scaffolds, and point-of-decision provenance, to guide authorship-preserving writing tools.

Understanding complex parameter dependencies is critical for effective configuration and maintenance of software systems across diverse domains - from Computer-Aided Engineering (CAE) to cloud infrastructure and database management. However, legacy tabular interfaces create a major bottleneck: engineers cannot easily comprehend how parameters relate across the system, leading to inefficient workflows, costly configuration errors, and reduced system trust - a fundamental program comprehension challenge in configuration-intensive software. This research evaluates whether interactive Sankey diagrams can improve comprehension of parameter dependencies compared to traditional spreadsheet interfaces. We employed a heuristic evaluation using the PURE method with three expert evaluators (UX design, simulation, and software development specialists) to compare a Sankey-based prototype to traditional tabular representations for core engineering tasks. Our key contribution demonstrates that flow-based parameter visualizations significantly reduce cognitive load (51% lower PURE scores) and interaction complexity (56% fewer steps) compared to traditional tables, while making parameter dependencies immediately visible rather than requiring mental reconstruction. By explicitly visualizing parameter relationships, Sankey diagrams address a core software visualization challenge: helping users comprehend complex system configurations without requiring deep tool-specific knowledge. While demonstrated through CAE software, this research contributes to program comprehension and software visualization by showing that dependency-aware visualizations can significantly improve understanding of configuration-intensive systems. The findings have implications for any software domain where comprehending complex parameter relationships is essential for effective system use and maintenance.

Recent advances in Large Language Models (LLMs) have improved multi-step reasoning. Most approaches rely on Chain-of-Thought (CoT) rationales. Previous studies have shown that LLMs often generate logically inconsistent reasoning steps even when their final answers are correct. These inconsistencies reduce the reliability of step-level reasoning. We propose GeoSteer, a manifold-based framework that improves the quality of intermediate reasoning. The method consists of: (1) constructing a CoT dataset with segment-level scores, (2) training a Variational Autoencoder (VAE) model and a quality estimation model to learn a low-dimensional manifold of high-quality CoT trajectories, and (3) steering hidden states of target LLMs toward higher-quality regions in the latent space. This update in a latent space behaves like a natural-gradient adjustment in the original hidden-state space. It ensures geometrically coherent steering. We evaluate GeoSteer on the GSM8k dataset using the Qwen3 series. We measure via answer accuracy and overall reasoning performance. GeoSteer improved the exact match accuracy by up to 2.6 points. It also enhanced the pairwise win rate by 5.3 points. These results indicate that GeoSteer provides an effective and controllable mechanism for improving the quality of intermediate reasoning in LLMs.

Optimization is central to both modern machine learning (ML) and scientific machine learning (SciML), yet the structure of the underlying optimization problems differs substantially across these domains. Classical ML typically relies on stochastic, sample-separable objectives that favor first-order and adaptive gradient methods. In contrast, SciML often involves physics-informed or operator-constrained formulations in which differential operators induce global coupling, stiffness, and strong anisotropy in the loss landscape. As a result, optimization behavior in SciML is governed by the spectral properties of the underlying physical models rather than by data statistics, frequently limiting the effectiveness of standard stochastic methods and motivating deterministic or curvature-aware approaches. This document provides a unified introduction to optimization methods in ML and SciML, emphasizing how problem structure shapes algorithmic choices. We review first- and second-order optimization techniques in both deterministic and stochastic settings, discuss their adaptation to physics-constrained and data-driven SciML models, and illustrate practical strategies through tutorial examples, while highlighting open research directions at the interface of scientific computing and scientific machine learning.

A new transformation is underway in software engineering, driven by the rapid adoption of generative AI in development workflows. Similar to how version control systems once automated manual coordination, AI tools are now beginning to automate many aspects of programming. For embedded software engineering organizations, however, this marks their first experience integrating AI into safety-critical and resource-constrained environments. The strict demands for determinism, reliability, and traceability pose unique challenges for adopting generative technologies. In this paper, we present findings from a qualitative study with ten senior experts from four companies who are evaluating generative AI-augmented development for embedded software. Through semi-structured focus group interviews and structured brainstorming sessions, we identified eleven emerging practices and fourteen challenges related to the orchestration, responsible governance, and sustainable adoption of generative AI tools. Our results show how embedded software engineering teams are rethinking workflows, roles, and toolchains to enable a sustainable transition toward agentic pipelines and generative AI-augmented development.

In enterprise datasets, documents are rarely pure. They are not just text, nor just numbers; they are a complex amalgam of narrative and structure. Current Retrieval-Augmented Generation (RAG) systems have attempted to address this complexity with a blunt tool: linearization. We convert rich, multidimensional tables into simple Markdown-style text strings, hoping that an embedding model will capture the geometry of a spreadsheet in a single vector. But it has already been shown that this is mathematically insufficient. This work presents Topo-RAG, a framework that challenges the assumption that "everything is text". We propose a dual architecture that respects the topology of the data: we route fluid narrative through traditional dense retrievers, while tabular structures are processed by a Cell-Aware Late Interaction mechanism, preserving their spatial relationships. Evaluated on SEC-25, a synthetic enterprise corpus that mimics real-world complexity, Topo-RAG demonstrates an 18.4% improvement in nDCG@10 on hybrid queries compared to standard linearization approaches. It's not just about searching better; it's about understanding the shape of information.

The prevalence of recommendation systems also brings privacy concerns to both the users and the sellers, as centralized platforms collect as much data as possible from them. To keep the data private, we propose PADER: a Paillier-based secure decentralized social recommendation system. In this system, the users and the sellers are nodes in a decentralized network. The training and inference of the recommendation model are carried out securely in a decentralized manner, without the involvement of a centralized platform. To this end, we apply the Paillier cryptosystem to the SoReg (Social Regularization) model, which exploits both user's ratings and social relations. We view the SoReg model as a two-party secure polynomial evaluation problem and observe that the simple bipartite computation may result in poor efficiency. To improve efficiency, we design secure addition and multiplication protocols to support secure computation on any arithmetic circuit, along with an optimal data packing scheme that is suitable for the polynomial computations of real values. Experiment results show that our method only takes about one second to iterate through one user with hundreds of ratings, and training with ~500K ratings for one epoch only takes <3 hours, which shows that the method is practical in real applications. The code is available at https://github.com/GarminQ/PADER.

Aligning multilingual assistants with culturally grounded user preferences is essential for serving India's linguistically diverse population of over one billion speakers across multiple scripts. However, existing benchmarks either focus on a single language or conflate retrieval with generation, leaving open the question of whether current embedding models can encode persona-instruction compatibility without relying on response synthesis. We present a unified benchmark spanning 12 Indian languages and four evaluation tasks: monolingual and cross-lingual persona-to-instruction retrieval, reverse retrieval from instruction to persona, and binary compatibility classification. Eight multilingual embedding models are evaluated in a frozen-encoder setting with a thin logistic regression head for classification. E5-Large-Instruct achieves the highest Recall@1 of 27.4\% on monolingual retrieval and 20.7\% on cross-lingual transfer, while BGE-M3 leads reverse retrieval at 32.1\% Recall@1. For classification, LaBSE attains 75.3\% AUROC with strong calibration. These findings offer practical guidance for model selection in Indic multilingual retrieval and establish reproducible baselines for future work\footnote{Code, datasets, and models are publicly available at https://github.com/aryashah2k/PI-Indic-Align.

Improving the reasoning abilities of Large Language Models (LLMs) has been a continuous topic recently. But most relevant works are based on outcome rewards at the trajectory level, missing fine-grained supervision during the reasoning process. Other existing training frameworks that try to combine process signals together to optimize LLMs also rely heavily on tedious additional steps like MCTS, training a separate reward model, etc., doing harm to the training efficiency. Moreover, the intuition behind the process signals design lacks rigorous theoretical support, leaving the understanding of the optimization mechanism opaque. In this paper, we propose Process Reward Learning (PRL), which decomposes the entropy regularized reinforcement learning objective into intermediate steps, with rigorous process rewards that could be assigned to models accordingly. Starting from theoretical motivation, we derive the formulation of PRL that is essentially equivalent to the objective of reward maximization plus a KL-divergence penalty term between the policy model and a reference model. However, PRL could turn the outcome reward into process supervision signals, which helps better guide the exploration during RL optimization. From our experiment results, we demonstrate that PRL not only improves the average performance for LLMs' reasoning ability measured by average @ n, but also broadens the reasoning boundary by improving the pass @ n metric. Extensive experiments show the effectiveness of PRL could be verified and generalized.

High-dimensional data often exhibit dependencies among variables that violate the isotropic-noise assumption under which principal component analysis (PCA) is optimal. For cases where the noise is not independent and identically distributed across features (i.e., the covariance is not spherical) we introduce Graph Regularized PCA (GR-PCA). It is a graph-based regularization of PCA that incorporates the dependency structure of the data features by learning a sparse precision graph and biasing loadings toward the low-frequency Fourier modes of the corresponding graph Laplacian. Consequently, high-frequency signals are suppressed, while graph-coherent low-frequency ones are preserved, yielding interpretable principal components aligned with conditional relationships. We evaluate GR-PCA on synthetic data spanning diverse graph topologies, signal-to-noise ratios, and sparsity levels. Compared to mainstream alternatives, it concentrates variance on the intended support, produces loadings with lower graph-Laplacian energy, and remains competitive in out-of-sample reconstruction. When high-frequency signals are present, the graph Laplacian penalty prevents overfitting, reducing the reconstruction accuracy but improving structural fidelity. The advantage over PCA is most pronounced when high-frequency signals are graph-correlated, whereas PCA remains competitive when such signals are nearly rotationally invariant. The procedure is simple to implement, modular with respect to the precision estimator, and scalable, providing a practical route to structure-aware dimensionality reduction that improves structural fidelity without sacrificing predictive performance.

Large Language Models (LLMs) have demonstrated remarkable capabilities in reasoning and generation, serving as the foundation for advanced persona simulation and Role-Playing Language Agents (RPLAs). However, achieving authentic alignment with human cognitive and behavioral patterns remains a critical challenge for these agents. We present HUMANLLM, a framework treating psychological patterns as interacting causal forces. We construct 244 patterns from ~12,000 academic papers and synthesize 11,359 scenarios where 2-5 patterns reinforce, conflict, or modulate each other, with multi-turn conversations expressing inner thoughts, actions, and dialogue. Our dual-level checklists evaluate both individual pattern fidelity and emergent multi-pattern dynamics, achieving strong human alignment (r=0.91) while revealing that holistic metrics conflate simulation accuracy with social desirability. HUMANLLM-8B outperforms Qwen3-32B on multi-pattern dynamics despite 4x fewer parameters, demonstrating that authentic anthropomorphism requires cognitive modeling--simulating not just what humans do, but the psychological processes generating those behaviors.

Group anomaly detection is crucial in many network applications, but faces challenges due to diverse anomaly patterns. Motivated by the success of large language models (LLMs) in natural language processing, graph foundation models (GFMs) is proposed to handle few-shot learning task with fewer labeling efforts. GFMs have been successfully applied to detection of individual anomalies but cannot be generalized to group anomalies, as group anomaly patterns must be detected as a whole and individuals in an abnormal group can look rather normal. Therefore, we propose GFM4GA, a novel graph foundation model for group anomaly detection. The pipeline is pretrained via dual-level contrastive learning based on feature-based estimation and group extraction, to capture potential group anomaly structure and feature inconsistencies. In the downstream tasks, the pipeline is finetuned in parameter-constrained and group-anomaly-proportion weighted few-shot settings, and its adaptive ability to unseen group anomalies expanded via group contexts determined by labeled anomaly neighbors. Experiments show that GFM4GA surpasses group anomaly detectors and GFMs for individual anomalies, achieving average improvements of 2.85% in AUROC and 2.55% in AUPRC.

Automated analysis of needle electromyography (nEMG) signals is emerging as a tool to support the detection of neuromuscular diseases (NMDs), yet the signals' high and heterogeneous sampling rates pose substantial computational challenges for feature-based machine-learning models, particularly for near real-time analysis. Downsampling offers a potential solution, but its impact on diagnostic signal content and classification performance remains insufficiently understood. This study presents a workflow for systematically evaluating information loss caused by downsampling in high-frequency time series. The workflow combines shape-based distortion metrics with classification outcomes from available feature-based machine learning models and feature space analysis to quantify how different downsampling algorithms and factors affect both waveform integrity and predictive performance. We use a three-class NMD classification task to experimentally evaluate the workflow. We demonstrate how the workflow identifies downsampling configurations that preserve diagnostic information while substantially reducing computational load. Analysis of shape-based distortion metrics showed that shape-aware downsampling algorithms outperform standard decimation, as they better preserve peak structure and overall signal morphology. The results provide practical guidance for selecting downsampling configurations that enable near real-time nEMG analysis and highlight a generalisable workflow that can be used to balance data reduction with model performance in other high-frequency time-series applications as well.

Large Language Models (LLMs) have achieved remarkable strides in multilingual translation but are hindered by a systemic cross-lingual verbosity bias, rendering them unsuitable for strict time-constrained tasks like subtitling and dubbing. Current prompt-engineering approaches struggle to resolve this conflict between semantic fidelity and rigid temporal feasibility. To bridge this gap, we first introduce Sand-Glass, a benchmark specifically designed to evaluate translation under syllable-level duration constraints. Furthermore, we propose HOMURA, a reinforcement learning framework that explicitly optimizes the trade-off between semantic preservation and temporal compliance. By employing a KL-regularized objective with a novel dynamic syllable-ratio reward, HOMURA effectively "tames" the output length. Experimental results demonstrate that our method significantly outperforms strong LLM baselines, achieving precise length control that respects linguistic density hierarchies without compromising semantic adequacy.

Climate prediction is a challenge due to the intricate spatiotemporal patterns within Earth systems. Global climate indices, such as the El Niño Southern Oscillation, are standard input features for long-term rainfall prediction. However, a significant gap persists regarding local-scale indices capable of improving predictive accuracy in specific regions of Thailand. This paper introduces a novel NorthEast monsoon climate index calculated from sea surface temperature to reflect the climatology of the boreal winter monsoon. To optimise the calculated areas used for this index, a Deep Q-Network reinforcement learning agent explores and selects the most effective rectangles based on their correlation with seasonal rainfall. Rainfall stations were classified into 12 distinct clusters to distinguish rainfall patterns between southern and upper Thailand. Experimental results show that incorporating the optimised index into Long Short-Term Memory models significantly improves long-term monthly rainfall prediction skill in most cluster areas. This approach effectively reduces the Root Mean Square Error for 12-month-ahead forecasts.

Pre-trained models operating directly on raw bytes have achieved promising performance in encrypted network traffic classification (NTC), but often suffer from shortcut learning-relying on spurious correlations that fail to generalize to real-world data. Existing solutions heavily rely on model-specific interpretation techniques, which lack adaptability and generality across different model architectures and deployment scenarios. In this paper, we propose BiasSeeker, the first semi-automated framework that is both model-agnostic and data-driven for detecting dataset-specific shortcut features in encrypted traffic. By performing statistical correlation analysis directly on raw binary traffic, BiasSeeker identifies spurious or environment-entangled features that may compromise generalization, independent of any classifier. To address the diverse nature of shortcut features, we introduce a systematic categorization and apply category-specific validation strategies that reduce bias while preserving meaningful information. We evaluate BiasSeeker on 19 public datasets across three NTC tasks. By emphasizing context-aware feature selection and dataset-specific diagnosis, BiasSeeker offers a novel perspective for understanding and addressing shortcut learning in encrypted network traffic classification, raising awareness that feature selection should be an intentional and scenario-sensitive step prior to model training.

Distributed linearly separable computation is a fundamental problem in large-scale distributed systems, requiring the computation of linearly separable functions over different datasets across distributed workers. This paper studies a heterogeneous distributed linearly separable computation problem, including one master and N distributed workers. The linearly separable task function involves Kc linear combinations of K messages, where each message is a function of one dataset. Distinguished from the existing homogeneous settings that assume each worker holds the same number of datasets, where the data assignment is carefully designed and controlled by the data center (e.g., the cyclic assignment), we consider a more general setting with arbitrary heterogeneous data assignment across workers, where `arbitrary' means that the data assignment is given in advance and `heterogeneous' means that the workers may hold different numbers of datasets. Our objective is to characterize the fundamental tradeoff between the computable dimension of the task function and the communication cost under arbitrary heterogeneous data assignment. Under the constraint of integer communication costs, for arbitrary heterogeneous data assignment, we propose a universal computing scheme and a universal converse bound by characterizing the structure of data assignment, where they coincide under some parameter regimes. We then extend the proposed computing scheme and converse bound to the case of fractional communication costs.

Customer Lifetime Value (LTV) prediction, a central problem in modern marketing, is characterized by a unique zero-inflated and long-tail data distribution. This distribution presents two fundamental challenges: (1) the vast majority of low-to-medium value users numerically overwhelm the small but critically important segment of high-value "whale" users, and (2) significant value heterogeneity exists even within the low-to-medium value user base. Common approaches either rely on rigid statistical assumptions or attempt to decouple ranking and regression using ordered buckets; however, they often enforce ordinality through loss-based constraints rather than inherent architectural design, failing to balance global accuracy with high-value precision. To address this gap, we propose \textbf{C}onditional \textbf{C}ascaded \textbf{O}rdinal-\textbf{R}esidual Networks \textbf{(CC-OR-Net)}, a novel unified framework that achieves a more robust decoupling through \textbf{structural decomposition}, where ranking is architecturally guaranteed. CC-OR-Net integrates three specialized components: a \textit{structural ordinal decomposition module} for robust ranking, an \textit{intra-bucket residual module} for fine-grained regression, and a \textit{targeted high-value augmentation module} for precision on top-tier users. Evaluated on real-world datasets with over 300M users, CC-OR-Net achieves a superior trade-off across all key business metrics, outperforming state-of-the-art methods in creating a holistic and commercially valuable LTV prediction solution.

Large Language Models (LLMs) have enabled the development of powerful agentic systems capable of automating complex workflows across various fields. However, these systems are highly vulnerable to indirect prompt injection attacks, where malicious instructions embedded in external data can hijack agent behavior. In this work, we present ReasAlign, a model-level solution to improve safety alignment against indirect prompt injection attacks. The core idea of ReasAlign is to incorporate structured reasoning steps to analyze user queries, detect conflicting instructions, and preserve the continuity of the user's intended tasks to defend against indirect injection attacks. To further ensure reasoning logic and accuracy, we introduce a test-time scaling mechanism with a preference-optimized judge model that scores reasoning steps and selects the best trajectory. Comprehensive evaluations across various benchmarks show that ReasAlign maintains utility comparable to an undefended model while consistently outperforming Meta SecAlign, the strongest prior guardrail. On the representative open-ended CyberSecEval2 benchmark, which includes multiple prompt-injected tasks, ReasAlign achieves 94.6% utility and only 3.6% ASR, far surpassing the state-of-the-art defensive model of Meta SecAlign (56.4% utility and 74.4% ASR). These results demonstrate that ReasAlign achieves the best trade-off between security and utility, establishing a robust and practical defense against prompt injection attacks in real-world agentic systems. Our code and experimental results could be found at https://github.com/leolee99/ReasAlign.

Compositionality is a cognitive mechanism that allows humans to systematically combine known concepts in novel ways. This study demonstrates how artificial neural agents acquire and utilize compositional generalization to describe previously unseen images. Our method, termed "Composition through Decomposition", involves two sequential training steps. In the 'Decompose' step, the agents learn to decompose an image into basic concepts using a codebook acquired during interaction in a multi-target coordination game. Subsequently, in the 'Compose' step, the agents employ this codebook to describe novel images by composing basic concepts into complex phrases. Remarkably, we observe cases where generalization in the `Compose' step is achieved zero-shot, without the need for additional training.

Open-vocabulary 3D Scene Graph (3DSG) generation can enhance various downstream tasks in robotics, such as manipulation and navigation, by leveraging structured semantic representations. A 3DSG is constructed from multiple images of a scene, where objects are represented as nodes and relationships as edges. However, existing works for open-vocabulary 3DSG generation suffer from both low object-level recognition accuracy and speed, mainly due to constrained viewpoints, occlusions, and redundant surface density. To address these challenges, we propose RAG-3DSG to mitigate aggregation noise through re-shot guided uncertainty estimation and support object-level Retrieval-Augmented Generation (RAG) via reliable low-uncertainty objects. Furthermore, we propose a dynamic downsample-mapping strategy to accelerate cross-image object aggregation with adaptive granularity. Experiments on Replica dataset demonstrate that RAG-3DSG significantly improves node captioning accuracy in 3DSG generation while reducing the mapping time by two-thirds compared to the vanilla version.

Debt collection is a critical function within the banking, financial services, and insurance (BFSI) sector, relying heavily on large-scale human-to-human conversational interactions conducted primarily in Vietnamese contact centers. These conversations involve informal spoken language, emotional variability, and complex domain-specific reasoning, which pose significant challenges for traditional natural language processing systems. This paper introduces Credit C-GPT, a domain-specialized large language model with seven billion parameters, fine-tuned for conversational understanding in Vietnamese debt collection scenarios. The proposed model integrates multiple conversational intelligence tasks, including dialogue understanding, sentiment recognition, intent detection, call stage classification, and structured slot-value extraction, within a single reasoning-based framework. We describe the data construction process, annotation strategy, and training methodology, and evaluate the model on proprietary human-annotated datasets. Experimental results show consistent improvements over traditional pipeline-based approaches, indicating that domain-specialized conversational language models provide a scalable and privacy-aware solution for real-time assistance and post-call analytics in enterprise contact centers.

The rapid growth of Cloud Computing and Internet of Things (IoT) has significantly increased the interconnection of computational resources, creating an environment where malicious software (malware) can spread rapidly. To address this challenge, researchers are increasingly utilizing Machine Learning approaches to identify malware through behavioral (i.e. dynamic) cues. However, current approaches are limited by their reliance on large labeled datasets, fixed model training, and the assumption that a trained model remains effective over time-disregarding the ever-evolving sophistication of malware. As a result, they often fail to detect evolving malware attacks that adapt over time. This paper proposes an online learning approach for dynamic malware detection, that overcomes these limitations by incorporating temporal information to continuously update its models using behavioral features, specifically process resource utilization metrics. By doing so, the proposed models can incrementally adapt to emerging threats and detect zero-day malware effectively. Upon evaluating our approach against traditional batch algorithms, we find it effective in detecting zero-day malware. Moreover, we demonstrate its efficacy in scenarios with limited data availability, where traditional batch-based approaches often struggle to perform reliably.

We introduce AWED-FiNER, an open-source ecosystem designed to bridge the gap in Fine-grained Named Entity Recognition (FgNER) for 36 global languages spoken by more than 6.6 billion people. While Large Language Models (LLMs) dominate general Natural Language Processing (NLP) tasks, they often struggle with low-resource languages and fine-grained NLP tasks. AWED-FiNER provides a collection of agentic toolkits, web applications, and several state-of-the-art expert models that provides FgNER solutions across 36 languages. The agentic tools enable to route multilingual text to specialized expert models and fetch FgNER annotations within seconds. The web-based platforms provide ready-to-use FgNER annotation service for non-technical users. Moreover, the collection of language specific extremely small sized open-source state-of-the-art expert models facilitate offline deployment in resource contraint scenerios including edge devices. AWED-FiNER covers languages spoken by over 6.6 billion people, including a specific focus on vulnerable languages such as Bodo, Manipuri, Bishnupriya, and Mizo. The resources can be accessed here: Agentic Tool (https://github.com/PrachuryyaKaushik/AWED-FiNER), Web Application (https://hf.co/spaces/prachuryyaIITG/AWED-FiNER), and 49 Expert Detector Models (https://hf.co/collections/prachuryyaIITG/awed-finer).

Pretraining corpora contain extensive discourse about AI systems, yet the causal influence of this discourse on downstream alignment remains poorly understood. If prevailing descriptions of AI behaviour are predominantly negative, LLMs may internalise corresponding behavioural priors, giving rise to self-fulfilling misalignment. This paper provides the first controlled study of this hypothesis by pretraining 6.9B-parameter LLMs with varying amounts of (mis)alignment discourse. We find that discussion of AI contributes to misalignment. Upsampling synthetic training documents about AI misalignment leads to a notable increase in misaligned behaviour. Conversely, upsampling documents about aligned behaviour reduces misalignment scores from 45% to 9%. We consider this evidence of self-fulfilling alignment. These effects are dampened, but persist through post-training. Our findings establish the study of how pretraining data shapes alignment priors, or alignment pretraining, as a complement to post-training. We recommend practitioners pretrain for alignment as well as capabilities. Our models and datasets are available at alignmentpretraining.ai

Most interpretability work focuses on layer- or neuron-level mechanisms in Transformers, leaving expert-level behavior in MoE LLMs underexplored. Motivated by functional specialization in the human brain, we analyze expert activation by distinguishing domain and driver experts. In this work, we study expert activation in MoE models across three public domains and address two key questions: (1) which experts are activated, and whether certain expert types exhibit consistent activation patterns; and (2) how tokens are associated with and trigger the activation of specific experts. To answer these questions, we introduce entropy-based and causal-effect metrics to assess whether an expert is strongly favored for a particular domain, and how strongly expert activation contributes causally to the model's output, thus identify domain and driver experts, respectively. Furthermore, we explore how individual tokens are associated with the activation of specific experts. Our analysis reveals that (1) Among the activated experts, some show clear domain preferences, while others exert strong causal influence on model performance, underscoring their decisive roles. (2) tokens occurring earlier in a sentence are more likely to trigger the driver experts, and (3) adjusting the weights of domain and driver experts leads to significant performance gains across all three models and domains. These findings shed light on the internal mechanisms of MoE models and enhance their interpretability.

Graph Neural Networks (GNNs) have been widely adopted for Protein Representation Learning (PRL), as residue interaction networks can be naturally represented as graphs. Current GNN-based PRL methods typically rely on single-perspective graph construction strategies, which capture partial properties of residue interactions, resulting in incomplete protein representations. To address this limitation, we propose MMPG, a framework that constructs protein graphs from multiple perspectives and adaptively fuses them via Mixture of Experts (MoE) for PRL. MMPG constructs graphs from physical, chemical, and geometric perspectives to characterize different properties of residue interactions. To capture both perspective-specific features and their synergies, we develop an MoE module, which dynamically routes perspectives to specialized experts, where experts learn intrinsic features and cross-perspective interactions. We quantitatively verify that MoE automatically specializes experts in modeling distinct levels of interaction from individual representations, to pairwise inter-perspective synergies, and ultimately to a global consensus across all perspectives. Through integrating this multi-level information, MMPG produces superior protein representations and achieves advanced performance on four different downstream protein tasks.

While LLM-based agents can interact with environments via invoking external tools, their expanded capabilities also amplify security risks. Monitoring step-level tool invocation behaviors in real time and proactively intervening before unsafe execution is critical for agent deployment, yet remains under-explored. In this work, we first construct TS-Bench, a novel benchmark for step-level tool invocation safety detection in LLM agents. We then develop a guardrail model, TS-Guard, using multi-task reinforcement learning. The model proactively detects unsafe tool invocation actions before execution by reasoning over the interaction history. It assesses request harmfulness and action-attack correlations, producing interpretable and generalizable safety judgments and feedback. Furthermore, we introduce TS-Flow, a guardrail-feedback-driven reasoning framework for LLM agents, which reduces harmful tool invocations of ReAct-style agents by 65 percent on average and improves benign task completion by approximately 10 percent under prompt injection attacks.

Compressing the KV cache is a required step to deploy large language models on edge devices. Current quantization methods compress storage but fail to reduce bandwidth as attention calculation requires dequantizing keys from INT4/INT8 to FP16 before use. We observe that attention scoring is mathematically equivalent to the inner product similarity search and we can apply some compression techniques from vector databases to compress KV-cache better. We propose LOOKAT, which applies product quantization and asymmetric distance computation, to transformer architecture by decomposing key vectors into subspaces, learning codebooks and computing attention tables via lookup tables. This transforms attention from memory-bound to compute-bound. LOOKAT achieves 64 $\times$ compression at 95.7\% output fidelity and 32 $\times$ compression at 95.0\% fidelity when tested on GPT-2. LOOKAT requires no architecture changes or training while maintaining rank correlation $ρ> 0.95$. Theoretical analysis confirms that rank correlation degrades as $O(d_k/mK)$, with guarantees validated across sequence lengths up to 1024 tokens.

Artificial intelligence (AI) has the potential to transform medical imaging by automating image analysis and accelerating clinical research. However, research and clinical use are limited by the wide variety of AI implementations and architectures, inconsistent documentation, and reproducibility issues. Here, we introduce MHub.ai, an open-source, container-based platform that standardizes access to AI models with minimal configuration, promoting accessibility and reproducibility in medical imaging. MHub.ai packages models from peer-reviewed publications into standardized containers that support direct processing of DICOM and other formats, provide a unified application interface, and embed structured metadata. Each model is accompanied by publicly available reference data that can be used to confirm model operation. MHub.ai includes an initial set of state-of-the-art segmentation, prediction, and feature extraction models for different modalities. The modular framework enables adaptation of any model and supports community contributions. We demonstrate the utility of the platform in a clinical use case through comparative evaluation of lung segmentation models. To further strengthen transparency and reproducibility, we publicly release the generated segmentations and evaluation metrics and provide interactive dashboards that allow readers to inspect individual cases and reproduce or extend our analysis. By simplifying model use, MHub.ai enables side-by-side benchmarking with identical execution commands and standardized outputs, and lowers the barrier to clinical translation.

The effectiveness of contrastive learning methods has been widely recognized in the field of graph learning, especially in contexts where graph data often lack labels or are difficult to label. However, the application of these methods to node classification tasks still faces a number of challenges. First, existing data enhancement techniques may lead to significant differences from the original view when generating new views, which may weaken the relevance of the view and affect the efficiency of model training. Second, the vast majority of existing graph comparison learning algorithms rely on the use of a large number of negative samples. To address the above challenges, this study proposes a novel node classification contrast learning method called Simple Network Graph Comparative Learning (SNGCL). Specifically, SNGCL employs a superimposed multilayer Laplace smoothing filter as a step in processing the data to obtain global and local feature smoothing matrices, respectively, which are thus passed into the target and online networks of the siamese network, and finally employs an improved triple recombination loss function to bring the intra-class distance closer and the inter-class distance farther. We have compared SNGCL with state-of-the-art models in node classification tasks, and the experimental results show that SNGCL is strongly competitive in most tasks.

