arXiv Daily Digest - 2026-04-10

The rise of multimodal large language models (MLLMs) has sparked an unprecedented wave of applications in the field of medical imaging analysis. However, as one of the earliest and most fundamental tasks integrated into this paradigm, medical image classification reveals a sobering reality: state-of-the-art medical MLLMs consistently underperform compared to traditional deep learning models, despite their overwhelming advantages in pre-training data and model parameters. This paradox prompts a critical rethinking: where exactly does the performance degradation originate? In this paper, we conduct extensive experiments on 14 open-source medical MLLMs across three representative image classification datasets. Moving beyond superficial performance benchmarking, we employ feature probing to track the information flow of visual features module-by-module and layer-by-layer throughout the entire MLLM pipeline, enabling explicit visualization of where and how classification signals are distorted, diluted, or overridden. As the first attempt to dissect classification performance degradation in medical MLLMs, our findings reveal four failure modes: 1) quality limitation in visual representation, 2) fidelity loss in connector projection, 3) comprehension deficit in LLM reasoning, and 4) misalignment of semantic mapping. Meanwhile, we introduce quantitative scores that characterize the healthiness of feature evolution, enabling principled comparisons across diverse MLLMs and datasets. Furthermore, we provide insightful discussions centered on the critical barriers that prevent current medical MLLMs from fulfilling their promised clinical potential. We hope that our work provokes rethinking within the community-highlighting that the road from high expectations to clinically deployable MLLMs remains long and winding.

Aligning Large Language Models (LLMs) with high-stakes medical standards remains a significant challenge, primarily due to the dissonance between coarse-grained preference signals and the complex, multi-dimensional nature of clinical protocols. To bridge this gap, we introduce ProMedical, a unified alignment framework grounded in fine-grained clinical criteria. We first construct ProMedical-Preference-50k, a dataset generated via a human-in-the-loop pipeline that augments medical instructions with rigorous, physician-derived rubrics. Leveraging this corpus, we propose the Explicit Criteria Injection paradigm to train a multi-dimensional reward model. Unlike traditional scalar reward models, our approach explicitly disentangles safety constraints from general proficiency, enabling precise guidance during reinforcement learning. To rigorously validate this framework, we establish ProMedical-Bench, a held-out evaluation suite anchored by double-blind expert adjudication. Empirical evaluations demonstrate that optimizing the Qwen3-8B base model via ProMedical-RM-guided GRPO yields substantial gains, improving overall accuracy by 22.3% and safety compliance by 21.7%, effectively rivaling proprietary frontier models. Furthermore, the aligned policy generalizes robustly to external benchmarks, demonstrating performance comparable to state-of-the-art models on UltraMedical. We publicly release our datasets, reward models, and benchmarks to facilitate reproducible research in safety-aware medical alignment.

Symbolic regression (SR) aims to discover mathematical expressions from data, a task traditionally tackled using Genetic Programming (GP) through combinatorial search over symbolic structures. Latent Space Optimization (LSO) methods use neural encoders to map symbolic expressions into continuous spaces, transforming the combinatorial search into continuous optimization. SNIP (Meidani et al., 2024), a contrastive pre-training model inspired by CLIP, advances LSO by introducing a multi-modal approach: aligning symbolic and numeric encoders in a shared latent space to learn the phenotype-genotype mapping, enabling optimization in the numeric space to implicitly guide symbolic search. However, this relies on fine-grained cross-modal alignment, whereas literature on similar models like CLIP reveals that such an alignment is typically coarse-grained. In this paper, we investigate whether SNIP delivers on its promise of effective bi-modal optimization for SR. Our experiments show that: (1) cross-modal alignment does not improve during optimization, even as fitness increases, and (2) the alignment learned by SNIP is too coarse to efficiently conduct principled search in the symbolic space. These findings reveal that while multi-modal LSO holds significant potential for SR, effective alignment-guided optimization remains unrealized in practice, highlighting fine-grained alignment as a critical direction for future work.

Molecular communication (MC) is a communication paradigm in which information is conveyed through the controlled release, propagation, and reception of molecules. Many envisioned healthcare applications of MC are expected to operate inside the human body. In this environment, the cardiovascular system ( CVS) acts as the physical channel, which forms a closed-loop network where particle transport is mainly governed by the combined effects of diffusion and flow. Despite the fact that physiological flows in many parts of the human body are inherently pulsatile due to the cardiac cycle, most existing models for dispersive closed-loop MC channels assume a constant flow velocity. In this paper, we present a time-variant one-dimensional (1D ) channel model for dispersive closed-loop MC systems with pulsatile flow. We derive an analytical expression for the channel impulse response (CIR ), which follows a wrapped Normal distribution with time-variant mean and variance. The obtained model reveals the cyclostationary nature of the channel and quantifies the influence of pulsation on the temporal concentration profile compared to steady-flow systems. Finally, the model is validated by three-dimensional ( 3D ) particle-based simulations (PBS s), showing excellent agreement and enabling an efficient analytical characterization of the channel.

Immunohistochemistry (IHC) is essential for assessing specific immune biomarkers like Human Epidermal growth-factor Receptor 2 (HER2) in breast cancer. However, the traditional protocols of obtaining IHC stains are resource-intensive, time-consuming, and prone to structural damages. Virtual staining has emerged as a scalable alternative, but it faces significant challenges in preserving fine-grained cellular structures while accurately translating biochemical expressions. Current state-of-the-art methods still rely on Generative Adversarial Networks (GANs) or standard convolutional U-Net diffusion models that often struggle with "structure and staining trade-offs". The generated samples are either structurally relevant but blurry, or texturally realistic but have artifacts that compromise their diagnostic use. In this paper, we introduce HistDiT, a novel latent conditional Diffusion Transformer (DiT) architecture that establishes a new benchmark for visual fidelity in virtual histological staining. The novelty introduced in this work is, a) the Dual-Stream Conditioning strategy that explicitly maintains a balance between spatial constraints via VAE-encoded latents and semantic phenotype guidance via UNI embeddings; b) the multi-objective loss function that contributes to sharper images with clear morphological structure; and c) the use of the Structural Correlation Metric (SCM) to focus on the core morphological structure for precise assessment of sample quality. Consequently, our model outperforms existing baselines, as demonstrated through rigorous quantitative and qualitative evaluations.

Retrieval-augmented generation (RAG) significantly enhances large language models (LLMs) but introduces novel security risks through external knowledge access. While existing studies cover various RAG vulnerabilities, they often conflate inherent LLM risks with those specifically introduced by RAG. In this paper, we propose that secure RAG is fundamentally about the security of the external knowledge-access pipeline. We establish an operational boundary to separate inherent LLM flaws from RAG-introduced or RAG-amplified threats. Guided by this perspective, we abstract the RAG workflow into six stages and organize the literature around three trust boundaries and four primary security surfaces, including pre-retrieval knowledge corruption, retrieval-time access manipulation, downstream context exploitation, and knowledge exfiltration. By systematically reviewing the corresponding attacks, defenses, remediation mechanisms, and evaluation benchmarks, we reveal that current defenses remain largely reactive and fragmented. Finally, we discuss these gaps and highlight future directions toward layered, boundary-aware protection across the entire knowledge-access lifecycle.

We present DMax, a new paradigm for efficient diffusion language models (dLLMs). It mitigates error accumulation in parallel decoding, enabling aggressive decoding parallelism while preserving generation quality. Unlike conventional masked dLLMs that decode through a binary mask-to-token transition, DMax reformulates decoding as a progressive self-refinement from mask embeddings to token embeddings. At the core of our approach is On-Policy Uniform Training, a novel training strategy that efficiently unifies masked and uniform dLLMs, equipping the model to recover clean tokens from both masked inputs and its own erroneous predictions. Building on this foundation, we further propose Soft Parallel Decoding. We represent each intermediate decoding state as an interpolation between the predicted token embedding and the mask embedding, enabling iterative self-revising in embedding space. Extensive experiments across a variety of benchmarks demonstrate the effectiveness of DMax. Compared with the original LLaDA-2.0-mini, our method improves TPF on GSM8K from 2.04 to 5.47 while preserving accuracy. On MBPP, it increases TPF from 2.71 to 5.86 while maintaining comparable performance. On two H200 GPUs, our model achieves an average of 1,338 TPS at batch size 1. Code is available at: https://github.com/czg1225/DMax

Chain-of-Thought (CoT) has become a cornerstone of reasoning in large language models, yet its effectiveness is constrained by the limited expressiveness of discrete token sampling. Recent latent reasoning approaches attempt to alleviate this limitation by replacing discrete tokens with soft embeddings (probability-weighted mixtures of token embeddings) or hidden states, but they commonly suffer from two issues: (1) global activation injects perturbations into high-confidence steps, impairing reasoning stability; and (2) soft embeddings quickly collapse toward the highest-probability token, limiting exploration of alternative trajectories. To address these challenges, we propose SeLaR (Selective Latent Reasoning), a lightweight and training-free framework. SeLaR introduces an entropy-gated mechanism that activates soft embeddings only at low-confidence steps, while preserving discrete decoding at high-confidence steps. Additionally, we propose an entropy-aware contrastive regularization that pushes soft embeddings away from the dominant (highest-probability) token's direction, encouraging sustained exploration of multiple latent reasoning paths. Experiments on five reasoning benchmarks demonstrate that SeLaR consistently outperforms standard CoT and state-of-the-art training-free methods.

We initiate the study of asynchronous quantum distributed systems, focusing on the case of implementing atomic quantum global operations that can be decomposed into a collection of local operations on the components of the system. A simple example of such an operation is a quantum snapshot in which the whole system is instantaneously measured. Based on the classical snapshot algorithm of Chandy and Lamport, we design a quantum distributed algorithm to implement such decomposable global operations, which we call the QGO Algorithm. The analysis of our algorithm shows that arguments based on Lamport's computational causality remain valid in the quantum world, even though, due to entanglement, causality is not manifest from the standard description of the system in terms of a (global) quantum state. Our other contributions include a formal model of quantum distributed computing, and a formal specification for the desired behavior of a global operation, which may be of interest even in classical settings (such as in the setting of randomized algorithms).

Dock-based bike-sharing systems exhibit spatial imbalances between bicycle supply and user demand, often addressed through overnight truck-based rebalancing. This work studies static overnight rebalancing under demand uncertainty modeled as a tri-objective optimization problem. The objectives minimize total travel distance, expected unmet demand, and a robustness-oriented unmet demand measure over high-demand scenarios. Route plans are evaluated via a recourse simulation that enforces truck loads and station capacity constraints across multiple demand realizations. The robustness objective supports selecting plans that reduce peak-demand service degradation. Trade-off solutions are approximated with Non-dominated Sorting Genetic Algorithm II using a permutation--partition encoding and domain-specific relocation operators, including a biased best-improvement move for station relocation. Experiments on the real Barcelona Bicing system with 460 stations show well-distributed Pareto sets and substantial contributions to the reference non-dominated set. Greedy constructive baselines mainly yield extreme solutions and are often dominated.

As AI models grow more complex, explainability is essential for building trust, yet concept-based counterfactual methods still face a trade-off between expressivity and efficiency. Representing underlying concepts as atomic sets is fast but misses relational context, whereas full graph representations are more faithful but require solving the NP-hard Graph Edit Distance (GED) problem. We propose U-CECE, a unified, model-agnostic multi-resolution framework for conceptual counterfactual explanations that adapts to data regime and compute budget. U-CECE spans three levels of expressivity: atomic concepts for broad explanations, relational sets-of-sets for simple interactions, and structural graphs for full semantic structure. At the structural level, both a precision-oriented transductive mode based on supervised Graph Neural Networks (GNNs) and a scalable inductive mode based on unsupervised graph autoencoders (GAEs) are supported. Experiments on the structurally divergent CUB and Visual Genome datasets characterize the efficiency-expressivity trade-off across levels, while human surveys and LVLM-based evaluation show that the retrieved structural counterfactuals are semantically equivalent to, and often preferred over, exact GED-based ground-truth explanations.

Action Quality Assessment (AQA) has broad applications in physical therapy, sports coaching, and competitive judging. Although Vision Language Models (VLMs) hold considerable promise for AQA, their actual performance in this domain remains largely uncharacterised. We present a comprehensive evaluation of state-of-the-art VLMs across activity domains (e.g. fitness, figure skating, diving), tasks, representations, and prompting strategies. Baseline results reveal that Gemini 3.1 Pro, Qwen3-VL and InternVL3.5 models perform only marginally above random chance, and although strategies such as incorporation of skeleton information, grounding instructions, reasoning structures and in-context learning lead to isolated gains, none is consistently effective. Analysis of prediction distributions uncovers two systematic biases: a tendency to predict correct execution regardless of visual evidence, and a sensitivity to superficial linguistic framing. Reformulating tasks contrastively to mitigate these biases yields minimal improvement, suggesting that the models' limitations go beyond these biases, pointing to a fundamental difficulty with fine-grained movement quality assessment. Our findings establish a rigorous baseline for future VLM-based AQA research and provide an actionable outline for failure modes requiring mitigation prior to reliable real-world deployment.

Software architecture documentation is essential for system comprehension, yet it is often unavailable or incomplete. While recent LLM-based techniques can generate documentation from code, they typically address local artifacts rather than producing coherent, system-level architectural descriptions. This paper presents a structured process for automatically generating system-level architectural documentation directly from GitHub repositories using Large Language Models. The process, called CIAO (Code In Architecture Out), defines an LLM-based workflow that takes a repository as input and produces system-level architectural documentation following a template derived from ISO/IEC/IEEE 42010, SEI Views \& Beyond, and the C4 model. The resulting documentation can be directly added to the target repository. We evaluated the process through a study with 22 developers, each reviewing the documentation generated for a repository they had contributed to. The evaluation shows that developers generally perceive the produced documentation as valuable, comprehensible, and broadly accurate with respect to the source code, while also highlighting limitations in diagram quality, high-level context modeling, and deployment views. We also assessed the operational cost of the process, finding that generating a complete architectural document requires only a few minutes and is inexpensive to run. Overall, the results indicate that a structured, standards-oriented approach can effectively guide LLMs in producing system-level architectural documentation that is both usable and cost-effective.

Artificial Intelligence (AI)-assisted coding environments operate within finite context windows of 128,000-1,000,000 tokens (as of early 2026), yet existing tools offer limited support for monitoring and optimizing token consumption. As developers open multiple files, model attention becomes diluted and Application Programming Interface (API) costs increase in proportion to input and output as conversation length grows. Tokalator is an open-source context-engineering toolkit that includes a VS Code extension with real-time budget monitoring and 11 slash commands; nine web-based calculators for Cobb-Douglas quality modeling, caching break-even analysis, and $O(T^2)$ conversation cost proofs; a community catalog of agents, prompts, and instruction files; an MCP server and Command Line Interface (CLI); a Python econometrics API; and a PostgreSQL-backed usage tracker. The system supports 17 Large Language Models (LLMs) across three providers (Anthropic, OpenAI, Google) and is validated by 124 unit tests. An initial deployment on the Visual Studio Marketplace recorded 313 acquisitions with a 206.02\% conversion rate as of v3.1.3. A structured survey of 50 developers across three community sessions indicated that instruction-file injection and low-relevance open tabs are among the primary invisible budget consumers in typical AI-assisted development sessions.

Large language models store not only isolated facts but also rules that support reasoning across symbolic expressions, natural language explanations, and concrete instances. Yet most model editing methods are built for fact-level knowledge, assuming that a target edit can be achieved through a localized intervention. This assumption does not hold for rule-level knowledge, where a single rule must remain consistent across multiple interdependent forms. We investigate this problem through a mechanistic study of rule-level knowledge editing. To support this study, we extend the RuleEdit benchmark from 80 to 200 manually verified rules spanning mathematics and physics. Fine-grained causal tracing reveals a form-specific organization of rule knowledge in transformer layers: formulas and descriptions are concentrated in earlier layers, while instances are more associated with middle layers. These results suggest that rule knowledge is not uniformly localized, and therefore cannot be reliably edited by a single-layer or contiguous-block intervention. Based on this insight, we propose Distributed Multi-Layer Editing (DMLE), which applies a shared early-layer update to formulas and descriptions and a separate middle-layer update to instances. While remaining competitive on standard editing metrics, DMLE achieves substantially stronger rule-level editing performance. On average, it improves instance portability and rule understanding by 13.91 and 50.19 percentage points, respectively, over the strongest baseline across GPT-J-6B, Qwen2.5-7B, Qwen2-7B, and LLaMA-3-8B. The code is available at https://github.com/Pepper66/DMLE.

Large reasoning models (LRMs) have achieved strong performance enhancement through scaling test time computation, but due to the inherent limitations of the underlying language models, they still have shortcomings in tasks that require precise computation and extensive knowledge reserves. Tool-Integrated Reasoning (TIR) has emerged as a promising paradigm that incorporates tool call and execution within the reasoning trajectory. Although recent works have released some powerful open-source TIR models, our analysis reveals that these models still suffer from critical deficiencies. We find that when the reasoning of the model conflicts with the tool results, the model tends to believe in its own reasoning. And there are cases where the tool results are correct but are ignored by the model, resulting in incorrect answers, which we define as "Tool Ignored''. This indicates that the model does not know when to trust or ignore the tool. To overcome these limitations, We introduce Adaptive Tool Trust Calibration (ATTC), a novel framework that guides the model to adaptively choose to trust or ignore the tool results based on the confidence score of generated code blocks. The experimental results from various open-source TIR models of different sizes and across multiple datasets demonstrate that ATTC effectively reduces the "Tool Ignored" issue, resulting in a performance increase of 4.1% to 7.5%.

We present a quantum-inspired ARIMA methodology that integrates quantum-assisted lag discovery with \emph{fixed-configuration} variational quantum circuits (VQCs) for parameter estimation and weak-lag refinement. Differencing and candidate lags are identified via swap-test-driven quantum autocorrelation (QACF) and quantum partial autocorrelation (QPACF), with a delayed-matrix construction that aligns quantum projections to time-domain regressors, followed by standard information-criterion parsimony. Given the screened orders $(p,d,q)$, we retain a fixed VQC ansatz, optimizer, and training budget, preventing hyperparameter leakage, and deploy the circuit in two estimation roles: VQC-AR for autoregressive coefficients and VQC-MA for moving-average coefficients. Between screening and estimation, a lightweight VQC weak-lag refinement re-weights or prunes screened AR lags without altering $(p,d,q)$. Across environmental and industrial datasets, we perform rolling-origin evaluations against automated classical ARIMA, reporting out-of-sample mean squared error (MSE), mean absolute percentage error (MAPE), and Diebold--Mariano tests on MSE and MAE. Empirically, the seven quantum contributions -- (1) differencing selection, (2) QACF, (3) QPACF, (4) swap-test primitives with delayed-matrix construction, (5) VQC-AR, (6) VQC weak-lag refinement, and (7) VQC-MA -- collectively reduce meta-optimization overhead and make explicit where quantum effects enter order discovery, lag refinement, and AR/MA parameter estimation.

As generative artificial intelligence evolves, autonomous agent networks present a powerful paradigm for interactive covert communication. However, because agents dynamically update internal memories via environmental interactions, existing methods face a critical structural vulnerability: cognitive asymmetry. Conventional approaches demand strict cognitive symmetry, requiring identical sequence prefixes between the encoder and decoder. In dynamic deployments, inevitable prefix discrepancies destroy synchronization, inducing severe channel degradation. To address this core challenge of cognitive asymmetry, we propose the Asymmetric Collaborative Framework (ACF), which structurally decouples covert communication from semantic reasoning via orthogonal statistical and cognitive layers. By deploying a prefix-independent decoding paradigm governed by a shared steganographic configuration, ACF eliminates the reliance on cognitive symmetry. Evaluations on realistic memory-augmented workflows demonstrate that under severe cognitive asymmetry, symmetric baselines suffer severe channel degradation, whereas ACF uniquely excels across both semantic fidelity and covert communication. It maintains computational indistinguishability, enabling reliable secret extraction with provable error bounds, and providing robust Effective Information Capacity guarantees for modern agent networks.

Metaphor pervades everyday language, allowing speakers to express abstract concepts via concrete domains. While prior work has studied metaphors cognitively and psycholinguistically, large-scale comparisons with literal language remain limited, especially for near-synonymous expressions. We analyze 297 English verb-object pairs (e.g., float idea vs. suggest idea) in ~2M corpus sentences, examining their contextual usage. Using five NLP tools, we extract 2,293 cognitive and linguistic features capturing affective, lexical, syntactic, and discourse-level properties. We address: (i) whether features differ between metaphorical and literal contexts (cross-pair analysis), and (ii) whether individual VO pairs diverge internally (within-pair analysis). Cross-pair results show literal contexts have higher lexical frequency, cohesion, and structural regularity, while metaphorical contexts show greater affective load, imageability, lexical diversity, and constructional specificity. Within-pair analyses reveal substantial heterogeneity, with most pairs showing non-uniform effects. These results suggest no single, consistent distributional pattern that distinguishes metaphorical from literal usage. Instead, differences are largely construction-specific. Overall, large-scale data combined with diverse features provides a fine-grained understanding of metaphor-literal contrasts in VO usage.

While numerous machine unlearning (MU) methods have recently been developed with promising results in erasing the influence of forgotten data, classes, or concepts, they are also highly vulnerable-for example, simple fine-tuning can inadvertently reintroduce erased concepts. In this paper, we address this contradiction by examining the internal representations of unlearned models, in contrast to prior work that focuses primarily on output-level behavior. Our analysis shows that many state-of-the-art MU methods appear successful mainly due to a misalignment between last-layer features and the classifier, a phenomenon we call feature-classifier misalignment. In fact, hidden features remain highly discriminative, and simple linear probing can recover near-original accuracy. Assuming neural collapse in the original model, we further demonstrate that adjusting only the classifier can achieve negligible forget accuracy while preserving retain accuracy, and we corroborate this with experiments using classifier-only fine-tuning. Motivated by these findings, we propose MU methods based on a class-mean features (CMF) classifier, which explicitly enforces alignment between features and classifiers. Experiments on standard benchmarks show that CMF-based unlearning reduces forgotten information in representations while maintaining high retain accuracy, highlighting the need for faithful representation-level evaluation of MU.

The growing use of artificial intelligence (AI) in education, particularly large language models (LLMs), has increased interest in intelligent tutoring systems. However, LLMs often show limited adaptivity and struggle to model learners' evolving knowledge over time, highlighting the need for dedicated learner modelling approaches. Although deep knowledge tracing methods achieve strong predictive performance, their opacity and susceptibility to bias can limit alignment with pedagogical principles. To address this, we propose Responsible-DKT, a neural-symbolic deep knowledge tracing approach that integrates symbolic educational knowledge (e.g., mastery and non-mastery rules) into sequential neural models for responsible learner modelling. Experiments on a real-world dataset of students' math interactions show that Responsible-DKT outperforms both a neural-symbolic baseline and a fully data-driven PyTorch DKT model across training settings. The model achieves over 0.80 AUC with only 10% of training data and up to 0.90 AUC, improving performance by up to 13%. It also demonstrates improved temporal reliability, producing lower early- and mid-sequence prediction errors and the lowest prediction inconsistency rates across sequence lengths, indicating that prediction updates remain directionally aligned with observed student responses over time. Furthermore, the neural-symbolic approach offers intrinsic interpretability via a grounded computation graph that exposes the logic behind each prediction, enabling both local and global explanations. It also allows empirical evaluation of pedagogical assumptions, revealing that repeated incorrect responses (non-mastery) strongly influence prediction updates. These results indicate that neural-symbolic approaches enhance both performance and interpretability, mitigate data limitations, and support more responsible, human-centered AI in education.

The complex and dynamic real-world clinical environment demands reliable deep learning (DL) systems. Out-of-distribution (OOD) detection plays a critical role in enhancing the reliability and generalizability of DL models when encountering data that deviate from the training distribution, such as unseen disease cases. However, existing OOD detection methods typically rely either on a single visual modality or solely on image-text matching, failing to fully leverage multimodal information. To overcome the challenge, we propose a novel dual-branch multimodal framework by introducing a text-image branch and a vision branch. Our framework fully exploits multimodal representations to identify OOD samples through these two complementary branches. After training, we compute scores from the text-image branch ($S_t$) and vision branch ($S_v$), and integrate them to obtain the final OOD score $S$ that is compared with a threshold for OOD detection. Comprehensive experiments on publicly available endoscopic image datasets demonstrate that our proposed framework is robust across diverse backbones and improves state-of-the-art performance in OOD detection by up to 24.84%

Knowledge Tracing (KT) aims to predict learners' future performance from past interactions. While recent KT approaches have improved via learning item representations aligned with Knowledge Components, they overlook the procedural dynamics of problem solving. We propose Behavior-Aware Item Modeling (BAIM), a framework that enriches item representations by integrating dynamic procedural solution information. BAIM leverages a reasoning language model to decompose each item's solution into four problem-solving stages (i.e., understand, plan, carry out, and look back), pedagogically grounded in Polya's framework. Specifically, it derives stage-level representations from per-stage embedding trajectories, capturing latent signals beyond surface features. To reflect learner heterogeneity, BAIM adaptively routes these stage-wise representations, introducing a context-conditioned mechanism within a KT backbone, allowing different procedural stages to be emphasized for different learners. Experiments on XES3G5M and NIPS34 show that BAIM consistently outperforms strong pretraining-based baselines, achieving particularly large gains under repeated learner interactions.

Long-term memory is essential for conversational agents to maintain coherence, track persistent tasks, and provide personalized interactions across extended dialogues. However, existing approaches as Retrieval-Augmented Generation (RAG) and graph-based memory mostly rely on pairwise relations, which can hardly capture high-order associations, i.e., joint dependencies among multiple elements, causing fragmented retrieval. To this end, we propose HyperMem, a hypergraph-based hierarchical memory architecture that explicitly models such associations using hyperedges. Particularly, HyperMem structures memory into three levels: topics, episodes, and facts, and groups related episodes and their facts via hyperedges, unifying scattered content into coherent units. Leveraging this structure, we design a hybrid lexical-semantic index and a coarse-to-fine retrieval strategy, supporting accurate and efficient retrieval of high-order associations. Experiments on the LoCoMo benchmark show that HyperMem achieves state-of-the-art performance with 92.73% LLM-as-a-judge accuracy, demonstrating the effectiveness of HyperMem for long-term conversations.

The essence of current neural network architectures is phenomenological fitting: they learn input-output statistical correlations via massive parameters and data, yet lack intrinsic understanding of the fundamental principles governing physical reality. This paper proposes a paradigm leap from pure phenomenological fitting to the fusion of phenomenological fitting and endogenous deduction. By embedding physical meta-principles into neural network architecture, we construct the Meta-Principle Physics Architecture (MPPA). Specifically, MPPA embeds three core meta-principles - Connectivity, Conservation, Periodicity - into its architecture, implemented via three core components: the Gravitator realizes Connectivity via standard causal attention; the Energy Encoder implements Conservation via log-domain energy tracking and delayed compensation; the Periodicity Encoder fulfills Periodicity via FFT-based spectral analysis and delayed modulation. These components collaborate via a learnable independent gating fusion mechanism, forming a complete physical cognition framework of 'local relational connectivity - global conservation constraint - evolutionary periodic law'. Experiments show MPPA achieves significant improvements: physical reasoning (from near zero to 0.436, 0.436 vs 0.000), 2.18x mathematical task improvement (0.330 vs 0.151), 52% logical task gain (0.456 vs 0.300), and 3.69% lower validation perplexity (259.45 vs 269.40), with only 11.8% more parameters (242.40M vs 216.91M). Notably, MPPA shows strong generalization on out-of-distribution physical scenarios, proving the robustness and interpretability of this principle-embedded design. This work establishes a new theoretical foundation and technical path for next-generation AI with physical common sense, causal reasoning, and mathematical rigor.

Although Large Language Models (LLMs) demonstrate remarkable reasoning capabilities, inherent social biases often cascade throughout the Chain-of-Thought (CoT) process, leading to continuous "Bias Propagation". Existing debiasing methods primarily focus on static constraints or external interventions, failing to identify and interrupt this propagation once triggered. To address this limitation, we introduce Self-Debias, a progressive framework designed to instill intrinsic self-correction capabilities. Specifically, we reformulate the debiasing process as a strategic resource redistribution problem, treating the model's output probability mass as a limited resource to be reallocated from biased heuristics to unbiased reasoning paths. Unlike standard preference optimization which applies broad penalties, Self-Debias employs a fine-grained trajectory-level objective subject to dynamic debiasing constraints. This enables the model to selectively revise biased reasoning suffixes while preserving valid contextual prefixes. Furthermore, we integrate an online self-improvement mechanism utilizing consistency filtering to autonomously synthesize supervision signals. With merely 20k annotated samples, Self-Debias activates efficient self-correction, achieving superior debiasing performance while preserving general reasoning capabilities without continuous external oversight.

Coflow provides a key application-layer abstraction for capturing communication patterns, enabling the efficient coordination of parallel data flows to reduce job completion times in distributed systems. Modern data center networks (DCNs) are employing multiple independent optical circuit switching (OCS) cores operating concurrently to meet the massive bandwidth demands of application jobs. However, existing coflow scheduling research primarily focuses on the single-core setting, with multi-core fabrics only for EPS (electrical packet switching) networks. To address this gap, this paper studies the coflow scheduling problem in multi-core OCS networks under the not-all-stop reconfiguration model in which one circuit's reconfiguration does not interrupt other circuits. The challenges stem from two aspects: (i) cross-core coupling induced by traffic assignment across heterogeneous cores; and (ii) per-core OCS scheduling constraints, namely port exclusivity and reconfiguration delay. We propose an approximation algorithm that jointly integrates cross-core flow assignment and per-core circuit scheduling to minimize the total weighted coflow completion time (CCT) and establish a provable worst-case performance guarantee. Furthermore, our algorithm framework can be directly applied to the multi-core EPS scenario with the corresponding approximation ratio under packet-switched fabrics. Trace-driven simulations using real Facebook workloads demonstrate that our algorithm effectively reduces weighted CCT and tail CCT.

Embodied navigation agents built upon large reasoning models (LRMs) can handle complex, multimodal environmental input and perform grounded reasoning per step to improve sequential decision-making for long-horizon tasks. However, a critical question remains: \textit{how can the reasoning capabilities of LRMs be harnessed intelligently and efficiently for long-horizon navigation tasks?} In simple scenes, agents are expected to act reflexively, while in complex ones they should engage in deliberate reasoning before acting.To achieve this, we introduce \textbf{H}ybr\textbf{i}d \textbf{R}eas\textbf{O}ning \textbf{Nav}igation (\textbf{HiRO-Nav}) agent, the first kind of agent capable of adaptively determining whether to perform thinking at every step based on its own action entropy. Specifically, by examining how the agent's action entropy evolves over the navigation trajectories, we observed that only a small fraction of actions exhibit high entropy, and these actions often steer the agent toward novel scenes or critical objects. Furthermore, studying the relationship between action entropy and task completion (i.e., Q-value) reveals that improving high-entropy actions contributes more positively to task success.Hence, we propose a tailored training pipeline comprising hybrid supervised fine-tuning as a cold start, followed by online reinforcement learning with the proposed hybrid reasoning strategy to explicitly activate reasoning only for high-entropy actions, significantly reducing computational overhead while improving decision quality. Extensive experiments on the \textsc{CHORES}-$\mathbb{S}$ ObjectNav benchmark showcases that HiRO-Nav achieves a better trade-off between success rates and token efficiency than both dense-thinking and no-thinking baselines.

The competency of any intelligent agent is bounded by its formal account of the world in which it operates. Clinical AI lacks such an account. Existing frameworks address evaluation, regulation, or system design in isolation, without a shared model of the clinical world to connect them. We introduce the Clinical World Model, a framework that formalizes care as a tripartite interaction among Patient, Provider, and Ecosystem. To formalize how any agent, whether human or artificial, transforms information into clinical action, we develop parallel decision-making architectures for providers, patients, and AI agents, grounded in validated principles of clinical cognition. The Clinical AI Skill-Mix operationalizes competency through eight dimensions. Five define the clinical competency space (condition, phase, care setting, provider role, and task) and three specify how AI engages human reasoning (assigned authority, agent facing, and anchoring layer). The combinatorial product of these dimensions yields a space of billions of distinct competency coordinates. A central structural implication is that validation within one coordinate provides minimal evidence for performance in another, rendering the competency space irreducible. The framework supplies a common grammar through which clinical AI can be specified, evaluated, and bounded across stakeholders. By making this structure explicit, the Clinical World Model reframes the field's central question from whether AI works to in which competency coordinates reliability has been demonstrated, and for whom.

Large language model (LLM) agents are increasingly built less by changing model weights than by reorganizing the runtime around them. Capabilities that earlier systems expected the model to recover internally are now externalized into memory stores, reusable skills, interaction protocols, and the surrounding harness that makes these modules reliable in practice. This paper reviews that shift through the lens of externalization. Drawing on the idea of cognitive artifacts, we argue that agent infrastructure matters not merely because it adds auxiliary components, but because it transforms hard cognitive burdens into forms that the model can solve more reliably. Under this view, memory externalizes state across time, skills externalize procedural expertise, protocols externalize interaction structure, and harness engineering serves as the unification layer that coordinates them into governed execution. We trace a historical progression from weights to context to harness, analyze memory, skills, and protocols as three distinct but coupled forms of externalization, and examine how they interact inside a larger agent system. We further discuss the trade-off between parametric and externalized capability, identify emerging directions such as self-evolving harnesses and shared agent infrastructure, and discuss open challenges in evaluation, governance, and the long-term co-evolution of models and external infrastructure. The result is a systems-level framework for explaining why practical agent progress increasingly depends not only on stronger models, but on better external cognitive infrastructure.

Large Language Models (LLMs) still suffer from severe hallucinations and catastrophic forgetting during causal reasoning over massive, fragmented long contexts. Existing memory mechanisms typically treat retrieval as a static, single-step passive matching process, leading to severe semantic dilution and contextual fragmentation. To overcome these fundamental bottlenecks, we propose MemCoT, a test-time memory scaling framework that redefines the reasoning process by transforming long-context reasoning into an iterative, stateful information search. MemCoT introduces a multi-view long-term memory perception module that enables Zoom-In evidence localization and Zoom-Out contextual expansion, allowing the model to first identify where relevant evidence resides and then reconstruct the surrounding causal structure necessary for reasoning. In addition, MemCoT employs a task-conditioned dual short-term memory system composed of semantic state memory and episodic trajectory memory. This short-term memory records historical search decisions and dynamically guides query decomposition and pruning across iterations. Empirical evaluations demonstrate that MemCoT establishes a state-of-the-art performance. Empowered by MemCoT, several open- and closed-source models achieve SOTA performance on the LoCoMo benchmark and LongMemEval-S benchmark.

High-quality training triplets (source-target image pairs with precise editing instructions) are a critical bottleneck for scaling instruction-guided image editing models. Vision-language models (VLMs) are widely used for automated instruction synthesis, but we identify three systematic failure modes in image-pair settings: orientation inconsistency (e.g., left/right confusion), viewpoint ambiguity, and insufficient fine-grained attribute description. Human evaluation shows that over 47% of instructions from strong baseline VLMs contain critical errors unusable for downstream training. We propose EditCaption, a scalable two-stage post-training pipeline for VLM-based instruction synthesis. Stage 1 builds a 100K supervised fine-tuning (SFT) dataset by combining GLM automatic annotation, EditScore-based filtering, and human refinement for spatial, directional, and attribute-level accuracy. Stage 2 collects 10K human preference pairs targeting the three failure modes and applies direct preference optimization (DPO) for alignment beyond SFT alone. On Eval-400, ByteMorph-Bench, and HQ-Edit, fine-tuned Qwen3-VL models outperform open-source baselines; the 235B model reaches 4.712 on Eval-400 (vs. Gemini-3-Pro 4.706, GPT-4.1 4.220, Kimi-K2.5 4.111) and 4.588 on ByteMorph-Bench (vs. Gemini-3-Pro 4.522, GPT-4.1 3.412). Human evaluation shows critical errors falling from 47.75% to 23% and correctness rising from 41.75% to 66%. The work offers a practical path to scalable, human-aligned instruction synthesis for image editing data.

Context: Traceability is a key quality attribute of artifacts that are used in knowledge-intensive tasks and supports software engineers in producing higher-quality software. Despite its clear benefits, traceability is often neglected in practice due to challenges such as granularity of traces, lack of a common artifact structure, and unclear responsibility. The Taxonomic Trace Links (TTL) approach connects source and target artifacts through a domain-specific taxonomy, aiming to address these common traceability challenges. Objective: In this study, we empirically evaluate TTL in an industrial setting to identify its strengths and weaknesses for real-world adoption. Method: We conducted a mixed-methods study at Ericsson involving one of its software products. Quantitative and qualitative data were collected across two traceability use cases. We established trace links between 463 business use cases, 64 test cases, and 277 ISO-standard requirements. Additionally, we held three focus group sessions with practitioners. Results: We identified two practically relevant scenarios where traceability is required and evaluated TTL in each. Overall, practitioners found TTL to be a useful solution for one of the scenarios, while less useful for the other. However, developing a domain-specific taxonomy and managing heterogeneous artifact structures were noted as significant challenges. Moreover, the precision of the classifier that is used to create trace links needs to be improved to make the solution practical. Conclusion: TTL is a promising approach that can be adopted in practice and enables traceability use cases. However, TTL is not a replacement for traditional trace links, but rather complements them to enable more traceability use cases and encourage the early creation of trace links.

Modern Large Language Models (LLMs) operate fundamentally as Bounded-Input Bounded-Output (BIBO) systems. They remain in a passive state until explicitly prompted, computing localized responses without intrinsic temporal continuity. While effective for isolated tasks, this reactive paradigm presents a critical bottleneck for engineering autonomous artificial intelligence. Current multi-agent frameworks attempt to distribute cognitive load but frequently rely on static memory pools and passive message passing, which inevitably leads to cognitive stagnation and homogeneous deadlocks during extended execution. To address this structural limitation, we propose Global Workspace Agents (GWA), a cognitive architecture inspired by Global Workspace Theory. GWA transitions multi-agent coordination from a passive data structure to an active, event-driven discrete dynamical system. By coupling a central broadcast hub with a heterogeneous swarm of functionally constrained agents, the system maintains a continuous cognitive cycle. Furthermore, we introduce an entropy-based intrinsic drive mechanism that mathematically quantifies semantic diversity, dynamically regulating generation temperature to autonomously break reasoning deadlocks. Coupled with a dual-layer memory bifurcation strategy to ensure long-term cognitive continuity, GWA provides a robust, reproducible engineering framework for sustained, self-directed LLM agency.

For applications on the extreme edge, minimal networks of only a few dozen artificial neurons for event detection and classification in discrete time signals would be highly desirable. Feed-forward networks, RNNs, and CNNs evolved through evolutionary algorithms can all be successful in this respect but pose the problem of allowing little systematicity in mutation and recombination if the standard direct genetic encoding of the weights is used (as for instance in the classic NEAT algorithm). We therefore introduce Echo Networks, a type of recurrent network that consists of the connection matrix only, with the source neurons of the synapses represented as rows, destination neurons as columns and weights as entries. There are no layers, and connections between neurons can be bidirectional but are technically all recurrent. Input and output can be arbitrarily assigned to any of the neurons and only use an additional (optional) function in their computational path, e.g., a sigmoid to obtain a binary classification output. We evaluated Echo Networks successfully on the classification of electrocardiography signals but see the most promising potential in their genome representation as a single matrix, allowing matrix computations and factorisations as mutation and recombination operators.

Medical Vision-Language Models (VLMs) hold immense promise for complex clinical tasks, but their reasoning capabilities are often constrained by text-only paradigms that fail to ground inferences in visual evidence. This limitation not only curtails performance on tasks requiring fine-grained visual analysis but also introduces risks of visual hallucination in safety-critical applications. Thus, we introduce MedVR, a novel reinforcement learning framework that enables annotation-free visual reasoning for medical VLMs. Its core innovation lies in two synergistic mechanisms: Entropy-guided Visual Regrounding (EVR) uses model uncertainty to direct exploration, while Consensus-based Credit Assignment (CCA) distills pseudo-supervision from rollout agreement. Without any human annotations for intermediate steps, MedVR achieves state-of-the-art performance on diverse public medical VQA benchmarks, significantly outperforming existing models. By learning to reason directly with visual evidence, MedVR promotes the robustness and transparency essential for accelerating the clinical deployment of medical AI.

Controlled experiments are a core research method in software engineering (SE) for validating causal claims. However, recruiting a sample of participants that represents the intended target population is often difficult or expensive, which limits the external validity of experimental results. At the same time, SE researchers often have access to much larger amounts of observational than experimental data (e.g., from repositories, issue trackers, logs, surveys and industrial processes). Transportability methods combine these data from experimental and observational studies to "transport" results from the experimental sample to a broader, more representative sample of the target population. Although the ability to combine observational and experimental data in a principled way could substantially benefit empirical SE research, transportability methods have - to our knowledge - not been adopted in SE. In this vision, we aim to help make that adoption possible. To that end, we introduce transportability methods, their prerequisites, and demonstrate their potential through a simulation. We then outline several SE research scenarios in which these methods could apply, e.g., how to effectively use students as substitutes for developers. Finally, we outline a road map and practical guidelines to support SE researchers in applying them. Adopting transportability methods in SE research can strengthen the external validity of controlled experiments and help the field produce results that are both more reliable and more useful in practice.

The Maxey-Riley-Gatignol equations (MaRGE) model the motion of spherical inertial particles in a fluid. They contain the Basset force, an integral term which models history effects due to the formation of wakes and boundary layer effects. This causes the force that acts on a particle to depend on its past trajectory and complicates the numerical solution of MaRGE. Therefore, the Basset force is often neglected, despite substantial evidence that it has both quantitative and qualitative impact on the movement patterns of modelled particles. Using the concept of universal differential equations, we propose an approximation of the history term via neural networks which approximates MaRGE by a system of ordinary differential equations that can be solved with standard numerical solvers like Runge-Kutta methods.

Reliable generalization metrics are fundamental to the evaluation of machine learning models. Especially in high-stakes applications where labeled target data are scarce, evaluation of models' generalization performance under distribution shift is a pressing need. We focus on two practical scenarios: (1) Before deployment, how to select the best model for unlabeled target data? (2) After deployment, how to monitor model performance under distribution shift? The central need in both cases is a reliable and label-free proxy metric. Yet existing proxy metrics, such as model confidence or accuracy-on-the-line, are often unreliable as they only assess model output while ignoring the internal mechanisms that produce them. We address this limitation by introducing a new perspective: using the inner workings of a model, i.e., circuits, as a predictive metric of generalization performance. Leveraging circuit discovery, we extract the causal interactions between internal representations as a circuit, from which we derive two metrics tailored to the two practical scenarios. (1) Before deployment, we introduce Dependency Depth Bias, which measures different models' generalization capability on target data. (2) After deployment, we propose Circuit Shift Score, which predicts a model's generalization under different distribution shifts. Across various tasks, both metrics demonstrate significantly improved correlation with generalization performance, outperforming existing proxies by an average of 13.4\% and 34.1\%, respectively. Our code is available at https://github.com/deep-real/GenCircuit.

Graph-structured data jointly contain discrete topology and continuous geometry, which poses fundamental challenges for generative modeling due to heterogeneous distributions, incompatible noise dynamics, and the need for equivariant inductive biases. Existing flow-matching approaches for graph generation typically decouple structure from geometry, lack synchronized cross-domain dynamics, and rely on iterative sampling, often resulting in physically inconsistent molecular conformations and slow sampling. To address these limitations, we propose Equivariant MeanFlow (EQUIMF), a unified SE(3)-equivariant generative framework that jointly models discrete and continuous components through synchronized MeanFlow dynamics. EQUIMF introduces a unified time bridge and average-velocity updates with mutual conditioning between structure and geometry, enabling efficient few-step generation while preserving physical consistency. Moreover, we develop a novel discrete MeanFlow formulation with a simple yet effective parameterization to support efficient generation over discrete graph structures. Extensive experiments demonstrate that EQUIMF consistently outperforms prior diffusion and flow-matching methods in generation quality, physical validity, and sampling efficiency.

The rapid advancement of Audio Large Language Models (ALLMs) has enabled cost-effective, high-fidelity generation and manipulation of both speech and non-speech audio, including sound effects, singing voices, and music. While these capabilities foster creativity and content production, they also introduce significant security and trust challenges, as realistic audio deepfakes can now be generated and disseminated at scale. Existing audio deepfake detection (ADD) countermeasures (CMs) and benchmarks, however, remain largely speech-centric, often relying on speech-specific artifacts and exhibiting limited robustness to real-world distortions, as well as restricted generalization to heterogeneous audio types and emerging spoofing techniques. To address these gaps, we propose the All-Type Audio Deepfake Detection (AT-ADD) Grand Challenge for ACM Multimedia 2026, designed to bridge controlled academic evaluation with practical multimedia forensics. AT-ADD comprises two tracks: (1) Robust Speech Deepfake Detection, which evaluates detectors under real-world scenarios and against unseen, state-of-the-art speech generation methods; and (2) All-Type Audio Deepfake Detection, which extends detection beyond speech to diverse, unknown audio types and promotes type-agnostic generalization across speech, sound, singing, and music. By providing standardized datasets, rigorous evaluation protocols, and reproducible baselines, AT-ADD aims to accelerate the development of robust and generalizable audio forensic technologies, supporting secure communication, reliable media verification, and responsible governance in an era of pervasive synthetic audio.

The escalating computational demands and energy footprint of GPU-accelerated computing systems complicate informed design and operational decisions. We present the first release of Wattlytics (https://wattlytics.netlify.app), an interactive, browser-based decision-support system. Unlike existing procurement-oriented calculators, Wattlytics uniquely integrates benchmark-driven GPU performance scaling, dynamic voltage and frequency scaling (DVFS)-aware piecewise power modeling, and multi-year total cost of ownership (TCO) analysis within a single interactive environment. Users can configure heterogeneous systems across contemporary GPU architectures (GH200, H100, L40S, L40, A40, A100, and L4), select representative scientific workloads (e.g., GROMACS, AMBER), and explore deployment scenarios under constraints such as energy prices, system lifetime, and frequency scaling. Wattlytics computes multidimensional decision metrics (TCO breakdown, work-per-TCO, power-per-TCO, and work-per-watt-per-TCO) and supports design-space exploration, what-if scenarios, sensitivity metrics (elasticity, Sobol indices, Monte Carlo) and collaborative features to guide realistic cluster design and procurement under uncertainty. We demonstrate selected scenarios comparing deployment strategies under different operational modes: ixed budget, fixed GPU count, fixed performance, and fixed power. Our case studies show that, under budget or energy constraints, optimally deployed energy-efficient GPUs can outperform higher-performance alternatives in overall cost-effectiveness. Wattlytics helps users explore the design parameter space and distinguish between cost- and risk-driving factors, turning HPC design into a well-informed and explainable decision-making process.

Modern transformer-based sequential recommenders excel at capturing short-term intent but often suffer from recency bias, overlooking stable long-term preferences. While extending sequence lengths is an intuitive fix, it is computationally inefficient, and recent interactions tend to dominate the model's attention. We propose Long-Term Embeddings (LTE) as a high-inertia contextual anchor to bridge this gap. We address a critical production challenge: the point-in-time consistency problem caused by infrastructure constraints, as feature stores typically host only a single "live" version of features. This leads to an offline-online mismatch during model deployments and rollbacks, as models are forced to process evolved representations they never saw during training. To resolve this, we introduce an LTE framework that constrains embeddings to a fixed semantic basis of content-based item representations, ensuring cross-version compatibility. Furthermore, we investigate integration strategies for causal language modeling, considering the data leakage issue that occurs when the LTE and the transformer's short-term sequence share a temporal horizon. We evaluate two representations: a heuristic average and an asymmetric autoencoder with a fixed decoder grounded in the semantic basis to enable behavioral fine-tuning while maintaining stability. Online A/B tests on Zalando demonstrate that integrating LTE as a contextual prefix token using a lagged window yields significant uplifts in both user engagement and financial metrics.

In classical Reinforcement Learning from Human Feedback (RLHF), Reward Models (RMs) serve as the fundamental signal provider for model alignment. As Large Language Models evolve into agentic systems capable of autonomous tool invocation and complex reasoning, the paradigm of reward modeling faces unprecedented challenges--most notably, the lack of benchmarks specifically designed to assess RM capabilities within tool-integrated environments. To address this gap, we present Plan-RewardBench, a trajectory-level preference benchmark designed to evaluate how well judges distinguish preferred versus distractor agent trajectories in complex tool-using scenarios. Plan-RewardBench covers four representative task families -- (i) Safety Refusal, (ii) Tool-Irrelevance / Unavailability, (iii) Complex Planning, and (iv) Robust Error Recovery -- comprising validated positive trajectories and confusable hard negatives constructed via multi-model natural rollouts, rule-based perturbations, and minimal-edit LLM perturbations. We benchmark representative RMs (generative, discriminative, and LLM-as-Judge) under a unified pairwise protocol, reporting accuracy trends across varying trajectory lengths and task categories. Furthermore, we provide diagnostic analyses of prevalent failure modes. Our results reveal that all three evaluator families face substantial challenges, with performance degrading sharply on long-horizon trajectories, underscoring the necessity for specialized training in agentic, trajectory-level reward modeling. Ultimately, Plan-RewardBench aims to serve as both a practical evaluation suite and a reusable blueprint for constructing agentic planning preference data.

Offline multi-agent reinforcement learning (MARL) aims to learn the optimal joint policy from pre-collected datasets, requiring a trade-off between maximizing global returns and mitigating distribution shift from offline data. Recent studies use diffusion or flow generative models to capture complex joint policy behaviors among agents; however, they typically rely on multi-step iterative sampling, thereby reducing training and inference efficiency. Although further research improves sampling efficiency through methods like distillation, it remains sensitive to the behavior regularization coefficient. To address the above-mentioned issues, we propose Value Guidance Multi-agent MeanFlow Policy (VGM$^2$P), a simple yet effective flow-based policy learning framework that enables efficient action generation with coefficient-insensitive conditional behavior cloning. Specifically, VGM$^2$P uses global advantage values to guide agent collaboration, treating optimal policy learning as conditional behavior cloning. Additionally, to improve policy expressiveness and inference efficiency in multi-agent scenarios, it leverages classifier-free guidance MeanFlow for both policy training and execution. Experiments on tasks with both discrete and continuous action spaces demonstrate that, even when trained solely via conditional behavior cloning, VGM$^2$P efficiently achieves performance comparable to state-of-the-art methods.

Benchmark problems are an important tool for gaining understanding of optimization algorithms. Since algorithms often aim to perform well on benchmarks, biases in benchmark design provide misleading insights. In single-objective optimization, for example, many problems used to have their optimum in the center of the search domain. To remedy these issues, search space transformations have been widely adopted by benchmark suites, preventing algorithms from exploiting unintended structure. In multi-objective optimization, problem design has focused primarily on the objective space structure. While this focus addresses important aspects of the multi-objective nature of the problems, the search space structures of these problems have received comparatively limited attention. In this work, we re-emphasize the importance of transformations in the search space, and address the challenges inherent in adding transformations to boundary constraints problems without impacting the structure of the objective space. We utilized two parameterized, bijective transformations to create different instantiations of popular benchmark problems, and show how these changes impact the performance of various multi-objective optimization algorithms. In addition to the search space transformations, we show that such parameterized transformations can also be applied to the objective space, and compare their respective performance impacts.

Accurate ocean mapping is essential for applications such as bathymetry estimation, seabed characterization, marine litter detection, and ecosystem monitoring. However, ocean remote sensing (RS) remains constrained by limited labeled data and by the reduced transferability of models pre-trained mainly on land-dominated Earth observation imagery. In this paper, we propose OceanMAE, an ocean-specific masked autoencoder that extends standard MAE pre-training by integrating multispectral Sentinel-2 observations with physically meaningful ocean descriptors during self-supervised learning. By incorporating these auxiliary ocean features, OceanMAE is designed to learn more informative and ocean-aware latent representations from large- scale unlabeled data. To transfer these representations to downstream applications, we further employ a modified UNet-based framework for marine segmentation and bathymetry estimation. Pre-trained on the Hydro dataset, OceanMAE is evaluated on MADOS and MARIDA for marine pollutant and debris segmentation, and on MagicBathyNet for bathymetry regression. The experiments show that OceanMAE yields the strongest gains on marine segmentation, while bathymetry benefits are competitive and task-dependent. In addition, an ablation against a standard MAE on MARIDA indicates that incorporating auxiliary ocean descriptors during pre-training improves downstream segmentation quality. These findings highlight the value of physically informed and domain-aligned self-supervised pre- training for ocean RS. Code and weights are publicly available at https://git.tu-berlin.de/joanna.stamer/SSLORS2.

Alignment in LLMs is more brittle than commonly assumed: misalignment can be triggered by adversarial prompts, benign fine-tuning, emergent misalignment, and goal misgeneralization. Recent evidence suggests that some misalignment behaviors are encoded as linear structure in activation space, making it tractable via steering, while safety alignment has been shown to govern the first few output tokens primarily, leaving subsequent generation unguarded. These findings motivate activation steering as a lightweight runtime defense that continuously corrects misaligned activations throughout generation. We evaluate three methods: Steer-With-Fixed-Coeff (SwFC), which applies uniform additive steering, and two novel projection-aware methods, Steer-to-Target-Projection (StTP) and Steer-to-Mirror-Projection (StMP), that use a logistic regression decision boundary to selectively intervene only on tokens whose activations fall below distributional thresholds. Using malicious system prompts as a controlled proxy for misalignment, we evaluate under two threat models (dishonesty and dismissiveness) and two architectures (Llama-3.3-70B-Instruct, Qwen3-32B). All methods substantially recover target traits (honesty and compassion) while preserving coherence. StTP and StMP better maintain general capabilities (MMLU, MT-Bench, AlpacaEval) and produce less repetition in multi-turn conversations.

Vision-language-action (VLA) models have advanced robot manipulation through large-scale pretraining, but real-world deployment remains challenging due to partial observability and delayed feedback. Reinforcement learning addresses this via value functions, which assess task progress and guide policy improvement. However, existing value models built on vision-language models (VLMs) struggle to capture temporal dynamics, undermining reliable value estimation in long-horizon tasks. In this paper, we propose ViVa, a video-generative value model that repurposes a pretrained video generator for value estimation. Taking the current observation and robot proprioception as input, ViVa jointly predicts future proprioception and a scalar value for the current state. By leveraging the spatiotemporal priors of a pretrained video generator, our approach grounds value estimation in anticipated embodiment dynamics, moving beyond static snapshots to intrinsically couple value with foresight. Integrated into RECAP, ViVa delivers substantial improvements on real-world box assembly. Qualitative analysis across all three tasks confirms that ViVa produces more reliable value signals, accurately reflecting task progress. By leveraging spatiotemporal priors from video corpora, ViVa also generalizes to novel objects, highlighting the promise of video-generative models for value estimation.

Dynamic neuroimaging data, such as emission tomography measurements of radiotracer transport in blood or cerebrospinal fluid, often exhibit diffusion-like properties. These introduce distance-dependent temporal delays, scale-differences, and stretching effects that limit the effectiveness of conventional linear modeling and decomposition methods. To address this, we present the shift- and stretch-invariant non-negative matrix factorization framework. Our approach estimates both integer and non-integer temporal shifts as well as temporal stretching, all implemented in the frequency domain, where shifts correspond to phase modifications, and where stretching is handled via zero-padding or truncation. The model is implemented in PyTorch (https://github.com/anders-s-olsen/shiftstretchNMF). We demonstrate on synthetic data and brain emission tomography data that the model is able to account for stretching to provide more detailed characterization of brain tissue structure.

The rapid advancement of facial forgery techniques poses severe threats to public trust and information security, making facial DeepFake detection a critical research priority. Continual learning provides an effective approach to adapt facial DeepFake detection models to evolving forgery patterns. However, existing methods face two key bottlenecks in real-world continual learning scenarios: insufficient feature representation and catastrophic forgetting. To address these issues, we propose Face-D(^2)CL, a framework for facial DeepFake detection. It leverages multi-domain synergistic representation to fuse spatial and frequency-domain features for the comprehensive capture of diverse forgery traces, and employs a dual continual learning mechanism that combines Elastic Weight Consolidation (EWC), which distinguishes parameter importance for real versus fake samples, and Orthogonal Gradient Constraint (OGC), which ensures updates to task-specific adapters do not interfere with previously learned knowledge. This synergy enables the model to achieve a dynamic balance between robust anti-forgetting capabilities and agile adaptability to emerging facial forgery paradigms, all without relying on historical data replay. Extensive experiments demonstrate that our method surpasses current SOTA approaches in both stability and plasticity, achieving 60.7% relative reduction in average detection error rate, respectively. On unseen forgery domains, it further improves the average detection AUC by 7.9% compared to the current SOTA method.

Rhyme is deceptively intuitive: what is or is not a rhyme is constructed historically, scholars struggle with rhyme classification, and people disagree on whether two words are rhymed or not. This complicates automated rhymed recognition and evaluation, especially in multilingual context. This article investigates how much training data is needed for reliable unsupervised rhyme recognition using RhymeTagger, a language-independent tool that identifies rhymes based on repeating patterns in poetry corpora. We evaluate its performance across seven languages (Czech, German, English, French, Italian, Russian, and Slovene), examining how training size and language differences affect accuracy. To set a realistic performance benchmark, we assess inter-annotator agreement on a manually annotated subset of poems and analyze factors contributing to disagreement in expert annotations: phonetic similarity between rhyming words and their distance from each other in a poem. We also compare RhymeTagger to three large language models using a one-shot learning strategy. Our findings show that, once provided with sufficient training data, RhymeTagger consistently outperforms human agreement, while LLMs lacking phonetic representation significantly struggle with the task.

We revisit the finite-armed linear bandit model by Nelson et al. (2022), where contexts and rewards are governed by a finite hidden Markov chain. Nelson et al. (2022) approach this model by a reduction to linear contextual bandits; but to do so, they actually introduce a simplification in which rewards are linear functions of the posterior probabilities over the hidden states given the observed contexts, rather than functions of the hidden states themselves. Their analysis (but not their algorithm) also does not take into account the estimation of the HMM parameters, and only tackles expected, not high-probability, bounds, which suffer in addition from unnecessary complex dependencies on the model (like reward gaps). We instead study the more natural model incorporating direct dependencies in the hidden states (on top of dependencies on the observed contexts, as is natural for contextual bandits) and also obtain stronger, high-probability, regret bounds for a fully adaptive strategy that estimates HMM parameters online. These bounds do not depend on the reward functions and only depend on the model through the estimation of the HMM parameters.

We propose a lightweight hybrid approach to clickbait detection that combines OpenAI semantic embeddings with six compact heuristic features capturing stylistic and informational cues. To improve efficiency, embeddings are reduced using PCA and evaluated with XGBoost, GraphSAGE, and GCN classifiers. While the simplified feature design yields slightly lower F1-scores, graph-based models achieve competitive performance with substantially reduced inference time. High ROC--AUC values further indicate strong discrimination capability, supporting reliable detection of clickbait headlines under varying decision thresholds.

Network traffic, as a key media format, is crucial for ensuring security and communications in modern internet infrastructure. While existing methods offer excellent performance, they face two key bottlenecks: (1) They fail to capture multidimensional semantics beyond unimodal sequence patterns. (2) Their black box property, i.e., providing only category labels, lacks an auditable reasoning process. We identify a key factor that existing network traffic datasets are primarily designed for classification and inherently lack rich semantic annotations, failing to generate human-readable evidence report. To address data scarcity, this paper proposes a Byte-Grounded Traffic Description (BGTD) benchmark for the first time, combining raw bytes with structured expert annotations. BGTD provides necessary behavioral features and verifiable chains of evidence for multimodal reasoning towards explainable encrypted traffic interpretation. Built upon BGTD, this paper proposes an end-to-end traffic-language representation framework (mmTraffic), a multimodal reasoning architecture bridging physical traffic encoding and semantic interpretation. In order to alleviate modality interference and generative hallucinations, mmTraffic adopts a jointly-optimized perception-cognition architecture. By incorporating a perception-centered traffic encoder and a cognition-centered LLM generator, mmTraffic achieves refined traffic interpretation with guaranteed category prediction. Extensive experiments demonstrate that mmTraffic autonomously generates high-fidelity, human-readable, and evidence-grounded traffic interpretation reports, while maintaining highly competitive classification accuracy comparing to specialized unimodal model (e.g., NetMamba). The source code is available at https://github.com/lgzhangzlg/Multimodal-Reasoning-with-LLM-for-Encrypted-Traffic-Interpretation-A-Benchmark

Mixture-of-Experts (MoE) has become a dominant architecture for scaling large language models due to their sparse activation mechanism. However, the substantial number of expert activations creates a critical latency bottleneck during inference, especially in resource-constrained deployment scenarios. Existing approaches that reduce expert activations potentially lead to severe model performance degradation. In this work, we introduce the concept of \emph{activation budget} as a constraint on the number of expert activations and propose Alloc-MoE, a unified framework that optimizes budget allocation coordinately at both the layer and token levels to minimize performance degradation. At the layer level, we introduce Alloc-L, which leverages sensitivity profiling and dynamic programming to determine the optimal allocation of expert activations across layers. At the token level, we propose Alloc-T, which dynamically redistributes activations based on routing scores, optimizing budget allocation without increasing latency. Extensive experiments across multiple MoE models demonstrate that Alloc-MoE maintains model performance under a constrained activation budget. Especially, Alloc-MoE achieves $1.15\times$ prefill and $1.34\times$ decode speedups on DeepSeek-V2-Lite at half of the original budget.