Long-sequence decision-making, which is usually addressed through reinforcement learning (RL), is a critical component for optimizing strategic operations in dynamic environments, such as real-time bidding in computational advertising. The Decision Transformer (DT) introduced a powerful paradigm by framing RL as an autoregressive sequence modeling problem. Concurrently, Large Language Models (LLMs) have demonstrated remarkable success in complex reasoning and planning tasks. This inspires us whether LLMs, which share the same Transformer foundation, but operate at a much larger scale, can unlock new levels of performance in long-horizon sequential decision-making problem. This work investigates the application of LLMs to offline decision making tasks. A fundamental challenge in this domain is the LLMs' inherent inability to interpret continuous values, as they lack a native understanding of numerical magnitude and order when values are represented as text strings. To address this, we propose treating trajectories as a distinct modality. By learning to align trajectory data with natural language task descriptions, our model can autoregressively predict future decisions within a cohesive framework we term DecisionLLM. We establish a set of scaling laws governing this paradigm, demonstrating that performance hinges on three factors: model scale, data volume, and data quality. In offline experimental benchmarks and bidding scenarios, DecisionLLM achieves strong performance. Specifically, DecisionLLM-3B outperforms the traditional Decision Transformer (DT) by 69.4 on Maze2D umaze-v1 and by 0.085 on AuctionNet. It extends the AIGB paradigm and points to promising directions for future exploration in online bidding.

In quantitative finance, the gap between training and real-world performance-driven by concept drift and distributional non-stationarity-remains a critical obstacle for building reliable data-driven systems. Models trained on static historical data often overfit, resulting in poor generalization in dynamic markets. The mantra "History Is Not Enough" underscores the need for adaptive data generation that learns to evolve with the market rather than relying solely on past observations. We present a drift-aware dataflow system that integrates machine learning-based adaptive control into the data curation process. The system couples a parameterized data manipulation module comprising single-stock transformations, multi-stock mix-ups, and curation operations, with an adaptive planner-scheduler that employs gradient-based bi-level optimization to control the system. This design unifies data augmentation, curriculum learning, and data workflow management under a single differentiable framework, enabling provenance-aware replay and continuous data quality monitoring. Extensive experiments on forecasting and reinforcement learning trading tasks demonstrate that our framework enhances model robustness and improves risk-adjusted returns. The system provides a generalizable approach to adaptive data management and learning-guided workflow automation for financial data.

Fine-tuning is an essential and pervasive functionality for applying large language models (LLMs) to downstream tasks. However, it has the potential to substantially degrade safety alignment, e.g., by greatly increasing susceptibility to jailbreak attacks, even when the fine-tuning data is entirely harmless. Despite garnering growing attention in defense efforts during the fine-tuning stage, existing methods struggle with a persistent safety-utility dilemma: emphasizing safety compromises task performance, whereas prioritizing utility typically requires deep fine-tuning that inevitably leads to steep safety declination. In this work, we address this dilemma by shedding new light on the geometric interaction between safety- and utility-oriented gradients in safety-aligned LLMs. Through systematic empirical analysis, we uncover three key insights: (I) safety gradients lie in a low-rank subspace, while utility gradients span a broader high-dimensional space; (II) these subspaces are often negatively correlated, causing directional conflicts during fine-tuning; and (III) the dominant safety direction can be efficiently estimated from a single sample. Building upon these novel insights, we propose safety-preserving fine-tuning (SPF), a lightweight approach that explicitly removes gradient components conflicting with the low-rank safety subspace. Theoretically, we show that SPF guarantees utility convergence while bounding safety drift. Empirically, SPF consistently maintains downstream task performance and recovers nearly all pre-trained safety alignment, even under adversarial fine-tuning scenarios. Furthermore, SPF exhibits robust resistance to both deep fine-tuning and dynamic jailbreak attacks. Together, our findings provide new mechanistic understanding and practical guidance toward always-aligned LLM fine-tuning.

Causal discovery aims to recover ``what causes what'', but classical constraint-based methods (e.g., PC, FCI) suffer from error propagation, and recent LLM-based causal oracles often behave as opaque, confidence-free black boxes. This paper introduces Tree-Query, a tree-structured, multi-expert LLM framework that reduces pairwise causal discovery to a short sequence of queries about backdoor paths, (in)dependence, latent confounding, and causal direction, yielding interpretable judgments with robustness-aware confidence scores. Theoretical guarantees are provided for asymptotic identifiability of four pairwise relations. On data-free benchmarks derived from Mooij et al. and UCI causal graphs, Tree-Query improves structural metrics over direct LLM baselines, and a diet--weight case study illustrates confounder screening and stable, high-confidence causal conclusions. Tree-Query thus offers a principled way to obtain data-free causal priors from LLMs that can complement downstream data-driven causal discovery. Code is available at https://anonymous.4open.science/r/Repo-9B3E-4F96.

Large Language Models (LLMs) have demonstrated impressive capabilities in reasoning and prediction across different domains. Yet, their ability to infer temporal regularities from structured behavioral data remains underexplored. This paper presents a systematic study investigating whether LLMs can predict time intervals between recurring user actions, such as repeated purchases, and how different levels of contextual information shape their predictive behavior. Using a simple but representative repurchase scenario, we benchmark state-of-the-art LLMs in zero-shot settings against both statistical and machine-learning models. Two key findings emerge. First, while LLMs surpass lightweight statistical baselines, they consistently underperform dedicated machine-learning models, showing their limited ability to capture quantitative temporal structure. Second, although moderate context can improve LLM accuracy, adding further user-level detail degrades performance. These results challenge the assumption that "more context leads to better reasoning". Our study highlights fundamental limitations of today's LLMs in structured temporal inference and offers guidance for designing future context-aware hybrid models that integrate statistical precision with linguistic flexibility.

Generating molecules that satisfy precise numeric constraints over multiple physicochemical properties is critical and challenging. Although large language models (LLMs) are expressive, they struggle with precise multi-objective control and numeric reasoning without external structure and feedback. We introduce \textbf{M olGen}, a fragment-level, retrieval-augmented, two-stage framework for molecule generation under multi-property constraints. Stage I : Prototype generation: a multi-agent reasoner performs retrieval-anchored, fragment-level edits to produce a candidate near the feasible region. Stage II : RL-based fine-grained optimization: a fragment-level optimizer trained with Group Relative Policy Optimization (GRPO) applies one- or multi-hop refinements to explicitly minimize the property errors toward our target while regulating edit complexity and deviation from the prototype. A large, automatically curated dataset with reasoning chains of fragment edits and measured property deltas underpins both stages, enabling deterministic, reproducible supervision and controllable multi-hop reasoning. Unlike prior work, our framework better reasons about molecules by leveraging fragments and supports controllable refinement toward numeric targets. Experiments on generation under two sets of property constraints (QED, LogP, Molecular Weight and HOMO, LUMO) show consistent gains in validity and precise satisfaction of multi-property targets, outperforming strong LLMs and graph-based algorithms.

Functional dependencies (FDs) are basic constraints in relational databases and are used for many data management tasks. Most FD discovery algorithms find all valid dependencies, but this causes two problems. First, the computational cost is prohibitive: computational complexity grows quadratically with the number of tuples and exponentially with the number of attributes, making discovery slow on large-scale and high-dimensional data. Second, the result set can be huge, making it hard to identify useful dependencies. We propose SDP (Selective-Discovery-and-Prune), which discovers the top-$k$ FDs ranked by redundancy count. Redundancy count measures how much duplicated information an FD explains and connects directly to storage overhead and update anomalies. SDP uses an upper bound on redundancy to prune the search space. It is proved that this upper bound is monotone: adding attributes refines partitions and thus decreases the bound. Once the bound falls below the top-$k$ threshold, the entire branch can be skipped. We improve SDP with three optimizations: ordering attributes by partition cardinality, using pairwise statistics in a Partition Cardinality Matrix to tighten bounds, and a global scheduler to explore promising branches first. Experiments on over 40 datasets show that SDP is much faster and uses less memory than exhaustive methods.

Current multimodal latent reasoning often relies on external supervision (e.g., auxiliary images), ignoring intrinsic visual attention dynamics. In this work, we identify a critical Perception Gap in distillation: student models frequently mimic a teacher's textual output while attending to fundamentally divergent visual regions, effectively relying on language priors rather than grounded perception. To bridge this, we propose LaViT, a framework that aligns latent visual thoughts rather than static embeddings. LaViT compels the student to autoregressively reconstruct the teacher's visual semantics and attention trajectories prior to text generation, employing a curriculum sensory gating mechanism to prevent shortcut learning. Extensive experiments show that LaViT significantly enhances visual grounding, achieving up to +16.9% gains on complex reasoning tasks and enabling a compact 3B model to outperform larger open-source variants and proprietary models like GPT-4o.

Scientific and engineering verticals often suffer from data scarcity and strict executability requirements: models must generate not only fluent text, but also syntactically valid, tool-compilable scripts. We present a schema-first alignment framework for building compact, executable domain-specific LLMs in low-resource settings. The framework integrates three core components: (i) large-scale synthetic QA data generation from expert documentation to instill foundational domain knowledge; (ii) a code-centric IR->DPO workflow that converts verified tool decks into interpretable intermediate representations (IR), performs equivalence-preserving diversification, and constructs preference pairs to directly optimize instruction compliance and code executability; and (iii) a controlled evaluation of Retrieval-Augmented Generation (RAG), showing that while RAG benefits general LLMs, it can marginally degrade the performance of already domain-aligned models. We demonstrate the framework by instantiating TcadGPT for semiconductor Technology Computer-Aided Design (TCAD). Using 1.5M synthetic QA pairs and an IR-driven DPO dataset, TcadGPT attains 85.6% semantic accuracy and an 80.0% syntax pass rate on SDE executability tests, substantially outperforming state-of-the-art general LLMs such as GPT-4o. To probe portability beyond TCAD, we apply the same recipe to the open-source FEM solver Elmer, observing consistent improvements in script-level success rates over general-purpose baselines. All datasets, benchmarks, and code (including P1, P2, and IR->DPO) are released for reproducibility. Together, these results suggest that the proposed framework provides a robust and reproducible path toward executable LLMs in specialized, data-scarce professional domains.

The Multi-Agent Path Finding (MAPF) problem aims at finding non-conflicting paths for multiple agents from their respective sources to destinations. This problem arises in multiple real-life situations, including robot motion planning and airspace assignment for unmanned aerial vehicle movement. The problem is computationally expensive, and adding to it, the agents are rational and can misreport their private information. In this paper, we study both variants of the problem under the realm of fairness. For the non-rational agents, we propose a heuristic solution for this problem. Considering the agents are rational, we develop a mechanism and demonstrate that it is a dominant strategy, incentive compatible, and individually rational. We employ various solution methodologies to highlight the effectiveness and efficiency of the proposed solution approaches.

In recent years, with the rapid advancement of large language models (LLMs), role-playing language agents (RPLAs) have emerged as a prominent research focus at the intersection of natural language processing (NLP) and human-computer interaction. This paper systematically reviews the current development and key technologies of RPLAs, delineating the technological evolution from early rule-based template paradigms, through the language style imitation stage, to the cognitive simulation stage centered on personality modeling and memory mechanisms. It summarizes the critical technical pathways supporting high-quality role-playing, including psychological scale-driven character modeling, memory-augmented prompting mechanisms, and motivation-situation-based behavioral decision control. At the data level, the paper further analyzes the methods and challenges of constructing role-specific corpora, focusing on data sources, copyright constraints, and structured annotation processes. In terms of evaluation, it collates multi-dimensional assessment frameworks and benchmark datasets covering role knowledge, personality fidelity, value alignment, and interactive hallucination, while commenting on the advantages and disadvantages of methods such as human evaluation, reward models, and LLM-based scoring. Finally, the paper outlines future development directions of role-playing agents, including personality evolution modeling, multi-agent collaborative narrative, multimodal immersive interaction, and integration with cognitive neuroscience, aiming to provide a systematic perspective and methodological insights for subsequent research.

Optimizing communication topology in LLM-based multi-agent system is critical for enabling collective intelligence. Existing methods mainly rely on spatio-temporal interaction paradigms, where the sequential execution of multi-round dialogues incurs high latency and computation. Motivated by the recent insights that evaluation and debate mechanisms can improve problem-solving in multi-agent systems, we propose TopoDIM, a framework for one-shot Topology generation with Diverse Interaction Modes. Designed for decentralized execution to enhance adaptability and privacy, TopoDIM enables agents to autonomously construct heterogeneous communication without iterative coordination, achieving token efficiency and improved task performance. Experiments demonstrate that TopoDIM reduces total token consumption by 46.41% while improving average performance by 1.50% over state-of-the-art methods. Moreover, the framework exhibits strong adaptability in organizing communication among heterogeneous agents. Code is available at: https://anonymous.4open.science/r/TopoDIM-8D35/

Large language models (LLMs) are challenging to deploy for domain-specific tasks due to their massive scale. While distilling a fine-tuned LLM into a smaller student model is a promising alternative, the capacity gap between teacher and student often leads to suboptimal performance. This raises a key question: when and how can a student model match or even surpass its teacher on domain-specific tasks? In this work, we propose a novel theoretical insight: a student can outperform its teacher if its advantage on a Student-Favored Subdomain (SFS) outweighs its deficit on the Teacher-Favored Subdomain (TFS). Guided by this insight, we propose Scheduled Checkpoint Distillation (SCD), which reduces the TFS deficit by emulating the teacher's convergence process during supervised fine-tuning (SFT) on the domain task, and a sample-wise Adaptive Weighting (AW) mechanism to preserve student strengths on SFS. Experiments across diverse domain tasks--including QA, NER, and text classification in multiple languages--show that our method consistently outperforms existing distillation approaches, allowing the student model to match or even exceed the performance of its fine-tuned teacher.

Repository aware coding agents often struggle to recover build and test structure, especially in multilingual projects where cross language dependencies are encoded across heterogeneous build systems and tooling. We introduce the Repository Intelligence Graph (RIG), a deterministic, evidence backed architectural map that represents buildable components, aggregators, runners, tests, external packages, and package managers, connected by explicit dependency and coverage edges that trace back to concrete build and test definitions. We also present SPADE, a deterministic extractor that constructs RIG from build and test artifacts (currently with an automatic CMake plugin based on the CMake File API and CTest metadata), and exposes RIG as an LLM friendly JSON view that agents can treat as the authoritative description of repository structure. We evaluate three commercial agents (Claude Code, Cursor, Codex) on eight repositories spanning low to high build oriented complexity, including the real world MetaFFI project. Each agent answers thirty structured questions per repository with and without RIG in context, and we measure accuracy, wall clock completion time, and efficiency (seconds per correct answer). Across repositories and agents, providing RIG improves mean accuracy by 12.2\% and reduces completion time by 53.9\%, yielding a mean 57.8\% reduction in seconds per correct answer. Gains are larger in multilingual repositories, which improve by 17.7\% in accuracy and 69.5\% in efficiency on average, compared to 6.6\% and 46.1\% in single language repositories. Qualitative analysis suggests that RIG shifts failures from structural misunderstandings toward reasoning mistakes over a correct structure, while rare regressions highlight that graph based reasoning quality remains a key factor.

This study proposes MCEMOL (Multi-Constrained Evolutionary Molecular Design Framework), a molecular optimization approach integrating rule-based evolution with molecular crossover. MCEMOL employs dual-layer evolution: optimizing transformation rules at rule level while applying crossover and mutation to molecular structures. Unlike deep learning methods requiring large datasets and extensive training, our algorithm evolves efficiently from minimal starting molecules with low computational overhead. The framework incorporates message-passing neural networks and comprehensive chemical constraints, ensuring efficient and interpretable molecular design. Experimental results demonstrate that MCEMOL provides transparent design pathways through its evolutionary mechanism while generating valid, diverse, target-compliant molecules. The framework achieves 100% molecular validity with high structural diversity and excellent drug-likeness compliance, showing strong performance in symmetry constraints, pharmacophore optimization, and stereochemical integrity. Unlike black-box methods, MCEMOL delivers dual value: interpretable transformation rules researchers can understand and trust, alongside high-quality molecular libraries for practical applications. This establishes a paradigm where interpretable AI-driven drug design and effective molecular generation are achieved simultaneously, bridging the gap between computational innovation and practical drug discovery needs.

Large reasoning models such as DeepSeek-R1 and their distilled variants achieve strong performance on complex reasoning tasks. Yet, distilling these models often demands large-scale data for supervised fine-tuning (SFT), motivating the pursuit of data-efficient training methods. To address this, we propose a skill-centric distillation framework that efficiently transfers reasoning ability to weaker models with two components: (1) Skill-based data selection, which prioritizes examples targeting the student model's weaker skills, and (2) Skill-aware fine-tuning, which encourages explicit skill decomposition during problem solving. With only 1,000 training examples selected from a 100K teacher-generated corpus, our method surpasses random SFT baselines by +1.6% on Qwen3-4B and +1.4% on Qwen3-8B across five mathematical reasoning benchmarks. Further analysis confirms that these gains concentrate on skills emphasized during training, highlighting the effectiveness of skill-centric training for efficient reasoning distillation.

Evaluating whether multimodal large language models truly understand long-form scientific papers remains challenging: answer-only metrics and synthetic "Needle-In-A-Haystack" tests often reward answer matching without requiring a causal, evidence-linked reasoning trace in the document. We propose the "Fish-in-the-Ocean" (FITO) paradigm, which requires models to construct explicit cross-modal evidence chains within native scientific documents. To operationalize FITO, we build SIN-Data, a scientific interleaved corpus that preserves the native interleaving of text and figures. On top of it, we construct SIN-Bench with four progressive tasks covering evidence discovery (SIN-Find), hypothesis verification (SIN-Verify), grounded QA (SIN-QA), and evidence-anchored synthesis (SIN-Summary). We further introduce "No Evidence, No Score", scoring predictions when grounded to verifiable anchors and diagnosing evidence quality via matching, relevance, and logic. Experiments on eight MLLMs show that grounding is the primary bottleneck: Gemini-3-pro achieves the best average overall score (0.573), while GPT-5 attains the highest SIN-QA answer accuracy (0.767) but underperforms on evidence-aligned overall scores, exposing a gap between correctness and traceable support.

The rapid integration of IoT with edge computing has revolutionized various domains, particularly healthcare, by enabling real-time data sharing, remote monitoring, and decision-making. However, it introduces critical challenges, including data privacy breaches, security vulnerabilities, especially in environments dealing with sensitive information. Traditional access control mechanisms and centralized security systems do not address these issues, leaving IoT environments exposed to unauthorized access and data misuse. This research proposes Fuzzychain-edge, a novel Fuzzy logic-based adaptive Access control model for Blockchain in Edge Computing framework designed to overcome these limitations by incorporating Zero-Knowledge Proofs (ZKPs), fuzzy logic, and smart contracts. ZKPs secure sensitive data during access control processes by enabling verification without revealing confidential details, thereby ensuring user privacy. Fuzzy logic facilitates adaptive, context-aware decision-making for access control by dynamically evaluating parameters such as data sensitivity, trust levels, and user roles. Blockchain technology, with its decentralized and immutable architecture, ensures transparency, traceability, and accountability using smart contracts that automate access control processes. The proposed framework addresses key challenges by enhancing security, reducing the likelihood of unauthorized access, and providing a transparent audit trail of data transactions. Expected outcomes include improved data privacy, accuracy in access control, and increased user trust in IoT systems. This research contributes significantly to advancing privacy-preserving, secure, and traceable solutions in IoT environments, laying the groundwork for future innovations in decentralized technologies and their applications in critical domains such as healthcare and beyond.

The automated extraction of structured questions from paper-based mathematics exams is fundamental to intelligent education, yet remains challenging in real-world settings due to severe visual noise. Existing benchmarks mainly focus on clean documents or generic layout analysis, overlooking both the structural integrity of mathematical problems and the ability of models to actively reject incomplete inputs. We introduce MathDoc, the first benchmark for document-level information extraction from authentic high school mathematics exam papers. MathDoc contains \textbf{3,609} carefully curated questions with real-world artifacts and explicitly includes unrecognizable samples to evaluate active refusal behavior. We propose a multi-dimensional evaluation framework covering stem accuracy, visual similarity, and refusal capability. Experiments on SOTA MLLMs, including Qwen3-VL and Gemini-2.5-Pro, show that although end-to-end models achieve strong extraction performance, they consistently fail to refuse illegible inputs, instead producing confident but invalid outputs. These results highlight a critical gap in current MLLMs and establish MathDoc as a benchmark for assessing model reliability under degraded document conditions. Our project repository is available at \href{https://github.com/winnk123/papers/tree/master}{GitHub repository}

Interactive humanoid video generation aims to synthesize lifelike visual agents that can engage with humans through continuous and responsive video. Despite recent advances in video synthesis, existing methods often grapple with the trade-off between high-fidelity synthesis and real-time interaction requirements. In this paper, we propose FlowAct-R1, a framework specifically designed for real-time interactive humanoid video generation. Built upon a MMDiT architecture, FlowAct-R1 enables the streaming synthesis of video with arbitrary durations while maintaining low-latency responsiveness. We introduce a chunkwise diffusion forcing strategy, complemented by a novel self-forcing variant, to alleviate error accumulation and ensure long-term temporal consistency during continuous interaction. By leveraging efficient distillation and system-level optimizations, our framework achieves a stable 25fps at 480p resolution with a time-to-first-frame (TTFF) of only around 1.5 seconds. The proposed method provides holistic and fine-grained full-body control, enabling the agent to transition naturally between diverse behavioral states in interactive scenarios. Experimental results demonstrate that FlowAct-R1 achieves exceptional behavioral vividness and perceptual realism, while maintaining robust generalization across diverse character styles.

Large language models are increasingly deployed in multi-agent systems for strategic tasks, yet how design choices such as role-based personas and payoff visibility affect reasoning remains poorly understood. We investigate whether multi-agent systems function as strategic reasoners capable of payoff optimization or as identity-driven actors that prioritize role alignment over explicit incentives. Using Nash equilibrium achievement as a diagnostic for strategic reasoning, we conduct systematic experiments across four LLM architectures (Qwen-7B, Qwen-32B, Llama-8B, Mistral-7B) in complex environmental decision-making games involving four agents. We show that role identity bias fundamentally alters strategic reasoning even when payoff-optimal equilibria exist and complete payoff information is available. Removing personas and providing explicit payoffs enables Qwen models to achieve high Nash equilibrium rates, indicating that both conditions are necessary for strategic reasoning. In contrast, personas systematically bias equilibrium selection toward socially preferred outcomes: with personas present, all of the achieved equilibria correspond to Green Transition, while models entirely fail to reach equilibrium when Tragedy of the Commons is payoff-optimal. The effect of explicit payoffs depends entirely on persona presence, revealing strong interactions between representational design choices. We also observe clear model-dependent patterns. Qwen architectures are highly sensitive to both personas and payoff visibility, whereas Llama and Mistral exhibit rigid reasoning behavior across conditions. These findings demonstrate that representational choices are substantive governance decisions that determine whether multi-agent systems act as strategic reasoners or identity-driven actors, with important implications for real-world deployment.

As knowledge and semantics on the web grow increasingly complex, enhancing Large Language Models (LLMs) comprehension and reasoning capabilities has become particularly important. Chain-of-Thought (CoT) prompting has been shown to enhance the reasoning capabilities of LLMs. However, it still falls short on logical reasoning tasks that rely on symbolic expressions and strict deductive rules. Neuro-symbolic methods address this gap by enforcing formal correctness through external solvers. Yet these solvers are highly format-sensitive, and small instabilities in model outputs can lead to frequent processing failures. LLM-driven approaches avoid parsing brittleness, but they lack structured representations and process-level error-correction mechanisms. To further enhance the logical reasoning capabilities of LLMs, we propose MatrixCoT, a structured CoT framework with a matrix-based plan. Specifically, we normalize and type natural language expressions, attach explicit citation fields, and introduce a matrix-based planning method to preserve global relations among steps. The plan becomes a verifiable artifact, making execution more stable. For verification, we also add a feedback-driven replanning mechanism. Under semantic-equivalence constraints, it identifies omissions and defects, rewrites and compresses the dependency matrix, and produces a more trustworthy final answer. Experiments on five logical-reasoning benchmarks and five LLMs show that, without relying on external solvers, MatrixCoT enhances both robustness and interpretability when tackling complex symbolic reasoning tasks, while maintaining competitive performance.

Multimodal models excel in English, supported by abundant image-text and audio-text data, but performance drops sharply for other languages due to limited multilingual multimodal resources. Existing solutions rely heavily on machine translation, while advances in multilingual text modeling remain underutilized. We introduce METAL, a lightweight alignment method that learns only a few linear layers using English text alone to map multilingual text embeddings into a multimodal space. Despite its simplicity, METAL matches baseline performance in English (94.9 percent Recall at 10) and achieves strong zero-shot transfer (89.5 percent Recall at 10 averaged across 11 languages, 10 unseen) on XTD text-to-image retrieval. Qualitative t-SNE visualizations show that multilingual embeddings align tightly with multimodal representations, while weight analysis reveals that the transformation reshapes embedding geometry rather than performing trivial rotations. Beyond image-text retrieval, METAL generalizes to audio-text retrieval and cross-lingual text-to-image generation. We release code and checkpoints at https://github.com/m2m-codebase/M2M , as well as multilingual evaluation datasets including MSCOCO Multilingual 30K (https://huggingface.co/datasets/piyushsinghpasi/mscoco-multilingual-30k ), AudioCaps Multilingual (https://huggingface.co/datasets/piyushsinghpasi/audiocaps-multilingual ), and Clotho Multilingual (https://huggingface.co/datasets/piyushsinghpasi/clotho-multilingual ), to facilitate further research.

Recent advances in multimodal learning have significantly enhanced the reasoning capabilities of vision-language models (VLMs). However, state-of-the-art approaches rely heavily on large-scale human-annotated datasets, which are costly and time-consuming to acquire. To overcome this limitation, we introduce V-Zero, a general post-training framework that facilitates self-improvement using exclusively unlabeled images. V-Zero establishes a co-evolutionary loop by instantiating two distinct roles: a Questioner and a Solver. The Questioner learns to synthesize high-quality, challenging questions by leveraging a dual-track reasoning reward that contrasts intuitive guesses with reasoned results. The Solver is optimized using pseudo-labels derived from majority voting over its own sampled responses. Both roles are trained iteratively via Group Relative Policy Optimization (GRPO), driving a cycle of mutual enhancement. Remarkably, without a single human annotation, V-Zero achieves consistent performance gains on Qwen2.5-VL-7B-Instruct, improving visual mathematical reasoning by +1.7 and general vision-centric by +2.6, demonstrating the potential of self-improvement in multimodal systems. Code is available at https://github.com/SatonoDia/V-Zero

Large programming courses struggle to provide timely, detailed feedback on student code. We developed Mark My Works, a local autograding system that combines traditional unit testing with LLM-generated explanations. The system uses role-based prompts to analyze submissions, critique code quality, and generate pedagogical feedback while maintaining transparency in its reasoning process. We piloted the system in a 191-student engineering course, comparing AI-generated assessments with human grading on 79 submissions. While AI scores showed no linear correlation with human scores (r = -0.177, p = 0.124), both systems exhibited similar left-skewed distributions, suggesting they recognize comparable quality hierarchies despite different scoring philosophies. The AI system demonstrated more conservative scoring (mean: 59.95 vs 80.53 human) but generated significantly more detailed technical feedback.

Multimodal clinical prediction is widely used to integrate heterogeneous data such as Electronic Health Records (EHR) and biosignals. However, existing methods tend to rely on static modality integration schemes and simple fusion strategies. As a result, they fail to fully exploit modality-specific representations. In this paper, we propose Level-guided Modal Fusion (LeMoF), a novel framework that selectively integrates level-guided representations within each modality. Each level refers to a representation extracted from a different layer of the encoder. LeMoF explicitly separates and learns global modality-level predictions from level-specific discriminative representations. This design enables LeMoF to achieve a balanced performance between prediction stability and discriminative capability even in heterogeneous clinical environments. Experiments on length of stay prediction using Intensive Care Unit (ICU) data demonstrate that LeMoF consistently outperforms existing state-of-the-art multimodal fusion techniques across various encoder configurations. We also confirmed that level-wise integration is a key factor in achieving robust predictive performance across various clinical conditions.

In this paper, we propose difficulty-guided sampling (DGS) to bridge the target gap between the distillation objective and the downstream task, therefore improving the performance of dataset distillation. Deep neural networks achieve remarkable performance but have time and storage-consuming training processes. Dataset distillation is proposed to generate compact, high-quality distilled datasets, enabling effective model training while maintaining downstream performance. Existing approaches typically focus on features extracted from the original dataset, overlooking task-specific information, which leads to a target gap between the distillation objective and the downstream task. We propose leveraging characteristics that benefit the downstream training into data distillation to bridge this gap. Focusing on the downstream task of image classification, we introduce the concept of difficulty and propose DGS as a plug-in post-stage sampling module. Following the specific target difficulty distribution, the final distilled dataset is sampled from image pools generated by existing methods. We also propose difficulty-aware guidance (DAG) to explore the effect of difficulty in the generation process. Extensive experiments across multiple settings demonstrate the effectiveness of the proposed methods. It also highlights the broader potential of difficulty for diverse downstream tasks.

The meta-analysis's utility is dependent on previous studies having accurately captured the variables of interest, but in medical studies, a key decision variable that impacts a physician's decisions was not captured. This results in an unknown effect size and unreliable conclusions. A Bayesian approach may allow analysis to determine if the claim of a positive effect is still warranted, and we build a Bayesian approach to this common medical scenario. To demonstrate its utility, we assist professional OBGYNs in evaluating Trial of Labor After a Cesarean-section (TOLAC) situations where few interventions are available for patients and find the support needed for physicians to advance patient care.

The past year has marked a turning point in the evolution and real-world use of large language models (LLMs). With the release of the first widely adopted reasoning model, o1, on December 5th, 2024, the field shifted from single-pass pattern generation to multi-step deliberation inference, accelerating deployment, experimentation, and new classes of applications. As this shift unfolded at a rapid pace, our empirical understanding of how these models have actually been used in practice has lagged behind. In this work, we leverage the OpenRouter platform, which is an AI inference provider across a wide variety of LLMs, to analyze over 100 trillion tokens of real-world LLM interactions across tasks, geographies, and time. In our empirical study, we observe substantial adoption of open-weight models, the outsized popularity of creative roleplay (beyond just the productivity tasks many assume dominate) and coding assistance categories, plus the rise of agentic inference. Furthermore, our retention analysis identifies foundational cohorts: early users whose engagement persists far longer than later cohorts. We term this phenomenon the Cinderella "Glass Slipper" effect. These findings underscore that the way developers and end-users engage with LLMs "in the wild" is complex and multifaceted. We discuss implications for model builders, AI developers, and infrastructure providers, and outline how a data-driven understanding of usage can inform better design and deployment of LLM systems.