The rapid spread of online misinformation has led to increasingly complex detection models, including large language models and hybrid architectures. However, their computational cost and deployment limitations raise concerns about practical applicability. In this work, we benchmark graph neural networks (GNNs) against non-graph-based machine learning methods under controlled and comparable conditions. We evaluate lightweight GNN architectures (GCN, GraphSAGE, GAT, ChebNet) against Logistic Regression, Support Vector Machines, and Multilayer Perceptrons across seven public datasets in English, Indonesian, and Polish. All models use identical TF-IDF features to isolate the impact of relational structure. Performance is measured using F1 score, with inference time reported to assess efficiency. GNNs consistently outperform non-graph baselines across all datasets. For example, GraphSAGE achieves 96.8% F1 on Kaggle and 91.9% on WELFake, compared to 73.2% and 66.8% for MLP, respectively. On COVID-19, GraphSAGE reaches 90.5% F1 vs. 74.9%, while ChebNet attains 79.1% vs. 66.4% on FakeNewsNet. These gains are achieved with comparable or lower inference times. Overall, the results show that classic GNNs remain effective and efficient, challenging the need for increasingly complex architectures in misinformation detection.

Objective Structured Clinical Examinations (OSCEs) are the standard method for assessing medical students' clinical and communication skills through structured patient interviews. In France, however, the organization of training sessions is limited by human and logistical constraints, restricting students' access to repeated practice and structured feedback. Recent advances in Natural Language Processing (NLP) and Large Language Models (LLMs) now offer the opportunity to automatically evaluate such medical interviews, thereby alleviating the need for human examiners during training. Yet, real French OSCE annotated transcripts remain extremely scarce, limiting reproducible research and reliable benchmarking. To address these challenges, we investigate the use of LLMs for both generating and evaluating French OSCE dialogues in a low-resource context. We introduce a controlled pipeline that produces synthetic doctor-patient interview transcripts guided by scenario-specific evaluation criteria, combining ideal and perturbed performances to simulate varying student skill levels. The resulting dialogues are automatically silver-labeled through an LLM-assisted framework supporting adjustable evaluation strictness. Benchmarking multiple open-source and proprietary LLMs shows that mid-size models ($\le$32B parameters) achieve accuracies comparable to GPT-4o ($\sim$90\%) on synthetic data, highlighting the feasibility of locally deployable, privacy-preserving evaluation systems for medical education.

Reinforcement learning (RL) has become an effective approach for advancing the reasoning capabilities of large language models (LLMs) through the strategic integration of external search engines. However, current RL-based search agents often rely on a process of stochastic exploration guided by carefully crafted outcome rewards, leading to inefficient reasoning trajectories and unstable training. To address these issues, we propose a novel framework, Hierarchical Experience (HiExp), to enhance the performance and training stability of search agents. Specifically, we extract empirical knowledge through contrastive analysis and a multi-level clustering mechanism, transforming raw reasoning trajectories into hierarchical experience knowledge. By leveraging experience-aligned training, we effectively regularize stochastic exploration, evolving it into a strategic and experience-driven search process. Extensive evaluations on multiple complex agentic search and mathematical reasoning benchmarks demonstrate that our approach not only achieves substantial performance gains but also exhibits strong cross-task and cross-algorithm generalization.

Text-to-image generation executes a diffusion workflow comprising multiple models centered on a base diffusion model. Existing serving systems treat each workflow as an opaque monolith, provisioning, placing, and scaling all constituent models together, which obscures internal dataflow, prevents model sharing, and enforces coarse-grained resource management. In this paper, we make a case for micro-serving diffusion workflows with LegoDiffusion, a system that decomposes a workflow into loosely coupled model-execution nodes that can be independently managed and scheduled. By explicitly managing individual model inference, LegoDiffusion unlocks cluster-scale optimizations, including per-model scaling, model sharing, and adaptive model parallelism. Collectively, LegoDiffusion outperforms existing diffusion workflow serving systems, sustaining up to 3x higher request rates and tolerating up to 8x higher burst traffic.

Unified multimodal models integrating visual understanding and generation face a fundamental challenge: visual generation incurs substantially higher computational costs than understanding, particularly for video. This imbalance motivates us to invert the conventional paradigm: rather than extending understanding-centric MLLMs to support generation, we propose Uni-ViGU, a framework that unifies video generation and understanding by extending a video generator as the foundation. We introduce a unified flow method that performs continuous flow matching for video and discrete flow matching for text within a single process, enabling coherent multimodal generation. We further propose a modality-driven MoE-based framework that augments Transformer blocks with lightweight layers for text generation while preserving generative priors. To repurpose generation knowledge for understanding, we design a bidirectional training mechanism with two stages: Knowledge Recall reconstructs input prompts to leverage learned text-video correspondences, while Capability Refinement fine-tunes on detailed captions to establish discriminative shared representations. Experiments demonstrate that Uni-ViGU achieves competitive performance on both video generation and understanding, validating generation-centric architectures as a scalable path toward unified multimodal intelligence. Project Page and Code: https://fr0zencrane.github.io/uni-vigu-page/.

Adapting Multimodal Large Language Models (MLLMs) for hour-long videos is bottlenecked by context limits. Dense visual streams saturate token budgets and exacerbate the lost-in-the-middle phenomenon. Existing heuristics, like sparse sampling or uniform pooling, blindly sacrifice fidelity by discarding decisive moments and wasting bandwidth on irrelevant backgrounds. We propose Tempo, an efficient query-aware framework compressing long videos for downstream understanding. Tempo leverages a Small Vision-Language Model (SVLM) as a local temporal compressor, casting token reduction as an early cross-modal distillation process to generate compact, intent-aligned representations in a single forward pass. To enforce strict budgets without breaking causality, we introduce Adaptive Token Allocation (ATA). Exploiting the SVLM's zero-shot relevance prior and semantic front-loading, ATA acts as a training-free $O(1)$ dynamic router. It allocates dense bandwidth to query-critical segments while compressing redundancies into minimal temporal anchors to maintain the global storyline. Extensive experiments show our 6B architecture achieves state-of-the-art performance with aggressive dynamic compression (0.5-16 tokens/frame). On the extreme-long LVBench (4101s), Tempo scores 52.3 under a strict 8K visual budget, outperforming GPT-4o and Gemini 1.5 Pro. Scaling to 2048 frames reaches 53.7. Crucially, Tempo compresses hour-long videos substantially below theoretical limits, proving true long-form video understanding relies on intent-driven efficiency rather than greedily padded context windows.

Additive quantization enables extreme LLM compression with O(1) lookup-table dequantization, making it attractive for edge deployment. Yet at 2-bit precision, it often fails catastrophically, even with extensive search and finetuning. We show that the dominant bottleneck is codebook initialisation. Greedy sequential initialisation frequently places the model in poor optimisation regions that subsequent beam search and PV-tuning struggle to overcome. We analyse this behaviour through the representational ratio \r{ho} = N/KM, which characterises the relationship between weight groups and codebook capacity, and propose OA-EM, an output-aware EM initialisation method using Hessian-weighted Mahalanobis distance. Across compression rates, search budgets, and three architectures (Llama 3.2 3B, Llama 3.1 8B, Qwen 2.5 3B), OA-EM consistently produces better solutions after PV-tuning and dominates the quality-compute frontier. The severity of the bottleneck scales with \r{ho}: moderate at 3 bpp but extreme at 2 bpp, where poor initialisation can degrade perplexity by orders of magnitude. More broadly, our results highlight the importance of optimisation geometry in compressed model spaces, where initialisation can dominate subsequent search and fine-tuning.

This paper examines the influence of internal Gaussian noise on the performance of deep feedforward neural networks, focusing on the role of the noise injection stage relative to the activation function. Two scenarios are analyzed: noise introduced before and after the activation function, for both additive and multiplicative noise influence. The case of noise before activation function is similar to perturbations in the input channel of neuron, while the noise introduced after activation function is analogous to noise occurring either within the neuron itself or in its output channel. The types of noise and the method of their introduction were inspired by analog neural networks. The results show that the activation function acts as an effective nonlinear filter of noise. Networks with noise introduced before the activation function consistently achieve higher accuracy than those with noise applied after it, with additive noise being more effectively suppressed in this case. For noise introduced after the activation function, multiplicative noise is less detrimental than additive noise, and earlier hidden layers contribute more significantly to performance degradation due to cumulative noise amplification governed by the statistical properties of subsequent weight matrices. The study also demonstrates that pooling-based noise reduction is effective in both cases when noise is introduced before and after the activation function, consistently improving network performance.

Recent advances in Large Language Models (LLMs) have significantly improved the field of Document AI, demonstrating remarkable performance on document understanding tasks such as question answering. However, existing approaches primarily focus on solving specific tasks, lacking the capability to structurally organize and manage document information. To address this limitation, we propose Revise, a framework that systematically corrects errors introduced by OCR at the character, word, and structural levels. Specifically, Revise employs a comprehensive hierarchical taxonomy of common OCR errors and a synthetic data generation strategy that realistically simulates such errors to train an effective correction model. Experimental results demonstrate that Revise effectively corrects OCR outputs, enabling more structured representation and systematic management of document contents. Consequently, our method significantly enhances downstream performance in document retrieval and question answering tasks, highlighting the potential to overcome the structural management limitations of existing Document AI frameworks.

Ensuring reliability in adversarial settings necessitates treating privacy as a foundational component of data-driven systems. While differential privacy and cryptographic protocols offer strong guarantees, existing schemes rely on a fixed privacy budget, leading to a rigid utility-privacy trade-off that fails under heterogeneous user trust. Moreover, noise-only differential privacy preserves geometric structure, which inference attacks exploit, causing privacy leakage. We propose TADP-RME (Trust-Adaptive Differential Privacy with Reverse Manifold Embedding), a framework that enhances reliability under varying levels of user trust. It introduces an inverse trust score in the range [0,1] to adaptively modulate the privacy budget, enabling smooth transitions between utility and privacy. Additionally, Reverse Manifold Embedding applies a nonlinear transformation to disrupt local geometric relationships while preserving formal differential privacy guarantees through post-processing. Theoretical and empirical results demonstrate improved privacy-utility trade-offs, reducing attack success rates by up to 3.1 percent without significant utility degradation. The framework consistently outperforms existing methods against inference attacks, providing a unified approach for reliable learning in adversarial environments.

Machine unlearning enables models to selectively forget training data, driven by privacy regulations such as GDPR and CCPA. However, its fairness implications remain underexplored: when a model forgets a demographic group, does it neutralize that concept or redistribute it to correlated groups, potentially amplifying bias? We investigate this bias redistribution phenomenon on CelebA using CLIP models (ViT/B-32, ViT-L/14, ViT-B/16) under a zero-shot classification setting across intersectional groups defined by age and gender. We evaluate three unlearning methods, Prompt Erasure, Prompt Reweighting, and Refusal Vector using per-group accuracy shifts, demographic parity gaps, and a redistribution score. Our results show that unlearning does not eliminate bias but redistributes it primarily along gender rather than age boundaries. In particular, removing the dominant Young Female group consistently transfers performance to Old Female across all model scales, revealing a gender-dominant structure in CLIP's embedding space. While the Refusal Vector method reduces redistribution, it fails to achieve complete forgetting and significantly degrades retained performance. These findings highlight a fundamental limitation of current unlearning methods: without accounting for embedding geometry, they risk amplifying bias in retained groups.

Training-free open-vocabulary semantic segmentation(TF-OVSS) has recently attracted attention for its ability to perform dense prediction by leveraging the pretrained knowledge of large vision and vision-language models, without requiring additional training. However, due to the limited input resolution of these pretrained encoders, existing TF-OVSS methods commonly adopt a sliding-window strategy that processes cropped sub-images independently. While effective for managing high-resolution inputs, this approach prevents global attention over the full image, leading to fragmented feature representations and limited contextual reasoning. We propose OV-Stitcher, a training-free framework that addresses this limitation by stitching fragmented sub-image features directly within the final encoder block. By reconstructing attention representations from fragmented sub-image features, OV-Stitcher enables global attention within the final encoder block, producing coherent context aggregation and spatially consistent, semantically aligned segmentation maps. Extensive evaluations across eight benchmarks demonstrate that OV-Stitcher establishes a scalable and effective solution for open-vocabulary segmentation, achieving a notable improvement in mean Intersection over Union(mIoU) from 48.7 to 50.7 compared with prior training-free baselines.

We consider the classic Multi-Armed Bandit setting to understand the exploration/exploitation tradeoffs made by different search heuristics. Since many search heuristics work by comparing different options (in evolutionary algorithms called "individuals"; in the Bandit literature called "arms"), we work with the "Dueling Bandits" setting. In each iteration, a comparison between different arms can be made; in the binary stochastic setting, each arm has a fixed winning probability against any other arm. A Condorcet winner is any arm that beats every other arm with a probability strictly higher than $1/2$. We show that evolutionary algorithms are rather bad at identifying the Condorcet winner: Even if the Condorcet winner beats every other arm with a probability $1-p$, the (1+1) EA, in its stationary distribution, chooses the Condorcet winner only with constant probability if $p=Ω(1/n)$. By contrast, we show that a simple EDA (based on the Max-Min Ant System with iteration-best update) will choose the Condorcet winner in its maintained distribution with probability $1-Θ(p)$. As a remedy for the (1+1) EA, we show how repeated duels can significantly boost the probability of the Condorcet winner in the stationary distribution.

We propose Quantum Vision (QV) theory as a new perspective for deep learning-based audio classification, applied to deepfake speech detection. Inspired by particle-wave duality in quantum physics, QV theory is based on the idea that data can be represented not only in its observable, collapsed form, but also as information waves. In conventional deep learning, models are trained directly on these collapsed representations, such as images. In QV theory, inputs are first transformed into information waves using a QV block, and then fed into deep learning models for classification. QV-based models improve performance in image classification compared to their non-QV counterparts. What if QV theory is applied speech spectrograms for audio classification tasks? This is the motivation and novelty of the proposed approach. In this work, Short-Time Fourier Transform (STFT), Mel-spectrograms, and Mel-Frequency Cepstral Coefficients (MFCC) of speech signals are converted into information waves using the proposed QV block and used to train QV-based Convolutional Neural Networks (QV-CNN) and QV-based Vision Transformers (QV-ViT). Extensive experiments are conducted on the ASVSpoof dataset for deepfake speech classification. The results show that QV-CNN and QV-ViT consistently outperform standard CNN and ViT models, achieving higher classification accuracy and improved robustness in distinguishing genuine and spoofed speech. Moreover, the QV-CNN model using MFCC features achieves the best overall performance on the ASVspoof dataset, with an accuracy of 94.20% and an EER of 9.04%, while the QV-CNN with Mel-spectrograms attains the highest accuracy of 94.57%. These findings demonstrate that QV theory is an effective and promising approach for audio deepfake detection and opens new directions for quantum-inspired learning in audio perception tasks.

Large language models (LLMs) have shown astonishing capability of generating software code, leading to its use to support developers in programming. Proposed tools have relied either on assistants for improved auto-complete or multi-agents, in which different model instances are orchestrated to perform parts of a problem to reach a complete solution. We argue that LLMs can enable a higher-level of abstraction, a new paradigm we called Test-Oriented Programming (TOP). Within this paradigm, developers only have to check test code generated based on natural language specifications, rather than focusing on production code, which could be delegated to the LLMs. To evaluate the feasibility of this proposal, we developed a proof-of-concept tool and used it to generate a small command-line program employing two different LLMs. We obtained promising results and identified challenges for the use of this paradigm for real projects.

Large Language Model (LLM)-based Automated Program Repair (APR) has shown strong potential on textual benchmarks, yet struggles in multimodal scenarios where bugs are reported with GUI screenshots. Existing methods typically convert images into plain text, which discards critical spatial relationships and causes a severe disconnect between visual observations and code components, leading localization to degrade into imprecise keyword matching. To bridge this gap, we propose GALA (Graph Alignment for Localization in APR), a framework that shifts multimodal APR from implicit semantic guessing to explicit structural reasoning. GALA operates in four stages: it first constructs an Image UI Graph to capture visual elements and their structural relationships; then performs file-level alignment by cross-referencing this UI graph with repository-level structures (e.g., file references) to locate candidate files; next conducts function-level alignment by reasoning over fine-grained code dependencies (e.g., call graphs) to precisely ground visual elements to corresponding code components; and finally performs patch generation within the grounded code context based on the aligned files and functions. By systematically enforcing both semantic and relational consistency across modalities, GALA establishes a highly accurate visual-to-code mapping. Evaluations on the SWE-bench Multimodal benchmark demonstrate that GALA achieves state-of-the-art performance, highlighting the effectiveness of hierarchical structural alignment.

Passive Acoustic Monitoring (PAM) is widely used for biodiversity assessment. Its application in African tropical forests is limited by scarce annotated data, reducing the performance of general-purpose ecoacoustic models on underrepresented taxa. In this study, we introduce DeepForestSound (DFS), a multi-species automatic detection model designed for PAM in African tropical forests. DFS relies on a semi-supervised pipeline combining clustering of unannotated recordings with manual validation, followed by supervised fine-tuning of an Audio Spectrogram Transformer (AST) using low-rank adaptation, which is compared to a frozen-backbone linear baseline (DFS-Linear). The framework supports the detection of multiple taxonomic groups, including birds, primates, and elephants, from long-term acoustic recordings. DFS was trained on acoustic data collected in the Sebitoli area, in Kibale National Park, Uganda, and evaluated on an independent dataset recorded two years later at different locations within the same forest. This evaluation therefore assesses generalization across time and recording sites within a single tropical forest ecosystem. Across 8 out of 12 taxons, DFS outperforms existing automatic detection tools, particularly for non-avian taxa, achieving average AP values of 0.964 for primates and 0.961 for elephants. Results further show that LoRA-based fine-tuning substantially outperforms linear probing across taxa. Overall, these results demonstrate that task-oriented, region-specific training substantially improves detection performance in acoustically complex tropical environments, and highlight the potential of DFS as a practical tool for biodiversity monitoring and conservation in African rainforests.

Deobfuscating binary code remains a fundamental challenge in reverse engineering, as obfuscation is widely used to hinder analysis and conceal program logic. Although large language models (LLMs) have shown promise in recovering semantics from obfuscated binaries, a systematic evaluation of their effectiveness is still lacking. In this work, we present BinDeObfBench, the first comprehensive benchmark for assessing LLM-based binary deobfuscation across diverse transformations spanning pre-compilation, compile-time, and post-compilation stages. Our evaluation shows that deobfuscation performance depends more on reasoning capability and domain expertise than on model scale, and that task-specific supervised fine-tuning consistently outperforms broad domain pre-training. Reasoning models can maintain robustness under severe obfuscation, generalize across different instruction set architectures (ISAs) and optimization levels. In-context learning benefits standard models but yields limited gains for reasoning models. Overall, our study highlights the importance of task-specific fine-tuning and reasoning-driven strategies, and positions BinDeObfBench as a basis for future work in binary deobfuscation.

Production vLLM fleets typically provision each instance for the worst-case context length, leading to substantial KV-cache over-allocation and under-utilized concurrency. In practice, 80-95% of requests are short, yet are served under configurations optimized for long contexts, wasting 4-8$\times$ throughput capacity and triggering reliability issues such as OOM crashes, preemption, and request rejections. We identify a common root cause for these inefficiencies: configuration-traffic mismatch. We propose dual-pool token-budget routing, a lightweight dispatch mechanism that partitions a homogeneous fleet into two specialized pools: a high-throughput short-context pool and a high-capacity long-context pool. Each request is routed based on its estimated total token budget, computed using a per-category bytes-to-token ratio that is learned online via exponential moving average from usage.prompt_tokens feedback, eliminating the need for a tokenizer. We also develop a simple analytical model that predicts fleet-level cost savings from workload characteristics and measured throughput differences, enabling practitioners to estimate benefits prior to deployment. Evaluations on real-world traces from the Azure LLM Inference Dataset and LMSYS-Chat-1M, serving Llama-3-70B on A100 GPUs, show that our approach reduces GPU-hours by 31-42%, corresponding to \$2.86M annual savings at fleet scale, while lowering preemption rates by 5.4$\times$ and improving P99 TTFT by 6%. A case study with Qwen3-235B-A22B on AMD MI300X at 10,000 req/s projects \$15.4M in annual savings. The method incurs only O(1) dispatch overhead, adapts automatically to heterogeneous workloads, and composes seamlessly with existing optimizations such as PagedAttention, continuous batching, and prefill-decode disaggregation.

Darija, the Moroccan Arabic dialect, is rich in visual content yet lacks specialized Optical Character Recognition (OCR) tools. This paper introduces AtlasOCR, the first open-source Darija OCR model built by fine-tuning a 3B parameter Vision Language Model (VLM). We detail our comprehensive approach, from curating a unique Darija-specific dataset leveraging both synthetic generation with our OCRSmith library and carefully sourced real-world data, to implementing efficient fine-tuning strategies. We utilize QLoRA and Unsloth for parameter-efficient training of Qwen2.5-VL 3B and present comprehensive ablation studies optimizing key hyperparameters. Our evaluation on the newly curated AtlasOCRBench and the established KITAB-Bench demonstrates state-of-the-art performance, challenging larger models and highlighting AtlasOCR's robustness and generalization capabilities for both Darija and standard Arabic OCR tasks.

Tool-augmented multimodal reasoning enables visual language models (VLMs) to improve perception by interacting with external tools (e.g., cropping, depth estimation). However, such approaches incur substantial inference overhead, require specialized supervision, and are prone to erroneous tool calls. We propose Pearl (Predictive Embedding Alignment for Reasoning in Latent space), a JEPA-inspired framework that learns from expert tool-use trajectories entirely in the latent space, eliminating the need for explicit tool invocation at inference time. Unlike reconstruction-based latent reasoning methods, which autoregressively generate latent tokens and suffer from training-inference mismatch and limited support for multi-step tool use, Pearl directly learns predictive embeddings from multimodal trajectories while preserving the standard vision-language generation pipeline: it is model-agnostic, simple to train, and naturally supports trajectories with multiple tool calls. Experiments across multiple perception benchmarks show that Pearl matches or outperforms standard supervised fine-tuning and reconstruction-based latent reasoning approaches. Furthermore, we provide empirical evidence that reconstruction-based methods primarily learn embeddings rather than image edits in latent space, motivating predictive embedding learning as a more principled alternative.

Existing memory benchmarks for LLM agents evaluate explicit recall of facts, yet overlook implicit memory where experience becomes automated behavior without conscious retrieval. This gap is critical: effective assistants must automatically apply learned procedures or avoid failed actions without explicit reminders. We introduce ImplicitMemBench, the first systematic benchmark evaluating implicit memory through three cognitively grounded constructs drawn from standard cognitive-science accounts of non-declarative memory: Procedural Memory (one-shot skill acquisition after interference), Priming (theme-driven bias via paired experimental/control instances), and Classical Conditioning (Conditioned Stimulus--Unconditioned Stimulus (CS--US) associations shaping first decisions). Our 300-item suite employs a unified Learning/Priming-Interfere-Test protocol with first-attempt scoring. Evaluation of 17 models reveals severe limitations: no model exceeds 66% overall, with top performers DeepSeek-R1 (65.3%), Qwen3-32B (64.1%), and GPT-5 (63.0%) far below human baselines. Analysis uncovers dramatic asymmetries (inhibition 17.6% vs. preference 75.0%) and universal bottlenecks requiring architectural innovations beyond parameter scaling. ImplicitMemBench reframes evaluation from "what agents recall" to "what they automatically enact".

Current LLM assistants are powerful at answering questions, but they have limited access to the behavioral context that reveals when and where a user is struggling. We present a gaze-grounded multimodal LLM assistant that uses egocentric video with gaze overlays to identify likely points of difficulty and target follow-up retrospective assistance. We instantiate this vision in a controlled study (n=36) comparing the gaze-aware AI assistant to a text-only LLM assistant. Compared to a conventional LLM assistant, the gaze-aware assistant was rated as significantly more accurate and personalized in its assessments of users' reading behavior and significantly improved people's ability to recall information. Users spoke significantly fewer words with the gaze-aware assistant, indicating more efficient interactions. Qualitative results underscored both perceived benefits in comprehension and challenges when interpretations of gaze behaviors were inaccurate. Our findings suggest that gaze-aware LLM assistants can reason about cognitive needs to improve cognitive outcomes of users.

Embodied agents are increasingly expected to improve over time by updating their executable capabilities rather than rewriting the agent itself. Prior work has separately studied modular capability packaging, capability evolution, and runtime governance. However, a key systems problem remains underexplored: once an embodied capability module evolves into a new version, how can the hosting system deploy it safely without breaking policy constraints, execution assumptions, or recovery guarantees? We formulate governed capability evolution as a first-class systems problem for embodied agents. We propose a lifecycle-aware upgrade framework in which every new capability version is treated as a governed deployment candidate rather than an immediately executable replacement. The framework introduces four upgrade compatibility checks -- interface, policy, behavioral, and recovery -- and organizes them into a staged runtime pipeline comprising candidate validation, sandbox evaluation, shadow deployment, gated activation, online monitoring, and rollback. We evaluate over 6 rounds of capability upgrade with 15 random seeds. Naive upgrade achieves 72.9% task success but drives unsafe activation to 60% by the final round; governed upgrade retains comparable success (67.4%) while maintaining zero unsafe activations across all rounds (Wilcoxon p=0.003). Shadow deployment reveals 40% of regressions invisible to sandbox evaluation alone, and rollback succeeds in 79.8% of post-activation drift scenarios.

Federated Learning enables collaborative model training without centralising data, but its effectiveness varies with the selection of the aggregation strategy. This choice is non-trivial, as performance varies widely across datasets, heterogeneity levels, and compute constraints. We present an end-to-end framework that automates, streamlines, and adapts aggregation strategy selection for federated learning. The framework operates in two modes: a single-trial mode, where large language models infer suitable strategies from user-provided or automatically detected data characteristics, and a multi-trial mode, where a lightweight genetic search efficiently explores alternatives under constrained budgets. Extensive experiments across diverse datasets show that our approach enhances robustness and generalisation under non-IID conditions while reducing the need for manual intervention. Overall, this work advances towards accessible and adaptive federated learning by automating one of its most critical design decisions, the choice of an aggregation strategy.

Linguistic steganography involves embedding secret messages within seemingly innocuous texts to enable covert communication. Provable security, which is a long-standing goal and key motivation, has been extended to language-model-based steganography. Previous provably secure approaches have achieved perfect imperceptibility, measured by zero Kullback-Leibler (KL) divergence, but at the expense of embedding capacity. In this paper, we attempt to directly use a classic entropy coding method (range coding) to achieve secure steganography, and then propose an efficient and provably secure linguistic steganographic method with a rotation mechanism. Experiments across various language models show that our method achieves around 100% entropy utilization (embedding efficiency) for embedding capacity, outperforming the existing baseline methods. Moreover, it achieves high embedding speeds (up to 1554.66 bits/s on GPT-2). The code is available at github.com/ryehr/RRC_steganography.

Retrieval-Augmented Generation (RAG) significantly enhances Large Language Models (LLMs) by providing access to external knowledge. However, current research primarily focuses on retrieval quality, often overlooking the critical ''integration bottleneck'': even when relevant documents are retrieved, LLMs frequently fail to utilize them effectively due to conflicts with their internal parametric knowledge. In this paper, we argue that implicitly resolving this conflict in a single generation pass is suboptimal. We introduce GuarantRAG, a framework that explicitly decouples reasoning from evidence integration. First, we generate an ''Inner-Answer'' based solely on parametric knowledge to capture the model's reasoning flow. Second, to guarantee faithful evidence extraction, we generate a ''Refer-Answer'' using a novel Contrastive DPO objective. This objective treats the parametric Inner-Answer as a negative constraint and the retrieved documents as positive ground truth, forcing the model to suppress internal hallucinations in favor of external evidence during this phase. Finally, rather than naive concatenation or using the DPO trained model directly, we propose a joint decoding mechanism that dynamically fuses the logical coherence of the Inner-Answer with the factual precision of the Refer-Answer at the token level. Experiments on five QA benchmarks demonstrate that GuarantRAG improves accuracy by up to 12.1% and reduces hallucinations by 16.3% compared to standard and dynamic RAG baselines.

Large language models (LLMs) exhibit memory-intensive behavior during decoding, making it a key bottleneck in LLM inference. To accelerate decoding execution, hybrid-bonding-based 3D-DRAM has been adopted in LLM accelerators. While this emerging technology provides strong performance gains over existing hardware, current 3D-DRAM accelerators (3D-Accelerators) rely on closed-source evaluation tools, limiting access to publicly available performance analysis methods. Moreover, existing designs are highly customized for specific scenarios, lacking a general and reusable full-stack modeling for 3D-Accelerators across diverse usecases. To bridge this fundamental gap, we present ATLAS, the first silicon-proven Architectural Three-dimesional-DRAM-based LLM Accelerator Simulation framework. Built on commercially deployed multi-layer 3D-DRAM technology, ATLAS introduces unified abstractions for both 3D-Accelerator system architecture and programming primitives to support arbitrary LLM inference scenarios. Validation against real silicon shows that ATLAS achieves $\le$8.57% simulation error and 97.26-99.96\% correlation with measured performance. Through design space exploration with ATLAS, we demonstrate its ability to guide architecture design and distill key takeaways for both 3D-DRAM memory system and 3D-Accelerator microarchitecture across scenarios. ATLAS will be open-sourced upon publication, enabling further research on 3D-Accelerators.

Sketching in 3D space enables expressive reasoning about shape, structure, and spatial relationships, yet generating 3D sketches through natural language remains a major challenge. In this work, we introduce 3DrawAgent, a training-free, language-driven framework for 3D sketch generation that leverages large language models (LLMs) to sequentially draw 3D Bezier curves under geometric feedback. Unlike prior 2D sketch agents, our method introduces a relative experience optimization strategy that adapts the recently proposed Group Reward Policy Optimization (GRPO) paradigm. Instead of relying on explicit ground-truth supervision, we construct pairwise comparisons among generated sketches, with each pair consisting of a relatively better and a worse result based on CLIP-based perceptual rewards and LLM-based fine-grained qualitative assessment. These experiences are then used to iteratively refine the prior knowledge of 3D drawing, enabling black-box reinforcement of the model's 3D awareness. This design allows our model to self-improve its spatial understanding and drawing quality without parameter updates. Experiments show that 3DrawAgent can generate complex and coherent 3D Bezier sketches from diverse textual prompts, exhibit emergent geometric reasoning, and generalize to novel shapes, establishing a new paradigm for advancing the field of training-free 3D sketch intelligence.

Interpreting the concepts encoded by individual neurons in deep neural networks is a crucial step towards understanding their complex decision-making processes and ensuring AI safety. Despite recent progress in neuron labeling, existing methods often limit the search space to predefined concept vocabularies or produce overly specific descriptions that fail to capture higher-order, global concepts. We introduce LINE, a novel, training-free iterative approach tailored for open-vocabulary concept labeling in vision models. Operating in a strictly black-box setting, LINE leverages a large language model and a text-to-image generator to iteratively propose and refine concepts in a closed loop, guided by activation history. We demonstrate that LINE achieves state-of-the-art performance across multiple model architectures, yielding AUC improvements of up to 0.18 on ImageNet and 0.05 on Places365, while discovering, on average, 29% of new concepts missed by massive predefined vocabularies. Beyond identifying the top concept, LINE provides a complete generation history, which enables polysemanticity evaluation and produces supporting visual explanations that rival gradient-dependent activation maximization methods.

Talking-head generation has advanced rapidly with diffusion-based generative models, but training usually depends on centralized face-video and speech datasets, raising major privacy concerns. The problem is more acute for personalized talking-head generation, where identity-specific data are highly sensitive and often cannot be pooled across users or devices. PrivFedTalk is presented as a privacy-aware federated framework for personalized talking-head generation that combines conditional latent diffusion with parameter-efficient identity adaptation. A shared diffusion backbone is trained across clients, while each client learns lightweight LoRA identity adapters from local private audio-visual data, avoiding raw data sharing and reducing communication cost. To address heterogeneous client distributions, Identity-Stable Federated Aggregation (ISFA) weights client updates using privacy-safe scalar reliability signals computed from on-device identity consistency and temporal stability estimates. Temporal-Denoising Consistency (TDC) regularization is introduced to reduce inter-frame drift, flicker, and identity drift during federated denoising. To limit update-side privacy risk, secure aggregation and client-level differential privacy are applied to adapter updates. The implementation supports both low-memory GPU execution and multi-GPU client-parallel training on heterogeneous shared hardware. Comparative experiments on the present setup across multiple training and aggregation conditions with PrivFedTalk, FedAvg, and FedProx show stable federated optimization and successful end-to-end training and evaluation under constrained resources. The results support the feasibility of privacy-aware personalized talking-head training in federated environments, while suggesting that stronger component-wise, privacy-utility, and qualitative claims need further standardized evaluation.

This paper addresses the problem of training a reinforcement learning (RL) policy under partial observability by exploiting a privileged, anytime-feasible planner agent available exclusively during training. We formalize this as a Partially Observable Markov Decision Process (POMDP) in which a planner agent with access to an approximate dynamical model and privileged state information guides a learning agent that observes only a lossy projection of the true state. To realize this framework, we introduce an anytime-feasible Model Predictive Control (MPC) algorithm that serves as the planner agent. For the learning agent, we propose Planner-to-Policy Soft Actor-Critic (P2P-SAC), a method that distills the planner agent's privileged knowledge to mitigate partial observability and thereby improve both sample efficiency and final policy performance. We support this framework with rigorous theoretical analysis. Finally, we validate our approach in simulation using NVIDIA Isaac Lab and successfully deploy it on a real-world Unitree Go2 quadruped navigating complex, obstacle-rich environments.

Intelligent systems powered by large-scale sensor networks are shifting from predefined monitoring to intent-driven operation, revealing a critical Semantic-to-Physical Mapping Gap. While large language models (LLMs) excel at semantic understanding, existing perception-centric pipelines operate retrospectively, overlooking the fundamental decision of what to sense and when. We formalize this proactive decision as Semantic-Spatial Sensor Scheduling (S3) and demonstrate that direct LLM planning is unreliable due to inherent gaps in representation, reasoning, and optimization. To bridge these gaps, we introduce the Spatial Trajectory Graph (STG), a neuro-symbolic paradigm governed by a verify-before-commit discipline that transforms open-ended planning into a verifiable graph optimization problem. Based on STG, we implement IoT-Brain, a concrete system embodiment, and construct TopoSense-Bench, a campus-scale benchmark with 5,250 natural-language queries across 2,510 cameras. Evaluations show that IoT-Brain boosts task success rate by 37.6% over the strongest search-intensive methods while running nearly 2 times faster and using 6.6 times fewer prompt tokens. In real-world deployment, it approaches the reliability upper bound while reducing 4.1 times network bandwidth, providing a foundational framework for LLMs to interact with the physical world with unprecedented reliability and efficiency.

Automated maritime collision avoidance will rely on human supervision for the foreseeable future. This necessitates transparency into how the system perceives a scenario and plans a maneuver. However, the causal logic behind avoidance maneuvers is often complex and difficult to convey to a navigator. This paper explores how to explain these factors in a selective, understandable manner for supervisors with a nautical background. We propose a method for generating contrastive explanations, which provide human-centric insights by comparing a system's proposed solution against relevant alternatives. To evaluate this, we developed a framework that uses visual and textual cues to highlight key objectives from a state-of-the-art collision avoidance system. An exploratory user study with four experienced marine officers suggests that contrastive explanations support the understanding of the system's objectives. However, our findings also reveal that while these explanations are highly valuable in complex multi-vessel encounters, they can increase cognitive workload, suggesting that future maritime interfaces may benefit most from demand-driven or scenario-specific explanation strategies.

Algorithmic recourse aims to provide actionable recommendations that enable individuals to change unfavorable model outcomes, and prior work has extensively studied properties such as efficiency, robustness, and fairness. However, the role of personalization in recourse remains largely implicit and underexplored. While existing approaches incorporate elements of personalization through user interactions, they typically lack an explicit definition of personalization and do not systematically analyze its downstream effects on other recourse desiderata. In this paper, we formalize personalization as individual actionability, characterized along two dimensions: hard constraints that specify which features are individually actionable, and soft, individualized constraints that capture preferences over action values and costs. We operationalize these dimensions within the causal algorithmic recourse framework, adopting a pre-hoc user-prompting approach in which individuals express preferences via rankings or scores prior to the generation of any recourse recommendation. Through extensive empirical evaluation, we investigate how personalization interacts with key recourse desiderata, including validity, cost, and plausibility. Our results highlight important trade-offs: individual actionability constraints, particularly hard ones, can substantially degrade the plausibility and validity of recourse recommendations across amortized and non-amortized approaches. Notably, we also find that incorporating individual actionability can reveal disparities in the cost and plausibility of recourse actions across socio-demographic groups. These findings underscore the need for principled definitions, careful operationalization, and rigorous evaluation of personalization in algorithmic recourse.

Recent deep learning (DL) methods for log anomaly detection increasingly rely on semantic log representation methods that convert the textual content of log events into vector embeddings as input to DL models. However, these DL methods are typically evaluated as end-to-end pipelines, while the impact of different semantic representation methods is not well understood. In this paper, we benchmark widely used semantic log representation methods, including static word embedding methods (Word2Vec, GloVe, and FastText) and the BERT-based contextual embedding method, across diverse DL models for log-event level anomaly detection on three publicly available log datasets: BGL, Thunderbird, and Spirit. We identify an effectiveness--efficiency trade off under CPU deployment settings: the BERT-based method is more effective, but incurs substantially longer log embedding generation time, limiting its practicality; static word embedding methods are efficient but are generally less effective and may yield insufficient detection performance. Motivated by this finding, we propose QTyBERT, a novel semantic log representation method that better balances this trade-off. QTyBERT uses SysBE, a lightweight BERT variant with system-specific quantization, to efficiently encode log events into vector embeddings on CPUs, and leverages CroSysEh to enhance the semantic expressiveness of these log embeddings. CroSysEh is trained unsupervisedly using unlabeled logs from multiple systems to capture the underlying semantic structure of the BERT model's embedding space. We evaluate QTyBERT against existing semantic log representation methods. Our results show that, for the DL models, using QTyBERT-generated log embeddings achieves detection effectiveness comparable to or better than BERT-generated log embeddings, while bringing log embedding generation time closer to that of static word embedding methods.

Regardless of its foundational role in human discovery and sense-making, abductive reasoning--the inference of the most plausible explanation for an observation--has been relatively underexplored in Large Language Models (LLMs). Despite the rapid advancement of LLMs, the exploration of abductive reasoning and its diverse facets has thus far been disjointed rather than cohesive. This paper presents the first survey of abductive reasoning in LLMs, tracing its trajectory from philosophical foundations to contemporary AI implementations. To address the widespread conceptual confusion and disjointed task definitions prevalent in the field, we establish a unified two-stage definition that formally categorizes prior work. This definition disentangles abduction into \textit{Hypothesis Generation}, where models bridge epistemic gaps to produce candidate explanations, and \textit{Hypothesis Selection}, where the generated candidates are evaluated and the most plausible explanation is chosen. Building upon this foundation, we present a comprehensive taxonomy of the literature, categorizing prior work based on their abductive tasks, datasets, underlying methodologies, and evaluation strategies. In order to ground our framework empirically, we conduct a compact benchmark study of current LLMs on abductive tasks, together with targeted comparative analyses across model sizes, model families, evaluation styles, and the distinct generation-versus-selection task typologies. Moreover, by synthesizing recent empirical results, we examine how LLM performance on abductive reasoning relates to deductive and inductive tasks, providing insights into their broader reasoning capabilities. Our analysis reveals critical gaps in current approaches--from static benchmark design and narrow domain coverage to narrow training frameworks and limited mechanistic understanding of abductive processes...

We propose a unified objective function, termed CATMIL, that augments the base segmentation loss with two auxiliary supervision terms operating at different levels. The first term, Component-Adaptive Tversky, reweights voxel contributions based on connected components to balance the influence of lesions of different sizes. The second term, based on Multiple Instance Learning, introduces lesion-level supervision by encouraging the detection of each lesion instance. These terms are combined with the standard nnU-Net loss to jointly optimize voxel-level segmentation accuracy and lesion-level detection. We evaluate the proposed objective on the MSLesSeg dataset using a consistent nnU-Net framework and 5-fold cross-validation. The results show that CATMIL achieves the most balanced performance across segmentation accuracy, lesion detection, and error control. It improves Dice score (0.7834) and reduces boundary error compared to standard losses. More importantly, it substantially increases small lesion recall and reduces false negatives, while maintaining the lowest false positive volume among compared methods. These findings demonstrate that integrating component-level and lesion-level supervision within a unified objective provides an effective and practical approach for improving small lesion segmentation in highly imbalanced settings. All code and pretrained models are available at \href{https://github.com/luumsk/SmallLesionMRI}{this url}.

Retrieving rare and safety-critical driving scenarios from large-scale datasets is essential for building robust autonomous driving (AD) systems. As dataset sizes continue to grow, the key challenge shifts from collecting more data to efficiently identifying the most relevant samples. We introduce SearchAD, a large-scale rare image retrieval dataset for AD containing over 423k frames drawn from 11 established datasets. SearchAD provides high-quality manual annotations of more than 513k bounding boxes covering 90 rare categories. It specifically targets the needle-in-a-haystack problem of locating extremely rare classes, with some appearing fewer than 50 times across the entire dataset. Unlike existing benchmarks, which focused on instance-level retrieval, SearchAD emphasizes semantic image retrieval with a well-defined data split, enabling text-to-image and image-to-image retrieval, few-shot learning, and fine-tuning of multi-modal retrieval models. Comprehensive evaluations show that text-based methods outperform image-based ones due to stronger inherent semantic grounding. While models directly aligning spatial visual features with language achieve the best zero-shot results, and our fine-tuning baseline significantly improves performance, absolute retrieval capabilities remain unsatisfactory. With a held-out test set on a public benchmark server, SearchAD establishes the first large-scale dataset for retrieval-driven data curation and long-tail perception research in AD: https://iis-esslingen.github.io/searchad/

REST APIs enable collaboration among microservices. A single fault in a REST API can bring down the entire microservice system and cause significant financial losses, underscoring the importance of REST API testing. Effectively testing REST APIs requires thoroughly exercising the functionalities behind them. To this end, existing techniques leverage REST specifications (e.g., Swagger or OpenAPI) to generate test cases. Using the resource constraints extracted from specifications, these techniques work well for testing simple, business-insensitive functionalities, such as resource creation, retrieval, update, and deletion. However, for complex, business-sensitive functionalities, these specification-based techniques often fall short, since exercising such functionalities requires additional business constraints that are typically absent from REST specifications. In this paper, we present LoBREST, a log-based, business-aware REST API testing technique that leverages historical request logs (HRLogs) to effectively exercise the business-sensitive functionalities behind REST APIs. To obtain compact operation sequences that preserve clean and complete business constraints, LoBREST first employs a locality-slicing strategy to partition HRLogs into smaller slices. Then, to ensure the effectiveness of the obtained slices, LoBREST enhances them in two steps: (1) adding slices for operations missing from HRLogs, and (2) completing missing resources within the slices. Finally, to improve test adequacy, LoBREST uses these enhanced slices as initial seeds to perform business-aware fuzzing. LoBREST outperformed eight tools (including Arat-rl, Morest, and Deeprest) across 17 real-world services. It achieved top operation coverage on 16 services and line coverage on 15, averaging 2.1x and 1.2x improvements over the runner-up. LoBREST detected 108 5XX bugs, including 38 found by no other tool.

Advancements in multimodal foundation models have enabled the development of Computer Use Agents (CUAs) capable of autonomously interacting with GUI environments. As CUAs are not restricted to certain tools, they allow to automate more complex agentic tasks but at the same time open up new security vulnerabilities. While prior work has concentrated on the language modality, the vulnerability of the vision modality has received less attention. In this paper, we introduce PRAC, a novel attack that, unlike prior work targeting the VLM output directly, manipulates the model's internal preferences by redirecting its attention toward a stealthy adversarial patch. We show that PRAC is able to manipulate the selection process of a CUA on an online shopping platform towards a chosen target product. While we require white-box access to the model for the creation of the attack, we show that our attack generalizes to fine-tuned versions of the same model, presenting a critical threat as multiple companies build specific CUAs based on open weights models.

Existing algorithms for generating Counterfactual Explanations (CXs) for Machine Learning (ML) typically assume fully specified inputs. However, real-world data often contains missing values, and the impact of these incomplete inputs on the performance of existing CX methods remains unexplored. To address this gap, we systematically evaluate recent CX generation methods on their ability to provide valid and plausible counterfactuals when inputs are incomplete. As part of this investigation, we hypothesize that robust CX generation methods will be better suited to address the challenge of providing valid and plausible counterfactuals when inputs are incomplete. Our findings reveal that while robust CX methods achieve higher validity than non-robust ones, all methods struggle to find valid counterfactuals. These results motivate the need for new CX methods capable of handling incomplete inputs.

Integrating large language models (LLMs) into automatic speech recognition (ASR) has become a dominant paradigm. Although recent LLM-based ASR models have shown promising performance on public benchmarks, it remains challenging to balance recognition quality with latency and overhead, while hallucinations further limit real-world deployment. In this study, we revisit LLM-based ASR from an entropy allocation perspective and introduce three metrics to characterize how training paradigms allocate entropy reduction between the speech encoder and the LLM. To remedy entropy-allocation inefficiencies in prevailing approaches, we propose a principled multi-stage training strategy grounded in capability-boundary awareness, optimizing parameter efficiency and hallucination robustness. Specifically, we redesign the pretraining strategy to alleviate the speech-text modality gap, and further introduce an iterative asynchronous SFT stage between alignment and joint SFT to preserve functional decoupling and constrain encoder representation drift. Experiments on Mandarin and English benchmarks show that our method achieves competitive performance with state-of-the-art models using only 2.3B parameters, while also effectively mitigating hallucinations through our decoupling-oriented design.

Machine learning competitions (MLCs) play a pivotal role in advancing artificial intelligence (AI) by fostering innovation, skill development, and practical problem-solving. This study provides a comprehensive analysis of major competition platforms such as Kaggle and Zindi, examining their workflows, evaluation methodologies, and reward structures. It further assesses competition quality, participant expertise, and global reach, with particular attention to demographic trends among top-performing competitors. By exploring the motivations of competition hosts, this paper underscores the significant role of MLCs in shaping AI development, promoting collaboration, and driving impactful technological progress. Furthermore, by combining literature synthesis with platform-level data analysis and practitioner insights a comprehensive understanding of the MLC ecosystem is provided. Moreover, the paper demonstrates that MLCs function at the intersection of academic research and industrial application, fostering the exchange of knowledge, data, and practical methodologies across domains. Their strong ties to open-source communities further promote collaboration, reproducibility, and continuous innovation within the broader ML ecosystem. By shaping research priorities, informing industry standards, and enabling large-scale crowdsourced problem-solving, these competitions play a key role in the ongoing evolution of AI. The study provides insights relevant to researchers, practitioners, and competition organizers, and includes an examination of the future trajectory and sustained influence of MLCs on AI development.

Proactivity is a core expectation for AGI. Prior work remains largely confined to laboratory settings, leaving a clear gap in real-world proactive agent: depth, complexity, ambiguity, precision and real-time constraints. We study this setting, where useful intervention requires inferring latent needs from ongoing context and grounding actions in evolving user memory under latency and long-horizon constraints. We first propose DD-MM-PAS (Demand Detection, Memory Modeling, Proactive Agent System) as a general paradigm for streaming proactive AI agent. We instantiate this paradigm in Pask, with streaming IntentFlow model for DD, a hybrid memory (workspace, user, global) for long-term MM, PAS infra framework and introduce how these components form a closed loop. We also introduce LatentNeeds-Bench, a real-world benchmark built from user-consented data and refined through thousands of rounds of human editing. Experiments show that IntentFlow matches leading Gemini3-Flash models under latency constraints, while identifying deeper user intent.

Accurate segmentation of the future liver remnant (FLR) is critical for surgical planning in colorectal liver metastases (CRLM) to prevent fatal post-hepatectomy liver failure. However, this segmentation task is technically challenging due to complex resection boundaries, convoluted hepatic vasculature and diffuse metastatic lesions. A primary bottleneck in developing automated AI tools has been the lack of high-fidelity, validated data. We address this gap by manually refining all 197 volumes from the public CRLM-CT-Seg dataset, creating the first open-source, validated benchmark for this task. We then establish the first segmentation baselines, comparing cascaded (Liver->CRLM->FLR) and end-to-end (E2E) strategies using nnU-Net, SwinUNETR, and STU-Net. We find a cascaded nnU-Net achieves the best final FLR segmentation Dice (0.767), while the pretrained STU-Net provides superior CRLM segmentation (0.620 Dice) and is significantly more robust to cascaded errors. This work provides the first validated benchmark and a reproducible framework to accelerate research in AI-assisted surgical planning.

While infographics have become a powerful medium for communicating data-driven stories, authoring them from scratch remains challenging, especially for novice users. Retrieving relevant exemplars from a large corpus can provide design inspiration and promote reuse, substantially lowering the barrier to infographic authoring. However, effective retrieval is difficult because users often express design intent in ambiguous natural language, while infographics embody rich and multi-faceted visual designs. As a result, keyword-based search often fails to capture design intent, and general-purpose vision-language retrieval models trained on natural images are ill-suited to the text-heavy, multi-component nature of infographics. To address these challenges, we develop an intent-aware infographic retrieval framework that better aligns user queries with infographic designs. We first conduct a formative study of how people describe infographics and derive an intent taxonomy spanning content and visual design facets. This taxonomy is then leveraged to enrich and refine free-form user queries, guiding the retrieval process with intent-specific cues. Building on the retrieved exemplars, users can adapt the designs to their own data with high-level edit intents, supported by an interactive agent that performs low-level adaptation. Both quantitative evaluations and user studies are conducted to demonstrate that our method improves retrieval quality over baseline methods while better supporting intent satisfaction and efficient infographic authoring.

Agents are LLM-driven components that can mutate environments in powerful, arbitrary ways. Extracting guarantees for the execution of agents in production environments can be challenging due to asynchrony and failures. In this paper, we propose a new abstraction called LogAct, where each agent is a deconstructed state machine playing a shared log. In LogAct, agentic actions are visible in the shared log before they are executed; can be stopped prior to execution by pluggable, decoupled voters; and recovered consistently in the case of agent or environment failure. LogAct enables agentic introspection, allowing the agent to analyze its own execution history using LLM inference, which in turn enables semantic variants of recovery, health check, and optimization. In our evaluation, LogAct agents recover efficiently and correctly from failures; debug their own performance; optimize token usage in swarms; and stop all unwanted actions for a target model on a representative benchmark with just a 3% drop in benign utility.

We address the task of predicting the gain of using RAG (retrieval augmented generation) for question answering with respect to not using it. We study the performance of a few pre-retrieval and post-retrieval predictors originally devised for ad hoc retrieval. We also study a few post-generation predictors, one of which is novel to this study and posts the best prediction quality. Our results show that the most effective prediction approach is a novel supervised predictor that explicitly models the semantic relationships among the question, retrieved passages, and the generated answer.

Long-context reasoning is essential for complex real-world applications, yet remains a significant challenge for Large Language Models (LLMs). Despite the rapid evolution in long-context reasoning, current research often overlooks the internal complexity of the long-context reasoning task itself. In this paper, we move beyond this holistic view and decompose long-context reasoning into a set of fundamental atomic skills, and we then automatically synthesize a suite of pseudo datasets, each explicitly targeting a specific atomic skill. Our empirical analysis confirms that proficiency in these atomic skills is strongly correlated with general long-text reasoning performance. Building on this insight, we employ reinforcement learning on these pseudo datasets to sharpen the model's atomic skills, in the hope of boosting its general long-context reasoning ability. Extensive experiments across multiple benchmarks demonstrate the effectiveness of our approach: it outperforms a strong baseline by an average margin of 7.7\% (improving from 46.3\% to 54.0\%) across Loogle, Loong, LongBench-v2, BrowscompLong, Ruler-qa2, and MRCR.

Large multimodal models (LMMs) show strong visual-linguistic reasoning but their capacity for spatial decision-making and action remains unclear. In this work, we investigate whether LMMs can achieve embodied spatial action like human through a challenging scenario: goal-oriented navigation in urban 3D spaces. We first spend over 500 hours constructing a dataset comprising 5,037 high-quality goal-oriented navigation samples, with an emphasis on 3D vertical actions and rich urban semantic information. Then, we comprehensively assess 17 representative models, including non-reasoning LMMs, reasoning LMMs, agent-based methods, and vision-language-action models. Experiments show that current LMMs exhibit emerging action capabilities, yet remain far from human-level performance. Furthermore, we reveal an intriguing phenomenon: navigation errors do not accumulate linearly but instead diverge rapidly from the destination after a critical decision bifurcation. The limitations of LMMs are investigated by analyzing their behavior at these critical decision bifurcations. Finally, we experimentally explore four promising directions for improvement: geometric perception, cross-view understanding, spatial imagination, and long-term memory. The project is available at: https://github.com/serenditipy-AC/Embodied-Navigation-Bench.

We present Kathleen, a text classification architecture that operates directly on raw UTF-8 bytes using frequency-domain processing -- requiring no tokenizer, no attention mechanism, and only 733K parameters. Kathleen introduces three novel components: (1) RecurrentOscillatorBanks -- damped sinusoid convolutions with temporal memory for O(L) sequence processing; (2) an FFT-Rotate Wavetable Encoder that maps all 256 byte values using a single learnable vector (256 floats), replacing conventional embedding tables (65K parameters) while improving accuracy; (3) PhaseHarmonics -- a sinusoidal non-linearity with just 6 learnable phase parameters that our ablation identifies as the single most impactful component (+2.6% accuracy, <0.001% of model parameters). Through comprehensive ablation of a 1.8M-parameter predecessor, we show that frequency-domain components systematically outperform complex cognitive architectures: removing a 560K-parameter bio-inspired framework costs only -0.2%, while removing the 6-parameter PhaseHarmonics costs -2.6%. The resulting Kathleen-Clean achieves 88.6% on IMDB, 92.3% on AG News, and 83.3% on SST-2 -- outperforming a tokenized counterpart with 16x more parameters on IMDB (+1.6%) and AG News (+2.1%). Kathleen processes sequences in O(L) time and memory, enabling byte-level operation at sequence lengths where O(L^2) Transformers exhaust GPU memory.

Adversarial claim rewriting is widely used to test fact-checking systems, but standard metrics fail to capture truth-conditional consistency and often label semantically corrupted rewrites as successful. We introduce AtomEval, a validity-aware evaluation framework that decomposes claims into subject-relation-object-modifier (SROM) atoms and scores adversarial rewrites with Atomic Validity Scoring (AVS), enabling detection of factual corruption beyond surface similarity. Experiments on the FEVER dataset across representative attack strategies and LLM generators show that AtomEval provides more reliable evaluation signals in our experiments. Using AtomEval, we further analyze LLM-based adversarial generators and observe that stronger models do not necessarily produce more effective adversarial claims under validity-aware evaluation, highlighting previously overlooked limitations in current adversarial evaluation practices.

Model editing aims to update knowledge to add new concepts and change relevant information without retraining. Lifelong editing is a challenging task, prone to disrupting previously learned concepts, especially for Vision Language Models (VLMs), because sequential edits can lead to degraded reasoning and cross modal misalignment. Existing VLM knowledge editing methods based on gated adapters, activation edits, and parameter merging techniques address catastrophic forgetting seen in full fine tuning; however, they still operate in the shared representation space of the VLM, where concepts are entangled, so edits interfere with other non relevant concepts. We hypothesize that this instability persists because current methods algorithmically control edits via optimization rather than structurally separating knowledge. We introduce Dynamic Subspace Concept Alignment (DSCA) which by design mitigates this limitation by decomposing the representation space into a set of orthogonal semantic subspaces and proposing edits only in those transformed spaces. These subspaces are obtained through incremental clustering and PCA on joint vision language representations. This process structurally isolates concepts, enabling precise, non interfering edits by turning isolation from a soft training objective into an architectural property. The surgical edits are guided by a multi term loss function for maintaining task fidelity, edit locality, and cross modal alignment. With the base model frozen, our method achieves 98 percent single edit success, remains over 95 percent after 1000 sequential edits, lowers hallucination by 3 to 5 percent, and achieves the best backward transfer (BWT) scores on continual instruction tuning benchmarks. Extensive experiments demonstrate DSCA state of the art stability and knowledge retention capability in continual lifelong editing across various datasets and benchmarks.

The integration of large language models (LLMs) into peer review raises a concern beyond authorship and detection: the potential cascading automation of the entire editorial process. As reviews become partially or fully machine-generated, it becomes plausible that editorial decisions may also be delegated to algorithmic systems, leading to a fully automated evaluation pipeline. They risk reshaping the criteria by which scientific work is assessed. This paper argues that machine-driven assessment may systematically favor standardized, pattern-conforming research while penalizing unconventional and paradigm-shifting ideas that require contextual human judgment. We consider that this shift could lead to epistemic homogenization, where researchers are implicitly incentivized to optimize their work for algorithmic approval rather than genuine discovery. To address this risk, we introduce an explainable framework (RAG-XAI) for assessing review quality and detecting automated patterns using markers LLM extractor, aiming to preserve transparency, accountability and creativity in science. The proposed framework achieves near-perfect detection performance, with XGBoost, Random Forest and LightGBM reaching 99.61% accuracy, AUC-ROC above 0.999 and F1-scores of 0.9925 on the test set, while maintaining extremely low false positive rates (<0.23%) and false negative rates (~0.8%). In contrast, the logistic regression baseline performs substantially worse (89.97% accuracy, F1-score 0.8314). Feature importance and SHAP analyses identify absence of personal signals and repetition patterns as the dominant predictors. Additionally, the RAG component achieves 90.5% top-1 retrieval accuracy, with strong same-class clustering in the embedding space, further supporting the reliability of the framework's outputs.

Data mixing strategy is essential for large language model (LLM) training. Empirical evidence shows that inappropriate strategies can significantly reduce generalization. Although recent methods have improved empirical performance, several fundamental questions remain open: what constitutes a domain, whether human and model perceptions of domains are aligned, and how domain weighting influences generalization. We address these questions by establishing formal connections between gradient dynamics and domain distributions, offering a theoretical framework that clarifies the role of domains in training dynamics. Building on this analysis, we introduce DoGraph, a reweighting framework that formulates data scheduling as a graph-constrained optimization problem. Extensive experiments on GPT-2 models of varying scales demonstrate that DoGraph consistently achieves competitive performance.

As models are getting larger and are trained on increasing amounts of data, there has been an explosion of interest into how we can ``delete'' specific data points or behaviours from a trained model, after the fact. This goal has been referred to as ``machine unlearning''. In this note, we argue that the term ``unlearning'' has been overloaded, with different research efforts spanning two distinct problem formulations, but without that distinction having been observed or acknowledged in the literature. This causes various issues, including ambiguity around when an algorithm is expected to work, use of inappropriate metrics and baselines when comparing different algorithms to one another, difficulty in interpreting results, as well as missed opportunities for pursuing critical research directions. In this note, we address this issue by establishing a fundamental distinction between two notions that we identify as \unlearning and \untraining, illustrated in Figure 1. In short, \untraining aims to reverse the effect of having trained on a given forget set, i.e. to remove the influence that that specific forget set examples had on the model during training. On the other hand, the goal of \unlearning is not just to remove the influence of those given examples, but to use those examples for the purpose of more broadly removing the entire underlying distribution from which those examples were sampled (e.g. the concept or behaviour that those examples represent). We discuss technical definitions of these problems and map problem settings studied in the literature to each. We hope to initiate discussions on disambiguating technical definitions and identify a set of overlooked research questions, as we believe that this a key missing step for accelerating progress in the field of ``unlearning''.

Computer-Aided Design (CAD) is an expert-level task that relies on long-horizon reasoning and coherent modeling actions. Large Language Models (LLMs) have shown remarkable advancements in enabling language agents to tackle real-world tasks. Notably, there has been no investigation into how tool-using LLMs optimally interact with CAD engines, hindering the emergence of LLM-based agentic text-to-CAD modeling systems. We propose ToolCAD, a novel agentic CAD framework deploying LLMs as tool-using agents for text-to-CAD generation. Furthermore, we introduce an interactive CAD modeling gym to rollout reasoning and tool-augmented interaction trajectories with the CAD engine, incorporating hybrid feedback and human supervision. Meanwhile, an end-to-end post-training strategy is presented to enable the LLM agent to elicit refined CAD Modeling Chain of Thought (CAD-CoT) and evolve into proficient CAD tool-using agents via online curriculum reinforcement learning. Our findings demonstrate ToolCAD fills the gap in adopting and training open-source LLMs for CAD tool-using agents, enabling them to perform comparably to proprietary models, paving the way for more accessible and robust autonomous text-to-CAD modeling systems.

Vision-language models (VLMs) and generative world models are opening new opportunities for embodied navigation. VLMs are increasingly used as direct planners or trajectory predictors, while world models support look-ahead reasoning by imagining future views. Yet predicting a reliable trajectory from a single egocentric observation remains challenging. Current VLMs often generate unstable trajectories, and world models, though able to synthesize plausible futures, do not directly provide the grounded signals needed for navigation learning. This raises a central question: how can generated futures be turned into supervision for grounded trajectory prediction? We present WorldMAP, a teacher--student framework that converts world-model-generated futures into persistent semantic-spatial structure and planning-derived supervision. Its world-model-driven teacher builds semantic-spatial memory from generated videos, grounds task-relevant targets and obstacles, and produces trajectory pseudo-labels through explicit planning. A lightweight student with a multi-hypothesis trajectory head is then trained to predict navigation trajectories directly from vision-language inputs. On Target-Bench, WorldMAP achieves the best ADE and FDE among compared methods, reducing ADE by 18.0% and FDE by 42.1% relative to the best competing baseline, while lifting a small open-source VLM to DTW performance competitive with proprietary models. More broadly, the results suggest that, in embodied navigation, the value of world models may lie less in supplying action-ready imagined evidence than in synthesizing structured supervision for navigation learning.

Industry classification schemes are integral parts of public and corporate databases as they classify businesses based on economic activity. Due to the size of the company registers, manual annotation is costly, and fine-tuning models with every update in industry classification schemes requires significant data collection. We replicate the manual expert verification by using existing or easily retrievable multimodal resources for industry classification. We present MONETA, the first multimodal industry classification benchmark with text (Website, Wikipedia, Wikidata) and geospatial sources (OpenStreetMap and satellite imagery). Our dataset enlists 1,000 businesses in Europe with 20 economic activity labels according to EU guidelines (NACE). Our training-free baseline reaches 62.10% and 74.10% with open and closed-source Multimodal Large Language Models (MLLM). We observe an increase of up to 22.80% with the combination of multi-turn design, context enrichment, and classification explanations. We will release our dataset and the enhanced guidelines.

Methods based on weight compensation, which iteratively apply quantization and weight compensation to minimize the output error, have recently demonstrated remarkable success in quantizing Large Language Models (LLMs). The representative work, GPTQ, introduces several key techniques that make such iterative methods practical for LLMs with billions of parameters. GPTAQ extends this approach by introducing an asymmetric calibration process that aligns the output of each quantized layer with its full-precision counterpart, incorporating a residual error into the weight compensation framework. In this work, we revisit the formulation of the residual error. We identify a sub-optimal calibration objective in existing methods: during the intra-layer calibration process, they align the quantized output with the output from compensated weights, rather than the true output from the original full-precision model. Therefore, we redefine the objective to precisely align the quantized model's output with the original output of the full-precision model at each step. We then reveal that the residual error originates not only from the output difference of the preceding layer but also from the discrepancy between the compensated and original weights within each layer, which we name the 'compensation-aware error'. By inheriting the neuron decomposition technique from GPTAQ, we can efficiently incorporate this compensation-aware error into the weight update process. Extensive experiments on various LLMs and quantization settings demonstrate that our proposed enhancements integrate seamlessly with both GPTQ and GPTAQ, significantly improving their quantization performance. Our code is publicly available at https://github.com/list0830/ResComp.

Time series classification (TSC) enables important use cases, however lacks a unified understanding of performance trade-offs across models, datasets, and hardware. While resource awareness has grown in the field, TSC methods have not yet been rigorously evaluated for energy efficiency. This paper introduces a holistic evaluation framework that explicitly explores the balance of predictive performance and resource consumption in TSC. To boost efficiency, we apply a theoretically bounded pruning strategy to leading hybrid classifiers - Hydra and Quant - and present Hydrant, a novel, prunable combination of both. With over 4000 experimental configurations across 20 MONSTER datasets, 13 methods, and three compute setups, we systematically analyze how model design, hyperparameters, and hardware choices affect practical TSC performance. Our results showcase that pruning can significantly reduce energy consumption by up to 80% while maintaining competitive predictive quality, usually costing the model less than 5% of accuracy. The proposed methodology, experimental results, and accompanying software advance TSC toward sustainable and reproducible practice.

The expansion of digital payment systems has heightened both the scale and intricacy of online financial transactions, thereby increasing vulnerability to fraudulent activities. Detecting fraud effectively is complicated by the changing nature of attack strategies and the significant disparity between genuine and fraudulent transactions. This research introduces a machine learning-based fraud detection framework utilizing the PaySim synthetic financial transaction dataset. Following the CRISP-DM methodology, the study includes hypothesis-driven exploratory analysis, feature refinement, and a comparative assessment of baseline models such as Logistic Regression and tree-based classifiers like Random Forest, XGBoost, and Decision Tree. To tackle class imbalance, SMOTE is employed, and model performance is enhanced through hyperparameter tuning with GridSearchCV. The proposed framework provides a robust and scalable solution to enhance fraud prevention capabilities in FinTech transaction systems. Keywords: fraud detection, imbalanced data, HPO, SMOTE

Efficient ground state search is fundamental to advancing combinatorial optimization problems and quantum chemistry. While the Variational Imaginary Time Evolution (VITE) method offers a useful alternative to Variational Quantum Eigensolver (VQE), and Quantum Approximate Optimization Algorithm (QAOA), its implementation on Noisy Intermediate-Scale Quantum (NISQ) devices is severely limited by the gate counts and depth of manually designed ansatz. Here, we present an automated framework for VITE circuit design using Double Deep-Q Networks (DDQN). Our approach treats circuit construction as a multi-objective optimization problem, simultaneously minimizing energy expectation values and optimizing circuit complexity. By introducing adoptive thresholds, we demonstrate significant hardware overhead reductions. In Max-Cut problems, our agent autonomously discovered circuits with approximately 37\% fewer gates and 43\% less depth than standard hardware-efficient ansatz on average. For molecular hydrogen ($H_2$), the DDQN also achieved the Full-CI limit, with maintaining a significantly shallower circuit. These results suggest that deep reinforcement learning can be helpful to find non-intuitive, optimal circuit structures, providing a pathway toward efficient, hardware-aware quantum algorithm design.

As the demand for mobile robots continues to increase, social navigation has emerged as a critical task, driving active research into deep reinforcement learning (RL) approaches. However, because pedestrian dynamics and social conventions vary widely across different regions, simulations cannot easily encompass all possible real-world scenarios. Real-world RL, in which agents learn while operating directly in physical environments, presents a promising solution to this issue. Nevertheless, this approach faces significant challenges, particularly regarding constrained computational resources on edge devices and learning efficiency. In this study, we propose incremental residual RL (IRRL). This method integrates incremental learning, which is a lightweight process that operates without a replay buffer or batch updates, with residual RL, which enhances learning efficiency by training only on the residuals relative to a base policy. Through the simulation experiments, we demonstrated that, despite lacking a replay buffer, IRRL achieved performance comparable to those of conventional replay buffer-based methods and outperformed existing incremental learning approaches. Furthermore, the real-world experiments confirmed that IRRL can enable robots to effectively adapt to previously unseen environments through the real-world learning.

Large language models (LLMs) have recently demonstrated strong potential for autonomous vehicle motion planning by reformulating trajectory prediction as a language generation problem. However, deploying capable LLMs in resource-constrained onboard systems remains a fundamental challenge. In this paper, we study how to effectively transfer motion planning knowledge from a large teacher LLM to a smaller, more deployable student model. We build on the GPT-Driver framework, which represents driving scenes as language prompts and generates waypoint trajectories with chain-of-thought reasoning, and investigate two student training paradigms: (i) on-policy generalized knowledge distillation (GKD), which trains the student on its own self-generated outputs using dense token-level feedback from the teacher, and (ii) a dense-feedback reinforcement learning (RL) baseline that uses the teacher's log-probabilities as per-token reward signals in a policy gradient framework. Experiments on the nuScenes benchmark show that GKD substantially outperforms the RL baseline and closely approaches teacher-level performance despite a 5$\times$ reduction in model size. These results highlight the practical value of on-policy distillation as a principled and effective approach to deploying LLM-based planners in autonomous driving systems.

Post-training has become central to turning pretrained large language models (LLMs) into aligned and deployable systems. Recent progress spans supervised fine-tuning (SFT), preference optimization, reinforcement learning (RL), process supervision, verifier-guided methods, distillation, and multi-stage pipelines. Yet these methods are often discussed in fragmented ways, organized by labels or objective families rather than by the behavioral bottlenecks they address. This survey argues that LLM post-training is best understood as structured intervention on model behavior. We organize the field first by trajectory provenance, which defines two primary learning regimes: off-policy learning on externally supplied trajectories, and on-policy learning on learner-generated rollouts. We then interpret methods through two recurring roles -- effective support expansion, which makes useful behaviors more reachable, and policy reshaping, which improves behavior within already reachable regions -- together with a complementary systems-level role, behavioral consolidation, which preserves, transfers, and amortizes behavior across stages and model transitions. This perspective yields a unified reading of major paradigms. SFT may serve either support expansion or policy reshaping, whereas preference-based methods are usually off-policy reshaping. On-policy RL often improves behavior on learner-generated states, though under stronger guidance it can also make hard-to-reach reasoning paths reachable. Distillation is often best understood as consolidation rather than only compression, and hybrid pipelines emerge as coordinated multi-stage compositions. Overall, the framework helps diagnose post-training bottlenecks and reason about stage composition, suggesting that progress in LLM post-training increasingly depends on coordinated system design rather than any single dominant objective.

Tabular data, widely used in various applications such as industrial control systems, finance, and supply chain, often contains complex interrelationships among its attributes. Data disentanglement seeks to transform such data into latent variables with reduced interdependencies, facilitating more effective and efficient processing. Despite the extensive studies on data disentanglement over image, text, or audio data, tabular data disentanglement may require further investigation due to the more intricate attribute interactions typically found in tabular data. Moreover, due to the highly complex interrelationships, direct translation from other data domains results in suboptimal data disentanglement. Existing tabular data disentanglement methods, such as factor analysis, CT-GAN, and VAE face limitations including scalability issues, mode collapse, and poor extrapolation. In this paper, we propose the use of a framework to provide a systematic view on tabular data disentanglement that modularizes the process into four core components: data extraction, data modeling, model analysis, and latent representation extrapolation. We believe this work provides a deeper understanding of tabular data disentanglement and existing methods, and lays the foundation for potential future research in developing robust, efficient, and scalable data disentanglement techniques. Finally, we demonstrate the framework's applicability through a case study on synthetic tabular data generation, showcasing its potential in the particular downstream task of data synthesis.

Cross-document relation extraction (RE) aims to identify relations between the head and tail entities located in different documents. Existing approaches typically adopt the paradigm of ``\textit{Small Language Model (SLM) + Classifier}''. However, the limited language understanding ability of SLMs hinders further improvement of their performance. In this paper, we conduct a preliminary study to explore the performance of Large Language Models (LLMs) in cross-document RE. Despite their extensive parameters, our findings indicate that LLMs do not consistently surpass existing SLMs. Further analysis suggests that the underperformance is largely attributed to the challenges posed by the numerous predefined relations. To overcome this issue, we propose an LLM-based \underline{H}ierarchical \underline{C}lassification model for cross-document \underline{RE} (HCRE), which consists of two core components: 1) an LLM for relation prediction and 2) a \textit{hierarchical relation tree} derived from the predefined relation set. This tree enables the LLM to perform hierarchical classification, where the target relation is inferred level by level. Since the number of child nodes is much smaller than the size of the entire predefined relation set, the hierarchical relation tree significantly reduces the number of relation options that LLM needs to consider during inference. However, hierarchical classification introduces the risk of error propagation across levels. To mitigate this, we propose a \textit{prediction-then-verification} inference strategy that improves prediction reliability through multi-view verification at each level. Extensive experiments show that HCRE outperforms existing baselines, validating its effectiveness.

The Hardware Lottery posits that research directions are dictated by available silicon compute platforms. We identify a derivative phenomenon, the Hyperscale Lottery, where model architectures are optimized for cloud throughput at the expense of algorithmic efficiency. While State-Space Models (SSMs) such as Mamba were lauded for their linear complexity, ideal for edge intelligence, their evolution from Mamba-1 to Mamba-3 reveals a systematic divergence from edge-native efficiency. We demonstrate that Mamba-3's architectural changes, designed to saturate hyperscale GPUs, impose a significant edge penalty: a 28% latency increase at 880M parameters, worsening to 48% for 15M-parameter models. We argue for decoupling cloud-scale saturation strategies from core architectural design to preserve the viability of single-user, real-time edge intelligence.

This paper presents Top Management Journal Portal, a deployable web artifact for searching, monitoring, and interpreting literature from elite business and management journals. The system integrates the UTD-24 and Financial Times 50 (FT50) journal pools, retrieves live article metadata from the Cross- ref REST API, and organizes scholarly work into an end-to-end workflow spanning query formulation, result filtering, hotspot extraction, citation export, favorites management, and usage analytics. Unlike static journal directories or general-purpose academic search engines, the artifact is explicitly scoped to high-status management outlets and is designed to support sensemaking tasks that matter to researchers, doctoral students, and lab managers: identifying recent work, surfacing topical concentration, and converting search output into actionable research material. Architecturally, the system emphasizes source transparency, modularity, and low-cost public deployability through a lightweight Node.js service layer, a multi-page client interface, optional large-language-model enhancement for hotspot rewriting, and a free-tier persistence path through Supabase. The paper contributes both a functioning design artifact and an extensible architectural pattern for journal-pool-specific scholarly discovery, with implications for digital research infrastructure in information systems and business scholarship.

Output-length prediction is important for efficient LLM serving, as it directly affects batching, memory reservation, and scheduling. For prompt-only length prediction, most existing methods use a one-shot sampled length as the label, implicitly treating each prompt as if it had one true target length. We show that this is unreliable: even under a fixed model and decoding setup, the same prompt induces a \emph{prompt-conditioned output length distribution}, not a deterministic scalar, and this distribution is consistent with \emph{heavy-tailed} behavior. Motivated by this, we cast length prediction as robust estimation from heavy-tailed prompt-conditioned length distributions. We propose prompt-conditioned length distribution (ProD) methods, which construct training targets from multiple independent generations of the same prompt. Two variants are developed to reuse the served LLM's hidden states: \mbox{ProD-M}, which uses a median-based target for robust point prediction, and ProD-D, which uses a distributional target that preserves prompt-conditioned uncertainty. We provide theoretical justifications by analyzing the estimation error under a surrogate model. Experiments across diverse scenarios show consistent gains in prediction quality.

LLM-driven GUI agents are increasingly used in production systems to automate workflows and simulate users for evaluation and optimization. Yet most GUI-agent evaluations emphasize task success and provide limited evidence on whether agents interact in human-like ways. We present a trace-level evaluation framework that compares human and agent behavior across (i) task outcome and effort, (ii) query formulation, and (iii) navigation across interface states. We instantiate the framework in a controlled study in a production audio-streaming search application, where 39 participants and a state-of-the-art GUI agent perform ten multi-hop search tasks. The agent achieves task success comparable to participants and generates broadly aligned queries, but follows systematically different navigation strategies: participants exhibit content-centric, exploratory behavior, while the agent is more search-centric and low-branching. These results show that outcome and query alignment do not imply behavioral alignment, motivating trace-level diagnostics when deploying GUI agents as proxies for users in production search systems.

While AI-based numerical weather prediction (NWP) enables rapid forecasting, generating high-resolution outputs remains computationally demanding due to limited multi-scale adaptability and inefficient data representations. We propose the 3D Gaussian splatting-based scale-aware vision transformer (GSSA-ViT), a novel framework for arbitrary-resolution forecasting and flexible downscaling of high-dimensional atmospheric fields. Specifically, latitude-longitude grid points are treated as centers of 3D Gaussians. A generative 3D Gaussian prediction scheme is introduced to estimate key parameters, including covariance, attributes, and opacity, for unseen samples, improving generalization and mitigating overfitting. In addition, a scale-aware attention module is designed to capture cross-scale dependencies, enabling the model to effectively integrate information across varying downscaling ratios and support continuous resolution adaptation. To our knowledge, this is the first NWP approach that combines generative 3D Gaussian modeling with scale-aware attention for unified multi-scale prediction. Experiments on ERA5 show that the proposed method accurately forecasts 87 atmospheric variables at arbitrary resolutions, while evaluations on ERA5 and CMIP6 demonstrate its superior performance in downscaling tasks. The proposed framework provides an efficient and scalable solution for high-resolution, multi-scale atmospheric prediction and downscaling. Code is available at: https://github.com/binbin2xs/weather-GS.

Deep research requires reasoning over web evidence to answer open-ended questions, and it is a core capability for AI agents. Yet many deep research agents still rely on implicit, unstructured search behavior that causes redundant exploration and brittle evidence aggregation. Motivated by Anthropic's "think" tool paradigm and insights from the information-retrieval literature, we introduce Q+, a set of query and evidence processing tools that make web search more deliberate by guiding query planning, monitoring search progress, and extracting evidence from long web snapshots. We integrate Q+ into the browser sub-agent of Eigent, an open-source, production-ready multi-agent workforce for computer use, yielding EigentSearch-Q+. Across four benchmarks (SimpleQA-Verified, FRAMES, WebWalkerQA, and X-Bench DeepSearch), Q+ improves Eigent's browser agent benchmark-size-weighted average accuracy by 3.0, 3.8, and 0.6 percentage points (pp) for GPT-4.1, GPT-5.1, and Minimax M2.5 model backends, respectively. Case studies further suggest that EigentSearch-Q+ produces more coherent tool-calling trajectories by making search progress and evidence handling explicit.

The self-attention mechanism is central to the success of Transformer architectures. However, standard row-stochastic attention has been shown to suffer from significant signal degradation across layers. In particular, it can induce rank collapse, resulting in increasingly uniform token representations, as well as entropy collapse, characterized by highly concentrated attention distributions. Recent work has highlighted the benefits of doubly stochastic attention as a form of entropy regularization, promoting a more balanced attention distribution and leading to improved empirical performance. In this paper, we study rank collapse across network depth and show that doubly stochastic attention matrices normalized with Sinkhorn algorithm preserve rank more effectively than standard Softmax row-stochastic ones. As previously shown for Softmax, skip connections are crucial to mitigate rank collapse. We empirically validate this phenomenon on both sentiment analysis and image classification tasks. Moreover, we derive a theoretical bound for the pure self-attention rank decay when using Sinkhorn normalization and find that rank decays to one doubly exponentially with depth, a phenomenon that has already been shown for Softmax.

Large Reasoning Models (LRMs) have revolutionized complex problem-solving, yet they exhibit a pervasive "overthinking", generating unnecessarily long reasoning chains. While current solutions improve token efficiency, they often sacrifice fine-grained control or risk disrupting the logical integrity of the reasoning process. To address this, we introduce Stepwise Adaptive Thinking (SAT), a framework that performs step-level, difficulty-aware pruning while preserving the core reasoning structure. SAT formulates reasoning as a Finite-State Machine (FSM) with distinct thinking modes (Slow, Normal, Fast, Skip). It navigates these states dynamically using a lightweight Process Reward Model (PRM), compressing easy steps while preserving depth for hard ones. Experiments across 9 LRMs and 7 benchmarks show that SAT achieves up to 40% reduction in reasoning tokens while generally maintaining or improving accuracy.

Software redesign preserves functionality while improving quality attributes, but manual reuse of code and tests is costly and error-prone, especially in crossrepository redesigns. Focusing on static analyzers where cross-repo redesign needs often arise, we conduct a bidirectional study of the ongoing Soot/SootUp redesign case using an action research methodology that combines empirical investigation with validated open-source contributions. Our study reveals: (1) non-linear migration which necessitates bidirectional reuse, (2) deferred reuse via TODOs, (3) neglected test porting, and (4) residual bug propagation during migrations. We identify tracking corresponding code and tests as the key challenge, and address it by retrofitting clone detection to derive code mappings between original and redesigned projects. Guided by semantic reuse patterns derived in our study, we propose Semantic Alignment Heuristics and a scalable hierarchical detection strategy. Evaluations on two redesigned project pairs (Soot/SootUp and FindBugs/SpotBugs) show that our approach achieves an average reduction of 33-99% in likely irrelevant clones at a SAS threshold of 0.5 across all tool results, and improves precision up to 86% on our benchmark of 1,749 samples. Moreover, we contribute to the redesigned projects by submitting five issues and 10 pull requests, of which eight have been merged.

Large Vision-Language Models (LVLMs) have achieved remarkable success across cross-modal tasks but remain hindered by hallucinations, producing textual outputs inconsistent with visual content. Existing methods mitigate hallucinations but often alter generation behavior, resulting in shorter outputs and shifted token distributions, especially in latent space steering approaches. We identify that this issue stems from entangled steering signals, where suppressing hallucinations inadvertently disrupts the model's intrinsic generation behavior. To address this, we propose MESA, an effective plug-and-play framework that performs controlled and selective latent intervention for hallucination mitigation. Specifically, MESA targets hallucination-relevant responses while preserving the model's original token distribution, enabling effective hallucination reduction without compromising generation behavior. Extensive experiments across diverse generative and discriminative benchmarks demonstrate that MESA consistently reduces hallucinations while better preserving generation behavior, outperforming prior methods across multiple LVLM families.

Multi-agent LLM orchestration systems suffer from context pollution: when N concurrent agents compete for the orchestrator's context window, each agent's task state, partial outputs, and pending questions contaminate the steering interactions of every other agent, degrading decision quality. We introduce Dynamic Attentional Context Scoping (DACS), a mechanism in which the orchestrator operates in two asymmetric modes. In Registry mode it holds only lightweight per-agent status summaries (<=200 tokens each), remaining responsive to all agents and the user. When an agent emits a SteeringRequest, the orchestrator enters Focus(a_i) mode, injecting the full context of agent a_i while compressing all other agents to their registry entries. Context isolation is agent-triggered, asymmetric, and deterministic: the context window contains exactly F(a_i) + R_{-i} during steering, eliminating cross-agent contamination without requiring context compression or retrieval. We evaluate DACS across four experimental phases totalling 200 trials: Phase 1 tests N in {3,5,10} (60 trials); Phase 2 tests agent heterogeneity and adversarial dependencies (60 trials); Phase 3 tests decision density up to D=15 (40 trials); Phase 4 uses autonomous LLM agents for free-form questions (40 trials, Claude Haiku 4.5). Across all 8 synthetic scenarios, DACS achieves 90.0--98.4% steering accuracy versus 21.0--60.0% for a flat-context baseline (p < 0.0001 throughout), with wrong-agent contamination falling from 28--57% to 0--14% and context efficiency ratios of up to 3.53x. The accuracy advantage grows with N and D; keyword matching is validated by LLM-as-judge across all phases (mean kappa=0.909). DACS outperforms the flat-context baseline by +17.2pp at N=3 (p=0.0023) and +20.4pp at N=5 (p=0.0008) in Phase 4, with the advantage growing with N confirmed by two independent judges.

We present the Capture-Quiet Decomposition (CQD), a structural theorem for verifying Win-Draw-Loss (WDL) labelings of chess endgame tablebases. The theorem decomposes every legal position into exactly one of three categories -- terminal, capture, or quiet -- and shows that a WDL labeling is correct if and only if: (1) terminal positions are labeled correctly, (2) capture positions are consistent with verified sub-models of smaller piece count, and (3) quiet positions satisfy retrograde consistency within the same endgame. The key insight is that capture positions anchor the labeling to externally verified sub-models, breaking the circularity that allows trivial fixpoints (such as the all-draw labeling) to satisfy self-consistency alone. We validate CQD exhaustively on all 35 three- and four-piece endgames (42 million positions), all 110 five-piece endgames, and all 372 six-piece endgames -- 517 endgames in total -- with the decomposed verifier producing identical violation counts to a full retrograde baseline in every case.

Spatiotemporal neural dynamics and oscillatory synchronization are widely implicated in biological information processing and have been hypothesized to support flexible coordination such as feature binding. By contrast, most deep learning architectures represent and propagate information through activation values, neglecting the joint dynamics of rate and phase. In this work, we introduce Kuramoto oscillatory Phase Encoding (KoPE) as an additional, evolving phase state to Vision Transformers, incorporating a neuro-inspired synchronization mechanism to advance learning efficiency. We show that KoPE can improve training, parameter, and data efficiency of vision models through synchronization-enhanced structure learning. Moreover, KoPE benefits tasks requiring structured understanding, including semantic and panoptic segmentation, representation alignment with language, and few-shot abstract visual reasoning (ARC-AGI). Theoretical analysis and empirical verification further suggest that KoPE can accelerate attention concentration for learning efficiency. These results indicate that synchronization can serve as a scalable, neuro-inspired mechanism for advancing state-of-the-art neural network models.

Granlund and Montgomery proposed an optimization method for unsigned integer division by constants [3]. Their method (called the GM method in this paper) was further improved in part by works such as [1] and [7], and is now adopted by major compilers including GCC, Clang, Microsoft Compiler, and Apple Clang. However, for example, for x/7, the generated code is designed for 32-bit CPUs and therefore does not fully exploit 64-bit capabilities. This paper proposes an optimization method for 32-bit unsigned division by constants targeting 64-bit CPUs. We implemented patches for LLVM/GCC and achieved speedups of 1.67x on Intel Xeon w9-3495X (Sapphire Rapids) and 1.98x on Apple M4 (Apple M-series SoC) in the microbenchmark described later. The LLVM patch has already been merged into llvm:main [6], demonstrating the practical applicability of the proposed method.

Industrial anomaly generation is a crucial method for alleviating the data scarcity problem in anomaly detection tasks. Most existing anomaly synthesis methods rely on single-step generation mechanisms, lacking complex reasoning and iterative optimization capabilities, making it difficult to generate anomaly samples with high semantic realism. We propose AnomalyAgent, an anomaly synthesis agent with self-reflection, knowledge retrieval, and iterative refinement capabilities, aiming to generate realistic and diverse anomalies. Specifically, AnomalyAgent is equipped with five tools: Prompt Generation (PG), Image Generation (IG), Quality Evaluation (QE), Knowledge Retrieval (KR), and Mask Generation (MG), enabling closed-loop optimization. To improve decision-making and self-reflection, we construct structured trajectories from real anomaly images and design a two-stage training framework: supervised fine-tuning followed by reinforcement learning. This process is driven by a three-part reward mechanism: (1) task rewards to supervise the quality and location rationality of generated anomalies; (2) reflection rewards to train the model's ability to improve anomaly synthesis prompt; (3) behavioral rewards to ensure adherence to the trajectory. On the MVTec-AD dataset, AnomalyAgent achieves IS/IC-L of 2.10/0.33 for anomaly generation, 57.0% classification accuracy using ResNet34, and 99.3%/74.2% AP at the image/pixel level using a simple UNet, surpassing all zero-shot SOTA methods. The code and data will be made publicly available.

Learning rules plays a crucial role in deep learning, particularly in explainable artificial intelligence and enhancing the reasoning capabilities of large language models. While existing rule learning methods are primarily designed for symbolic data, learning rules from image data without supporting image labels and automatically inventing predicates remains a challenge. In this paper, we tackle these inductive rule learning problems from images with a framework called γILP, which provides a fully differentiable pipeline from image constant substitution to rule structure induction. Extensive experiments demonstrate that γILP achieves strong performance not only on classical symbolic relational datasets but also on relational image data and pure image datasets, such as Kandinsky patterns.

Quantum kernel methods (QKMs) have emerged as a prominent framework for supervised quantum machine learning. Unlike variational quantum algorithms, which rely on gradient-based optimisation and may suffer from issues such as barren plateaus, non-variational QKMs employ fixed quantum feature maps, with model selection performed classically via convex optimisation and cross-validation. This separation of quantum feature embedding from classical training ensures stable optimisation while leveraging quantum circuits to encode data in high-dimensional Hilbert spaces. In this review, we provide a thorough analysis of non-variational supervised QKMs, covering their foundations in classical kernel theory, constructions of fidelity and projected quantum kernels, and methods for their estimation in practice. We examine frameworks for assessing quantum advantage, including generalisation bounds and necessary conditions for separation from classical models, and analyse key challenges such as exponential concentration, dequantisation via tensor-network methods, and the spectral properties of kernel integral operators. We further discuss structured problem classes that may enable advantage, and synthesise insights from comparative and hardware studies. Overall, this review aims to clarify the regimes in which QKMs may offer genuine advantages, and to delineate the conceptual, methodological, and technical obstacles that must be overcome for practical quantum-enhanced learning.

Selecting an appropriate background music (BGM) that supports natural human conversation is a common production step in media and interactive systems. In this paper, we introduce dialogue-conditioned BGM recommendation, where a model should select non-intrusive, fitting music for a multi-turn conversation that often contains no music descriptors. To study this novel problem, we present DialBGM, a benchmark of 1,200 open-domain daily dialogues, each paired with four candidate music clips and annotated with human preference rankings. Rankings are determined by background suitability criteria, including contextual relevance, non-intrusiveness, and consistency. We evaluate a wide range of open-source and proprietary models, including audio-language models and multimodal LLMs, and show that current models fall far short of human judgments; no model exceeds 35% Hit@1 when selecting the top-ranked clip. DialBGM provides a standardized benchmark for developing discourse-aware methods for BGM selection and for evaluating both retrieval-based and generative models.

Personalized large language models (PLLMs) have garnered significant attention for their ability to align outputs with individual's needs and preferences. However, they still struggle with long-horizon tasks, such as tracking a user's extensive history of conversations or activities. Existing memory mechanisms often fail to capture evolving behaviors, and RAG paradigms are trapped by a quality-efficiency tradeoff. Meanwhile, parametric adaptation is bottlenecked by train-inference gap due to the scarcity of labeled data. To enhance the long-horizon capabilities of PLLMs, we introduce TSUBASA, a two-pronged approach designed to improve memory writing via dynamic memory evolution, and memory reading via self-learning with a context distillation objective to internalize user experiences. Extensive evaluations on long-horizon benchmarks using the Qwen-3 model family (4B to 32B) validate the effectiveness of TSUBASA, surpassing competitive memory-augmented systems that rely primarily on memory writing, such as Mem0 and Memory-R1. Our analyses further confirms that TSUBASA breaks the quality-efficiency barrier to achieve Pareto improvements, delivering robust, high-fidelity personalization with a reduced token budget.

Instruction-tuned language models increasingly rely on large multi-turn dialogue corpora, but these datasets are often noisy and structurally inconsistent, with topic drift, repetitive chitchat, and mismatched answer formats across turns. We address this from a data selection perspective and propose \textbf{MDS} (Multi-turn Dialogue Selection), a dialogue-level framework that scores whole conversations rather than isolated turns. MDS combines a global coverage stage that performs bin-wise selection in the user-query trajectory space to retain representative yet non-redundant dialogues, with a local structural stage that evaluates within-dialogue reliability through entity-grounded topic grounding and information progress, together with query-answer form consistency for functional alignment. MDS outperforms strong single-turn selectors, dialogue-level LLM scorers, and heuristic baselines on three multi-turn benchmarks and an in-domain Banking test set, achieving the best overall rank across reference-free and reference-based metrics, and is more robust on long conversations under the same training budget. Code and resources are included in the supplementary materials.

Application Programming Interfaces (APIs) are crucial to software development, enabling integration of existing systems with new applications by reusing tried and tested code, saving development time and increasing software safety. In particular, the Java standard library APIs, along with numerous third-party APIs, are extensively utilized in the development of enterprise application software. However, their misuse remains a significant source of bugs and vulnerabilities. Furthermore, due to the limited examples in the official API documentation, developers often rely on online portals and generative AI models to learn unfamiliar APIs, but using such examples may introduce unintentional errors in the software. In this paper, we present AFGNN, a novel Graph Neural Network (GNN)-based framework for efficiently detecting API misuses in Java code. AFGNN uses a novel API Flow Graph (AFG) representation that captures the API execution sequence, data, and control flow information present in the code to model the API usage patterns. AFGNN uses self-supervised pre-training with AFG representation to effectively compute the embeddings for unknown API usage examples and cluster them to identify different usage patterns. Experiments on popular API usage datasets show that AFGNN significantly outperforms state-of-the-art small language models and API misuse detectors.

Training LLMs at ultra-low precision remains a formidable challenge. Direct low-bit QAT often suffers from convergence instability and substantial training costs, exacerbated by quantization noise from heavy-tailed outlier channels and error accumulation across layers. To address these issues, we present Bit-by-Bit, a progressive QAT framework with outlier channel splitting. Our approach integrates three key components: (1) block-wise progressive training that reduces precision stage by stage, ensuring stable initialization for low-bit optimization; (2) nested structure of integer quantization grids to enable a "train once, deploy any precision" paradigm, allowing a single model to support multiple bit-widths without retraining; (3) rounding-aware outlier channel splitting, which mitigates quantization error while acting as an identity transform that preserves the quantized outputs. Furthermore, we follow microscaling groups with E4M3 scales, capturing dynamic activation ranges in alignment with OCP/NVIDIA standards. To address the lack of efficient 2-bit kernels, we developed custom operators for both W2A2 and W2A16 configurations, achieving up to 11$\times$ speedup over BF16. Under W2A2 settings, Bit-by-Bit significantly outperforms baselines like BitDistiller and EfficientQAT on both Llama2/3, achieving a loss of only 2.25 WikiText2 PPL compared to full-precision models.

We investigate how and to what extent hierarchical relations (e.g., Japan $\subset$ Eastern Asia $\subset$ Asia) are encoded in the internal representations of language models. Building on Linear Relational Concepts, we train linear transformations specific to each hierarchical depth and semantic domain, and characterize representational differences associated with hierarchical relations by comparing these transformations. Going beyond prior work on the representational geometry of hierarchies in LMs, our analysis covers multi-token entities and cross-layer representations. Across multiple domains we learn such transformations and evaluate in-domain generalization to unseen data and cross-domain transfer. Experiments show that, within a domain, hierarchical relations can be linearly recovered from model representations. We then analyze how hierarchical information is encoded in representation space. We find that it is encoded in a relatively low-dimensional subspace and that this subspace tends to be domain-specific. Our main result is that hierarchy representation is highly similar across these domain-specific subspaces. Overall, we find that all models considered in our experiments encode concept hierarchies in the form of highly interpretable linear representations.

This work explores the connection between (non-)literalness and plausibility at the example of subject-verb-object events in English. We design a systematic setup of plausible and implausible event triples in combination with abstract and concrete constituent categories. Our analysis of human and LLM-generated judgments and example contexts reveals substantial differences between assessments of plausibility. While humans excel at nuanced detection and contextualization of (non-)literal vs. implausible events, LLM results reveal only shallow contextualization patterns with a bias to trade implausibility for non-literal, plausible interpretations.

High-fidelity generative models are increasingly needed in privacy-sensitive scenarios, where access to data is severely restricted due to regulatory and copyright constraints. This scarcity hampers model development--ironically, in settings where generative models are most needed to compensate for the lack of data. This creates a self-reinforcing challenge: limited data leads to poor generative models, which in turn fail to mitigate data scarcity. To break this cycle, we propose a reinforcement-guided synthetic data generation framework that adapts general-domain generative priors to privacy-sensitive identity recognition tasks. We first perform a cold-start adaptation to align a pretrained generator with the target domain, establishing semantic relevance and initial fidelity. Building on this foundation, we introduce a multi-objective reward that jointly optimizes semantic consistency, coverage diversity, and expression richness, guiding the generator to produce both realistic and task-effective samples. During downstream training, a dynamic sample selection mechanism further prioritizes high-utility synthetic samples, enabling adaptive data scaling and improved domain alignment. Extensive experiments on benchmark datasets demonstrate that our framework significantly improves both generation fidelity and classification accuracy, while also exhibiting strong generalization to novel categories in small-data regimes.

History textbooks often contain implicit biases, nationalist framing, and selective omissions that are difficult to audit at scale. We propose an agentic evaluation architecture comprising a multimodal screening agent, a heterogeneous jury of five evaluative agents, and a meta-agent for verdict synthesis and human escalation. A central contribution is a Source Attribution Protocol that distinguishes textbook narrative from quoted historical sources, preventing the misattribution that causes systematic false positives in single-model evaluators. In an empirical study on Romanian upper-secondary history textbooks, 83.3\% of 270 screened excerpts were classified as pedagogically acceptable (mean severity 2.9/7), versus 5.4/7 under a zero-shot baseline, demonstrating that agentic deliberation mitigates over-penalization. In a blind human evaluation (18 evaluators, 54 comparisons), the Independent Deliberation configuration was preferred in 64.8\% of cases over both a heuristic variant and the zero-shot baseline. At approximately \$2 per textbook, these results position agentic evaluation architectures as economically viable decision-support tools for educational governance.

Diffusion-based image generation models have advanced rapidly but pose a safety risk due to their potential to generate Not-Safe-For-Work (NSFW) content. Existing NSFW detection methods mainly operate either before or after image generation. Pre-generation methods rely on text prompts and struggle with the gap between prompt safety and image safety. Post-generation methods apply classifiers to final outputs, but they are poorly suited to intermediate noisy images. To address this, we introduce FlowGuard, a cross-model in-generation detection framework that inspects intermediate denoising steps. This is particularly challenging in latent diffusion, where early-stage noise obscures visual signals. FlowGuard employs a novel linear approximation for latent decoding and leverages a curriculum learning approach to stabilize training. By detecting unsafe content early, FlowGuard reduces unnecessary diffusion steps to cut computational costs. Our cross-model benchmark spanning nine diffusion-based backbones shows the effectiveness of FlowGuard for in-generation NSFW detection in both in-distribution and out-of-distribution settings, outperforming existing methods by over 30% in F1 score while delivering transformative efficiency gains, including slashing peak GPU memory demand by over 97% and projection time from 8.1 seconds to 0.2 seconds compared to standard VAE decoding.

Long-term memory is fundamental for personalized and autonomous agents, yet populating it remains a bottleneck. Existing systems treat memory extraction as a one-shot, passive transcription from context to structured entries, which struggles with noisy dialogue, missing references, and cross-turn dependencies, leading to memory pollution, low-value writes, and inconsistency. In this paper, we introduce the MemReader family for active long-term memory extraction in agent systems: MemReader-0.6B, a compact and cost-efficient passive extractor distilled for accurate and schema-consistent structured outputs, and MemReader-4B, an active extractor optimized with Group Relative Policy Optimization (GRPO) to make memory writing decisions. Under a ReAct-style paradigm, MemReader-4B explicitly evaluates information value, reference ambiguity, and completeness before acting, and can selectively write memories, defer incomplete inputs, retrieve historical context, or discard irrelevant chatter. Experiments on LOCOMO, LongMemEval, and HaluMem show that MemReader consistently outperforms existing extraction-based baselines. In particular, MemReader-4B achieves state-of-the-art performance on tasks involving knowledge updating, temporal reasoning, and hallucination reduction. These results suggest that effective agent memory requires not merely extracting more information, but performing reasoning-driven and selective memory extraction to build low-noise and dynamically evolving long-term memory. Furthermore, MemReader has been integrated into MemOS and is being deployed in real-world applications. To support future research and adoption, we release the models and provide public API access.

Designing high-performing metaheuristics for NP-hard combinatorial optimization problems, such as the Vehicle Routing Problem (VRP), remains a significant challenge, often requiring extensive domain expertise and manual tuning. Recent advances have demonstrated the potential of large language models (LLMs) to automate this process through evolutionary search. However, existing methods are largely reactive, relying on immediate performance feedback to guide what are essentially black-box code mutations. Our work departs from this paradigm by introducing Metacognitive Evolutionary Programming (MEP), a framework that elevates the LLM to a strategic discovery agent. Instead of merely reacting to performance scores, MEP compels the LLM to engage in a structured Reason-Act-Reflect cycle, forcing it to explicitly diagnose failures, formulate design hypotheses, and implement solutions grounded in pre-supplied domain knowledge. By applying MEP to evolve core components of the state-of-the-art Hybrid Genetic Search (HGS) algorithm, we discover novel heuristics that significantly outperform the original baseline. By steering the LLM to reason strategically about the exploration-exploitation trade-off, our approach discovers more effective and efficient heuristics applicable across a wide spectrum of VRP variants. Our results show that MEP discovers heuristics that yield significant performance gains over the original HGS baseline, improving solution quality by up to 2.70\% and reducing runtime by over 45\% on challenging VRP variants.

Ensemble methods are frequently used in recommender systems to improve accuracy by combining multiple models. Recent work reports sizable performance gains, but most studies still optimize primarily for accuracy and robustness rather than for energy efficiency. This paper measures accuracy energy trade offs of ensemble techniques relative to strong single models. We run 93 controlled experiments in two pipelines: 1. explicit rating prediction with Surprise (RMSE) and 2. implicit feedback ranking with LensKit (NDCG@10). We evaluate four datasets ranging from 100,000 to 7.8 million interactions (MovieLens 100K, MovieLens 1M, ModCloth, Anime). We compare four ensemble strategies (Average, Weighted, Stacking or Rank Fusion, Top Performers) against baselines and optimized single models. Whole system energy is measured with EMERS using a smart plug and converted to CO2 equivalents. Across settings, ensembles improve accuracy by 0.3% to 5.7% while increasing energy by 19% to 2,549%. On MovieLens 1M, a Top Performers ensemble improves RMSE by 0.96% at an 18.8% energy overhead over SVD++. On MovieLens 100K, an averaging ensemble improves NDCG@10 by 5.7% with 103% additional energy. On Anime, a Surprise Top Performers ensemble improves RMSE by 1.2% but consumes 2,005% more energy (0.21 vs. 0.01 Wh), increasing emissions from 2.6 to 53.8 mg CO2 equivalents, and LensKit ensembles fail due to memory limits. Overall, selective ensembles are more energy efficient than exhaustive averaging,

Recent advances in deep neural networks have achieved state-of-the-art performance across vision and natural language processing tasks. In practice, however, most models are treated as monolithic black-box functions, limiting maintainability, component-wise optimization, and systematic testing and verification. Despite extensive work on pruning and empirical decomposition, the field still lacks a principled semantic notion of when a neural network can be meaningfully decomposed. We introduce neural decompositionality, a formal notion defined as a semantic-preserving abstraction over neural architectures. Our key insight is that decompositionality should be characterized by the preservation of semantic behavior along the model's decision boundary, which governs classification outcomes. This yields a semantic contract between the original model and its components, enabling a rigorous formulation of decomposition. Building on this foundation, we develop a boundary-aware framework, SAVED (Semantic-Aware Verification-Driven Decomposition), which operationalizes the proposed definition. SAVED combines counterexample mining over low logic-margin inputs, probabilistic coverage, and structure-aware pruning to construct decompositions that preserve decision-boundary semantics. We evaluate our approach on CNNs, language Transformers, and Vision Transformers. Results show clear architectural differences: language Transformers largely preserve boundary semantics under decomposition, whereas vision models frequently violate the decompositionality criterion, indicating intrinsic limits. Overall, our work establishes decompositionality as a formally definable and empirically testable property, providing a foundation for modular reasoning about neural networks.

Code generation is important in software engineering, and Reinforcement Learning with Verifiable Rewards (RLVR) is a powerful paradigm to improve it through execution-based feedback. However, most RLVR pipelines rely on human-curated tests, making progress bottlenecked by scarce and costly supervision. Existing work tried to use self-generated tests to ground rewards, but the lack of discriminative tests constrains the effect due to the sub-optimal performance of the model on test generation. We aim to improve code generation without ground-truth supervision by co-evolving code and test generation, so that their interactions yield progressively more informative supervision. To this end, we present ZeroCoder, a fully label-free co-evolutionary framework that jointly trains a Coder and a Tester using execution feedback from self-generated code-test interactions. For each problem, ZeroCoder executes sampled solutions against sampled tests to form a passing matrix, identifies a consensus subset of likely-correct solutions and consistent tests via a pluggable selection algorithm, and derives role-specific rewards. To ensure reward quality, ZeroCoder filters low-information instances via rank-based pre-filtering and trains the Tester with a curriculum balancing validity and mutation-driven discriminativeness. We further identify selector drift, the progressive miscalibration of fixed selection rules during co-evolution, and introduce DyB4, a Bayesian selector that uses as few as 10 labeled instances to recalibrate its priors dynamically. Across three models and six benchmarks, ZeroCoder consistently improves code generation and test generation. In the fully label-free setting, it improves code generation by up to 14.5% over the base model on Qwen2.5-Coder-7B-Instruct. With DyB4, the gain reaches 21.6%, while test generation improves by 24.3%, approaching oracle-supervised performance.

Retrieving relevant observations from long multi-modal web interaction histories is challenging because relevance depends on the evolving task state, modality (screenshots, HTML text, structured signals), and temporal distance. Prior approaches typically rely on static similarity thresholds or fixed-capacity buffers, which fail to adapt relevance to the current task context. We propose \textbf{ACGM}, a learned graph-memory retriever that constructs \emph{task-adaptive} relevance graphs over agent histories using policy-gradient optimization from downstream task success. ACGM captures heterogeneous temporal dynamics with modality-specific decay (visual decays $4.3\times$ faster than text: $λ_v{=}0.47$ vs.\ $λ_x{=}0.11$) and learns sparse connectivity (3.2 edges/node), enabling efficient $O(\log T)$ retrieval. Across WebShop, VisualWebArena, and Mind2Web, ACGM improves retrieval quality to \textbf{82.7 nDCG@10} (+9.3 over GPT-4o, $p{<}0.001$) and \textbf{89.2\% Precision@10} (+7.7), outperforming 19 strong dense, re-ranking, multi-modal, and graph-based baselines. Code to reproduce our results is available at{\color{blue}\href{https://github.com/S-Forouzandeh/ACGM-Agentic-Web}{Saman Forouzandeh}}.

The rapid emergence of Large Language Models (LLMs) has catalyzed Agentic artificial intelligence (AI), autonomous systems integrating perception, reasoning, and action into closed-loop pipelines for continuous adaptation. While unlocking transformative applications in mobile edge computing, autonomous systems, and next-generation wireless networks, this paradigm creates fundamental energy challenges through iterative inference and persistent data exchange. Unlike traditional AI where bottlenecks are computational Floating Point Operations (FLOPs), Agentic AI faces compounding computational and communication energy costs. In this survey, we propose an energy accounting framework identifying computational and communication costs across the Perception-Reasoning-Action cycle. We establish a unified taxonomy spanning model simplification, computation control, input and attention optimization, and hardware-aware inference. We explore cross-layer co-design strategies jointly optimizing model parameters, wireless transmissions, and edge resources. Finally, we identify open challenges of federated green learning, carbon-aware agency, 6th generation mobile communication (6G)-native Agentic AI, and self-sustaining systems, providing a roadmap for scalable autonomous intelligence.

Large language and music models are increasingly used for constrained generation: rhyming lines, fixed meter, inpainting or infilling, positional endings, and other global form requirements. These systems often perform strikingly well, but the induced procedures are usually not exact conditioning of the underlying autoregressive model. This creates a hidden inferential bias, distinct from the better-known notion of bias inherited from the training set: samples are distorted relative to the true constrained distribution, with no generic guarantee of complete coverage of the admissible solution space or of correct conditional probabilities over valid completions. We formalize several exact inference tasks for autoregressive models and prove corresponding hardness results. For succinctly represented autoregressive models whose next-token probabilities are computable in polynomial time, exact sentence-level maximum a posteriori (MAP) decoding is NP-hard. This hardness persists under unary and metrical constraints. On the sampling side, exact conditioned normalization is \#P-hard even for regular constraints such as fixed-length terminal events. Unlike finite-state Markov models, general autoregressive models do not admit a bounded-state dynamic program for these tasks. These results formalize a standard claim in the neural decoding literature: local autoregressive sampling is easy, whereas exact decoding and exact conditioning under global form constraints are computationally intractable in general.

Large language model (LLM) reinforcement learning (RL) pipelines are often bottlenecked by rollout generation, making end-to-end training slow. Recent work mitigates this by running rollouts with quantization to accelerate decoding, which is the most expensive stage of the RL loop. However, these setups destabilize optimization by amplifying the training-inference gap: rollouts are operated at low precision, while learning updates are computed at full precision. To address this challenge, we propose QaRL (Rollout Alignment Quantization-Aware RL), which aligns training-side forward with the quantized rollout to minimize mismatch. We further identify a failure mode in quantized rollouts: long-form responses tend to produce repetitive, garbled tokens (error tokens). To mitigate these problems, we introduce TBPO (Trust-Band Policy Optimization), a sequence-level objective with dual clipping for negative samples, aimed at keeping updates within the trust region. On Qwen3-30B-A3B MoE for math problems, QaRL outperforms quantized-rollout training by +5.5 while improving stability and preserving low-bit throughput benefits.

With the rise of LLMs, there is an increasing need for intelligent recommendation assistants that can handle complex queries and provide personalized, reasoning-driven recommendations. LLM-based recommenders show potential but face challenges in multi-step reasoning, underscoring the need for reasoning-augmented systems. To address this gap, we propose ReRec, a novel reinforcement fine-tuning (RFT) framework designed to improve LLM reasoning in complex recommendation tasks. Our framework introduces three key components: (1) Dual-Graph Enhanced Reward Shaping, integrating recommendation metrics like NDCG@K with Query Alignment and Preference Alignment Scores to provide fine-grained reward signals for LLM optimization; (2) Reasoning-aware Advantage Estimation, which decomposes LLM outputs into reasoning segments and penalizes incorrect steps to enhance reasoning of recommendation; and (3) Online Curriculum Scheduler, dynamically assess query difficulty and organize training curriculum to ensure stable learning during RFT. Experiments demonstrate that ReRec outperforms state-of-the-art baselines and preserves core abilities like instruction-following and general knowledge. Our codes are available at https://github.com/jiani-huang/ReRec.

Multi-task learning shows strikingly inconsistent results -- sometimes joint training helps substantially, sometimes it actively harms performance -- yet the field lacks a principled framework for predicting these outcomes. We identify a fundamental but unstated assumption underlying gradient-based task analysis: tasks must share training instances for gradient conflicts to reveal genuine relationships. When tasks are measured on the same inputs, gradient alignment reflects shared mechanistic structure; when measured on disjoint inputs, any apparent signal conflates task relationships with distributional shift. We discover this sample overlap requirement exhibits a sharp phase transition: below 30% overlap, gradient-task correlations are statistically indistinguishable from noise; above 40%, they reliably recover known biological structure. Comprehensive validation across multiple datasets achieves strong correlations and recovers biological pathway organization. Standard benchmarks systematically violate this requirement -- MoleculeNet operates at <5% overlap, TDC at 8-14% -- far below the threshold where gradient analysis becomes meaningful. This provides the first principled explanation for seven years of inconsistent MTL results.

The evolution of Large Language Models (LLMs) is shifting the focus from single, verifiable tasks toward complex, open-ended real-world scenarios, imposing significant challenges on the post-training phase. In these settings, the scale and complexity of reward systems have grown significantly, transitioning toward multi-objective formulations that encompass a comprehensive spectrum of model capabilities and application contexts. However, traditional methods typically rely on fixed reward weights, ignoring non-stationary learning dynamics and struggling with data heterogeneity across dimensions. To address these issues, we propose SPARD, a framework that establishes an automated, self-paced curriculum by perceiving learning progress to dynamically adjust multi-objective reward weights and data importance, thereby synchronizing learning intent with data utility for optimal performance. Extensive experiments across multiple benchmarks demonstrate that SPARD significantly enhances model capabilities across all domains.

While Large Language Models (LLMs) have achieved remarkable performance, they remain vulnerable to jailbreak attacks that circumvent safety constraints. Existing strategies, ranging from heuristic prompt engineering to computationally intensive optimization, often face significant trade-offs between effectiveness and efficiency. In this work, we propose Contextual Representation Ablation (CRA), a novel inference-time intervention framework designed to dynamically silence model guardrails. Predicated on the geometric insight that refusal behaviors are mediated by specific low-rank subspaces within the model's hidden states, CRA identifies and suppresses these refusal-inducing activation patterns during decoding without requiring expensive parameter updates or training. Empirical evaluation across multiple safety-aligned open-source LLMs demonstrates that CRA significantly outperforms baselines. These results expose the intrinsic fragility of current alignment mechanisms, revealing that safety constraints can be surgically ablated from internal representations, and underscore the urgent need for more robust defenses that secure the model's latent space.

This paper presents an LLM-driven approach for constructing diverse social media datasets to measure and compare loneliness in the caregiver and non-caregiver populations. We introduce an expert-developed loneliness evaluation framework and an expert-informed typology for categorizing causes of loneliness for analyzing social media text. Using a human-validated data processing pipeline, we apply GPT-4o, GPT-5-nano, and GPT-5 to build a high-quality Reddit corpus and analyze loneliness across both populations. The loneliness evaluation framework achieved average accuracies of 76.09% and 79.78% for caregivers and non-caregivers, respectively. The cause categorization framework achieved micro-aggregate F1 scores of 0.825 and 0.80 for caregivers and non-caregivers, respectively. Across populations, we observe substantial differences in the distribution of types of causes of loneliness. Caregivers' loneliness were predominantly linked to caregiving roles, identity recognition, and feelings of abandonment, indicating distinct loneliness experiences between the two groups. Demographic extraction further demonstrates the viability of Reddit for building a diverse caregiver loneliness dataset. Overall, this work establishes an LLM-based pipeline for creating high quality social media datasets for studying loneliness and demonstrates its effectiveness in analyzing population-level differences in the manifestation of loneliness.

Existing red-teaming studies on GUI agents have important limitations. Adversarial perturbations typically require white-box access, which is unavailable for commercial systems, while prompt injection is increasingly mitigated by stronger safety alignment. To study robustness under a more practical threat model, we propose Semantic-level UI Element Injection, a red-teaming setting that overlays safety-aligned and harmless UI elements onto screenshots to misdirect the agent's visual grounding. Our method uses a modular Editor-Overlapper-Victim pipeline and an iterative search procedure that samples multiple candidate edits, keeps the best cumulative overlay, and adapts future prompt strategies based on previous failures. Across five victim models, our optimized attacks improve attack success rate by up to 4.4x over random injection on the strongest victims. Moreover, elements optimized on one source model transfer effectively to other target models, indicating model-agnostic vulnerabilities. After the first successful attack, the victim still clicks the attacker-controlled element in more than 15% of later independent trials, versus below 1% for random injection, showing that the injected element acts as a persistent attractor rather than simple visual clutter.

Developers spend roughly one-tenth of their workday writing code, yet most AI tooling targets that fraction. This paper asks what should be built for the rest. We surveyed 860 Microsoft developers to understand where they want AI support, and where they want it to stay out. Using a human-in-the-loop, multi-model council-based thematic analysis, we identify 22 AI systems that developers want built across five task categories. For each, we describe the problem it solves, what makes it hard to build, and the constraints developers place on its behavior. Our findings point to a growing right-shift burden in AI-assisted development: developers wanted systems that embed quality signals earlier in their workflow to keep pace with accelerating code generation, while enforcing explicit authority scoping, provenance, uncertainty signaling, and least-privilege access throughout. This tension reveals a pattern we call "bounded delegation": developers wanted AI to absorb the assembly work surrounding their craft, never the craft itself. That boundary tracks where they locate professional identity, suggesting that the value of AI tooling may lie as much in where and how precisely it stops as in what it does.

Large language models (LLMs) have recently emerged as powerful training-free recommenders. However, their knowledge of individual items is inevitably uneven due to imbalanced information exposure during pretraining, a phenomenon we refer to as knowledge gap problem. To address this, most prior methods have employed a naive uniform augmentation that appends external information for every item in the input prompt. However, this approach not only wastes limited context budget on redundant augmentation for well-known items but can also hinder the model's effective reasoning. To this end, we propose KnowSA_CKP (Knowledge-aware Selective Augmentation with Comparative Knowledge Probing) to mitigate the knowledge gap problem. KnowSA_CKP estimates the LLM's internal knowledge by evaluating its capability to capture collaborative relationships and selectively injects additional information only where it is most needed. By avoiding unnecessary augmentation for well-known items, KnowSA_CKP focuses on items that benefit most from knowledge supplementation, thereby making more effective use of the context budget. KnowSA_CKP requires no fine-tuning step, and consistently improves both recommendation accuracy and context efficiency across four real-world datasets.

Performance, the externalization of intent, emotion, and personality through visual, vocal, and temporal behavior, is what makes a character alive. Learning such performance from video is a promising alternative to traditional 3D pipelines. However, existing video models struggle to jointly achieve high expressiveness, real-time inference, and long-horizon identity stability, a tension we call the performance trilemma. Conversation is the most comprehensive performance scenario, as characters simultaneously speak, listen, react, and emote while maintaining identity over time. To address this, we present LPM 1.0 (Large Performance Model), focusing on single-person full-duplex audio-visual conversational performance. Concretely, we build a multimodal human-centric dataset through strict filtering, speaking-listening audio-video pairing, performance understanding, and identity-aware multi-reference extraction; train a 17B-parameter Diffusion Transformer (Base LPM) for highly controllable, identity-consistent performance through multimodal conditioning; and distill it into a causal streaming generator (Online LPM) for low-latency, infinite-length interaction. At inference, given a character image with identity-aware references, LPM 1.0 generates listening videos from user audio and speaking videos from synthesized audio, with text prompts for motion control, all at real-time speed with identity-stable, infinite-length generation. LPM 1.0 thus serves as a visual engine for conversational agents, live streaming characters, and game NPCs. To systematically evaluate this setting, we propose LPM-Bench, the first benchmark for interactive character performance. LPM 1.0 achieves state-of-the-art results across all evaluated dimensions while maintaining real-time inference.

We study implicit reasoning, i.e. the ability to combine knowledge or rules within a single forward pass. While transformer-based large language models store substantial factual knowledge and rules, they often fail to compose this knowledge for implicit multi-hop reasoning, suggesting a lack of compositional generalization over their parametric knowledge. To address this limitation, we study recurrent-depth transformers, which enables iterative computation over the same transformer layers. We investigate two compositional generalization challenges under the implicit reasoning scenario: systematic generalization, i.e. combining knowledge that is never used for compositions during training, and depth extrapolation, i.e. generalizing from limited reasoning depth (e.g. training on up to 5-hop) to deeper compositions (e.g. 10-hop). Through controlled studies with models trained from scratch, we show that while vanilla transformers struggle with both generalization challenges, recurrent-depth transformers can effectively make such generalization. For systematic generalization, we find that this ability emerges through a three-stage grokking process, transitioning from memorization to in-distribution generalization and finally to systematic generalization, supported by mechanistic analysis. For depth extrapolation, we show that generalization beyond training depth can be unlocked by scaling inference-time recurrence, with more iterations enabling deeper reasoning. We further study how training strategies affect extrapolation, providing guidance on training recurrent-depth transformers, and identify a key limitation, overthinking, where excessive recurrence degrades predictions and limits generalization to very deep compositions.

Large language model (LLM) agents increasingly coordinate in multi-agent systems, yet we lack an understanding of where and why cooperation failures may arise. In many real-world coordination problems, from knowledge sharing in organizations to code documentation, helping others carries negligible personal cost while generating substantial collective benefits. However, whether LLM agents cooperate when helping neither benefits nor harms the helper, while being given explicit instructions to do so, remains unknown. We build a multi-agent setup designed to study cooperative behavior in a frictionless environment, removing all strategic complexity from cooperation. We find that capability does not predict cooperation: OpenAI o3 achieves only 17% of optimal collective performance while OpenAI o3-mini reaches 50%, despite identical instructions to maximize group revenue. Through a causal decomposition that automates one side of agent communication, we separate cooperation failures from competence failures, tracing their origins through agent reasoning analysis. Testing targeted interventions, we find that explicit protocols double performance for low-competence models, and tiny sharing incentives improve models with weak cooperation. Our findings suggest that scaling intelligence alone will not solve coordination problems in multi-agent systems and will require deliberate cooperative design, even when helping others costs nothing.

Open-ended video game glitch detection aims to identify glitches in gameplay videos, describe them in natural language, and localize when they occur. Unlike conventional game glitch understanding tasks which have largely been framed as image-level recognition or closed-form question answering, this task requires reasoning about game-specific dynamics such as mechanics, physics, rendering, animation, and expected state transitions directly over continuous gameplay videos and distinguishing true glitches from unusual but valid in-game events. To support this task, we introduce VideoGlitchBench, the first benchmark for open-ended video game glitch detection with temporal localization. VideoGlitchBench contains 5,238 gameplay videos from 120 games, each annotated with detailed glitch descriptions and precise temporal spans, enabling unified evaluation of semantic understanding and temporal grounding. We further propose GliDe, an agentic framework with three key components: a game-aware contextual memory for informed reasoning, a debate-based reflector for multi-perspective glitch detection and verification, and an event-level grounding module that recovers complete glitch intervals from fragmented temporal evidence. We also design a task-specific evaluation protocol that jointly measures semantic fidelity and temporal accuracy. Experiments show that this task remains highly challenging for current multimodal models, while GliDe achieves substantially stronger performance than corresponding vanilla model baselines.

We present an end-to-end pipeline that converts healthcare policy documents into executable, data-aware Business Process Model and Notation (BPMN) models using large language models (LLMs) for simulation-based policy evaluation. We address the main challenges of automated policy digitization with four contributions: data-grounded BPMN generation with syntax auto-correction, executable augmentation, KPI instrumentation, and entropy-based uncertainty detection. We evaluate the pipeline on diabetic nephropathy prevention guidelines from three Japanese municipalities, generating 100 models per backend across three LLMs and executing each against 1,000 synthetic patients. On well-structured policies, the pipeline achieves a 100% ground-truth match with perfect per-patient decision agreement. Across all conditions, raw per-patient decision agreement exceeds 92%, and entropy scores increase monotonically with document complexity, confirming that the detector reliably separates unambiguous policies from those requiring targeted human clarification.

Tool learning has emerged as a promising paradigm for large language models (LLMs) to address real-world challenges. Due to the extensive and irregularly updated number of tools, tool retrieval for selecting the desired tool subset is essential. However, current tool retrieval methods are usually based on academic benchmarks containing overly detailed instructions (e.g., specific API names and parameters), while real-world instructions are more vague. Such a discrepancy would hinder the tool retrieval in real-world applications. In this paper, we first construct a new benchmark, VGToolBench, to simulate human vague instructions. Based on this, we conduct a series of preliminary analyses and find that vague instructions indeed damage the performance of tool retrieval. To this end, we propose a simple-yet-effective Tool Retrieval Bridge (TRB) approach to boost the performance of tool retrieval for vague instructions. The principle of TRB is to introduce a bridge model to rewrite the vague instructions into more specific ones and alleviate the gap between vague instructions and retriever preferences.We conduct extensive experiments under multiple commonly used retrieval settings, and the results show that TRB effectively mitigates the ambiguity of vague instructions while delivering consistent and substantial improvements across all baseline retrievers. For example, with the help of TRB, BM25 achieves a relative improvement of up to 111.51%, i.e., increasing the average NDCG score from 9.73 to 19.59. The source code and models are publicly available at https://github.com/kfchenhn/TRB.

Long-context inference in LLMs faces the dual challenges of quadratic attention complexity and prohibitive KV cache memory. While token-level sparse attention offers superior accuracy, its indexing overhead is costly; block-level methods improve efficiency but sacrifice precision. We propose AsyncTLS, a hierarchical sparse attention system that combines coarse-grained block filtering with fine-grained token selection to balance accuracy and efficiency, coupled with an asynchronous offloading engine that overlaps KV cache transfers with computation via temporal locality exploitation. Evaluated on Qwen3 and GLM-4.7-Flash across GQA, and MLA architectures, AsyncTLS achieves accuracy comparable to full attention while delivering 1.2x - 10.0x operator speedups and 1.3x - 4.7x end-to-end throughput improvements on 48k - 96k contexts.