Large Language Models (LLMs) are increasingly used in mental health-related settings, yet they struggle to sustain realistic, goal-directed dialogue over extended interactions. While LLMs generate fluent responses, they optimize locally for the next turn rather than maintaining a coherent model of therapeutic progress, leading to brittleness and long-horizon drift. We introduce CALM-IT, a framework for generating and evaluating long-form Motivational Interviewing (MI) dialogues that explicitly models dual-actor conversational dynamics. CALM-IT represents therapist-client interaction as a bidirectional state-space process, in which both agents continuously update inferred alignment, mental states, and short-term goals to guide strategy selection and utterance generation. Across large-scale evaluations, CALM-IT consistently outperforms strong baselines in Effectiveness and Goal Alignment and remains substantially more stable as conversation length increases. Although CALM-IT initiates fewer therapist redirections, it achieves the highest client acceptance rate (64.3%), indicating more precise and therapeutically aligned intervention timing. Overall, CALM-IT provides evidence for modeling evolving conversational state being essential for generating high-quality long-form synthetic conversations.

Machine learning classification systems are susceptible to poor performance when trained with incorrect ground truth labels, even when data is well-curated by expert annotators. As machine learning becomes more widespread, it is increasingly imperative to identify and correct mislabeling to develop more powerful models. In this work, we motivate and describe Adaptive Label Error Detection (ALED), a novel method of detecting mislabeling. ALED extracts an intermediate feature space from a deep convolutional neural network, denoises the features, models the reduced manifold of each class with a multidimensional Gaussian distribution, and performs a simple likelihood ratio test to identify mislabeled samples. We show that ALED has markedly increased sensitivity, without compromising precision, compared to established label error detection methods, on multiple medical imaging datasets. We demonstrate an example where fine-tuning a neural network on corrected data results in a 33.8% decrease in test set errors, providing strong benefits to end users. The ALED detector is deployed in the Python package statlab.

Communication in times of crisis is essential. However, there is often a mismatch between the language of governments, aid providers, doctors, and those to whom they are providing aid. Commercial MT systems are reasonable tools to turn to in these scenarios. But how effective are these tools for translating to and from low resource languages, particularly in the crisis or medical domain? In this study, we evaluate four commercial MT systems using the TICO-19 dataset, which is composed of pandemic-related sentences from a large set of high priority languages spoken by communities most likely to be affected adversely in the next pandemic. We then assess the current degree of ``readiness'' for another pandemic (or epidemic) based on the usability of the output translations.

Role-playing (RP) agents rely on behavioral profiles to act consistently across diverse narrative contexts, yet existing profiles are largely unstructured, non-executable, and weakly validated, leading to brittle agent behavior. We propose Codified Decision Trees (CDT), a data-driven framework that induces an executable and interpretable decision structure from large-scale narrative data. CDT represents behavioral profiles as a tree of conditional rules, where internal nodes correspond to validated scene conditions and leaves encode grounded behavioral statements, enabling deterministic retrieval of context-appropriate rules at execution time. The tree is learned by iteratively inducing candidate scene-action rules, validating them against data, and refining them through hierarchical specialization, yielding profiles that support transparent inspection and principled updates. Across multiple benchmarks, CDT substantially outperforms human-written profiles and prior profile induction methods on $85$ characters across $16$ artifacts, indicating that codified and validated behavioral representations lead to more reliable agent grounding.

Reinforcement Learning (RL) has become essential for eliciting complex reasoning capabilities in Large Language Models (LLMs). However, the substantial memory overhead of storing Key-Value (KV) caches during long-horizon rollouts acts as a critical bottleneck, often prohibiting efficient training on limited hardware. While existing KV compression techniques offer a remedy for inference, directly applying them to RL training induces a severe policy mismatch, leading to catastrophic performance collapse. To address this, we introduce Sparse-RL empowers stable RL training under sparse rollouts. We show that instability arises from a fundamental policy mismatch among the dense old policy, the sparse sampler policy, and the learner policy. To mitigate this issue, Sparse-RL incorporates Sparsity-Aware Rejection Sampling and Importance-based Reweighting to correct the off-policy bias introduced by compression-induced information loss. Experimental results show that Sparse-RL reduces rollout overhead compared to dense baselines while preserving the performance. Furthermore, Sparse-RL inherently implements sparsity-aware training, significantly enhancing model robustness during sparse inference deployment.

In 1888, Vincent van Gogh wrote, "I am seeking exaggeration in the essential." This principle, amplifying structural form while suppressing photographic detail, lies at the core of Post-Impressionist art. However, most existing 3D style transfer methods invert this philosophy, treating geometry as a rigid substrate for surface-level texture projection. To authentically reproduce Post-Impressionist stylization, geometric abstraction must be embraced as the primary vehicle of expression. We propose a flow-guided geometric advection framework for 3D Gaussian Splatting (3DGS) that operationalizes this principle in a mesh-free setting. Our method extracts directional flow fields from 2D paintings and back-propagates them into 3D space, rectifying Gaussian primitives to form flow-aligned brushstrokes that conform to scene topology without relying on explicit mesh priors. This enables expressive structural deformation driven directly by painterly motion rather than photometric constraints. Our contributions are threefold: (1) a projection-based, mesh-free flow guidance mechanism that transfers 2D artistic motion into 3D Gaussian geometry; (2) a luminance-structure decoupling strategy that isolates geometric deformation from color optimization, mitigating artifacts during aggressive structural abstraction; and (3) a VLM-as-a-Judge evaluation framework that assesses artistic authenticity through aesthetic judgment instead of conventional pixel-level metrics, explicitly addressing the subjective nature of artistic stylization.

We introduce ReaMIL (Reasoning- and Evidence-Aware MIL), a multiple instance learning approach for whole-slide histopathology that adds a light selection head to a strong MIL backbone. The head produces soft per-tile gates and is trained with a budgeted-sufficiency objective: a hinge loss that enforces the true-class probability to be $\geq τ$ using only the kept evidence, under a sparsity budget on the number of selected tiles. The budgeted-sufficiency objective yields small, spatially compact evidence sets without sacrificing baseline performance. Across TCGA-NSCLC (LUAD vs. LUSC), TCGA-BRCA (IDC vs. Others), and PANDA, ReaMIL matches or slightly improves baseline AUC and provides quantitative evidence-efficiency diagnostics. On NSCLC, it attains AUC 0.983 with a mean minimal sufficient K (MSK) $\approx 8.2$ tiles at $τ= 0.90$ and AUKC $\approx 0.864$, showing that class confidence rises sharply and stabilizes once a small set of tiles is kept. The method requires no extra supervision, integrates seamlessly with standard MIL training, and naturally yields slide-level overlays. We report accuracy alongside MSK, AUKC, and contiguity for rigorous evaluation of model behavior on WSIs.

Assessment of sperm morphological quality remains a critical yet subjective component of male fertility evaluation, often limited by inter-observer variability and resource constraints. This study presents a comparative biomedical artificial intelligence framework evaluating an image-based deep learning model (HuSHeM) alongside a clinically grounded baseline derived from World Health Organization criteria augmented with the Systemic Inflammation Response Index (WHO(+SIRI)). The HuSHeM model was trained on high-resolution sperm morphology images and evaluated using an independent clinical cohort. Model performance was assessed using discrimination, calibration, and clinical utility analyses. The HuSHeM model demonstrated higher discriminative performance, as reflected by an increased area under the receiver operating characteristic curve with relatively narrow confidence intervals compared to WHO(+SIRI). Precision-recall analysis further indicated improved performance under class imbalance, with higher precision-recall area values across evaluated thresholds. Calibration analysis indicated closer agreement between predicted probabilities and observed outcomes for HuSHeM, while decision curve analysis suggested greater net clinical benefit across clinically relevant threshold probabilities. These findings suggest that image-based deep learning may offer improved predictive reliability and clinical utility compared with traditional rule-based and inflammation-augmented criteria. The proposed framework supports objective and reproducible assessment of sperm morphology and may serve as a decision-support tool within fertility screening and referral workflows. The proposed models are intended as decision-support or referral tools and are not designed to replace clinical judgment or laboratory assessment.

Hybrid testing that integrates fuzzing, symbolic execution, and sampling has demonstrated superior testing efficiency compared to individual techniques. However, the state-of-the-art (SOTA) hybrid testing tools do not fully exploit the capabilities of symbolic execution and sampling in two key aspects. First, the SOTA hybrid testing tools employ tailored symbolic execution engines that tend to over-prune branches, leading to considerable time wasted waiting for seeds from the fuzzer and missing opportunities to discover crashes. Second, existing methods do not apply sampling to the appropriate branches and therefore cannot utilize the full capability of sampling. To address these two limitations, we propose a novel hybrid testing architecture that combines the precision of conventional symbolic execution with the scalability of tailored symbolic execution engines. Based on this architecture, we propose several principles for combining fuzzing, symbolic execution, and sampling. We implement our method in a hybrid testing tool S$^2$F. To evaluate its effectiveness, we conduct extensive experiments on 15 real-world programs. Experimental results demonstrate that S$^2$F outperforms the SOTA tool, achieving an average improvement of 6.14% in edge coverage and 32.6% in discovered crashes. Notably, our tool uncovers three previously unknown crashes in real-world programs.

Content-based recommendation systems (CRSs) utilize content features to predict user-item interactions, serving as essential tools for helping users navigate information-rich web services. However, ensuring the effectiveness of CRSs requires large-scale and even continuous model training to accommodate diverse user preferences, resulting in significant computational costs and resource demands. A promising approach to this challenge is coreset selection, which identifies a small but representative subset of data samples that preserves model quality while reducing training overhead. Yet, the selected coreset is vulnerable to the pervasive noise in user-item interactions, particularly when it is minimally sized. To this end, we propose Noise-aware Coreset Selection (NaCS), a specialized framework for CRSs. NaCS constructs coresets through submodular optimization based on training gradients, while simultaneously correcting noisy labels using a progressively trained model. Meanwhile, we refine the selected coreset by filtering out low-confidence samples through uncertainty quantification, thereby avoid training with unreliable interactions. Through extensive experiments, we show that NaCS produces higher-quality coresets for CRSs while achieving better efficiency than existing coreset selection techniques. Notably, NaCS recovers 93-95\% of full-dataset training performance using merely 1\% of the training data. The source code is available at \href{https://github.com/chenxing1999/nacs}{https://github.com/chenxing1999/nacs}.

Large Language Models (LLMs) have demonstrated remarkable reasoning capabilities, particularly in solving complex mathematical problems. Recent studies show that distilling long reasoning trajectories can effectively enhance the reasoning performance of small-scale student models. However, teacher-generated reasoning trajectories are often excessively long and structurally complex, making them difficult for student models to learn. This mismatch leads to a gap between the provided supervision signal and the learning capacity of the student model. To address this challenge, we propose Prefix-ALIGNment distillation (P-ALIGN), a framework that fully exploits teacher CoTs for distillation through adaptive prefix alignment. Specifically, P-ALIGN adaptively truncates teacher-generated reasoning trajectories by determining whether the remaining suffix is concise and sufficient to guide the student model. Then, P-ALIGN leverages the teacher-generated prefix to supervise the student model, encouraging effective prefix alignment. Experiments on multiple mathematical reasoning benchmarks demonstrate that P-ALIGN outperforms all baselines by over 3%. Further analysis indicates that the prefixes constructed by P-ALIGN provide more effective supervision signals, while avoiding the negative impact of redundant and uncertain reasoning components. All code is available at https://github.com/NEUIR/P-ALIGN.

Recent video generation models have revealed the emergence of Chain-of-Frame (CoF) reasoning, enabling frame-by-frame visual inference. With this capability, video models have been successfully applied to various visual tasks (e.g., maze solving, visual puzzles). However, their potential to enhance text-to-image (T2I) generation remains largely unexplored due to the absence of a clearly defined visual reasoning starting point and interpretable intermediate states in the T2I generation process. To bridge this gap, we propose CoF-T2I, a model that integrates CoF reasoning into T2I generation via progressive visual refinement, where intermediate frames act as explicit reasoning steps and the final frame is taken as output. To establish such an explicit generation process, we curate CoF-Evol-Instruct, a dataset of CoF trajectories that model the generation process from semantics to aesthetics. To further improve quality and avoid motion artifacts, we enable independent encoding operation for each frame. Experiments show that CoF-T2I significantly outperforms the base video model and achieves competitive performance on challenging benchmarks, reaching 0.86 on GenEval and 7.468 on Imagine-Bench. These results indicate the substantial promise of video models for advancing high-quality text-to-image generation.

Large language models (LLMs) exhibit impressive in-context learning (ICL) capabilities, yet the quality of their predictions is fundamentally limited by the few costly labeled demonstrations that can fit into a prompt. Meanwhile, there exist vast and continuously growing amounts of unlabeled data that may be closely related to the ICL task. How to utilize such unlabeled data to provably enhance the performance of ICL thus becomes an emerging fundamental question. In this work, we propose a novel augmented ICL framework, in which the prompt includes a small set of labeled examples alongside a block of unlabeled inputs. We focus on the multi-class linear classification setting and demonstrate that, with chain-of-thought (CoT) prompting, a multi-layer transformer can effectively emulate an expectation-maximization (EM) algorithm. This enables the transformer to implicitly extract useful information from both labeled and unlabeled data, leading to provable improvements in ICL accuracy. Moreover, we show that such a transformer can be trained via teacher forcing, with its parameters converging to the desired solution at a linear rate. Experiments demonstrate that the augmented ICL framework consistently outperforms conventional few-shot ICL, providing empirical support for our theoretical findings. To the best of our knowledge, this is the first theoretical study on the impact of unlabeled data on the ICL performance of transformers.

Particularly, financial named-entity recognition (NER) is one of the many important approaches to translate unformatted reports and news into structured knowledge graphs. However, free, easy-to-use large language models (LLMs) often fail to differentiate organisations as people, or disregard an actual monetary amount entirely. This paper takes Meta's Llama 3 8B and applies it to financial NER by combining instruction fine-tuning and Low-Rank Adaptation (LoRA). Each annotated sentence is converted into an instruction-input-output triple, enabling the model to learn task descriptions while fine-tuning with small low-rank matrices instead of updating all weights. Using a corpus of 1,693 sentences, our method obtains a micro-F1 score of 0.894 compared with Qwen3-8B, Baichuan2-7B, T5, and BERT-Base. We present dataset statistics, describe training hyperparameters, and perform visualizations of entity density, learning curves, and evaluation metrics. Our results show that instruction tuning combined with parameter-efficient fine-tuning enables state-of-the-art performance on domain-sensitive NER.

Artificial intelligence is rapidly transforming astronomical research, yet the scientific community has largely treated this transformation as an engineering challenge rather than an epistemological one. This perspective article argues that philosophy of science offers essential tools for navigating AI's integration into astronomy--conceptual clarity about what "understanding" means, critical examination of assumptions about data and discovery, and frameworks for evaluating AI's roles across different research contexts. Drawing on an interdisciplinary workshop convening astronomers, philosophers, and computer scientists, we identify several tensions. First, the narrative that AI will "derive fundamental physics" from data misconstrues contemporary astronomy as equation-derivation rather than the observation-driven enterprise it is. Second, scientific understanding involves more than prediction--it requires narrative construction, contextual judgment, and communicative achievement that current AI architectures struggle to provide. Third, because narrative and judgment matter, human peer review remains essential--yet AI-generated content flooding the literature threatens our capacity to identify genuine insight. Fourth, while AI excels at well-defined problem-solving, the ill-defined problem-finding that drives breakthroughs appears to require capacities beyond pattern recognition. Fifth, as AI accelerates what is feasible, pursuitworthiness criteria risk shifting toward what AI makes easy rather than what is genuinely important. We propose "pragmatic understanding" as a framework for integration--recognizing AI as a tool that extends human cognition while requiring new norms for validation and epistemic evaluation. Engaging with these questions now may help the community shape the transformation rather than merely react to it.

Neuromorphic computers hold the potential to vastly improve the speed and efficiency of a wide range of computational kernels with their asynchronous, compute-memory co-located, spatially distributed, and scalable nature. However, performance models that are simple yet sufficiently expressive to predict runtime on actual neuromorphic hardware are lacking, posing a challenge for researchers and developers who strive to design fast algorithms and kernels. As breaking the memory bandwidth wall of conventional von-Neumann architectures is a primary neuromorphic advantage, modeling communication time is especially important. At the same time, modeling communication time is difficult, as complex congestion patterns arise in a heavily-loaded Network-on-Chip. In this work, we introduce the first max-affine lower-bound runtime model -- a multi-dimensional roofline model -- for Intel's Loihi 2 neuromorphic chip that quantitatively accounts for both compute and communication based on a suite of microbenchmarks. Despite being a lower-bound model, we observe a tight correspondence (Pearson correlation coefficient greater than or equal to 0.97) between our model's estimated runtime and the measured runtime on Loihi 2 for a neural network linear layer, i.e., matrix-vector multiplication, and for an example application, a Quadratic Unconstrained Binary Optimization solver. Furthermore, we derive analytical expressions for communication-bottlenecked runtime to study scalability of the linear layer, revealing an area-runtime tradeoff for different spatial workload configurations with linear to superliner runtime scaling in layer size with a variety of constant factors. Our max-affine runtime model helps empower the design of high-speed algorithms and kernels for Loihi 2.

This paper introduces EmplifAI, a Japanese empathetic dialogue dataset designed to support patients coping with chronic medical conditions. They often experience a wide range of positive and negative emotions (e.g., hope and despair) that shift across different stages of disease management. EmplifAI addresses this complexity by providing situation-based dialogues grounded in 28 fine-grained emotion categories, adapted and validated from the GoEmotions taxonomy. The dataset includes 280 medically contextualized situations and 4125 two-turn dialogues, collected through crowdsourcing and expert review. To evaluate emotional alignment in empathetic dialogues, we assessed model predictions on situation--dialogue pairs using BERTScore across multiple large language models (LLMs), achieving F1 scores of 0.83. Fine-tuning a baseline Japanese LLM (LLM-jp-3.1-13b-instruct4) with EmplifAI resulted in notable improvements in fluency, general empathy, and emotion-specific empathy. Furthermore, we compared the scores assigned by LLM-as-a-Judge and human raters on dialogues generated by multiple LLMs to validate our evaluation pipeline and discuss the insights and potential risks derived from the correlation analysis.

The scientific computation of large deformations in elastic-plastic solids is crucial in various manufacturing applications. Traditional numerical methods exhibit several inherent limitations, prompting Deep Learning (DL) as a promising alternative. The effectiveness of current DL techniques typically depends on the availability of high-quantity and high-accuracy datasets, which are yet difficult to obtain in large deformation problems. During the dataset construction process, a dilemma stands between data quantity and data accuracy, leading to suboptimal performance in the DL models. To address this challenge, we focus on a representative application of large deformations, the stretch bending problem, and propose FilDeep, a Fidelity-based Deep Learning framework for large Deformation of elastic-plastic solids. Our FilDeep aims to resolve the quantity-accuracy dilemma by simultaneously training with both low-fidelity and high-fidelity data, where the former provides greater quantity but lower accuracy, while the latter offers higher accuracy but in less quantity. In FilDeep, we provide meticulous designs for the practical large deformation problem. Particularly, we propose attention-enabled cross-fidelity modules to effectively capture long-range physical interactions across MF data. To the best of our knowledge, our FilDeep presents the first DL framework for large deformation problems using MF data. Extensive experiments demonstrate that our FilDeep consistently achieves state-of-the-art performance and can be efficiently deployed in manufacturing.

Academic paper search is a fundamental task in scientific research, yet most existing approaches rely on rigid, predefined workflows that struggle with complex, conditional queries. To address this limitation, we propose PaperScout, an autonomous agent that reformulates paper search as a sequential decision-making process. Unlike static workflows, PaperScout dynamically decides whether, when, and how to invoke search and expand tools based on accumulated retrieval context. However, training such agents presents a fundamental challenge: standard reinforcement learning methods, typically designed for single-turn tasks, suffer from a granularity mismatch when applied to multi-turn agentic tasks, where token-level optimization diverges from the granularity of sequence-level interactions, leading to noisy credit assignment. We introduce Proximal Sequence Policy Optimization (PSPO), a process-aware, sequence-level policy optimization method that aligns optimization with agent-environment interaction. Comprehensive experiments on both synthetic and real-world benchmarks demonstrate that PaperScout significantly outperforms strong workflow-driven and RL baselines in both recall and relevance, validating the effectiveness of our adaptive agentic framework and optimization strategy.

Coded polynomial aggregation (CPA) enables the master to directly recover a weighted aggregation of polynomial evaluations without individually decoding each term, thereby reducing the number of required worker responses. In this paper, we extend CPA to straggler-aware distributed computing systems and introduce a straggler-aware CPA framework with pre-specified non-straggler patterns, where exact recovery is required only for a given collection of admissible non-straggler sets. Our main result shows that exact recovery of the desired aggregation is achievable with fewer worker responses than required by polynomial coded computing based on individual decoding, and that feasibility is fundamentally characterized by the intersection structure of the non-straggler patterns. In particular, we establish necessary and sufficient conditions for exact recovery in straggler-aware CPA and identify an intersection-size threshold that is sufficient to guarantee exact recovery. We further prove that this threshold becomes both necessary and sufficient when the number of admissible non-straggler sets is sufficiently large. We also provide an explicit construction of feasible CPA schemes whenever the intersection size exceeds the derived threshold. Finally, simulations reveal a sharp feasibility transition at the predicted threshold, providing empirical evidence that the bound is tight in practice.

Large Language Models (LLMs) are increasingly shaping human-computer interaction (HCI), from personalized assistants to social simulations. Beyond language competence, researchers are exploring whether LLMs can exhibit human-like characteristics that influence engagement, decision-making, and perceived realism. Personality, in particular, is critical, yet existing approaches often struggle to achieve both nuanced and adaptable expression. We present a framework that models LLM personality via Jungian psychological types, integrating three mechanisms: a dominant-auxiliary coordination mechanism for coherent core expression, a reinforcement-compensation mechanism for temporary adaptation to context, and a reflection mechanism that drives long-term personality evolution. This design allows the agent to maintain nuanced traits while dynamically adjusting to interaction demands and gradually updating its underlying structure. Personality alignment is evaluated using Myers-Briggs Type Indicator questionnaires and tested under diverse challenge scenarios as a preliminary structured assessment. Findings suggest that evolving, personality-aware LLMs can support coherent, context-sensitive interactions, enabling naturalistic agent design in HCI.

Ensemble learning is widely recognized as a pivotal strategy for pushing the boundaries of predictive performance. Traditional static ensemble methods, such as Stacking, typically assign weights by treating each base learner as a holistic entity, thereby overlooking the fact that individual models exhibit varying degrees of competence across different regions of the instance space. To address this limitation, Dynamic Ensemble Selection (DES) was introduced. However, both static and dynamic approaches predominantly rely on the divergence among different models as the basis for integration. This inter-model perspective neglects the intrinsic characteristics of the models themselves and necessitates a heavy reliance on validation sets for competence estimation. In this paper, we propose the Behavioral Profiling Ensemble (BPE) framework, which introduces a novel paradigm shift. Unlike traditional methods, BPE constructs a ``behavioral profile'' intrinsic to each model and derives integration weights based on the deviation between the model's response to a specific test instance and its established behavioral profile. Extensive experiments on both synthetic and real-world datasets demonstrate that the algorithm derived from the BPE framework achieves significant improvements over state-of-the-art ensemble baselines. These gains are evident not only in predictive accuracy but also in computational efficiency and storage resource utilization across various scenarios.

Clinical decision-making increasingly relies on timely and context-aware access to patient information within Electronic Health Records (EHRs), yet most existing natural language question-answering (QA) systems are evaluated solely on benchmark datasets, limiting their practical relevance. To overcome this limitation, we introduce EHRNavigator, a multi-agent framework that harnesses AI agents to perform patient-level question answering across heterogeneous and multimodal EHR data. We assessed its performance using both public benchmark and institutional datasets under realistic hospital conditions characterized by diverse schemas, temporal reasoning demands, and multimodal evidence integration. Through quantitative evaluation and clinician-validated chart review, EHRNavigator demonstrated strong generalization, achieving 86% accuracy on real-world cases while maintaining clinically acceptable response times. Overall, these findings confirm that EHRNavigator effectively bridges the gap between benchmark evaluation and clinical deployment, offering a robust, adaptive, and efficient solution for real-world EHR question answering.

This paper studies time aggregation features for XGBoost models in click-through rate prediction. The setting is the Avazu click-through rate prediction dataset with strict out-of-time splits and a no-lookahead feature constraint. Features for hour H use only impressions from hours strictly before H. This paper compares a strong time-aware target encoding baseline to models augmented with entity history time aggregation under several window designs. Across two rolling-tail folds on a deterministic ten percent sample, a trailing window specification improves ROC AUC by about 0.0066 to 0.0082 and PR AUC by about 0.0084 to 0.0094 relative to target encoding alone. Within the time aggregation design grid, event count windows provide the only consistent improvement over trailing windows, and the gain is small. Gap windows and bucketized windows underperform simple trailing windows in this dataset and protocol. These results support a practical default of trailing windows, with an optional event count window when marginal ROC AUC gains matter.

While high-quality technology support can assist older adults in using digital applications, many struggle to articulate their issues due to unfamiliarity with technical terminology and age-related cognitive changes. This study examines these communication challenges and explores AI-based approaches to mitigate them. We conducted a diary study with English-speaking, community-dwelling older adults to collect asynchronous, technology-related queries and used reflexive thematic analysis to identify communication barriers. To address these barriers, we evaluated how foundation models can paraphrase older adults' queries to improve solution accuracy. Two controlled experiments followed: one with younger adults evaluating AI-rephrased queries and another with older adults evaluating AI-generated solutions. We also developed a pipeline using large language models to generate the first synthetic dataset of how older adults request tech support (OATS). We identified four key communication challenges: verbosity, incompleteness, over-specification, and under-specification. Our prompt-chaining approach using the large language model, GPT-4o, elicited contextual details, paraphrased the original query, and generated a solution. AI-rephrased queries significantly improved solution accuracy (69% vs. 46%) and Google search results (69% vs. 35%). Younger adults better understood AI-rephrased queries (93.7% vs. 65.8%) and reported greater confidence and ease. Older adults reported high perceived ability to answer contextual questions (89.8%) and follow solutions (94.7%), with high confidence and ease. OATS demonstrated strong fidelity and face validity. This work shows how foundation models can enhance technology support for older adults by addressing age-related communication barriers. The OATS dataset offers a scalable resource for developing equitable AI systems that better serve aging populations.

Personalized Federated Learning (PFL) enables collaboratively model training on decentralized, heterogeneous data while tailoring them to each client's unique distribution. However, existing PFL methods produce static models with a fixed tradeoff between accuracy and efficiency, limiting their applicability in environments where inference requirements vary with contexts and resource availability. Early-exit networks (EENs) offer adaptive inference by attaching intermediate classifiers. Yet integrating them into PFL is challenging due to client-wise heterogeneity and depth-wise interference arising from conflicting exit objectives. Prior studies fail to resolve both conflicts simultaneously, leading to suboptimal performance. In this paper, we propose CAFEDistill, a Conflict-Aware Federated Exit Distillation framework that jointly addresses these conflicts and extends PFL to early-exit networks. Through a progressive, depth-prioritized student coordination mechanism, CAFEDistill mitigates interference among shallow and deep exits while allowing effective personalized knowledge transfer across clients. Furthermore, it reduces communication overhead via a client-decoupled formulation. Extensive evaluations show that CAFEDistill outperforms the state-of-the-arts, achieving higher accuracy and reducing inference costs by 30.79%-46.86%.

Federated learning (FL) enables collaborative model training without sharing raw user data, but conventional simulations often rely on unrealistic data partitioning and current user selection methods ignore data correlation among users. To address these challenges, this paper proposes a metadatadriven FL framework. We first introduce a novel data partition model based on a homogeneous Poisson point process (HPPP), capturing both heterogeneity in data quantity and natural overlap among user datasets. Building on this model, we develop a clustering-based user selection strategy that leverages metadata, such as user location, to reduce data correlation and enhance label diversity across training rounds. Extensive experiments on FMNIST and CIFAR-10 demonstrate that the proposed framework improves model performance, stability, and convergence in non-IID scenarios, while maintaining comparable performance under IID settings. Furthermore, the method shows pronounced advantages when the number of selected users per round is small. These findings highlight the framework's potential for enhancing FL performance in realistic deployments and guiding future standardization.

Multimodal decentralized federated learning (DFL) is challenging because agents differ in available modalities and model architectures, yet must collaborate over peer-to-peer (P2P) networks without a central coordinator. Standard multimodal pipelines learn a single shared embedding across all modalities. In DFL, such a monolithic representation induces gradient misalignment between uni- and multimodal agents; as a result, it suppresses heterogeneous sharing and cross-modal interaction. We present PARSE, a multimodal DFL framework that operationalizes partial information decomposition (PID) in a server-free setting. Each agent performs feature fission to factorize its latent representation into redundant, unique, and synergistic slices. P2P knowledge sharing among heterogeneous agents is enabled by slice-level partial alignment: only semantically shareable branches are exchanged among agents that possess the corresponding modality. By removing the need for central coordination and gradient surgery, PARSE resolves uni-/multimodal gradient conflicts, thereby overcoming the multimodal DFL dilemma while remaining compatible with standard DFL constraints. Across benchmarks and agent mixes, PARSE yields consistent gains over task-, modality-, and hybrid-sharing DFL baselines. Ablations on fusion operators and split ratios, together with qualitative visualizations, further demonstrate the efficiency and robustness of the proposed design.