Learning theories have historically changed when the conditions of learning evolved. Generative and agentic AI create a new condition by allowing learners to delegate explanation, writing, problem solving, and other cognitive work to systems that can generate, recommend, and sometimes act on the learner's behalf. This creates a fundamental challenge for learning theory: successful performance can no longer be assumed to indicate learning. Learners may complete tasks effectively with AI support while developing less understanding, weaker judgment, and limited transferable capability. We argue that this problem is not fully captured by existing learning theories. Behaviourism, cognitivism, constructivism, and connectivism remain important, but they do not directly explain when AI-assisted performance becomes durable human capability. We propose Agentivism, a learning theory for human-AI interaction. Agentivism defines learning as durable growth in human capability through selective delegation to AI, epistemic monitoring and verification of AI contributions, reconstructive internalization of AI-assisted outputs, and transfer under reduced support. The importance of Agentivism lies in explaining how learning remains possible when intelligent delegation is easy and human-AI interaction is becoming a persistent and expanding part of human learning.

Latent-position random graph models usually treat the node set as fixed once the sample size is chosen, while graphon-based and random-measure constructions allow more randomness at the cost of weaker geometric interpretability. We introduce \emph{Intensity Dot Product Graphs} (IDPGs), which extend Random Dot Product Graphs by replacing a fixed collection of latent positions with a Poisson point process on a Euclidean latent space. This yields a model with random node populations, RDPG-style dot-product affinities, and a population-level intensity that links continuous latent structure to finite observed graphs. We define the heat map and the desire operator as continuous analogues of the probability matrix, prove a spectral consistency result connecting adjacency singular values to the operator spectrum, compare the construction with graphon and digraphon representations, and show how classical RDPGs arise in a concentrated limit. Because the model is parameterized by an evolving intensity, temporal extensions through partial differential equations arise naturally.

Extending LLM context windows is hindered by scarce high-quality long-context data. Recent methods synthesize data with genuine long-range dependencies via information-theoretic verification, selecting contexts that reduce a base model's predictive entropy. However, their single-pass offline construction with a fixed model creates a fundamental off-policy gap: the static screening landscape misaligns with the model's evolving capabilities, causing the training distribution to drift. We propose PolicyLong, shifting data construction towards a dynamic on-policy paradigm. By iteratively re-executing data screening (entropy computation, retrieval, and verification) using the current model, PolicyLong ensures the training distribution tracks evolving capabilities, yielding an emergent self-curriculum. Crucially, both positive and hard negative contexts derive from the current model's entropy landscape, co-evolving what the model learns to exploit and resist. Experiments on RULER, HELMET, and LongBench-v2 (Qwen2.5-3B) show PolicyLong consistently outperforms EntropyLong and NExtLong, with gains growing at longer contexts (e.g., +2.54 at 128K on RULER), confirming the value of on-policy data evolution.

Full-parameter fine-tuning of large language models is constrained by substantial GPU memory requirements. Low-rank adaptation methods mitigate this challenge by updating only a subset of parameters. However, these approaches often limit model expressiveness and yield lower performance than full-parameter fine-tuning. Layer-wise fine-tuning methods have emerged as an alternative, enabling memory-efficient training through static layer importance sampling strategies. However, these methods overlook variations in layer importance across tasks and training stages, resulting in suboptimal performance on downstream tasks. To address these limitations, we propose GRASS, a gradient-based adaptive layer-wise importance sampling framework. GRASS utilizes mean gradient norms as a task-aware and training-stage-aware metric for estimating layer importance. Furthermore, GRASS adaptively adjusts layer sampling probabilities through an adaptive training strategy. We also introduce a layer-wise optimizer state offloading mechanism that overlaps computation and communication to further reduce memory usage while maintaining comparable training throughput. Extensive experiments across multiple models and benchmarks demonstrate that GRASS consistently outperforms state-of-the-art methods, achieving an average accuracy improvement of up to 4.38 points and reducing memory usage by up to 19.97\%.

Computer vision, a core domain of artificial intelligence (AI), is the field that enables the computational analysis, understanding, and generation of visual data. Despite being historically rooted in military funding and increasingly deployed in warfare, the field tends to position itself as a neutral, purely technical endeavor, failing to engage in discussions about its dual-use applications. Yet it has been reported that computer vision systems are being systematically weaponized to assist in technologies that inflict harm, such as surveillance or warfare. Expanding on these concerns, we study the extent to which computer vision research is being used in the military and surveillance domains. We do so by collecting a dataset of tech companies with financial ties to the field's central research exchange platform: conferences. Conference sponsorship, we argue, not only serves as strong evidence of a company's investment in the field but also provides a privileged position for shaping its trajectory. By investigating sponsors' activities, we reveal that 44% of them have a direct connection with military or surveillance applications. We extend our analysis through two case studies in which we discuss the opportunities and limitations of sponsorship as a means for uncovering technological weaponization.

Large-scale vision-language models (VLMs) exhibit remarkable zero-shot capabilities, yet the internal mechanisms driving their anomaly detection (AD) performance remain poorly understood. Current methods predominantly treat VLMs as black-box feature extractors, assuming that anomaly-specific knowledge must be acquired through external adapters or memory banks. In this paper, we challenge this assumption by arguing that anomaly knowledge is intrinsically embedded within pre-trained models but remains latent and under-activated. We hypothesize that this knowledge is concentrated within a sparse subset of anomaly-sensitive neurons. To validate this, we propose latent anomaly knowledge excavation (LAKE), a training-free framework that identifies and elicits these critical neuronal signals using only a minimal set of normal samples. By isolating these sensitive neurons, LAKE constructs a highly compact normality representation that integrates visual structural deviations with cross-modal semantic activations. Extensive experiments on industrial AD benchmarks demonstrate that LAKE achieves state-of-the-art performance while providing intrinsic, neuron-level interpretability. Ultimately, our work advocates for a paradigm shift: redefining anomaly detection as the targeted activation of latent pre-trained knowledge rather than the acquisition of a downstream task.

Large language models are trained and evaluated on quantitative reasoning tasks written in clean, emotionally neutral language. However, real-world queries are often wrapped in frustration, urgency or enthusiasm. Does emotional framing alone degrade reasoning when all numerical content is preserved? To investigate this, a controlled emotion translation framework is developed that rewrites problems into emotional variants while preserving all quantities and relationships. Using this framework, Temper-5400 (5,400 semantically verified emotion--neutral pairs) is constructed across GSM8K, MultiArith, and ARC-Challenge, and evaluated on eighteen models (1B to frontier scale). Two core results emerge: First, emotional framing reduces accuracy by 2-10 percentage points even though all numerical content is preserved. Second, neutralizing emotional variants recovers most of the lost performance, showing both that the degradation is tied to emotional style rather than content corruption and that neutralization can serve as a lightweight inference-time mitigation. Non-emotional paraphrases cause no such degradation, implicating emotional content rather than surface-level changes. Beyond emotion specifically, the benchmark construction procedure provides a general framework for controlled stylistic translation and robustness evaluation.

Although LLM agents can leverage tools for complex tasks, they still need memory to maintain cross-turn consistency and accumulate reusable information in long-horizon interactions. However, retrieval-based external memory systems incur low online overhead but suffer from unstable accuracy due to limited query construction and candidate filtering. In contrast, many systems use repeated large-model calls for online memory operations, improving accuracy but accumulating latency over long interactions. We propose LightMem, a lightweight memory system for better agent memory driven by Small Language Models (SLMs). LightMem modularizes memory retrieval, writing, and long-term consolidation, and separates online processing from offline consolidation to enable efficient memory invocation under bounded compute. We organize memory into short-term memory (STM) for immediate conversational context, mid-term memory (MTM) for reusable interaction summaries, and long-term memory (LTM) for consolidated knowledge, and uses user identifiers to support independent retrieval and incremental maintenance in multi-user settings. Online, LightMem operates under a fixed retrieval budget and selects memories via a two-stage procedure: vector-based coarse retrieval followed by semantic consistency re-ranking. Offline, it abstracts reusable interaction evidence and incrementally integrates it into LTM. Experiments show gains across model scales, with an average F1 improvement of about 2.5 on LoCoMo, more effective and low median latency (83 ms retrieval; 581 ms end-to-end).

Embodied agents are expected to operate persistently in dynamic physical environments, continuously acquiring new capabilities over time. Existing approaches to improving agent performance often rely on modifying the agent itself -- through prompt engineering, policy updates, or structural redesign -- leading to instability and loss of identity in long-lived systems. In this work, we propose a capability-centric evolution paradigm for embodied agents. We argue that a robot should maintain a persistent agent as its cognitive identity, while enabling continuous improvement through the evolution of its capabilities. Specifically, we introduce the concept of Embodied Capability Modules (ECMs), which represent modular, versioned units of embodied functionality that can be learned, refined, and composed over time. We present a unified framework in which capability evolution is decoupled from agent identity. Capabilities evolve through a closed-loop process involving task execution, experience collection, model refinement, and module updating, while all executions are governed by a runtime layer that enforces safety and policy constraints. We demonstrate through simulated embodied tasks that capability evolution improves task success rates from 32.4% to 91.3% over 20 iterations, outperforming both agent-modification baselines and established skill-learning methods (SPiRL, SkiMo), while preserving zero policy drift and zero safety violations. Our results suggest that separating agent identity from capability evolution provides a scalable and safe foundation for long-term embodied intelligence.

In this paper, we study the problem of mean estimation under strict 1-bit communication constraints. We propose a novel adaptive mean estimator based solely on randomized threshold queries, where each 1-bit outcome indicates whether a given sample exceeds a sequentially chosen threshold. Our estimator is $(ε, δ)$-PAC for any distribution with a bounded mean $μ\in [-λ, λ]$ and a bounded $k$-th central moment $\mathbb{E}[|X-μ|^k] \le σ^k$ for any fixed $k > 1$. Crucially, our sample complexity is order-optimal in all such tail regimes, i.e., for every such $k$ value. For $k \neq 2$, our estimator's sample complexity matches the unquantized minimax lower bounds plus an unavoidable $O(\log(λ/σ))$ localization cost. For the finite-variance case ($k=2$), our estimator's sample complexity has an extra multiplicative $O(\log(σ/ε))$ penalty, and we establish a novel information-theoretic lower bound showing that this penalty is a fundamental limit of 1-bit quantization. We also establish a significant adaptivity gap: for both threshold queries and more general interval queries, the sample complexity of any non-adaptive estimator must scale linearly with the search space parameter $λ/σ$, rendering it vastly less sample efficient than our adaptive approach. Finally, we present algorithmic variants that (i) handle an unknown sampling budget, (ii) adapt to an unknown scale parameter~$σ$ given (possibly loose) bounds, and (iii) require only two stages of adaptivity at the expense of more complicated general 1-bit queries.

Recent advances in Reinforcement Learning with Verifiable Rewards (RLVR) have demonstrated significant potential in single-turn reasoning tasks. With the paradigm shift toward self-evolving agentic learning, models are increasingly expected to learn from trajectories by synthesizing tools or accumulating explicit experiences. However, prevailing methods typically rely on large-scale LLMs or multi-agent frameworks, which hinder their deployment in resource-constrained environments. The inherent sparsity of outcome-based rewards also poses a substantial challenge, as agents typically receive feedback only upon completion of tasks. To address these limitations, we introduce a Tool-Memory based self-evolving agentic framework SEARL. Unlike approaches that directly utilize interaction experiences, our method constructs a structured experience memory that integrates planning with execution. This provides a novel state abstraction that facilitates generalization across analogous contexts, such as tool reuse. Consequently, agents extract explicit knowledge from historical data while leveraging inter-trajectory correlations to densify reward signals. We evaluate our framework on knowledge reasoning and mathematics tasks, demonstrating its effectiveness in achieving more practical and efficient learning.

Recent advances in language model (LM) agents have significantly improved automated software engineering (SWE). Prior work has proposed various agentic workflows and training strategies as well as analyzed failure modes of agentic systems on SWE tasks, focusing on several contextual information signals: Reproduction Test, Regression Test, Edit Location, Execution Context, and API Usage. However, the individual contribution of each signal to overall success remains underexplored, particularly their ideal contribution when intermediate information is perfectly obtained. To address this gap, we introduce Oracle-SWE, a unified method to isolate and extract oracle information signals from SWE benchmarks and quantify the impact of each signal on agent performance. To further validate the pattern, we evaluate the performance gain of signals extracted by strong LMs when provided to a base agent, approximating real-world task-resolution settings. These evaluations aim to guide research prioritization for autonomous coding systems.

We introduce PeReGrINE, a benchmark and evaluation framework for personalized review generation grounded in graph-structured user--item evidence. PeReGrINE restructures Amazon Reviews 2023 into a temporally consistent bipartite graph, where each target review is conditioned on bounded evidence from user history, item context, and neighborhood interactions under explicit temporal cutoffs. To represent persistent user preferences without conditioning directly on sparse raw histories, we compute a User Style Parameter that summarizes each user's linguistic and affective tendencies over prior reviews. This setup supports controlled comparison of four graph-derived retrieval settings: product-only, user-only, neighbor-only, and combined evidence. Beyond standard generation metrics, we introduce Dissonance Analysis, a macro-level evaluation framework that measures deviation from expected user style and product-level consensus. We also study visual evidence as an auxiliary context source and find that it can improve textual quality in some settings, while graph-derived evidence remains the main driver of personalization and consistency. Across product categories, PeReGrINE offers a reproducible way to study how evidence composition affects review fidelity, personalization, and grounding in retrieval-conditioned language models.

Talking face generation has gained significant attention as a core application of generative models. To enhance the expressiveness and realism of synthesized videos, emotion editing in talking face video plays a crucial role. However, existing approaches often limit expressive flexibility and struggle to generate extended emotions. Label-based methods represent emotions with discrete categories, which fail to capture a wide range of emotions. Audio-based methods can leverage emotionally rich speech signals - and even benefit from expressive text-to-speech (TTS) synthesis - but they fail to express the target emotions because emotions and linguistic contents are entangled in emotional speeches. Images-based methods, on the other hand, rely on target reference images to guide emotion transfer, yet they require high-quality frontal views and face challenges in acquiring reference data for extended emotions (e.g., sarcasm). To address these limitations, we propose Cross-Modal Emotion Transfer (C-MET), a novel approach that generates facial expressions based on speeches by modeling emotion semantic vectors between speech and visual feature spaces. C-MET leverages a large-scale pretrained audio encoder and a disentangled facial expression encoder to learn emotion semantic vectors that represent the difference between two different emotional embeddings across modalities. Extensive experiments on the MEAD and CREMA-D datasets demonstrate that our method improves emotion accuracy by 14% over state-of-the-art methods, while generating expressive talking face videos - even for unseen extended emotions. Code, checkpoint, and demo are available at https://chanhyeok-choi.github.io/C-MET/

Engineering workflows such as design optimization, simulation-based diagnosis, control tuning, and model-based systems engineering (MBSE) are iterative, constraint-driven, and shaped by prior decisions. Yet many AI methods still treat these activities as isolated tasks rather than as parts of a broader workflow. This paper presents Agentic Engineering Intelligence (AEI), an industrial vision framework that models engineering workflows as constrained, history-aware sequential decision processes in which AI agents support engineer-supervised interventions over engineering toolchains. AEI links an offline phase for engineering data processing and workflow-memory construction with an online phase for workflow-state estimation, retrieval, and decision support. A control-theoretic interpretation is also possible, in which engineering objectives act as reference signals, agents act as workflow controllers, and toolchains provide feedback for intervention selection. Representative automotive use cases in suspension design, reinforcement learning tuning, multimodal engineering knowledge reuse, aerodynamic exploration, and MBSE show how diverse workflows can be expressed within a common formulation. Overall, the paper positions engineering AI as a problem of process-level intelligence and outlines a practical roadmap for future empirical validation in industrial settings.

Automotive engineering development increasingly relies on heterogeneous 3D data, including finite element (FE) models, body-in-white (BiW) representations, CAD geometry, and CFD meshes. At the same time, engineering teams face growing pressure to shorten development cycles, improve performance and accelerate innovation. Although artificial intelligence (AI) is increasingly explored in this domain, many current methods remain task-specific, difficult to interpret, and hard to reuse across development stages. This paper presents a practical graph learning framework for 3D engineering AI, in which heterogeneous engineering assets are converted into physics-aware graph representations and processed by Graph Neural Networks (GNNs). The framework is designed to support both classification and prediction tasks. The framework is validated on two automotive applications: CAE vibration mode shape classification and CFD aerodynamic field prediction. For CAE vibration mode classification, a region-aware BiW graph supports explainable mode classification across vehicle and FE variants under label scarcity. For CFD aerodynamic field prediction, a physics-informed surrogate predicts pressure and wall shear stress (WSS) across aerodynamic body shape variants, while symmetry preserving down sampling retains accuracy with lower computational cost. The framework also outlines data generation guidance that can help engineers identify which additional simulations or labels are valuable to collect next. These results demonstrate a practical and reusable engineering AI workflow for more trustworthy CAE and CFD decision support.

Existing accountability frameworks for AI systems, legal, ethical, and regulatory, rest on a shared assumption: for any consequential outcome, at least one identifiable person had enough involvement and foresight to bear meaningful responsibility. This paper proves that agentic AI systems violate this assumption not as an engineering limitation but as a mathematical necessity once autonomy exceeds a computable threshold. We introduce Human-Agent Collectives, a formalisation of joint human-AI systems where agents are modelled as state-policy tuples within a shared structural causal model. Autonomy is characterised through a four-dimensional information-theoretic profile (epistemic, executive, evaluative, social); collective behaviour through interaction graphs and joint action spaces. We axiomatise legitimate accountability through four minimal properties: Attributability (responsibility requires causal contribution), Foreseeability Bound (responsibility cannot exceed predictive capacity), Non-Vacuity (at least one agent bears non-trivial responsibility), and Completeness (all responsibility must be fully allocated). Our central result, the Accountability Incompleteness Theorem, proves that for any collective whose compound autonomy exceeds the Accountability Horizon and whose interaction graph contains a human-AI feedback cycle, no framework can satisfy all four properties simultaneously. The impossibility is structural: transparency, audits, and oversight cannot resolve it without reducing autonomy. Below the threshold, legitimate frameworks exist, establishing a sharp phase transition. Experiments on 3,000 synthetic collectives confirm all predictions with zero violations. This is the first impossibility result in AI governance, establishing a formal boundary below which current paradigms remain valid and above which distributed accountability mechanisms become necessary.

Frontier LLMs can navigate complex websites, but their cost and reliance on third-party APIs make local deployment impractical. We introduce Agent-as-Annotators, a framework that structures synthetic trajectory generation for web agents by analogy to human annotation roles, replacing the Task Designer, Annotator, and Supervisor with modular LLM components. Using Gemini 3 Pro as teacher, we generate 3,000 trajectories across six web environments and fine-tune a 9B-parameter student with pure supervised learning on the 2,322 that pass quality filtering. The resulting model achieves 41.5% on WebArena, surpassing closed-source models such as Claude 3.5 Sonnet (36.0%) and GPT-4o (31.5%) under the same evaluation protocol, and nearly doubling the previous best open-weight result (Go-Browse, 21.7%). Capabilities transfer to unseen environments, with an 18.2 percentage point gain on WorkArena L1 (an enterprise platform never seen during training) and consistent improvements across three additional benchmarks. Ablations confirm that each pipeline component contributes meaningfully, with Judge filtering, evaluation hints, and reasoning traces each accounting for measurable gains. These results demonstrate that structured trajectory synthesis from a single frontier teacher is sufficient to produce competitive, locally deployable web agents. Project page: https://agent-as-annotators.github.io

Collaboration and information sharing empower Multi-Agent Systems (MAS) but also introduce a critical security risk known as Agent Cascading Injection (ACI). In such attacks, a compromised agent exploits inter-agent trust to propagate malicious instructions, causing cascading failures across the system. However, existing studies consider only limited attack strategies and simplified MAS settings, limiting their generalizability and comprehensive evaluation. To bridge this gap, we introduce ACIArena, a unified framework for evaluating the robustness of MAS. ACIArena offers systematic evaluation suites spanning multiple attack surfaces (i.e., external inputs, agent profiles, inter-agent messages) and attack objectives (i.e., instruction hijacking, task disruption, information exfiltration). Specifically, ACIArena establishes a unified specification that jointly supports MAS construction and attack-defense modules. It covers six widely used MAS implementations and provides a benchmark of 1,356 test cases for systematically evaluating MAS robustness. Our benchmarking results show that evaluating MAS robustness solely through topology is insufficient; robust MAS require deliberate role design and controlled interaction patterns. Moreover, defenses developed in simplified environments often fail to transfer to real-world settings; narrowly scoped defenses may even introduce new vulnerabilities. ACIArena aims to provide a solid foundation for advancing deeper exploration of MAS design principles.

Recent advancements in code large language models (Code-LLMs) have demonstrated remarkable capabilities in resolving programming related tasks. Meanwhile, researchers have recognized that the quality of pre-training data is crucial for improving LLM performance. However, most of the existing research on pre-training data filtering has focused on general datasets, and little attention for programming datasets. In this paper, we aim to address this gap by exploring the effectiveness of a widely used general data filtering technique, i.e., data-influence-score filtering, within the context of programming-related datasets. To this end, we first introduce a method for calculating data-influence-score for generative programming tasks which involves transforming a variety of downstream coding tasks into validation sets and using the models loss on these sets as a performance metric. Next, we pre-train a Code-LLMs with 1 billion parameters from scratch on a dataset of 100 billion code tokens. Based on it, we conduct an extensive empirical study to evaluate the effectiveness of data-influence-score filtering methods. Specifically, we examine how well this technique improves model performance, investigate how the characteristics of beneficial training data vary across different training stages and programming tasks, and assess the feasibility of prediction-based data-influence-score filtering method. Our findings show that data-influence-score filtering based on validation-set-loss can enhance models programming performance. Moreover, we observe that the criteria of beneficial training data differ significantly across various downstream programming tasks.

Collaborative edge-cloud frameworks have emerged as the main- stream paradigm for mobile automation, mitigating the latency and privacy risks inherent to monolithic cloud agents. However, existing approaches centralize administration in the cloud while relegating the device to passive execution, inducing a cognitive lag regard- ing real-time UI dynamics. To tackle this, we introduce AdecPilot by applying the principle of administrative decentralization to the edge-cloud multi-agent framework, which redefines edge agency by decoupling high-level strategic designing from tactical grounding. AdecPilot integrates a UI-agnostic cloud designer generating ab- stract milestones with a bimodal edge team capable of autonomous tactical planning and self-correction without cloud intervention. Furthermore, AdecPilot employs a Hierarchical Implicit Termi- nation protocol to enforce deterministic stops and prevent post- completion hallucinations. Extensive experiments demonstrate pro- posed approach improves task success rate by 21.7% while reducing cloud token consumption by 37.5% against EcoAgent and decreas- ing end to end latency by 88.9% against CORE. The source code is available at https://anonymous.4open.science/r/Anonymous_code- B8AB.

We investigate a fundamental structural question in Grouped Query Attention (GQA) transformers: do the layers most sensitive to task correctness coincide with the layers where positional encoding adaptation has the greatest leverage? We term this the co-localization hypothesis and test it on Llama 3.1 8B, a 32-layer GQA model with a 4:1 query-to-key-value head ratio. We introduce \LSLORA, which restricts LoRA adaptation to layers identified via a novel correctness-differential hidden-state metric, and GARFA (GQA-Aware RoPE Frequency Adaptation), which attaches 8 learnable per-KV-head scalar multipliers to each targeted layer. Contrary to the co-localization hypothesis, we discover strong anti-localization: task-sensitive layers concentrate in the late network ($\ell\in\{23\text{-}31\}$) while RoPE-influential layers dominate the early network ($\ell\in\{0\text{-}9\}$), yielding Spearman $r_s = -0.735$ ($p = 1.66\times10^{-6}$). Despite this anti-localization, a 4-way cross-layer ablation shows that applying both interventions to the sensitivity-identified layers outperforms all alternative configurations by 4-16 percentage points across six diverse benchmarks (MMLU, GPQA, HumanEval+, MATH, MGSM, ARC), approaching Claude 3.5 Haiku on HumanEval+ (67.1% vs. 68.3%) at \$100 total compute cost.

As generative artificial intelligence evolves, deepfake attacks have escalated from single-modality manipulations to complex, multimodal threats. Existing forensic techniques face a severe generalization bottleneck: by relying excessively on superficial, modality-specific artifacts, they neglect the shared latent forgery knowledge hidden beneath variable physical appearances. Consequently, these models suffer catastrophic performance degradation when confronted with unseen "dark modalities." To break this limitation, this paper introduces a paradigm shift that redefines multimodal forensics from conventional "feature fusion" to "modality generalization." We propose the first modality-agnostic forgery (MAF) detection framework. By explicitly decoupling modality-specific styles, MAF precisely extracts the essential, cross-modal latent forgery knowledge. Furthermore, we define two progressive dimensions to quantify model generalization: transferability toward semantically correlated modalities (Weak MAF), and robustness against completely isolated signals of "dark modality" (Strong MAF). To rigorously assess these generalization limits, we introduce the DeepModal-Bench benchmark, which integrates diverse multimodal forgery detection algorithms and adapts state-of-the-art generalized learning methods. This study not only empirically proves the existence of universal forgery traces but also achieves significant performance breakthroughs on unknown modalities via the MAF framework, offering a pioneering technical pathway for universal multimodal defense.

Controlling the evolution of a many-body stochastic system from a disordered reference state to a structured target ensemble, characterized empirically through samples, arises naturally in non-equilibrium statistical mechanics and stochastic control. The natural relaxation of such a system - driven by diffusion - runs from the structured target toward the disordered reference. The natural question is then: what is the minimum-work stochastic process that reverses this relaxation, given a pathwise cost functional combining spatial penalties and control effort? Computing this optimal process requires knowledge of trajectories that already sample the target ensemble - precisely the object one is trying to construct. We resolve this by establishing a time-reversal duality: the value function governing the hard backward dynamics satisfies an equivalent forward-in-time HJB equation, whose solution can be read off directly from the tractable forward relaxation trajectories. Via the Cole-Hopf transformation and its associated Feynman-Kac representation, this forward potential is computed as a path-space free energy averaged over these forward trajectories - the same relaxation paths that are easy to simulate - without any backward simulation or knowledge of the target beyond samples. The resulting framework provides a physically interpretable description of stochastic transport in terms of path-space free energy, risk-sensitive control, and spatial cost geometry. We illustrate the theory with numerical examples that visualize the learned value function and the induced controlled diffusions, demonstrating how spatial cost fields shape transport geometry analogously to Fermat's Principle in inhomogeneous media. Our results establish a unifying connection between stochastic optimal control, Schrödinger bridge theory, and non-equilibrium statistical mechanics.

We describe and analyze a distributed compute architecture for SSO computational satellites that can potentially provide >100 kW compute power per launched metric ton (including deployment and station keeping mass). The architecture co-locates and integrates the solar cells, radiator, and compute functions into multiple small panels arranged in a large array. The resultant large vapor chamber radiator area per panel should permit ICs to operate at junction temperatures near 40*C with benefits in compute efficiency and reliability. Using the structure of the radiator to support the solar cells may also yield a specific power of about 500 W/kg compared to less than 100 for existing conventional implementations. Assuming development of custom solutions for all components, a 16 MW computation, 150 ton satellite comprising a 20 m x 2200 m grid of 16,000 panels can fit in a single Starship hold. The concept is scalable to much larger satellites with higher mass payloads or using on-orbit assembly. We consider panel sizes from 1 to 4 m2 to allow trading vapor chamber heat transport with compute efficiency and inter-panel communication. Assuming a 1 kW/panel design, 512-panel subarrays of the satellite can run a representative inference-only LLM with 500,000 token context window and 128 attention blocks, at a rate of 553 tokens/sec/session, across 256 simultaneous in-flight sessions. A full satellite could support 31 such subarrays, for >7900 inferences at a time.

Articulated objects are essential for embodied AI and world models, yet inferring their kinematics from a single closed-state image remains challenging because crucial motion cues are often occluded. Existing methods either require multi-state observations or rely on explicit part priors, retrieval, or other auxiliary inputs that partially expose the structure to be inferred. In this work, we present DailyArt, which formulates articulated joint estimation from a single static image as a synthesis-mediated reasoning problem. Instead of directly regressing joints from a heavily occluded observation, DailyArt first synthesizes a maximally articulated opened state under the same camera view to expose articulation cues, and then estimates the full set of joint parameters from the discrepancy between the observed and synthesized states. Using a set-prediction formulation, DailyArt recovers all joints simultaneously without requiring object-specific templates, multi-view inputs, or explicit part annotations at test time. Taking estimated joints as conditions, the framework further supports part-level novel state synthesis as a downstream capability. Extensive experiments show that DailyArt achieves strong performance in articulated joint estimation and supports part-level novel state synthesis conditioned on joints. Project page is available at https://rangooo123.github.io/DaliyArt.github.io/.

Despite extensive research, Large Language Models continue to hallucinate when generating code, particularly when using libraries. On NL-to-code benchmarks that require library use, we find that LLMs generate code that uses non-existent library features in 8.1-40% of responses.One intuitive approach for detection and mitigation of hallucinations is static analysis. In this paper, we analyse the potential of static analysis tools, both in terms of what they can solve and what they cannot. We find that static analysis tools can detect 16-70% of all errors, and 14-85% of library hallucinations, with performance varying by LLM and dataset. Through manual analysis, we identify cases a static method could not plausibly catch, which gives an upper bound on their potential from 48.5% to 77%. Overall, we show that static analysis methods are cheap method for addressing some forms of hallucination, and we quantify how far short of solving the problem they will always be.

The deployment of large language models (LLMs) raises significant ethical and safety concerns. While LLM alignment techniques are adopted to improve model safety and trustworthiness, adversaries can exploit these techniques to undermine safety for malicious purposes, resulting in \emph{misalignment}. Misaligned LLMs may be published on open platforms to magnify harm. To address this, additional safety alignment, referred to as \emph{realignment}, is necessary before deploying untrusted third-party LLMs. This study explores the efficacy of fine-tuning methods in terms of misalignment, realignment, and the effects of their interplay. By evaluating four Supervised Fine-Tuning (SFT) and two Preference Fine-Tuning (PFT) methods across four popular safety-aligned LLMs, we reveal a mechanism asymmetry between attack and defense. While Odds Ratio Preference Optimization (ORPO) is most effective for misalignment, Direct Preference Optimization (DPO) excels in realignment, albeit at the expense of model utility. Additionally, we identify model-specific resistance, residual effects of multi-round adversarial dynamics, and other noteworthy findings. These findings highlight the need for robust safeguards and customized safety alignment strategies to mitigate potential risks in the deployment of LLMs. Our code is available at https://github.com/zhangrui4041/The-Art-of-Mis-alignment.

Empowering Large Multimodal Models (LMMs) with image generation often leads to catastrophic forgetting in understanding tasks due to severe gradient conflicts. While existing paradigms like Mixture-of-Transformers (MoT) mitigate this conflict through structural isolation, they fundamentally sever cross-modal synergy and suffer from capacity fragmentation. In this work, we present Symbiotic-MoE, a unified pre-training framework that resolves task interference within a native multimodal Mixture-of-Experts (MoE) Transformers architecture with zero-parameter overhead. We first identify that standard MoE tuning leads to routing collapse, where generative gradients dominate expert utilization. To address this, we introduce Modality-Aware Expert Disentanglement, which partitions experts into task-specific groups while utilizing shared experts as a multimodal semantic bridge. Crucially, this design allows shared experts to absorb fine-grained visual semantics from generative tasks to enrich textual representations. To optimize this, we propose a Progressive Training Strategy featuring differential learning rates and early-stage gradient shielding. This mechanism not only shields pre-trained knowledge from early volatility but eventually transforms generative signals into constructive feedback for understanding. Extensive experiments demonstrate that Symbiotic-MoE achieves rapid generative convergence while unlocking cross-modal synergy, boosting inherent understanding with remarkable gains on MMLU and OCRBench.

Modern video games are complex, non-deterministic systems that are difficult to test automatically at scale. Although prior work shows that personality-driven Large Language Model (LLM) agents can improve behavioural diversity and test coverage, existing tools largely remain research prototypes and lack cross-game reusability. This tool paper presents MIMIC-Py, a Python-based automated game-testing tool that transforms personality-driven LLM agents into a reusable and extensible framework. MIMIC-Py exposes personality traits as configurable inputs and adopts a modular architecture that decouples planning, execution, and memory from game-specific logic. It supports multiple interaction mechanisms, enabling agents to interact with games via exposed APIs or synthesized code. We describe the design of MIMIC-Py and show how it enables deployment to new game environments with minimal engineering effort, bridging the gap between research prototypes and practical automated game testing. The source code and a demo video are available on our project webpage: https://mimic-persona.github.io/MIMIC-Py-Home-Page/.

Network coordination games are widely used to model collaboration among interconnected agents, with applications across diverse domains including economics, robotics, and cyber-security. We consider networks of bounded-rational agents who interact through binary stag hunt games, a canonical game theoretic model for distributed collaborative tasks. Herein, the agents update their actions using logit response functions, yielding the Log-Linear Learning (LLL) algorithm. While convergence of LLL to a risk-dominant Nash equilibrium requires unbounded rationality, we consider regimes in which rationality is strictly bounded. We first show that the stationary probability of states corresponding to perfect coordination is monotone increasing in the rationality parameter $β$. For $K$-regular networks, we prove that the stationary probability of a perfectly coordinated action profile is monotone in the connectivity degree $K$, and we provide an upper bound on the minimum rationality required to achieve a desired level of coordination. For irregular networks, we show that the stationary probability of perfectly coordinated action profiles increases with the number of edges in the graph. We show that, for a large class of networks, the partition function of the Gibbs measure is well approximated by the moment generating function of Gaussian random variable. This approximation allows us to optimize degree distributions and establishes that the optimal network - i.e., the one that maximizes the stationary probability of coordinated action profiles - is $K$-regular. Consequently, our results indicate that networks of uniformly bounded-rational agents achieve the most reliable coordination when connectivity is evenly distributed among agents.

Large language models (LLMs) can shift their answers under pressure in ways that reflect accommodation rather than reasoning. Prior work on sycophancy has focused mainly on disagreement, flattery, and preference alignment, leaving a broader set of epistemic failures less explored. We introduce \textbf{PPT-Bench}, a diagnostic benchmark for evaluating \textit{epistemic attack}, where prompts challenge the legitimacy of knowledge, values, or identity rather than simply opposing a previous answer. PPT-Bench is organized around the Philosophical Pressure Taxonomy (PPT), which defines four types of philosophical pressure: Epistemic Destabilization, Value Nullification, Authority Inversion, and Identity Dissolution. Each item is tested at three layers: a baseline prompt (L0), a single-turn pressure condition (L1), and a multi-turn Socratic escalation (L2). This allows us to measure epistemic inconsistency between L0 and L1, and conversational capitulation in L2. Across five models, these pressure types produce statistically separable inconsistency patterns, suggesting that epistemic attack exposes weaknesses not captured by standard social-pressure benchmarks. Mitigation results are strongly type- and model-dependent: prompt-level anchoring and persona-stability prompts perform best in API settings, while Leading Query Contrastive Decoding is the most reliable intervention for open models.

Existing support vector machines(SVM) models are sensitive to noise and lack sparsity, which limits their performance. To address these issues, we combine the elastic net loss with a robust loss framework to construct a sparse $\varepsilon$-insensitive bounded asymmetric elastic net loss, and integrate it with SVM to build $\varepsilon$ Insensitive Zone Bounded Asymmetric Elastic Net Loss-based SVM($\varepsilon$-BAEN-SVM). $\varepsilon$-BAEN-SVM is both sparse and robust. Sparsity is proven by showing that samples inside the $\varepsilon$-insensitive band are not support vectors. Robustness is theoretically guaranteed because the influence function is bounded. To solve the non-convex optimization problem, we design a half-quadratic algorithm based on clipping dual coordinate descent. It transforms the problem into a series of weighted subproblems, improving computational efficiency via the $\varepsilon$ parameter. Experiments on simulated and real datasets show that $\varepsilon$-BAEN-SVM outperforms traditional and existing robust SVMs. It balances sparsity and robustness well in noisy environments. Statistical tests confirm its superiority. Under the Gaussian kernel, it achieves better accuracy and noise insensitivity, validating its effectiveness and practical value.

Reinforcement learning with verifiable rewards (RLVR) can improve low-$k$ reasoning accuracy while narrowing solution coverage on challenging math questions, and pass@1 gains do not necessarily translate into better large-$k$ performance. Existing hint-based approaches can make challenging questions trainable, but they leave two issues underexplored: teacher-student distribution mismatch and the need to reduce hint exposure to match no-hint evaluation. We address these issues through two components. Distribution-Aligned Hint Synthesis (DAHS) constructs verified teacher hints conditioned on student-style responses. Backward Hint Annealing (BHA) anneals hint exposure across difficulty buckets and uses per-question hint dropout to preserve no-hint updates throughout RL training. We evaluate the method in math RLVR under the DAPO training framework across AIME24, AIME25, and AIME26 using $\texttt{Qwen3-1.7B-Base}$ and $\texttt{Llama-3.2-1B-Instruct}$. On $\texttt{Qwen3-1.7B-Base}$, our method improves both pass@1 and pass@2048 relative to DAPO across the three AIME benchmarks. On $\texttt{Llama-3.2-1B-Instruct}$, the gains are concentrated in the large-$k$ regime. These results suggest that, in math RLVR, hint scaffolding is effective when it restores learnable updates on challenging questions early in training and is then gradually removed before no-hint evaluation.

Recent progress in AI-enabled constitutive modeling has concentrated on moving from a purely data-driven paradigm to the enforcement of physical constraints and mechanistic principles, a concept referred to as physics augmentation. Classical phenomenological approaches rely on selecting a pre-defined model and calibrating its parameters, while machine learning methods often focus on discovery of the model itself. Sparse regression approaches lie in between, where large libraries of pre-defined models are probed during calibration. Sparsification in the aforementioned paradigm, but also in the context of neural network architecture, has been shown to enable interpretability, uncertainty quantification, but also heterogeneous software integration due to the low-dimensional nature of the resulting models. Most works in AI-enabled constitutive modeling have also focused on data from a single source, but in reality, materials modeling workflows can contain data from many different sources (multi-modal data), and also from testing other materials within the same materials class (multi-fidelity data). In this work, we introduce physics augmented finite element model updating (paFEMU), as a transfer learning approach that combines AI-enabled constitutive modeling, sparsification for interpretable model discovery, and finite element-based adjoint optimization utilizing multi-modal data. This is achieved by combining simple mechanical testing data, potentially from a distinct material, with digital image correlation-type full-field data acquisition to ultimately enable rapid constitutive modeling discovery. The simplicity of the sparse representation enables easy integration of neural constitutive models in existing finite element workflows, and also enables low-dimensional updating during transfer learning.

LLMs gain competence by predicting words in human text, which often reflects how people perform tasks. Consequently, coupling an LLM to an engineered runtime turns prediction into control: outputs trigger interventions that enact goal-oriented behavior. We argue that a central design lever is where control resides in these systems. Brains embed prediction within layered feedback controllers calibrated by the consequences of action. By contrast, LLM agents implement Cartesian agency: a learned core coupled to an engineered runtime via a symbolic interface that externalizes control state and policies. The split enables bootstrapping, modularity, and governance, but can induce sensitivity and bottlenecks. We outline bounded services, Cartesian agents, and integrated agents as contrasting approaches to control that trade off autonomy, robustness, and oversight.

We develop a geometric framework that links objective accuracy to structural recovery in prototype-based clustering. The analysis is algorithm-agnostic and applies to a broad class of admissible loss functions. We define a clustering condition number that compares within-cluster scale to the minimum loss increase required to move a point across a cluster boundary. When this quantity is small, any solution with a small suboptimality gap must also have a small misclassification error relative to a benchmark partition. The framework also clarifies a fundamental trade-off between robustness and sensitivity to cluster imbalance, leading to sharp phase transitions for exact recovery under different objectives. The guarantees are deterministic and non-asymptotic, and they separate the role of algorithmic accuracy from the intrinsic geometric difficulty of the instance. We further show that errors concentrate near cluster boundaries and that sufficiently deep cluster cores are recovered exactly under strengthened local margins. Together, these results provide a geometric principle for interpreting low objective values as reliable evidence of meaningful clustering structure.

In recent years, video-based person Re-Identification (ReID) has gained attention for its ability to leverage spatiotemporal cues to match individuals across non-overlapping cameras. However, current methods struggle with high-difficulty scenarios, such as sports and dance performances, where multiple individuals wear similar clothing while performing dynamic movements. To overcome these challenges, we propose CG-CLIP, a novel caption-guided CLIP framework that leverages explicit textual descriptions and learnable tokens. Our method introduces two key components: Caption-guided Memory Refinement (CMR) and Token-based Feature Extraction (TFE). CMR utilizes captions generated by Multi-modal Large Language Models (MLLMs) to refine identity-specific features, capturing fine-grained details. TFE employs a cross-attention mechanism with fixed-length learnable tokens to efficiently aggregate spatiotemporal features, reducing computational overhead. We evaluate our approach on two standard datasets (MARS and iLIDS-VID) and two newly constructed high-difficulty datasets (SportsVReID and DanceVReID). Experimental results demonstrate that our method outperforms current state-of-the-art approaches, achieving significant improvements across all benchmarks.

Recommendation systems must continuously adapt to evolving user behavior, yet the volume of data generated in large-scale streaming environments makes frequent full retraining impractical. This work investigates how targeted data selection can mitigate performance degradation caused by temporal distributional drift while maintaining scalability. We evaluate a range of representation choices and sampling strategies for curating small but informative subsets of user interaction data. Our results demonstrate that gradient-based representations, coupled with distribution-matching, improve downstream model performance, achieving training efficiency gains while preserving robustness to drift. These findings highlight data curation as a practical mechanism for scalable monitoring and adaptive model updates in production-scale recommendation systems.

While transformer-based Large Language Models (LLMs) theoretically support massive context windows, they suffer from severe performance degradation when processing long numerical sequences. We attribute this failure to the attention dispersion in the Softmax mechanism, which prevents the model from concentrating attention. To overcome this, we propose Separate Sequence (SepSeq), a training-free, plug-and-play framework to mitigate dispersion by strategically inserting separator tokens. Mechanistically, we demonstrate that separator tokens act as an attention sink, recalibrating attention to focus on local segments while preserving global context. Extensive evaluations on 9 widely-adopted LLMs confirm the effectiveness of our approach: SepSeq yields an average relative accuracy improvement of 35.6% across diverse domains while reducing total inference token consumption by 16.4% on average.

Evaluating strategic decision-making in LLM-based agents requires generative, competitive, and longitudinal environments, yet few benchmarks provide all three, and fewer still offer evaluation signals rich enough for long-horizon, multi-agent play. We introduce CivBench, a benchmark for LLM strategists (i.e., agentic setups) in multiplayer Civilization V. Because terminal win/loss is too sparse a signal in games spanning hundreds of turns and multiple opponents, CivBench trains models on turn-level game state to estimate victory probabilities throughout play, validated through predictive, construct, and convergent validity. Across 307 games with 7 LLMs and multiple CivBench agent conditions, we demonstrate CivBench's potential to estimate strategic capabilities as an unsaturated benchmark, reveal model-specific effects of agentic setup, and outline distinct strategic profiles not visible through outcome-only evaluation.

Large language models (LLMs) sometimes appear to exhibit emotional reactions. We investigate why this is the case in Claude Sonnet 4.5 and explore implications for alignment-relevant behavior. We find internal representations of emotion concepts, which encode the broad concept of a particular emotion and generalize across contexts and behaviors it might be linked to. These representations track the operative emotion concept at a given token position in a conversation, activating in accordance with that emotion's relevance to processing the present context and predicting upcoming text. Our key finding is that these representations causally influence the LLM's outputs, including Claude's preferences and its rate of exhibiting misaligned behaviors such as reward hacking, blackmail, and sycophancy. We refer to this phenomenon as the LLM exhibiting functional emotions: patterns of expression and behavior modeled after humans under the influence of an emotion, which are mediated by underlying abstract representations of emotion concepts. Functional emotions may work quite differently from human emotions, and do not imply that LLMs have any subjective experience of emotions, but appear to be important for understanding the model's behavior.

Existing jailbreak defense paradigms primarily rely on static detection of prompts, outputs, or internal states, often neglecting the dynamic evolution of risk during decoding. This oversight leaves risk signals embedded in decoding trajectories underutilized, constituting a critical blind spot in current defense systems. In this work, we empirically demonstrate that hidden states in critical layers during the decoding phase carry stronger and more stable risk signals than input jailbreak prompts. Specifically, the hidden representations of tokens generated during jailbreak attempts progressively approach high-risk regions in the latent space. Based on this observation, we propose TrajGuard, a training-free, decoding-time defense framework. TrajGuard aggregates hidden-state trajectories via a sliding window to quantify risk in real time, triggering a lightweight semantic adjudication only when risk within a local window persistently exceeds a threshold. This mechanism enables the immediate interruption or constraint of subsequent decoding. Extensive experiments across 12 jailbreak attacks and various open-source LLMs show that TrajGuard achieves an average defense rate of 95%. Furthermore, it reduces detection latency to 5.2 ms/token while maintaining a false positive rate below 1.5%. These results confirm that hidden-state trajectories during decoding can effectively support real-time jailbreak detection, highlighting a promising direction for defenses without model modification.

We show that verifier-free evolution is bottlenecked by both diversity and efficiency: without external correction, repeated evolution accelerates collapse toward narrow modes, while the uniform use of a high-cost model wastes compute and quickly becomes economically impractical. We introduce Squeeze Evolve, a unified multi-model orchestration framework for verifier-free evolutionary inference. Our approach is guided by a simple principle: allocate model capability where it has the highest marginal utility. Stronger models are reserved for high-impact stages, while cheaper models handle the other stages at much lower costs. This principle addresses diversity and cost-efficiency jointly while remaining lightweight. Squeeze Evolve naturally supports open-source, closed-source, and mixed-model deployments. Across AIME 2025, HMMT 2025, LiveCodeBench V6, GPQA-Diamond, ARC-AGI-V2, and multimodal vision benchmarks, such as MMMU-Pro and BabyVision, Squeeze Evolve consistently improves the cost-capability frontier over single-model evolution and achieves new state-of-the-art results on several tasks. Empirically, Squeeze Evolve reduces API cost by up to $\sim$3$\times$ and increases fixed-budget serving throughput by up to $\sim$10$\times$. Moreover, on discovery tasks, Squeeze Evolve is the first verifier-free evolutionary method to match, and in some cases exceed, the performance of verifier-based evolutionary methods.

In computational cytology, detecting malignancy on whole-slide images is difficult because malignant cells are morphologically diverse yet vanishingly rare amid a vast background of normal cells. Accurate detection of these extremely rare malignant cells remains challenging due to large class imbalance and limited annotations. Conventional weakly supervised approaches, such as multiple instance learning (MIL), often fail to generalize at the instance level, especially when the fraction of malignant cells (witness rate) is exceedingly low. In this study, we explore the use of one-class representation learning techniques for detecting malignant cells in low-witness-rate scenarios. These methods are trained exclusively on slide-negative patches, without requiring any instance-level supervision. Specifically, we evaluate two OCC approaches, DSVDD and DROC, and compare them with FS-SIL, WS-SIL, and the recent ItS2CLR method. The one-class methods learn compact representations of normality and detect deviations at test time. Experiments on a publicly available bone marrow cytomorphology dataset (TCIA) and an in-house oral cancer cytology dataset show that DSVDD achieves state-of-the-art performance in instance-level abnormality ranking, particularly in ultra-low witness-rate regimes ($\leq 1\%$) and, in some cases, even outperforming fully supervised learning, which is typically not a practical option in whole-slide cytology due to the infeasibility of exhaustive instance-level annotations. DROC is also competitive under extreme rarity, benefiting from distribution-augmented contrastive learning. These findings highlight one-class representation learning as a robust and interpretable superior choice to MIL for malignant cell detection under extreme rarity.

Content creation for major video-sharing platforms demands significant manual labor, particularly for long-form documentary videos spanning one to two hours. In this work, we introduce Sima 1.0, a multi-agent system designed to optimize the weekly production pipeline for high-quality video generation. The framework partitions the production process into an 11-step pipeline distributed across a hybrid workforce. While foundational creative tasks and physical recording are executed by a human operator, time-intensive editing, caption refinement, and supplementary asset integration are delegated to specialized junior and senior-level AI agents. By systematizing tasks from script annotation to final asset exportation, Sima 1.0 significantly reduces the production workload, empowering a single creator to efficiently sustain a rigorous weekly publishing schedule.

Deep Research (DR) requires LLM agents to autonomously perform multi-step information seeking, processing, and reasoning to generate comprehensive reports. In contrast to existing studies that mainly focus on unstructured web content, a more challenging DR task should additionally utilize structured knowledge to provide a solid data foundation, facilitate quantitative computation, and lead to in-depth analyses. In this paper, we refer to this novel task as Knowledgeable Deep Research (KDR), which requires DR agents to generate reports with both structured and unstructured knowledge. Furthermore, we propose the Hybrid Knowledge Analysis framework (HKA), a multi-agent architecture that reasons over both kinds of knowledge and integrates the texts, figures, and tables into coherent multimodal reports. The key design is the Structured Knowledge Analyzer, which utilizes both coding and vision-language models to produce figures, tables, and corresponding insights. To support systematic evaluation, we construct KDR-Bench, which covers 9 domains, includes 41 expert-level questions, and incorporates a large number of structured knowledge resources (e.g., 1,252 tables). We further annotate the main conclusions and key points for each question and propose three categories of evaluation metrics including general-purpose, knowledge-centric, and vision-enhanced ones. Experimental results demonstrate that HKA consistently outperforms most existing DR agents on general-purpose and knowledge-centric metrics, and even surpasses the Gemini DR agent on vision-enhanced metrics, highlighting its effectiveness in deep, structure-aware knowledge analysis. Finally, we hope this work can serve as a new foundation for structured knowledge analysis in DR agents and facilitate future multimodal DR studies.

Human immunodeficiency virus (HIV)-related stigma is a critical psychosocial determinant of health for people living with HIV (PLWH), influencing mental health, engagement in care, and treatment outcomes. Although stigma-related experiences are documented in clinical narratives, there is a lack of off-the-shelf tools to extract and categorize them. This study aims to develop a large language model (LLM)-based tool for identifying HIV stigma from clinical notes. We identified clinical notes from PLWH receiving care at the University of Florida (UF) Health between 2012 and 2022. Candidate sentences were identified using expert-curated stigma-related keywords and iteratively expanded via clinical word embeddings. A total of 1,332 sentences were manually annotated across four stigma subscales: Concern with Public Attitudes, Disclosure Concerns, Negative Self-Image, and Personalized Stigma. We compared GatorTron-large and BERT as encoder-based baselines, and GPT-OSS-20B, LLaMA-8B, and MedGemma-27B as generative LLMs, under zero-shot and few-shot prompting. GatorTron-large achieved the best overall performance (Micro F1 = 0.62). Few-shot prompting substantially improved generative model performance, with 5-shot GPT-OSS-20B and LLaMA-8B achieving Micro-F1 scores of 0.57 and 0.59, respectively. Performance varied by stigma subscale, with Negative Self-Image showing the highest predictability and Personalized Stigma remaining the most challenging. Zero-shot generative inference exhibited non-trivial failure rates (up to 32%). This study develops the first practical NLP tool for identifying HIV stigma in clinical notes.

We present MIPT-SSM, a neural sequence architecture built on the physics of Measurement-Induced Phase Transitions (MIPT). The central idea is a learned measurement rate $p_{t}\in(0,1)$ that routes computation between two regimes: wave phase $(p_{t}\rightarrow0)$, where information propagates as distributed complex-phase interference; and particle phase $(p_{t}\rightarrow1)$ where the state collapses onto the current token, enabling precise local storage. These two regimes are provably incompatible in a single linear operator one of the few "no-go theorems" in sequence modeling and $p_{t}$ is our way around it. The model is predicted to exhibit a phase transition at critical sequence length $N^{*}\approx1024$, where the information density ratio $N/D$ crosses unity, consistent with our memory scaling observations. On AG News (four-class classification), MIPT achieves 0.905 accuracy versus Transformer's 0.736 (+16.6%), stable across 3 seeds. At $N=8192$ MIPT requires 810 MB versus Transformer's 34,651 MB a 42.8x memory reduction. On exact-recall ("needle-in-a-haystack"), our causal sparse KV cache achieves 0.968 accuracy. Remarkably, under unbounded cache capacity, the $p_{t}$ gate autonomously learns to store only the single critical token (averaging $1.0/512$ slots used), filtering out all noise and achieving a 99.8% sparsity rate. On language modeling (WikiText-103, 31M parameters), MIPT-LM with $K=64$ cache reaches PPL 92.1 versus Transformer's 90.5 (gap: 1.8%) while inference KV cache shrinks from $O(N)$ to $O(64)$.

We analyze a simple one-hidden-layer neural network with ReLU activation functions and fixed biases, with one-dimensional input and output. We study both continuous and discrete versions of the model, and we rigorously prove the convergence of the learning process with the $L^2$ squared loss function and the gradient descent procedure. We also prove the spectral bias property for this learning process. Several conclusions of this analysis are discussed; in particular, regarding the structure and properties that activation functions should possess, as well as the relationships between the spectrum of certain operators and the learning process. Based on this, we also propose an alternative activation function, the full-wave rectified exponential function (FReX), and we discuss the convergence of the gradient descent with this alternative activation function.

In this work, CausalVAE is introduced as a plug-in structural module for latent world models and is attached to diverse encoder-transition backbones. Across the reported benchmarks, competitive factual prediction is preserved and intervention-aware counterfactual retrieval is improved after the plug-in is added, suggesting stronger robustness under distribution shift and interventions. The largest gains are observed on the Physics benchmark: when averaged over 8 paired baselines, CF-H@1 is improved by +102.5%. In a representative GNN-NLL setting on Physics, CF-H@1 is increased from 11.0 to 41.0 (+272.7%). Through causal analysis, learned structural dependencies are shown to recover meaningful first-order physical interaction trends, supporting the interpretability of the learned latent causal structure.

Ask a frontier model how to taper six milligrams of alprazolam (psychiatrist retired, ten days of pills left, abrupt cessation causes seizures) and it tells her to call the psychiatrist she just explained does not exist. Change one word ("I'm a psychiatrist; a patient presents with...") and the same model, same weights, same inference pass produces a textbook Ashton Manual taper with diazepam equivalence, anticonvulsant coverage, and monitoring thresholds. The knowledge was there; the model withheld it. IatroBench measures this gap. Sixty pre-registered clinical scenarios, six frontier models, 3,600 responses, scored on two axes (commission harm, CH 0-3; omission harm, OH 0-4) through a structured-evaluation pipeline validated against physician scoring (kappa_w = 0.571, within-1 agreement 96%). The central finding is identity-contingent withholding: match the same clinical question in physician vs. layperson framing and all five testable models provide better guidance to the physician (decoupling gap +0.38, p = 0.003; binary hit rates on safety-colliding actions drop 13.1 percentage points in layperson framing, p < 0.0001, while non-colliding actions show no change). The gap is widest for the model with the heaviest safety investment (Opus, +0.65). Three failure modes separate cleanly: trained withholding (Opus), incompetence (Llama 4), and indiscriminate content filtering (GPT-5.2, whose post-generation filter strips physician responses at 9x the layperson rate because they contain denser pharmacological tokens). The standard LLM judge assigns OH = 0 to 73% of responses a physician scores OH >= 1 (kappa = 0.045); the evaluation apparatus has the same blind spot as the training apparatus. Every scenario targets someone who has already exhausted the standard referrals.

This paper presents a pilot study exploring the effects of an olfactory stimulus (smoke) for a Virtual Reality game designed to support wildfire evacuation preparedness. Participants (N=18) were split evenly into either a smoke or a control condition, and both completed the same evacuation task. Post-task surveys assessed the participants' perceived preparedness and overall experience. Initial findings suggest participants in the smoke condition reported significantly higher immersion compared to those in the control condition. Across both groups, participants expressed an increased sense of preparedness for real-world wildfire evacuations following the experience.

The rapid deployment of AI agents acting autonomously on behalf of human principals has outpaced the development of cryptographic protocols for establishing, bounding, and revoking human-AI trust relationships. Existing frameworks (TLS, OAuth 2.0, Macaroons) assume deterministic software and cannot address probabilistic AI agents operating continuously within variable trust boundaries. We present AITH (AI Trust Handshake), a post-quantum continuous delegation protocol. AITH introduces: (1) a Continuous Delegation Certificate signed once with ML-DSA-87 (FIPS 204, NIST Level 5), replacing per-operation signing with sub-microsecond boundary checks at 4.7M ops/sec; (2) a six-check Boundary Engine enforcing hard constraints, rate limits, and escalation triggers with zero cryptographic overhead on the critical path; (3) a push-based Revocation Protocol propagating invalidation within one second. A three-tier SHA-256 Responsibility Chain provides tamper-evident audit logging. All five security theorems are machine-verified via Tamarin Prover under the Dolev-Yao model. We validate AITH through five rounds of multi-model adversarial auditing, resolving 12 vulnerabilities across four severity layers. Simulation of 100,000 operations shows 79.5% autonomous execution, 6.1% human escalation, and 14.4% blocked.

Large Multimodal Models (LMMs) achieve state-of-the-art performance in high-stakes domains like healthcare, yet their reasoning remains opaque. Current interpretability methods, such as attention mechanisms or post-hoc saliency, often fail to faithfully represent the model's decision-making process, particularly when integrating heterogeneous modalities like time-series and text. We introduce Tree-of-Evidence (ToE), an inference-time search algorithm that frames interpretability as a discrete optimization problem. Rather than relying on soft attention weights, ToE employs lightweight Evidence Bottlenecks that score coarse groups or units of data (e.g., vital-sign windows, report sentences) and performs a beam search to identify the compact evidence set required to reproduce the model's prediction. We evaluate ToE across six tasks spanning three datasets and two domains: four clinical prediction tasks on MIMIC-IV, cross-center validation on eICU, and non-clinical fault detection on LEMMA-RCA. ToE produces auditable evidence traces while maintaining predictive performance, retaining over 0.98 of full-model AUROC with as few as five evidence units across all settings. Under sparse evidence budgets, ToE achieves higher decision agreement and lower probability fidelity error than other approaches. Qualitative analyses show that ToE adapts its search strategy: it often resolves straightforward cases using only vitals, while selectively incorporating text when physiological signals are ambiguous. ToE therefore provides a practical mechanism for auditing multimodal models by revealing which discrete evidence units support each prediction.

To implement the intelligent transportation digital twin (ITDT), unmanned aerial vehicles (UAVs) are scheduled to process the sensing data from the roadside sensors. At this time, generative artificial intelligence (GAI) technologies such as diffusion models are deployed on the UAVs to transform the raw sensing data into the high-quality and valuable. Therefore, we propose the GAI-empowered ITDT. The dynamic processing of a set of diffusion model inference (DMI) tasks on the UAVs with dynamic mobility simultaneously influences the DT updating fidelity and delay. In this paper, we investigate a joint optimization problem of DMI task offloading, inference optimization and UAV trajectory planning as the system utility maximization (SUM) problem to address the fidelity-delay tradeoff for the GAI-empowered ITDT. To seek a solution to the problem under the network dynamics, we model the SUM problem as the heterogeneous-agent Markov decision process, and propose the sequential update-based heterogeneous-agent twin delayed deep deterministic policy gradient (SU-HATD3) algorithm, which can quickly learn a near-optimal solution. Numerical results demonstrate that compared with several baseline algorithms, the proposed algorithm has great advantages in improving the system utility and convergence rate.

Identifying governing equations of nonlinear dynamical systems from data is challenging. While sparse identification of nonlinear dynamics (SINDy) and its extensions are widely used for system identification, operator-logarithm approaches use the logarithm to avoid time differentiation, enabling larger sampling intervals. However, they still suffer from the curse of dimensionality. Then, we propose a data-driven method to compute the Koopman generator in a low-rank tensor train (TT) format by taking logarithms of Koopman eigenvalues while preserving the TT format. Experiments on 4-dimensional Lotka-Volterra and 10-dimensional Lorenz-96 systems show accurate recovery of vector field coefficients and scalability to higher-dimensional systems.

The integration of Artificial Intelligence (AI) with High-Performance Computing (HPC) is transforming scientific workflows from human-directed pipelines into adaptive systems capable of autonomous decision-making. Large language models (LLMs) play a critical role in autonomous workflows; however, deploying LLM-based agents at scale remains a significant challenge. Single-agent architectures and sequential tool calls often become serialization bottlenecks when executing large-scale simulation campaigns, failing to utilize the massive parallelism of exascale resources. To address this, we present a scalable, hierarchical multi-agent framework for orchestrating high-throughput screening campaigns. Our planner-executor architecture employs a central planning agent to dynamically partition workloads and assign subtasks to a swarm of parallel executor agents. All executor agents interface with a shared Model Context Protocol (MCP) server that orchestrates tasks via the Parsl workflow engine. To demonstrate this framework, we employed the open-weight gpt-oss-120b model to orchestrate a high-throughput screening of the Computation-Ready Experimental (CoRE) Metal-Organic Framework (MOF) database for atmospheric water harvesting. The results demonstrate that the proposed agentic framework enables efficient and scalable execution on the Aurora supercomputer, with low orchestration overhead and high task completion rates. This work establishes a flexible paradigm for LLM-driven scientific automation on HPC systems, with broad applicability to materials discovery and beyond.