Existing NL2SQL systems face two critical limitations: (1) they rely on in-context learning with only correct examples, overlooking the rich signal in historical error-fix pairs that could guide more robust self-correction; and (2) test-time scaling approaches often decompose questions arbitrarily, producing near-identical SQL candidates across runs and diminishing ensemble gains. Moreover, these methods suffer from a stark accuracy-efficiency trade-off: high performance demands excessive computation, while fast variants compromise quality. We present Memo-SQL, a training-free framework that addresses these issues through two simple ideas: structured decomposition and experience-aware self-correction. Instead of leaving decomposition to chance, we apply three clear strategies, entity-wise, hierarchical, and atomic sequential, to encourage diverse reasoning. For correction, we build a dynamic memory of both successful queries and historical error-fix pairs, and use retrieval-augmented prompting to bring relevant examples into context at inference time, no fine-tuning or external APIs required. On BIRD, Memo-SQL achieves 68.5% execution accuracy, setting a new state of the art among open, zero-fine-tuning methods, while using over 10 times fewer resources than prior TTS approaches.

Video Large Language Models (VideoLLMs) exhibit various types of hallucinations. Existing research has primarily focused on hallucinations involving the presence of events, objects, and scenes in videos, while largely neglecting event relation hallucination. In this paper, we introduce a novel benchmark for evaluating the Video Event Relation Hallucination, named VERHallu. This benchmark focuses on causal, temporal, and subevent relations between events, encompassing three types of tasks: relation classification, question answering, and counterfactual question answering, for a comprehensive evaluation of event relation hallucination. Additionally, it features counterintuitive video scenarios that deviate from typical pretraining distributions, with each sample accompanied by human-annotated candidates covering both vision-language and pure language biases. Our analysis reveals that current state-of-the-art VideoLLMs struggle with dense-event relation reasoning, often relying on prior knowledge due to insufficient use of frame-level cues. Although these models demonstrate strong grounding capabilities for key events, they often overlook the surrounding subevents, leading to an incomplete and inaccurate understanding of event relations. To tackle this, we propose a Key-Frame Propagating (KFP) strategy, which reallocates frame-level attention within intermediate layers to enhance multi-event understanding. Experiments show it effectively mitigates the event relation hallucination without affecting inference speed.

We present a hybrid transformer architecture that replaces discrete middle layers with a continuous-depth Neural Ordinary Differential Equation (ODE) block, enabling inference-time control over generation attributes via a learned steering signal. Unlike standard transformers that process representations through fixed discrete layers, our approach treats depth as a continuous variable governed by a learned vector field $F_θ(H, τ, u)$, where $u$ is a low-dimensional control signal injected via explicit concatenation. We validate the architecture through four experiments: (1) gradient flow stability with zero exploding/vanishing gradient events, (2) semantic steering achieving 98\%/88\% accuracy for positive/negative sentiment control, (3) continuous interpolation validated by a negligible 0.068\% trajectory divergence between fixed and adaptive solvers, and (4) efficiency benchmarking demonstrating latency parity with standard discrete baselines. Additionally, we show that adaptive ODE solvers reveal geometric structure in the learned dynamics: the control signal partitions the vector field into distinct dynamical regimes with different curvature characteristics. The adjoint method enables $O(1)$ memory training regardless of integration depth. Our results demonstrate that continuous-depth dynamics with learned control signals provide a viable, efficient mechanism for steerable language generation.

Large Language Models (LLMs) are increasingly adopted in healthcare to support clinical decision-making, summarize electronic health records (EHRs), and enhance patient care. However, this integration introduces significant privacy and security challenges, driven by the sensitivity of clinical data and the high-stakes nature of medical workflows. These risks become even more pronounced across heterogeneous deployment environments, ranging from small on-premise hospital systems to regional health networks, each with unique resource limitations and regulatory demands. This Systematization of Knowledge (SoK) examines the evolving threat landscape across the three core LLM phases: Data preprocessing, Fine-tuning, and Inference within realistic healthcare settings. We present a detailed threat model that characterizes adversaries, capabilities, and attack surfaces at each phase, and we systematize how existing privacy-preserving techniques (PPTs) attempt to mitigate these vulnerabilities. While existing defenses show promise, our analysis identifies persistent limitations in securing sensitive clinical data across diverse operational tiers. We conclude with phase-aware recommendations and future research directions aimed at strengthening privacy guarantees for LLMs in regulated environments. This work provides a foundation for understanding the intersection of LLMs, threats, and privacy in healthcare, offering a roadmap toward more robust and clinically trustworthy AI systems.

Constructing Knowledge Graphs (KGs) from unstructured text provides a structured framework for knowledge representation and reasoning, yet current LLM-based approaches struggle with a fundamental trade-off: factual coverage often leads to relational fragmentation, while premature consolidation causes information loss. To address this, we propose SocraticKG, an automated KG construction method that introduces question-answer pairs as a structured intermediate representation to systematically unfold document-level semantics prior to triple extraction. By employing 5W1H-guided QA expansion, SocraticKG captures contextual dependencies and implicit relational links typically lost in direct KG extraction pipelines, providing explicit grounding in the source document that helps mitigate implicit reasoning errors. Evaluation on the MINE benchmark demonstrates that our approach effectively addresses the coverage-connectivity trade-off, achieving superior factual retention while maintaining high structural cohesion even as extracted knowledge volume substantially expands. These results highlight that QA-mediated semantic scaffolding plays a critical role in structuring semantics prior to KG extraction, enabling more coherent and reliable graph construction in subsequent stages.

We establish a majorization-based theory for bounding observables of waves with varied coherence. For any measurement, exact bounds are attained by the maximal and minimal elements in the set of input coherence spectra. The set's supremum and infimum, which may lie outside the set, provide optimal universal bounds: any alternative spectrum yielding universal bounds produces weaker constraints. We present an algorithm to compute the supremum and infimum, and prove that they lie either at singular boundary points or strictly outside the set of coherence spectra.

The impact of disorder on collective phenomena depends crucially on whether it is quenched or annealed. In synchronization problems, quenched disorder in higher dimensional Kuramoto models is known to produce unconventional dimensional effects, including a striking odd even dichotomy: synchronization transitions are continuous in even dimensions and discontinuous in odd dimensions. By contrast, the impact of annealed disorder has received comparatively little attention. Here we study a D dimensional Kuramoto model with both fundamental and higher-harmonic interactions under annealed disorder, and develop an arbitrary dimensional center-manifold framework to analyze the nonlinear dynamics near the onset of collective behavior. We show that annealed disorder fundamentally alters the role of dimensionality. With fundamental coupling alone, it completely removes the odd even dichotomy, yielding continuous synchronization transitions with universal mean-field scaling in all dimensions. Higher-harmonic interactions preserve this universality while rendering the synchronization transition tunable between continuous and discontinuous. At the same time, they give rise to a novel, correlation-driven transition between a symmetry-protected incoherent phase and a symmetry broken state lacking global synchronization, which is therefore invisible to the conventional Kuramoto order parameter. This transition is continuous in two dimensions but discontinuous in higher dimensions, revealing an emergent and previously-unrecognized special role of two dimensions.

The dynamic motion of topological defects in magnets induces an emergent electric field, as exemplified by the continuous flow of skyrmion vortices. However, the electrodynamics underlying this emergent field remains poorly understood. In this context, magnetic domain walls - one dimensional topological defects with two collective modes, sliding and spin tilt - offer a promising platform for exploration. Here, we demonstrate that the dissipative motion of domain walls under oscillatory current excitation generates an emergent electric field. We image domain patterns and quantify domain wall length under applied magnetic fields in mesoscopic devices based on the magnetic Weyl semimetal NdAlSi. These devices exhibit exceptionally strong domain wall scattering and a pronounced emergent electric field, observed in the imaginary component of the complex impedance. Spin dynamics simulations reveal that domain wall sliding dominates over spin tilting, where the phase delay of the domain wall motion with respect to the driving force impacts the emergent electric field. Our findings establish domain-wall dynamics as a platform for studying emergent electromagnetic fields and motivate further investigations on the coupled motion of magnetic solitons and conduction electrons.

Two-dimensional polyaramids exhibit strong hydrogen bonding to create molecularly thin nanosheets analogous to graphene. Here, we report the first nanomechanical resonators made out of a two-dimensional polyaramid, 2DPA-1, with thicknesses as small as 8 nm. To fabricate these molecular-scale resonators, we transferred nanofilms of 2DPA-1 onto chips with previously etched arrays of circular microwells. We then characterized the thermal resonances of these resonators under different conditions. When there is no residual gas inside the 2DPA-1-covered microwells, the eigenfrequencies are well-described by a tensioned plate theory, providing the Young's modulus and tension of the 2DPA-1 nanofilms. With gas present, the nanofilms bulge up and mechanical resonances are modified due to the adhesion, bulging and slack present in the system. The fabrication and mechanical characterization of these first 2DPA-1 nanomechanical resonators represent a convincing path toward molecular-scale polymeric NEMS with high mechanical strength, low density, and synthetic processability.

Advances in the next generation of mesoscopic electronics require an understanding of topological phases in inhomogeneous media and the principles that govern them. Motivated by the nature of motifs available in printable conducting inks, we introduce and study quantum transport in a minimal model that describes a bundle of one-dimensional metallic wires that are randomly interconnected by semiconducting chains. Each of these interconnects is represented by a Su-Schrieffer-Heeger chain, which can reside in either a trivial or a topological phase. Using a tight-binding approach, we show that such a system can transit from an insulating phase to a robust metallic phase as the interconnects undergo a transition from a trivial to a topological phase. In the latter, despite the random interconnectedness, the metal evades Anderson localization and exhibits a ballistic conductance that scales linearly with the number of wires. We show that this behavior originates from hopping renormalization in the wire network. The zero-energy modes of the topological interconnects act as effective random dimers, giving rise to an energy-dependent localization length that diverges as $\sim 1/E^2$. Our work establishes that random networks provide a yet-unexplored platform to host intriguing phases of topological quantum matter.

Plasmons are collective oscillations of mobile electrons. Using terahertz spacetime metrology, we probe plasmon dynamics of mono- and bi-layer graphene. In both systems, the experimentally measured Drude weight systematically exceeds the prediction based on non-interacting electronic system. This enhancement is most pronounced at ultra-low carrier densities. We attribute the observed deviation to pseudospin dynamics of the Dirac fermions in multi-layer graphene, which leads to a breakdown of Galilean invariance. Our results establish that pseudospin structure of the single-particle electronic wave function can directly govern collective excitations, with implications that extend beyond graphene to a broad class of quantum materials.

We study the high-dimensional training dynamics of a shallow neural network with quadratic activation in a teacher-student setup. We focus on the extensive-width regime, where the teacher and student network widths scale proportionally with the input dimension, and the sample size grows quadratically. This scaling aims to describe overparameterized neural networks in which feature learning still plays a central role. In the high-dimensional limit, we derive a dynamical characterization of the gradient flow, in the spirit of dynamical mean-field theory (DMFT). Under l2-regularization, we analyze these equations at long times and characterize the performance and spectral properties of the resulting estimator. This result provides a quantitative understanding of the effect of overparameterization on learning and generalization, and reveals a double descent phenomenon in the presence of label noise, where generalization improves beyond interpolation. In the small regularization limit, we obtain an exact expression for the perfect recovery threshold as a function of the network widths, providing a precise characterization of how overparameterization influences recovery.

We address the spatially nonlocal dielectric functions of graphene at any frequency derived starting fromthe first principles of thermal quantum field theory using the formalism of the polarization tensor. After a brief review of this formalism, the longitudinal and transverse dielectric functions are considered at any relationship between the frequency and the wave vector. The analytic properties of their real and imaginary parts are investigated at low and high frequencies. Emphasis is given to the double pole at zero frequency which arises in the transverse dielectric function. The role of this unusual property for solving the problem of disagreement between experiment and theory in the Casimir effect is discussed. We guess that a more complete dielectric response of ordinary metals should also be spatially nonlocal and its transverse part may possess the double pole in the region of evanescent waves.

We analyze quantum systems under a broad class of protocols in which the temperature and a Hamiltonian control parameter are ramped simultaneously and, in general, in a nonlinear fashion toward a quantum critical point. Using an open-system version of a Kitaev quantum wire as an example, we show that, unlike finite-temperature protocols at fixed temperature, these protocols allow us to probe, in an out-of-equilibrium situation and at finite temperature, the universality class (characterized by the critical exponents $ν$ and $z$) of an equilibrium quantum phase transition at zero temperature. Key to this is the identification of ramps in which both coherent and incoherent parts of the open-system dynamics affect the excitation density in a non-negligible way. We also identify the specific ramps for which subleading corrections to the asymptotic scaling laws are suppressed, which serves as a guide to dynamically probing quantum critical exponents in experimentally realistic finite-temperature situations.

By extending the Sanchez-Lacombe lattice-fluid model for mixtures to the case of polymer blends containing solid fillers, we calculate the excess thermodynamic quantities arising from the presence of fillers. These results are then used to derive the spinodal stability condition of a filled polymer blend. In the low-compressibility limit, this condition reduces to a remarkably simple analytical expression that is derived self-consistently within the present framework. Comparison between the exact and approximate spinodal curves shows excellent agreement, with deviations in the spinodal temperature of less than 4 K, thereby validating the proposed approximation. The obtained analytical approximation enables a straightforward evaluation of the spinodal temperature without the extensive numerical calculations required to determine the exact spinodal condition. Both the exact and approximate spinodal conditions yield good quantitative agreement with experimental data for filled and unfilled blends.

We study finite-rank normal deformations of rotationally invariant non-Hermitian random matrices. Extending the classical Baik-Ben Arous-Péché (BBP) framework, we characterize the emergence and fluctuations of outlier eigenvalues in models of the form $\mathbf{A} + \mathbf{T}$, where $\mathbf{A}$ is a large rotationally invariant non-Hermitian random matrix and $\mathbf{T}$ is a finite-rank normal perturbation. We also describe the corresponding eigenvector behavior. Our results provide a unified framework encompassing both Hermitian and non-Hermitian settings, thereby generalizing several known cases.

Advanced manufacturing with new bio-derived materials can be achieved faster and more economically with first-principle-based artificial intelligence and machine learning (AI/ML)-derived models and process optimization. Not only is this motivated by increased industry profitability, but it can also be optimized to reduce waste generation, energy consumption, and gas emissions through additive manufacturing (AM) and AI/ML-directed self-driving laboratory (SDL) process optimization. From this perspective, the benefits of using 3D printing technology to manufacture durable, sustainable materials will enable high-value reuse and promote a better circular economy. Using AI/ML workflows at different levels, it is possible to optimize the synthesis and adaptation of new bio-derived materials with self-correcting 3D printing methods, and in-situ characterization. Working with training data and hypotheses derived from Large Language Models (LLMs) and algorithms, including ML-optimized simulation, it is possible to demonstrate more field convergence. The combination of SDL and AI/ML Workflows can be the norm for improved use of biobased and renewable materials towards advanced manufacturing. This should result in faster and better structure, composition, processing, and properties (SCPP) correlation. More agentic AI tasks, as well as supervised or unsupervised learning, can be incorporated to improve optimization protocols continuously. Deep Learning (DL), Reinforcement Learning (RL), and Deep Reinforcement Learning (DRL) with Deep Neural Networks (DNNs) can be applied to more generative AI directions in both AM and SDL, with bio-based materials.

Springs can be found in many applications and biological systems, and when these are soft, they easily deform. At small scales, capillarity can induce a force leading to spring deformations when the elastocapillary number is small. We demonstrate through experiments the non-trivial equilibrium shape liquid droplets adopt in these soft springs, which form an annulus, Eruciform, and spherical shapes. When these droplets are set in motion, they display different flow regimes with significant dissipation generated by the internal rotational flow. The static and dynamics of droplets in such a capillary slinky is also used to demonstrate how surface tension can actuate springs by stretching/compression, while providing a way for active flow control in soft springs.

Disordered spatial networks are model systems that describe structures and interactions across multiple length scales. Scattering and interference of waves in these networks can give rise to structural phase transitions, localization, diffusion, and band gaps. The study of these complex phenomena requires efficient numerical methods to computer-generate disordered networks with targeted structural properties. In the established Wooten-Weaire-Winer algorithm, a series of bond switch moves introduces disorder into an initial network. Conventional strain energies that govern this evolution are limited to 3D networks with coordination numbers of no more than four. We extend the algorithm to arbitrary coordination number statistics by introducing bond repulsion in the Keating strain energy. We tune the degree and type of disorder introduced into initially crystalline networks by varying the bond-bending force constant in the strain energy and the temperature profile. The effects of these variables are analyzed using a list of order metrics that capture both direct and reciprocal space. A feedforward neural network is trained to predict the structural characteristics from the algorithm inputs, enabling targeted network generation. As a case study, we statistically reproduce four disordered biophotonic networks exhibiting structural color. This work presents a versatile method for generating disordered networks with tailored structural properties. It will enable new insights into structure-property relations, such as photonic band gaps in disordered networks.

Quantum bianisotropy and chirality are fundamental concepts in light matter interaction that describe how materials with broken symmetries respond to electromagnetic fields at the level of macroscopic quantum electrodynamics. In quantum bianisotropy, magnetoelectric (ME) energy plays a critical role in mediating and enhancing light matter interactions. This concept is essential for bridging the gap between classical electromagnetics (where bianisotropy often involves field nonlocality) and quantum mechanics in metamaterials. The precise manipulation of a quantum emitter's properties at a subwavelength scale is due to near fields, which effectively function as a tunable environment. We show that the ME near field, interpreted as a structure combining the effect of bianisotropy (chirality) with a quantum atmosphere, is a nonMaxwellian field with spacetime symmetry breaking. Quantum ME fields arise from the dynamic modulation and topological coupling of magnetization and electric polarization within ME meta atoms, specific subwavelength structural elements with magnetic and dielectric subsystems in magnetic insulators.

We investigate the open system dynamics of a micromaser quantum battery operating in the ultrastrong coupling (USC) regime under environmental dissipation. The battery consists of a single-mode electromagnetic cavity sequentially interacting, via the Rabi Hamiltonian, with a stream of qubits acting as chargers. Dissipative effects arise from the weak coupling of the qubit-cavity system to a thermal bath. Non-negligible in the USC regime, the counter-rotating terms substantially improve the charging speed, but also lead, in the absence of dissipation, to unbounded energy growth and highly mixed cavity states. Dissipation during each qubit-cavity interaction mitigates these detrimental effects, yielding steady-state of finite energy and ergotropy. Optimal control on qubit preparation and interaction times enhances battery's performance in: (i) Maximizing the stored ergotropy trhough an optimized charging protocol; (ii) Stabilizing the stored ergotropy against dissipative losses through an optimized measurement-based passive-feedback strategy. Overall, our numerical results demonstrate that the interplay of ultrastrong light-matter coupling, controlled dissipation, and optimized control strategies enables micromaser quantum batteries to achieve both enhanced charging performance and long-term stability under realistic conditions.

The variable range hopping (VRH) model has been widely applied to describe electrical transport in disordered systems, providing theoretical formulas to fit temperature-dependent electric conductivity. These models rely on oversimplified assumptions that restrict their applicability and result in problematic fitting behaviors, yet their overusing situation is becoming increasingly serious. In this work we formulate an integral variable range hopping (IVRH) model, which replaces the empirical temperature power-law dependence in standard VRH theories with a physics-inspired integral formulation. The model builds upon the standard hopping probability $ω(R)$ w.r.t. hopping distance $R$ and incorporates the density of accessible electronic states through an effective volume function $V(R)$, which reflects the influence of system geometry. The IVRH formulation inherently reproduces both the Mott behavior at low temperatures and the Arrhenius behavior at high temperatures, respectively, and enables a smooth transition between the two regimes. We apply the IVRH model to two-dimensional, three-dimensional, and multi-layered systems. Monte Carlo simulations validate the model's predictions and yield consistent values for the fitting parameters, with substantially reduced variances compared to fitting using the standard VRH model. Furthermore, the improved robustness of IVRH also extends to the transport measurements in monolayer MoS$_2$ system and monolayer WS$_2$ system, enabling more physically meaningful interpretation.IVRH model offers a more stable and physically sound framework for interpreting hopping transport in low-dimensional amorphous materials, providing deeper insights into the universal geometric scaling factors that govern charge transport in disordered systems.

Quantum chaotic systems exhibit certain universal statistical properties that closely resemble predictions from random matrix theory (RMT). With respect to observables, it has recently been conjectured that, when truncated to a sufficiently narrow energy window, their statistical properties can be described by an unitarily invariant ensemble, and testable criteria have been introduced, which are based on the scaling behavior of free cumulants. In this paper, we investigate the conjecture numerically in translationally invariant Heisenberg spin chains with spin quantum number $S =\frac{1}{2},1,\frac{3}{2}$. Combining a quantum-typicality-based numerical method with the exploitation of the system's symmetries, we study the spin current operator and find clear evidence of consistency with the proposed criteria in chaotic cases. Our findings further support the conjecture of the existence of RMT universality as manifest in the observable properties in quantum chaotic systems.

We investigate the coherence properties of parity-protected $\cos(2\varphi)$ qubits based on interferences between two Josephson elements in a superconducting loop. We show that qubit implementations of a $\cos(2\varphi)$ potential using a single loop, such as those employing semiconducting junctions, rhombus circuits, flowermon and KITE structures, can be described by the same Hamiltonian as two multi-harmonic Josephson junctions in a SQUID geometry. We find that, despite the parity protection arising from the suppression of single Cooper pair tunneling, there exists a fundamental trade-off between charge and flux noise dephasing channels. Using numerical simulations, we examine how relaxation and dephasing rates depend on external flux and circuit parameters, and we identify the best compromise for maximum coherence. With currently existing circuit parameters, the qubit lifetime $T_1$ can exceed milliseconds while the dephasing time $T_\varphi$ remains limited to only a few microseconds due to either flux or charge noise. Our findings establish practical limits on the coherence of this class of qubits and raise questions about the long-term potential of this approach.

In this work, 3-atom clusters, Ru3 and Pt3, were deposited onto radio frequency RF-sputter deposited TiO2, treated with Ar+ ion sputtering. Ru3 was deposited by both solution submersion and chemical vapor deposition of Ru3(CO)12, while Pt3 was deposited under ultra-high vacuum using a laser vaporisation cluster source. The valence electronic density of states (DOS) of the deposited clusters were analysed after heat treatment using ultraviolet photoelectron spectroscopy (UPS) and metastable impact electron spectroscopy (MIES), where UPS measures the top several layers while MIES measures only the top atomic layer. XPS was used to determine the cluster surface coverages. The DOS were found to be very similar between Ru3 deposited by solution submersion and chemical vapor deposition. MIES results for Ru3 had contributions from titania O 2p sites due to encapsulation by a reduced titania overlayer. For Pt3 clusters the UPS and MIES results provided evidence that Pt was present on the topmost layer, and encapsulation did not occur. The proposed reason for the encapsulation of Ru3 but not of Pt3 is the higher surface energy of Ru over Pt. It is concluded that Pt clusters deposited onto TiO2 can modify the outermost layer by adding discrete energy levels on the surface, whereas the Ru clusters being encapsulated just below the surface generate a broad distribution of energy states close to the Fermi level. The outcome of this work is that Pt3-cluster-modified surfaces could be used as catalysts for reactions where the Pt3 energy levels are suitable for the respective reaction. The implication of the DOS found for photocatalytic water splitting are discussed.

Bright electroluminescence (EL) from dopant-free ambipolar lateral p-n junctions in GaAs/AlGaAs single heterointerface (SH) heterostructures is used to probe neutral free excitons arising from two-dimensional electron and hole gases (2DEGs and 2DHGs). The EL spectra reveal both the heavy-hole neutral free exciton (X$^0$) and the high-energy free exciton of the H band (HE). A combination of transition energies, lifetimes, spatial emission profiles, and temperature dependences points to a predominantly two-dimensional character for these excitons at the SH. For X$^0$, the EL peak energies (1515.5-1515.7 meV) lie slightly above the corresponding bulk GaAs photoluminescence (PL) line at 1515.3 meV, while time-resolved measurements yield markedly shorter lifetimes for EL than for PL (337 ps vs. 1610 ps), consistent with recombination in a confined interfacial layer. The HE exciton exhibits a Stark blueshift under forward bias below threshold, and its energies and lifetimes (down to 575 ps) are tuned by the topgate voltage; above threshold, HE emission is quenched in favor of X$^0$. Finally, the tidal effect $-$ a form of pulsed EL generated by swapping the topgate voltage polarity in ambipolar field-effect transistors $-$ produces an X$^0$ line at the same energy as in the lateral p-n junction and reproduces the characteristic nonmonotonic frequency dependence of the brightness previously observed in quantum-well heterostructures, again indicating a 2D-like origin. Taken together, these results show electrically generated and controllable 2D-like excitons (HE and X$^0$), thereby bridging 2D exciton physics and 2DEG/2DHG platforms in dopant-free GaAs/AlGaAs SH devices.

Quasiperiodic systems host exotic transport regimes that are distinct from those found in periodic or disordered lattices. In this work, we study quantum transport in the Aubry-André-Harper lattice in a two-terminal setup coupled to zero-temperature reservoirs, where the conductance is evaluated via the nonequilibrium Green's function method. In the extended phase, we uncover a universal subdiffusive transport when the bath chemical potential aligns with the band edges. Specifically, the typical conductance displays a scaling of $\mathcal{G}_{\text{typ}}\sim L^{-2}$ with system size $L$. We attribute this behavior to the emergence of an exceptional point (Jordan normal form) in the transfer matrix in the thermodynamic limit. In the localized phase, the conductance shows exponential decay governed by the Lyapunov exponent. Intriguingly, in the critical phase, we identify pronounced log-periodic oscillations of the conductance as a function of system size, arising from the discrete scale invariance inherent to the singular-continuous spectrum. We further extend our analysis to the generalized Aubry-André-Harper model and provide numerical evidence suggesting that the exact mobility edge resides within a finite spectral gap. This results in a counter-intuitive exponential suppression of conductance precisely at the mobility edge. Our work highlights the distinct transport behaviors in quasiperiodic systems and elucidates how they are rigorously dictated by the underlying local spectral structure.

We present a model of active particles interacting through a dynamic, heterogeneous environment, leading to emergent collective behaviors without direct agent-to-agent communication. Expanding the resource-dependent framework introduced in Briozzo et al., 2025, arXiv:2512.08762, agents perform a persistent random walk combined with chemotaxis, directing toward nutrient-rich patches, whose resources are generated by logistic regrowth. We identify distinct phases of collective organization, ranging from disordered gas-like states to polar traveling waves and nematic independent clusters, depending on the interplay between chemotactic sensitivity and angular noise. The system exhibits spontaneous symmetry breaking and density waves driven purely by the coupling between population dynamics (birth-death processes) and environmental feedback. Our results bridge active matter physics and movement ecology, demonstrating that complex spatiotemporal patterns can arise without direct interaction between agents, but solely from the maximization of resource intake in a reactive environment.

The presence of correlated noise, arising from a mixture of independent fluctuations and a common noisy input shared across the neural population, is a ubiquitous feature of neural circuits, yet its impact on collective network dynamics remains poorly understood. We analyze a network of quadratic integrate-and-fire neurons driven by Gaussian noise with a tunable degree of correlation. Using the cumulant expansion method, we derive a reduced set of effective mean-field equations that accurately describe the evolution of the population's mean firing rate and membrane potential. Our analysis reveals a counterintuitive phenomenon: increasing the noise correlation strength suppresses the mean network activity, an effect we term correlated-noise-inhibited spiking. Furthermore, within a specific parameter regime, the network exhibits metastability, manifesting itself as spontaneous, noise-driven transitions between distinct high- and low-activity states. These results provide a theoretical framework for reducing the dynamics of complex stochastic networks and demonstrate how correlated noise can fundamentally regulate macroscopic neural activity, with implications for understanding state transitions in biological systems.

We report microwave spectroscopy of Josephson junctions chains made from an epitaxial Al/InAs heterostructure. The chains exhibit superinductance, with characteristic wave impedance exceeding $R_{Q} = \hbar/(2e)^{2}$. The planar nature of the junctions results in a large plasma frequency, with no measurable deviations from ideal dispersion up to $12~\mathrm{GHz}$. Internal quality factors decrease sharply with frequency, which we describe with a simple loss model. The possibility of a loss mechanism intrinsic to the superconductor-semiconductor junction is considered.

Memory effects are essential for dynamics of condensed materials and are responsible for non-exponential relaxation of correlation functions of dynamic variables through the memory function. Memory functions of dipole rotations for polar liquids have never been calculated. We present here calculations of memory functions for single-dipole rotations and for the overall dipole moment of the sample for SPC/E water. The memory functions for single-particle and collective dipole dynamics turn out to be nearly identical. This result validates theories of dielectric spectroscopy in terms of single-particle time correlation functions and the connection between the collective and single-particle relaxation times through the Kirkwood factor. The dielectric function in this formalism contains no new dynamic information that does not exist in the single-dipole correlation function. A short memory time, $\lesssim 1$ fs, justifies the use of rotational diffusion model to describe dynamics of a single molecular dipole moment in bulk water.

Weyl degeneracies in spectra of magnetoplasma waves enable nonreciprocal energy flow and topologically protected modes, yet conventional materials require impractical magnetic fields to operate. Developing an effective Hamiltonian framework for magnetic Weyl semimetals, we show that these systems overcome the limit, hosting Weyl magnetoplasma physics at zero field due to their giant intrinsic anomalous Hall response. The resulting topology supports nonreciprocal modes localized at magnetic domain walls, including a pair of topological "Fermi-arc-like modes and additional bound states. These effects are fully developed across a broad THz window, and we propose feasible experimental routes for their detection.

Recent studies of moiré physics have unveiled a wealth of opportunities for significantly advancing the field of quantum phase transitions. However, properties of reentrant phase transitions driven by moiré strength are poorly understood. Here, we investigate the reentrant sequence of phase transitions and the invariant of universality class in moiré-modulated extended Su-Schrieffer-Heeger (SSH) model. For the simplified case with intercell hopping $w=0$, we analytically derive renormalization relations of Hamiltonian parameters to explain the reentrant phenomenon. For the general case, numerical phase boundaries are calculated in the thermodynamic limit. The bulk boundary correspondence between zero-energy edge modes and entanglement spectrum is revealed from the degeneracy of both quantities. We also address the correspondence between the central charge obtained from entanglement entropy and the change in winding number during the phase transition. Our results shed light on the understanding of universal characteristics and bulk-boundary correspondence for moiré induced reentrant phase transitions in 1D condensed-matter systems.