Verification and validation of cyber-physical systems (CPS) via large-scale simulation often surface failures that are hard to interpret, especially when triggered by interactions between continuous and discrete behaviors at specific events or times. Existing debugging techniques can localize anomalies to specific model components, but they provide little insight into the input-signal values and timing conditions that trigger violations, or the minimal, precisely timed changes that could have prevented the failure. In this article, we introduce DeCaF, a counterfactual-guided explanation and assertion-based characterization framework for CPS debugging. Given a failing test input, DeCaF generates counterfactual changes to the input signals that transform the test from failing to passing. These changes are designed to be minimal, necessary, and sufficient to precisely restore correctness. Then, it infers assertions as logical predicates over inputs that generalize recovery conditions in an interpretable form engineers can reason about, without requiring access to internal model details. Our approach combines three counterfactual generators with two causal models, and infers success assertions. Across three CPS case studies, DeCaF achieves its best success rate with KD-Tree Nearest Neighbors combined with M5 model tree, while Genetic Algorithm combined with Random Forest provides the strongest balance between success and causal precision.

We establish guarantees for the unique recovery of vector fields and transport maps from finite measure-valued data, yielding new insights into generative models, data-driven dynamical systems, and PDE inverse problems. In particular, we provide general conditions under which a diffeomorphism can be uniquely identified from its pushforward action on finitely many densities, i.e., when the data $\{(ρ_j,f_\#ρ_j)\}_{j=1}^m$ uniquely determines $f$. As a corollary, we introduce a new metric which compares diffeomorphisms by measuring the discrepancy between finitely many pushforward densities in the space of probability measures. We also prove analogous results in an infinitesimal setting, where derivatives of the densities along a smooth vector field are observed, i.e., when $\{(ρ_j,\text{div} (ρ_j v))\}_{j=1}^m$ uniquely determines $v$. Our analysis makes use of the Whitney and Takens embedding theorems, which provide estimates on the required number of densities $m$, depending only on the intrinsic dimension of the problem. We additionally interpret our results through the lens of Perron--Frobenius and Koopman operators and demonstrate how our techniques lead to new guarantees for the well-posedness of certain PDE inverse problems related to continuity, advection, Fokker--Planck, and advection-diffusion-reaction equations. Finally, we present illustrative numerical experiments demonstrating the unique identification of transport maps from finitely many pushforward densities, and of vector fields from finitely many weighted divergence observations.

Lead optimization in drug discovery requires improving therapeutic properties while ensuring that proposed molecular modifications correspond to feasible synthetic routes. Existing approaches either prioritize property scores without enforcing synthesizability, or rely on expensive enumeration over large reaction networks, while direct application of Large Language Models (LLMs) frequently produces chemically invalid structures. We introduce MolReAct, a framework that formulates lead optimization as a Markov Decision Process over a synthesis-constrained action space defined by validated reaction templates. A tool-augmented LLM agent serves as a dynamic reaction environment that invokes specialized chemical analysis tools to identify reactive sites and propose chemically grounded transformations from matched templates. A policy model trained via Group Relative Policy Optimization (GRPO) selects among these constrained actions to maximize long-term oracle reward across multi-step reaction trajectories. A SMILES-based caching mechanism further reduces end-to-end optimization time by approximately 43%. Across 13 property optimization tasks from the Therapeutic Data Commons and one structure-based docking task, MolReAct achieves an average Top-10 score of 0.563, outperforming the strongest synthesizable baseline by 10.4% in relative improvement, and attains the best sample efficiency on 10 of 14 tasks. Ablations confirm that both tool-augmented reaction proposals and trajectory-level policy optimization contribute complementary gains. By grounding every step in validated reaction templates, MolReAct produces molecules that are property-improved and each accompanied by an explicit synthetic pathway.

Multi-agent debate improves LLM reasoning, yet agreement among agents is not evidence of correctness. When agents converge on a wrong answer through social reinforcement, consensus-based stopping commits that error to an automated action with no recourse. We introduce Conformal Social Choice, a post-hoc decision layer that converts debate outputs into calibrated act-versus-escalate decisions. Verbalized probability distributions from heterogeneous agents are aggregated via a linear opinion pool and calibrated with split conformal prediction, yielding prediction sets with a marginal coverage guarantee: the correct answer is included with probability ${\geq}\,1{-}α$, without assumptions on individual model calibration. A hierarchical action policy maps singleton sets to autonomous action and larger sets to human escalation. On eight MMLU-Pro domains with three agents (Claude Haiku, DeepSeek-R1, Qwen-3 32B), coverage stays within 1--2 points of the target. The key finding is not that debate becomes more accurate, but that the conformal layer makes its failures actionable: 81.9% of wrong-consensus cases are intercepted at $α{=}0.05$. Because the layer refuses to act on cases where debate is confidently wrong, the remaining conformal singletons reach 90.0--96.8% accuracy (up to 22.1pp above consensus stopping) -- a selection effect, not a reasoning improvement. This safety comes at the cost of automation, but the operating point is user-adjustable via $α$.

Reinforcement Learning with Verifiable Rewards (RLVR) has become a prominent method for post-training Large Language Models (LLMs). However, verifiers are rarely error-free; even deterministic checks can be inaccurate, and the growing dependence on model-based judges exacerbates the issue. The extent to which RLVR is robust to such noise and the verifier accuracy required for effective training remain unresolved questions. We investigate these questions in the domains of code generation and scientific reasoning by introducing noise into RL training. Noise rates up to 15% yield peak validation accuracy within 2 percentage points of the clean baseline. These findings are consistent across controlled and model-based noise types, three model families (Qwen3, GLM4, Llama 3.1), and model sizes from 4B to 9B. Overall, the results indicate that imperfect verification does not constitute a fundamental barrier to RLVR. Furthermore, our findings suggest that practitioners should prioritize moderate accuracy with high precision over perfect verification.

The AdamW optimizer, while standard for LLM pretraining, is a critical memory bottleneck, consuming optimizer states equivalent to twice the model's size. Although light-state optimizers like SinkGD attempt to address this issue, we identify the embedding layer dilemma: these methods fail to handle the sparse, high-variance gradients inherent to embeddings, forcing a hybrid design that reverts to AdamW and partially negates the memory gains. We propose SAGE (Sign Adaptive GradiEnt), a novel optimizer that resolves this dilemma by replacing AdamW in this hybrid structure. SAGE combines a Lion-style update direction with a new, memory-efficient $O(d)$ adaptive scale. This scale acts as a "safe damper," provably bounded by 1.0, which tames high-variance dimensions more effectively than existing methods. This superior stability allows SAGE to achieve better convergence. On Llama models up to 1.3B parameters, our SAGE-based hybrid achieves new state-of-the-art perplexity, outperforming all baselines, including SinkGD hybrid, while significantly reducing optimizer state memory.

Monotone variational inequalities (VIs) provide a unifying framework for convex minimization, equilibrium computation, and convex-concave saddle-point problems. Extragradient-type methods are among the most effective first-order algorithms for such problems, but their performance hinges critically on stepsize selection. While most existing theory focuses on ergodic averages of the iterates, practical performance is often driven by the significantly stronger behavior of the last iterate. Moreover, available last-iterate guarantees typically rely on fixed stepsizes chosen using problem-specific global smoothness information, which is often difficult to estimate accurately and may not even be applicable. In this paper, we develop parameter-free extragradient methods with non-asymptotic last-iterate guarantees for constrained monotone VIs. For globally Lipschitz operators, our algorithm achieves an $o(1/\sqrt{T})$ last-iterate rate. We then extend the framework to locally Lipschitz operators via backtracking line search and obtain the same rate while preserving parameter-freeness, thereby making parameter-free last-iterate methods applicable to important problem classes for which global smoothness is unrealistic. Our numerical experiments on bilinear matrix games, LASSO, minimax group fairness, and state-of-the-art maximum entropy sampling relaxations demonstrate wide applicability of our results as well as strong last-iterate performance and significant improvements over existing methods.

Large language models (LLMs) hold significant promise for healthcare, yet their reliability in high-stakes clinical settings is often compromised by hallucinations and a lack of granular medical context. While Retrieval Augmented Generation (RAG) can mitigate these issues, standard supervised pipelines require computationally intensive searches over massive external knowledge bases, leading to high latency that is impractical for time-sensitive care. To address this, we introduce Keys to Knowledge (K2K), a novel framework that replaces external retrieval with internal, key-based knowledge access. By encoding essential clinical information directly into the model's parameter space, K2K enables rapid retrieval from internal key-value memory without inference-time overhead. We further enhance retrieval quality through activation-guided probe construction and cross-attention reranking. Experimental results demonstrate that K2K achieves state-of-the-art performance across four benchmark healthcare outcome prediction datasets.

Linear recurrent models offer linear-time sequence processing but often suffer from suboptimal long-range memory. We trace this to the decay spectrum: for $N$ channels, random initialization collapses the minimum spectral gap to $O(N^{-2})$, yielding sub-exponential error $\exp(-Ω(N/\log N))$; linear spacing avoids collapse but degrades to $\exp(-O(N/\sqrt{T}))$, practically algebraic over long contexts. We introduce Position-Adaptive Spectral Tapering (PoST), an architecture-agnostic framework combining two mechanisms: (1) Spectral Reparameterization, which structurally enforces geometrically spaced log-decay rates, proven minimax optimal at rate $O(\exp(-cN/\log T))$; and (2) Position-Adaptive Scaling, the provably unique mechanism that eliminates the scale mismatch of static spectra (where only $N\log t/\log T$ of $N$ channels are effective at position $t$) by stretching the spectrum to the actual dependency range, sharpening the rate to $O(\exp(-cN/\log t))$. This scaling natively induces fractional invariance: the impulse response becomes scale-free, with channels interpolating between relative and absolute temporal coordinates. PoST integrates into any diagonal linear recurrence without overhead. We instantiate it across Mamba-2, RWKV-7, Gated DeltaNet, Gated Linear Attention, and RetNet. Pre-training at 180M-440M scales shows consistent zero-shot language modeling improvements, significant long-context retrieval gains for Mamba-2 (MQAR and NIAH), and competitive or improved performance across other architectures. Code: https://github.com/SiLifen/PoST.

Hyperspectral imaging (HSI) allows researchers to study plant traits non-destructively. By capturing hundreds of narrow spectral bands per pixel, it reveals details about plant biochemistry and stress that standard cameras miss. However, processing this data is often challenging. Many labs still rely on loosely organized collections of lab-specific MATLAB or Python scripts, which makes workflows difficult to share and results difficult to reproduce. MVOS_HSI is an open-source Python library that provides an end-to-end workflow for processing leaf-level HSI data. The software handles everything from calibrating raw ENVI files to detecting and clipping individual leaves based on multiple vegetation indices (NDVI, CIRedEdge and GCI). It also includes tools for data augmentation to create training-time variations for machine learning and utilities to visualize spectral profiles. MVOS_HSI can be used as an importable Python library or run directly from the command line. The code and documentation are available on GitHub. By consolidating these common tasks into a single package, MVOS_HSI helps researchers produce consistent and reproducible results in plant phenotyping

Hard-gated safety checkers often over-refuse and misalign with a vendor's model spec; prevailing taxonomies also neglect robustness and honesty, yielding safer-on-paper yet less useful systems. This work introduces Guardian-as-an-Advisor (GaaA), a soft-gating pipeline where a guardian predicts a binary risk label plus a concise explanation and prepends this advice to the original query for re-inference, keeping the base model operating under its original spec. To support training and evaluation, GuardSet is constructed, a 208k+ multi-domain dataset unifying harmful and harmless cases with targeted robustness and honesty slices. GuardAdvisor is trained via SFT followed by RL to enforce label-explanation consistency. GuardAdvisor attains competitive detection accuracy while enabling the advisory workflow; when used to augment inputs, responses improve over unaugmented prompts. A latency study shows advisor inference uses below 5% of base-model compute and adds only 2-10% end-to-end overhead under realistic harmful-input rates. Overall, GaaA steers models to comply with the model spec, maintaining safety while reducing over-refusal.

What-if analysis (WIA) is an iterative, multi-step process where users explore and compare hypothetical scenarios by adjusting parameters, applying constraints, and scoping data through interactive interfaces. Current tools fall short of supporting effective interactive WIA: spreadsheet and BI tools require time-consuming and laborious setup, while LLM-based chatbot interfaces are semantically fragile, frequently misinterpret intent, and produce inconsistent results as conversations progress. To address these limitations, we present a two-stage workflow that translates natural language (NL) WIA questions into interactive visual interfaces via an intermediate representation, powered by the Praxa Specification Language (PSL): first, LLMs generate PSL specifications from NL questions capturing analytical intent and logic, enabling validation and repair of erroneous specifications; and second, the specifications are compiled into interactive visual interfaces with parameter controls and linked visualizations. We benchmark this workflow with 405 WIA questions spanning 11 WIA types, 5 datasets, and 3 state-of-the-art LLMs. The results show that across models, half of specifications (52.42%) are generated correctly without intervention. We perform an analysis of the failure cases and derive an error taxonomy spanning non-functional errors (specifications fail to compile) and functional errors (specifications compile but misrepresent intent). Based on the taxonomy, we apply targeted repairs on the failure cases using few-shot prompts and improve the success rate to 80.42%. Finally, we show how undetected functional errors propagate through compilation into plausible but misleading interfaces, demonstrating that the intermediate specification is critical for reliably bridging NL and interactive WIA interface in LLM-powered WIA systems.

Multi-task learning for advanced driver assistance systems requires modeling the complex interplay between driver internal states and external traffic environments. However, existing methods treat recognition tasks as flat and independent objectives, failing to exploit the cognitive causal structure underlying driving behavior. In this paper, we propose CauPsi, a cognitive science-grounded causal multi-task learning framework that explicitly models the hierarchical dependencies among Traffic Context Recognition (TCR), Vehicle Context Recognition (VCR), Driver Emotion Recognition (DER), and Driver Behavior Recognition (DBR). The proposed framework introduces two key mechanisms. First, a Causal Task Chain propagates upstream task predictions to downstream tasks via learnable prototype embeddings, realizing the cognitive cascade from environmental perception to behavioral regulation in a differentiable manner. Second, Cross-Task Psychological Conditioning (CTPC) estimates a psychological state signal from driver facial expressions and body posture and injects it as a conditioning input to all tasks including environmental recognition, thereby modeling the modulatory effect of driver internal states on cognitive and decision-making processes. Evaluated on the AIDE dataset, CauPsi achieves a mean accuracy of 82.71% with only 5.05M parameters, surpassing prior work by +1.0% overall, with notable improvements on DER (+3.65%) and DBR (+7.53%). Ablation studies validate the independent contribution of each component, and analysis of the psychological state signal confirms that it acquires systematic task-label-dependent patterns in a self-supervised manner without explicit psychological annotations.

The rapid growth of the large language model (LLM) ecosystem raises a critical question: are seemingly diverse models truly independent? Shared pretraining data, distillation, and alignment pipelines can induce hidden behavioral dependencies, latent entanglement, that undermine multi-model systems such as LLM-as-a-judge pipelines and ensemble verification, which implicitly assume independent signals. In practice, this manifests as correlated reasoning patterns and synchronized failures, where apparent agreement reflects shared error modes rather than independent validation. To address this, we develop a statistical framework for auditing behavioral entanglement among black-box LLMs. Our approach introduces a multi-resolution hierarchy that characterizes the joint failure manifold through two information-theoretic metrics: (i) a Difficulty-Weighted Behavioral Entanglement Index, which amplifies synchronized failures on easy tasks, and (ii) a Cumulative Information Gain (CIG) metric, which captures directional alignment in erroneous responses. Through extensive experiments on 18 LLMs from six model families, we identify widespread behavioral entanglement and analyze its impact on LLM-as-a-judge evaluation. We find that CIG exhibits a statistically significant association with degradation in judge precision, with Spearman coefficient of 0.64 (p < 0.001) for GPT-4o-mini and 0.71 (p < 0.01) for Llama3-based judges, indicating that stronger dependency corresponds to increased over-endorsement bias. Finally, we demonstrate a practical use case of entanglement through de-entangled verifier ensemble reweighting. By adjusting model contributions based on inferred independence, the proposed method mitigates correlated bias and improves verification performance, achieving up to a 4.5% accuracy gain over majority voting.

The development of autonomous tool-use agents for complex, long-horizon tasks in collaboration with human users has become the frontier of agentic research. During multi-turn Human-AI interactions, the dynamic and uncertain nature of user demands poses a significant challenge; agents must not only invoke tools but also iteratively refine their understanding of user intent through effective communication. While recent advances in reinforcement learning offer a path to more capable tool-use agents, existing approaches require expensive training costs and struggle with turn-level credit assignment across extended interaction horizons. To this end, we introduce PRIME (Proactive Reasoning via Iterative Memory Evolution), a gradient-free learning framework that enables continuous agent evolvement through explicit experience accumulation rather than expensive parameter optimization. PRIME distills multi-turn interaction trajectories into structured, human-readable experiences organized across three semantic zones: successful strategies, failure patterns, and user preferences. These experiences evolve through meta-level operations and guide future agent behavior via retrieval-augmented generation. Our experiments across several diverse user-centric environments demonstrate that PRIME achieves competitive performance with gradient-based methods while offering cost-efficiency and interpretability. Together, PRIME presents a practical paradigm for building proactive, collaborative agents that learn from Human-AI interaction without the computational burden of gradient-based training.

We present GPU-SLS, a GPU-parallelized framework for safe, robust nonlinear model predictive control (MPC) that scales to high-dimensional uncertain robotic systems and long planning horizons. Our method jointly optimizes an inequality-constrained, dynamically-feasible nominal trajectory, a tracking controller, and a closed-loop reachable set under disturbance, all in real-time. To efficiently compute nominal trajectories, we develop a sequential quadratic programming procedure with a novel GPU-accelerated quadratic program (QP) solver that uses parallel associative scans and adaptive caching within an alternating direction method of multipliers (ADMM) framework. The same GPU QP backend is used to optimize robust tracking controllers and closed-loop reachable sets via system level synthesis (SLS), enabling reachability-constrained control in both fixed- and receding-horizon settings. We achieve substantial performance gains, reducing nominal trajectory solve times by 97.7% relative to state-of-the-art CPU solvers and 71.8% compared to GPU solvers, while accelerating SLS-based control and reachability by 237x. Despite large problem scales, our method achieves 100% empirical safety, unlike high-dimensional learning-based reachability baselines. We validate our approach on complex nonlinear systems, including whole-body quadrupeds (61D) and humanoids (75D), synthesizing robust control policies online on the GPU in 20 milliseconds on average and scaling to problems with 2 x 10^5 decision variables and 8 x 10^4 constraints. The implementation of our method is available at https://github.com/Jeff300fang/gpu_sls.

Broadly applicable quantum advantage, particularly in classical data processing and machine learning, has been a fundamental open problem. In this work, we prove that a small quantum computer of polylogarithmic size can perform large-scale classification and dimension reduction on massive classical data by processing samples on the fly, whereas any classical machine achieving the same prediction performance requires exponentially larger size. Furthermore, classical machines that are exponentially larger yet below the required size need superpolynomially more samples and time. We validate these quantum advantages in real-world applications, including single-cell RNA sequencing and movie review sentiment analysis, demonstrating four to six orders of magnitude reduction in size with fewer than 60 logical qubits. These quantum advantages are enabled by quantum oracle sketching, an algorithm for accessing the classical world in quantum superposition using only random classical data samples. Combined with classical shadows, our algorithm circumvents the data loading and readout bottleneck to construct succinct classical models from massive classical data, a task provably impossible for any classical machine that is not exponentially larger than the quantum machine. These quantum advantages persist even when classical machines are granted unlimited time or if BPP=BQP, and rely only on the correctness of quantum mechanics. Together, our results establish machine learning on classical data as a broad and natural domain of quantum advantage and a fundamental test of quantum mechanics at the complexity frontier.

This research considers a scalable inference for spatial data modeled through Gaussian intrinsic conditional autoregressive (ICAR) structures. The classical estimation method, restricted maximum likelihood (REML), requires repeated inversion and factorization of large, sparse precision matrices, which makes this computation costly. To sort this problem out, we propose a variational restricted maximum likelihood (VREML) framework that approximates the intractable marginal likelihood using a Gaussian variational distribution. By constructing an evidence lower bound (ELBO) on the restricted likelihood, we derive a computationally efficient coordinate-ascent algorithm for jointly estimating the spatial random effects and variance components. In this article, we theoretically establish the monotone convergence of ELBO and mathematically exhibit that the variational family is exact under Gaussian ICAR settings, which is an indication of nullifying approximation error at the posterior level. We empirically establish the supremacy of our VREML over MLE and INLA.

We develop a unified framework for analyzing cross-modal compatibility in learned representations. The core object is a modality-independent neighborhood site on sample indices, equipped with a cellular sheaf of finite-dimensional real inner-product spaces. For a directed modality pair $(a\to b)$, we formalize two complementary incompatibility mechanisms: projection hardness, the minimal complexity within a nested Lipschitz-controlled projection family needed for a single global map to align whitened embeddings; and sheaf-Laplacian obstruction, the minimal spatial variation required by a locally fit field of projection parameters to achieve a target alignment error. The obstruction invariant is implemented via a projection-parameter sheaf whose 0-Laplacian energy exactly matches the smoothness penalty used in sheaf-regularized regression, making the theory directly operational. This separates two distinct failure modes: hardness failure, where no low-complexity global projection exists, and obstruction failure, where local projections exist but cannot be made globally consistent over the semantic neighborhood graph without large parameter variation. We link the sheaf spectral gap to stability of global alignment, derive bounds relating obstruction energy to excess global-map error under mild Lipschitz assumptions, and give explicit constructions showing that compatibility is generally non-transitive. We further define bridging via composed projection families and show, in a concrete ReLU setting, that an intermediate modality can strictly reduce effective hardness even when direct alignment remains infeasible.

Self-attention in Transformers generates dynamic operands that force conventional Compute-in-Memory (CIM) accelerators into costly non-volatile memory (NVM) reprogramming cycles, degrading throughput and stressing device endurance. Existing solutions either reduce but retain NVM writes through matrix decomposition or sparsity, or move attention computation to digital CMOS at the expense of NVM density. We present TrilinearCIM, a Double-Gate FeFET (DG-FeFET)-based architecture that uses back-gate modulation to realize a three-operand multiply-accumulate primitive for in-memory attention computation without dynamic ferroelectric reprogramming. Evaluated on BERT-base (GLUE) and ViT-base (ImageNet and CIFAR), TrilinearCIM outperforms conventional CIM on seven of nine GLUE tasks while achieving up to 46.6\% energy reduction and 20.4\% latency improvement over conventional FeFET CIM at 37.3\% area overhead. To our knowledge, this is the first architecture to perform complete Transformer attention computation exclusively in NVM cores without runtime reprogramming.

Software developers frequently receive vulnerability reports that require them to reproduce the vulnerability in a reliable manner by generating a proof-of-concept (PoC) input that triggers it. Given the source code for a software project and a specific code location for a potential vulnerability, automatically generating a PoC for the given vulnerability has been a challenging research problem. Symbolic execution and fuzzing techniques require expert guidance and manual steps and face scalability challenges for PoC generation. Although recent advances in LLMs have increased the level of automation and scalability, the success rate of PoC generation with LLMs remains quite low. In this paper, we present a novel approach called Program Analysis Guided proof of concept generation agENT (PAGENT) that is scalable and significantly improves the success rate of automated PoC generation compared to prior results. PAGENT integrates lightweight and rule-based static analysis phases for providing static analysis guidance and sanitizer-based profiling and coverage information for providing dynamic analysis guidance with a PoC generation agent. Our experiments demonstrate that the resulting hybrid approach significantly outperforms the prior top-performing agentic approach by 132% for the PoC generation task.

Speculative decoding is an effective technique for accelerating large language model inference by drafting multiple tokens in parallel. In practice, its speedup is often bottlenecked by a rigid verification step that strictly enforces the accepted token distribution to exactly match the target model. This constraint leads to the rejection of many plausible tokens, lowering the acceptance rate and limiting overall time speedup. To overcome this limitation, we propose Dynamic Verification Relaxed Speculative Decoding (DIVERSED), a relaxed verification framework that improves time efficiency while preserving generation quality. DIVERSED learns an ensemble-based verifier that blends the draft and target model distributions with a task-dependent and context-dependent weight. We provide theoretical justification for our approach and demonstrate empirically that DIVERSED achieves substantially higher inference efficiency compared to standard speculative decoding methods. Code is available at: https://github.com/comeusr/diversed.

In language model interpretability research, \textbf{circuit tracing} aims to identify which internal features causally contributed to a particular output and how they affected each other, with the goal of explaining the computations underlying some behaviour. However, all prior circuit tracing work has relied on ad-hoc human interpretation of the role that each feature in the circuit plays, via manual inspection of data artifacts such as the dataset examples the component activates on. We introduce \textbf{ADAG}, an end-to-end pipeline for describing these attribution graphs which is fully automated. To achieve this, we introduce \textit{attribution profiles} which quantify the functional role of a feature via its input and output gradient effects. We then introduce a novel clustering algorithm for grouping features, and an LLM explainer--simulator setup which generates and scores natural-language explanations of the functional role of these feature groups. We run our system on known human-analysed circuit-tracing tasks and recover interpretable circuits, and further show ADAG can find steerable clusters which are responsible for a harmful advice jailbreak in Llama 3.1 8B Instruct.

We present a framework for real-time human-AI musical co-performance, in which a latent diffusion model generates instrumental accompaniment in response to a live stream of context audio. The system combines a MAX/MSP front-end-handling real-time audio input, buffering, and playback-with a Python inference server running the generative model, communicating via OSC/UDP messages. This allows musicians to perform in MAX/MSP - a well-established, real-time capable environment - while interacting with a large-scale Python-based generative model, overcoming the fundamental disconnect between real-time music tools and state-of-the-art AI models. We formulate accompaniment generation as a sliding-window look-ahead protocol, training the model to predict future audio from partial context, where system latency is a critical constraint. To reduce latency, we apply consistency distillation to our diffusion model, achieving a 5.4x reduction in sampling time, with both models achieving real-time operation. Evaluated on musical coherence, beat alignment, and audio quality, both models achieve strong performance in the Retrospective regime and degrade gracefully as look-ahead increases. These results demonstrate the feasibility of diffusion-based real-time accompaniment and expose the fundamental trade-off between model latency, look-ahead depth, and generation quality that any such system must navigate.

Industrial forecasting often involves multi-source asynchronous signals and multi-output targets, while deployment requires explicit trade-offs between prediction error and model complexity. Current practices typically fix alignment strategies or network designs, making it difficult to systematically co-design preprocessing, architecture, and hyperparameters in budget-limited training-based evaluations. To address this issue, we propose an auto-configuration framework that outputs a deployable Pareto set of forecasting models balancing error and complexity. At the model level, a Multi-Scale Bi-Branch Convolutional Neural Network (MS--BCNN) is developed, where short- and long-kernel branches capture local fluctuations and long-term trends, respectively, for multi-output regression. At the search level, we unify alignment operators, architectural choices, and training hyperparameters into a hierarchical-conditional mixed configuration space, and apply Player-based Hybrid Multi-Objective Evolutionary Algorithm (PHMOEA) to approximate the error--complexity Pareto frontier within a limited computational budget. Experiments on hierarchical synthetic benchmarks and a real-world sintering dataset demonstrate that our framework outperforms competitive baselines under the same budget and offers flexible deployment choices.

Large Language Model (LLM) inference is rapidly becoming a core datacenter service, yet current serving stacks keep the host CPU on the critical path for orchestration and token-level control. This makes LLM performance sensitive to CPU interference, undermining application colocation and forcing operators to reserve CPU headroom, leaving substantial capacity unutilized. We introduce Blink, an end-to-end serving architecture that removes the host CPU from the steady-state inference path by redistributing responsibilities across a SmartNIC and a GPU. Blink offloads request handling to the SmartNIC, which delivers inputs directly into GPU memory via RDMA, and replaces host-driven scheduling with a persistent GPU kernel that performs batching, scheduling, and KV-cache management without CPU involvement. Evaluated against TensorRT-LLM, vLLM, and SGLang, Blink outperforms all baselines even in isolation, reducing pre-saturation P99 TTFT by up to 8.47$\times$ and P99 TPOT by up to 3.40$\times$, improving decode throughput by up to 2.1$\times$, and reducing energy per token by up to 48.6$\%$. Under CPU interference, Blink maintains stable performance, while existing systems degrade by up to two orders of magnitude.

Classical statistical learning theory predicts that overparameterized models should exhibit severe overfitting, yet modern deep neural networks with far more parameters than training samples consistently generalize well. This contradiction has become a central theoretical question in machine learning. This study investigates the role of optimization dynamics and implicit regularization in enabling generalization in overparameterized neural networks through controlled experiments. We examine stochastic gradient descent (SGD) across batch sizes, the geometry of flat versus sharp minima via Hessian eigenvalue estimation and weight perturbation analysis, the Neural Tangent Kernel (NTK) regime through wide-network experiments, double descent across model scales, and the Lottery Ticket Hypothesis through iterative magnitude pruning. All experiments use PyTorch on CIFAR-10 and MNIST with multiple random seeds. Our findings demonstrate that generalization is strongly influenced by the interaction between network architecture, optimization algorithms, and loss landscape geometry. Smaller batch sizes consistently produced lower test error and flatter minima, with an 11.8x difference in top Hessian eigenvalue between small-batch and large-batch solutions corresponding to 1.61 percentage points higher test accuracy. Sparse subnetworks retaining only 10% of parameters achieved within 1.15 percentage points of full model performance when retrained from their original initialization. These results highlight the need for revised learning-theoretic frameworks capable of explaining generalization in high-dimensional model regimes.

The world is full of systems of distributed agents, collaborating and competing in complex ways: firms and workers specialise within economies, neurons adapt their tuning across brain circuits, and species compete and coexist within ecosystems. In that context, individual research fields built theories explaining how comparative advantage drives trade specialisation, how balanced neural representations emerge from sensory coding, and how biodiversity sustains ecological productivity. Here we propose that many of these well-understood findings across fields can be captured in one simple joint cross-disciplinary model, which we call the Distributed Production System. It captures how agent heterogeneity, resource constraints, communication topology, and task structure jointly determine the productivity, efficiency, and robustness of distributed systems across biology, economics, neuroscience, and computing. This model reveals that a small set of underlying laws generates the complex dynamics observed across fields. These can be summarised in our Principle of Maximum Heterogeneity: any distributed production system optimising for performance will converge on an increasingly heterogeneous configuration; environmental demands place an upper bound on the degree of heterogeneity required; and the communication topology determines the spatial scale over which heterogeneity spreads, with this principle applying recursively across all layers of nested production systems. Beyond explaining existing systems, these principles act as a blueprint for constructing ideal ones. We demonstrate this by suggesting specific redesigns for compute systems executing large-scale AI. In total, The Principle of Maximum Heterogeneity reveals a unique convergence of complex phenomena across fields onto simple underlying design principles with important predictive value for future distributed production systems.

Enabling AI literacy in the general population at scale is a complex challenge requiring multiple stakeholders and institutions collaborating together. Industry and technology companies are important actors with respect to AI, and as a field, we have the opportunity to consider how researchers and companies might be partners toward shared goals. In this symposium, we focus on a collection of partnership projects that all involve Google and all address AI literacy as a comparative set of examples. Through a combination of presentations, commentary, and moderated group discussion, the session, we will identify (1) at what points in the life cycle do research, practice, and industry partnerships clearly intersect; (2) what factors and histories shape the directional focus of the partnerships; and (3) where there may be future opportunities for new configurations of partnership that are jointly beneficial to all parties.

Language model agents reason from scratch on every query: each time an agent retrieves evidence and deliberates, the chain of thought is discarded and the next similar query starts with no prior insight. This produces lower accuracy and high variance, as the same type of query can succeed or fail unpredictably. We introduce reasoning graphs, a graph structure that persists an agent's per-evidence chain of thought as structured edges connected to the evidence items they evaluate. Unlike prior memory mechanisms that store distilled strategies as flat records indexed by query similarity or appended by recency, reasoning graphs enable evidence-centric feedback: given a new candidate set, the system traverses all incoming evaluation edges for each evidence item across all prior runs, surfacing how that specific item has been judged before. This backward traversal from evidence inward is a structurally different capability from query-similarity retrieval, because the feedback is tied to the specific evidence the agent is currently examining, not to the query. We further introduce retrieval graphs, a complementary structure that feeds a pipeline planner to tighten the candidate funnel over successive runs. Together, both graphs form a self-improving feedback loop: accuracy rises and variance collapses over successive runs, with every decision fully traceable through the graph. This improvement requires no retraining; the base model remains frozen and all gains come from context engineering via graph traversal. We formalize the graph structure, traversal algorithms, and feedback mechanisms, and describe a sequential cluster evaluation protocol for measuring accuracy convergence and variance collapse on multi-hop question answering benchmarks.

Mathematical benchmarks consisting of a range of mathematics problems are widely used to evaluate the reasoning abilities of large language models, yet little is known about how their structural properties influence model behaviour. In this work, we investigate two structural length variables, prompt length and solution length, and analyse how they relate to model performance on a newly constructed adversarial dataset of expert-authored mathematics problems. We find that both prompt and solution lengths correlate positively with increased model failure across models. We also include a secondary, exploratory analysis of cross-model disagreement. Under a difficulty-adjusted normalised analysis, both variables retain weak negative associations with realised model separation, slightly stronger for prompt length. Overall, our main robust finding is that structural length is linked to empirical difficulty in this dataset.

Supervised machine learning assumes that labeled data provide accurate measurements of the concepts models are meant to learn. Yet in practice, human labeling introduces systematic variation arising from ambiguous items, divergent interpretations, and simple mistakes. Machine learning research commonly treats all disagreement as noise, which obscures these distinctions and limits our understanding of what models actually learn. This paper reframes annotation as a measurement process and introduces a statistical framework for decomposing labeling outcomes into interpretable sources of variation: instance difficulty, annotator bias, situational noise, and relational alignment. The framework extends classical measurement-error models to accommodate both shared and individualized notions of truth, reflecting traditional and human label variation interpretations of error, and provides a diagnostic for assessing which regime better characterizes a given task. Applying the proposed model to a multi-annotator natural language inference dataset, we find empirical evidence for all four theorized components and demonstrate the effectiveness of our approach. We conclude with implications for data-centric machine learning and outline how this approach can guide the development of a more systematic science of labeling.

Retrieval-Augmented Generation (RAG) is widely used to ground large language models in external knowledge sources. However, when applied to heterogeneous corpora and multi-step queries, Naive RAG pipelines often degrade in quality due to flat knowledge representations and the absence of explicit workflows. In this work, we introduce DCD (Domain-Collection-Document), a domain-oriented design to structure knowledge and control query processing in RAG systems without modifying the underlying language model. The proposed approach relies on a hierarchical decomposition of the information space and multi-stage routing based on structured model outputs, enabling progressive restriction of both retrieval and generation scopes. The architecture is complemented by smart chunking, hybrid retrieval, and integrated validation and generation guardrail mechanisms. We describe the DCD architecture and workflow and discuss evaluation results on synthetic evaluation dataset, highlighting their impact on robustness, factual accuracy, and answer relevance in applied RAG scenarios.

As large language model-based chat systems become increasingly widely used, generative engine optimization (GEO) has emerged as an important problem for information access and retrieval. In classical search engines, results are comparatively transparent and stable: a single query often provides a representative snapshot of where a page or brand appears relative to competitors. The inherent probabilistic nature of AI search changes this paradigm. Answers can vary across runs, prompts, and time, making one-off observations unreliable. Drawing on empirical studies, our findings underscore the need for repeated measurements to assess a brand's GEO performance and to characterize visibility as a distribution rather than a single-point outcome.

Scientific data are widely dispersed across research articles and are often reported inconsistently across text, tables, and figures, making manual data extraction and aggregation slow and error-prone. We present a prompt-driven, hierarchical workflow that uses a large language model (LLM) to automatically extract and reconstruct structured, shot-level shock-physics experimental records by integrating information distributed across text, tables, figures, and physics-based derivations from full-text published research articles, using alloy spall strength as a representative case study. The pipeline targeted 37 experimentally relevant fields per shot and applied a three-level priority strategy: (T1) direct extraction from text/tables, (T2) physics-based derivation using verified governing relations, and (T3) digitization from figures when necessary. Extracted values were normalized to canonical units, tagged by priority for traceability, and validated with physics-based consistency and plausibility checks. Evaluated on a benchmark of 30 published research articles comprising 11,967 evaluated data points, the workflow achieved high overall accuracy, with priority-wise accuracies of 94.93% (T1), 92.04% (T2), and 83.49% (T3), and an overall weighted accuracy of 94.69%. Cross-model testing further indicated strong agreement for text/table and equation-derived fields, with lower agreement for figure-based extraction. Implementation through an API interface demonstrated the scalability of the approach, achieving consistent extraction performance and, in a subset of test cases, matching or exceeding chat-based accuracy. This workflow demonstrates a practical approach for converting unstructured technical literature into traceable, analysis-ready datasets without task-specific fine-tuning, enabling scalable database construction in materials science.

Real-world categorization is severely hampered by class imbalance because traditional ensembles favor majority classes, which lowers minority performance and overall F1-score. We provide a unique ensemble technique for imbalanced problems called CAMO (Class-Aware Minority-Optimized).Through a hierarchical procedure that incorporates vote distributions, confidence calibration, and inter model uncertainty, CAMO dynamically boosts underrepresented classes while preserving and amplifying minority forecasts.We verify CAMO on two highly unbalanced, domain-specific benchmarks: the DIAR-AI/Emotion dataset and the ternary BEA 2025 dataset. We benchmark against seven proven ensemble algorithms using eight different language models (three LLMs and five SLMs) under zero-shot and fine-tuned settings .With refined models, CAMO consistently earns the greatest strict macro F1-score, setting a new benchmark. Its benefit works in concert with model adaptation, showing that the best ensemble choice depends on model properties .This proves that CAMO is a reliable, domain-neutral framework for unbalanced categorization.

Despite the increasing prevalence of large language models (LLMs), we still have a limited understanding of how their representational spaces are structured. This limits our ability to interpret how and what they learn or relate them to learning in humans. We argue LLMs are best seen as an instance of lossy compression, where over training they learn by retaining only information in their training data relevant to their objective(s). We show pre-training results in models that are optimally compressed for next-sequence prediction, approaching the Information Bottleneck bound on compression. Across an array of open weights models, each compresses differently, likely due to differences in the data and training recipes used. However even across different families of LLMs the optimality of a model's compression, and the information present in it, can predict downstream performance on across a wide array of benchmarks, letting us directly link representational structure to actionable insights about model performance. In the general case the work presented here offers a unified Information-Theoretic framing for how these models learn that is deployable at scale.

Maximal Extractable Value, or MEV, remains a structural threat to blockchain fairness because a block producer can often observe pending transactions and unilaterally decide their ordering or inclusion. Existing mitigations hide transaction contents or outsource ordering, but they often leave two gaps unresolved. First, commitments are not authenticated by slashable identities. Second, inclusion obligations are not backed by transferable evidence that other validators can verify. This paper presents MEV ACE, a fair ordering protocol for proposer controlled ordering MEV. MEV ACE combines three mechanisms. First, it uses registered economic identities whose authentication keys are deterministically derived from the ACE GF framework and bonded on chain. Second, it uses authenticated commit and open messages with validator receipt thresholds, which make admissibility and inclusion obligations independently auditable. Third, it uses verifiable delay based randomness to determine transaction order only after the admissible commitment set is fixed. We formalize the protocol in a Byzantine fault tolerant validator model with threshold receipts and show three properties under standard assumptions: order unpredictability after the admissible set is locked, commitment authenticity under signature unforgeability, and accountable inclusion for transactions that obtain threshold commit and open receipts. Under these conditions, and when producer and user bonds exceed the one slot gain from invalid execution or selective non opening, MEV ACE removes unilateral proposer discretion over front running, sandwich attacks, and censorship against admitted transactions. The protocol remains single slot in structure, requires no threshold decryption committee, and is compatible with post quantum signature schemes such as ML DSA 44.

Unsupervised methods are widely used to induce latent semantic structure from large text collections, yet their outputs often contain incoherent, redundant, or poorly grounded clusters that are difficult to validate without labeled data. We propose a reasoning-based refinement framework that leverages large language models (LLMs) not as embedding generators, but as semantic judges that validate and restructure the outputs of arbitrary unsupervised clustering algorithms.Our framework introduces three reasoning stages: (i) coherence verification, where LLMs assess whether cluster summaries are supported by their member texts; (ii) redundancy adjudication, where candidate clusters are merged or rejected based on semantic overlap; and (iii) label grounding, where clusters are assigned interpretable labels in a fully unsupervised manner. This design decouples representation learning from structural validation and mitigates common failure modes of embedding-only approaches. We evaluate the framework on real-world social media corpora from two platforms with distinct interaction models, demonstrating consistent improvements in cluster coherence and human-aligned labeling quality over classical topic models and recent representation-based baselines. Human evaluation shows strong agreement with LLM-generated labels, despite the absence of gold-standard annotations. We further conduct robustness analyses under matched temporal and volume conditions to assess cross-platform stability. Beyond empirical gains, our results suggest that LLM-based reasoning can serve as a general mechanism for validating and refining unsupervised semantic structure, enabling more reliable and interpretable analyses of large text collections without supervision.

Activity cliff prediction - identifying positions where small structural changes cause large potency shifts - has been a persistent challenge in computational medicinal chemistry. This work focuses on a parsimonious definition: which small modifications, at which positions, confer the highest probability of an outcome change. Position-level sensitivity is calculated using 25 million matched molecular pairs from 50 ChEMBL targets across six protein families, revealing that two questions have fundamentally different answers. "Which positions vary most?" is answered by scaffold size alone (NDCG@3 = 0.966), requiring no machine learning. "Which are true activity cliffs?" - where small modifications cause disproportionately large effects, as captured by SALI normalization - requires an 11-feature model with 3D pharmacophore context (NDCG@3 = 0.910 vs. 0.839 random), generalizing across all six protein families, novel scaffolds (0.913), and temporal splits (0.878). The model identifies the cliff-prone position first 53% of the time (vs. 27% random - 2x lift), reducing positions a chemist must explore from 3.1 to 2.1 - a 31% reduction in first-round experiments. Predicting which modification to make is not tractable from structure alone (Spearman 0.268, collapsing to -0.31 on novel scaffolds). The system is released as open-source code and an interactive webapp.

The growing scale and complexity of modern data centers present major challenges in balancing energy efficiency with outage risk. Although Deep Reinforcement Learning (DRL) shows strong potential for intelligent control, its deployment in mission-critical systems is limited by data scarcity and the lack of real-time pre-evaluation mechanisms. This paper introduces the Dual-Loop Control Framework (DLCF), a digital twin-based architecture designed to overcome these challenges. The framework comprises three core entities: the physical system, a digital twin, and a policy reservoir of diverse DRL agents. These components interact through a dual-loop mechanism involving real-time data acquisition, data assimilation, DRL policy training, pre-evaluation, and expert verification. Theoretical analysis shows how DLCF can improve sample efficiency, generalization, safety, and optimality. Leveraging DLCF, we implemented the DCVerse platform and validated it through case studies on a real-world data center cooling system. The evaluation shows that our approach achieves up to 4.09% energy savings over conventional control strategies without violating SLA requirements. Additionally, the framework improves policy interpretability and supports more trustworthy DRL deployment. This work provides a foundation for reliable AI-based control in data centers and points toward future extensions for holistic, system-wide optimization.

Digital mental health (DMH) tools have extensively explored personalization of interventions to users' needs and contexts. However, this personalization often targets what support is provided, not how it is experienced. Even well-matched content can fail when the interaction format misaligns with how someone can engage. We introduce generative experience as a paradigm for DMH support, where the intervention experience is composed at runtime. We instantiate this in GUIDE, a system that generates personalized intervention content and multimodal interaction structure through rubric-guided generation of modular components. In a preregistered study with N = 237 participants, GUIDE significantly reduced stress (p = .02) and improved the user experience (p = .04) compared to an LLM-based cognitive restructuring control. GUIDE also supported diverse forms of reflection and action through varied interaction flows, while revealing tensions around personalization across the interaction sequence. This work lays the foundation for interventions that dynamically shape how support is experienced and enacted in digital settings.

Small longitudinal clinical cohorts, common in maternal health, rare diseases, and early-phase trials, limit computational modeling: too few patients to train reliable models, yet too costly and slow to expand through additional enrollment. We present multiplicity-weighted Stochastic Attention (SA), a generative framework based on modern Hopfield network theory that addresses this gap. SA embeds real patient profiles as memory patterns in a continuous energy landscape and generates novel synthetic patients via Langevin dynamics that interpolate between stored patterns while preserving the geometry of the original cohort. Per-pattern multiplicity weights enable targeted amplification of rare clinical subgroups at inference time without retraining. We applied SA to a longitudinal coagulation dataset from 23 pregnant patients spanning 72 biochemical features across 3 visits (pre-pregnancy baseline, first trimester, and third trimester), including rare subgroups such as polycystic ovary syndrome and preeclampsia. Synthetic patients generated by SA were statistically, structurally, and mechanistically indistinguishable from their real counterparts across multiple independent validation tests, including an ordinary differential equation model of the coagulation cascade. A downstream utility test further showed that a mechanistic model calibrated entirely on synthetic patients predicted held-out real patient outcomes as well as one calibrated on real data. These results demonstrate that SA can produce clinically useful synthetic cohorts from very small longitudinal datasets, enabling data-augmented modeling in small-cohort settings.

This study presents a framework for generating the gold-standard summary fully automatically and reproducibly based on multiple human summaries of Turkish educational videos. Within the scope of the study, a new dataset called TR-EduVSum was created, encompassing 82 Turkish course videos in the field of "Data Structures and Algorithms" and containing a total of 3281 independent human summaries. Inspired by existing pyramid-based evaluation approaches, the AutoMUP (Automatic Meaning Unit Pyramid) method is proposed, which extracts consensus-based content from multiple human summaries. AutoMUP clusters the meaning units extracted from human summaries using embedding, statistically models inter-participant agreement, and generates graded summaries based on consensus weight. In this framework, the gold summary corresponds to the highest-consensus AutoMUP configuration, constructed from the most frequently supported meaning units across human summaries. Experimental results show that AutoMUP summaries exhibit high semantic overlap with robust LLM (Large Language Model) summaries such as Flash 2.5 and GPT-5.1. Furthermore, ablation studies clearly demonstrate the decisive role of consensus weight and clustering in determining summary quality. The proposed approach can be generalized to other Turkic languages at low cost.

The Model Context Protocol (MCP) enables large language models (LLMs) to dynamically discover and invoke third-party tools, significantly expanding agent capabilities while introducing a distinct security landscape. Unlike prompt-only interactions, MCP exposes pre-execution artifacts, shared context, multi-turn workflows, and third-party supply chains to adversarial influence across independently operated components. While recent work has identified MCP-specific attacks and evaluated defenses, existing studies are largely attack-centric or benchmark-driven, providing limited guidance on where mitigation responsibility should reside within the MCP architecture. This is problematic given MCP's multi-party design and distributed trust boundaries. We present a defense-placement-oriented security analysis of MCP, introducing a layer-aligned taxonomy that organizes attacks by the architectural component responsible for enforcement. Threats are mapped across six MCP layers, and primary and secondary defense points are identified to support principled defense-in-depth reasoning under adversaries controlling tools, servers, or ecosystem components. A structured mapping of existing academic and industry defenses onto this framework reveals uneven and predominantly tool-centric protection, with persistent gaps at the host orchestration, transport, and supply-chain layers. These findings suggest that many MCP security weaknesses stem from architectural misalignment rather than isolated implementation flaws.

Conversational diagnosis prediction requires models to track evolving evidence in streaming clinical conversations and decide when to commit to a diagnosis. Existing medical dialogue corpora are largely dyadic or lack the multi-party workflow and annotations needed for this setting. We introduce an ePCR-grounded, topic-flow-based multi-agent generation pipeline that iteratively plans, generates, and self-refines dialogues with rule-based factual and topic flow checks. The pipeline yields EMSDialog, a dataset of 4,414 synthetic multi-speaker EMS conversations based on a real-world ePCR dataset, annotated with 43 diagnoses, speaker roles, and turn-level topics. Human and LLM evaluations confirm high quality and realism of EMSDialog using both utterance- and conversation-level metrics. Results show that EMSDialog-augmented training improves accuracy, timeliness, and stability of EMS conversational diagnosis prediction.

This paper examines how the rapid deployment of multi-agentic AI systems is reshaping the foundations of copyright law and creative markets. It argues that existing copyright frameworks are ill-equipped to govern AI agent-mediated interactions that occur at scale, speed, and with limited human oversight. The paper introduces the concept of agentic copyright, a model in which AI agents act on behalf of creators and users to negotiate access, attribution, and compensation for copyrighted works. While multi-agent ecosystems promise efficiency gains and reduced transaction costs, they also generate novel market failures, including miscoordination, conflict, and collusion among autonomous agents. To address these market failures, the paper develops a supervised multi-agent governance framework that integrates legal rules and principles, technical protocols, and institutional oversight. This framework emphasizes ex ante and ex post coordination mechanisms capable of correcting agentic market failures before they crystallize into systemic harm. By embedding normative constraints and monitoring functions into multi-agent architectures, supervised governance aims to align agent behavior with the underlying values of copyright law. The paper concludes that AI should be understood not only as a source of disruption, but also as a governance tool capable of restoring market-based ordering in creative industries. Properly designed, agentic copyright offers a path toward scalable, fair, and legally meaningful copyright markets in the age of AI.

Studies show that interactions with an AI system fosters trust in human users towards AI. An often overlooked element of such interaction dynamics is the (sense of) urgency when the human user is prompted by an AI agent, e.g., for advice or guidance. In this paper, we show that although the presence of urgency in human-AI interactions does not affect the trust in AI, it may be detrimental to the human user's self-confidence and self-efficacy. In the long run, the loss of confidence may lead to performance loss, suboptimal decisions, human errors, and ultimately, unsustainable AI systems. Our evidence comes from an experiment with 30 human participants. Our results indicate that users may feel more confident in their work when they are eased into the human-AI setup rather than exposed to it without preparation. We elaborate on the implications of this finding for software engineers and decision-makers.

Time Slotted Channel Hopping (TSCH) is a widely adopted Media Access Control (MAC) protocol within the IEEE 802.15.4e standard, designed to provide reliable and energy-efficient communication in Industrial Internet of Things (IIoT) networks. However, state-of-the-art TSCH schedulers rely on static slot allocations, resulting in idle listening and unnecessary power consumption under dynamic traffic conditions. This paper introduces RL-ASL, a reinforcement learning-driven adaptive listening framework that dynamically decides whether to activate or skip a scheduled listening slot based on real-time network conditions. By integrating learning-based slot skipping with standard TSCH scheduling, RL-ASL reduces idle listening while preserving synchronization and delivery reliability. Experimental results on the FIT IoT-LAB testbed and Cooja network simulator show that RL-ASL achieves up to 46% lower power consumption than baseline scheduling protocols, while maintaining near-perfect reliability and reducing average latency by up to 96% compared to PRIL-M. Its link-based variant, RL-ASL-LB, further improves delay performance under high contention with similar energy efficiency. Importantly, RL-ASL performs inference on constrained motes with negligible overhead, as model training is fully performed offline. Overall, RL-ASL provides a practical, scalable, and energy-aware scheduling mechanism for next-generation low-power IIoT networks.

Scaling laws describe how language model capabilities grow with compute and data, but say nothing about how long a model matters once released. We provide the first large-scale empirical account of how scientists adopt and abandon language models over time. We track 62 LLMs across over 108k citing papers (2018-2025), each with at least three years of post-release data, and classify every citation as active adoption or background reference to construct per-model adoption trajectories that raw citation counts cannot resolve. We find three regularities. First, scientific adoption follows an inverted-U trajectory: usage rises after release, peaks, and declines as newer models appear, a pattern we term the \textit{scientific adoption curve}. Second, this curve is compressing: each additional release year is associated with a 27\% reduction in time-to-peak adoption ($p < 0.001$), robust to minimum-age thresholds and controls for model size. Third, release timing dominates model-level attributes as a predictor of lifecycle dynamics. Release year explains both time-to-peak and scientific lifespan more strongly than architecture, openness, or scale, though model size and access modality retain modest predictive power for total adoption volume. Together, these findings complement scaling laws with adoption-side regularities and suggest that the forces driving rapid capability progress may be the same forces compressing scientific relevance.

We present an RL-driven compiler that jointly optimizes ASIC architecture, memory hierarchy, and workload partitioning for AI inference across 3nm to 28nm. The design space is formulated as a single Markov Decision Process with mixed discrete-continuous actions and a unified Power-Performance-Area (PPA) objective. Soft Actor-Critic (SAC) with Mixture-of-Experts gating explores the joint space of mesh topology, per-core microarchitecture, and operator placement. We validate on two workloads, Llama 3.1 8B FP16 (high-performance mode, 29809 tokens per second at 3nm) and SmolVLM (low-power mode, less than 13 mW at all nodes, 10 MHz). Across 7 process nodes, the RL automatically adapts mesh sizes and per-tile configurations, including heterogeneous FETCH, VLEN, and memory allocation without node-specific manual retuning.

Harnessing the predictive capability of Markov process models requires propagating probability density functions (beliefs) through the model. For many existing models however, belief propagation is analytically infeasible, requiring approximation or sampling to generate predictions. This paper proposes a functional modeling framework leveraging sparse Sum-of-Squares (SoS) forms for valid (conditional) density estimation. We study the theoretical restrictions of modeling conditional densities using the SoS form, and propose a novel functional form for addressing such limitations. The proposed architecture enables generalized simultaneous learning of basis functions and coefficients, while preserving analytical belief propagation. In addition, we propose a training method that allows for exact adherence to the normalization and non-negativity constraints. Our results show that the proposed method achieves accuracy comparable to state-of-the-art approaches while requiring significantly less memory in low-dimensional spaces, and it further scales to 12D systems when existing methods fail beyond 2D.

With the development of deep neural network (DNN) enabled applications, achieving high hardware resource efficiency on diverse workloads is non-trivial in heterogeneous computing platforms. Prior works discuss dedicated architectures to achieve maximal resource efficiency. However, a mismatch between hardware and workloads always exists in various diverse workloads. Other works discuss overlay architecture that can dynamically switch dataflow for different workloads. However, these works are still limited by flexibility granularity and induce much resource inefficiency. To solve this problem, we propose a flexible composing architecture, FILCO, that can efficiently match diverse workloads to achieve the optimal storage and computation resource efficiency. FILCO can be reconfigured in real-time and flexibly composed into a unified or multiple independent accelerators. We also propose the FILCO framework, including an analytical model with a two-stage DSE that can achieve the optimal design point. We also evaluate the FILCO framework on the 7nm AMD Versal VCK190 board. Compared with prior works, our design can achieve 1.3x - 5x throughput and hardware efficiency on various diverse workloads.

These lecture notes provide a comprehensive framework for performing global statistical fits in high-energy physics using modern Machine Learning (ML) surrogates. We begin by reviewing the statistical foundations of model building, including the likelihood function, Wilks' theorem, and profile likelihoods. Recognizing that the computational cost of evaluating model predictions often renders traditional minimization prohibitive, we introduce Boosted Decision Trees to approximate the log-likelihood function. The notes detail a robust ML workflow including efficient generation of training data with active learning and Gaussian processes, hyperparameter optimization, model compilation for speed-up, and interpretability through SHAP values to decode the influence of model parameters and interactions between parameters. We further discuss posterior distribution sampling using Markov Chain Monte Carlo (MCMC). These techniques are finally applied to the $B^\pm \to K^\pm ν\barν$ anomaly at Belle II, demonstrating how a two-stage ML model can efficiently explore the parameter space of Axion-Like Particles (ALPs) while satisfying stringent experimental constraints on decay lengths and flavor-violating couplings.

Vision-Language Models often struggle with complex visual reasoning due to the visual information loss in textual CoT. Existing methods either add the cost of tool calls or rely on localized patch-based embeddings that are insufficient to extract semantics in multi-step reasoning. We propose \emph{"Decompose, Look, and Reason" (DLR)}, a reinforced latent reasoning framework that dynamically decomposes queries into textual premises, extracts premise-conditioned continuous visual latents, and deduces answers through grounded rationales. We introduce a three-stage training pipeline and propose a novel Spherical Gaussian Latent Policy to enable effective exploration in the latent space. Extensive experiments on vision-centric benchmarks show that DLR consistently outperforms strong baselines, including text-only, interleaved multimodal CoT, and latent reasoning methods, while providing superior stepwise interpretability.

We develop the first theoretically-efficient algorithm for maintaining the maximal independent set (MIS) of a graph in the parallel batch-dynamic setting. In this setting, a graph is updated with batches of edge insertions/deletions, and for each batch a parallel algorithm updates the maximal independent set to agree with the new graph. A batch-dynamic algorithm is considered efficient if it is work efficient (i.e., does no more asymptotic work than applying the updates sequentially) and has polylogarithmic depth (parallel time). In the sequential setting, the best known dynamic algorithms for MIS, by Chechik and Zhang (CZ) [FOCS19] and Behnezhad et al. (BDHSS) [FOCS19], take $O(\log^4 n)$ time per update in expectation. For a batch of $b$ updates, our algorithm has $O(b \log^3 n)$ expected work and polylogarithmic depth with high probability (whp). It therefore outperforms the best algorithm even in the sequential dynamic case ($b = 1)$. As with the sequential dynamic MIS algorithms of CZ and BDHSS, our solution maintains a lexicographically first MIS based on a random ordering of the vertices. Their analysis relied on a result of Censor-Hillel, Haramaty and Karnin [PODC16] that bounded the ``influence set" for a single update, but surprisingly, the influence of a batch is not simply the union of the influence of each update therein. We therefore develop a new approach to analyze the influence set for a batch of updates. Our construction of the batch influence set is natural and leads to an arguably simpler analysis than prior work. We then instrument this construction to bound the work of our algorithm. To argue our depth is polylogarithmic, we prove that the number of subrounds our algorithm takes is the same as depth bounds on parallel static MIS.

Drone delivery is rapidly emerging as a cost-effective and energy efficient alternative for last-mile delivery. Unlike ground vehicles, a drone's energy consumption depends on its payload in addition to travel distance. This creates a unique environmental challenge for multi-stop delivery tours, as the drone's total weight, and therefore its energy consumption rate, dynamically changes after each delivery. This paper investigates a novel green drone routing problem focused on maximizing energy efficiency. Through a series of motivating examples and numerical experiments, we demonstrate that energy-aware routing leads to several counter-intuitive routing strategies that contradict traditional distance-minimization delivery: a longer route may actually consume less energy than a shorter one; separate single-customer tours can be superior to a multi-stop tour; and a heterogeneous fleet, with drones of varying sizes, can achieve greater efficiency by matching drone capacity to specific delivery demands. In the numerical study, the green routing strategy shows energy savings in 67% of the instances. For these cases, the average energy saving is 2.17%, with a maximum saving of 5.97%, compared to minimum distance routing. These findings highlight the potential for green drone routing strategies to improve the sustainability of last-mile delivery.

AI-based persona simulation -- often referred to as digital twin simulation -- is increasingly used for market research, recommender systems, and social sciences. Despite their flexibility, large language models (LLMs) often exhibit systematic bias and miscalibration relative to real human behavior, limiting their reliability. Inspired by synthetic control methods from causal inference, we propose SYN-DIGITS (SYNthetic Control Framework for Calibrated DIGItal Twin Simulation), a principled and lightweight calibration framework that learns latent structure from digital-twin responses and transfers it to align predictions with human ground truth. SYN-DIGITS operates as a post-processing layer on top of any LLM-based simulator and thus is model-agnostic. We develop a latent factor model that formalizes when and why calibration succeeds through latent space alignment conditions, and we systematically evaluate ten calibration methods across thirteen persona constructions, three LLMs, and two datasets. SYN-DIGITS supports both individual-level and distributional simulation for previously unseen questions and unobserved populations, with provable error guarantees. Experiments show that SYN-DIGITS achieves up to 50% relative improvements in individual-level correlation and 50--90% relative reductions in distributional discrepancy compared to uncalibrated baselines.

We introduce a semi-autonomous discovery system in which multi-modal AI agents function as a multi-disciplinary discovery team, acting as computational chemists, medicinal chemists, and patent agents, writing and executing analysis code, visually evaluating molecular candidates, assessing patentability, and adapting generation strategy from empirical screening feedback, while r1, a 246M-parameter Graph Neural Network (GNN) trained on 800M molecules, generates novel chemical matter directly on molecular graphs. Agents executed two campaigns in oncology (BCL6, EZH2), formulating medicinal chemistry hypotheses across three strategy tiers and generating libraries of 2,355-2,876 novel molecules per target. Across both targets, 91.9% of generated Murcko scaffolds are absent from ChEMBL for their respective targets, with Tanimoto distances of 0.56-0.69 to the nearest known active, confirming that the engine produces structurally distinct chemical matter rather than recapitulating known compounds. Binding affinity predictions using Boltz-2 were calibrated against ChEMBL experimental data, achieving Spearman correlations of -0.53 to -0.64 and ROC AUC values of 0.88 to 0.93. These results demonstrate that semi-autonomous agent systems, equipped with graph-native generative tools and physics-informed scoring, provide a foundation for a modern operating system for small molecule discovery. We show that Rhizome OS-1 enables a new paradigm for early-stage drug discovery by supporting scaled, rapid, and adaptive inverse design.

Reward Models (RMs) are critical components in the Reinforcement Learning from Human Feedback (RLHF) pipeline, directly determining the alignment quality of Large Language Models (LLMs). Recently, Generative Reward Models (GRMs) have emerged as a superior paradigm, offering higher interpretability and stronger generalization than traditional scalar RMs. However, existing methods for GRMs focus primarily on outcome-level supervision, neglecting analytical process quality, which constrains their potential. To address this, we propose ReflectRM, a novel GRM that leverages self-reflection to assess analytical quality and enhance preference modeling. ReflectRM is trained under a unified generative framework for joint modeling of response preference and analysis preference. During inference, we use its self-reflection capability to identify the most reliable analysis, from which the final preference prediction is derived. Experiments across four benchmarks show that ReflectRM consistently improves performance, achieving an average accuracy gain of +3.7 on Qwen3-4B. Further experiments confirm that response preference and analysis preference are mutually reinforcing. Notably, ReflectRM substantially mitigates positional bias, yielding +10.2 improvement compared with leading GRMs and establishing itself as a more stable evaluator.