We study stochastic transport of interacting particles on a disordered network described by the random comb geometry. The model is defined on a one-dimensional backbone from which branches of random lengths emanate, providing a minimal model of percolation networks beyond the critical percolation probability. The dynamics obeys local detailed balance with respect to a Bose-Hubbard Hamiltonian containing both an external bias and on-site repulsion. This choice yields an analytically tractable steady state through a mapping to the zero-range-process. We compute the backbone current, branch density profiles, and macroscopic drift velocity, and analyze how bias and interactions compete to shape transport. The backbone current increases monotonically with density, while the drift velocity displays a non-monotonic dependence on the external field, remaining finite for any nonzero bias, in contrast to the vanishing drift velocity of noninteracting particles beyond a threshold bias. Density profiles along branches exhibit stepwise plateaus governed by the ratio of interaction to bias energy. These results highlight how repulsive interactions suppress trapping and restore transport in disordered geometries, bridging earlier studies of field induced drift in random networks with the physics of disordered Bose-Hubbard systems.

We determine the complete structure of the symmetry algebras associated with the N-body Calogero-Moser system and its maximally superintegrable discretization. We prove that the discretization naturally leads to a nontrivial deformation of the continuous symmetry algebra, with the discretization parameter playing the rôle of a deformation parameter. This phenomenon illustrates how discrete superintegrable systems can be viewed as natural sources of deformed polynomial algebraic structures. As a byproduct of these results, we also reveal a connection between the above-mentioned symmetry algebras and the Bell polynomials, as a consequence of the trace properties.

Urban crises - floods, pandemics, economic shocks, and conflicts - function as accelerators of urban change, exposing structural vulnerabilities while creating windows for reinvention. Building on a prior theoretical contribution that identified fifteen principles of urban antifragility, this paper tests and operationalizes the framework through an empirical assessment of 26 cities selected for their post-crisis adaptation trajectories. Using a tailored diagnostic methodology, we benchmark cities' Stress Response Strategies (SRS) and then evaluate Urban Development Trajectories (UDT) across four weighted dimensions, positioning each case along a fragility-robustness-resilience-antifragility continuum and applying a balanced-threshold rule to confirm antifragile status. Results show that "resilience enhanced by innovation and technology" is the most effective response typology (86.9/100), and that six cities meet the antifragile trajectory criteria. By mapping best practices to activated principles and analysing co-activations, the study identifies a robust "hard core" of principles - Sustainable Resilience (O), Strategic Diversity (F), Proactive Innovation (I), and Active Prevention (N) - supplemented by operational enablers (e.g., anticipation, mobilization, shock absorption). The paper concludes by proposing an evidence-based, SDG-aligned operational model that links high-impact principle pairings to measurable indicators, offering a practical roadmap for cities seeking to convert crises into sustained transformation. Keywords: Post-crisis strategies, Urban antifragility, Sustainable cities and communities, Disaster resilience and urban regeneration, Risk governance and Black Swan adaptation.

Higher-order interactions can dramatically reshape collective dynamics, yet how their microscopic organization controls macroscopic critical behavior remains unclear. Here we develop a new theory to study contagion dynamics on hypergraphs and show that nested hyperedges not only facilitate the onset of spreading, but also suppress backward bifurcations, thereby inhibiting explosive behavior. By disentangling contagion pathways, we find that overlap redirects transmission from external links to internal, group-embedded routes -- boosting early activation but making dyadic and triadic channels increasingly redundant. This loss of structural independence quenches the nonlinear amplification required for bistability, progressively smoothing the transition as hyperedges become nested. We observe the same phenomenology in Kuramoto dynamics, pointing to a broadly universal mechanism by which nested higher-order structure governs critical transitions in complex systems.

Gold nanoclusters possess multiple competing structural motifs with small energy differences, enabling structural coexistence and interconversion. Using a high-accuracy machine learned potential trained on some 20'000 density functional theory reference data points, we investigate transformation pathways connecting both high-symmetry and amorphous cuboctahedra, Ino's decahedra and icosahedra for Au55, Au147, Au309 and Au561 nanoclusters. Our saddle point searches reveal that high-symmetry transformations from cuboctahedra and Ino's decahedra to icosahedra proceed through a single barrier and represent soft-mode-driven jitterbug-type and slip-dislocation motions. In addition, we identify lower-barrier asymmetric transformation pathways that drive the system into disordered, Jahn-Teller-stabilized amorphous icosahedra. Minima Hopping sampling further uncovers, in this context, many such low-symmetry minima. Some of the newly identified global minima for Au309 and Au561 have energies that are up to 2.8 eV lower than the previously reported global minima. Hence, both the shapes and the transformation pathways studied in previous investigations are not the physically relevant ones. In contrast to the previously studied pathways, our transformation pathways give reasonable transformation times that are in rough agreement with experiments.

The accurate identification of historical restoration traces and material degradation is essential for the scientific preservation of ancient bronzes. In this study, the prestigious FUHAO bronze artifact (late Shang period, 13th-11th century BCE) was non-destructively examined using pulsed thermal imaging (PT). By combining single- and double-layer heat conduction models with Thermal Tomography (TT), this approach allowed for precise spatial localization of repair crevices, patches, and filler materials, while also distinguishing restorative interventions from the original bronze substrate. The artifact was revealed to have been assembled from multiple fragments, exhibiting uneven surface corrosion and clear evidence of prior conservation. The results not only provide direct insights for conservation strategy and historical interpretation but also demonstrate the capability of pulsed thermal imaging as an effective diagnostic tool for the integrated surface and subsurface assessment of cultural heritage objects.

Despite the significant role of turbomachinery in fluid-based energy transfer, precise simulation of rotating solid objects with complex geometry is a challenging task. In the present study, the volume penalization method (VPM) is combined with multiple reference frame (MRF) and sliding mesh (SLM), respectively, so as to develop immersed-boundary approaches for simulating flows around a rotating solid. The level-set function is adopted to represent arbitrary geometries embedded in Cartesian grids. The VPM body-forcing terms in the momentum equation are proposed for MRF and SLM, respectively, so as to build unified governing equations for both fluid and solid regions. The flows around a rotating cuboid under various rotating speeds are simulated by the present schemes, namely, VPM with MRF, and VPM with SLM, and compared to corresponding simulations by the body-fitted method (BFM). The results suggest the relative deviations of predicted pressure drop and torque between the present VPM and BFM are around 5%, demonstrating the validity of the present VPM.

Innovation emerges from complex collaboration patterns - among inventors, firms, or institutions. However, not much is known about the overall mesoscopic structure around which inventive activity self-organizes. Here, we tackle this problem by employing patent data to analyze both individual (\emph{co-inventorship}) and organization (\emph{co-ownership}) networks in three strategic domains (\emph{artificial intelligence}, \emph{biotechnology} and \emph{semiconductors}). We characterize the mesoscale structure (in terms of clusters) of each domain by comparing two alternative methods: a standard baseline - modularity maximization - and one based on the minimization of the Bayesian Information Criterion, within the Stochastic Block Model and its degree-corrected variant. We find that, across sectors, inventor networks are denser and more clustered than organization ones - consistent with the presence of small recurrent teams embedded into broader institutional hierarchies - whereas organization networks have neater hierarchical role-based structures, with few bridging firms coordinating the most peripheral ones. We also find that the discovered meso-structures are connected to innovation output. In particular, Lorenz curves of forward citations show a pervasive inequality in technological influence: across sectors and methods, both inventor (especially) and organization networks consistently show high levels of concentration of citations in a few of the discovered clusters. Our results demonstrate that the baseline modularity-based method may not be capable of fully capturing the way collaborations drive the spreading of inventive impact across technological domains. This is due to the presence of local hierarchies that call for more refined tools based on Bayesian inference.

We review and study the correspondence between discrete lattice/chain models of interacting particles and their continuous counterparts represented by partial differential equations. We study the correspondence problem for nearest neighbour interaction lattice models as well as for multiple-neighbour interaction lattice models, and we gradually proceed from infinite lattices to periodic lattices and finally to finite lattices with fixed ends/zero Dirichlet boundary conditions. The whole study is framed as systematic specialisation of Fourier analysis tools from the continuous to the discrete setting and vice versa, and the correspondence between the discrete and continuous models is examined primarily with regard to the dispersion relation.

Quantum algorithms are gaining attention in Computational Fluid Dynamics (CFD) for their favorable scaling, as encoding physical fields into quantum probability amplitudes enables representation of two to the power of n spatial points with only n qubits. A key challenge in Quantum CFD is the efficient readout of simulation results, a topic that has received limited attention in literature. This work presents methods to extract statistical properties of spatial velocity fields, such as central moments and structure functions, directly from parameterized ansatz circuits, avoiding full quantum state tomography. As a proof of concept, we implement our approach for 1D velocity fields, encoding 16 spatial points with 4 qubits, and analyze both a sine wave signal and four snapshots from Burgers' equation evolution. Using Qedma's error mitigation software QESEM, we demonstrate that such computations achieve high accuracy on current quantum devices, specifically IBMQ's Heron2 system ibm_fez.

Conjugate scalar transport with interfacial jump conditions on complex interfacial geometries is common in thermal and chemical processes, while its accurate and efficient simulations are still quite challenging. In the present study, a novel treatment of a two-phase interface in the volume penalization method, a kind of immersed boundary method, for solving conjugate scalar transport with general interfacial boundary conditions is developed. We first propose an interfacial treatment for solving an advection-diffusion equation with a Neumann boundary condition, and then extend it to general conjugate scalar transport with both interfacial flux and scalar jumps. A one-dimensional diffusion problem is solved to verify the present scheme and demonstrate the advantage of the present scheme in improving accuracy and unifying the governing equations in the two phases with an additional source term representing the local jump condition of the interfacial scalar flux. Then, the present scheme is further applied to fluid-solid coupled scalar diffusion and advection-diffusion problems with the scalar and its flux jumps across the interface. The simulation results of the present scheme generally show good agreement with reference results obtained by body-fitted mesh simulations with average relative deviations less than 3.0%.

Superconductivity in sulfur superhydride H$_{3}$S under extreme pressures has been explained theoretically, but it requires a peaked concentration of the electronic density of states (DOS), which has been found in first-principles calculations. The mechanism of this peak formation, though vital for its high transition temperature, has however remained obscure. We address this problem through detailed analysis of the first-principles electronic wave functions. The valence wave functions are shown to be significantly plane-wave-like. From the Fourier-mode analysis of the self-consistent potential and atomic pseudopotentials, we extract the nearly uniform models that accurately reproduce the first-principles band structure with very few parameters. The DOS peak is shown to be the consequence of the hybridization of specific plane waves. Adjacency of Jones' large zone to the plane-wave spherical Fermi surface is posited to be the root cause of the multiple plane-wave hybridization, the DOS peak formation and its proximity to the Fermi level. The present theory resolves the minimal modeling problem of electronic states in H$_{3}$S, as well as establishes a mechanism that may help to boost the transition temperatures in pressure induced superconductors.

Introduction Sporadic Creutzfeldt Jakob disease sCJD is a rapidly progressive neurodegenerative disease without effective treatment that usually results in death within one year. The recently applied methods have improved the accuracy of the disease diagnosis and the specific radiological findings provide the necessary information for differential diagnosis. Research question The research is aimed to provide a different perspective on the development of CJD and associated literature review. Materials and methods The study presents a case who presented cognitive deficits, gait instability, and urinary and fecal incontinence suffered from traumatic brain injury eight months ago before admission with cerebral ventricle dilation on CT images. Furthermore, studies describe relevant cases are also included. Results The patients symptoms got deteriorated. Further examinations, including 14-3-3 and tau proteins in the cerebrospinal fluid CSF, MRI, and EEG, confirmed the patients diagnosis of sCJD. He returned to the local hospital for the conservative treatment without effective medical intervention. Conclusion This case illustrates the diagnostic process of CJD and underscores the importance of distinguishing rare disorders from common conditions to achieve a comprehensive understanding of the disease.

Acute stress alters cognitive performance, yet competing models make divergent predictions regarding the mechanisms, scope, and temporal dynamics of these effects. This large-scale randomized controlled trial tested predications from three influential stress-effect models using a broad cognitive task battery embedded within a psychopharmacological stress paradigm. Across 606 testing sessions, 303 healthy male participants completed both the Maastricht Acute Stress Test (MAST) and its non-stress control condition. To independently manipulate acute stress and stress hormone availability, participants were additionally randomized to receive atomoxetine (40 mg; to prolong norepinephrine availability), hydrocortisone (10 mg; to increase cortisol availability), or placebo. Cognitive performance was assessed over 80-minutes (post-stress) using tasks targeting visual perception (rapid serial visual presentation), response inhibition (stop-signal), and cognitive flexibility (dual and switch tasks). MAST exposure selectively impaired response inhibition, reflected in shorter stop-signal delays, lower probabilities of successful stopping and prolonged stop-signal reaction times, particularly during later testing phases (40-80 minutes post-stress). MAST exposure did not affect visual perception or task-switching performance but buffered time-related declines in processing efficiency at the expense of task prioritization in the dual task. Pharmacological manipulation of norepinephrine or cortisol availability was effective but did not moderate cognitive stress effects. Overall, this pattern of task-specific impairment alongside stabilized processing efficiency cannot be fully explained by any tested model, highlighting the need to refine existing models and adopt more integrative approaches to advance our mechanistic understanding of cognitive stress-effects in laboratory and real-world contexts.

Why do neurons encode information the way they do? Normative answers to this question model neural activity as the solution to an optimisation problem; for example, the celebrated efficient coding hypothesis frames neural activity as the optimal encoding of information under efficiency constraints. Successful normative theories have varied dramatically in complexity, from simple linear models (Atick & Redlich '90), to complex deep neural networks (Lindsay '21). What complex models gain in flexibility, they lose in tractability and often understandability. Here, we split the difference by constructing a set of tractable but flexible normative representational theories. Instead of optimising the neural activities directly, following Sengupta et al. '18, we optimise the representational similarity, a matrix formed from the dot products of each pair of neural responses. Using this, we show that a large family of interesting optimisation problems are convex. This family includes problems corresponding to linear and some non-linear neural networks, and problems from the literature not previously recognised as convex, such as modified versions of semi-nonnegative matrix factorisation or nonnegative sparse coding. We put these findings to work in three ways. First, we provide the first necessary and sufficient identifiability result for a form of semi-nonnegative matrix factorisation. Second, we show that if neural tunings are `different enough' then they are uniquely linked to the optimal representational similarity, partially justifying the use of single neuron tuning analysis in neuroscience. Finally, we use the tractable nonlinearity of some of our problems to explain why dense retinal codes, but not sparse cortical codes, optimally split the coding of a single variable into ON & OFF channels. In sum, we identify a space of convex problems, and use them to derive neural coding results.

We present a mathematical model that describes the synergistic mechanism of combined Vesicular Stomatitis Virus (VSV) and Vaccinia Virus (VV). The model captures the dynamic interplay between tumor cells, viral replication, and the interferon-mediated immune response, revealing a `ping-pong' synergy where VV-infected cells produce B18R protein that neutralizes interferon-$α$, thereby enhancing VSV replication within the tumor. Numerical simulations demonstrate that this combination achieves complete tumor clearance in approximately 50 days, representing an 11\% acceleration compared to VV monotherapy (56 days), while VSV alone fails to eradicate tumors. Through bifurcation analysis, we identify critical thresholds for viral burst size and B18R inhibition, while sensitivity analysis highlights infection rates and burst sizes as the most influential parameters for treatment efficacy. Temporal optimization reveals that therapeutic outcomes are maximized through immediate VSV administration followed by delayed VV injection within a 1-19 day window, offering a strategic approach to overcome the timing and dosing challenges inherent in OVT.

Short-term synaptic plasticity (STP) is traditionally viewed as a purely presynaptic filter of incoming spike trains, independent of postsynaptic activity. Recent experiments, however, reveal an associative form of STP in which presynaptic release probability changes alongside long-term potentiation, implying a richer computational role for presynaptic plasticity. Here we develop a normative theory of associative STP using an information-theoretic framework. Extending Fisher-information-based learning to Tsodyks-Markram synapses, we derive analytic update rules for baseline synaptic strength and release probability that maximize encoded stimulus information under resource constraints. The learning rules separate into a conventional postsynaptic term tracking local firing and a distinct presynaptic term with a phase-advanced structure that selectively detects stimulus onset; critically, differences between plasticity of baseline strength and release probability arise within this presynaptic component. For stimulus variations slower than the EPSP time constant, onset sensitivity biases optimal connectivity toward anti-causal associations, strengthening synapses from neurons activated later to those activated earlier. In recurrent circuits, these rules yield ramp-like sustained representations and reverse replay after drive removal. Linear-response analysis further shows that STP confers frequency-dependent phase selectivity on presynaptic drive and that constraints on total release probability systematically tune temporal asymmetry. Together, our results provide a principled account of associative STP and identify presynaptic plasticity of release probability as a substrate for rapidly reconfigurable temporal coding.

Recruitment dynamics, or the distribution of the number of offspring among individuals, is central for understanding ecology and evolution. Sweepstakes reproduction (heavy right-tailed offspring number distribution) is central for understanding the ecology and evolution of highly fecund natural populations. Sweepstakes reproduction can induce jumps in type frequencies and multiple mergers in gene genealogies of sampled gene copies. We take sweepstakes reproduction to be skewed offspring number distribution due to mechanisms not involving natural selection, such as in chance matching of broadcast spawning with favourable environmental conditions. Here, we consider population genetic models of sweepstakes reproduction in a diploid panmictic populations absent selfing and evolving in a random environment. Our main results are {\it (i)} continuous-time Beta and Poisson-Dirichlet coalescents, when combining the results the skewness parameter $α$ of the Beta-coalescent ranges from $0$ to $2$, and the Beta-coalescents may be incomplete due to an upper bound on the number of potential offspring produced by any pair of parents; {\it (ii)} in large populations time is measured in units proportional to either $N/\log N$ or $N$ generations (where $2N$ is the population size when constant); {\it (iii)} it follows that incorporating population size changes leads to time-changed coalescents with the time-change independent of $α$; {\it (iv)} using simulations we show that the ancestral process is not well approximated by the corresponding coalescent (as measured through certain functionals of the processes); {\it (v)} whenever the skewness of the offspring number distribution is increased the conditional (conditioned on the population ancestry) and the unconditional ancestral processes are not in good agreement.

Although most theories posit that natural behavior can be explained as maximizing some form of extrinsic reward, often called utility, some behaviors appear to be reward independent. For instance, spontaneous motor babbling in human newborns and curiosity in little kids and other animals seem to elude a simple explanation in terms of extrinsic reward maximization. Rooted in these observations, intrinsic motivation has emerged as a potentially major driver of behavior. However, only recently have several quantitative and foundational theories of intrinsic motivation been put forward. We first provide a general framework to understand behavior as being organized hierarchically: objective--intrinsic reward, or motivation--drives, goals and extrinsic reward. We next review the main formalizations of intrinsic motivation, including empowerment, the free energy principle, information-gain maximization, and the maximum occupancy principle. These theories produce complex behavior by promoting, in various ways, entropic action-state paths. The presence of a single intrinsic motivation objective breaks infinite regress, as drives and goals act only temporarily to serve the objective. Extrinsic rewards, such as sugar or protein, are just a means to achieve the objective. Bounded cognition and embodiment impose constraints and boundary conditions for the intrinsic motivation objective. By virtue of their capability to generate complex behavior in a task-agnostic manner, theories of intrinsic motivation promise to become successful generative models of open-ended, embodied behavior.

The classification of microscopy videos capturing complex cellular behaviors is crucial for understanding and quantifying the dynamics of biological processes over time. However, it remains a frontier in computer vision, requiring approaches that effectively model the shape and motion of objects without rigid boundaries, extract hierarchical spatiotemporal features from entire image sequences rather than static frames, and account for multiple objects within the field of view. To this end, we organized the Cell Behavior Video Classification Challenge (CBVCC), benchmarking 35 methods based on three approaches: classification of tracking-derived features, end-to-end deep learning architectures to directly learn spatiotemporal features from the entire video sequence without explicit cell tracking, or ensembling tracking-derived with image-derived features. We discuss the results achieved by the participants and compare the potential and limitations of each approach, serving as a basis to foster the development of computer vision methods for studying cellular dynamics.

Learning, inference, and emergence in biological and artificial systems are often studied within disparate theoretical frameworks, ranging from energy-based models to recurrent and attention-based architectures. Here we develop a unified dynamical field theory in which learning and inference are governed by a minimal stochastic dynamical equation admitting a Martin--Siggia--Rose--Janssen--de Dominicis formulation. Within this framework, inference corresponds to saddle-point trajectories of the associated action, while fluctuation-induced loop corrections render collective modes dynamically emergent and generate nontrivial dynamical time scales. A central result of this work is that cognitive function is controlled not by microscopic units or precise activity patterns, but by the collective organization of dynamical time scales. We introduce the \emph{time-scale density of states} (TDOS) as a compact diagnostic that characterizes the distribution of collective relaxation modes governing inference dynamics. Learning and homeostatic regulation are naturally interpreted as processes that reshape the TDOS, selectively generating slow collective modes that support stable inference, memory, and context-dependent computation despite stochasticity and structural irregularity. This framework unifies energy-based models, recurrent neural networks, transformer architectures, and biologically motivated homeostatic dynamics within a single physical description, and provides a principled route toward understanding cognition as an emergent dynamical phenomenon.

Atrial fibrillation (AF) is a common cardiac arrhythmia that significantly increases the risk of stroke and heart failure, necessitating reliable and generalizable detection methods from electrocardiogram (ECG) recordings. Although deep learning has advanced automated AF diagnosis, existing approaches often struggle to exploit complementary time-frequency information effectively, limiting both robustness under intra-dataset and generalization across diverse clinical datasets. To address these challenges, we propose a cross-modal deep learning framework comprising two key components: a Bidirectional Gating Module (BGM) and a Cross-modal Supervised Contrastive Learning (CSCL) strategy. The BGM facilitates dynamic, reciprocal refinement between time and frequency domain features, enhancing model robustness to signal variations within a dataset. Meanwhile, CSCL explicitly structures the joint embedding space by pulling together label-consistent samples and pushing apart different ones, thereby improving inter-class separability and enabling strong cross-dataset generalization. We evaluate our method through five-fold cross-validation on the AFDB and the CPSC2021 dataset, as well as bidirectional cross-dataset experiments (training on one and testing on the other). Results show consistent improvements over state-of-the-art methods across multiple metrics, demonstrating that our approach achieves both high intra-dataset robustness and excellent cross-dataset generalization. We further demonstrate that our method achieves high computational efficiency and anti-interference capability, making it suitable for edge deployment.

We examine the pertinent geometric characteristics of entanglement that arise from stationary Hamiltonian evolutions transitioning from separable to maximally entangled two-qubit quantum states. From a geometric perspective, each evolution is characterized by means of geodesic efficiency, speed efficiency, and curvature coefficient. Conversely, from the standpoint of entanglement, these evolutions are quantified using various metrics, such as concurrence, entanglement power, and entangling capability. Overall, our findings indicate that time-optimal evolution trajectories are marked by high geodesic efficiency, with no energy resource wastage, no curvature (i.e., zero bending), and an average path entanglement that is less than that observed in time-suboptimal evolutions. Additionally, when analyzing separable-to-maximally entangled evolutions between nonorthogonal states, time-optimal evolutions demonstrate a greater short-time degree of nonlocality compared to time-suboptimal evolutions between the same initial and final states. Interestingly, the reverse is generally true for separable-to-maximally entangled evolutions involving orthogonal states. Our investigation suggests that this phenomenon arises because suboptimal trajectories between orthogonal states are characterized by longer path lengths with smaller curvature, which are traversed with a higher energy resource wastage compared to suboptimal trajectories between nonorthogonal states. Consequently, a higher initial degree of nonlocality in the unitary time propagators appears to be essential for achieving the maximally entangled state from a separable state. Furthermore, when assessing optimal and suboptimal evolutions...

The Landau--Zener (LZ) model describes a two-level quantum system that undergoes an avoided crossing. In the adiabatic limit, the transition probability vanishes. An auxiliary control field $H_\text{CD}$ can be reverse-engineered so that the full Hamiltonian $H_0 + H_\text{CD}$ reproduces adiabaticity for all parameter values. Our aim is to construct a single control field $H_1$ that drives an ensemble of LZ-type Hamiltonians with a distribution of energy gaps. $H_1$ works best statistically, minimizing the average transition probability. We restrict our attention to a special class of $H_1$ controls, motivated by $H_\text{CD}$. We found a systematic trade-off between instantaneous adiabaticity and the final transition probability. Certain limiting cases with a linear sweep can be treated analytically; one of them being the LZ system with Dirac $δ(t)$ function. Comprehensive and systematic numerical simulations support and extend the analytic results.

It is known that the continuous-time variant of Grover's search algorithm is characterized by quantum search frameworks that are governed by stationary Hamiltonians, which result in search trajectories confined to the two-dimensional subspace of the complete Hilbert space formed by the source and target states. Specifically, the search approach is ineffective when the source and target states are orthogonal. In this paper, we employ normalization, orthogonality, and energy limitations to demonstrate that it is unfeasible to breach time-optimality between orthogonal states with constant Hamiltonians when the evolution is limited to the two-dimensional space spanned by the initial and final states. Deviations from time-optimality for unitary evolutions between orthogonal states can only occur with time-dependent Hamiltonian evolutions or, alternatively, with constant Hamiltonian evolutions in higher-dimensional subspaces of the entire Hilbert space. Ultimately, we employ our quantitative analysis to provide meaningful insights regarding the relationship between time-optimal evolutions and analog quantum search methods. We determine that the challenge of transitioning between orthogonal states with a constant Hamiltonian in a sub-optimal time is closely linked to the shortcomings of analog quantum search when the source and target states are orthogonal and not interconnected by the search Hamiltonian. In both scenarios, the fundamental cause of the failure lies in the existence of an inherent symmetry within the system.

The entanglement distance of evolutionary quantum states of a two-spin system with the XXZ model has been studied. The analysis has been conducted both analytically and using quantum computing. An analytical dependence of the entanglement distance on the values of the model coupling constants and the parameters of the initial states has been obtained. The speed of evolution of a two-spin system has been investigated. The analysis has been performed analytically and using quantum computing. An explicit dependence of the speed of evolution on the coupling constants and on the parameters of the initial state has been obtained. The results of quantum computations are in good agreement with the theoretical predictions.

We investigate late-time cosmology in the context of modified $f(Q,L_m)$ gravity, considering a non-linear model$ f(Q,L_m) = αQ + βL_m^n + λ$ where, $α$, $β$, $λ$, and $n$ are some free parameters. The modified Friedmann equations are derived for a barotropic cosmic fluid, and an analytical solution for the Hubble parameter $H(z)$ is obtained. Using the latest DESI DR2 BAO data, previous BAO compilations (P-BAO), and cosmic chronometer (CC) datasets, we constrain the model parameters through a Markov Chain Monte Carlo analysis. Our results show that the model successfully describes the observed late-time cosmic acceleration with slightly tighter constraints from the inclusion of DESI dataset. The present-day Hubble constant is determined as $H_0 \simeq 69.5\ \mathrm{km\ s^{-1}\ Mpc^{-1}}$, while the deceleration parameter confirms accelerated expansion with $q_0 \simeq -0.57$. The transition redshift, where the universe switches from deceleration to acceleration, occurs in the range $z_{\rm tr} \sim 0.56 - 0.77$. Similarly, a smooth and physically consistent transition from a matter-dominated decelerated period at high redshifts to an accelerated phase at late times is revealed by the evolution of $ω_{eff}(z)$. While statefinder diagnostic shows the model favours a Chaplygin gas like nature for DESI and DESI+CC, whereas the model favours as quintessence dominated evolution for P-BAO+CC in the late time regime. Conclusively, all these results along with the study of the energy conditions and stability analysis showcases the given $f(Q,L_m)$ model offers a viable alternative to GR-based cosmology

We develop quantitative surgery, which extends the classical constructions of Gromov--Lawson and Lawson--Michelsohn. As an application, we prove a conjecture of Gromov on the total mean curvature of fill-ins.

Quantum embedding methods, such as dynamical mean-field theory (DMFT), provide a powerful framework for investigating strongly correlated materials. A central computational bottleneck in DMFT is in solving the Anderson impurity model (AIM), whose exact solution is classically intractable for large bath sizes. In this work, we develop and benchmark a quantum-classical hybrid solver tailored for DMFT applications, using the variational quantum eigensolver (VQE) to prepare the ground state of the AIM with shallow quantum circuits. The solver uses a unified ansatz framework to prepare the particle and hole excitations of the ground-state from parameter-shifted circuits, enabling the reconstruction of the impurity Green's function through a continued-fraction expansion. We evaluate the performance of this approach across a few bath sizes and interaction strengths under noisy, shot-limited conditions. We compare three optimization routines (COBYLA, Adam, and L-BFGS-B) in terms of convergence and fidelity, assess the benefits of estimating a quantum-computed moment (QCM) correction to the variational energies, and benchmark the approach by comparing the reconstructed density of states (DOS) against that obtained using a classical pipeline. Our results demonstrate the feasibility of Green's function reconstruction on near-term devices and establish practical benchmarks for quantum impurity solvers embedded within self-consistent DMFT loops.

We demonstrate a Doppler thermometer based on direct optical frequency comb spectroscopy of an $^{85}$Rb vapor with a chirped electro-optic frequency comb (EOFC). The direct EOFC Doppler thermometer is accurate to within its approximately 1 K statistical uncertainty. We experimentally compare direct EOFC spectroscopy with conventional Doppler spectroscopy using a single-frequency, step-scanned laser probe. Our results show that direct EOFC spectroscopy mitigates transit-induced optical pumping distortion of the atomic lineshape, which is the dominant systematic temperature shift in alkali atom Doppler thermometry. Optical Bloch equation simulations of conventional and direct EOFC Doppler spectroscopy confirm that EOFC spectroscopy can use higher optical power to reduce statistical noise without optical pumping distortion. Our results indicate that EOFC Doppler thermometry is a promising approach to realizing a primary thermometer with size and measurement rate sufficient for applications including pharmaceutical manufacturing and nuclear waste monitoring.

The scalable and deterministic preparation of large Fock-number states represents a long-standing frontier in quantum science, with direct implications for quantum metrology, communication, and simulation. Despite significant progress in small-scale implementations, extending such state generation to large excitation numbers while maintaining high fidelity remains a formidable challenge. Here, we present a scalable protocol for generating large Fock states with fidelities exceeding 0.9 up to photon numbers on the order of 100, achieved using only native control operations and, when desired, further enhanced by an optional post-selection step. Our method employs a hybrid Genetic-Adam optimization framework that combines the global search efficiency of genetic algorithms with the adaptive convergence of Adam to optimize multi-pulse control sequences comprising Jaynes-Cummings interactions and displacement operations, both of which are native to leading experimental platforms. The resulting control protocols achieve high fidelities with shallow circuit depths and strong robustness against parameter variations. These results establish an efficient and scalable pathway toward high-fidelity non-classical state generation for precision metrology and fault-tolerant quantum technologies.