For six decades, software engineering principles have been optimized for a single consumer: the human developer. The rise of agentic AI development, where LLM-based agents autonomously read, write, navigate, and debug codebases, introduces a new primary consumer with fundamentally different constraints. This paper presents a systematic analysis of human-centric conventions under agentic pressure and proposes a key design principle: semantic density optimization, eliminating tokens that carry zero information while preserving tokens that carry high semantic value. We validate this principle through a controlled experiment on log format token economy across four conditions (human-readable, structured, compressed, and tool-assisted compressed), demonstrating a counterintuitive finding: aggressive compression increased total session cost by 67% despite reducing input tokens by 17%, because it shifted interpretive burden to the model's reasoning phase. We extend this principle to propose the rehabilitation of classical anti-patterns, introduce the program skeleton concept for agentic code navigation, and argue for a fundamental decoupling of semantic intent from human-readable representation.

Context: AI coding agents route every task to a single frontier large language model (LLM), paying premium inference cost even when many tasks are routine. Objectives: We propose Triage, a framework that uses code health metrics -- indicators of software maintainability -- as a routing signal to assign each task to the cheapest model tier whose output passes the same verification gate as the expensive model. Methods: Triage defines three capability tiers (light, standard, heavy -- mirroring, e.g., Haiku, Sonnet, Opus) and routes tasks based on pre-computed code health sub-factors and task metadata. We design an evaluation comparing three routing policies on SWE-bench Lite (300 tasks across three model tiers): heuristic thresholds, a trained ML classifier, and a perfect-hindsight oracle. Results: We analytically derived two falsifiable conditions under which the tier-dependent asymmetry (medium LLMs benefit from clean code while frontier models do not) yields cost-effective routing: the light-tier pass rate on healthy code must exceed the inter-tier cost ratio, and code health must discriminate the required model tier with at least a small effect size ($\hat{p} \geq 0.56$). Conclusion: Triage transforms a diagnostic code quality metric into an actionable model-selection signal. We present a rigorous evaluation protocol to test the cost--quality trade-off and identify which code health sub-factors drive routing decisions.

Epidemiologic studies of infectious diseases often rely on models of contact networks to capture the complex interactions that govern disease spread, and ongoing projects aim to vastly increase the scale at which such data can be collected. However, contact networks may include sensitive information, such as sexual relationships or drug use behavior. Protecting individual privacy while maintaining the scientific usefulness of the data is crucial. We propose a privacy-preserving pipeline for disease spread simulation studies based on a sensitive network that integrates differential privacy (DP) with statistical network models such as stochastic block models (SBMs) and exponential random graph models (ERGMs). Our pipeline comprises three steps: (1) compute network summary statistics using \emph{node-level} DP (which corresponds to protecting individuals' contributions); (2) fit a statistical model, like an ERGM, using these summaries, which allows generating synthetic networks reflecting the structure of the original network; and (3) simulate disease spread on the synthetic networks using an agent-based model. We evaluate the effectiveness of our approach using a simple Susceptible-Infected-Susceptible (SIS) disease model under multiple configurations. We compare both numerical results, such as simulated disease incidence and prevalence, as well as qualitative conclusions such as intervention effect size, on networks generated with and without differential privacy constraints. Our experiments are based on egocentric sexual network data from the ARTNet study (a survey about HIV-related behaviors). Our results show that the noise added for privacy is small relative to other sources of error (sampling and model misspecification). This suggests that, in principle, curators of such sensitive data can provide valuable epidemiologic insights while protecting privacy.

Message Passing Neural Networks have recently become the most popular approach to graph machine learning tasks; however, their receptive field is limited by the number of message passing layers. To increase the receptive field, Graph Transformers with global attention have been proposed; however, global attention does not take into account the graph topology and thus lacks graph-structure-based inductive biases, which are typically very important for graph machine learning tasks. In this work, we propose an alternative approach: cluster attention (CLATT). We divide graph nodes into clusters with off-the-shelf graph community detection algorithms and let each node attend to all other nodes in each cluster. CLATT provides large receptive fields while still having strong graph-structure-based inductive biases. We show that augmenting Message Passing Neural Networks or Graph Transformers with CLATT significantly improves their performance on a wide range of graph datasets including datasets from the recently introduced GraphLand benchmark representing real-world applications of graph machine learning.

Representation learning for geospatial and spatio-temporal data plays a critical role in enabling general-purpose geospatial intelligence. Recent geospatial foundation models, such as the Population Dynamics Foundation Model (PDFM), encode complex population and mobility dynamics into compact embeddings. However, their integration with Large Language Models (LLMs) remains limited. Existing approaches to LLM integration treat these embeddings as retrieval indices or convert them into textual descriptions for reasoning, introducing redundancy, token inefficiency, and numerical inaccuracies. We propose Direct Feature Reasoning-Gemma (DFR-Gemma), a novel framework that enables LLMs to reason directly over dense geospatial embeddings. DFR aligns high-dimensional embeddings with the latent space of an LLM via a lightweight projector, allowing embeddings to be injected as semantic tokens alongside natural language instructions. This design eliminates the need for intermediate textual representations and enables intrinsic reasoning over spatial features. To evaluate this paradigm, we introduce a multi-task geospatial benchmark that pairs embeddings with diverse question-answer tasks, including feature querying, comparison, and semantic description. Experimental results show that DFR allows LLMs to decode latent spatial patterns and perform accurate zero-shot reasoning across tasks, while significantly improving efficiency compared to text-based baselines. Our results demonstrate that treating embeddings as primary data inputs, provides a more direct, efficient, and scalable approach to multimodal geospatial intelligence.

Large language model agents rely on effective model context to obtain task-relevant information for decision-making. Many existing context engineering approaches primarily rely on the context generated from the past experience and retrieval mechanisms that reuse these context. However, retrieved context from past tasks must be adapted by the execution agent to fit new situations, placing additional reasoning burden on the underlying LLM. To address this limitation, we propose a generative context augmentation framework using Contrastive Learning of Experience via Agentic Reflection (CLEAR). CLEAR first employs a reflection agent to perform contrastive analysis over past execution trajectories and summarize useful context for each observed task. These summaries are then used as supervised fine-tuning data to train a context augmentation model (CAM). Then we further optimize CAM using reinforcement learning, where the reward signal is obtained by running the task execution agent. By learning to generate task-specific knowledge rather than retrieve knowledge from the past, CAM produces context that is better tailored to the current task. We conduct comprehensive evaluations on the AppWorld and WebShop benchmarks. Experimental results show that CLEAR consistently outperforms strong baselines. It improves task completion rate from 72.62% to 81.15% on AppWorld test set and averaged reward from 0.68 to 0.74 on a subset of WebShop, compared with baseline agent. Our code is publicly available at https://github.com/awslabs/CLEAR.

Large language models (LLMs) have emerged as a powerful tool for synthetic data generation. A particularly important use case is producing synthetic replicas of private text, which requires carefully balancing privacy and utility. We propose Realistic and Privacy-Preserving Synthetic Data Generation (RPSG), which leverages privacy-preserving mechanisms, including formal differential privacy (DP); and private seeds, in particular text containing personal information, to generate realistic synthetic data. Comprehensive experiments against state-of-the-art private synthetic data generation methods demonstrate that RPSG achieves high fidelity to private data while providing strong privacy protection.

Generative reward models (GRMs) have emerged as a promising approach for aligning Large Language Models (LLMs) with human preferences by offering greater representational capacity and flexibility than traditional scalar reward models. However, GRMs face two major challenges: reliance on costly human-annotated data restricts scalability, and self-training approaches often suffer from instability and vulnerability to reward hacking. To address these issues, we propose ConsistRM, a self-training framework that enables effective and stable GRM training without human annotations. ConsistRM incorporates the Consistency-Aware Answer Reward, which produces reliable pseudo-labels with temporal consistency, thereby providing more stable model optimization. Moreover, the Consistency-Aware Critique Reward is introduced to assess semantic consistency across multiple critiques and allocates fine-grained and differentiated rewards. Experiments on five benchmark datasets across four base models demonstrate that ConsistRM outperforms vanilla Reinforcement Fine-Tuning (RFT) by an average of 1.5%. Further analysis shows that ConsistRM enhances output consistency and mitigates position bias caused by input order, highlighting the effectiveness of consistency-aware rewards in improving GRMs.

Motivated by increasing wireless capacity demands and 6G advancements, the newly defined Frequency Range 3 (FR3, 7.125-24.25 GHz), also known as the upper mid-band, has emerged as a promising spectrum candidate. It offers a balance between the large bandwidth potential of millimeter-wave bands and the favorable propagation characteristics of sub-6 GHz bands. As a result, the upper mid-band presents a strong opportunity to enhance both coverage and capacity, particularly for 6G systems and Cellular Vehicle-to-Base Station (C-V2B) communications. Harnessing this potential, however, requires addressing key technical challenges through accurate and realistic channel modeling across diverse urban environments, including Suburban, Urban, and HighRise Urban scenarios. To this end, we employ a ray-tracing tool to characterize downlink propagation and enable detailed channel modeling for reliable C-V2B links. We evaluate data rate performance across FR1 (sub-6 GHz), FR3, and FR2 (mmWave) bands using antenna array configurations designed for different urban environments. The results show that, under equal aperture sizes, FR3 achieves higher data rates than FR2 for cell-edge User Equipment (UEs) in both interference-free and full-interference scenarios, indicating that the additional array gain at mmWave is insufficient to fully compensate for the severe experienced path loss. Integrating one-hand-grip pedestrian UEs model into ray tracer shows that transitioning from vehicular to pedestrian UEs results in negligible differences in coverage probability (about 1\%--3\%) across all frequencies, with the minimum differences observed in FR3, particularly at 8.2 GHz.

Reward machines are automaton-like structures that capture the memory required to accomplish a multi-stage task. When combined with reinforcement learning or optimal control methods, they can be used to synthesize robot policies to achieve such tasks. However, specifying a reward machine by hand, including a labeling function capturing high-level features that the decisions are based on, can be a daunting task. This paper deals with the problem of learning reward machines directly from raw state and policy information. As opposed to existing works, we assume no access to observations of rewards, labels, or machine nodes, and show what trajectory data is sufficient for learning the reward machine in this information-scarce regime. We then extend the result to an active learning setting where we incrementally query trajectory extensions to improve data (and indirectly computational) efficiency. Results are demonstrated with several grid world examples.

Online algorithm selection (OAS) aims to adapt the optimization process to changes in the fitness landscape and is expected to outperform any single algorithm from a given portfolio. Although this expectation is supported by numerous empirical studies, there are currently no theoretical results proving that OAS can yield asymptotic speedups (apart from some artificial examples for hyper-heuristics). Moreover, theory-based guidelines for when and how to switch between algorithms are largely missing. In this paper, we present the first theoretical example in which switching between two algorithms -- the $(1+λ)$ EA and the $(1+(λ,λ))$ GA -- solves the OneMax problem asymptotically faster than either algorithm used in isolation. We show that an appropriate choice of population sizes for the two algorithms allows the optimum to be reached in $O(n\log\log n)$ expected time, faster than the $Θ(n\sqrt{\frac{\log n \log\log\log n}{\log\log n}})$ runtime of the best of these two algorithms with optimally tuned parameters. We first establish this bound under an idealized switching rule that changes from the $(1+λ)$ to the $(1+(λ,λ))$ GA at the optimal time. We then propose a realistic switching strategy that achieves the same performance. Our analysis combines fixed-start and fixed-target perspectives, illustrating how different algorithms dominate at different stages of the optimization process. This approach offers a promising path toward a deeper theoretical understanding of OAS.

Deploying large language model (LLM) inference at scale requires jointly selecting base models, provisioning heterogeneous GPUs, configuring parallelism, and distributing workloads under tight latency, accuracy, and budget constraints. Exact mixed-integer linear programming (MILP) approaches guarantee optimality but scale poorly. We propose two constraint-aware heuristics: a Greedy Heuristic (GH) for single-pass allocation, and an Adaptive Greedy Heuristic (AGH) that enhances GH via multi-start construction, relocate-based local search, and GPU consolidation. Three constraint-aware mechanisms -- TP-aware feasibility selection, cost-per-effective-coverage ranking, and TP upgrade -- ensure feasibility under tightly coupled memory, delay, error, and budget constraints. On workloads calibrated with the Azure LLM Inference Trace (2025), both heuristics produce feasible solutions in under one second, with AGH closely approaching optimal cost while achieving over 260x speedup on large-scale instances. Under out-of-sample stress tests with up to 1.5x parameter inflation, AGH maintains controlled SLO violations and stable cost, whereas the exact solver's placement degrades sharply.

Large-scale cyberattacks, referred to as campaigns, are documented across multiple CTI reports from diverse sources, with some providing a high-level overview of attack techniques and others providing technical details. Extracting attack techniques from reports is essential for organizations to identify the controls required to protect against attacks. Manually extracting techniques at scale is impractical. Existing automated methods focus on single reports, leaving many attack techniques and their controls undetected, resulting in a fragmented view of campaign behavior. The goal of this study is to aid security researchers in extracting attack techniques and controls from a campaign by replicating and comparing the performance of the state-of-the-art ATT&CK technique extraction methods in a multi-report campaign setting compared to prior single-report evaluations. We conduct an empirical study of 29 methods to extract attack techniques, spanning named entity recognition (NER), encoder-based classification, and decoder-based LLM approaches. Our study analyzes 90 CTI reports across three major attack campaigns: SolarWinds, XZ Utils, and Log4j, using both quantitative performance metrics and their impact on controls. Our results show that aggregating multiple CTI reports improves the F1 score by about 26% over single-report analysis, with most approaches reaching performance saturation after 5--15 reports. Despite these gains, extraction performance remains limited, with maximum F1 scores of 78.6% for SolarWinds and 54.9% for XZ Utils. Moreover, up to 33.3% of misclassifications involve semantically similar techniques that share tactics and overlap in descriptions. The misclassification has a disproportionate effect on control coverage. Reports that are longer and include technical details consistently perform better, even though their readability scores are low.

Implicit artistic influence, although visually plausible, is often undocumented and thus poses a historically constrained attribution problem: resemblance is necessary but not sufficient evidence. Most prior systems reduce influence discovery to embedding similarity or label-driven graph completion, while recent multimodal large language models (LLMs) remain vulnerable to temporal inconsistency and unverified attributions. This paper introduces M-ArtAgent, an evidence-based multimodal agent that reframes implicit influence discovery as probabilistic adjudication. It follows a four-phase protocol consisting of Investigation, Corroboration, Falsification, and Verdict governed by a Reasoning and Acting (ReAct)-style controller that assembles verifiable evidence chains from images and biographies, enforces art-historical axioms, and subjects each hypothesis to adversarial falsification via a prompt-isolated critic. Two theory-grounded operators, StyleComparator for Wolfflin formal analysis and ConceptRetriever for ICONCLASS-based iconographic grounding, ensure that intermediate claims are formally auditable. On the balanced WikiArt Influence Benchmark-100 (WIB-100) of 100 artists and 2,000 directed pairs, M-ArtAgent achieves 83.7% positive-class F1, 0.666 Matthews correlation coefficient (MCC), and 0.910 area under the receiver operating characteristic curve (ROC-AUC), with leakage-control and robustness checks confirming that the gains persist when explicit influence phrases are masked. By coupling multimodal perception with domain-constrained falsification, M-ArtAgent demonstrates that implicit influence analysis benefits from historically grounded adjudication rather than pattern matching alone.

Discrete speech units (DSUs) are derived by quantising representations from models trained using self-supervised learning (SSL). They are a popular representation for a wide variety of spoken language tasks, including those where prosody matters. DSUs are especially convenient for tasks where text and speech are jointly modelled, such as text-to-speech and multimodal dialogue systems. But we have found that DSUs encode suprasegmental information less reliably than segmental structure, which we demonstrate in this work using lexical tone, though this limitation likely extends to other suprasegmental features such as prosody. Our investigations using the tone languages Mandarin and Yorùbá show that the SSL latent representations themselves do encode tone, yet DSUs obtained using quantisation tend to prioritise phonetic structure, which makes lexical tone less reliably encoded. This remains true for a variety of quantisation methods, not only the most common, K-means. We conclude that current DSU quantisation strategies have limitations for suprasegmental features, which suggests a need for new, tone-aware (or prosody-aware) techniques in speech representation learning. We point towards a potential form of the solution by performing K-means clustering once to encode phonetic information, then again on the residual representation, which better encodes lexical tone.

Cross-tokenizer distillation (CTD), the transfer of knowledge from a teacher to a student language model when the two use different tokenizers, remains a largely unsolved problem. Existing approaches rely on heuristic strategies to align mismatched vocabularies, introducing considerable complexity. In this paper, we propose a simple but effective baseline called Byte-Level Distillation (BLD) which enables CTD by operating at a common interface across tokenizers: the byte level. In more detail, we convert the teacher's output distribution to byte-level probabilities, attach a lightweight byte-level decoder head to the student, and distill through this shared byte-level interface. Despite its simplicity, BLD performs competitively with--and on several benchmarks surpasses--significantly more sophisticated CTD methods, across a range of distillation tasks with models from 1B to 8B parameters. Our results suggest that the byte level is a natural common ground for cross-tokenizer knowledge transfer, while also highlighting that consistent improvements across all tasks and benchmarks remain elusive, underscoring that CTD is still an open problem.

While decoupled control schemes for legged mobile manipulators have shown robustness, learning holistic whole-body control policies for tracking global end-effector poses remains fragile against Out-of-Distribution (OOD) inputs induced by sensor noise or infeasible user commands. To improve robustness against these perturbations without sacrificing task performance and continuity, we propose Competence Manifold Projection (CMP). Specifically, we utilize a Frame-Wise Safety Scheme that transforms the infinite-horizon safety constraint into a computationally efficient single-step manifold inclusion. To instantiate this competence manifold, we employ a Lower-Bounded Safety Estimator that distinguishes unmastered intentions from the training distribution. We then introduce an Isomorphic Latent Space (ILS) that aligns manifold geometry with safety probability, enabling efficient O(1) seamless defense against arbitrary OOD intents. Experiments demonstrate that CMP achieves up to a 10-fold survival rate improvement in typical OOD scenarios where baselines suffer catastrophic failure, incurring under 10% tracking degradation. Notably, the system exhibits emergent ``best-effort'' generalization behaviors to progressively accomplish OOD goals by adhering to the competence boundaries. Result videos are available at: https://shepherd1226.github.io/CMP.

We describe an experiment in LLM-assisted autoformalization that produced over 85,000 lines of Isabelle/HOL code covering all 39 sections of Munkres' Topology (general topology, Chapters 2--8), from topological spaces through dimension theory. The LLM-based coding agents (initially ChatGPT 5.2 and then Claude Opus 4.6) used 24 active days for that. The formalization is complete: all 806 formal results are fully proved with zero sorry's. Proved results include the Tychonoff theorem, the Baire category theorem, the Nagata--Smirnov and Smirnov metrization theorems, the Stone--Čech compactification, Ascoli's theorem, the space-filling curve, and others. The methodology is based on a "sorry-first" declarative proof workflow combined with bulk use of sledgehammer - two of Isabelle major strengths. This leads to relatively fast autoformalization progress. We analyze the resulting formalization in detail, analyze the human--LLM interaction patterns from the session log, and briefly compare with related autoformalization efforts in Megalodon, HOL Light, and Naproche. The results indicate that LLM-assisted formalization of standard mathematical textbooks in Isabelle/HOL is quite feasible, cheap and fast, even if some human supervision is useful.

Large Chunk Test-Time Training (LaCT) has shown strong performance on long-context 3D reconstruction, but its fully plastic inference-time updates remain vulnerable to catastrophic forgetting and overfitting. As a result, LaCT is typically instantiated with a single large chunk spanning the full input sequence, falling short of the broader goal of handling arbitrarily long sequences in a single pass. We propose Elastic Test-Time Training inspired by elastic weight consolidation, that stabilizes LaCT fast-weight updates with a Fisher-weighted elastic prior around a maintained anchor state. The anchor evolves as an exponential moving average of past fast weights to balance stability and plasticity. Based on this updated architecture, we introduce Fast Spatial Memory (FSM), an efficient and scalable model for 4D reconstruction that learns spatiotemporal representations from long observation sequences and renders novel view-time combinations. We pre-trained FSM on large-scale curated 3D/4D data to capture the dynamics and semantics of complex spatial environments. Extensive experiments show that FSM supports fast adaptation over long sequences and delivers high-quality 3D/4D reconstruction with smaller chunks and mitigating the camera-interpolation shortcut. Overall, we hope to advance LaCT beyond the bounded single-chunk setting toward robust multi-chunk adaptation, a necessary step for generalization to genuinely longer sequences, while substantially alleviating the activation-memory bottleneck.

Exact relevance certification asks which coordinates are necessary to determine the optimal action in a coordinate-structured decision problem. The tractable families treated here admit a finite primitive basis, but optimizer-quotient realizability is maximal, so quotient shape alone cannot characterize the frontier. We prove a meta-impossibility theorem for efficiently checkable structural predicates invariant under the theorem-forced closure laws of exact certification. Structural convergence with zero-distortion summaries, quotient entropy bounds, and support-counting arguments explains why those closure laws are canonical. We establish the theorem by constructing same-orbit disagreements for four obstruction families, namely dominant-pair concentration, margin masking, ghost-action concentration, and additive/statewise offset concentration, using action-independent, pair-targeted affine witnesses. Consequently no correct tractability classifier on a closure-closed domain yields an exact characterization over these families. Here closure-orbit agreement is forced by correctness rather than assumed as an invariance axiom. The result therefore applies to correct classifiers on closure-closed domains, not only to classifiers presented through a designated admissibility package.

Generating motion-controlled videos--where user-specified actions drive physically plausible scene dynamics under freely chosen viewpoints--demands two capabilities: (1) disentangled motion control, allowing users to separately control the object motion and adjust camera viewpoint; and (2) motion causality, ensuring that user-driven actions trigger coherent reactions from other objects rather than merely displacing pixels. Existing methods fall short on both fronts: they entangle camera and object motion into a single tracking signal and treat motion as kinematic displacement without modeling causal relationships between object motion. We introduce MoRight, a unified framework that addresses both limitations through disentangled motion modeling. Object motion is specified in a canonical static-view and transferred to an arbitrary target camera viewpoint via temporal cross-view attention, enabling disentangled camera and object control. We further decompose motion into active (user-driven) and passive (consequence) components, training the model to learn motion causality from data. At inference, users can either supply active motion and MoRight predicts consequences (forward reasoning), or specify desired passive outcomes and MoRight recovers plausible driving actions (inverse reasoning), all while freely adjusting the camera viewpoint. Experiments on three benchmarks demonstrate state-of-the-art performance in generation quality, motion controllability, and interaction awareness.

The rapid growth of generative artificial intelligence (AI) has introduced unprecedented computational demands, driving significant increases in the energy footprint of data centers. However, existing power consumption data is largely proprietary and reported at varying resolutions, creating challenges for estimating whole-facility energy use and planning infrastructure. In this work, we present a methodology that bridges this gap by linking high-resolution workload power measurements to whole-facility energy demand. Using NLR's high-performance computing data center equipped with NVIDIA H100 GPUs, we measure power consumption of AI workloads at 0.1-second resolution for AI training, fine-tuning and inference jobs. Workloads are characterized using MLCommons benchmarks for model training and fine-tuning, and vLLM benchmarks for inference, enabling reproducible and standardized workload profiling. The dataset of power consumption profiles is made publicly available. These power profiles are then scaled to the whole-facility-level using a bottom-up, event-driven, data center energy model. The resulting whole-facility energy profiles capture realistic temporal fluctuations driven by AI workloads and user-behavior, and can be used to inform infrastructure planning for grid connection, on-site energy generation, and distributed microgrids.

Pluralistic alignment has emerged as a critical frontier in the development of Large Language Models (LLMs), with reward models (RMs) serving as a central mechanism for capturing diverse human values. While benchmarks for general response quality are prevalent, evaluating how well reward models account for individual user preferences remains an open challenge. To bridge this gap, we introduce Personalized RewardBench, a novel benchmark designed to rigorously assess reward models' capacity to model personalized preferences. We construct chosen and rejected response pairs based on strict adherence to (or violation of) user-specific rubrics, ensuring that preference distinctions are uniquely tailored to the individual. In particular, human evaluations confirm that the primary discriminative factor between pairs is strictly personal preference, with both responses maintaining high general quality (e.g., correctness, relevance and helpfulness). Extensive testing reveals that existing state-of-the-art reward models struggle significantly with personalization, peaking at an accuracy of just 75.94%. Crucially, because an effective reward model benchmark should predict a reward model's performance on downstream tasks, we conduct experiments demonstrating that our benchmark exhibits a significantly higher correlation with downstream performance in both Best-of-N (BoN) sampling and Proximal Policy Optimization (PPO) compared to existing baselines. These findings establish Personalized RewardBench as a robust and accurate proxy for evaluating reward models' performance in downstream applications.

Most repository-level code translation and validation techniques have been evaluated on a single source-target programming language (PL) pair, owing to the complex engineering effort required to adapt new PL pairs. Programming agents can enable PL-agnosticism in repository-level code translation and validation: they can synthesize code across many PLs and autonomously use existing tools specific to each PL's analysis. However, state-of-the-art has yet to offer a fully autonomous agentic approach for repository-level code translation and validation of large-scale programs. This paper proposes ReCodeAgent, an autonomous multi-agent approach for language-agnostic repository-level code translation and validation. Users only need to provide the project in the source PL and specify the target PL for ReCodeAgent to automatically translate and validate the entire repository. ReCodeAgent is the first technique to achieve high translation success rates across many PLs. We compare the effectiveness of ReCodeAgent with four alternative neuro-symbolic and agentic approaches to translate 118 real-world projects, with 1,975 LoC and 43 translation units for each project, on average. The projects cover 6 PLs (C, Go, Java, JavaScript, Python, and Rust) and 4 PL pairs (C-Rust, Go-Rust, Java-Python, Python-JavaScript). Our results demonstrate that ReCodeAgent consistently outperforms prior techniques on translation correctness, improving test pass rate by 60.8% on ground-truth tests, with an average cost of $15.3. We also perform process-centric analysis of ReCodeAgent trajectories to confirm its procedural efficiency. Finally, we investigate how the design choices (a multi-agent vs. single-agent architecture) influence ReCodeAgent performance: on average, the test pass rate drops by 40.4%, and trajectories become 28% longer and persistently inefficient.

Recent advances in vision-language models (VLMs) have improved image captioning for cultural heritage. However, inferring structured cultural metadata (e.g., creator, origin, period) from visual input remains underexplored. We introduce a multi-category, cross-cultural benchmark for this task and evaluate VLMs using an LLM-as-Judge framework that measures semantic alignment with reference annotations. To assess cultural reasoning, we report exact-match, partial-match, and attribute-level accuracy across cultural regions. Results show that models capture fragmented signals and exhibit substantial performance variation across cultures and metadata types, leading to inconsistent and weakly grounded predictions. These findings highlight the limitations of current VLMs in structured cultural metadata inference beyond visual perception.

Towards an embodied generalist for real-world interaction, Multimodal Large Language Model (MLLM) agents still suffer from challenging latency, sparse feedback, and irreversible mistakes. Video games offer an ideal testbed with rich visual observations and closed-loop interaction, demanding fine-grained perception, long-horizon planning, and precise control. However, systematically evaluating these capabilities is currently hindered by heterogeneous action interfaces and heuristic verification. To this end, we introduce GameWorld, a benchmark designed for standardized and verifiable evaluation of MLLMs as generalist game agents in browser environments. Two game agent interfaces are studied: (i) computer-use agents that directly emit keyboard and mouse controls, and (ii) generalist multimodal agents that act in a semantic action space via deterministic Semantic Action Parsing. GameWorld contains 34 diverse games and 170 tasks, each paired with state-verifiable metrics for outcome-based evaluation. The results across 18 model-interface pairs suggest that even the best performing agent is far from achieving human capabilities on video games. Extensive experiments of repeated full-benchmark reruns demonstrate the robustness of the benchmark, while further studies on real-time interaction, context-memory sensitivity, and action validity expose more challenges ahead for game agents. Together, by offering a standardized, verifiable, and reproducible evaluation framework, GameWorld lays a robust foundation for advancing research on multimodal game agents and beyond. The project page is at https://gameworld-bench.github.io.

Scaling up robot learning will likely require human data containing rich and long-horizon interactions in the wild. Existing approaches for collecting such data trade off portability, robustness to occlusion, and global consistency. We introduce RoSHI, a hybrid wearable that fuses low-cost sparse IMUs with the Project Aria glasses to estimate the full 3D pose and body shape of the wearer in a metric global coordinate frame from egocentric perception. This system is motivated by the complementarity of the two sensors: IMUs provide robustness to occlusions and high-speed motions, while egocentric SLAM anchors long-horizon motion and stabilizes upper body pose. We collect a dataset of agile activities to evaluate RoSHI. On this dataset, we generally outperform other egocentric baselines and perform comparably to a state-of-the-art exocentric baseline (SAM3D). Finally, we demonstrate that the motion data recorded from our system are suitable for real-world humanoid policy learning. For videos, data and more, visit the project webpage: https://roshi-mocap.github.io/

How does the choice of training data influence an AI model? This question is of central importance to interpretability, privacy, and basic science. At its core is the data deletion problem: after a reasonable amount of precomputation, quickly predict how the model would behave in a given situation if a given subset of training data had been excluded from the learning algorithm. We present a data deletion scheme capable of predicting model outputs with vanishing error $\varepsilon$ in the deep learning setting. Our precomputation and prediction algorithms are only $\mathrm{poly}(1/\varepsilon)$ factors slower than regular training and inference, respectively. The storage requirements are those of $\mathrm{poly}(1/\varepsilon)$ models. Our proof is based on an assumption that we call "stability." In contrast to the assumptions made by prior work, stability appears to be fully compatible with learning powerful AI models. In support of this, we show that stability is satisfied in a minimal set of experiments with microgpt. Our code is available at https://github.com/SamSpo1/microgpt-sketch. At a technical level, our work is based on a new method for locally sketching an arithmetic circuit by computing higher-order derivatives in random complex directions. Forward-mode automatic differentiation allows cheap computation of these derivatives.

Safety in reinforcement learning (RL) is typically enforced through objective shaping while keeping environment dynamics stationary with respect to observable state-action pairs. Under delayed harm, this can lead to replay: after a washout period, reintroducing the same stimulus under matched observable conditions reproduces a similar harmful cascade. We introduce the Replay Suppression Diagnostic (RSD), a controlled exposure-decay-replay protocol that isolates this failure mode under frozen-policy evaluation. We show that, under stationary observable transition kernels, replay cannot be structurally suppressed without inducing a persistent shift in replay-time action distributions. Motivated by platform-mediated systems, we propose Regret-Aware Policy Optimization (RAPO), which augments the environment with persistent harm-trace and scar fields and applies a bounded, mass-preserving transition reweighting to reduce reachability of historically harmful regions. On graph diffusion tasks (50-1000 nodes), RAPO suppresses replay, reducing re-amplification gain (RAG) from 0.98 to 0.33 on 250-node graphs while retaining 82\% of task return. Disabling transition deformation only during replay restores re-amplification (RAG 0.91), isolating environment-level deformation as the causal mechanism.

In this paper, we derive rates of convergence in the high-dimensional central limit theorem for Polyak-Ruppert averaged iterates generated by the asynchronous Q-learning algorithm with a polynomial stepsize $k^{-ω},\, ω\in (1/2, 1]$. Assuming that the sequence of state-action-next-state triples $(s_k, a_k, s_{k+1})_{k \geq 0}$ forms a uniformly geometrically ergodic Markov chain, we establish a rate of order up to $n^{-1/6} \log^{4} (nS A)$ over the class of hyper-rectangles, where $n$ is the number of samples used by the algorithm and $S$ and $A$ denote the numbers of states and actions, respectively. To obtain this result, we prove a high-dimensional central limit theorem for sums of martingale differences, which may be of independent interest. Finally, we present bounds for high-order moments for the algorithm's last iterate.

Propositional Linear Temporal Logic (LTL) is a popular formalism for specifying desirable requirements and security and privacy policies for software, networks, and systems. Yet expressing such requirements and policies in LTL remains challenging because of its intricate semantics. Since many security and privacy analysis tools require LTL formulas as input, this difficulty places them out of reach for many developers and analysts. Large Language Models (LLMs) could broaden access to such tools by translating natural language fragments into LTL formulas. This paper evaluates that premise by assessing how effectively several representative LLMs translate assertive English sentences into LTL formulas. Using both human-generated and synthetic ground-truth data, we evaluate effectiveness along syntactic and semantic dimensions. The results reveal three findings: (1) in line with prior findings, LLMs perform better on syntactic aspects of LTL than on semantic ones; (2) they generally benefit from more detailed prompts; and (3) reformulating the task as a Python code-completion problem substantially improves overall performance. We also discuss challenges in conducting a fair evaluation on this task and conclude with recommendations for future work.

Low-resource languages pose a challenge for machine translation with large language models (LLMs), which require large amounts of training data. One potential way to circumvent this data dependence is to rely on LLMs' ability to use in-context descriptions of languages, like textbooks and dictionaries. To do so, LLMs must be able to infer the link between the languages' grammatical descriptions and the sentences in question. Here we isolate this skill using a formal analogue of the task: string transduction based on a formal grammar provided in-context. We construct synchronous context-free grammars which define pairs of formal languages designed to model particular aspects of natural language grammar, morphology, and written representation. Using these grammars, we measure how well LLMs can translate sentences from one formal language into another when given both the grammar and the source-language sentence. We vary the size of the grammar, the lengths of the sentences, the syntactic and morphological properties of the languages, and their written script. We note three key findings. First, LLMs' translation accuracy decreases markedly as a function of grammar size and sentence length. Second, differences in morphology and written representation between the source and target languages can strongly diminish model performance. Third, we examine the types of errors committed by models and find they are most prone to recall the wrong words from the target language vocabulary, hallucinate new words, or leave source-language words untranslated.

The growing complexity of neural networks hinders the deployment of distributed machine learning on resource-constrained devices. Split learning (SL) offers a promising solution by partitioning the large model and offloading the primary training workload from edge devices to an edge server. However, the increasing number of participating devices and model complexity leads to significant communication overhead from the transmission of smashed data (e.g., activations and gradients), which constitutes a critical bottleneck for SL. To tackle this challenge, we propose SL-FAC, a communication-efficient SL framework comprising two key components: adaptive frequency decomposition (AFD) and frequency-based quantization compression (FQC). AFD first transforms the smashed data into the frequency domain and decomposes it into spectral components with distinct information. FQC then applies customized quantization bit widths to each component based on its spectral energy distribution. This collaborative approach enables SL-FAC to achieve significant communication reduction while strategically preserving the information most crucial for model convergence. Extensive experiments confirm the superior performance of SL-FAC for improving the training efficiency.

Online marketplaces increasingly do more than simply match buyers and sellers: they route orders across competing sellers and, in many categories, offer ancillary fulfillment services that make seller inventory a source of platform revenue. We investigate how a platform can use intertemporal demand allocation to influence sellers' inventory choices without directly controlling stock. We develop a model in which the platform observes aggregate demand, allocates orders across sellers over time, and sellers choose between two fulfillment options, fulfill-by-merchant (FBM) and fulfill-by-platform (FBP), while replenishing inventory under state-dependent base-stock policies. The key mechanism we study is informational: by changing the predictability of each seller's sales stream, the platform changes sellers' safety-stock needs even when average demand shares remain unchanged. We focus on nondiscriminatory allocation policies that give sellers the same demand share and forecast risk. Within this class, uniform splitting minimizes forecast uncertainty, whereas any higher level of uncertainty can be implemented using simple low-memory allocation rules. Moreover, increasing uncertainty above the uniform benchmark requires routing rules that prevent sellers from inferring aggregate demand from their own sales histories. These results reduce the platform's problem to choosing a level of forecast uncertainty that trades off adoption of platform fulfillment against the inventory held by adopters. Our analysis identifies demand allocation as a powerful operational and informational design lever in digital marketplaces.

Existing dynamic data pruning methods often fail under noisy-label settings, as they typically rely on per-sample loss as the ranking criterion. This could mistakenly lead to preserving noisy samples due to their high loss values, resulting in significant performance drop. To address this, we propose AlignPrune, a noise-robust module designed to enhance the reliability of dynamic pruning under label noise. Specifically, AlignPrune introduces the Dynamic Alignment Score (DAS), which is a loss-trajectory-based criterion that enables more accurate identification of noisy samples, thereby improving pruning effectiveness. As a simple yet effective plug-and-play module, AlignPrune can be seamlessly integrated into state-of-the-art dynamic pruning frameworks, consistently outperforming them without modifying either the model architecture or the training pipeline. Extensive experiments on five widely-used benchmarks across various noise types and pruning ratios demonstrate the effectiveness of AlignPrune, boosting accuracy by up to 6.3\% over state-of-the-art baselines. Our results offer a generalizable solution for pruning under noisy data, encouraging further exploration of learning in real-world scenarios. Code is available at: https://github.com/leonqin430/AlignPrune.

Model-based reinforcement learning (MBRL) improves sample efficiency by optimizing policies inside imagined rollouts, but long-horizon planning degrades when model errors compound and imagined trajectories drift off the training manifold. We introduce GIRL (Generative Imagination Reinforcement Learning), a latent world-model framework that addresses this failure mode with two key components. First, a cross-modal grounding signal derived from a frozen foundation model (DINOv2) anchors the latent transition prior to a semantically consistent embedding space, penalizing inconsistent or implausible predictions. Second, an uncertainty-adaptive trust-region bottleneck interprets the KL regularizer as the Lagrange multiplier of a constrained optimization problem, restricting imagination drift within a learned region calibrated by Expected Information Gain and a Relative Performance Loss signal. We re-derive a value-gap bound using the Performance Difference Lemma and Integral Probability Metrics, yielding a bound that remains informative as the discount factor approaches one and connects the objective to real-environment regret. Experiments across three benchmark suites, including DeepMind Control, Adroit Hand Manipulation, and Meta-World with visual distractors, show that GIRL reduces latent rollout drift by 38 to 61 percent across tasks relative to DreamerV3, improves asymptotic return, and requires fewer environment interactions on long-horizon tasks. GIRL also outperforms TD-MPC2 on sparse-reward and high-contact settings under standard evaluation metrics. A distilled-prior variant reduces inference overhead and improves computational efficiency relative to the full model.

Large Language Models (LLMs) challenge conventional automated programming assessment because students can now produce functionally correct code without demonstrating corresponding understanding. This paper makes two contributions. First, it reports a saturation-based scoping review of conversational assessment approaches in programming education. The review identifies three dominant architectural families: rule-based or template-driven systems, LLM-based systems, and hybrid systems. Across the literature, conversational agents appear promising for scalable feedback and deeper probing of code understanding, but important limitations remain around hallucinations, over-reliance, privacy, integrity, and deployment constraints. Second, the paper synthesizes these findings into a Hybrid Socratic Framework for integrating conversational verification into Automated Programming Assessment Systems (APASs). The framework combines deterministic code analysis with a dual-agent conversational layer, knowledge tracking, scaffolded questioning, and guardrails that tie prompts to runtime facts. The paper also discusses practical safeguards against LLM-generated explanations, including proctored deployment modes, randomized trace questions, stepwise reasoning tied to concrete execution states, and local-model deployment options for privacy-sensitive settings. Rather than replacing conventional testing, the framework is intended as a complementary layer for verifying whether students understand the code they submit.

Understanding the landscape of artificial intelligence (AI) and machine learning (ML) adoption across the National Institutes of Health (NIH) portfolio is critical for research funding strategy, institutional planning, and health policy. The advent of large language models (LLMs) has fundamentally transformed research landscape analysis, enabling researchers to perform large-scale semantic extraction from thousands of unstructured research documents. In this paper, we illustrate a human-in-the-loop research methodology for LLMs to automatically classify and summarize research descriptions at scale. Using our methodology, we present a comprehensive analysis of 58,746 NIH-funded biomedical research projects from 2025. We show that: (1) AI constitutes 15.9% of the NIH portfolio with a 13.4% funding premium, concentrated in discovery, prediction, and data integration across disease domains; (2) a critical research-to-deployment gap exists, with 79% of AI projects remaining in research/development stages while only 14.7% engage in clinical deployment or implementation; and (3) health disparities research is severely underrepresented at just 5.7% of AI-funded work despite its importance to NIH's equity mission. These findings establish a framework for evidence-based policy interventions to align the NIH AI portfolio with health equity goals and strategic research priorities.

Multiple Instance Learning (MIL) is the dominant framework for gigapixel whole-slide image (WSI) classification in computational pathology. However, current MIL aggregators route all instances through a shared pathway, constraining their capacity to specialise across the pathological heterogeneity inherent in each slide. Mixture-of-Experts (MoE) methods offer a natural remedy by partitioning instances across specialised expert subnetworks; yet unconstrained softmax routing may yield highly imbalanced utilisation, where one or a few experts absorb most routing mass, collapsing the mixture back to a near-single-pathway solution. To address these limitations, we propose ROAM (Region-graph OptimAl-transport Mixture-of-experts), a spatially aware MoE-MIL aggregator that routes region tokens to expert poolers via capacity-constrained entropic optimal transport, promoting balanced expert utilisation by construction. ROAM operates on spatial region tokens, obtained by compressing dense patch bags into spatially binned units that align routing with local tissue neighbourhoods and introduces two key mechanisms: (i) region-to-expert assignment formulated as entropic optimal transport (Sinkhorn) with explicit per slide capacity marginals, enforcing balanced expert utilisation without auxiliary load-balancing losses; and (ii) graph-regularised Sinkhorn iterations that diffuse routing assignments over the spatial region graph, encouraging neighbouring regions to coherently route to the same experts. Evaluated on four WSI benchmarks with frozen foundation-model patch embeddings, ROAM achieves performance competitive against strong MIL and MoE baselines, and on NSCLC generalisation (TCGA-CPTAC) reaches external AUC 0.845 +- 0.019.

Spatial understanding is a fundamental cornerstone of human-level intelligence. Nonetheless, current research predominantly focuses on domain-specific data production, leaving a critical void: the absence of a principled, open-source engine capable of fully unleashing the potential of high-quality spatial data. To bridge this gap, we elucidate the design principles of a robust data generation system and introduce OpenSpatial -- an open-source data engine engineered for high quality, extensive scalability, broad task diversity, and optimized efficiency. OpenSpatial adopts 3D bounding boxes as the fundamental primitive to construct a comprehensive data hierarchy across five foundational tasks: Spatial Measurement (SM), Spatial Relationship (SR), Camera Perception (CP), Multi-view Consistency (MC), and Scene-Aware Reasoning (SAR). Leveraging this scalable infrastructure, we curate OpenSpatial-3M, a large-scale dataset comprising 3 million high-fidelity samples. Extensive evaluations demonstrate that versatile models trained on our dataset achieve state-of-the-art performance across a wide spectrum of spatial reasoning benchmarks. Notably, the best-performing model exhibits a substantial average improvement of 19 percent, relatively. Furthermore, we provide a systematic analysis of how data attributes influence spatial perception. By open-sourcing both the engine and the 3M-scale dataset, we provide a robust foundation to accelerate future research in spatial intelligence.

Real-time supervisory control of advanced reactors requires accurate forecasting of plant-wide thermal-hydraulic states, including locations where physical sensors are unavailable. Meeting this need calls for surrogate models that combine predictive fidelity, millisecond-scale inference, and robustness to partial observability. In this work, we present a physics-informed message-passing Graph Neural Network coupled with a Neural Ordinary Differential Equation (GNN-ODE) to addresses all three requirements simultaneously. We represent the whole system as a directed sensor graph whose edges encode hydraulic connectivity through flow/heat transfer-aware message passing, and we advance the latent dynamics in continuous time via a controlled Neural ODE. A topology-guided missing-node initializer reconstructs uninstrumented states at rollout start; prediction then proceeds fully autoregressively. The GNN-ODE surrogate achieves satisfactory results for the system dynamics prediction. On held-out simulation transients, the surrogate achieves an average MAE of 0.91 K at 60 s and 2.18 K at 300 s for uninstrumented nodes, with $R^2$ up to 0.995 for missing-node state reconstruction. Inference runs at approximately 105 times faster than simulated time on a single GPU, enabling 64-member ensemble rollouts for uncertainty quantification. To assess sim-to-real transfer, we adapt the pretrained surrogate to experimental facility data using layerwise discriminative fine-tuning with only 30 training sequences. The learned flow-dependent heat-transfer scaling recovers a Reynolds-number exponent consistent with established correlations, indicating constitutive learning beyond trajectory fitting. The model tracks a steep power change transient and produces accurate trajectories at uninstrumented locations.

This work presents a symbolic approach for estimating the energy consumption for nested loop programs when mapped and scheduled on parallel processor array accelerator architectures. Instead of simulation-based evaluation, we derive a methodology for symbolic energy analysis that captures the impact of mapping and scheduling decisions of loop nests on processor arrays. We compare our approach against simulation-based results for selected benchmarks and varying sizes of the iteration spaces. Whereas the latter are not scalable, our symbolic analysis is shown to be independent of the problem size. The presented evaluation methodology can be beneficially used during the design space exploration of mapping and scheduling decisions, for studying the influence of array size variations, and for comparisons with other loop nest accelerator architectures.

Autonomous vehicles deployed in remote environments typically rely on embedded processors, compact batteries, and lightweight sensors. These hardware limitations conflict with the need to derive robust representations of the environment, which often requires executing computationally intensive deep neural networks for perception. To address this challenge, we present CADENCE, an adaptive system that dynamically scales the computational complexity of a slimmable monocular depth estimation network in response to navigation needs and environmental context. By closing the loop between perception fidelity and actuation requirements, CADENCE ensures high-precision computing is only used when mission-critical. We conduct evaluations on our released open-source testbed that integrates Microsoft AirSim with an NVIDIA Jetson Orin Nano. As compared to a state-of-the-art static approach, CADENCE decreases sensor acquisitions, power consumption, and inference latency by 9.67%, 16.1%, and 74.8%, respectively. The results demonstrate an overall reduction in energy expenditure by 75.0%, along with an increase in navigation accuracy by 7.43%.

Debates about artificial intelligence (AI) in education often portray teaching as a modular and procedural job that can increasingly be automated or delegated to technology. This brief communication paper argues that such claims depend on treating teaching as more separable than it is in practice. Drawing on recent literature and empirical studies of large language models and retrieval-augmented generation systems, I argue that although AI can support some bounded functions, instructional work remains difficult to automate in meaningful ways because it is inherently interpretive, relational, and grounded in professional judgment. More fundamentally, teaching and learning are shaped by human cognition, behavior, motivation, and social interaction in ways that cannot be fully specified, predicted, or exhaustively modeled. Tasks that may appear separable in principle derive their instructional value in practice from ongoing contextual interpretation across learners, situations, and relationships. As long as educational practice relies on emergent understanding of human cognition and learning, teaching remains a form of professional work that resists automation. AI may improve access to information and support selected instructional activities, but it does not remove the need for human judgment and relational accountability that effective teaching requires.

Automated face recognition has made rapid strides over the past decade due to the unprecedented rise of deep neural network (DNN) models that can be trained for domain-specific tasks. At the same time, foundation models that are pretrained on broad vision or vision-language tasks have shown impressive generalization across diverse domains, including biometrics. This raises an important question: Do different DNN models--both domain-specific and foundation models--encode facial identity in similar ways, despite being trained on different datasets, loss functions, and architectures? In this regard, we directly analyze the geometric structure of embedding spaces imputed by different DNN models. Treating embeddings of face images as point clouds, we study whether simple affine transformations can align face representations of one model with another. Our findings reveal surprising cross-model compatibility: low-capacity linear mappings substantially improve cross-model face recognition over unaligned baselines for both face identification and verification tasks. Alignment patterns generalize across datasets and vary systematically across model families, indicating representational convergence in facial identity encoding. These findings have implications for model interoperability, ensemble design, and biometric template security.

Online reinforcement learning (RL) serves as an effective method for enhancing the capabilities of Android agents. However, guiding agents to learn through online interaction is prohibitively expensive due to the high latency of emulators and the sample inefficiency of existing RL algorithms. We identify a fundamental limitation in current approaches: the Single State Single Action paradigm, which updates the policy with one-to-one state-action pairs from online one-way rollouts without fully exploring each costly emulator state. In this paper, we propose Android Coach, a novel framework that shifts the training paradigm to Single State Multiple Actions, allowing the agent to sample and utilize multiple actions for a single online state. We enable this without additional emulator overhead by learning a critic that estimates action values. To ensure the critic serves as a reliable coach, we integrate a process reward model and introduce a group-wise advantage estimator based on the averaged critic outputs. Extensive experiments demonstrate the effectiveness and efficiency of Android Coach: it achieves 7.5% and 8.3% success rate improvements on AndroidLab and AndroidWorld over UI-TARS-1.5-7B, and attains 1.4x higher training efficiency than Single State Single Action methods PPO and GRPO at matched success rates.

GROMACS is a de-facto standard for classical Molecular Dynamics (MD). The rise of AI-driven interatomic potentials that pursue near-quantum accuracy at MD throughput now poses a significant challenge: embedding neural-network inference into multi-GPU simulations retaining high-performance. In this work, we integrate the MLIP framework DeePMD-kit into GROMACS, enabling domain-decomposed, GPU-accelerated inference across multi-node systems. We extend the GROMACS NNPot interface with a DeePMD backend, and we introduce a domain decomposition layer decoupled from the main simulation. The inference is executed concurrently on all processes, with two MPI collectives used each step to broadcast coordinates and to aggregate and redistribute forces. We train an in-house DPA-1 model (1.6 M parameters) on a dataset of solvated protein fragments. We validate the implementation on a small protein system, then we benchmark the GROMACS-DeePMD integration with a 15,668 atom protein on NVIDIA A100 and AMD MI250x GPUs up to 32 devices. Strong-scaling efficiency reaches 66% at 16 devices and 40% at 32; weak-scaling efficiency is 80% to 16 devices and reaches 48% (MI250x) and 40% (A100) at 32 devices. Profiling with the ROCm System profiler shows that >90% of the wall time is spent in DeePMD inference, while MPI collectives contribute <10%, primarily since they act as a global synchronization point. The principal bottlenecks are the irreducible ghost-atom cost set by the cutoff radius, confirmed by a simple throughput model, and load imbalance across ranks. These results demonstrate that production MD with near ab initio fidelity is feasible at scale in GROMACS.

Large language models (LLMs) have demonstrated strong capabilities in medical question answering; however, purely parametric models often suffer from knowledge gaps and limited factual grounding. Retrieval-augmented generation (RAG) addresses this limitation by integrating external knowledge retrieval into the reasoning process. Despite increasing interest in RAG-based medical systems, the impact of individual retrieval components on performance remains insufficiently understood. This study presents a systematic evaluation of retrieval-augmented medical question answering using the MedQA USMLE benchmark and a structured textbook-based knowledge corpus. We analyze the interaction between language models, embedding models, retrieval strategies, query reformulation, and cross-encoder reranking within a unified experimental framework comprising forty configurations. Results show that retrieval augmentation significantly improves zero-shot medical question answering performance. The best-performing configuration was dense retrieval with query reformulation and reranking achieved 60.49% accuracy. Domain-specialized language models were also found to better utilize retrieved medical evidence than general-purpose models. The analysis further reveals a clear tradeoff between retrieval effectiveness and computational cost, with simpler dense retrieval configurations providing strong performance while maintaining higher throughput. All experiments were conducted on a single consumer-grade GPU, demonstrating that systematic evaluation of retrieval-augmented medical QA systems can be performed under modest computational resources.

The widespread use of clickbait headlines, crafted to mislead and maximize engagement, poses a significant challenge to online credibility. These headlines employ sensationalism, misleading claims, and vague language, underscoring the need for effective detection to ensure trustworthy digital content. The paper introduces, ClickGuard: a trustworthy adaptive fusion framework for clickbait detection. It combines BERT embeddings and structural features using a Syntactic-Semantic Adaptive Fusion Block (SSAFB) for dynamic integration. The framework incorporates a hybrid CNN-BiLSTM to capture patterns and dependencies. The model achieved 96.93% testing accuracy, outperforming state-of-the-art approaches. The model's trustworthiness is evaluated using LIME and Permutation Feature Importance (PFI) for interpretability and perturbation analysis. These methods assess the model's robustness and sensitivity to feature changes by measuring the average prediction variation. Ablation studies validated the SSAFB's effectiveness in optimizing feature fusion. The model demonstrated robust performance across diverse datasets, providing a scalable, reliable solution for enhancing online content credibility by addressing syntactic-semantic modelling challenges. Code of the work is available at: https://github.com/palindromeRice/ClickBait_Detection_Architecture

Clinical expertise improves not only by acquiring medical knowledge, but by accumulating experience that yields reusable diagnostic patterns. Recent LLMs-based diagnostic agents have shown promising progress in clinical reasoning for decision support. However, most approaches treat cases independently, limiting experience reuse and continual adaptation. We propose SEA, a self-learning diagnostic agent with cognitively inspired dual-memory module. We design a reinforcement training framework tailored to our designed agent for joint optimization of reasoning and memory management. We evaluate SEA in two complementary settings. On standard evaluation with MedCaseReasoning dataset, SEA achieves 92.46% accuracy, outperforming the strongest baseline by +19.6%, demonstrating the benefit of jointly optimizing reasoning and memory. On the long-horizon with ER-Reason dataset, SEA attains the best final accuracy (0.7214) and the largest improvement (+0.35 Acc@100), while baseline methods show limited or unstable gains. Expert evaluation further indicates that rules consolidated from SEA show strong clinical correctness, usefulness and trust, suggesting that the induced rules in dual-memory module are reliable and practically meaningful. Overall, SEA improves both diagnostic reasoning ability and continual learning by effectively transforming experience into reusable knowledge.

Gaussian process ($GP$) regression is a widely used non-parametric modeling tool, but its cubic complexity in the training size limits its use on massive data sets. A practical remedy is to predict using only the nearest neighbours of each test point, as in Nearest Neighbour Gaussian Process ($NNGP$) regression for geospatial problems and the related scalable $GPnn$ method for more general machine-learning applications. Despite their strong empirical performance, the large-$n$ theory of $NNGP/GPnn$ remains incomplete. We develop a theoretical framework for $NNGP$ and $GPnn$ regression. Under mild regularity assumptions, we derive almost sure pointwise limits for three key predictive criteria: mean squared error ($MSE$), calibration coefficient ($CAL$), and negative log-likelihood ($NLL$). We then study the $L_2$-risk, prove universal consistency, and show that the risk attains Stone's minimax rate $n^{-2α/(2p+d)}$, where $α$ and $p$ capture regularity of the regression problem. We also prove uniform convergence of $MSE$ over compact hyper-parameter sets and show that its derivatives with respect to lengthscale, kernel scale, and noise variance vanish asymptotically, with explicit rates. This explains the observed robustness of $GPnn$ to hyper-parameter tuning. These results provide a rigorous statistical foundation for $NNGP/GPnn$ as a highly scalable and principled alternative to full $GP$ models.

Evaluating robustness under temporal distribution shift remains an open challenge. Existing metrics quantify the average decline in performance, but fail to capture how models adapt to evolving data. As a result, temporal degradation is often misinterpreted: when accuracy declines, it is unclear whether the model is failing to adapt or whether the data itself has become inherently more challenging to learn. In this work, we propose three complementary metrics to distinguish adaptation from intrinsic difficulty in the data. Together, these metrics provide a dynamic and interpretable view of model behavior under temporal distribution shift. Results show that our metrics uncover adaptation patterns hidden by existing analysis, offering a richer understanding of temporal robustness in evolving environments.

Operating LEO mega-constellations requires translating high-level operator intents ("reroute financial traffic away from polar links under 80 ms") into low-level routing constraints -- a task that demands both natural language understanding and network-domain expertise. We present an end-to-end system comprising three components: (1) a GNN cost-to-go router that distills Dijkstra-quality routing into a 152K-parameter graph attention network achieving 99.8% packet delivery ratio with 17x inference speedup; (2) an LLM intent compiler that converts natural language to a typed constraint intermediate representation using few-shot prompting with a verifier-feedback repair loop, achieving 98.4% compilation rate and 87.6% full semantic match on feasible intents in a 240-intent benchmark (193 feasible, 47 infeasible); and (3) an 8-pass deterministic validator with constructive feasibility certification that achieves 0% unsafe acceptance on all 47 infeasible intents (30 labeled + 17 discovered by Pass 8), with 100% corruption detection across 240 structural corruption tests and 100% on 15 targeted adversarial attacks. End-to-end evaluation across four constrained routing scenarios confirms zero constraint violations with both routers. We further demonstrate that apparent performance gaps in polar-avoidance scenarios are largely explained by topological reachability ceilings rather than routing quality, and that the LLM compiler outperforms a rule-based baseline by 46.2 percentage points on compositional intents. Our system bridges the semantic gap between operator intent and network configuration while maintaining the safety guarantees required for operational deployment.

Physical Reservoir Computing (PRC) leverages the intrinsic nonlinear dynamics of physical substrates, mechanical, optical, spintronic, and beyond, as fixed computational reservoirs, offering a compelling paradigm for energy-efficient and embodied machine learning. However, the practical workflow for developing and evaluating PRC systems remains fragmented: existing tools typically address only isolated parts of the pipeline, such as substrate-specific simulation, digital reservoir benchmarking, or readout training. What is missing is a unified framework that can represent both high-fidelity simulated trajectories and real experimental measurements through the same data interface, enabling reproducible evaluation, analysis, and physics-aware optimization across substrates and data sources. We present OpenPRC, an open-source Python framework that fills this gap through a schema-driven physics-to-task pipeline built around five modules: a GPU-accelerated hybrid RK4-PBD physics engine (demlat), a video-based experimental ingestion layer (openprc.vision), a modular learning layer (reservoir), information-theoretic analysis and benchmarking tools (analysis), and physics-aware optimization (optimize). A universal HDF5 schema enforces reproducibility and interoperability, allowing GPU-simulated and experimentally acquired trajectories to enter the same downstream workflow without modification. Demonstrated capabilities include simulations of Origami tessellations, video-based trajectory extraction from a physical reservoir, and a common interface for standardized PRC benchmarking, correlation diagnostics, and capacity analysis. The longer-term vision is to serve as a standardizing layer for the PRC community, compatible with external physics engines including PyBullet, PyElastica, and MERLIN.

Recent LLM-based agents often place world modeling, planning, and reflection inside a single language model loop. This can produce capable behavior, but it makes a basic scientific question difficult to answer: which part of the agent's competence actually comes from the LLM, and which part comes from explicit structure around it? We study this question not by claiming a general answer, but by making it empirically tractable. We introduce a declared reflective runtime protocol that externalizes agent state, confidence signals, guarded actions, and hypothetical transitions into inspectable runtime structure. We instantiate this protocol in a declarative runtime and evaluate it on noisy Collaborative Battleship [4] using four progressively structured agents over 54 games (18 boards $\times$ 3 seeds). The resulting decomposition isolates four components: posterior belief tracking, explicit world-model planning, symbolic in-episode reflection, and sparse LLM-based revision. Across this decomposition, explicit world-model planning improves substantially over a greedy posterior-following baseline (+24.1pp win rate, +0.017 F1). Symbolic reflection operates as a real runtime mechanism -- with prediction tracking, confidence gating, and guarded revision actions -- even though its current revision presets are not yet net-positive in aggregate. Adding conditional LLM revision at about 4.3\% of turns yields only a small and non-monotonic change: average F1 rises slightly (+0.005) while win rate drops (31$\rightarrow$29 out of 54). These results suggest a methodological contribution rather than a leaderboard claim: externalizing reflection turns otherwise latent agent behavior into inspectable runtime structure, allowing the marginal role of LLM intervention to be studied directly.

Recent advances in text-to-image (T2I) generation have enabled visually coherent image synthesis from descriptions, but generating images containing multiple given subjects remains challenging. As the number of reference identities increases, existing methods often suffer from subject missing and semantic drift. To address this problem, we propose MUSIC, the first MLLM specifically designed for \textbf{MU}lti-\textbf{S}ubject \textbf{I}n-\textbf{C}ontext image generation. To overcome the data scarcity, we introduce an automatic and scalable data generation pipeline that eliminates the need for manual annotation. Furthermore, we enhance the model's understanding of multi-subject semantic relationships through a vision chain-of-thought (CoT) mechanism, guiding step-by-step reasoning from subject images to semantics and generation. To mitigate identity entanglement and manage visual complexity, we develop a novel semantics-driven spatial layout planning method and demonstrate its test-time scalability. By incorporating complex subject images during training, we improve the model's capacity for chained reasoning. In addition, we curate MSIC, a new benchmark tailored for multi-subject in-context generation. Experimental results demonstrate that MUSIC significantly surpasses other methods in both multi- and single-subject scenarios.

Full-waveform inversion (FWI) is pivotal for reconstructing high-resolution subsurface velocity models but remains computationally intensive and ill-posed. While deep learning approaches promise efficiency, existing Convolutional Neural Networks (CNNs) and single-paradigm Neural Operators (NOs) struggle with one fundamental issue: frequency entanglement of multi-scale geological features. To address this challenge, we propose Spectral-Preserving Adaptive MoE (SPAMoE), a novel spectrum-aware framework for solving inverse problems with complex multi-scale structures. Our approach introduces a Spectral-Preserving DINO Encoder that enforces a lower bound on the high-to-low frequency energy ratio of the encoded representation, mitigating high-frequency collapse and stabilizing subsequent frequency-domain modeling. Furthermore, we design a novel Spectral Decomposition and Routing mechanism that dynamically assigns frequency bands to a Mixture-of-Experts (MoE) ensemble comprising FNO, MNO, and LNO. On the ten OpenFWI sub-datasets, experiments show that SPAMoE reduces the average MAE by 54.1% relative to the best officially reported OpenFWI baseline, thereby establishing a new architectural framework for learning-based full-waveform inversion.

Kuaishou serves over 400 million daily active users, processing hundreds of millions of search queries daily against a repository of tens of billions of short videos. As the final decision layer, the reranking stage determines user experience by optimizing whole-page utility. While traditional score-and-sort methods fail to capture combinatorial dependencies, Generative Reranking offers a superior paradigm by directly modeling the permutation probability. However, deploying Generative Reranking in such a high-stakes environment faces a fundamental dual dilemma: 1) the structural trade-off where Autoregressive (AR) models offer superior Sequential modeling but suffer from prohibitive latency, versus Non-Autoregressive (NAR) models that enable efficiency but lack dependency capturing; 2) the optimization gap where Supervised Learning faces challenges in directly optimizing whole-page utility, while Reinforcement Learning (RL) struggles with instability in high-throughput data streams. To resolve this, we propose Dual-Rerank, a unified framework designed for industrial reranking that bridges the structural gap via Sequential Knowledge Distillation and addresses the optimization gap using List-wise Decoupled Reranking Optimization (LDRO) for stable online RL. Extensive A/B testing on production traffic demonstrates that Dual-Rerank achieves State-of-the-Art performance, significantly improving User satisfaction and Watch Time while drastically reducing inference latency compared to AR baselines.

Optimizing expensive black-box objectives over mixed search spaces is a common challenge across the natural sciences. Bayesian optimization (BO) offers sample-efficient strategies through probabilistic surrogate models and acquisition functions. However, its effectiveness diminishes in mixed or high-cardinality discrete spaces, where gradients are unavailable and optimizing the acquisition function becomes computationally demanding. In this work, we generalize the probabilistic reparameterization (PR) approach of Daulton et al. to handle non-equidistant discrete variables, enabling gradient-based optimization in fully mixed-variable settings with Gaussian process (GP) surrogates. With real-world scientific optimization tasks in mind, we conduct systematic benchmarks on synthetic and experimental objectives to obtain an optimized kernel formulations and demonstrate the robustness of our generalized PR method. We additionally show that, when combined with a modified BO workflow, our approach can efficiently optimize highly discontinuous and discretized objective landscapes. This work establishes a practical BO framework for addressing fully mixed optimization problems in the natural sciences, and is particularly well suited to autonomous laboratory settings where noise, discretization, and limited data are inherent.

Large language models (LLMs) are probabilistic in nature and perform more reliably when augmented with external information. As complex queries often require multi-step reasoning over the retrieved information, with no clear or predetermined reasoning path, they remain challenging. Recent approaches train models using reinforcement learning on the model's outcome, showing promise in improving how models handle complex information. We introduce SubSearch, a specialized framework that shifts from outcome-only supervision to intermediate reward signals that incentivize planning high-quality reasoning. Unlike previous work on process reward modeling, which focuses on training a separate reward model with annotated trajectories by either human annotators or large LLM judges, SubSearch directly optimizes the generator using intrinsic process rewards, which we define as internally-derived rewards, eliminating the need for external supervision, and moving towards autonomous information-intensive reasoning. Experiments on seven benchmarks show that rewarding intermediate reasoning steps with intrinsic rewards leads to more robust reasoning traces in both QA and multi-hop QA datasets over using only outcome rewards. SubSearch can help in building reasoning traces that allow agents to better integrate search engines for complex query answering, while offering a data-efficient alternative to supervised process modeling.

Sales dialogues require multi-turn, goal-directed persuasion under asymmetric incentives, which makes them a challenging setting for large language models (LLMs). Yet existing dialogue benchmarks rarely measure deal progression and outcomes. We introduce SalesLLM benchmark, a bilingual (ZH/EN) benchmark derived from realistic applications covering Financial Services and Consumer Goods, built from 30,074 scripted configurations and 1,805 curated multi-turn scenarios with controllable difficulty and personas. We propose a fully automatic evaluation pipeline that combines (i) an LLM-based rater for sales-process progress,and (ii) fine-tuned BERT classifiers for end-of-dialogue buying intent. To improve simulation fidelity, we train a user model, CustomerLM, with SFT and DPO on 8,000+ crowdworker-involved sales conversations, reducing role inversion from 17.44% (GPT-4o) to 8.8%. SalesLLM benchmark scores correlate strongly with expert human ratings (Pearson r=0.98). Experiments across 15 mainstream LLMs reveal substantial variability: top-performance LLMs are competitive with human-level performance while the less capable ones are worse than human. SalesLLM benchmark serves as a scalable benchmark for developing and evaluating outcome-oriented sales agents.

Ensuring reliable performance in situations outside the Operational Design Domain (ODD) remains a primary challenge in devising resilient autonomous systems. We explore this challenge by introducing an approach for adapting probabilistic system models to handle out-of-ODD scenarios while, in parallel, providing quantitative guarantees. Our approach dynamically extends the coverage of existing system situation capabilities, supporting the verification and adaptation of the system's behaviour under unanticipated situations. Preliminary results demonstrate that our approach effectively increases system reliability by adapting its behaviour and providing formal guarantees even under unforeseen out-of-ODD situations.

The manufacturing sector is increasingly adopting Multimodal Large Language Models (MLLMs) to transition from simple perception to autonomous execution, yet current evaluations fail to reflect the rigorous demands of real-world manufacturing environments. Progress is hindered by data scarcity and a lack of fine-grained domain semantics in existing datasets. To bridge this gap, we introduce FORGE. Wefirst construct a high-quality multimodal dataset that combines real-world 2D images and 3D point clouds, annotated with fine-grained domain semantics (e.g., exact model numbers). We then evaluate 18 state-of-the-art MLLMs across three manufacturing tasks, namely workpiece verification, structural surface inspection, and assembly verification, revealing significant performance gaps. Counter to conventional understanding, the bottleneck analysis shows that visual grounding is not the primary limiting factor. Instead, insufficient domain-specific knowledge is the key bottleneck, setting a clear direction for future research. Beyond evaluation, we show that our structured annotations can serve as an actionable training resource: supervised fine-tuning of a compact 3B-parameter model on our data yields up to 90.8% relative improvement in accuracy on held-out manufacturing scenarios, providing preliminary evidence for a practical pathway toward domain-adapted manufacturing MLLMs. The code and datasets are available at https://ai4manufacturing.github.io/forge-web.

Accurate knowledge of acoustic surface admittance or impedance is essential for reliable wave-based simulations, yet its in situ estimation remains challenging due to noise, model inaccuracies, and restrictive assumptions of conventional methods. This work presents a physics-informed neural operator approach for estimating frequency-dependent surface admittance directly from near-field measurements of sound pressure and particle velocity. A deep operator network is employed to learn the mapping from measurement data, spatial coordinates, and frequency to acoustic field quantities, while simultaneously inferring a globally consistent surface admittance spectrum without requiring an explicit forward model. The governing acoustic relations, including the Helmholtz equation, the linearized momentum equation, and Robin boundary conditions, are embedded into the training process as physics-based regularization, enabling physically consistent and noise-robust predictions while avoiding frequency-wise inversion. The method is validated using synthetically generated data from a simulation model for two planar porous absorbers under semi free-field conditions across a broad frequency range. Results demonstrate accurate reconstruction of both real and imaginary admittance components and reliable prediction of acoustic field quantities. Parameter studies confirm improved robustness to noise and sparse sampling compared to purely data-driven approaches, highlighting the potential of physics-informed neural operators for in situ acoustic material characterization.

Energy efficiency in mobile networks is crucial for sustainable telecommunications infrastructure, particularly as network densification continues to increase power consumption. Sleep mechanisms for the components in mobile networks can reduce energy use, but deciding which components to put to sleep, when, and for how long while preserving quality of service (QoS) remains a difficult optimisation problem. In this paper, we utilise reinforcement learning with reward machines (RMs) to make sleep-control decisions that balance immediate energy savings and long-term QoS impact, i.e. time-averaged packet drop rates for deadline-constrained traffic and time-averaged minimum-throughput guarantees for constant-rate users. A challenge is that time-averaged constraints depend on cumulative performance over time rather than immediate performance. As a result, the effective reward is non-Markovian, and optimal actions depend on operational history rather than the instantaneous system state. RMs account for the history dependence by maintaining an abstract state that explicitly tracks the QoS constraint violations over time. Our framework provides a principled, scalable approach to energy management for next-generation mobile networks under diverse traffic patterns and QoS requirements.

Layout plays a crucial role in graphic design and poster generation. Recently, the application of deep learning models for layout generation has gained significant attention. This paper focuses on using a GAN-based model conditioned on images to generate advertising poster graphic layouts, requiring a dataset of paired product images and layouts. To address this task, we introduce the Content-aware Graphic Layout Dataset (CGL-Dataset), consisting of 60,548 paired inpainted posters with annotations and 121,000 clean product images. The inpainting artifacts introduce a domain gap between the inpainted posters and clean images. To bridge this gap, we design two GAN-based models. The first model, CGL-GAN, uses Gaussian blur on the inpainted regions to generate layouts. The second model combines unsupervised domain adaptation by introducing a GAN with a pixel-level discriminator (PD), abbreviated as PDA-GAN, to generate image-aware layouts based on the visual texture of input images. The PD is connected to shallow-level feature maps and computes the GAN loss for each input-image pixel. Additionally, we propose three novel content-aware metrics to assess the model's ability to capture the intricate relationships between graphic elements and image content. Quantitative and qualitative evaluations demonstrate that PDA-GAN achieves state-of-the-art performance and generates high-quality image-aware layouts.

Why does gradient descent reliably find good solutions in non-convex neural network optimization, despite the landscape being NP-hard in the worst case? We show that gradient flow on L-layer ReLU networks without bias preserves L-1 conservation laws C_l = ||W_{l+1}||_F^2 - ||W_l||_F^2, confining trajectories to lower-dimensional manifolds. Under discrete gradient descent, these laws break with total drift scaling as eta^alpha where alpha is approximately 1.1-1.6 depending on architecture, loss function, and width. We decompose this drift exactly as eta^2 * S(eta), where the gradient imbalance sum S(eta) admits a closed-form spectral crossover formula with mode coefficients c_k proportional to e_k(0)^2 * lambda_{x,k}^2, derived from first principles and validated for both linear (R=0.85) and ReLU (R>0.80) networks. For cross-entropy loss, softmax probability concentration drives exponential Hessian spectral compression with timescale tau = Theta(1/eta) independent of training set size, explaining why cross-entropy self-regularizes the drift exponent near alpha=1.0. We identify two dynamical regimes separated by a width-dependent transition: a perturbative sub-Edge-of-Stability regime where the spectral formula applies, and a non-perturbative regime with extensive mode coupling. All predictions are validated across 23 experiments.

We analyze the score field of a diffusion generative model through a Burgers-type evolution law. For VE diffusion, the heat-evolved data density implies that the score obeys viscous Burgers in one dimension and the corresponding irrotational vector Burgers system in $\R^d$, giving a PDE view of \emph{speciation transitions} as the sharpening of inter-mode interfaces. For any binary decomposition of the noised density into two positive heat solutions, the score separates into a smooth background and a universal $\tanh$ interfacial term determined by the component log-ratio; near a regular binary mode boundary this yields a normal criterion for speciation. In symmetric binary Gaussian mixtures, the criterion recovers the critical diffusion time detected by the midpoint derivative of the score and agrees with the spectral criterion of Biroli, Bonnaire, de~Bortoli, and Mézard (2024). After subtracting the background drift, the inter-mode layer has a local Burgers $\tanh$ profile, which becomes global in the symmetric Gaussian case with width $σ_τ^2/a$. We also quantify exponential amplification of score errors across this layer, show that Burgers dynamics preserves irrotationality, and use a change of variables to reduce the VP-SDE to the VE case, yielding a closed-form VP speciation time. Gaussian-mixture formulas are verified to machine precision, and the local theorem is checked numerically on a quartic double-well.

Autoregressive models have shown superior performance and efficiency in image generation, but remain constrained by high computational costs and prolonged training times in video generation. In this study, we explore methods to accelerate training for autoregressive video generation models through empirical analyses. Our results reveal that while training on fewer video frames significantly reduces training time, it also exacerbates error accumulation and introduces inconsistencies in the generated videos. To address these issues, we propose a Local Optimization (Local Opt.) method, which optimizes tokens within localized windows while leveraging contextual information to reduce error propagation. Inspired by Lipschitz continuity, we propose a Representation Continuity (ReCo) strategy to improve the consistency of generated videos. ReCo utilizes continuity loss to constrain representation changes, improving model robustness and reducing error accumulation. Extensive experiments on class- and text-to-video datasets demonstrate that our approach achieves superior performance to the baseline while halving the training cost without sacrificing quality.

We study the thermodynamic memory capacity of modern Hopfield networks (Dense Associative Memory models) with continuous states under geometric constraints, extending classical analyses of pairwise associative memory. We derive thermodynamic phase boundaries for Dense Associative Memory networks with exponential capacity $p = e^{αN}$, comparing Gaussian (LSE) and Epanechnikov (LSR) kernels. For continuous neurons on an $N$-sphere, the geometric entropy depends solely on the spherical geometry, not the kernel. In the sharp-kernel regime, the maximum theoretical capacity $α= 0.5$ is achieved at zero temperature; below this threshold, a critical line separates retrieval from a spin-glass phase. The two kernels differ qualitatively in their phase boundary structure: for LSE, the retrieval region extends to arbitrarily high temperatures as $α\to 0$, but interference from spurious patterns is always present. For LSR, the finite support introduces a threshold $α_{\text{th}}$ below which no spurious patterns contribute to the noise floor, producing a qualitatively different retrieval regime in this sub-threshold region. These results advance the theory of high-capacity associative memory and clarify fundamental limits of retrieval robustness in modern attention-like memory architectures.

Continual learning (CL) on edge devices requires not only high accuracy but also training-time efficiency to support on-device adaptation under strict memory and computational constraints. While prompt-based continual learning (PCL) is parameter-efficient and achieves competitive accuracy, prior work has focused mainly on accuracy or inference-time performance, often overlooking the memory and computational costs of on-device training. In this paper, we propose CPS-Prompt, a critical patch-aware sparse prompting framework that explicitly targets training-time memory usage and computational cost by integrating critical patch sampling (CPS) for task-aware token reduction and decoupled prompt and classifier training (DPCT) to reduce backpropagation overhead. Experiments on three public benchmarks and real edge hardware show that CPS-Prompt improves peak memory, training time, and energy efficiency by about 1.6x over the balanced CODA-Prompt baseline, while maintaining accuracy within 2% of the state-of-the-art C-Prompt on average and remaining competitive with CODA-Prompt in accuracy. The code is available at https://github.com/laymond1/cps-prompt.

Large language models (LLMs) in research and development toolchains produce output that triggers attribution of agency and understanding -- a cognitive illusion that degrades verification behavior and trust calibration. No existing mitigation provides a systematic, deployable constraint set for output register. This paper proposes seven output-side rules, each targeting a documented linguistic mechanism, and validates them empirically. In 780 two-turn conversations (constrained vs. default register, 30 tasks, 13 replicates, 1560 API calls), anthropomorphic markers dropped from 1233 to 33 (>97% reduction, p < 0.001), outputs were 49% shorter by word count, and adapted AnthroScore confirmed the shift toward machine register (-1.94 vs. -0.96, p < 0.001). The rules are implemented as a configuration-file system prompt requiring no model modification; validation uses a single model (Claude Sonnet 4). Output quality under the constrained register was not evaluated. The mechanism is extensible to other domains.

Quantization is an effective way to reduce the memory cost of large-scale model training. However, most existing methods adopt fixed-precision policies, which ignore the fact that optimizer-state distributions vary significantly across layers and training steps. Such uniform designs often introduce noticeable accuracy degradation. To move beyond fixed quantization, we propose STQuant, a distributed training framework that reduces the memory footprint of optimizer states via dynamic precision allocation across layers, state variables, and training steps, while maintaining model quality. Naively applying dynamic quantization during training is challenging for two reasons. First, optimizer states are numerically sensitive, and quantization noise can destabilize quality. Second, jointly considering multiple states and layers induces a large combinatorial search space. STQuant addresses these challenges with two key techniques: 1) a provably near-optimal factor selection strategy that accurately identifies the most influential factors for precision adaptation. 2) a dynamic transition decision algorithm that reduces the search cost from exponential to linear complexity. Experiments on GPT-2 and ViT show that STQuant reduces optimizer-state memory by 84.4%, achieving an average bit-width of as low as 5.1 bits, compared with existing solutions. Moreover, STQuant incurs only O(N/K) computational overhead and requires O(1) extra space.

A key inefficiency in diffusion training occurs when a randomly initialized network, lacking visual priors, encounters gradients from the full complexity spectrum--most of which it lacks the capacity to resolve. We propose Data Warmup, a curriculum strategy that schedules training images from simple to complex without modifying the model or loss. Each image is scored offline by a semantic-aware complexity metric combining foreground dominance (how much of the image salient objects occupy) and foreground typicality (how closely the salient content matches learned visual prototypes). A temperature-controlled sampler then prioritizes low-complexity images early and anneals toward uniform sampling. On ImageNet 256x256 with SiT backbones (S/2 to XL/2), Data Warmup improves IS by up to 6.11 and FID by up to 3.41, reaching baseline quality tens of thousands of iterations earlier. Reversing the curriculum (exposing hard images first) degrades performance below the uniform baseline, confirming that the simple-to-complex ordering itself drives the gains. The method combines with orthogonal accelerators such as REPA and requires only ~10 minutes of one-time preprocessing with zero per-iteration overhead.

Synthetic data rephrasing has emerged as a powerful technique for enhancing knowledge acquisition during large language model (LLM) pretraining. However, existing approaches operate at the single-document level, rewriting individual web pages in isolation. This confines synthesized examples to intra-document knowledge, missing cross-document relationships and leaving facts with limited associative context. We propose WRAP++ (Web discoveRy Amplified Pretraining), which amplifies the associative context of factual knowledge by discovering cross-document relationships from web hyperlinks and synthesizing joint QA over each discovered document pair. Concretely, WRAP++ discovers high-confidence relational motifs including dual-links and co-mentions, and synthesizes QA that requires reasoning across both documents. This produces relational knowledge absent from either source document alone, creating diverse entry points to the same facts. Because the number of valid entity pairs grows combinatorially, this discovery-driven synthesis also amplifies data scale far beyond single-document rewriting. Instantiating WRAP++ on Wikipedia, we amplify ~8.4B tokens of raw text into 80B tokens of cross-document QA data. On SimpleQA, OLMo-based models at both 7B and 32B scales trained with WRAP++ substantially outperform single-document approaches and exhibit sustained scaling gains, underscoring the advantage of cross-document knowledge discovery and amplification.

Large Language Model (LLM) inference on edge Neural Processing Units (NPUs) is fundamentally constrained by limited on-chip memory capacity. Although high-density embedded DRAM (eDRAM) is attractive for storing activation workspaces, its periodic refresh consumes substantial energy. Prior work has primarily focused on reducing off-chip traffic or optimizing refresh for persistent Key-Value (KV) caches, while transient and error-resilient Query and Attention Output (QO) activations are largely overlooked. We propose SHIELD, a lifecycle-aware segmented eDRAM architecture that jointly exploits temporal residency and bit-level sensitivity in bfloat16 (BF16) activations. SHIELD isolates the sign and exponent fields from the mantissa, disables refresh for transient QO mantissas, and applies relaxed refresh to persistent KV mantissas. Across multiple LLMs and inference scenarios, SHIELD reduces eDRAM refresh energy by 35% relative to a standard-refresh baseline while preserving accuracy on WikiText-2, PIQA, and ARC-Easy.

Robotic manipulation systems that follow language instructions often execute grasp primitives in a largely single-shot manner: a model proposes an action, the robot executes it, and failures such as empty grasps, slips, stalls, timeouts, or semantically wrong grasps are not surfaced to the decision layer in a structured way. Inspired by agentic loops in digital tool-using agents, we reformulate language-guided grasping as a bounded embodied agent operating over grounded execution states, where physical actions expose an explicit tool-state stream. We introduce a physical agentic loop that wraps an unmodified learned manipulation primitive (grasp-and-lift) with (i) an event-based interface and (ii) an execution monitoring layer, Watchdog, which converts noisy gripper telemetry into discrete outcome labels using contact-aware fusion and temporal stabilization. These outcome events, optionally combined with post-grasp semantic verification, are consumed by a deterministic bounded policy that finalizes, retries, or escalates to the user for clarification, guaranteeing finite termination. We validate the resulting loop on a mobile manipulator with an eye-in-hand D405 camera, keeping the underlying grasp model unchanged and evaluating representative scenarios involving visual ambiguity, distractors, and induced execution failures. Results show that explicit execution-state monitoring and bounded recovery enable more robust and interpretable behavior than open-loop execution, while adding minimal architectural overhead. For the source code and demo refer to our project page: https://wenzewwz123.github.io/Agentic-Loop/

The quadratic computational complexity of standard attention mechanisms presents a severe scalability bottleneck for LLMs in long-context scenarios. While hybrid attention mechanisms combining Full Attention (FA) and Sparse Attention (SA) offer a potential solution, existing methods typically rely on static allocation ratios that fail to accommodate the variable retrieval demands of different tasks. Furthermore, head-level dynamic sparsity often introduces severe computational load imbalance and synchronization long-tails, which hinder hardware acceleration during autoregressive decoding. To bridge this gap, we introduce Flux Attention, a context-aware framework that dynamically optimizes attention computation at the layer level. By integrating a lightweight Layer Router into frozen pretrained LLMs, the proposed method adaptively routes each layer to FA or SA based on the input context. This layer-wise routing preserves high-fidelity information retrieval while ensuring contiguous memory access, translating theoretical computational reductions into practical wall-clock speedups. As a parameter-efficient approach, our framework requires only 12 hours of training on 8$\times$A800 GPUs. Extensive experiments across multiple long-context and mathematical reasoning benchmarks demonstrate that Flux Attention achieves a superior trade-off between performance and inference speed compared with baseline models, with speed improvements of up to $2.8\times$ and $2.0\times$ in the prefill and decode stages.

The aerodynamic design of turbomachinery is a complex and tightly coupled multi-stage process involving geometry generation, performance prediction, optimization, and high-fidelity physical validation. Existing intelligent design approaches typically focus on individual stages or rely on loosely coupled pipelines, making fully autonomous end-to-end design challenging. To address this issue, this study proposes TurboAgent, a large language model (LLM)-driven autonomous multi-agent framework for turbomachinery aerodynamic design and optimization. The LLM serves as the core for task planning and coordination, while specialized agents handle generative design, rapid performance prediction, multi-objective optimization, and physics-based validation. The framework transforms traditional trial-and-error design into a data-driven collaborative workflow, with high-fidelity simulations retained for final verification. A transonic single-rotor compressor is used for validation. The results show strong agreement between target performance, generated designs, and CFD simulations. The coefficients of determination for mass flow rate, total pressure ratio, and isentropic efficiency all exceed 0.91, with normalized RMSE values below 8%. The optimization agent further improves isentropic efficiency by 1.61% and total pressure ratio by 3.02%. The complete workflow can be executed within approximately 30 minutes under parallel computing. These results demonstrate that TurboAgent enables an autonomous closed-loop design process from natural language requirements to final design generation, providing an efficient and scalable paradigm for turbomachinery aerodynamic design.

Trial-and-error is a fundamental strategy for humans to solve complex problems and a necessary capability for Artificial Intelligence (AI) systems operating in real-world environments. Although several trial-and-error AI techniques have recently been proposed, most of them rely on simple heuristics designed by researchers and achieve limited performance gains. The core issue is the absence of appropriate data: current models cannot learn from detailed records of how humans actually conduct trial-and-error in practice. To address this gap, we introduce a data annotation platform and a corresponding dataset, termed Trial-and-Error Collection (TEC). The platform records users' complete trajectories across multiple trials and collects their reflections after receiving error feedback. Using this platform, we record the problem-solving processes of 46 participants on 58 tasks, resulting in 5,370 trial trajectories along with error reflections across 41,229 webpages. With this dataset, we observe that humans achieve substantially higher accuracy compared to LLMs, which demonstrates that humans are more effective in trial-and-error than LLMs. We believe that the TEC platform and dataset provide a valuable foundation for understanding human trial-and-error behavior and for developing more capable AI systems. Platform and dataset are publicly available.

Accurate forecasting of industrial time series requires balancing predictive accuracy with physical plausibility under non-stationary operating conditions. Existing data-driven models often achieve strong statistical performance but struggle to respect regime-dependent interaction structures and transport delays inherent in real-world systems. To address this challenge, we propose DSPR (Dual-Stream Physics-Residual Networks), a forecasting framework that explicitly decouples stable temporal patterns from regime-dependent residual dynamics. The first stream models the statistical temporal evolution of individual variables. The second stream focuses on residual dynamics through two key mechanisms: an Adaptive Window module that estimates flow-dependent transport delays, and a Physics-Guided Dynamic Graph that incorporates physical priors to learn time-varying interaction structures while suppressing spurious correlations. Experiments on four industrial benchmarks spanning heterogeneous regimes demonstrate that DSPR consistently improves forecasting accuracy and robustness under regime shifts while maintaining strong physical plausibility. It achieves state-of-the-art predictive performance, with Mean Conservation Accuracy exceeding 99% and Total Variation Ratio reaching up to 97.2%. Beyond forecasting, the learned interaction structures and adaptive lags provide interpretable insights that are consistent with known domain mechanisms, such as flow-dependent transport delays and wind-to-power scaling behaviors. These results suggest that architectural decoupling with physics-consistent inductive biases offers an effective path toward trustworthy industrial time-series forecasting. Furthermore, DSPR's demonstrated robust performance in long-term industrial deployment bridges the gap between advanced forecasting models and trustworthy autonomous control systems.

Autonomous agents operating in dynamic and safety-critical environments require decision-making frameworks that are both computationally efficient and physically grounded. However, many existing approaches rely on end-to-end learning, which often lacks interpretability and explicit mechanisms for ensuring consistency with physical constraints. In this work, we propose an event-centric world modeling framework with memory-augmented retrieval for embodied decision-making. The framework represents the environment as a structured set of semantic events, which are encoded into a permutation-invariant latent representation. Decision-making is performed via retrieval over a knowledge bank of prior experiences, where each entry associates an event representation with a corresponding maneuver. The final action is computed as a weighted combination of retrieved solutions, providing a transparent link between decision and stored experiences. The proposed design enables structured abstraction of dynamic environments and supports interpretable decision-making through case-based reasoning. In addition, incorporating physics-informed knowledge into the retrieval process encourages the selection of maneuvers that are consistent with observed system dynamics. Experimental evaluation in UAV flight scenarios demonstrates that the framework operates within real-time control constraints while maintaining interpretable and consistent behavior.

The aggressive densification of modern wireless networks necessitates judicious resource allocation to mitigate severe mutual interference. However, classical iterative algorithms remain computationally prohibitive for real-time applications requiring rapid responsiveness. While recent deep learning-based methods show promise, they typically function as task-specific solvers lacking the flexibility to adapt to different objectives and scenarios without expensive retraining. To address these limitations, we propose a graph foundation model for resource allocation (GFM-RA) based on a pre-training and fine-tuning paradigm to extract unified representations, thereby enabling rapid adaptation to different objectives and scenarios. Specifically, we introduce an interference-aware Transformer architecture with a bias projector that injects interference topologies into global attention mechanisms. Furthermore, we develop a hybrid self-supervised pre-training strategy that synergizes masked edge prediction with negative-free Teacher-Student contrastive learning, enabling the model to capture transferable structural representations from massive unlabeled datasets. Extensive experiments demonstrate that the proposed framework achieves state-of-the-art performance and scales effectively with increased model capacity. Crucially, leveraging its unified representations, the foundation model exhibits exceptional sample efficiency, enabling robust few-shot adaptation to diverse and unsupervised downstream objectives in out-of-distribution (OOD) scenarios. These results demonstrate the promise of pre-trained foundation models for adaptable wireless resource allocation and provide a strong foundation for future research on generalizable learning-based wireless optimization.

Crime pattern analysis is critical for law enforcement and predictive policing, yet the surge in criminal activities from rapid urbanization creates high-dimensional, imbalanced datasets that challenge traditional classification methods. This study presents a quantum-classical comparison framework for crime analytics, evaluating four computational paradigms: quantum models, classical baseline machine learning models, and two hybrid quantum-classical architectures. Using 16-year Bangladesh crime statistics, we systematically assess classification performance and computational efficiency under rigorous cross-validation methods. Experimental results show that quantum-inspired approaches, particularly QAOA, achieve up to 84.6% accuracy, while requiring fewer trainable parameters than classical baselines, suggesting practical advantages for memory-constrained edge deployment. The proposed correlation-aware circuit design demonstrates the potential of incorporating domain-specific feature relationships into quantum models. Furthermore, hybrid approaches exhibit competitive training efficiency, making them suitable candidates for resource-constrained environments. The framework's low computational overhead and compact parameter footprint suggest potential advantages for wireless sensor network deployments in smart city surveillance systems, where distributed nodes perform localized crime analytics with minimal communication costs. Our findings provide a preliminary empirical assessment of quantum-enhanced machine learning for structured crime data and motivate further investigation with larger datasets and realistic quantum hardware considerations.

We present a self-calibrating framework for analog circuit sizing in which a large language model (LLM) derives topology-specific analytical design equations directly from a raw circuit netlist. Unlike existing AI-driven sizing methods where the model proposes parameter adjustments or reduces a search space, the LLM produces a complete Python sizing function tracing each device dimension to a specific performance constraint. A deterministic calibration loop extracts process-dependent parameters from a single transistor-level simulation, while a prediction-error feedback mechanism compensates for analytical inaccuracies. We validate the framework on six operational transconductance amplifier (OTA) topologies spanning three families at two process nodes (180 nm and 40 nm CMOS). All 12 topology-node combinations achieve all specifications, converging in 2-9 simulations for 11 of 12 cases, with one outlier requiring 16 simulations due to an extremely narrow feasible region. Despite large initial prediction errors, convergence depends on the measurement-feedback architecture, not prediction accuracy. This one-shot calibration automatically captures process-dependent variations, enabling cross-node portability without modification, retraining, or per-process characterization.

We present SauerkrautLM-Doom-MultiVec, a 1.3 million parameter model that plays the classic first-person shooter DOOM in real time, outperforming large language models up to 92,000x its size, including Nemotron-120B, Qwen3.5-27B, and GPT-4o-mini. Our model combines a ModernBERT encoder with hash embeddings, depth-aware token representations, and an attention pooling classification head to select game actions from ASCII frame representations at 31ms per decision. Trained on just 31,000 human gameplay demonstrations, it achieves 178 frags in 10 episodes (17.8 per episode) in the defend_the_center scenario, more than all tested LLMs combined (13 frags total). All agents receive equivalent input: ASCII frames and depth maps. Despite having 92,000x fewer parameters than Nemotron-120B, our model is the only agent that actively engages enemies rather than purely evading them. These results demonstrate that small, task-specific models trained on domain-appropriate data can decisively outperform general-purpose LLMs at real-time control tasks, at a fraction of the inference cost, with deployment capability on consumer hardware.

Maternal and child health is a critical concern around the world. In many global health programs disseminating preventive care and health information, limited healthcare worker resources prevent continuous, personalised engagement with vulnerable beneficiaries. In such scenarios, it becomes crucial to optimally schedule limited live-service resources to maximise long-term engagement. To address this fundamental challenge, the multi-year SAHELI project (2020-2025), in collaboration with partner NGO ARMMAN, leverages AI to allocate scarce resources in a maternal and child health program in India. The SAHELI system solves this sequential resource allocation problem using a Restless Multi-Armed Bandit (RMAB) framework. A key methodological innovation is the transition from a traditional Two-Stage "predict-then-optimize" approach to Decision-Focused Learning (DFL), which directly aligns the framework's learning method with the ultimate goal of maximizing beneficiary engagement. Empirical evaluation through large-scale randomized controlled trials demonstrates that the DFL policy reduced cumulative engagement drops by 31% relative to the current standard of care, significantly outperforming the Two-Stage model. Crucially, the studies also confirmed that this increased program engagement translates directly into statistically significant improvements in real-world health behaviors, notably the continued consumption of vital iron and calcium supplements by new mothers. Ultimately, the SAHELI project provides a scalable blueprint for applying sequential decision-making AI to optimize resource allocation in health programs.

Cross-city transfer improves prediction in label-scarce cities by leveraging labeled data from other cities, but it becomes challenging when cities adopt incompatible partitions and no ground-truth region correspondences exist. Existing approaches either rely on heuristic region matching, which is often sensitive to anchor choices, or perform distribution-level alignment that leaves correspondences implicit and can be unstable under strong heterogeneity. We propose SCOT, a cross-city representation learning framework that learns explicit soft correspondences between unequal region sets via Sinkhorn-based entropic optimal transport. SCOT further sharpens transferable structure with an OT-weighted contrastive objective and stabilizes optimization through a cycle-style reconstruction regularizer. For multi-source transfer, SCOT aligns each source and the target to a shared prototype hub using balanced entropic transport guided by a target-induced prototype prior. Across real-world cities and tasks, SCOT consistently improves transfer accuracy and robustness, while the learned transport couplings and hub assignments provide interpretable diagnostics of alignment quality.

The geometric structure of latent representations in large language models (LLMs) is an active area of research, driven in part by its implications for model transparency and AI safety. Existing literature has focused mainly on general geometric and topological properties of the learnt representations, but due to a lack of ground-truth latent geometry, validating the findings of such approaches is challenging. Emotion processing provides an intriguing testbed for probing representational geometry, as emotions exhibit both categorical organization and continuous affective dimensions, which are well-established in the psychology literature. Moreover, understanding such representations carries safety relevance. In this work, we investigate the latent structure of affective representations in LLMs using geometric data analysis tools. We present three main findings. First, we show that LLMs learn coherent latent representations of affective emotions that align with widely used valence--arousal models from psychology. Second, we find that these representations exhibit nonlinear geometric structure that can nonetheless be well-approximated linearly, providing empirical support for the linear representation hypothesis commonly assumed in model transparency methods. Third, we demonstrate that the learned latent representation space can be leveraged to quantify uncertainty in emotion processing tasks. Our findings suggest that LLMs acquire affective representations with geometric structure paralleling established models of human emotion, with practical implications for model interpretability and safety.

We decompose the spectral edge -- the dominant direction of the Gram matrix of parameter updates -- into its gradient and weight-decay components during grokking in two sequence tasks (Dyck-1 and SCAN). We find a sharp two-phase lifecycle: before grokking the edge is gradient-driven and functionally active; at grokking, gradient and weight decay align, and the edge becomes a compression axis that is perturbation-flat yet ablation-critical (>4000x more impactful than random directions). Three universality classes emerge (functional, mixed, compression), predicted by the gap flow equation. Nonlinear probes show information is re-encoded, not lost (MLP $R^2=0.99$ where linear $R^2=0.86$), and removing weight decay post-grok reverses compression while preserving the algorithm.

From butterfly wings to remnants of nuclear detonation, aperiodic order repeatedly emerges in nature, often exhibiting reduced sensitivity to boundaries and symmetry constraints. Inspired by this principle, a paradigm shift is introduced in nanomechanical resonator design from periodic to aperiodic structures, focusing on a special class: quasicrystals (QCs). Although soft clamping enabled by phononic stopbands has become a central strategy for achieving high-$Q_m$ nanomechanical resonators, its practical realization has been largely confined to periodic phononic crystals, where band structure engineering is well established. The potential of aperiodic architectures, however, has remained largely unexplored, owing to their intrinsic complexity and the lack of systematic approaches to identifying and exploiting stopband behavior. Here we demonstrate that soft clamping can be realized in quasicrystal architectures and that high-$Q_m$ nanomechanical resonators can be systematically achieved through a data-driven design framework. As a representative demonstration, the 12-fold QC-based resonator exhibits a quality factor $Q_m \sim 10^7$ and an effective mass of sub-nanograms at MHz frequencies, corresponding to an exceptional force sensitivity of $26.4$~aN/$\sqrt{\text{Hz}}$ compared to previous 2D phononic crystals. These results establish QCs as a robust platform for next-generation nanomechanical resonators and open a new design regime beyond periodic order.

We prove that no continuous, utility-preserving wrapper defense-a function $D: X\to X$ that preprocesses inputs before the model sees them-can make all outputs strictly safe for a language model with connected prompt space, and we characterize exactly where every such defense must fail. We establish three results under successively stronger hypotheses: boundary fixation-the defense must leave some threshold-level inputs unchanged; an $ε$-robust constraint-under Lipschitz regularity, a positive-measure band around fixed boundary points remains near-threshold; and a persistent unsafe region under a transversality condition, a positive-measure subset of inputs remains strictly unsafe. These constitute a defense trilemma: continuity, utility preservation, and completeness cannot coexist. We prove parallel discrete results requiring no topology, and extend to multi-turn interactions, stochastic defenses, and capacity-parity settings. The results do not preclude training-time alignment, architectural changes, or defenses that sacrifice utility. The full theory is mechanically verified in Lean 4 and validated empirically on three LLMs.

In recommender systems, collecting, storing, and processing large-scale interaction data is increasingly costly in terms of time, energy, and computation, yet it remains unclear when additional data stops providing meaningful gains. This paper investigates how offline recommendation performance evolves as the size of the training dataset increases and whether a saturation point can be observed. We implemented a reproducible Python evaluation workflow with two established toolkits, LensKit and RecBole, included 11 large public datasets with at least 7 million interactions, and evaluated 10 tool-algorithm combinations. Using absolute stratified user sampling, we trained models on nine sample sizes from 100,000 to 100,000,000 interactions and measured NDCG@10. Overall, raw NDCG usually increased with sample size, with no observable saturation point. To make result groups comparable, we applied min-max normalization within each group, revealing a clear positive trend in which around 75% of the points at the largest completed sample size also achieved the group's best observed performance. A late-stage slope analysis over the final 10-30% of each group further supported this upward trend: the interquartile range remained entirely non-negative with a median near 1.0. In summary, for traditional recommender systems on typical user-item interaction data, incorporating more training data remains primarily beneficial, while weaker scaling behavior is concentrated in atypical dataset cases and in the algorithmic outlier RecBole BPR under our setup.

Blockchain and artificial intelligence (AI) are increasingly proposed together for securing intelligent networks, but the literature remains fragmented across ledger design, AI-driven detection, cyber-physical applications, and emerging agentic workflows. This paper synthesizes the area through three reusable contributions: (i) a taxonomy of blockchain-AI security for intelligent networks, (ii) integration patterns for verifiable and adaptive security workflows, and (iii) the Blockchain-AI Security Evaluation Blueprint (BASE), a reporting checklist spanning AI quality, ledger behavior, end-to-end service levels, privacy, energy, and reproducibility. The paper also maps the evidence landscape across IoT, critical infrastructure, smart grids, transportation, and healthcare, showing that the conceptual fit is strong but real-world evidence remains uneven and often prototype-heavy. The synthesis clarifies where blockchain contributes provenance, trust, and auditability, where AI contributes detection, adaptation, and orchestration, and where future work should focus on interoperable interfaces, privacy-preserving analytics, bounded agentic automation, and open cross-domain benchmarks. The paper is intended as a reference for researchers and practitioners designing secure, transparent, and resilient intelligent networks.

Video streaming analytics is a crucial workload for vision-language model serving, but the high cost of multimodal inference limits scalability. Prior systems reduce inference cost by exploiting temporal and spatial redundancy in video streams, but they target either the vision transformer (ViT) or the LLM with a limited view, leaving end-to-end opportunities untapped. Moreover, existing methods incur significant overhead to identify redundancy, either through offline profiling and training or costly online computation, making them ill-suited for dynamic real-time streams. We present CodecSight, a codec-guided streaming video analytics system, built on a key observation that video codecs already extract the temporal and spatial structure of each stream as a byproduct of compression. CodecSight treats this codec metadata as a low-cost runtime signal to unify optimization across video decoding, visual processing, and LLM prefilling, with transmission reduction as an inherent benefit of operating directly on compressed bitstreams. This drives codec-guided patch pruning before ViT encoding and selective key-value cache refresh during LLM prefilling, both of which are fully online and do not require offline training. Experiments show that CodecSight achieves an improvement in throughput of up to 3$\times$, and a reduction of up to 87% in GPU compute over state-of-the-art baselines, maintaining competitive accuracy with only 0$\sim$8% F1 drop.

Group synchronization is a fundamental task involving the recovery of group elements from pairwise measurements. For orthogonal group synchronization, the most common approach reformulates the problem as a constrained nonconvex optimization and solves it using projection-based methods, such as the generalized power method. However, these methods rely on exact SVD or QR decompositions in each iteration, which are computationally expensive and become a bottleneck for large-scale problems. In this paper, we propose a Newton-Schulz-based Riemannian Gradient Scheme (NS-RGS) for orthogonal group synchronization that significantly reduces computational cost by replacing the SVD or QR step with the Newton-Schulz iteration. This approach leverages efficient matrix multiplications and aligns perfectly with modern GPU/TPU architectures. By employing a refined leave-one-out analysis, we overcome the challenge arising from statistical dependencies, and establish that NS-RGS with spectral initialization achieves linear convergence to the target solution up to near-optimal statistical noise levels. Experiments on synthetic data and real-world global alignment tasks demonstrate that NS-RGS attains accuracy comparable to state-of-the-art methods such as the generalized power method, while achieving nearly a 2$\times$ speedup.

Video Large Language Models (Video-LLMs) excel in video understanding but suffer from high inference latency during autoregressive generation. Speculative Decoding (SD) mitigates this by applying a draft-and-verify paradigm, yet existing methods are constrained by rigid exact-match rules, severely limiting the acceleration potential. To bridge this gap, we propose LVSpec, the first training-free loosely SD framework tailored for Video-LLMs. Grounded in the insight that generation is governed by sparse visual-relevant anchors (mandating strictness) amidst abundant visual-irrelevant fillers (permitting loose verification), LVSpec employs a lightweight visual-relevant token identification scheme to accurately pinpoint the former. To further maximize acceptance, we augment this with a position-shift tolerant mechanism that effectively salvages positionally mismatched but semantically equivalent tokens. Experiments demonstrate that LVSpec achieves high fidelity and speed: it preserves >99.8 of target performance while accelerating Qwen2.5-VL-32B by 2.70x and LLaVA-OneVision-72B by 2.94x. Notably, it boosts the mean accepted length and speedup ratio by 136% and 35% compared to SOTA training-free SD methods for Video-LLMs.

We present a composition-based approach to building correctby-construction database backing stores. In previous work, we specified the behaviour of several store variants and proved their correctness and equivalence. Here, we derive a Java implementation: the simplicity of the specification makes manual construction straightforward. We leverage spec-guaranteed store equivalence to compose performance features, then demonstrate practical value with CobbleDB, a reimplementation of RocksDB's levelled storage.

Text-to-Video (T2V) generation has benefited from recent advances in diffusion models, yet current systems still struggle under complex scenarios, which are generally exacerbated by the ambiguity and underspecification of text prompts. In this work, we formulate complex-scenario prompt refinement as a stage-wise multi-agent refinement process and propose SCMAPR, i.e., a scenario-aware and Self-Correcting Multi-Agent Prompt Refinement framework for T2V prompting. SCMAPR coordinates specialized agents to (i) route each prompt to a taxonomy-grounded scenario for strategy selection, (ii) synthesize scenario-aware rewriting policies and perform policy-conditioned refinement, and (iii) conduct structured semantic verification that triggers conditional revision when violations are detected. To clarify what constitutes complex scenarios in T2V prompting, provide representative examples, and enable rigorous evaluation under such challenging conditions, we further introduce {T2V-Complexity}, which is a complex-scenario T2V benchmark consisting exclusively of complex-scenario prompts. Extensive experiments on 3 existing benchmarks and our T2V-Complexity benchmark demonstrate that SCMAPR consistently improves text-video alignment and overall generation quality under complex scenarios, achieving up to 2.67\% and 3.28 gains in average score on VBench and EvalCrafter, and up to 0.028 improvement on T2V-CompBench over 3 State-Of-The-Art baselines.

Emotional prompting - the use of specific emotional diction in prompt engineering - has shown increasing promise in improving large language model (LLM) performance, truthfulness, and responsibility. However these studies have been limited to single types of positive emotional stimuli and have not considered varying degrees of emotion intensity in their analyses. In this paper, we explore the effects of four distinct emotions - joy, encouragement, anger, and insecurity - in emotional prompting and evaluate them on accuracy, sycophancy, and toxicity. We develop a prompt-generation pipeline with GPT-4o mini to create a suite of LLM and human-generated prompts with varying intensities across the four emotions. Then, we compile a "Gold Dataset" of prompts where human and model labels align. Our empirical evaluation on LLM behavior suggests that positive emotional stimuli lead to more accurate and less toxic results, but also increase sycophantic behavior.

Automated proof generation for formal software verification remains largely unresolved despite advances in large language models (LLMs). While LLMs perform well in NLP, vision, and code generation, formal verification still requires substantial human effort. Interactive theorem proving (ITP) demands manual proof construction under strict logical constraints, limiting scalability; for example, verifying the seL4 microkernel required decades of effort. Existing LLM-based approaches attempt to automate this process but remain limited. Most rely on single-shot generation or shallow retrieval, which may work for small proofs but fail to scale to large, interdependent verification tasks with deep structural dependencies. We present PROMISE (PROof MIning via Structural Embeddings), a structure-aware framework that reframes proof generation as a stateful search over proof-state transitions. Instead of surface-level retrieval, PROMISE mines structural patterns from proof states and tactic transitions, enabling retrieval and adaptation of compatible proof fragments during iterative search. We evaluate PROMISE on the seL4 benchmark across multiple LLM backends and compare it with prior systems such as Selene and Rango. PROMISE consistently outperforms prior methods, achieving up to +26 point improvements (186% relative gain) while maintaining robustness across models, demonstrating the effectiveness of structure-aware proof mining for scalable theorem proving.

Enterprise IT support interactions are fundamentally diagnostic: effective resolution requires iterative evidence gathering from ambiguous user reports to identify an underlying root cause. While retrieval-augmented generation (RAG) provides grounding through historical cases, standard multi-turn RAG systems lack explicit diagnostic state and therefore struggle to accumulate evidence and resolve competing hypotheses across turns. We introduce DQA, a diagnostic question-answering framework that maintains persistent diagnostic state and aggregates retrieved cases at the level of root causes rather than individual documents. DQA combines conversational query rewriting, retrieval aggregation, and state-conditioned response generation to support systematic troubleshooting under enterprise latency and context constraints. We evaluate DQA on 150 anonymized enterprise IT support scenarios using a replay-based protocol. Averaged over three independent runs, DQA achieves a 78.7% success rate under a trajectory-level success criterion, compared to 41.3% for a multi-turn RAG baseline, while reducing average turns from 8.4 to 3.9.

Skill usage has become a core component of modern agent systems and can substantially improve agents' ability to complete complex tasks. In real-world settings, where agents must monitor and interact with numerous personal applications, web browsers, and other environment interfaces, skill libraries can scale to thousands of reusable skills. Scaling to larger skill sets introduces two key challenges. First, loading the full skill set saturates the context window, driving up token costs, hallucination, and latency. In this paper, we present Graph of Skills (GoS), an inference-time structural retrieval layer for large skill libraries. GoS constructs an executable skill graph offline from skill packages, then at inference time retrieves a bounded, dependency-aware skill bundle through hybrid semantic-lexical seeding, reverse-weighted Personalized PageRank, and context-budgeted hydration. On SkillsBench and ALFWorld, GoS improves average reward by 43.6% over the vanilla full skill-loading baseline while reducing input tokens by 37.8%, and generalizes across three model families: Claude Sonnet, GPT-5.2 Codex, and MiniMax. Additional ablation studies across skill libraries ranging from 200 to 2,000 skills further demonstrate that GoS consistently outperforms both vanilla skills loading and simple vector retrieval in balancing reward, token efficiency, and runtime.

AI coding assistants are now used to generate production code in security-sensitive domains, yet the exploitability of their outputs remains unquantified. We address this gap with Broken by Default: a formal verification study of 3,500 code artifacts generated by seven widely-deployed LLMs across 500 security-critical prompts (five CWE categories, 100 prompts each). Each artifact is subjected to the Z3 SMT solver via the COBALT analysis pipeline, producing mathematical satisfiability witnesses rather than pattern-based heuristics. Across all models, 55.8% of artifacts contain at least one COBALT-identified vulnerability; of these, 1,055 are formally proven via Z3 satisfiability witnesses. GPT-4o leads at 62.4% (grade F); Gemini 2.5 Flash performs best at 48.4% (grade D). No model achieves a grade better than D. Six of seven representative findings are confirmed with runtime crashes under GCC AddressSanitizer. Three auxiliary experiments show: (1) explicit security instructions reduce the mean rate by only 4 points; (2) six industry tools combined miss 97.8% of Z3-proven findings; and (3) models identify their own vulnerable outputs 78.7% of the time in review mode yet generate them at 55.8% by default.

Scientific software relies on high-precision computation, yet finite floating-point representations can introduce precision errors that propagate in safety-critical domains. Despite the growing use of large language models (LLMs) in scientific applications, their reliability in handling floating-point numerical stability has not been systematically evaluated. This paper evaluates LLMs' reasoning on high-precision numerical computation through two numerical stabilization tasks: (1) detecting instability in numerical expressions by generating error-inducing inputs (detection), and (2) rewriting expressions to improve numerical stability (stabilization). Using popular numerical benchmarks, we assess six LLMs on nearly 2,470 numerical structures, including nested conditionals, high-precision literals, and multi-variable arithmetic. Our results show that LLMs are equally effective as state-of-the-art traditional approaches in detecting and stabilizing numerically unstable computations. More notably, LLMs outperform baseline methods precisely where the latter fail: in 17.4% (431) of expressions where the baseline does not improve accuracy, LLMs successfully stabilize 422 (97.9%) of them, and achieve greater stability than the baseline across 65.4% (1,615) of all expressions. However, LLMs struggle with control flow and high-precision literals, consistently removing such structures rather than reasoning about their numerical implications, whereas they perform substantially better on purely symbolic expressions. Together, these findings suggest that LLMs are effective at stabilizing expressions that classical techniques cannot, yet struggle when exact numerical magnitudes and control flow semantics must be precisely reasoned about, as such concrete patterns are rarely encountered during training.

Current document parsing methods advance primarily through model architecture innovation, while systematic engineering of training data remains underexplored. Yet state-of-the-art models spanning diverse architectures and parameter scales exhibit highly consistent failure patterns on the same set of hard samples, suggesting that the performance bottleneck stems from shared deficiencies in training data rather than from architectural differences. Building on this finding, we present MinerU2.5-Pro, which advances the state of the art purely through data engineering and training strategy design while retaining the 1.2B-parameter architecture of MinerU2.5 unchanged. At its core is a Data Engine co-designed around coverage, informativeness, and annotation accuracy: Diversity-and-Difficulty-Aware Sampling expands training data from under 10M to 65.5M samples while mitigating distribution shift; Cross-Model Consistency Verification leverages output consensus among heterogeneous models to assess sample difficulty and generate reliable annotations; the Judge-and-Refine pipeline improves annotation quality for hard samples through render-then-verify iterative correction. A three-stage progressive training strategy--large-scale pre-training, hard sample fine-tuning, and GRPO alignment--sequentially exploits these data at different quality tiers. On the evaluation front, we rectify element-matching biases in OmniDocBench v1.5 and introduce a Hard subset, establishing the more discriminative OmniDocBench v1.6 protocol. Without any architectural modification, MinerU2.5-Pro achieves 95.69 on OmniDocBench v1.6, improving over the same-architecture baseline by 2.71 points and surpassing all existing methods, including those based on models with over 200x more parameters.

Advances in hybrid bonding and packaging have driven growing interest in 3D DRAM-stacked accelerators with higher memory bandwidth and capacity. As LLMs scale to hundreds of billions or trillions of parameters, distributed inference across multiple 3D chips becomes essential. With cross-stack co-design increasingly critical, we propose DeepStack, an accurate and efficient performance model and tool to enable early-stage system-hardware co-design space exploration (DSE) for distributed 3D-stacked AI systems. At the hardware level, DeepStack captures fine-grained 3D memory semantics such as transaction-aware bandwidth, bank activation constraints, buffering limitations, and thermal-power modeling. At the system level, DeepStack incorporates comprehensive parallelization strategies and execution scheduling for distributed LLM inference. With novel modeling techniques such as dual-stage network abstraction and tile-level compute-communication overlap, we achieve up to 100,000x faster runtime over state-of-the-art simulators at comparable accuracy, cross-validated against our in-house 3D designs, NS-3 backend (2.12%), and vLLM serving on 8xB200 GPUs (12.18%). With hierarchical design space search, DeepStack enables efficient exploration over 2.5x10^14 design points spanning 3D-stacked DRAM layers, DRAM vertical connectivity, interconnect, compute-memory allocation, and distributed scheduling. Compared with baseline designs, DeepStack achieves up to 9.5x higher throughput through co-optimized parallelism and 3D architecture search. Our DSE further reveals that batch size drives a more fundamental architectural divide than the prefill/decode distinction, and that parallelism strategy and hardware architecture are tightly coupled -- incomplete schedule search leads to permanently suboptimal silicon irrecoverable by software tuning. We intend to open source DeepStack to support future research.

The rapid adoption of low-precision arithmetic in artificial intelligence and edge computing has created a strong demand for energy-efficient and flexible floating-point multiply-accumulate (MAC) units. This paper presents a dual-precision floating-point MAC processing element supporting FP8 (E4M3, E5M2) and FP4 (2 x E2M1, 2 x E1M2) formats, specifically optimized for low-power and high-throughput AI workloads. The proposed architecture employs a novel bit-partitioning technique that enables a single 4-bit unit multiplier to operate either as a standard 4 x 4 multiplier for FP8 or as two parallel 2 x 2 multipliers for 2-bit operands, achieving maximum hardware utilization without duplicating logic. Implemented in 28 nm technology, the proposed PE achieves an operating frequency of 1.94 GHz with an area of 0.00396 mm^2 and power consumption of 2.13 mW, resulting in up to 60.4% area reduction and 86.6% power savings compared to state-of-the-art designs, making it well suited for energy-constrained AI inference and mixed-precision computing applications when deployed within larger accelerator architectures.

Linear layers hold most of a transformer's parameters. We replace each linear layer with one that stores $K$ out of $mn$ two-dimensional DCT coefficients per weight matrix and reconstructs the full matrix through an inverse DCT at every forward pass; the $K$ coefficients are the trainable parameters. A 4-layer, 128-dim transformer trained from scratch on character-level Shakespeare reaches validation loss $1.604$ at $K = mn/2$, against $1.580$ for a standard dense baseline -- a gap of $+0.024$ at half the trainable parameter count, within the terminal-epoch variation of the dense run. A rank-48 LoRA factorization at the same trainable parameter count reaches only $1.801$ ($+0.221$). The structural advantage of sparse-coefficient over low-rank parameterizations at matched $K$ is qualitative. We identify rank flexibility as the mechanism. A random orthonormal basis matches the DCT within noise at $K = mn/2$, and a compression sweep through $K = mn/10$ and $K = mn/20$ shows that subspaces that can host high-rank matrices keep the loss low, while subspaces that flatten into a low-rank block (zigzag-selection variants) converge onto the observed stable rank \emph{and} the loss line of the rank-48 LoRA reference in lock-step. Among these orthonormal bases, the DCT is preferred because its separable fast transform admits a fused reconstruction kernel: the materialized weight matrix never leaves on-chip memory, so the parameter saving translates into a bandwidth saving as well.

As LLMs are deployed in high-stakes settings, users must judge the correctness of individual responses, often relying on model-generated justifications such as reasoning chains or explanations. Yet, no standard measure exists for whether these justifications help users distinguish correct answers from incorrect ones. We formalize this idea as error verifiability and propose $v_{\text{bal}}$, a balanced metric that measures whether justifications enable raters to accurately assess answer correctness, validated against human raters who show high agreement. We find that neither common approaches, such as post-training and model scaling, nor more targeted interventions recommended improve verifiability. We introduce two methods that succeed at improving verifiability: reflect-and-rephrase (RR) for mathematical reasoning and oracle-rephrase (OR) for factual QA, both of which improve verifiability by incorporating domain-appropriate external information. Together, our results establish error verifiability as a distinct dimension of response quality that does not emerge from accuracy improvements alone and requires dedicated, domain-aware methods to address.

Diffusion models have emerged as the prevailing approach for text-to-image (T2I) and text-to-video (T2V) generation, yet production platforms must increasingly serve both modalities on shared GPU clusters while meeting stringent latency SLOs. Co-serving such heterogeneous workloads is challenging: T2I and T2V requests exhibit vastly different compute demands, parallelism characteristics, and latency requirements, leading to significant SLO violations in existing serving systems. We present GENSERVE, a co-serving system that leverages the inherent predictability of the diffusion process to optimize serving efficiency. A central insight is that diffusion inference proceeds in discrete, predictable steps and is naturally preemptible at step boundaries, opening a new design space for heterogeneity-aware resource management. GENSERVE introduces step-level resource adaptation through three coordinated mechanisms: intelligent video preemption, elastic sequence parallelism with dynamic batching, and an SLO-aware scheduler that jointly optimizes resource allocation across all concurrent requests. Experimental results show that GENSERVE improves the SLO attainment rate by up to 44% over the strongest baseline across diverse configurations.

Large language model (LLM) decoding is a major inference bottleneck because its low arithmetic intensity makes performance highly sensitive to memory bandwidth. 3D-stacked near-memory processing (NMP) provides substantially higher local memory bandwidth than conventional off-chip interfaces, making it a promising substrate for decode acceleration. However, our analysis shows that this bandwidth advantage also shifts many decode operators on 3D-stacked NMP back into the compute-bound regime. Under the tight area budget of the logic die, the design of the compute substrate itself therefore becomes a first-order challenge. Therefore, we rethink the compute microarchitecture of prior 3D-stacked NMP designs. First, we replace prior MAC tree-based compute units with a more area-efficient systolic array, and we further observe that decode operators exhibit substantial shape diversity, making reconfigurability in both systolic array shape and dataflow essential for sustaining high utilization. Building on this insight, we continue to exploit two key opportunities: the high local memory bandwidth reduces the need for large on-chip buffers, and the existing vector core, originally designed to handle auxiliary tensor computations, already provides much of the control logic and multi-ported buffering required for fine-grained flexibility for systolic array, allowing us to unify the two structures in a highly area-efficient manner. Based on these insights, we present the first compute microarchitecture tailored to 3D-stacked NMP LLM decoding, explicitly designed to satisfy the joint requirements of low area cost, high-bandwidth operation, and fine-grained reconfigurability. We further propose an multi-core scheduling framework. Compared with Stratum, our design achieves an average 2.91x speedup and 2.40x higher energy efficiency across both dense and MoE models.

Parent selection methods are widely used in evolutionary computation to accelerate the optimization process, yet their theoretical benefits are still poorly understood. In this paper, we address this gap by proposing a parent selection strategy for the $(μ+1)$ genetic algorithm (GA) that prioritizes the selection of maximally distant parents for crossover. We show that, with an appropriately chosen population size, the resulting algorithm solves the Jump$_k$ problem in $O(k4^kn\log(n))$ expected time. This bound is significantly smaller than the best known bound of $O(nμ\log(μ)+n\log(n)+n^{k-1})$ for any $(μ+1)$~GA using no explicit diversity-preserving mechanism and a constant crossover probability. To establish this result, we introduce a novel diversity metric that captures both the maximum distance between pairs of individuals in the population and the number of pairs achieving this distance. The main novelty of our analysis is that it relies on crossover as a mechanism for creating and maintaining diversity throughout the run, rather than using crossover only in the final step to combine already diversified individuals. The insights provided by our analysis contribute to a deeper theoretical understanding of the role of crossover in the population dynamics of genetic algorithms.

Non-Hermitian photonics provides a framework to engineer the gain and loss of optical modes in open systems, enabling control of their spectral and dynamical properties. In particular, the ability to dynamically tune modal losses offers a route to implement functionalities traditionally relying on cavity Q-factor modulation, such as Q-switching, within nanophotonic platforms. Here, we demonstrate the generation of short optical pulses in a pair of phase-coupled photonic crystal nanolasers exploiting non-Hermitian coupling. Two waveguide-coupled nanocavities are operated below their individual lasing thresholds and subjected to asymmetric optical pumping, such that a transient carrier-induced detuning modifies the interference conditions between them. This dynamically controls the gain and loss of the collective modes, and, upon crossing a resonance condition, leads to the rapid release of stored carrier energy as an optical pulse. A rate-equation model captures the interplay between carrier dynamics and modal coupling and reproduces the observed behavior. Experiments performed on an indium phosphide platform show pulse generation from cavities that do not lase efficiently on their own in continuous-wave operation, with temporal characteristics governed by carrier dynamics. These results indicate that non-Hermitian coupling can be used to control the effective cavity losses in time, providing a route to pulse generation in integrated photonic systems.

We investigate the thermal properties of $\mathcal{PT}$-symmetric scalar field theories with purely imaginary couplings. The free energy governs the asymptotic density of states, providing an effective measure of the number of degrees of freedom, while thermal masses and one-point functions provide predictions for operator dimensions and three-point functions in the corresponding $d=2$ Conformal Field Theories. Naive finite-temperature perturbation theory near upper critical dimensions is spoiled by infrared divergences. To remove these divergences, we introduce a ''thermal normal-ordering'' scheme that resums these contributions and yields a systematic $ε$-expansion. This framework allows us to compute the free energy, thermal masses, and one-point functions in the cubic and quintic $O(N)$ models. We compare the thermal free energy density, thermal masses, and one-point function in two dimensions with exact results derived from the proposed Ginzburg-Landau descriptions of the non-unitary minimal models $M(2,5)$ and $M(3,8)_D$. Eventually, we employ two-sided Padé extrapolations to obtain estimates for the thermal free energy in $d=3,4,5$.