We study the computation of the $α$-Rényi capacity of a classical-quantum (c-q) channel for $α\in(0,1)$. We propose an exponentiated-gradient (mirror descent) iteration that generalizes the Blahut-Arimoto algorithm. Our analysis establishes relative smoothness with respect to the entropy geometry, guaranteeing a global sublinear convergence of the objective values. Furthermore, under a natural tangent-space nondegeneracy condition (and a mild spectral lower bound in one regime), we prove local linear (geometric) convergence in Kullback-Leibler divergence on a truncated probability simplex, with an explicit contraction factor once the local curvature constants are bounded.

We present an analytic framework demonstrating that a spinning black hole endowed with a net magnetic charge exhibits a dramatically amplified superradiant instability against charged scalar fields, enhanced by several orders of magnitude compared with the neutral Kerr case. The amplification arises from a monopole induced reduction of the centrifugal barrier. This shift deepens the gravitational bound-state potential well and produces a parametrically larger instability growth rate. This resulting rapid growth yields a macroscopic boson cloud that acts as a coherent source of near monochromatic continuous gravitational waves (GWs). We find an enhanced GW power. Monopole harmonic selection rules restrict the emission from the north (south) clouds corresponding to opposite helicities. Their superposition generates an (approximately) circularly polarized continuous GWs at a fixed sky location within even parity general relativity, distinct from the generic elliptical polarization of the Kerr case. In light of these new findings, we propose a potential smoking-gun search strategy for magnetic monopole and ultralight boson: the rapid post-merger follow-up GW signals from binary-black-hole merger remnants through ground-based and space-based GW experiments. In contrast to the Kerr case, where the signal turn-on can be delayed to decades-centuries, a magnetic remnant can form a cloud and emit a stronger, circularly polarized continuous GWs within weeks to months. Taking the magnetic supermassive remnants as an example, we demonstrate that the rapid follow-up GW signal in the mHz band appears just in few weeks after binary black hole mergers. Moreover, future polarization (ellipticity) measurements can distinguish the magnetic scenario from Kerr while providing a parity-even mechanism for circularly polarized GWs in general relativity.

The remnants of binary neutron star mergers are expected to be massive, rapidly rotating stars whose oscillations produce gravitational waves in the kilohertz band. The degree of differential rotation and the rotation profiles strongly influence their structure, stability and oscillation spectrum, and must therefore be taken into account when modeling their dynamics. We extend the pseudospectral code ROXAS (Relativistic Oscillations of non-aXisymmetric neutron stArS) to enable the dynamical evolution of oscillating, differentially rotating neutron stars. Using the updated code, we aim to study the star's oscillation frequencies. We extend the previous formalism, based on primitive variables and the conformal flatness approximation, to differential rotation. Within this framework, we run a series of axisymmetric and non-axisymmetric simulations of perturbed, differentially rotating neutron stars with different rotation rates, and extract their oscillation frequencies. Axisymmetric modes, as well as those under the Cowling approximation, show excellent agreement with published results. We show that the secondary fundamental mode in the Cowling approximation is an artifact that does not appear in dynamical spacetimes. In addition, we provide, for the first time, frequency values for non-axisymmetric modes in differentially rotating configurations evolved in conformal flatness. This extension broadens the range of physical scenarios that can be studied with ROXAS, and represents a step toward more realistic modeling of post-merger remnants and their gravitational-wave emission.

We show how our Universe can emerge from a symmetry breaking of a multicomponent $W_3$ algebra, where the components in addition form a Jordan algebra. We discuss how symmetry breaking related to the Jordan algebras $H_3(C)$ and $H_3(O)$ over the complex and octonion numbers can lead to an extended four-dimensional spacetime, where the expansion of the Universe is governed by a modified Friedmann equation. We finally discuss how this modified Friedmann equation might explain a number of puzzling cosmological observations.

Neutral atom quantum processors have emerged as a promising platform for scalable quantum information processing, offering high-fidelity operations and exceptional qubit scalability. A key challenge in realizing practical applications is efficiently extracting readout outcomes while maintaining high system throughput, i.e., the rate of quantum task executions. In this work, we develop a theoretical framework to quantify the trade-off between readout fidelity and atomic retention. Moreover, we introduce a metric of quantum circuit iteration rate (qCIR) and employ normalized quantum Fisher information to characterize system overall performance. Further, by carefully balancing fidelity and retention, we demonstrate a readout strategy for optimizing information acquisition efficiency. Considering the experimentally feasible parameters for 87Rb atoms, we demonstrate that qCIRs of 197.2Hz and 154.5Hz are achievable using single photon detectors and cameras, respectively. These results provide practical guidance for constructing scalable and high-throughput neutral atom quantum processors for applications in sensing, simulation, and near-term algorithm implementation.

Quantum Amplitude Amplification (QAA), the generalization of Grover's algorithm, is capable of yielding optimal solutions to combinatorial optimization problems with high probabilities. In this work we extend the conventional 2-dimensional representation of Grover's (orthogonal collective states) to oracles which encode cost functions such as QUBO, and show that linear cost functions are a special case whereby an exact formula exists for determining optimal oracle parameter settings. Using simulations of problem sizes up to 40 qubits we demonstrate QAA's algorithmic performance across all possible solutions, with an emphasis on the closeness in Grover-like performance for solutions near the global optimum. We conclude with experimental demonstrations of generalized QAA on both IBMQ (superconducting) and IonQ (trapped ion) qubits, showing that the observed probabilities of each basis state match our equations as a function of varying the free parameters in the oracle and diffusion operators.

In this article, we apply the Simpson-Visser (SV) regularization scheme to Anti-de Sitter (AdS) charged black holes and investigate the resulting spacetime geometry in detail, with emphasis on both geodesic structure and thermodynamic behavior. In particular, we analyze the motion of massless particle, focusing on key features such as the photon sphere, black hole shadow, photon trajectory and the dynamics of charged particles, including the characteristics of the circular and type of orbits. Furthermore, we compare the theoretical predictions of the charged SV-AdS black hole with recent observations reported by the Event Horizon telescope (EHT) for M87* and Sgr~A*. Beyond the geodesic analysis, we explore the thermodynamics of the regularized charged SV-AdS black hole by deriving essential quantities such as the Hawking temperature, Gibbs free energy, and specific heat capacity. Through a systematic examination of these thermodynamic variables, we demonstrate how the regularization parameter inherent in the SV regularization influences particle dynamics, stability conditions, and the overall thermal properties of the modified black hole solution. This comprehensive study highlights the interplay between regularization effects and the physical observables associated with charged AdS black holes.

Fast and high fidelity shuttling of spin qubits has been demonstrated in semiconductor quantum dot devices. Several architectures based on shuttling have been proposed; it has been suggested that singlet-triplet (dual-spin) qubits could be optimal for the highest shuttling fidelities. Here we present a fault-tolerant framework for quantum error correction based on such dual-spin qubits, establishing them as a natural realisation of erasure qubits within semiconductor architectures. We introduce a hardware-efficient leakage-detection protocol that automatically projects leaked qubits back onto the computational subspace, without the need for measurement feedback or increased classical control overheads. When combined with the XZZX surface code and leakage-aware decoding, we demonstrate a twofold increase in the error correction threshold and achieve orders-of-magnitude reductions in logical error rates. This establishes the singlet-triplet encoding as a practical route toward high-fidelity shuttling and erasure-based, fault-tolerant quantum computation in semiconductor devices.

We propose a generalization of the Filoche--Mayboroda localization landscape that extends the theory well beyond the static, elliptic and Hermitian settings while preserving its geometric interpretability. Using the positive operator $H^\dagger H$, we obtain a landscape that predicts localization across non-Hermitian, Floquet, and topological systems without computing eigenstates. Singular-value collapse reveals spectral instabilities and skin effects, the Sambe formulation captures coherent destruction of tunneling, and topological zero modes emerge directly from the landscape. Applications to Hatano--Nelson chains, driven two-level systems, and driven Aubry--André--Harper models confirm quantitative accuracy, establishing a unified predictor for localization in equilibrium and driven quantum matter.

We present a unified scheme for generating high-fidelity entangling gates in superconducting platforms by continuous dynamical decoupling (CDD) combined with variational minimal-energy optimal control. During the CDD stage, we suppress residual couplings, calibration drifting, and quasistatic noise, resulting in a stable effective Hamiltonian that preserves the designed ZZ interaction intended for producing tunable couplers. In this stable $\mathrm{SU}(4)$ manifold, we calculate smooth low-energy single-quibt control functions using a variational geodesic optimization process that directly minimizes gate infidelity. We illustrate the methodology by applying it to CZ, CX, and generic engangling gates, achieving virtually unit fidelity and robustness under restricted single-qubit action, with experimentally realistic control fields. These results establish CDD-enhanced variational geometric optimal control as a practical and noise-resilient scheme for designing superconducting entangling gates.

We introduce SpinPulse, an open-source python package for simulating spin qubit-based quantum computers at the pulse-level. SpinPulse models the specific physics of spin qubits, particularly through the inclusion of classical non-Markovian noise. This enables realistic simulations of native gates and quantum circuits, in order to support hardware development. In SpinPulse, a quantum circuit is first transpiled into the native gate set of our model and then converted to a pulse sequence. This pulse sequence is subsequently integrated numerically in the presence of a simulated noisy experimental environment. We showcase workflows including transpilation, pulse-level compilation, hardware benchmarking, quantum error mitigation, and large-scale simulations via integration with the tensor-network library quimb. We expect SpinPulse to be a valuable open-source tool for the quantum computing community, fostering efforts to devise high-fidelity quantum circuits and improved strategies for quantum error mitigation and correction.

The thermodynamic uncertainty relation (TUR) imposes a fundamental constraint between current fluctuations and entropy production, providing a refined formulation of the second law for micro- and nanoscale systems. Quantum violations of the classical TUR reveal genuinely quantum thermodynamic effects, which are essential for improving performance and enabling optimization in quantum technologies. In this work, we analyze the TUR in a class of paradigmatic quantum thermal-machine models whose operation is enabled by coherent coupling between two energy levels forming a virtual qubit. Steady-state coherences are confined to this virtual-qubit subspace, while in the absence of coherent coupling the system satisfies detailed balance with the thermal reservoirs and supports no steady-state heat currents. We show that the steady-state currents and entropy production can be fully reproduced by an effective classical Markov process, whereas current fluctuations acquire an additional purely quantum correction originating from coherence. As a result, the thermodynamic uncertainty naturally decomposes into a classical (diagonal) contribution and a coherent contribution. The latter becomes negative under resonant conditions and reaches its minimum at the coupling strength that maximizes steady-state coherence. We further identify the optimization conditions and the criteria for surpassing the classical TUR bound in the vicinity of the reversible limit.

With the growing number of gravitational-wave detections, particularly from binary black hole mergers, there is increasing anticipation that an astrophysical background, formed by an ensemble of faint, high-redshift events, will be observed in the near future by the ground-based detector network. This background is anticipated to exhibit non-Gaussian statistical properties. To develop a robust method for detecting such a non-Gaussian gravitational-wave background, we revisit optimal detection strategies based on the Gaussian-mixture likelihood model. In this work, we demonstrate that properly accounting for the detector antenna pattern is essential. Current approaches typically rely on the overlap reduction function averaged over the sky. Through simulations, we show that using such an averaged response introduces significant biases in parameter estimation. In addition, we propose a computationally feasible method that incorporates second-order corrections as an approximation of the full integral over the source distribution. Our results indicate that this approach effectively eliminates these biases. We also show that our method remains robust even when considering anisotropic backgrounds.

Classical and quantum physics represent two distinct theories; however, quantum physics is regarded as the more fundamental of the two. It is posited that classical mechanics should arise from quantum mechanics under certain limiting conditions. Nevertheless, this remains a challenging objective. In this work, we explore the potential for unifying the dynamics of classical and quantum physics. This discussion does not suggest that classical behavior emerges from quantum mechanics; rather, it demonstrates the exact equivalence between the dynamics of quantum observables and their classical counterparts. It is shown that the Heisenberg equations of motion can be cast in a form that is identical to Newton's equations of motion, with $\hbar$ being absent from the formulation. This implies that both quantum and classical dynamics are governed by the same equations, with the Heisenberg operators substituting the classical observables.

Experiments on Bose-Einstein condensates at finite temperature typically extract the system parameters, such as temperature, atom number, and condensed fraction from time-of-flight images taken after a free expansion time. This paper systematically examines the effect of repulsive interactions between the condensed and thermal atoms in partially condensed clouds on the expansion profile of the thermal cloud. An analytical expression for the expansion can be obtained only if the interactions between the Bose-Einstein condensate and thermal atoms are neglected, resulting in a Bose-enhanced distribution for the thermal component. Here, the deformation of the cloud due to interactions and the effects on estimated parameters are investigated by simulating the expansion using a ballistic approximation. By fitting the simulated expansion profiles with a Bose-enhanced distribution, the errors of using such a fit are estimated, and the results are explained phenomenologically. The simulation was also used as a fitting function for experimental data, showing better agreement of the extracted condensed fraction with the semi-ideal model than results from a Bose-enhanced fit.

The evolution of an isolated quantum system inevitably exhibits recurrence: the state returns to the vicinity of its initial condition after finite time. Despite its fundamental nature, a rigorous quantitative understanding of recurrence has been lacking. We establish upper bounds on the recurrence time, $t_{\mathrm{rec}} \lesssim t_{\mathrm{exit}}(ε)(1/ε)^d$, where $d$ is the Hilbert-space dimension, $ε$ the neighborhood size, and $t_{\mathrm{exit}}(ε)$ the escape time from this neighborhood. For pure states evolving under a Hamiltonian $H$, estimating $t_{\mathrm{exit}}$ is equivalent to an inverse quantum speed limit problem: finding upper bounds on the time a time-evolved state $ψ_t$ needs to depart from the $ε$-vicinity of the initial state $ψ_0$. We provide a partial solution, showing that under mild assumptions $t_{\mathrm{exit}}(ε) \approx ε/\sqrt{ Δ(H^2)}$, with $Δ(H^2)$ the Hamiltonian variance in $ψ_0$. We show that our upper bound on $t_{\mathrm{rec}}$ is generically saturated for random Hamiltonians. Finally, we analyze the impact of coherence of the initial state in the eigenbasis of $H$ on recurrence behavior.

Calculating bounds of properties of many-body quantum systems is of paramount importance, since they guide our understanding of emergent quantum phenomena and complement the insights obtained from estimation methods. Recent semidefinite programming approaches enable probabilistic bounds from finite-shot measurements of easily accessible, yet informationally incomplete, observables. Here we render these methods scalable in the number of qubits by instead utilizing moment-matrix relaxations. After introducing the general formalism, we show how the approach can be adapted with specific knowledge of the system, such as it being the ground state of a given Hamiltonian, possessing specific symmetries or being the steady state of a given Lindbladian. Our approach defines a scalable real-world certification scheme leveraging semidefinite programming relaxations and experimental estimations which, unavoidably, contain shot noise.

Pinsker's inequality sets a lower bound on the Umegaki divergence of two quantum states in terms of their trace distance. In this work, we formulate corresponding estimates for a variety of quantum and classical divergences including $f$-divergences like Hellinger and $χ^2$-divergences as well as Rényi divergences and special cases thereof like the Umegaki divergence, collision divergence, max divergence. We further provide a strategy on how to adapt these bounds to smoothed divergences.

High-Q bosonic memories are central to hardware-efficient quantum error correction, but their isolation makes fast, high-fidelity reset a persistent bottleneck. Existing approaches either rely on weak intermode cross-Kerr conversion or on measurement-based sequences with substantial latency. Here we demonstrate a hardware-efficient Rabi-Driven Reset (RDR) that implements continuous, measurement-free cooling of a superconducting cavity mode. A strong resonant Rabi drive on a transmon, together with sideband drives on the memory and readout modes detuned by the Rabi frequency, converts the dispersive interaction into an effective Jaynes-Cummings coupling between the qubit dressed states and each mode. This realizes a tunable dissipation channel from the memory to the cold readout bath. Crucially, the engineered coupling scales with the qubit-mode dispersive interaction and the drive amplitude, rather than with the intermode cross-Kerr, enabling fast cooling even in very weakly coupled architectures that deliberately suppress direct mode-mode coupling. We demonstrate RDR of a single photon with a decay time of $1.2 μs$, more than two orders of magnitude faster than the intrinsic lifetime. Furthermore, we reset about 30 thermal photons in about $80 μs$ to a steady-state average photon number of $\bar{n} = 0.045 \pm 0.025$.

Learning the Hamiltonian governing a quantum system is a central task in quantum metrology, sensing, and device characterization. Existing Heisenberg-limited Hamiltonian learning protocols either require multi-qubit operations that are prone to noise, or single-qubit operations whose frequency or strength increases with the desired precision. These two requirements limit the applicability of Hamiltonian learning on near-term quantum platforms. We present a protocol that learns a quantum Hamiltonian with the optimal Heisenberg-limited scaling using only single-qubit control in the form of static fields with strengths that are independent of the target precision. Our protocol is robust against the state preparation and measurement (SPAM) error. By overcoming these limitations, our protocol provides new tools for device characterization and quantum sensing. We demonstrate that our method achieves the Heisenberg-limited scaling through rigorous mathematical proof and numerical experiments. We also prove an information-theoretic lower bound showing that a non-vanishing static field strength is necessary for achieving the Heisenberg limit unless one employs an extensive number of discrete control operations.

We investigate the experimental prospects for testing a relativistic local quantum measurement inequality that quantifies the trade-off between vacuum insensitivity and responsiveness to excitations for finite-size detectors. Building on the Reeh--Schlieder approximation for coherent states, we derive an explicit and practically applicable bound for arbitrary coherent states. To connect with realistic photodetection scenarios, we model the detection region as a square prism operating over a finite time window and consider a normally incident single-mode coherent state. Numerical results exhibit the expected qualitative behavior: suppressing dark counts necessarily tightens the achievable click probability.

The definition of weak equivalence principle of quantum matter is an open problem at present. In order to reflect the probability of quantum system in the quantum version of weak equivalence principle, we proposed a quantum weak equivalence principle based on differential scattering cross section at tree level, that is, the differential scattering cross section does not depend on the mass and properties of the scattered particles when the target particles take the large mass limit. This version of the quantum equivalence principle we proposed will be broken by the spin properties of quantum matter. In the non-relativistic case, the difference of differential scattering cross sections of scattered particles with different spin properties scattered by target particles is mainly reflected in the order of $ \mathcal O (p _ {\mathrm{cm}} ^2) $. In the relativistic case , we studied the asymptotic behavior of differential scattering cross sections at small angles. When the target particles are scalar particles, the difference of light particles with different spin properties is mainly reflected in the $ \mathcal O (1/θ^2) $ order. When the target particles are Dirac particles, the difference of light particles with different spin properties is mainly reflected in the $ \mathcal O (1/θ^4) $ order. The polarization of differential scattering cross section when scattered particles are Dirac particles is investigated. The result of the degree of polarization depends on the polarization direction of the incident particles.

The manipulation of distinct degrees of freedom of photons plays a critical role in both classical and quantum information processing. While the principles of wave optics provide elegant and scalable control over classical light in spatial and temporal domains, engineering quantum states in Fock space has been largely restricted to few-photon regimes, hindered by the computational and experimental challenges of large Hilbert spaces. Here, we introduce ``Fock-space optics", establishing a conceptual framework of wave propagation in the quantum domain by treating photon number as a synthetic dimension. Using a superconducting microwave resonator, we experimentally demonstrate Fock-space analogues of optical propagation, refraction, lensing, dispersion, and interference with up to 180 photons. These results establish a fundamental correspondence between Schrödinger evolution in a single bosonic mode and classical paraxial wave propagation. By mapping intuitive optical concepts onto high-dimensional quantum state engineering, our work opens a path toward scalable control of large-scale quantum systems with thousands of photons and advanced bosonic information processing.

This paper presents a theoretical study of the light-induced shift of the coherent population trapping resonance. An analytical model is proposed that describes the interaction of two radiation components with an atomic system using a $Λ$ scheme and takes into account an additional level of excited state. Both weak and strong coupling regimes with off-resonant transitions are considered. It is shown that, in addition to the conventional dynamic Stark shift, an extra shift arises due to the distortion of the resonance line shape when bichromatic laser radiation interacts with off-resonant atomic transitions. An analytical expression for this additional shift is derived in the weak-coupling limit, and its significant impact on the resonance shape and sensitivity to the intensities of the laser field components is demonstrated. It is found that under strong coupling conditions, the additional shift can deviate substantially from a linear dependence on light intensity, suggesting new opportunities for controlling light shifts in precision atomic devices such as quantum frequency standards.

In this study, we investigate the geodesic motion of massless particles -- specifically photons -- in the spacetime of a charged anti-de Sitter (AdS) black hole (BH) surrounded by a cloud of strings (CoS) within the framework of Kalb-Ramond (KR) gravity. We analyze the effective potential that governs photon trajectories, explore the properties and location of the photon sphere (PS), and examine the effective radial force acting on photons. The resulting BH shadow is also studied, highlighting the roles of both the CoS parameter $α$ and the KR field parameter $\ell$ in shaping its geometry. We constrain these parameters using observational data from M87* and Sgr A* obtained by the Event Horizon Telescope (EHT). Furthermore, we extend our investigation to the motion of neutral test particles in the same gravitational background. By examining the impact of the CoS and KR field, we show how these additional fields modify the dynamics relative to standard charged BH scenarios. Finally, we study the fundamental frequencies associated with quasiperiodic oscillations (QPOs) of test particles, demonstrating how these frequencies are affected by the presence of the CoS and KR field. Our results reveal the rich structure of AdS-BH spacetimes influenced by string clouds and antisymmetric tensor fields, with potential observational consequences in gravitational wave and BH imaging astronomy.

A relativistic equation for a neutral complex field as a probability amplitude is proposed. The continuity equation for the probability density is obtained. It is shown that there are two types of excitations of this field, which describe particles with positive energy and different dispersion laws. Based on the Lagrangian formalism, conservation laws are obtained. The transition to secondary quantization is considered.

Several photonic quantum technologies rely on the ability to generate multiple indistinguishable photons. Benchmarking the level of indistinguishability of these photons is essential for scalability. The Hong-Ou-Mandel dip provides a benchmark for the indistinguishability between two photons, and extending this test to the multi-photon setting has so far resulted in a protocol that computes the genuine n-photon indistinguishability (GI). However, this protocol has a sample complexity that increases exponentially with the number of input photons for an estimation of GI up to a given additive error. To address this problem, we introduce new theorems that strengthen our understanding of the relationship between distinguishability and the suppression laws of the quantum Fourier transform interferometer (QFT). Building on this, we propose a protocol using the QFT for benchmarking GI that achieves constant sample complexity for the estimation of GI up to a given additive error for prime photon numbers, and sub-polynomial scaling otherwise, representing an exponential improvement over the state of the art. We prove the optimality of our protocol in many relevant scenarios and validate our approach experimentally on Quandela's reconfigurable photonic quantum processor, where we observe a clear advantage in runtime and precision over the state of the art. We therefore establish the first scalable method for computing multi-photon indistinguishability, which applies naturally to current and near-term photonic quantum hardware.

We consider the question of counting the degrees of freedom in theoretical models, with an emphasis on theories of fields and gravity. Among the possible approaches, the Hamiltonian formulation remains one of the most systematic and robust tools. However, it can easily become long and technically involved. In this work, we present a broadly applicable recipe to find the degrees of freedom directly, based on the Lagrangian formulation. We compare it to the standard approaches, highlight the challenges that may arise in the latter, and demonstrate that the proposed method leads to transparent insights about the dynamical nature of theory in a quick, simple, and straight-forward way.

This paper presents a systematic study of adversarial hypothesis testing for both quantum-quantum (QQ) and classical-quantum (CQ) channels. Unlike conventional channel discrimination, we consider a framework where the sender, Alice, selects the channel input adversarially to minimize Bob's distinguishability. We analyze this problem across four settings based on whether Alice employs i.i.d. or general inputs and whether the receiver, Bob, is informed of the specific input choice (allowing his measurement to depend on the input). We characterize the Stein exponents for each setting and reveal a striking distinction in behavior: for QQ channels with i.i.d. inputs, Bob's knowledge of the input significantly enhances distinguishability, yet this advantage vanishes when general inputs are permitted. In contrast, for CQ channels, Bob being informed provides a consistent advantage over the corresponding entanglement-breaking channels for both i.i.d. and general inputs. These results demonstrate a unique phenomenon in adversarial hypothesis testing where the CQ channel does not merely behave as a special case of the QQ channel.

Waves propagating through a gravitational potential exhibit wave-optics effects when their wavelength is not significantly smaller than the lensing scales. We study the propagation of a scalar wave, governed by the Klein-Gordon equation in curved spacetime, to focus on effects on amplitude and phase, while leaving aside the issue of wave polarization which affects electromagnetic and gravitational waves. Using the Newman-Penrose formalism, we obtain the first corrections beyond the geometric optics in the expansion in the inverse frequency. In vacuum, that is for Weyl tensor lensing, there is no wave effect at first order in $G$ and wave effects start at order $G^2$. Conversely, if the wave travels through a non-vanishing matter density, the first corrections start at order $G$. We check these analytic results by solving numerically the equations dictating the evolution of the corrections either in the vicinity of a Schwarzschild black hole or through a transparent star.

In the thermodynamic limit, the Dicke-Ising chain maps exactly onto an effective self-consistent matter Hamiltonian with the photon field acting solely as a self-consistent effective field. As a consequence, no quantum correlations between photons and spins are needed to understand the quantum phase diagram. This enables us to determine the quantum phase diagram in the thermodynamic limit using numerical linked-cluster expansions combined with density matrix renormalization group calculations (NLCE+DMRG) to solve the resulting self-consistent matter Hamiltonian. This includes magnetically ordered phases with significantly improved accuracy compared to previous estimates. For ferromagnetic Ising couplings, we refine the location of the multicritical point governing the change in the order of the superradiant phase transition, reaching a relative accuracy of $10^{-4}$. For antiferromagnetic Ising couplings, we confirm the existence of the narrow antiferromagnetic superradiant phase in the thermodynamic limit. The effective matter Hamiltonian framework identifies the antiferromagnetic superradiant phase as the many-body ground state of an antiferromagnetic transverse-field Ising model with longitudinal field. This phase emerges through continuous Dicke-type polariton condensation from the antiferromagnetic normal phase, followed by a first-order transition to the paramagnetic superradiant phase. Thus, NLCE+DMRG provides a precise determination of the Dicke-Ising phase diagram in one dimension by solving the self-consistent effective matter Hamiltonian.

We analyze quantum state preservation in open quantum systems using quantum error-correcting (QEC) codes that are explicitly embedded into microscopic system-bath models. Instead of abstract quantum channels, we consider multi-qubit registers coupled to bosonic thermal environments, derive a second-order master equation for the reduced dynamics, and use it to benchmark the five-qubit, Steane, and toric codes under local and collective noise. We compute state fidelities for logical qubits as functions of coupling strength, bath temperature, and the number of correction cycles. In the low-temperature regime, we find that repeated error-correction with the five-qubit code strongly suppresses decoherence and relaxation, while in the high-temperature regime, thermal excitations dominate the dynamics and reduce the benefit of all codes, though the five-qubit code still outperforms the Steane and toric codes. For two-qubit Werner states, we identify a critical evolution time before which QEC does not improve fidelity, and this time increases as entanglement grows. After this critical time, QEC does improve fidelity. Comparative analysis further reveals that the five-qubit code (the smallest perfect code) offers consistently higher fidelities than topological and concatenated architectures in these open-system settings. These findings establish a quantitative framework for evaluating QEC under realistic noise environments and provide guidance for developing noise-resilient quantum architectures in near-term quantum technologies.

Pre-execution calibration is a major bottleneck for operating superconducting quantum processors, and qubit frequency allocation is especially challenging due to crosstalk-coupled objectives. We establish that the widely-used Snake optimizer is mathematically equivalent to Block Coordinate Descent (BCD), providing a rigorous theoretical foundation for this calibration strategy. Building on this formalization, we present a topology-aware block ordering obtained by casting order selection as a Sequence-Dependent Traveling Salesman Problem (SD-TSP) and solving it efficiently with a nearest-neighbor heuristic. The SD-TSP cost reflects how a given block choice expands the reduced-circuit footprint required to evaluate the block-local objective, enabling orders that minimize per-epoch evaluation time. Under local crosstalk/bounded-degree assumptions, the method achieves linear complexity in qubit count per epoch, while retaining calibration quality. We formalize the calibration objective, clarify when reduced experiments are equivalent or approximate to the full objective, and analyze convergence of the resulting inexact BCD with noisy measurements. Simulations on multi-qubit models show that the proposed BCD-NNA ordering attains the same optimization accuracy at markedly lower runtime than graph-based heuristics (BFS, DFS) and random orders, and is robust to measurement noise and tolerant to moderate non-local crosstalk. These results provide a scalable, implementation-ready workflow for frequency calibration on NISQ-era processors.

Studying the complexity of states sampled from various ensembles is a central component of quantum information theory. In this work we establish the average-case hardness of learning, in the statistical query model, the Born distributions of states sampled uniformly from the circular and (fermionic) Gaussian ensembles. These ensembles of states are induced variously by the uniform measures on the compact symmetric spaces of type AI, AII, and DIII. This finding complements analogous recent results for states sampled from the classical compact groups. On the technical side, we employ a somewhat unconventional approach to integrating over the compact groups which may be of some independent interest. For example, our approach allows us to exactly evaluate the total variation distances between the output distributions of Haar random unitary and orthogonal circuits and the constant distribution, which were previously known only approximately.

We demonstrate that large language model (LLM) agents can autonomously perform tensor network simulations of quantum many-body systems, achieving approximately 90% success rate across representative benchmark tasks. Tensor network methods are powerful tools for quantum simulation, but their effective use requires expertise typically acquired through years of graduate training. By combining in-context learning with curated documentation and multi-agent decomposition, we create autonomous AI agents that can be trained in specialized computational domains within minutes. We benchmark three configurations (baseline, single-agent with in-context learning, and multi-agent with in-context learning) on problems spanning quantum phase transitions, open quantum system dynamics, and photochemical reactions. Systematic evaluation using DeepSeek-V3.2, Gemini 2.5 Pro, and Claude Opus 4.5 demonstrates that both in-context learning and multi-agent architecture are essential. Analysis of failure modes reveals characteristic patterns across models, with the multi-agent configuration substantially reducing implementation errors and hallucinations compared to simpler architectures.

Historically, the focus in entanglement distillation has predominantly been on the distillable entanglement, and the framework assumes complete knowledge of the initial state. In this paper, we study the reliability function of entanglement distillation, which specifies the optimal exponent of the decay of the distillation error when the distillation rate is below the distillable entanglement. Furthermore, to capture greater operational significance, we extend the framework from the standard setting of known states to a black-box setting, where distillation is performed from a set of possible states. We establish an exact finite blocklength result connecting to composite correlated hypothesis testing without any redundant correction terms. Based on this, the reliability function of entanglement distillation is characterized by the regularized quantum Hoeffding divergence. In the special case of a pure initial state, our result reduces to the error exponent for entanglement concentration derived by Hayashi et al. in 2003. Given full prior knowledge of the state, we construct a concrete optimal distillation protocol. Additionally, we analyze the strong converse exponent of entanglement distillation. While all the above results assume the free operations to be non-entangling, we also investigate other free operation classes, including PPT-preserving, dually non-entangling, and dually PPT-preserving operations.

Recent studies have extensively explored chaotic dynamics in quantum optical systems through the mean-field approximation, which corresponds to an ideal, fluctuation-free scenario. However, the inherent sensitivity of chaos to initial conditions implies that even minute fluctuations can be amplified, thereby questioning the applicability of this approximation. Here, we analyze these chaotic effects using stochastic Langevin equations or the Lindblad master equation. For systems operating at frequencies of $10^5$ to $10^7$ Hz, we demonstrate that room-temperature thermal fluctuations are sufficient to suppress chaos at the level of expectation values, even under weak nonlinearity. Furthermore, nonlinearity induces deviations from Gaussian phase-space distributions of the quantum state, revealing attractor-like features in the Wigner function. With increasing nonlinearity, the noise threshold for chaos suppression decreases, approaching the scale of vacuum fluctuations. These results provide a bidirectional validation of the quantum mechanical suppression of chaos.

We present a comprehensive study of warm hybrid inflation within the framework of $α$-attractor models, where an axionic inflaton is coupled to a waterfall field in the presence of thermal dissipation. The model is analyzed for both linear ($Υ\propto T$) and cubic ($Υ\propto T^{3}$) dissipation regimes. Confronting the theoretical predictions with the latest observational data from Planck+BICEP/Keck, P-ACT-LB-BK18 and SPT, and , we find that in the weak dissipative regime ($Q_{*} \lesssim 10^{-5}$), the scalar spectral index $n_{s} \simeq 0.965$ lies at the boundary of the combined P-ACT-LB-BK18 constraints, while the tensor-to-scalar ratio $r$ remains within observable ranges. For stronger dissipation ($Q_{*} \gtrsim 10^{-5}$), the model predicts values of $n_{s}$ well within the $1$--$2σ$ confidence region of all datasets, with tensor modes remaining fully observable in both dissipation scenarios. These results indicate that forthcoming CMB polarization experiments may be capable of detecting primordial gravitational waves, thereby providing a robust observational test of warm hybrid inflation across different dissipative regimes.

The transition to the fault-tolerant era exposes the limitations of homogeneous quantum systems, where no single qubit modality simultaneously offers optimal operation speed, connectivity, and scalability. In this work, we propose a strategic approach to Heterogeneous Quantum Architectures (HQA) that synthesizes the distinct advantages of the superconducting (SC) and neutral atom (NA) platforms. We explore two architectural role assignment strategies based on hardware characteristics: (1) We offload the latency-critical Magic State Factory (MSF) to fast SC devices while performing computation on scalable NA arrays, a design we term MagicAcc, which effectively mitigates the resource-preparation bottleneck. (2) We explore a Memory-Compute Separation (MCSep) paradigm that utilizes NA arrays for high-density qLDPC memory storage and SC devices for fast surface-code processing. Our evaluation, based on a comprehensive end-to-end cost model, demonstrates that principled heterogeneity yields significant performance gains. Specifically, our designs achieve $752\times$ speedup over NA-only baselines on average and reduce the physical qubit footprint by over $10\times$ compared to SC-only systems. These results chart a clear pathway for leveraging cross-modality interconnects to optimize the space-time efficiency of future fault-tolerant quantum computers.

We explore the cosmological dynamics of a teleparallel Gauss-Bonnet gravity model defined by the torsion scalar $T$ and the torsion-based Gauss-Bonnet invariant $T_{\mathcal{G}}$, deriving modified Friedmann equations for a flat FLRW Universe and corresponding linear scalar perturbation equations. Using a numerical approach, we solve these equations for pressureless matter, predicting the redshift evolution of the Hubble parameter $H(z)$. Bayesian Markov chain Monte Carlo analysis, incorporating late-time observations from Cosmic Chronometers, Pantheon$^+$ with SH0ES, and DESI BAO (Data Release 1 and Data Release 2), constrains the model parameters, revealing that $f(T, T_{\mathcal{G}})$ mimics dark energy in the absence of a cosmological constant, presenting a viable alternative to $Λ$CDM paradigm. Stability is confirmed via scalar perturbation analysis of Hubble and matter density fluctuations, positioning $f(T, T_{\mathcal{G}})$ gravity as a robust framework to address cosmic acceleration challenges. The model yields a present-day effective equation of state $ω_{\mathrm{eff}}(z=0) \approx -0.664$ to \(-0.693\), consistent with observations, and partially alleviates the Hubble tension with $H_0$ estimates of 69 to 71.5\kms. These findings highlight the potential of $f(T, T_{\mathcal{G}})$ gravity to resolve fundamental cosmological puzzles while aligning with late-time observational data.

Casimir forces arise from quantum electromagnetic fluctuations and depend on the dielectric response of interacting materials across the entire frequency spectrum. Although this dependence is central to Lifshitz theory of the Casimir effect, the formulation of the force in terms of dielectric functions evaluated at imaginary frequencies has largely obscured its connection to real-frequency optical properties, limiting the use of Casimir interactions as a probe of materials. Here we demonstrate that Casimir force measurements encode sufficient information to reconstruct a material's broadband optical response. Using supervised machine learning to invert Lifshitz theory, we determine the complex permittivity of a material over more than seven orders of magnitude in frequency from a single force-distance curve. We show that measurements at different separations selectively constrain distinct frequency ranges of the dielectric response, providing direct physical insight into how quantum fluctuations sample the electromagnetic spectrum. These results establish Casimir interactions as a physically constrained, broadband spectroscopic tool and open new opportunities for optical characterization in regimes inaccessible to conventional techniques.

Magic states are essential for universal quantum computation and are widely viewed as a key source of quantum advantage, yet in realistic devices they are inevitably noisy. In this work, we characterize how noise on injected magic resources changes the classical simulability of quantum circuits and when it induces a transition from classically intractable behavior to efficient classical simulation. We adopt a resource-centric noise model in which only the injected magic components are noisy, while the baseline states, operations, and measurements belong to an efficiently simulable family. Within this setting, we develop an approximate classical sampling algorithm with controlled error and prove explicit noise-dependent conditions under which the algorithm runs in polynomial time. Our framework applies to both qubit circuits with Clifford baselines and fermionic circuits with matchgate baselines, covering representative noise channels such as dephasing and particle loss. We complement the analysis with numerical estimates of the simulation cost, providing concrete thresholds and runtime scaling across practically relevant parameter regimes.

We study the Fano effect in dissipative cavity quantum electrodynamics, which originates from the interference between the emitter's direct radiation and that mediated by a cavity mode. Starting from a two-level system coupled to a structured reservoir, we show that a quantum master equation previously derived within the Born-Markov approximation can be rederived by introducing a single auxiliary mode via pseudomode approach. We identify the corresponding spectral function of the system--environment interaction and demonstrate that it consists of a constant and a non-Lorentzian contribution forming the Fano profile. The constant term is shown to be essential for obtaining a Lindblad master equation and is directly related to the rate associated with this Fano interference. Furthermore, by applying Fano diagonalization to a common-environment setup including an explicit cavity mode, we independently derive the same spectral function in the strongest-interference regime. Our results establish a unified framework for describing the Fano effect in single-mode cavity QED systems and clarify its non-Markovian origin encoded in the spectral function.

Static phase detuning fundamentally constrains coherent state transfer in asymmetric classical and quantum systems. We introduce a Bloch-sphere formulation for piecewise-coherent modulation that recasts coupled-mode dynamics as geometric trajectories, transforming algebraic control into path optimization. The approach reveals a cone of inaccessibility at the target pole and yields exact geodesic criteria for complete mode conversion in detuned systems. Leveraging this framework, we break time-reversal symmetry to realize a magnet-free optical isolator with near-unity contrast. Furthermore, for detuning larger than coupling between modes, we develop a recursive multi-step protocol enabling deterministic transfer for arbitrary detunings and derive a universal geometric lower bound on the required number of coupling-switching events.

We study SU(3) gauge theory on small lattices in the minimal (qutrit) electric field truncation retaining only the ${\bf 1}, {\bf 3}, {\bf \overline{3}}$ representations for the link variables. Explicit expressions are given for the Kogut-Susskind Hamiltonian for the square plaquette chain and the two-dimensional honeycomb lattice. Our formalism can be easily extended to the minimally truncated general SU($N_c$) gauge theory. The addition of (static) quarks is discussed. We present results for the energy spectrum of the gauge field on these lattices by exact diagonalization of the Hamiltonian and analyze its statistical properties. We also compute the SU(3) string tension and discuss how it is modified by vacuum fluctuations. Finally, we calculate the potential energies of a static quark-antiquark pair and three static quarks and study their screening at finite temperature.

Quantum state tomography is essential for characterizing quantum systems, but it becomes infeasible for large systems due to exponential resource scaling. Overlapping tomography addresses this challenge by reconstructing all $k$-body marginals using few measurement settings, enabling the efficient extraction of key information for many quantum tasks. While optimal schemes are known for qubits, the extension to higher-dimensional qudit systems remains largely unexplored. Here, we investigate optimal qudit overlapping tomography, constructing local measurement settings from generalized Gell-Mann matrices. By establishing a correspondence with combinatorial covering arrays, we present two explicit constructions of optimal measurement schemes. For $n$-qutrit systems, we prove that pairwise tomography requires at most $8 + 56\left\lceil \log_{8} n \right\rceil$ measurement settings, and provide an explicit scheme achieving this bound. Furthermore, we develop an efficient algorithm to determine the optimal order of these measurement settings, minimizing the experimental overhead associated with switching configurations. Compared to the worst-case ordering, our optimized schedule reduces switching costs by approximately 50\%. These results provide a practical pathway for efficient characterization of qudit systems, facilitating their application in quantum communication and computation.

The high overhead of fault-tolerant measurement sequences (FTMSs) poses a major challenge for implementing quantum stabilizer codes. Here, we address this problem by constructing efficient FTMSs for the class of quantum Hamming codes $[\![2^r-1, 2^r-1-2r, 3]\!]$ with $r=3k+1$ ($k \in \mathbb{Z}^+$). Our key result demonstrates that the sequence length can be reduced to exactly $2r+1$-only one additional measurement beyond the original non-fault-tolerant sequence, establishing a tight lower bound. The proposed method leverages cyclic matrix transformations to systematically combine rows of the initial stabilizer matrix and preserving a self-dual CSS-like symmetry analogous to that of the original quantum Hamming codes. This induced symmetry enables hardware-efficient circuit reuse: the measurement circuits for the first $r$ stabilizers are transformed into circuits for the remaining $r$ stabilizers simply by toggling boundary Hadamard gates, eliminating redundant hardware. For distance-3 fault-tolerant error correction, our approach simultaneously reduces the time overhead via shorting the FTMS length and the hardware overhead through symmetry-enabled circuit multiplexing. These results provide an important advance towards the important open problem regarding the design of minimal FTMSs for quantum Hamming codes and may shed light on similar challenges in other quantum stabilizer codes.

Decision making often exhibits context dependence that challenges classical probability theory. While quantum cognition has successfully modeled such phenomena, it remains unclear whether quantum probability is merely a convenient assumption or a necessary consequence of decision dynamics. Here we present a theoretical framework in which contextuality arises generatively from physically grounded constraints on decision making. By developing a quantum extension of the Tug-of-War (TOW) model, we show that conservation-based internal state updates and measurement-induced disturbance preclude any non-contextual classical description with a single, unified internal state. Contextuality therefore emerges as a structural consequence of adaptive learning dynamics. We further show that the resulting measurement structure admits Klyachko-Can-Binicioglu-Shumovsky (KCBS)-type contextuality witnesses in a minimal single-system setting. These results indicate that quantum probability is not merely a descriptive convenience, but an unavoidable effective theory for adaptive decision dynamics.

Our objective is to illuminate the global structure of non-orientable manifolds with signature-changing metrics. Using explicit constructions based on the topology of the Möbius strip, we produce examples of crosscap manifolds where the gluing junction serves as the locus of signature change. In another set of examples, we convert the Möbius strip into a singular signature-type changing manifold. For these resulting manifolds, we test whether the metric can be expressed as $\tilde{g}=g+fV^{\flat}\otimes V^{\flat}$, with $g$ a Lorentzian metric and $f$ a smooth interpolation function between the Lorentzian and Riemannian regions, separated by the signature change hypersurface $\mathcal{H}$. Our analysis reveals that the radical of the metric can transition from transverse to tangent at $\mathcal{H}$, pseudo-space orientability is obstructed by the Euler characteristic, and pseudo-time orientability may still hold. These examples illustrate subtle obstructions to applying standard transformation prescriptions for signature change and highlight novel phenomena in compact, non-orientable semi-Riemannian manifolds.

Accurately capturing electron correlation in large-scale molecular systems remains one of the foremost challenges in quantum chemistry and a primary driver for the development of quantum algorithms. Classical configuration-interaction methods, while rigorous, suffer from exponential scaling, rendering them impractical for large or strongly correlated systems. Overcoming this limitation is central to realizing the promise of quantum computing in chemistry. Here, we investigate the pyridine-Li ion complex using three quantum algorithms: the variational quantum eigensolver (VQE), the subspace quantum diagonalization (SQD) method, and the recently introduced handover iterative VQE (HI-VQE). Our results demonstrate how new generations of hybrid quantum-classical frameworks overcome the scalability and noise sensitivity that constrain conventional VQE approaches. SQD and HI-VQE achieve ground-state energy calculations for problem sizes inaccessible to classical computation, marking a clear advance toward quantum advantage. In particular, HI-VQE enables calculations within active spaces as large as (24e,22o), requiring 44 qubits-well beyond the reach of classical CASCI and VQE. This capability provides a systematic pathway for incorporating increasing numbers of electrons into quantum treatment, thereby approaching exact molecular energies. Importantly, both SQD and HI-VQE exhibit robustness against hardware noise, a critical improvement over earlier approaches. By enabling quantum simulations of molecular systems previously deemed intractable, SQD and HI-VQE offer a realistic route toward practical quantum advantage in computational chemistry. The comparison between HI-VQE and SQD shows that optimizing circuit parameters is crucial for accurate simulation.

The subadditivity-doubling-rotation (SDR) technique is a powerful route to Gaussian optimality in classical information theory and relies on strict subadditivity and its equality-case analysis, where double Markovity is a standard tool. We establish quantum analogues of double Markovity. For tripartite states, we characterize the simultaneous Markov conditions A-B-C and A-C-B via compatible projective measurements on B and C that induce a common classical label J yielding A-J-(BC). For strictly positive four-party states, we show that A-(BD)-C and A-(CD)-B hold if and only if A-D-(BC) holds. These results remove a key bottleneck in extending SDR-type arguments to quantum systems.

Quantities computed by minimal cuts, such as entanglement entropies achievable by the Ryu-Takayanagi proposal in the AdS/CFT correspondence, are constrained by linear inequalities. We prove a previously conjectured property of all such constraints: Any $k$ systems on the "greater-than" side of the inequality are subsumed in some $k$ systems on its "less-than" side (accounting for multiplicity). This finding adds evidence that the same inequalities also constrain the entropies under time-dependent conditions because it preempts a large class of potential counterexamples. We prove several other properties of holographic entropy inequalities and comment on their relation to quantum erasure correction and the Renormalization Group.

A holographic entropy inequality (HEI) is a linear inequality obeyed by Ryu-Takayanagi holographic entanglement entropies, or equivalently by the minimum cut function on weighted graphs. We establish a new combinatorial framework for studying HEIs, and use it to prove several properties they share, including two majorization-related properties as well as a necessary and sufficient condition for an inequality to be an HEI. We thereby resolve all the conjectures presented in [arXiv:2508.21823], proving two of them and disproving the other two. In particular, we show that the null reduction of any superbalanced HEI passes the majorization test defined in [arXiv:2508.21823], thereby providing strong new evidence that all HEIs are obeyed in time-dependent holographic states.

Photon statistics plays a governing role in multi-photon interference. While interference visibility in the standard two-photon case, known as Hong-Ou-Mandel interference, monotonically degrades with higher intensity correlation functions, we show that this monotonicity does not hold for three-photon interference in symmetric circuits. We reveal that, in the discrete Fourier transform circuit, engineered super-Poissonian photon-number fluctuations, realized using a modulated laser, maximize the visibility, surpassing the magnitude of the single-photon signature. In addition, by tuning the symmetric circuit parameters, we demonstrate that the visibility hierarchy inverts relative to the benchmark of Poissonian statistics. This trade-off implies that quantum and classical advantages are mutually exclusive resources for interference, indicating a form of statistical complementarity.

There exist conformally invariant, higher-derivative, variational analogs of the Yang-Mills condition for connections on vector bundles over a conformal manifold of even dimension greater than or equal to six. We give a compact formula for these analogs and prove that they are a strict weakening of the Yang-Mills condition with respect to an Einstein metric. We also show that the conformal Yang-Mills condition for the tractor connection of an even dimensional conformal manifold is equivalent to vanishing of its Fefferman-Graham obstruction tensor. This result uses that the tractor connection on a Poincaré-Einstein manifold is itself Yang-Mills.

Interface circuits are the key components that enable the hybrid integration of superconductor and semiconductor digital electronics. The design requirements of superconductor-semiconductor interface circuits vary depending on the application, such as high-performance classical computing, superconducting quantum computing, and digital signal processing. In this survey, various interface circuits are categorized based on the working principle and structure. The superconducting output drivers are explored, which are capable of converting and amplifying, e.g., single flux quantum (SFQ) voltage pulses, to voltage levels that semiconductor circuits can process. Several trade-offs between circuit- and system-level design parameters are examined. Accordingly, parameters such as the data rate, output voltage, power dissipation, layout area, thermal/heat load of cryogenic cables, and bit-error rate are considered.

The Variational Quantum Eigensolver (VQE) is a hybrid quantum-classical algorithm for computing ground state energies of molecular systems. We implement VQE to calculate the potential energy surface of the hydrogen molecule (H$_2$) across 100 bond lengths using the PennyLane quantum computing framework on an HPC cluster featuring 4$\times$ NVIDIA H100 GPUs (80GB each). We present a comprehensive parallelization study with four phases: (1) Optimizer + JIT compilation achieving 4.13$\times$ speedup, (2) GPU device acceleration achieving 3.60$\times$ speedup at 4 qubits scaling to 80.5$\times$ at 26 qubits, (3) MPI parallelization achieving 28.5$\times$ speedup, and (4) Multi-GPU scaling achieving 3.98$\times$ speedup with 99.4% parallel efficiency across 4 H100 GPUs. The combined effect yields 117$\times$ total speedup for the H$_2$ potential energy surface (593.95s $\rightarrow$ 5.04s). We conduct a CPU vs GPU scaling study from 4--26 qubits, finding GPU advantage at all scales with speedups ranging from 10.5$\times$ to 80.5$\times$. Multi-GPU benchmarks demonstrate near-perfect scaling with 99.4% efficiency and establish that a single H100 can simulate up to 29 qubits before hitting memory limits. The optimized implementation reduces runtime from nearly 10 minutes to 5 seconds, enabling interactive quantum chemistry exploration.

Quantum Computing (QC) has evolved from a few custom quantum computers, which were only accessible to their creators, to an array of commercial quantum computers that can be accessed on the cloud by anyone. Accessing these cloud quantum computers requires a complex chain of tools that facilitate connecting, programming, simulating algorithms, estimating resources, submitting quantum computing jobs, retrieving results, and more. Some steps in the chain are hardware dependent and subject to change as both hardware and software tools, such as available gate sets and optimizing compilers, evolve. Understanding the trade-offs inherent in this process is essential for evaluating the power and utility of quantum computers. ARLIS has been systematically investigating these environments to understand these complexities. The work presented here is a detailed summary of three months of using such quantum programming environments. We show metadata obtained from these environments, including the connection metrics to the different services, the execution of algorithms, the testing of the effects of varying the number of qubits, comparisons to simulations, execution times, and cost. Our objective is to provide concrete data and insights for those who are exploring the potential of quantum computing. It is not our objective to present any new algorithms or optimize performance on any particular machine or cloud platform; rather, this work is focused on providing a consistent view of a single algorithm executed using out-of-the-box settings and tools across machines, cloud platforms, and time. We present insights only available from these carefully curated data.

We analyze a static open string in a general stationary spacetime, which can represent a heavy quark-antiquark pair within the holographic framework or effective theory. We establish that for a simple U-shaped string with only radial dependence on the space string coordinate, $x_r'(σ) \neq 0$, the string is generally not symmetric about its turning point, and the symmetry restores only for backgrounds with $h_{pr} = G_{00} G_{pr} - G_{0p} G_{0r} = 0$. Consequently, such asymmetric strings directly probe a possibility of the parity violation in the quark-antiquark interaction. Nevertheless, we identify a wide family of metrics for which the symmetry is preserved, enabling a direct isolation of the linear-in-distance term in the static interquark potential for simple symmetric string configurations, even in non-diagonal backgrounds. Applying the holographic framework, we further study the Rindler-AdS spacetime dual to an accelerated $\mathcal{N}=4$ super Yang-Mills plasma. We show that the distance between quarks decreases, the static potential between them increases, and the deconfinement phase transition temperature, $T_{\rm dec} = (π/3) T_H = a_c/6$, increases with an acceleration. However, we observe that an acceleration-scaled potential as a function of the acceleration-scaled distance does not depend on the certain value of the acceleration This result, reflecting the scale invariance and self-similarity of the holographic setup, can be also obtained in the dimensionless metric after scaling of the coordinates onto the acceleration, $\tilde{x}_i = a_c x_i$, for which one obtains an universal value of the phase transition temperature, $\tilde{T}_{\rm dec} = (π/3) \tilde{T}_H = 1/6$.

We give an overview of the construction of Borcherds algebras, particularly the Monstrous Lie algebras $\mathfrak m_g$ constructed by Carnahan, where $g$ is an element of the Monster finite simple group. When $g$ is the identity element, $\mathfrak m_g$ is the Monster Lie algebra of Borcherds. We discuss the appearance of the $\mathfrak m_g$ in compactified models of the Heterotic String. We also summarize recent work on associating Lie group analogs to the Lie algebras $\mathfrak m_g$. We include a discussion of some open problems.

We investigate the in-medium modification and dissociation of heavy quarkonium in a hot QCD medium at finite quark chemical potential and in the weak magnetic-field regime. Starting from the one-loop resummed gluon propagator in the imaginary-time formalism, and incorporating non-perturbative effects through a phenomenological correction to the HTL description, we compute the real and imaginary parts of the dielectric permittivity. This, in turn, leads to a complex heavy-quark potential: the real part is used to determine binding energies by solving the nonrelativistic Schrödinger equation, while the imaginary part generates thermal decay widths, dominated by Landau damping. Within the explored parameter range, temperature has the greatest control over Debye screening, potential modification, and quarkonium stability, whereas finite density and weak magnetic fields introduce comparatively smaller quantitative changes. As the temperature increases, binding energies decrease and thermal widths grow, giving rise to the expected hierarchy between ground and excited states and a sequential suppression pattern in the dissociation temperatures. Overall, our results indicate that while finite chemical potential and weak magnetic fields can shift quarkonium properties in a measurable way, thermal effects remain the primary driver of dissociation, with direct relevance for heavy-ion collision phenomenology.

In the minimal helicity-chirality formalism, we systematically construct higher-point massive amplitudes from the fundamental building blocks: the contact three-point and four-point massive amplitudes. The inclusion of four-point contact amplitudes is essential to maintain gauge invariance in the spontaneously broken Standard Model. We construct all the standard model massive contact amplitudes and identify the physical light-cone gauge nature of massive amplitudes. Then only using the contact minimal helicity-chirality amplitudes at the leading order, we show both bootstrap techniques and on-shell recursion relations can be utilized to compute higher-point massive amplitudes. This provides a systematic framework for constructing various higher-point electroweak amplitudes, analogous to established on-shell methods for massless theories. Finally by deforming the gauge-invariant $n$-point amplitudes, we extend the massless-massive correspondence from three-and-four point contact amplitudes to general $n$-point factorized amplitudes.

In this work, the massless-massive correspondence for the on-shell scattering amplitudes is constructed so the massive amplitudes could inherit advantageous techniques developed in the massless calculation. This correspondence is established by matching massless amplitudes to Minimal Helicity-Chirality (MHC) amplitudes, which arise from an expansion of massive spin-spinor amplitudes in terms of the chirality-flip $mη$ order by order. The primary MHC amplitude deforms into a massless amplitude of the same helicity; if a vector boson is involved, it may instead vanish due to the associated conserved current. In cases where the primary amplitude vanishes, the leading contributions originate from descendant MHC amplitudes, each corresponding to a distinct massless amplitude in the ultraviolet theory containing either a transverse gauge boson or a Goldstone boson. We propose a systematic amplitude deformation procedure for three-point massless-massive matching based on helicity-chirality unification and the scaling properties of $mη$. Sub-leading MHC amplitudes are matched to massless amplitudes with additional on-shell Higgs splitting, a process known as on-shell Higgsing. In this work, we extend and reinterpret on-shell Higgsing as a transversality flip between different MHC states, and obtain all the 3-point massless-massive matching results in the spontaneous broken standard model.

Soft current algebras capture the infrared structure of scattering in asymptotically flat spacetimes, but an analogous algebraic description of finite-energy dynamics has been missing. We uncover an infinite-dimensional hard current algebra that encodes finite-energy contributions to scattering and implies novel Ward identities. The soft current algebras are not independent but arise naturally from the hard ones. This provides a unified algebraic framework underlying quantum theory in flat spacetime.

Using (10087$\pm$44)$\times$10$^{6}$ $J/ψ$ events collected with the BESIII detector at the BEPCII collider at the center-of-mass energy of $\sqrt{s}=3.097~\rm{GeV}$, we report the first search for $η\toπ^0S\toπ^0χ\barχ$ with $S$ denotes an on-shell dark scalar boson and $χ$ an invisible dark matter particle. No significant signals are observed with $S$ mass ranging from 0 to 400 $\rm{MeV}/c^2$. The upper limits on the branching fractions and the new physics coupling strengths between $S$ and quarks are set to be $(1.8\sim5.5)\times10^{-5}$ and $(1.3\sim3.2)\times10^{-5}$ at the 90% confidence level, respectively. The constraints on the dark-matter-nucleon scattering cross section is improved by approximately 5 orders of magnitude over previous dark-matter-nucleon scattering experiments, providing unique insights into sub-GeV dark matter.

We examine the first-order Einstein-Cartan (EC) action in 2+1 dimensions, including a cosmological term and its supersymmetric extension. In this setting the spin connection can be expressed as an axial vector, yielding an action that is bilinear in the quantum fields and allows quantization without background fields. We identify the complete set of first-class constraints and derive the associated gauge transformations, which differ from the standard diffeomorphism and local Lorentz invariances. Using the closed gauge algebra, we construct the Faddeev-Popov-Nielsen path integral and show how a Lagrange multiplier field can be introduced to remove higher-loop contributions while preserving unitarity and gauge invariance.

The common eigenfunctions of the twisted Cherednik operators can be first analyzed in the limit of $q\longrightarrow 1$. Then, the polynomial eigenfunctions form a simple set originating from the symmetric ground state of non-vanishing degree and excitations over it, described by non-symmetric polynomials of higher degrees and enumerated by weak compositions. This pattern is inherited by the full spectrum at $q\neq 1$, which can be considered as a deformation. The whole story looks like a typical NP problem: the Cherednik equations are difficult to solve, but easy to check the solution once it is somehow found.

We investigate six singly Cabibbo-suppressed decay channels in $D^0\to VP$ ( $V$ and $P$ stand for the ground state vector and pseudoscalar mesons, respectively), i.e. $D^{0}\to ρ^{+}π^{-}$, $ρ^{-}π^{+}$, $K^{*+}K^{-}$, $K^{*-}K^{+}$, $K^{*0}\bar{K}^{0}$, and $\bar{K}^{*0}K^{0}$. These decay channels share the similar transition mechanisms involving only the direct emission (DE) and internal conversion (IC) processes. We show that a combined analysis of these channels can explicitly highlight the role played by the IC processes which contribute to the amplitudes at the same order of magnitude as the DE processes.

Reconstruction of one-dimensional kinematic distributions from calculations based on high-dimensional Monte-Carlo integration is a standard problem in high-energy physics. Traditionally, this is done by collecting randomly-generated events in histograms. In this article, we explore an alternative approach, whose main idea is to approximate the target distribution by a weighted sum of orthogonal basis functions whose coefficients are calculated using the Monte-Carlo integration. This method has the advantage of directly yielding smooth approximations to target distributions. Furthermore, in the context of high-order perturbative calculations with local subtractions, it eliminates the so-called bin-to-bin fluctuations, which often severely affect the quality of conventional histograms. We also demonstrate that the availability of a high-quality approximation to the target distribution, for example the leading-order result in the perturbative expansion, can be exploited to construct an optimized orthonormal basis. We compare the performance of this method to conventional histograms in both toy-model and real Monte-Carlo settings, applying it to Higgs boson production in weak boson fusion as an example.

We present a comparative study of three Monte Carlo simulation frameworks -SRIM, GEANT4, and PHITS- for modeling the transport, stopping, and atomic cascade of negative muons in micrometer-scale, multilayer systems relevant to Muon-Induced X-ray Emission (MIXE) experiments at the Paul Scherrer Institute (PSI). Using a lithium-ion battery as a benchmark target, simulated implantation profiles are compared with experimental data from the GIANT spectrometer. All three codes reproduce the overall muon depth distributions with good consistency, even across sharp density contrasts. SRIM provides reliable implantation estimates for compact geometries, whereas PHITS reproduces GEANT4 results with comparable accuracy and additionally generates muonic X-ray spectra. These spectra, however, exhibit a systematic energy offset in the K-line transitions of medium- and high-Z elements relative to theoretical and experimental values. Despite this bias, PHITS accurately captures relative intensities and spectral shapes, enabling element-specific line identification. The results demonstrate that SRIM and PHITS constitute practical tools for rapid estimation of muon implantation and stopping profiles, and that PHITS holds strong potential for predictive MIXE spectroscopy once its transition-energy bias is corrected.

We perform a two-loop calculation in Light Cone Perturbation Theory (LCPT) to evaluate the next-to-leading order nonsinglet splitting function. Our calculation demonstrates the methodology and feasibility of performing higher order calculations in LCPT. Since in Hamiltonian perturbation theory the longitudinal $k^+$ momentum is always positive, poles in $1/k^+$ can be regularized by a simple cutoff which cancels in physical results, without any associated ambiguities. For transverse momentum integrals we use dimensional regularization. Developing methods for loop calculations in LCPT paves the way for a systematical, automatizable procedure for precision calculations in this framework with a transparent physical partonic interpretation. This can provide a standard framework in higher order calculations in the gluon saturation regime of QCD.

In the braneworld scenario, we introduce a uniformly distributed cloud of infinitely long strings in the five dimensional AdS bulk spacetime. The end points of the strings are attached to the brane and becomes the source of the four dimensional matter on the brane, while the body of the strings hang onto the radial direction of the bulk and act as the gluonic field on the brane. The presence of matter in the brane induces a nonsingular cosmological evolution for the scale factor of the brane world under certain conditions of mass and cosmological parameters. However, the nonsingular nature is unstable since the bounce occurs inside the Cauchy horizon. Further, we consider the shellworld or the dark bubble scenario for the same bulk spacetime. It shows stable nonsingular cosmological nature of the bubble universe under certain conditions on the bulk and bubble parameters.

It has recently been shown that when $N_f$ generations of chiral fermions are coupled in a specific manner to $N$ (with $N \geq 2N_f-1$) pairs of vectorlike fermions whose mass terms form a one-dimensional lattice-like structure in theory space, locality along the lattice ensures that only a single fermion generation acquires a mass at tree level. Radiative corrections can induce controlled departures from locality in the latticed space, thereby generating suppressed but non-vanishing masses for the remaining $N_f-1$ generations. In this work, we present an explicit implementation of this mechanism to address the flavour hierarchies of the Standard Model. After delineating the minimal extensions of the gauge, scalar, and Yukawa sectors required for feasible implementation of the mechanism, we demonstrate that the framework successfully reproduces the observed charged-fermion mass spectrum and quark mixing pattern. We analyse the new-physics effects arising from the extended sectors and confront them with existing constraints from direct, indirect searches and precision measurements. It is shown that a viable realisation of the mechanism allows the spectrum of vectorlike fermions and additional gauge boson to lie at scales as low as $\mathcal{O}(5)\,\mathrm{TeV}$ with the lightest states typically corresponding to top partners. This stands in sharp contrast to conventional radiative mass-generation scenarios, in which phenomenological constraints typically impose a lower bound on the new-physics scale of order a few hundred to several thousand TeV.

The neutrino masses ordering remains one of the most important open questions in neutrino physics. While upcoming oscillation experiments aim to resolve this problem at low energies, complementary approaches are highly desirable. In this Letter, we show that the neutrino mass ordering can be probed at high-energy colliders through the lepton-flavor structure of heavy neutral lepton (HNL) interactions. In the minimal Type-I seesaw scenario with two nearly degenerate HNLs, the heavy--light neutrino mixings are strongly correlated with the light-neutrino mass spectrum, leading to distinct flavor patterns for the normal and inverted hierarchies. We demonstrate that future $Z$ factories, such as CEPC and FCC-ee, can probe the neutrino mass ordering for total HNL mixings as small as $U_{\rm tot}^2 \gtrsim 4 \times 10^{-9}$, and discriminate between the two hierarchies for $U_{\rm tot}^2 \gtrsim 10^{-6}$. Our results establish collider searches for HNLs as a powerful and complementary probe of the neutrino mass ordering.

In this proceedings we explore the physics potential of the ESSnuSBplus setup to study beam and non-beam based physics scenarios in both standard and new physics cases. The ESSnuSBplus setup consists of three neutrino sources: the main ESS linac, a low energy monitored neutrino beam and a low energy nuSTORM facility and three detectors: the main far detector and two near detectors. The goal of this facility is to measure the leptonic CP phase with extremely high precision and the neutrino nucleus cross-section in the few hundred MeV region.

In this work, we investigate the effects of $\text{T}\bar{\text{T}}$ and root-$\text{T}\bar{\text{T}}$ deformations on reflected and entanglement entropy in the context of both pure and mixed state entanglement measures. Utilizing a mixed boundary condition framework, we analyze how these deformations modify entanglement structures and explore their implications in three-dimensional AdS space. Our results provide insights into the interplay between solvable irrelevant deformations and quantum information-theoretic quantities, shedding light on the entanglement structure of deformed theories.

Advances in fabrication techniques and high-performance electronics have facilitated the development of fine-pitch Gas Electron Multipliers (GEMs). Earlier experimental and simulation findings suggest that these reduced-pitch GEMs can outperform the standard configuration in terms of effective gain, collection efficiency, and position resolution. However, a noticeable fraction of avalanche electrons is lost within the GEM systems, resulting in a degradation of charge collection efficiency. Therefore, a comprehensive simulation-based study is essential to provide deeper insights into the extent of degradation and its contributing factors. In this context, we employ ANSYS and Garfield++ to model the Triple GEM detectors with reduced pitch sizes of 90 and 60 $μ$m, and perform a comparative performance analysis with the standard configuration (pitch size: 140 $μ$m). At first, the simulation framework is validated by comparing the results of the standard configuration with available experimental data and previously reported simulation outcomes. Despite the characteristic gain offset, the framework remains physically consistent and reliable in capturing microscopic avalanche dynamics, reproducing the experimental trend. Following validation, we investigate electron losses at the metal electrodes and within the Kapton holes, electron transmission through the transfer and induction regions, electron diffusion on the induction electrode, and the overall collection efficiency. These parameters are analyzed as functions of GEM potential, outer hole diameter, inner hole diameter, Kapton thickness, metal thickness, and gas composition, thereby offering insights for designing efficient GEM detectors.

Theoretical developments in relativistic spin hydrodynamics, which describes the macroscopic transport of spin angular momentum alongside other fundamental conserved quantities, have progressed rapidly since the experimental observation of the global spin polarization of $Λ$ hyperons in relativistic heavy-ion collision experiments. However, numerical simulations of relativistic spin hydrodynamics remain largely unaddressed due to computational challenges, particularly the accurate numerical conservation of total angular momentum. In this work, we propose the use of physics-informed neural networks (PINNs) for computational relativistic spin hydrodynamics. As a concrete application, we consider a rotating fluid confined within a cylindrical container. We show that angular-momentum conservation can be accurately achieved in the PINNs-based numerical framework. Furthermore, we investigate the spin-orbit conversion induced by the rotational viscous effect, which is the intrinsic dissipative process of relativistic spin hydrodynamics. Our analysis numerically identifies the mismatch between the transverse thermal vorticity and the spin potential as the driving mechanism of the spin-orbit conversion.

Based on the kinetic theory with relaxation time approximation, we investigate the bulk viscosity ($ζ$) and its ratio to shear viscosity ($ζ/η$) of quark matter at finite temperature and chemical potential with the in-medium particle masses derived in the 2+1 flavor Polyakov-loop improved Nambu--Jona-Lasinio (PNJL) model. We explore the behaviors of specific bulk viscosity ($ζ/s$) and $ζ/η$ across different QCD phase transitions, including the Mott phase transition, the chiral crossover, and the first-order transition with the associated metastable phase. The calculation shows that both $ζ/s$ and $ζ/η$ are extremely small at high temperatures, approaching the nature of a conformal theory. Larger $ζ/s$ and $ζ/η$ are derived near the chiral phase transition at finite temperature. Along the chiral crossover line, $ζ/s$ and $ζ/η$ generally increase with decreasing temperature, though $ζ/η$ exhibits a slight decline near the critical endpoint (CEP). On the boundary of the first-order transition, $ζ/s$ shows a non-monotonic variation with temperature. Furthermore, an additional peak structure emerges beyond the chiral phase boundary for both $ζ/s$ and $ζ/η$, with magnitudes even exceeding those near the chiral crossover of $u, d$ quarks. Our analysis indicates this peak originates from the chiral crossover transformation of strange quark.

The type I seesaw mechanism is one of the leading proposed explanations for how neutrinos acquire their tiny masses. However, the mass scale of the undiscovered right-handed neutrinos required by this mechanism remains undetermined. Assuming vanilla leptogenesis in the two-flavor regime, we work backwards to find the required general textures of the Dirac mass matrix from which we determine the mass of the lightest right-handed neutrino to be around $10^9 {\rm GeV}$ to $10^{12} {\rm GeV}$.

The study is devoted to comparing the parameters of the interstellar medium of HII regions in the Kongiu and Groves catalogs for the galaxy NGC 628. The article analyzes the characteristics of star-forming regions, including a comparison of radiation fluxes in the ranges of 7.7 $μ$m and 21 $μ$m and in the H$α$, H$β$, OIII and CO lines, calculating the kinematic parameters (FWHM) for the lines, and analyzing the spatial distribution of regions for both catalogs. The results of the study showed that the regions from the Groves catalog demonstrate higher line widths compared to the Kongiu catalog. Signs of possible misidentified classification of some regions from the Groves catalog were revealed: there is a possibility that some of them are not HII regions, but shock ionization regions.

We present a four-component high-resolution model of thermal dust emission for microwave and sub-mm frequencies derived from Planck HFI, WHAM and Gaia. The resulting high-resolution model derived here forms the basis for the thermal dust model employed in the Cosmoglobe DR2 reanalysis of COBE-DIRBE. The four dust components are called ``cold dust'', ``hot dust'', ``nearby dust'', and ``Ha correlated dust'', respectively, and trace different physical environments. The spatial distributions of the nearby dust and Ha dust components are defined by the Edenhofer et al. Gaia 3D extinction model and the WHAM survey, respectively, while the hot and cold dust components are fit freely pixel-by-pixel to the Planck HFI data. We use a global parameter grid search coupled to an amplitude map Gibbs sampler to fit this model to Planck HFI data. In agreement with the companion low-resolution analysis, we find that the hot dust component is strongly correlated with the FIRAS Cii map, while the cold dust component is strongly correlated with the HI4PI Hi map. Despite its fewer degrees of freedom per pixel compared to the Planck 2015 legacy dust model, we find that this new model performs competitively in terms of overall residuals, capturing over 98% of the full-sky dust variance for all channels. When fitting a spatially varying 3-parameter MBB model to the new dust model with isotropic SEDs, we find very similar spatial distributions to those of the official Planck analysis, and this new model thus represents an economical decomposition of previously published spatially varying spectral parameter maps. We conclude that this new model represents both a statistically more efficient summary of thermal dust in the microwave and far-infrared regimes and a physically more realistic decomposition of the sky compared to the traditional 3-parameter MBB model. (abridged)

Observations of active galactic nuclei have revealed a correlation between millimeter luminosity, X-ray luminosity, and mass, suggesting the emission in each of these bands is powered by a common source. Starting with a set of five general relativistic magnetohydrodynamic simulations with dynamically important magnetic fields, we perform ray-tracing calculations to produce spectra including synchrotron emission, bremsstrahlung emission, and Compton scattering. Our models with similar Eddington ratios to the objects for which the relationship was inferred naturally reproduce observations without tuning. Our lower Eddington ratio models depart from this relationship, likely attributable to an observational bias against extremely low accretion rates. We find that inverse Compton scattering dominates the production of X-rays over bremsstrahlung radiation in almost all models, and in all models consistent with the observed correlation. We find only a modest spin dependence in this relationship. This study demonstrates that a compact, hot accretion flow with dynamically important magnetic fields can naturally explain observed millimeter and X-ray properties in low-luminosity active galactic nuclei. Future work should explore the impacts of non-thermal electron populations, weaker magnetic fields, and radiative cooling.

The scarcity of high-fidelity extinction measurements remains a bottleneck in deriving accurate stellar properties from Gaia parallaxes. In this work, we aim to derive precision extinction estimates for APOGEE DR19 stars, establishing a new benchmark for Galactic stellar population studies. We first determine reddening by comparing observed colorsr, etrieved from photometric surveys or standardized synthetic magnitudes from Gaia BP/RP spectra, to intrinsic colors predicted via an XGBoost model. The model is trained on minimally reddened stars to infer intrinsic colors and their associated uncertainties, using APOGEE stellar parameters (Teff, logg, [Fe/H], and [alpha/Fe]). The derived reddening values are then converted into extinctions using an anchor ratio of A_BP / A_RP = 1.694 +/- 0.004, derived from red-clump-like stars. Here, we provide extinction measurements in 39 filters across 10 photometric systems and introduce a new empirical extinction curve optimized for broadband passbands. Our extinction estimates (Av) outperform existing results (Bayestar19, StarHorse, SEDEX), achieving a typical precision of 0.03 mag in Av. Notably, we identify systematic deviations of up to 30% between monochromatic and passband-integrated extinction ratios at wavelengths greater than 700 nm. This result highlights the necessity of adopting passband-specific coefficients when correcting extinction to derive stellar parameters. As the foundation for a forthcoming series of papers, these benchmark measurements will be used to (1) revise asteroseismic scaling relations, (2) calibrate differential reddening in open clusters, and (3) reconcile heterogeneous dust maps into a unified, all-sky extinction scheme.

We present a statistical study on the origins of the UV continuum and narrow/broad emission lines in little red dots (LRDs), a newly discovered class of active galactic nuclei (AGNs). Leveraging all archived JWST/NIRSpec data, we build a sample of 28 spectroscopically-confirmed LRDs at $5<z_{\rm spec}<7.2$, by requiring broad H$α$ emission, blue UV colors, V-shaped continua, and compact morphologies. We define a control sample of 9 blue, compact, broad-line AGNs without red optical continua (hereafter little blue dots; LBDs), and examine correlations between rest UV and the narrow/broad H$α$ luminosities in these populations. In LRDs, both narrow and broad H$α$ components are tightly correlated with the UV continuum, and the luminosity ratios are consistent with those in young starburst galaxies. In contrast, the UV to broad H$α$ ratios in LBDs closely match local unobscured AGNs and are statistically different from LRDs. The Ly$α$ occurrence rates and strengths do not differ between LRDs and LBDs and are comparable to normal star-forming galaxies. These results are consistent with a scenario where the central BH in LRDs is enshrouded by a dense opaque gas envelope -- in this model, the UV continuum as well as narrow and even broad H$α$ emissions are not powered by AGNs but predominantly by young massive stars surrounding the envelope, while the envelope radiates as a $\sim 5000$ K blackbody. As the envelope dissipates, direct AGN emission can emerge, potentially transforming LRDs into LBDs and marking the end of a short-lived phase of rapid black hole growth.

A substantial fraction of metal-poor stars in the local Milky Way halo exhibit large overabundances of carbon. These stars, dubbed Carbon-Enhanced Metal-Poor (CEMP) stars, provide crucial constraints on the nature of the early universe including the earliest nucleosynthetic events. Whether these stars exist at similar rates in nearby galaxies is a major open question with implications for the environmental dependence of early chemical evolution. Here, we present the discovery of the first five CEMP stars in the Milky Way's largest dwarf companion, the LMC, using SDSS-V spectra from the BOSS instrument. We measure metallicities ranging from [Fe/H] = -2.1 to -3.2 and evolutionary state corrected carbon enhancements of [C/Fe] = +1.2 to +2.4, placing these stars among the most metal-poor and carbon-rich ever identified in the LMC. This discovery demonstrates that CEMP stars do exist in the LMC despite previous null detections, and establishes the foundation for measuring the CEMP occurrence rate in this system. Such measurements will provide critical tests of whether environmental differences affect the formation channels and frequencies of these ancient, carbon-rich stars.

Bridging the spatiotemporal scales of magnetic seed field generation and subsequent dynamo amplification in the weakly collisional intracluster medium presents an extreme numerical challenge. We perform collisionless turbulence simulations with initially unmagnetized electrons that capture both magnetic seed generation via the electron Weibel instability and the ensuing dynamo amplification. Going beyond existing pair-plasma studies, we use an ion-to-electron mass ratio of 100 for which we find electron and ion dynamics are sufficiently decoupled. These simulations are enabled by the 10-moment collisionless fluid solver of Gkeyll, which evolves the full pressure tensor for all species. The electron heat-flux closure regulates pressure isotropization and effectively sets the magnetic Reynolds number. We investigate how the strength of of the closure influences the transition between a regime reminiscent of previous kinetic pair-plasma simulations and a more MHD-like dynamo regime.

Long Period Transients (LPTs) are a recently identified class of sources characterized by periodic radio bursts lasting seconds to minutes, with flux densities that might reach several tens of Jy. These radio bursts repeat with periodicity from minutes to hours, and they exhibit strong polarization and transient activity periods. To date, about 12 such sources have been identified, which might encompass the same or different physical scenarios. Proposed explanations include binary systems with a white dwarf and a low-mass star companion, slow-spinning magnetars, highly magnetized isolated white dwarfs, and other exotic objects. In a few cases the optical counterpart indeed points toward a white dwarf with a low-mass companion, while in other cases, transient X-ray emission was detected, very common in magnetars. However, despite being able to reproduce partially some of the characteristics of LPTs, all the proposed scenarios find difficulty in explaining the exact physical origin of their bright, highly polarized and periodic radio emission. We review here the state-of-the-art in the observations and interpretation of this puzzling class of radio transients.

We investigate the correlation between ionized gas outflows, jets, and star formation in a sample of 42 local type-I active galactic nuclei (AGNs) exhibiting significant [O III] outflows. This study uses both new submillimeter (sub-mm) observations and archival data from the James Clerk Maxwell Telescope. Our analysis, which includes a correction for jet emission in the sub-mm bands, fitting spectral energy distribution, and analyzing spectra, enables us to derive star-formation rates (SFRs) through various methods. By comparing radio power and SFRs, we select a sub-sample of jetted AGNs of which radio emission is mostly from the jets. We find that jetted AGNs predominantly lie above the main sequence of star-forming galaxies, suggesting a correlation between jet activity and star formation. By comparing dust extinction, we demonstrate that jetted AGNs do not have more dust which is the fuel of both star formation and AGN activity. Therefore, this correlation is more likely to arise from AGN feedback. We also find that the Eddington ratio does not impact the specific SFRs (sSFRs) of our sample. Additionally, for jetted AGNs, stronger radio emission corresponds to higher sSFRs, suggesting that jet emission may promote star formation, i.e., positive feedback. Our results not only shed light on the feedback mechanisms of AGNs but also underscore the complex interplay between black hole activity and star formation in galaxy evolution.

The bending of radio galaxies in galaxy clusters is expected to be caused by interactions with the local environment. The physical processes responsible for jet bending, and their influence on the polarization properties of radio galaxies, remain poorly understood, leading to the question of whether jet properties in bent radio galaxies differ from those in linear radio galaxies. Using a sample of 24 polarized bent radio galaxies, observed with the Karl G. Jansky Very Large Array at 1--2 GHz, we test for correlation of bending angle with polarization parameters measuring Faraday rotation, intrinsic fractional polarization, and Faraday rotation dispersion, used here as a measure of turbulence along the line of sight. We find no statistically significant correlations. At the spatial resolution of our dataset (3--46 kpc, median 18.4 kpc), our results indicate that we are primarily probing larger-scale intracluster medium effects not related to bending angle. The absence of a statistically significant correlation suggests that bent radio galaxies are reliable probes of intracluster magnetic fields, because their intrinsic properties do not appear to introduce systematic biases into measured polarization parameters. We do detect a preference for source magnetic field vectors to align with the direction of jet bending. Finally, we estimate that the POSSUM and SKA surveys will contain $\gtrsim$300 and $\gtrsim$1000 polarized radio galaxies, respectively, providing large future samples with a range of bending angles and similar redshift distribution and number of beams per source as in our sample, enabling our results to be tested with greater statistical power.

Low--luminosity gamma-ray bursts (LLGRBs) are promising candidates for high-energy neutrinos, yet no coincident neutrino events have been detected so far. Recent advances in X-ray time-domain astronomy, together with the development of next-generation neutrino telescopes, open new opportunities for joint X-ray and neutrino observations of these transients. We calculate the jet dynamical evolution and the associated neutrino production for both non-magnetized and magnetized outflows. For individual events, joint X-ray and neutrino detection is generally limited to nearby LLGRBs or sources with high luminosities. Thus, we consider a next-generation neutrino telescope with an effective area enhanced by a factor of $\sim30$ relative to IceCube. In the non-magnetized scenario, joint detection of individual events is enabled for sources with typical isotropic luminosities of $L_{\mathrm{iso}}\sim10^{47}\,\mathrm{erg\,s^{-1}}$ out to luminosity distances of $D_L\sim1.6\times10^{2}\,\mathrm{Mpc}$, corresponding to an expected detection rate of order $1$ per year. In contrast, for the magnetized scenario at the same luminosity, the accessible distance is significantly reduced, with joint observations confined to sources within $D_L\sim6.5\times10^{1}\,\mathrm{Mpc}$ and an expected detection rate of order $0.5$ per year. For stacked samples of $\sim100$ magnetized LLGRBs, stacking substantially enlarges the accessible distance range, enabling joint observations for sources with representative luminosities of $L_{\mathrm{iso}}\sim1\times10^{47}\,\mathrm{erg\,s^{-1}}$ out to $D_L\lesssim7.0\times10^{2}\,\mathrm{Mpc}$ and corresponding to an expected detection rate of order $2$ per year. These results demonstrate that joint X-ray and next-generation neutrino observations enable a practical multimessenger probe of LLGRBs.

We present the implementation of a Self-Calibration of Intrinsic Alignments of galaxies as an extension to the Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST) Dark Energy Science Collaboration (DESC)'s weak lensing 3x2pt pipeline (TXPipe). As a demonstration, we have run this pipeline on the Dark Energy Survey (DES) year one data set. We find indications of a non-zero intrinsic alignment signal in the higher redshift bins, while in the lower bins our results look more uncertain. We believe this is caused by known issues with the individual galaxies photo-z estimation. This effect is particularly harmful for the self-calibration method, since it has high requirements for reliable estimation of the photo-$z$s, and the need for individual galaxy point estimates and tomographic binning to match. We show how different methods of recreating the redshift probability distribution can affect the detection of intrinsic alignment.

The 21-cm global signal, a treasure trove of information about the nature of the first luminous sources of the Universe, has traditionally been modelled assuming that these early sources were predominantly star-forming galaxies. However, recent observations by the James Webb Space Telescope (JWST) have revealed several AGNs as early as z ~ 10 - 10.4 . In light of this, it is important to investigate the contribution of such AGNs to the 21-cm signal. Assuming that these AGNs are seeded by Primordial Black Holes (PBHs) and employing an analytical PBH model, consistent with existing cosmological and astrophysical constraints, we show that these exotic objects can have a significant impact on the redshift evolution of the global signal.

To date, Galactic Astronomy has largely concerned itself with astrophysical processes, and with the locations, space motions and compositions of objects. Consider, for example, the elucidation of the components of the Galaxy over the past decades, its mapping as enabled by Gaia and its predecessors, the photometric and spectroscopic characterization of innumerable astrophysical objects in various wavelength ranges, both from the ground and from space, and the expanding discovery and characterization of exoplanets; all focused on the current, static Galaxy. This White Paper proposes a dedicated program to derive stellar ages from time-series spectroscopy to hasten the transformation of this static conception into a dynamical one with age-labeled objects and events.

Building on the transformative success of optical redshift surveys, the emerging technique of neutral hydrogen (HI) intensity mapping (IM) offers a novel probe of large-scale structure (LSS) growth and the late-time accelerated expansion of the universe. We present cosmological forecasts for the Baryon Acoustic Oscillations from Integrated Neutral Gas Observations (BINGO), a pioneering HI IM experiment, quantifying its potential to constrain the \textit{Planck}-calibrated $Λ$CDM cosmology and extensions to the $w_0w_a$CDM dark energy model. For BINGO's Phase~1 configuration, we simulate the HI IM signal using a lognormal model and incorporate three dominant systematics: foreground residuals, thermal noise, and beam resolution effects. Using Bayesian inference, we derive joint constraints on six cosmological parameters ($Ω_b h^2$, $Ω_c h^2$, $100θ_s$, $n_s$, $\ln 10^{10} A_s$, and $τ_r$) alongside 60 HI parameters ($b_{\rm HI}^i$, $Ω_{\rm HI}^i b_{\rm HI}^i$) across 30 frequency channels. Our results demonstrate that combining BINGO with the Planck 2018 CMB dataset tightens the confidence regions of cosmological parameters to $\sim$40\% the size of those from Planck alone, significantly improving the precision of parameter estimation. Furthermore, BINGO constrains the redshift evolution of HI density and delivers competitive measurements of the dark energy equation of state parameters ($w_0$, $w_a$). These results demonstrate BINGO's potential to extract significant cosmological information from the HI distribution and provide constraints competitive with current and future cosmological surveys.

A central goal in supernova (SN) research is to identify and characterize their progenitors. However, this is very difficult due to the limited archival images with sufficient depth and spatial resolution required for direct progenitor detection and due to the circumstellar dust which often biases the estimate of their intrinsic parameters. This field will be revolutionized by Euclid and the upcoming Chinese Space Station Survey Telescope (CSST), which conduct deep, wide-field, high-resolution and multi-band imaging surveys. We analyze their detection capability by comparing the model magnitudes of red supergiant (RSG) progenitors with the detection limits under different conditions, and we estimate the annual detection rates with Monte-Carlo simulations. We explore how to recover the intrinsic properties of SN progenitors with the help of radiation transfer calculations in circumstellar dust. We find the optical and near-infrared filters of the Euclid and CSST are highly effective for detecting RSG progenitors. We predict that archival images from the completed 2 surveys will enable $\lesssim13$ (or 24) progenitor detections per year within the mass range of 8--16 (or 8--25)M_\odot, an order of magnitude higher than the current detection rate of $\sim1$ detection per year. In the presence of circumstellar dust, the emerging spectral energy distribution (SED) of the progenitor is mainly affected by the optical depth and is almost independent of dust temperature in the Euclid and CSST filters. Our mock tests demonstrate that one can derive the progenitor mass and dust optical depth simultaneously by fitting the observed SED over the 11 filters of the 2 surveys while fixing the dust temperature to a typical value. Euclid and CSST will significantly enlarge the sample of direct progenitor detections with accurate mass measurements, which is crucial to resolve the long-standing RSG problem.

Recent works have identified that rapidly rotating stars are predominantly binaries with separations of a few to a few tenths of au. This is a crucial range of separation that is often inaccessible to searches of binary stars, providing a unique opportunity to examine their statistical properties. In particular, we have performed an analysis of rapid rotators in young moving groups. We examined their fraction as a function of the stellar density of the population in which they are found. We find that there is a deficit of rapid rotators in dense clusters such as the Orion Nebula in comparison to the more diffuse parts of the Orion Complex, as intracluster interactions with neighboring stars likely dissolve binaries with intermediate separations before they had a chance to fully form. In contrast, in older populations with an age of $\sim100$ Myr, mass segregation redistributes binaries relative to single stars, thus in such older regions, rapid rotators are predominantly found in the regions of higher stellar density. This work sheds light on both the conditions that lead to the formation of binary stars and their dynamical evolution.

We report the discovery of two swallowtail image formations at $z=2.91$ and $z=6.70$ behind the galaxy cluster MS 0451.6-0305 in JWST-NIRCam imaging. We find that in both of the above lensed systems, the complex image morphology cannot be reproduced by simple fold/cusp caustics, and detailed lens modeling reveals higher-order swallowtail caustic configurations. In the $z=2.91$ lens system, a small part of the source galaxy (which itself is part of a galaxy group) containing atleast two compact knots sits inside the swallowtail caustic, producing a quadruply imaged arc. At two of the image positions of these knots, we infer point source magnifications of $\gtrsim 300$, implying lensing-corrected effective radii of $\lesssim 0.8-1.5$ pc. The $z=6.70$ system exhibits even more complex image formation. We therefore only use the most confidently identified counter-images of knots in this system as constraints in our lens modeling. The resulting model predicts magnifications $\sim20-200$ and lensing-corrected effective radii of $\lesssim 0.8-18.5$ pc for various knots. Together, these two systems represent the first example of observations of multiple swallowtail image formations in a single galaxy cluster and demonstrate the ability of swallowtail caustics to magnify individual substructures at sub-parsec scales, from intermediate redshifts to the first billion years of the Universe.

We present new measurements of galaxy luminosity functions (LFs) from JWST/NIRCam imaging over the redshift range z=4.5-9.7, using photometric catalogs from JADES and public extragalactic fields. Our analysis includes rest-frame UV and B-band LFs, as well as apparent LFs in F090W, F115W, F200W, F356W, and F444W. We present the first constraints on the rest-frame B-band LF at z~7-8 and extend existing measurements at z~5 to M(B) = -18 mag. The B-band LFs evolve more strongly with redshift than UV LFs, though both decline more gradually than predicted by simulations at z>5. No single existing simulation reproduces all observed trends, with discrepancies likely driven by assumptions about binary evolution and stellar population synthesis models. The apparent LFs in F356W and F444W show hints of a bright-end excess at all redshifts, extending to fainter magnitudes at higher redshift. While extreme emission line galaxies may partially account for it, the excess may also indicate a population of moderately red, optically bright sources - potentially dusty star-forming galaxies or obscured AGNs. Finally, we find that rest-frame B-band luminosity correlates more tightly with stellar mass than UV, making it a powerful tracer of mass assembly and reinforcing the diagnostic value of rest-frame optical LFs in uncovering the physical processes that drive early galaxy formation.