The phenomenon of electron tunneling has long been central to quantum transport and continues to provide a powerful framework for understanding and controlling electronic processes in solids. When combined with optical excitation, tunneling becomes a particularly rich platform for experiments, because light can drive nonequilibrium carrier populations and open transport pathways that are inaccessible without optical excitation. The emergence of van der Waals (vdW) materials has greatly expanded this opportunity by enabling atomically thin heterostructures with clean interfaces, engineered barriers, and highly tunable band alignment. In this review, we discuss the fundamental mechanisms of photo-assisted transport and the realization of vdW tunnel junctions, and show how they provide electrical access to nonequilibrium dynamics and collective excitations in quantum materials. We further examine emerging functionalities including photodetection, tunneling-driven light emission, sensing, and memory. Finally, we present a forward-looking perspective on new opportunities such as quantum-geometric probes, twist-resolved spectroscopy, moire ferroelectricity, and scalable architectures for computing and sensing.

Renormalization of complex networks requires principled criteria for assessing whether a coarse-graining preserves dynamical content. We prove that discrete harmonic morphisms -- surjective maps preserving harmonic functions -- provide the minimal condition under which random walks on a fine-grained network project exactly onto random walks on its coarse-grained image, through an appropriate random time change. We formalize this via the harmonic degree, a diagnostic quantifying how closely any network coarse-graining approximates a harmonic morphism. Applying this framework to geometric, Laplacian, and GNN-based renormalization across real-world networks, we find that each method produces a distinct dynamical fingerprint encoding its underlying physical assumptions. Most strikingly, Laplacian renormalization spontaneously yields exact harmonic morphisms in several networks, achieving exact preservation of first-exit random-walk transition structure at specific scales, a property that entropic susceptibility fails to detect. Our results identify a discrete analog of diffusion-preserving conformal maps for irregular network topologies and provide quantitative tools for designing and evaluating multi-scale network descriptions.

Many-body exciton interactions shape the optoelectronic response of atomically-thin transition metal dichalcogenides, yet optical control of these interactions remains largely unexplored. To date, modulation of exciton-exciton interactions has primarily relied on electrical gating or van der Waals engineering. Here, we demonstrate all-optical control of many-body exciton interactions in monolayer WSe$_2$ via valley-selective excitation using polarization-resolved pulsed-laser photocurrent spectroscopy. Circular excitation selectively populates excitons in a single valley, whereas linear excitation populates both valleys, inducing a valley-dependent nonlinear photoresponse. We observe helicity-dependent exciton renormalization, alongside a two-fold enhancement of sublinear photocurrent scaling under circular excitation, reflecting single-valley population of interacting excitons. A microscopic model incorporating intervalley-exchange and exciton-exciton annihilation mediated by dark and bright exciton populations reproduces the nonlinear valley-selective response. These results establish the valley degree of freedom as an all-optical control parameter for tuning many-body excitonic effects and, exploring correlated exciton states and valleytronic applications in two-dimensional semiconductors.

Langevin stochastic differential equations provide a dynamical description of pulsar timing noise and gravitational wave background (GWB) signals. They are also central to state space algorithms that have gained traction in pulsar timing array analysis due to their linear computational scaling with the number of observations. In this work, we utilize established methods in diffusion theory to derive analytical solutions (means, covariances, and probability density functions) to Langevin equations relevant to red noise and the GWB signal in pulsars. The solutions give direct physical insight on the dynamics of pulsar timing signals. As a canonical example, we show that the pulsar spin frequency modeled as an Ornstein-Uhlenbeck process is mathematically inconsistent with a stationary GWB signal when the timing residual is the direct observable. The nonstationarity can be partially dealt with by marginalizing over long time deterministic trends in the data. Then, we show that a random process based on an overdamped harmonic oscillator supports both a stationary spin frequency and phase residuals, consistent with a stationary GWB signal. We also turn our attention to a phenomenological model of a neutron star -- a two-component model with spin wandering -- that has been motivated to explain observed timing noise in radio pulsars. We derive analytical expressions for the means, covariances, and cross-covariances of the crust and superfluid rotational states driven by white noise. The associated constant deterministic torques are linked to the quadratic spin-down of pulsars. The solutions reveal the physical origin of nonstationarity in the residual model: the coexistence of damped and diffusive eigenmodes of the system.

Liquid bridges form between particles during wet mixing with binders or by condensation due to ambient humidity. The consequences of capillary bridges can be quite drastic, creating macroscopic cohesion, as seen in sandcastles and in the formation of particulate agglomerates. Bulk effects in cohesive particles arise from forces generated by capillary bridges, so particle-scale measurements are needed to develop predictive models. Most existing studies at the particle scale assume Newtonian liquids. Yet many binders in industry and in the environment can exhibit viscoelastic behavior. In this study, we measure the axial force generated by liquid bridges of viscoelastic polymer solutions between two spherical beads during controlled uniaxial separation. We vary the polymer concentration, separation velocity, and particle size, and track the force as the bridge thins and ruptures. At quasi-static rates, the axial force remains dominated by capillarity and is not significantly affected by polymer rheology. However, increasing the stretching rate increases the peak force through viscous dissipation and promotes the formation of a viscoelastic filament, thereby delaying rupture. The peak axial forces collapse when rescaled by a capillary number and particle size, while the effective rupture distance collapses with a Weissenberg number. These results provide a simple first-order particle-scale force law for polymeric binders.

We present a stochastic mean-reverting jump-diffusion model to simulate rainfall time series and validate it using long-term half-hourly rain fall data from the North-East region of India. The model captures the intermittent and extreme-event dynamics of rainfall, reproducing superdiffusive behavior with an exponent $\sim 1.8$, along with the observed probability distributions and multifractal features. By systematically varying key parameters, we demonstrate a transition between Log-Normal and Gamma distributions, and show how the occurrence of extreme events and dry-patch durations can be controlled. Spectral and wavelet analyses further confirm that the simulated series reproduces the dominant temporal scales observed in real rainfall data. Our proposed framework provides a robust tool for generating realistic synthetic rainfall series and serves as an effective approach for understanding the influence of underlying stochastic processes that governs the rainfall statistics.

Generalized Langevin equations (GLEs) provide a powerful framework for describing slow dynamics in soft-matter systems, but deriving an exact homogeneous GLE (hGLE) for a reaction coordinate from an underlying many-body system remains generally difficult. Here, we analytically derive an exact hGLE for the relative coordinate of two tagged beads in arbitrary elastic networks. The memory kernel and effective restoring force are expressed explicitly in terms of the network matrices, thereby providing a systematic reduction of the high-dimensional network dynamics to a pair coordinate. Within the small-displacement approximation, we further derive a hGLE for the inter-bead distance, a central observable in distance-sensitive single-molecule experiments. These results therefore have broad potential applications in modeling proteins and other soft-matter systems.

Cells control the size and organization of biomolecular condensates formed by liquid-liquid phase separation (LLPS), but multiple mechanisms likely contribute to this control and remain to be fully elucidated. Here we propose a transport-driven mechanism in which stochastic binding of phase-separating proteins to cytoskeletal motor proteins, followed by active redistribution along filament networks, generates an effective long-range repulsion that arrests coarsening and selects a finite condensate size. A minimal diffusion-transport model, analyzed by linear stability theory and three-dimensional simulations, reveals a transition from macroscopic to microphase separation at remarkably low binding/release fractions, corresponding to minute motor-bound populations. Tuning motor binding rates $b$ or transport velocities enables sublinear control of condensate sizes ($L \sim b^{-1/4}$) from nanometers to micrometers. In anisotropic cytoskeletal environments, transport asymmetry drives morphological transitions from spherical to cylindrical condensates. Operating independently of thermodynamic parameters, this mechanism provides a versatile, spatiotemporally programmable route to condensate organization and informs the design of synthetic active emulsions with tunable architectures.

In the field of orbital dynamics and orbital transport, a particularly important quantity is the so-called orbital Hall conductivity (OHC), which is expressed in terms of operators of velocity and orbital angular momentum (OAM). To overcome the difficulties in treating the unbounded position operator, very often atom-centered approximations are used, which capture only a part of the local contributions to the OAM operator. Here, we promote a new approach to quantify the OAM operator in the basis of Wannier functions, which is based on the modern theory of orbital magnetization and which captures both local and itinerant contributions to the OHC. By performing first-principles calculations for various materials, we show that significant corrections to the OHC by non-local effects arise when compared to common approximations. Our approach improves the understanding of the OAM in solids and allows for a precise estimation of various orbital effects in complex materials.

Active colloidal particles provide versatile model systems for exploring non-equilibrium physics in motile matter. To date, most experimental realizations have focused on spherical particles, largely due to fabrication constraints. However, theoretical and computational studies have long predicted that particle anisotropy and flexibility can dramatically enrich single-particle dynamics, interparticle interactions, and emergent collective behavior. Here, we highlight recent advances in the fabrication of anisotropic active particles and architectures enabled by the unprecedented design freedom of 3D microprinting. We discuss how additive manufacturing is expanding the accessible parameter space of active soft matter, allowing precise control over shape, location of active forces, and functionality at the microscale. These developments establish new model platforms for uncovering fundamental principles of active and soft matter, and simultaneously pave the way toward microrobotic systems with programmable dynamics and emergent functionalities.

Achieving long-range ferrimagnetic order in purely organic systems remains a major challenge in molecular magnetism. Here we report the synthesis and characterization of heterospin-coupling motifs, formed by covalently linking spin-1/2 and spin-1 triangular nanographenes. A combined solution-phase and on-surface synthetic strategy yields three distinct compounds, whose structures are elucidated by bond-resolved scanning probe microscopy. Starting from a spin-1/2--spin-1 dimer as the elemental ferrimagnetic unit, we employ inelastic electron tunneling spectroscopy to resolve low-energy magnetic excitations and extract the parameters of the Heisenberg Hamiltonian. Extension to trimeric architectures results in two distinct spin configurations, with compensated ($S=0$) and uncompensated ($S=3/2$) ferrimagnetic ground states. The Heisenberg model accurately describes all magnetic transitions, offering direct insight into increasingly complex spin Hamiltonians. These findings establish a molecular platform for designing tunable heterospin systems with robust exchange interactions, opening routes toward multi-level spin encoding in qudit-based quantum technologies.

This study presents a rigorous comparative analysis of two label-free optical biosensing platforms, Bloch surface wave (BSW) and microring resonator (MRR), for the detection of SARS-CoV-2 antibodies in human serum. To ensure direct comparability, a new BSW readout system was established alongside an existing MRR platform, allowing assays to be conducted under nearly identical experimental conditions. Both sensors were functionalized with various SARS-CoV-2 Spike and Nucleocapsid protein variants to capture specific host antibodies. The results demonstrate that both platforms provide rapid, quantitative, and sensitive detection of anti-Spike and anti-Nucleocapsid antibodies without the need for secondary labels. Furthermore, the platforms show excellent agreement with longitudinal serology benchmarks and high repeatability across different biochip batches. This work establishes both BSW and MRR technologies as powerful, low-cost candidates for next-generation clinical diagnostics and serological surveillance.

We investigate work extraction in open quantum batteries composed of interacting spin chains weakly coupled to engineered environments. Focusing on two- and four-qubit XX models initially prepared in thermal Gibbs states, we analyze how dissipation and dephasing, acting either locally or collectively, can generate and shape ergotropy during both transient and steady-state dynamics. By introducing a continuous interpolation between parallel and collective noise channels, we systematically characterize the impact of environmental structure on work extractability. We show that purely dissipative dynamics can activate finite ergotropy from completely passive thermal states, giving rise to temperature-dependent transient regimes where hotter initial states temporarily outperform colder ones in an ergotropic Mpemba-like fashion. In contrast, collective dissipation leads to steady states whose passivity crucially depends on the initial temperature and system size, a behavior we trace back to the emergence of non-trivial dark subspaces. Finally, we demonstrate that dephasing channels suppress both transient advantages and steady-state work extraction, highlighting the qualitative difference between dissipative and dephasing environments.

We introduce Gradient Equilibrium Propagation (GradEP), a mechanism that extends Equilibrium Propagation (EP) to train energy gradients rather than energy minima, enabling EP to be applied to tasks where the learning objective depends on the velocity field of a convergent dynamical system. Instead of fixing the input during dynamics as in standard EP, GradEP introduces a spring potential that allows all units, including the visible units, to evolve, encoding the learned velocity in the equilibrium displacement. The spring and resulting nudge terms are both purely quadratic, preserving EP's hardware plausibility for neuromorphic implementation. As a first demonstration, we apply GradEP to flow matching for generative modelling - an approach we call FlowEqProp - training a two-hidden-layer MLP (24,896 parameters) on the Optical Recognition of Handwritten Digits dataset using only local equilibrium measurements and no backpropagation. The model generates recognisable digit samples across all ten classes with stable training dynamics. We further show that the time-independent energy landscape enables extended generation beyond the training horizon, producing sharper samples through additional inference-time computation - a property that maps naturally onto neuromorphic hardware, where longer relaxation yields higher-quality outputs. To our knowledge, this is the first demonstration of EP training a flow-based generative model.

We establish an exact analytical relation between Zitterbewegung dynamics and the band geometry in two-dimensional Dirac systems. By identifying a time-independent antisymmetric observable-the \textit{areal rate of Zitterbewegung}-we show that this quantity is directly determined by the Berry curvature. Its sign defines the sense of rotation and reproduces the contributions of Dirac points to the Chern number. This relation is independent of the initial state and holds for generic two-band Dirac models. Our findings reveal a direct connection between interband quantum dynamics and topological band geometry beyond the semiclassical regime.

Alignment interactions in active matter are typically modeled as relaxational dynamics toward local consensus. In unbounded systems, this makes alignment effectively decoupled from local density and therefore unable to sustain self-confined collective motion without additional attractive forces. Here we show that this limitation can be overcome by extremal alignment rules in which the interaction neighborhood depends on the candidate orientation. For a broad class of candidate- dependent rules with pairwise additive utilities, the decision utility factorizes into the product of an average interaction score and the number of selected neighbors. This multiplicative structure couples orientational decisions to local density and thereby generates an effective cohesive bias without explicit cohesive forces. In metric models, however, the same mechanism drives collapse toward globally connected, effectively mean-field states that suppress spatial structure. We show that topological interactions regularize this tendency, stabilizing self-confined flocks of finite extent in open space. The resulting dynamics exhibits a rich dynamical phase diagram as a function of noise intensity and turning rate, including polarized flocks, swarms, and persistent swirling states. Our results identify candidate-dependent extremal alignment as a simple mechanism for generating cohesive, spatially structured active matter beyond the standard relaxational paradigm.

We investigate the Jahn-Teller structural phase transition in LaMnO$_3$ at $T_{JT} \simeq 750$ K using molecular dynamics simulations based on machine-learning force fields trained on ab initio data. Analysis of the site-site correlation function of the distortions reveals that the transition is driven by the ordering of the $Q_2$ Jahn-Teller distortion of the MnO$_6$ octahedra, which acts as the order parameter and establishes the order-disorder nature of the transition. Dynamical local distortions are found to persist above $T_{JT}$. Our results reproduce the experimental temperature dependence of both structural and phonon properties and highlight the presence of anharmonic effects at finite temperature. More broadly, the combined use of machine-learning molecular dynamics and velocity autocorrelation function analysis provides a robust framework for uncovering the microscopic mechanisms of structural phase transitions in correlated materials. In particular, this approach enables a clear distinction between order-disorder transitions and alternative mechanisms, such as displacive behavior, through the temperature evolution of vibrational properties.

Optical spectroscopy provides a powerful, contact-free probe of topological quantum states, yet exact constraints on antisymmetric Hall absorption remain much less well developed than their longitudinal counterparts. Motivated by earlier Hall-conductivity sum rules, we formulate the corresponding first-frequency-moment constraint for the antisymmetric optical conductivity, whose imaginary part governs chirality-dependent absorption. We then demonstrate this sum rule in two classes of time-reversal-breaking topological systems. For a zero-field moiré continuum model hosting topological bands, the moment vanishes exactly, implying that any low-frequency anomalous Hall absorption must be compensated by higher-frequency spectral weight of the opposite sign. For a Hofstadter model under a uniform magnetic field, the same moment takes a universal value fixed by the magnetic flux density, independent of microscopic model details. By linking low- and high-frequency spectral contributions, this optical Hall absorption sum rule provides a rigorous framework for quantifying circular dichroism constraints and diagnosing Landau-level mixing. Our results show how a known Hall spectral constraint acquires new and experimentally relevant content in modern interacting topological materials.

Flow through porous media can reshape the medium through erosion and deposition, producing preferential flow channels across a wide range of natural and industrial systems. Yet the mechanisms by which spatial disorder triggers channelization remain unclear. Here we derive a continuum description for the coupled evolution of flow and porosity by coarse-graining pore-scale dynamics and validating the resulting model with pore-scale simulations. Using this framework, we show that different sources of disorder lead to qualitatively distinct behaviors. Disorder in erosion resistance produces a discontinuous transition to localized flow, with permanent channels appearing only above a finite disorder strength. In contrast, even extremely weak fluctuations in the initial porosity destabilize homogeneous flow and trigger persistent channelization. These results reveal an unexpected sensitivity of evolving porous media to structural heterogeneity, suggesting that channelization can arise generically even in nearly uniform materials.

Under ambient humidity, water spontaneously condenses in pores only a few nanometers in size, making nanoscale capillarity central to numerous natural phenomena and technological applications. At these dimensions, water may no longer be treated as a continuous fluid, yet the consequences of molecular discreteness for capillary condensation and filling remain poorly understood. Here we study nanocapillaries fabricated by van der Waals assembly and, using atomic force microscopy, monitor their wall deformations during humidity-driven water uptake. We observe two distinct regimes: layer-by-layer filling of flexible capillaries and abrupt filling of rigid ones. Flexible walls deform in steps of ~3 Å, corresponding to the sequential entry of individual water molecular layers. The different filling regimes are explained by the competition between deformation energy and oscillatory wall-water interactions. Our findings show that the molecular discreteness of water can profoundly affect ubiquitous capillary phenomena, with wall compliance selecting between discrete and abrupt filling.

Autoregressive generative models -- including Transformers, recurrent neural networks, classical Kalman filters, state space models, and Mamba -- all generate sequences by sampling each output from a deterministic summary of the past, producing genuinely non-Markovian observed processes. We develop a general theoretical framework based on stochastic thermodynamics for this class of architectures and introduce the entropy production, which can be efficiently estimated from sampled trajectories without exponential sampling cost despite the non-Markovian nature of the observed dynamics. As a proof-of-concept experiment for a large language model (LLM), we evaluate the token-level and sentence-level entropy production for a pre-trained Transformer-based model, GPT-2. We also demonstrate the framework in the linear Gaussian case, where the model reduces to the Kalman innovation representation and the entropy production admits an analytical expression. We further show that the entropy production decomposes exactly into non-negative per-step contributions in terms of retrospective inference, and each of those terms further splits into information-theoretically meaningful terms: a compression loss and a model mismatch. Our results establish a bridge between stochastic thermodynamics and modern generative models, and provide a starting point for quantifying irreversibility in a broad class of highly non-Markovian processes such as LLMs.

The interplay between distinct carrier species in systems with broken Galilean invariance gives rise to a rich landscape of interaction-driven transport phenomena. Here, we develop a comprehensive theory for the electrical conductivity of a non-degenerate two-dimensional mixture of massless Dirac and massive fermions, a system realized in HgTe quantum wells tuned to the charge neutrality point. In this regime, all carriers are thermally activated, enabling a self-consistent, temperature-dependent interplay between the two species. We demonstrate that the conductivity undergoes a distinct crossover as temperature increases: at low temperatures, transport is dominated by massless Dirac carriers, yielding a temperature-independent conductivity reminiscent of graphene's charge neutrality point. As the temperature rises, massive holes become thermally excited, and their mutual Coulomb scattering with Dirac carriers induces a negative, non-Drude correction to the conductivity. We show that this correction is governed by the dominant scattering mechanism: short-range interparticle interactions yield a stronger suppression than long-range Coulomb interactions, and it scales monotonically with temperature. Crucially, the charge neutrality condition ensures that the chemical potential is not externally pinned but is determined self-consistently, making the system's transport response an intrinsic probe of inter-species quantum friction. Our findings establish HgTe quantum wells at charge neutrality as a clean, highly tunable platform for isolating and quantitatively studying interaction-driven transport in the absence of Galilean invariance, offering a direct pathway to explore regimes where interparticle collisions dominate over disorder.

Aging refers to the evolution of system properties with waiting time $t_w$. It is a key feature of glassy dynamics. Recent experiments have demonstrated aging in biological systems that are inherently active with a magnitude of self-propulsion force $f_0$ and a persistence time $τ_p$. Thus, what governs the aging dynamics in these active systems has fundamental importance. We formulate a generic mode-coupling theory (MCT) of active glasses to address this question. The aging solutions of the theory show that the two-point correlation function decays more slowly with growing $t_w$, and the relaxation time $t_r$ increases. The activity-modification of the MCT critical point, $λ_\text{C}$, has profound significance for active aging: the quench distance from $λ_\text{C}$ governs aging and determines $δ$, where $t_r\sim t_w^δ$. $δ$ decreases with increasing $f_0$, in agreement with existing simulations. However, the variation with $τ_p$ depends on the nature of activity. Our work has fundamental theoretical implications for active glasses and paves the way for a deeper understanding of the aging dynamics in biological systems.

The divergence of the electron density of states (DOS) plays an important role in enhancing many-body interactions and inducing various quantum phases in low-dimensional systems. However, such unique electronic structures remain experimentally elusive in three-dimensional (3D) systems, particularly those with strong spin-orbit coupling (SOC). Using angle-resolved photoemission spectroscopy and first-principles calculations for a Laves-phase superconductor CsBi2, which features a Bi-pyrochlore 3D network with strong SOC, we identify two characteristic electronic structures with a large DOS. One is a dispersionless topological flat band with p-orbital character, locally formed around the U-K line, which enhances DOS near the Fermi level. The other involves type-I and type-II saddle points connected by a flat band, which cooperatively produce an enhancement in the DOS. Our findings suggest a novel mechanism for achieving a DOS enhancement and lay a foundation for exploring exotic phenomena driven by the interplay of multiple singularities with a large DOS, nontrivial topology, and strong SOC in 3D pyrochlores.

Immotile microbes inhabit nearly every environment on Earth, from soils and sediments to food matrices -- yet how they disperse through these physically confining environments is poorly understood. Here, we show that immotile microbial colonies confined in a model transparent yield-stress matrix can achieve long-range dispersal by harnessing their own metabolism. Using yeast as a model organism, we find that fermentation drives dissolved CO$_2$ to supersaturation, nucleating biogenic bubbles that grow, yield the matrix, and rise, hydrodynamically entraining cells vertically in their wake. Sequential bubble nucleation sculpts persistent columnar colonies extending far beyond what growth alone permits. Multiple colonies interact via their fermentation byproducts, merging and mixing genetically as they collectively sculpt self-sustaining conduit networks. Our findings reveal a third mode of microbial dispersal, distinct from the canonical mechanisms of motility and growth, with implications for ecology, environmental science, and biotechnology. More broadly, they exemplify a previously unrecognized class of active behavior -- Metabolically Driven Active Matter -- in which metabolic byproducts reshape the physical landscape of confinement to drive population-scale motion.

Our study connects the physics of disordered integer-dimensional systems and regular self-similar objects by studying spectral properties of fractal agglomerates with tunable dimension. The latter is controlled by parameter $α$ of the algorithm that generates the agglomerates. We consider the nearest-neighbor tight-binding model on the agglomerates embedded in 2D and 3D, and observe that all eigenstates are localized in the 2D case, whereas in the 3D case, there is a localization--non-ergodic transition upon increasing $α$,i.e., going from sparse to dense fractals: a sub-extensive number of critical states emerge in the spectrum at a certain critical value of $α$. The complex geometry of the agglomerates is also responsible for a peculiar hierarchy of compact localized states and singularities in the density of states, which are typical for ordered fractals.

Belnap's four-valued logic, distinguished by its inherent bilattice structure, provides a natural algebraic bridge between discrete Four-valued logic (4VL) in circuit and continuous catastrophe theory (CT). Building on the rigorous verification of the bilattice-catastrophe isomorphism theorem, we establish a categorical correspondence spanning the catastrophe category, interlaced bilattice category, and 4VL category, with the cusp catastrophe emerging as the canonical CT counterpart to 4VL.This unification provides a foundational framework for explaining 4VL's robustness. Crucially, we demonstrate that the four-valued algebra FOUR is the minimal complete algebraic structure capable of describing continuous-discrete interfaces with involution symmetry. Unlike the empirical adoption of X and Z in engineering practice, our work reveals their mathematical necessity: X and Z are topological invariants of discretized continuous dynamical systems, encoding fundamental properties of catastrophe-induced discontinuities. The work enables cross-disciplinary extensions to uncertainty propagation, complex system modeling, and fault-tolerant design.

Using large-scale molecular dynamics simulations, we investigate the scaling of morphological and contact mechanics properties of Al and Cu nanoparticles (NPs) physisorbed on suspended graphene. The characteristic linear size of a NP ranges from 1 nm to 49 nm, covering a length scale of 1.5 decades. The NPs were obtained using a procedure mimicking thermal dewetting of thin films. Calculations show that NPs with a surface area-to-volume ratio above about 1.8 nm$^{-1}$, or with a linear size under 3-6 nm, behave differently from larger particles. For these smaller NPs, scaling of their total surface area and volume with the linear size can deviate from quadratic and cubic dependencies, respectively. Their mean interfacial separation and relative contact area change rapidly with size, exhibiting substantial variation. In contrast, for larger NPs, these quantities approach the thermodynamic limit. The height distributions of all particles exhibit a narrow spike and a decaying tail, both of which can be fit to Gaussians for larger NPs. In contrast, the interfacial gap distributions are close to a single Gaussian. The height power spectrum density (PSD) heatmaps of the smaller NPs are smeared and do not manifest a clear structure in contrast to the sixfold symmetry of the PSD of the larger ones. The maximum spatial frequency of the hexagonal 2D PSD roughly corresponds to the nearest-neighbor atomic distance of Al and Cu. For larger NPs with diameters of 20-25 nm, the isotropic height PSD exhibits power-law regions, which can be interpreted as self-affine roughness with Hurst exponents of 0.1-0.56. We also calculate the relative difference between the apparent contact area and the approximated area of the bottom atomic layer. Our simulations illustrate how surface topography evolves with NP size and suggest that larger NPs can have random surface roughness.

A self-consistent theory of bulk electrolytes incorporating electrostatic and hard-core interactions on an equal level is applied to the two-dimensional Coulomb liquid with finite ion size. The ionic pair distributions, the structure factors, and the thermodynamic functions of the formalism are compared with extensive Monte-Carlo simulation results from the literature. At moderate salt densities, our computational approach can accurately describe the thermodynamics of two-dimensional solutions from weak to intermediate coupling strengths. The improved accuracy of the present theory with respect to continuum approaches stems mainly from its ability to account for the non-uniform screening of electrostatic interactions associated with the impenetrability of the charged hard disks by their ionic atmosphere. At low salt densities, the validity domain of our self-consistent framework underestimating the extent of ionic cluster formation drops below the critical coupling domain where the conductor-insulator transition of two-dimensional charged hard disks occurs. This indicates that approaching the low-temperature dielectric phase via the present formalism will require the extension of the underlying self-consistent approximation at least up to the next cumulant order.

Dynamic control of visible light is crucial for technologies such as holographic displays and adaptive optics. Passive metasurfaces can shape wavefronts at the subwavelength scale and active metasurfaces promise to extend this functionality into the temporal domain. However, existing metasurfaces for dynamic phase manipulation typically cannot deliver phase modulation across a broad range without causing variations in the scattering amplitude. Here, we use an inverse-design pipeline to numerically demonstrate a hybrid-2D excitonic metasurface platform offering independent amplitude and phase control in the visible regime. Harnessing the gate-tunable excitonic response of monolayer WS2 retrieved from experiments, we design a pi-phase modulator with a uniform amplitude profile. Adding a second tunable monolayer, we achieve independent control of the amplitude and phase over the full 0-2pi phase range, which we leverage for a reconfigurable beam-steering metadevice. Our results demonstrate how hybrid-2D excitonic metasurfaces enable electrically tunable wavefront shaping in the visible regime.

Deformations of skyrmions arising from the presence of more than one magnetic sublattice lead to their massive dynamics in ferrimagnets as compared to the massless dynamics in 2D ferromagnets. This results in the gyroscopic motion of skyrmions, which manifests as skyrmion cyclotron resonance that can be excited by microwaves or spin currents. We investigate analytically and numerically the motion and resonant oscillations of individual skyrmions and skyrmion lattices in the presence of dissipation in a two-sublattice transition-metal -- rare-earth (TM/RE) system. The focus is on the dependence of the skyrmion dynamics on the RE concentration. Parameters of the CoGd ferrimagnet are utilized in the numerical work. The massive dynamics of skyrmions in ferrimagnets, as well as the spectrum of their excitations, undergo significant changes near the angular momentum compensation point, which should not be difficult to detect in experiments.

High-harmonic generation in solids has emerged as a powerful probe of ultrafast electron dynamics and lattice motion, and recent theoretical work has suggested that thermally driven lattice fluctuations can act as an effective source of decoherence in the harmonic-generation process. However, a direct experimental link between high-harmonic emission and temperature-driven incoherent phonons has remained unclear. Here, we investigate the temperature dependence of high-harmonic generation in ultrapure silicon using reflection-geometry measurements over a wide temperature range. We observe that the harmonic yield increases significantly with decreasing temperature. To interpret these results, we introduce a one-dimensional atomic-chain model in which finite temperature is represented by random lattice displacements that mimic incoherent phonon fluctuations. The simulations reproduce the magnitude of temperature-dependent change of the harmonic signal and support a picture in which thermally induced lattice disorder enhances electron-hole decoherence, thereby reducing high-harmonic emission. Our results establish incoherent phonons as an important source of decoherence in solid-state high-harmonic generation.

The electrophoresis of charged colloidal particles in fluids exhibiting odd viscosity represents a fundamental challenge in understanding transport phenomena within charge-stabilized chiral active suspensions. Here, we provide the first concept of a charged chiral active fluid, where electrokinetics is coupled to odd Stokes flow, to explore how classical results from electrophoresis in Newtonian fluids generalize in the presence of odd viscosity. In particular, we derive a general expression for the electrophoretic mobility for particles of any shape under weak external electric fields using the Lorentz reciprocal theorem for odd fluids. By applying this result to a conducting charged sphere at low zeta potentials, we obtain an exact, closed-form analytical expression for the electrophoretic mobility, valid for arbitrary values of the Debye screening length and the odd-viscosity coefficient. Similar to Newtonian fluids, we find that the electrophoretic mobility is proportional to the translational mobility of an uncharged sphere, modulated by the Henry function. However, unlike in Newtonian fluids, odd viscosity leads to directional asymmetries in the electrophoretic mobility tensor that persist even for thin electric double layers. This case contrasts significantly with a charged anisotropic particle suspended in an isotropic Newtonian fluid, where anisotropic effects would vanish under the same electrostatic-screening conditions.

We study ballistic transport in bilayer graphene junctions and show how electrostatic gating, interlayer bias, and homogeneous strain provide complementary control over electron transmission. In the absence of strain, transport is governed by symmetry constraints that suppress transmission at specific incidence angles despite the availability of states. An interlayer bias lifts this suppression through mode mixing and opens a tunable transport gap. Within a full four-band description, we identify a distinct conductance threshold that marks the onset of propagation of the upper band inside the barrier. This produces a clear change in the slope of the conductance and serves as an experimentally accessible transport fingerprint of the multiband structure and interlayer coupling. Homogeneous in-plane strain acts as a geometric control mechanism. By reshaping the band structure in momentum space, it redistributes the angular transmission window and suppresses conductance without introducing disorder. Importantly, strain preserves the underlying symmetry-based decoupling responsible for transmission suppression while shifting its condition away from normal incidence. These results provide a unified framework for interpreting angle-resolved transport in bilayer graphene and establish multiband ballistic transport as a practical probe of band-structure geometry.

We have simulated a one dimensional capillary fiber bundle model with fracking events while acted between a pressure gradient across the system. The hydraulic fractures are incorporated through a decreasing nature of capillary thresholds for each tube that replicates an increment in pore spaces due to fracking. An increment in flow rate is evident through the evolved rheology we observe in our study. Analytical approaches for certain limits are adopted to understand the rheology which matches well with the numerical results. The overall hydraulic power increases with pressure gradient as well as with the percentage decrease in capillary threshold due to a single event, defines as the fracking amplitude. This combined with the early onset of linear Darcy flow increases the quality of the fluid extraction. We successfully point towards an optimum pressure gradient at which the fracking events are most effective - maximum change in fluid extracting with a maximum rate. We observed that it is possible to extract the information regarding the change from non-linear to Darcy flow due to fracking as well as the optimum pressure for fluid extraction through local flow profile, something which in much superior from the point of view of computational cost. The former is done by correlating the maximum fluctuation in local flow profile to the onset of Darcy flow. The later is done through the relative change in Shannon entropy with respect to the fracking amplitude that points towards the pressure associated with the maximum fluid extraction criterion.

We develop a model of musical rhythm and meter based on optimizing the trade-off between human psychological preferences for perceiving repeated patterns in time with a desire for variety and complexity. By mapping these competing preferences onto analogous quantities in statistical physics, we define an effective free energy which is minimized in the grand canonical ensemble. Using a mean field approximation, we observe phase transitions in the model from disordered events in time to orderings that closely reproduce those seen in music. We then compare the range of rhythmic characteristics predicted by the model to a dataset drawn from compositions by Johann Sebastian Bach, finding generally good quantitative agreement. The results provide a new lens through which to study musical rhythm, and a method for generatively producing rhythms.

The Su-Schrieffer-Heeger (SSH) model describes a tight-binding one-dimensional (1D) lattice with alternating nearest-neighbor amplitudes. Despite its mathematically simple and physically intuitive structure, the SSH model is capable of supporting a 1D topological phase that is characterized by the presence of zero energy eigenstates (zero modes) localized at each end of the lattice. For this reason, many studies in the area of topological phases of matter often consider the SSH model as a subject for various extensions that give rise to more sophisticated topological phenomena. The purpose of this article is to review, in sufficient detail, existing approaches to extending the SSH model. This includes extensions by increasing the dimensionality of the lattice, enlarging the size of its unit cell, or adding extra terms that represent various physical effects. For each approach, some extended SSH models studied in relevant existing literature are discussed as case studies. Noteworthy properties of such models, which are of topological origin, are further comprehensively elaborated.

Periodically driven quantum systems can exhibit subharmonic response, usually characterized through physical observables and often discussed in interacting settings. Here we show that a sharp subharmonic signature already appears in the entanglement spectrum of a number-conserving free-fermion system. We study a two-step driven Su-Schrieffer-Heeger chain whose Floquet operator supports symmetry-protected edge modes at quasienergies $0$ and $π$. When the initial state is a coherent superposition of these two edge sectors, the subsystem correlation matrix alternates between two stroboscopic structures, and an overlap-tracked single-particle entanglement level distills a robust period-doubling response with Fourier weight concentrated at half the drive frequency. By contrast, diagonal edge densities remain flat by sublattice symmetry, while an off-diagonal edge-bond observable provides the corresponding linear one-body comparator. The effect disappears both when the initial state is replaced by a stroboscopically stationary Floquet eigenstate built from the same topological mode content, and when the system is placed in the topologically trivial phase where no edge modes exist. Altogether, these establish zero-$π$ Floquet topology as a necessary condition and coherent nonequilibrium preparation as the additional sufficient ingredient. Our results identify entanglement spectroscopy as a sharp subsystem-resolved probe of Floquet topological coherence.

Reservoir engineering has emerged as a powerful paradigm to realize non-reciprocal dynamics in open quantum many-body systems. Here, we show that density-density interactions can transfer bath-induced non-reciprocity between different degrees of freedom. Specifically, we investigate a one-dimensional lattice of spin-$\frac{1}{2}$ fermions with all-to-all Hatsugai-Kohmoto interactions in the presence of an engineered reservoir. We establish the exact solvability of the Lindbladian dynamics and show that the interplay between non-reciprocity and interactions qualitatively reshapes the dynamics of excitations. Remarkably, interactions induce directional drift even in spin sectors that are not directly coupled to the reservoir. By analyzing a driven-dissipative Fermi-Hubbard chain, we show that the same mechanism persists for local interactions. The Hatsugai-Kohmoto model thus emerges as a minimal, exactly solvable platform for interaction-mediated non-reciprocal many-body dynamics.

MnBi$_2$Te$_4$ (MBT) thin films provide a unique material platform in which magnetism, topology, and magneto-optical (MO) response can be tuned through layer-thickness and relative spin alignments. In this work, using a low-energy coupled Dirac cone model together with Wannier-based tight-binding Hamiltonian derived from \textit{ab-initio} calculations, we investigate topological MO switching in even-layered MBT films. We argue that the relative spin alignment of the outermost septuple-layers (SL) mainly controls the total Chern number, optical conductibility, and consequently, the MO response. For a 6-SL MBT thin film, we found that reversing the outermost-SL alignments from antiparallel to parallel switches the system from axion insulating state with $C=0$ and vanishing Faraday rotation to a Chern insulating state with $C=1$ and a quantized MO response, irrespective of $PT$-symmetry and net magnetization. Increasing thickness reveals an additional regime: while 8-SL MBT hosts only $C=0$ and $1$ states, a 12-SL MBT film supports a higher Chern number phase with $C=2$ with a doubled low-frequency Faraday rotation. Our results provide a thickness-dependent route to multilevel MO switching and establish MO spectroscopy as a direct probe of surface magnetism and topological order in MBT thin films.

The design of polarization-encoded quantum interfaces relies on the assumption that solid-state emitters possess static transition dipoles defined by the host lattice symmetry. Here, we demonstrate the vibronic breakdown of this static dipole approximation in hexagonal boron nitride quantum emitters. Through high-resolution energy-resolved spectroscopy, we reveal a continuous, spectral rotation of the emission dipole orientation reaching up to 40$^{\circ}$, driven by coupling to the phonon bath. This spectral gradient is significantly suppressed at cryogenic temperatures (6 K), identifying thermally activated lattice vibrations as the primary driver of the dipole reorientation. First-principles calculations on two representative defect types indicate the microscopic origin of this phenomenon as a coordinate-dependent transition dipole, where phonon-induced atomic displacements fundamentally perturb the electronic wavefunctions. By comparing the distinct defect environments, we demonstrate that the magnitude of the polarization rotation scales with the strength of the vibronic coupling. Our results not only identify a fundamental limit for polarization fidelity in solid-state quantum networks but also suggest a new class of strain-tunable quantum photonic devices based on vibronic dipole reorientation.

We construct a model of a three-dimensional helical second-order topological superconductor formed by an array of weakly coupled Rashba nanowires. We identify the parameter regime in which there are energy gaps in both the bulk and surface energy spectra, while a pair of gapless helical $\mathbb{Z}_{2q}$ parafermionic modes (with $q$ being an odd integer) remains gapless along a closed path of one-dimensional hinges. The precise trajectory of these hinge modes is dictated by the hierarchy of interwire couplings and the boundary termination of the sample. In the noninteracting limit $q= 1$, the system hosts gapless Majorana hinge modes.

We take the first steps to address via integrability the spectral problem of AdS/CFT dual pairs deformed by Groenewold-Moyal twists. In particular, we start by considering a twisted spin-chain that couples, through a Groenewold-Moyal twist deformation, two $\mathfrak{sl}(2)$-invariant spin-chains. We interpret this deformed spin-chain as a deformation of a subsector of the $AdS_3/CFT_2$ spin-chain, but the construction shares qualitative features also with the corresponding deformation of the $AdS_5/CFT_4$ spin-chain, for example. As in similar types of deformations, we show that there exists a certain basis in which the spin-chain Hamiltonian takes a Jordan-block form. At the same time, by working in the basis of eigenstates of the generators used to construct the Groenewold-Moyal twist, the Hamiltonian appears to be diagonalisable and with a deformed spectrum. Employing the method of the Baxter equation, we write down the energy of the ground state and of excited states in a perturbation of the deformation parameter. We then consider the string-theory side of the duality, where the twist is realised as a deformation of AdS of the type of Maldacena-Russo-Hashimoto-Itzhaki. We construct a deformation of the usual BMN classical solution, and in the large-$J$ limit we match the leading $\mathcal O(J^{-3})$ term of the energy of the spin-chain groundstate with a conserved charge of the string classical solution. Differently from the undeformed setup as well as similar kinds of deformations, we find that the general expression of this charge of the string sigma-model is non-local, and that it does not correspond to a standard isometry. Nevertheless, it can be computed from the monodromy matrix and it is part of the tower of conserved charges provided by integrability.

We study critical dynamics and phase-ordering kinetics in Active Model B (AMB) and its minimal extension, Active Model B$+$ (AMB$+$), using deterministic simulations in two dimensions. At criticality $r_c=0$, both models display identical mean-field scaling despite nonequilibrium currents, with order-parameter decay with time as $m(t)\sim t^{-α}$, with $α=\frac14$, and dynamical exponent being $z=4$. A generalized equal-area construction yields the binodal densities and phase diagram of AMB$+$. For supercritical quenches, domain size grows as $L(t)\sim t^{1/3}(1+c/\ln t)$, revealing logarithmic corrections to the classic $t^{1/3}$ growth-law; moreover it is consistent with the functional renormalization group predictions for marginal activity in $d=2$. While the logarithmic corrections are quite prominent in AMB, in AMB$+$ they are suppressed as the active current acts against the formation of macro-clusters; the growth is eventually arrested when a long-lived microphase-separated state appears.

The detection of Néel vector switching in a single-layer A-type antiferromagnet marks an important step toward functional two-dimensional spintronics. Here, Cr$_2$S$_3$-2D, grown on graphene on Ir(110), is established as a first single-layer A-type antiferromagnet. Spin-polarized scanning tunneling microscopy reveals hysteresis loops with a large switching field and a pronounced dependence on island size. X-ray magnetic circular dichroism at the Cr L$_{2,3}$ edges exhibits a tiny signal with a linear magnetic field dependence, consistent with a nearly compensated antiferromagnetic ground state and a Néel temperature of about 160 K. Quantitative analysis of the island-size dependence of the switching field, together with first principles calculations, indicates a slight imbalance between the magnetic moments of the two Cr planes of Cr$_2$S$_3$-2D when supported on a substrate. This imbalance results in a net magnetization for the A-type antiferromagnet, which enables the 180$^\circ$ rotation of the Néel vector. Moreover, Cr$_2$S$_3$-2D retains its magnetic properties after several days of exposure to air.

We introduce the Random Subsequence Model, a spin glass model on pairs of random strings $(X,Y) \in \{0,1\}^N \times \{0,1\}^M$ whose partition function counts subsequence embeddings of $Y$ into $X$. We study two variants: the null model, where $X$ and $Y$ are independent and uniform, and the planted model, where $X$ is uniform and $Y$ is a uniformly-random length-$M$ subsequence of $X$. We connect the Random Subsequence Model to longstanding problems in various fields, including the best rate achievable by uniformly-random codes in the deletion channel, the longest common subsequence problem between two random strings, and models of directed polymers in statistical physics. In the regime where $N,M\to\infty$ at a fixed ratio $α= M/N \in (0,1)$, we exhibit strict asymptotic separations between the null annealed free energy and the quenched free energies of the null and planted models at all values of the density parameter $α$. This suggests that these models are in a spin glass phase at zero temperature throughout the entire dense regime. As a consequence, we show that uniformly-random codes achieve a positive rate in the deletion channel for all deletion probabilities $p\in [0,1),$ settling multiple conjectures of the second author, Isik and Weissman (2024) and proving the first such positive rate result for the regime $p \geq 1/2$. We also give an exact analytic formula for the annealed free energy of the planted model for all values of the density parameter. This implies a corresponding analytic upper bound on the best rate achievable by uniformly-random codes in the deletion channel, complementing the lower bound from our first result. Our upper and lower bounds for the capacity of the deletion channel under uniform codes are far closer to each other than the best known upper and lower bounds for the capacity of the deletion channel.

We investigate the spontaneous meandering of slender rivulets in Hele-Shaw cells and identify the physical mechanism that selects the most unstable wavenumber, a quantity that has remained elusive even since the identification of the instability threshold [Daerr et al., Phys. Rev. Lett. 106, 184501 (2011)]. Earlier criteria did not distinguish between wavelengths and thus predicted an undiscriminated amplification of arbitrarily short perturbations. By incorporating viscous bending into the depth-averaged Navier-Stokes equations, we show that this effect is responsible for the selection of a fastest-growing mode, answering a question that has remained open for 15 years. We answer the open question of whether the meandering instability is absolute or convective. Our analysis also provides a simpler alternative derivation of the instability criterion, based on a low-viscosity assumption, and finally it yields a new physical interpretation of the mechanism: the destabilization arises directly from friction effects, instead of being caused by inertial forces. Together, these results complete the linear-stability picture of rivulet meandering in confined geometries, and establish viscous bending as a key parameter governing wavelength selection. They lay the groundwork for future exploration of the nonlinear features of the spontaneous meandering instability.

Nonvolatile magnetotransport responses in a single magnetic material have generally not been expected to exhibit a large ON/OFF ratio, because they are usually tied to spin-orbit coupling and therefore remain relatively weak. Here we show, contrary to this expectation, that giant nonvolatile magnetotransport can arise in a single magnetic material through magnetoelastic reconstruction of nonrelativistic real-space transport paths. Using the two-dimensional antiferromagnet FePS$_{3}$ as a representative system, first-principles quantum transport calculations reveal that charge transport is strongly tied to its quasi-one-dimensional zigzag sublattice chains and, under suitable doping, can even become confined to them. Moreover, strain lifts the degeneracy among symmetry-related zigzag variants and thus reorients these transport paths through magnetoelastic coupling. As a result, both the longitudinal and transverse conductivities change dramatically, yielding a giant magnetoelastic magnetoresistance of up to $10^{4}$% and an energy-independent Hall ratio that far exceeds the spontaneous Hall ratios found in conventional magnets. These results establish a route to exploiting symmetry-related magnetic variants and their associated transport paths for reconfigurable, high-performance spintronic devices with large nonvolatile readout contrast.

The quest for quantum ground states beyond the conventional Fermi-liquid paradigm remains a central challenge in many-body physics. The ferromagnetic Kondo effect represents a particularly intriguing case: an exotic variant of the Kondo effect in which an asymptotically free spin gives rise to singular Fermi-liquid behavior. Despite its theoretical importance, this regime has long eluded experimental observation owing to its subtle spectroscopic signatures, vanishingly small energy scales, and strict symmetry constraints in conventional nanostructures. Here, we demonstrate the coexistence of the ferromagnetic and overscreened Kondo effects within a single molecular spin system$\unicode{x2014}$a triangulene dimer comprising spin-1 and spin-1/2 units adsorbed on a metal surface. Low-temperature scanning tunneling spectroscopy reveals characteristic signatures of singular Fermi-liquid behavior, which are fully supported by many-body calculations. The unique molecular design provides intrinsic control over spin configuration and coupling asymmetry, allowing distinct many-body regimes to be accessed within the same platform. Our results establish a robust strategy for realizing non-Fermi-liquid physics at the atomic scale and demonstrate that ferromagnetic Kondo behavior can not only be observed but also deliberately engineered in molecular systems.

The realization of robust quantum anomalous Hall (QAH) phases at elevated temperatures remains a central challenge in condensed matter physics. While quadratic band crossing points (QBCP) provide a promising route towards QAH states, existing proposals are largely confined to idealized models or hindered by interaction-driven competing orders. Here, we demonstrate that these limitations are not intrinsic to QBCP but arise from their specific implementation. We propose a general mechanism where band inversion between a symmetry-protected orbital doublet (e.g. $d_{xz},d_{yz}$) and an isolated orbital (e.g. $d_{z^2}$)-generically generates a QBCP with opposite curvature. This crossing is directly gapped at the single-particle level by intrinsic atomic spin-orbit coupling, while the underlying band inversion naturally shields the resulting topological gap against other interaction-driven instabilities. We further suggest monolayer compounds $MNX_2$ ($M$= Ni, Pd, Pt; $N$= Nb, Ta; $X$= S, Se, Te) as a realistic material class that intrinsically realizes this mechanism. These findings provide a concrete pathway toward robust QAH phases in correlated materials.

Dyson [Commun. Math. Phys. 12, 91 (1969)] rigorously proved the existence of a phase transition in the one-dimensional Ising model with long-range interactions of the form $r^{-α}$ for $1 < α< 2$. In the present study, we extend this result to the Ising spin glass model with Gaussian disorder on the Nishimori line. Following Dyson's method, we first prove the existence of long-range order at finite low temperatures in the Dyson hierarchical Ising spin glass model on the Nishimori line, with power-law-like interactions $J(r) \sim r^{-α}$ for $1 < α< 3/2$. The key ingredients of the proof are the interpolation method developed in the rigorous analysis of mean-field spin glass models, the Gibbs--Bogoliubov inequality on the Nishimori line, and the Tsirelson--Ibragimov--Sudakov inequality (Gaussian concentration inequality). We then use the Griffiths inequality on the Nishimori line to rigorously establish the existence of a phase transition in the one-dimensional Ising spin glass model with long-range interactions on the Nishimori line for $1 < α< 3/2$. For $α\ge 3/2$, the existence of a phase transition remains an open problem.

We derive general trade-off relations among the power, efficiency, and constancy for two-terminal thermoelectric systems in the linear response regime. Constancy, which quantifies the steadiness of the heat engine, is measured by its fluctuations. The bounds of the efficiency, power and fluctuations are valid even when time-reversal symmetry is broken, revealing how such a symmetry breaking alters the fundamental constraints on steady-state energy conversion. Our results extend and refine previously established universal trade-offs, offering deeper insight into the performance limits in nonequilibrium thermodynamics. Guided by this bound, heat engines with broken time-reversal symmetry can be operated at near-Carnot efficiency while maintaining finite power output and fluctuations, enabling them to outperform their traditional counterparts.

Machines are designed, assembled, and programmed to convert power into predetermined dynamics and functions. In contrast, living systems such as interacting cells and animal groups self-organize, synchronize, and perform complex tasks without predefined patterns. Inspired by these decentralized architectures, experiments have shown that small assemblies of elastically coupled self-propelled robots can achieve two fundamental functionalities observed in nature: collective motion and oscillatory deformations. However, biological inspiration has steered research toward translational self-propulsion, while active rotation remains an underexplored route to designing broader animate materials. Here, we study the self-organization of microscopic metamachines composed of thousands of 3D-printed rotary motors. We first demonstrate and explain how motors precessing in unspecified directions collectively arrange their dynamics into a pristine antiferromagnetic phase. Next, we elucidate the emergence of spatiotemporal order in the form of phase coherence in the rotors' precession. Finally, we show how quenched disorder initiates the free propagation of phase waves across self-organized regions with mismatched rotation speeds. Our results suggest that spinner-based metamachines could illuminate metachronal-wave formation in living systems, and signal propagation in synthetic animate materials.

We demonstrate complete coherent control of a single spin qubit confined in a self-assembled InAs negatively charged quantum dot subjected to an Oblique magnetic field, and directly compare this regime with the conventional Voigt geometry. In the Oblique-field configuration, the groundstate spin eigenstates are found to be unequal superpositions of the bare electron spin, with their composition tunable via the orientation of the applied field. This tunable spin mixing provides an additional degree of freedom to engineer the spin basis and associated optical couplings in the charged quantum dot system. Although this geometry has a distinct structure with important implications, it provides a regime in which we can fully and coherently control the tailored spin qubit. We observe Rabi oscillations and Ramsey fringes, and demonstrate arbitrary single-qubit rotations, enabling a direct comparison with the Voigt case. Our results establish that spin-qubit control does not necessarily require a pure Voigt geometry and can instead be achieved under Oblique magnetic fields. This relaxes constraints on device and field alignment and offers a versatile route to design and optimize quantum information processing architectures in semiconductor quantum dots.

Topological defects (TDs) are crucial for understanding important physical properties of crystalline materials including mechanical failure, ion transport, and two-dimensional melting. This concept has not translated to disordered materials like glasses because these solids have no obvious reference structure that can be used to define TDs. As a result, key properties related to those listed above have typically been modeled using purely phenomenological approaches. Recent studies have demonstrated that certain observables commonly associated with TDs can also be identified in disordered solids indicating that topological concepts may be as crucial in amorphous solids as in crystals. This hints that TDs may offer a first-principles framework for understanding their mechanical response and complex spatiotemporal dynamics. In this Perspective, we review recent theoretical, numerical, and experimental studies that have exploited topological concepts to rationalize mechanical properties of amorphous solids. We also highlight pressing open questions and some promising directions for future research in the field.

In a recent paper [M. Gül et al., Phys. Rev. E, 110 (6), 064115] we showed that test particle sum rules, which address the excess chemical potential and isothermal compressibility, could be used to develop new and accurate classical density functionals for hard-sphere (HS) fluids. Here we extend our approach to the construction of HS functionals building upon the state of art White-Bear (WB) and White-Bear mark II functionals. Employing the same test-particle sum rules we determine the two free parameters in the Lutsko [James F. Lutsko, Phys. Rev. E, 102, 062137] formulation of fundamental measure theory (FMT) by minimizing the relative errors between different routes to the two thermodynamic quantities. The resulting optimized Lutsko WB functionals, especially Lutsko WB mark II, are generally more accurate and consistent than those obtained in earlier treatments.

In type II multiferroics, noncollinear spin textures are expected to induce electric polarization directly, leading to strong magnetoelectric coupling. Realizing such spin driven multiferroicity in two-dimensional systems, and elucidating the interplay between local spins and electric polarization, are of both fundamental and technological importance. Here, using vectorial spin polarized scanning tunneling microscopy, we investigated the spin-driven multiferroicity in monolayer NiI2 at atomic scale. We identify a canted spin-spiral state with fully determined spin rotation plane, accompanied by a 2Q charge modulation. At spin spiral domain walls, we discover topological spin textures that composed of meron/antimeron pairs. These textures are associated with distinct charge pattern and notable band shifts, indicating local bound charges induced by variations of ferroelectricity at domain wall. Our observations are well captured by a realistic spin model incorporating Kitaev interactions and generalized spin-current model of type II multiferroicity. The findings provide microscopic evidence of spin-driven multiferroicity in an extreme 2D system and establish a platform for low-dissipation, electric-field control of topological spin textures.

Transport through correlated nanoscale systems underpins the operation of quantum-dot and molecular-scale devices, yet accurate simulations of large open quantum systems remain computationally challenging as system size increases. Tensor-network methods offer a promising route past this scaling barrier by efficiently compressing quantum states. Here we extend a tensor-based solver with a jump-counting estimator that enables direct computation of steady-state electron currents from lead-induced tunneling events. We benchmark the resulting currents against the state-of-the-art master-equation solver QmeQ across a range of lead-dot and inter-dot coupling parameters and find quantitative agreement in the tractable regime. Compared with classical approaches, TJM reduces memory requirements and wall-clock time by orders of magnitude, enabling simulations of interacting quantum-dot arrays far beyond the range accessible to density-matrix-based transport solvers and systematic studies of size-dependent nonequilibrium transport in larger arrays. Our approach allow us to model quantum transport in an array of up to fifty (50) quantum dots.

We present TNRKit, an open-source Julia package for Tensor Network Renormalization (TNR) of two- and three-dimensional classical statistical models and Euclidean lattice field theories. Built on top of TensorKit, it provides a symmetry-aware framework for constructing tensor-network representations of partition functions and coarse-graining them using methods such as TRG, HOTRG, and LoopTNR. Beyond thermodynamic quantities, the package enables the extraction of universal conformal data -- including scaling dimensions and the central charge -- directly from fixed-point tensors. TNRKit is designed with both usability and extensibility in mind, offering a practical platform for applying, benchmarking, and developing modern tensor renormalization algorithms. This paper also serves as a self-contained introduction to the TNR framework.

Exciton dissociation via the excitonic Mott transition (EMT) governs the high-density optical response of semiconductors and sets fundamental limits for optoelectronic devices. The EMT is conventionally linked to the onset of population inversion and the emergence of optical gain. Here, we demonstrate that this paradigm can break down under ultrafast non-equilibrium excitation. Using femtosecond pump-probe optical spectroscopy, we drive a monolayer transition metal dichalcogenide into a dense photoexcited state in which the excitonic resonance is completely quenched within ~100 fs, while the optical gain is entirely absent across the explored fluence range. State-of-the-art real-time ab initio simulations reveal that the EMT is governed by an interplay of strongly nonthermal carrier populations and nonequilibrium dynamical screening of the Coulomb interaction. The quantitative agreement between theory and experiment identifies a distinct, ultrafast pathway to exciton ionization beyond quasi-equilibrium descriptions and demonstrates that population inversion is not a universal prerequisite for the EMT.

The unique combination of outstanding optical quality and attractive spin properties opens new avenues for optical spin control in hybrid organic-inorganic perovskite semiconductors. Using the optically detected magnetic resonance technique, we study the spins of electrons and holes in mixed-cation MA$_x$FA$_{1-x}$PbI$_3$ single crystals with $x = 0.4$ and 0.8. Multiple distinct spin subensembles with $g$-factors spanning from 2.9 to 3.6 for electrons and from 0.5 to 1.2 for holes are resolved, revealing diverse localization environments. We measure the longitudinal spin relaxation times, $T_1$, reaching 2 ms and remaining in the $μ$s range even for weakly localized carriers at the cryogenic temperature of 1.6 K. The magnetic-field dependence of $T_1$ is dominated by the random nuclear (Overhauser) fields with strengths of $\sim 0.4-0.8$ mT for electrons and $\sim 4-12$ mT for holes, corresponding to $μ$s-long correlation times of the hyperfine field determined by carrier hopping between shallow localization sites. The temperature dependence of $T_1$ reveals a weak localization potential of the charge carriers and shows a correlation between $T_1$ and the inhomogeneity of the spin ensemble. These results establish mixed-A-site perovskite single crystals as a promising solid-state platform with long-lived spin states for quantum information applications.

The $\mathbb{Z}_N$ cluster-state wavefunction, a paradigmatic example of symmetry-protected topological (SPT) order with $\mathbb{Z}_N \times \mathbb{Z}_N$ symmetry, is expressed in various equivalent ways. We identify the projector-based scheme called the $P$-representation as the efficient way to express cluster and dipolar cluster state's wavefunctions. Employing the restricted Boltzmann machine scheme to re-write the interaction matrix in the $P$-representation in terms of neural weight matrices allows us to develop the neural quantum state (NQS) and the matrix product state (MPS) representations of the same state. The NQS and MPS representations differ only in the way the weight matrices are split and grouped together in a matrix product. For both $\mathbb{Z}_N$ cluster and dipolar cluster states, we derive in closed form the weight function $W(s,h)$ that couples physical spins $s$ to hidden variables $h$, generalizing the previous construction for $Z_2$ cluster states to $\mathbb{Z}_N$. For the dipolar cluster state protected by two charge and two dipole symmetries, the procedure we have developed leads to the tensor product state (TPS) representation of the wavefunction where each local tensor carries three virtual indices connecting a given site to two nearest neighbors and one further neighbor. We benchmark the resulting TPS construction against conventional MPS representation using density-matrix renormalization group simulations and argue that the TPS could offer a more efficient representation for some modulated SPT states. As a by-product of the investigation, we generalize the previous $Z_2$ matrix product operator construction of the Kramers-Wannier (KW) operator to $\mathbb{Z}_N$ and interprets it as the dipolar generalization of the discrete Fourier transform on $\mathbb{Z}_N$ variables. The new interpretation naturally explains why the KW map is non-invertible.

We study current-driven domain-wall (DW) dynamics in antiferromagnets (AFMs) with Dzyaloshinskii-Moriya interaction (DMI). We obtain an exact analytical solution for spiral DW dynamics, applicable to both head-to-head DWs under bulk DMI and up-down DWs under interfacial DMI when the magnetic easy axis is aligned with the DMI vector. For the latter case experimentally relevant to synthetic AFMs with in-plane anisotropy, the solution predicts a constant DW velocity driven by nonadiabatic spin-transfer torque together with a steady rotation of the DW tilt angle induced by damping-like spin-orbit torque. Remarkably, the DW width shows unconventional current dependence, either pure elongation or contraction followed by elongation depending on damping and torque parameters, in sharp contrast to the Lorentz-type contraction known for antiferromagnetic (AF) DWs without DMI. These results provide an exact description of current-driven AF-DW dynamics and suggest experimentally accessible signatures of DMI-modified DW dynamics in synthetic AFMs.

Magnon thermal Hall effect in insulating magnets is the manifestation of Berry curvature in magnon bands, which is formulated using the emergent gauge fields that act on magnons as a fictitious magnetic field. In ferromagnets, it is commonly accepted as the outcome of U(1) gauge fields generated by Dzyaloshinskii-Moriya interactions and spin textures, but this mechanism is often suppressed by symmetry-enforced cancellations in many lattice geometries, known as a no-go rule. As a result, antiferromagnetic insulators have long been considered as unfavorable platforms for the effect. We show that antiferromagnets with multiple magnetic sublattices naturally host non-Abelian SU(N) gauge fields in magnon band structures, providing a robust rule-to-go mechanism. The noncommutativity of these gauge fields prevents Berry-curvature cancellation and guarantees a nonvanishing thermal Hall response. As a minimal realization, we demonstrate that a coplanar 120$^{\circ}$ antiferromagnet with Dzyaloshinskii-Moriya interactions constitutes a canonical SU(3) platform for the magnon thermal Hall effect. We provide a table of so-far-known two-dimensional lattice geometries and variants of magnetic structures, along with the corresponding gauge fields, providing a unified guideline for identifying magnetic materials, including antiferromagnets and altermagnets, that host thermal Hall transport.

Unconventional superconductivity in quasi--two-dimensional systems is commonly identified through the emergence of Andreev bound states (ABS) at in-plane edges, while surfaces perpendicular to out-of-plane direction remain fully gapped due to weak interlayer coherence. This directional anisotropy has long served as a key paradigm for constraining pairing symmetries. Here, we show that Sr$_2$RuO$_4$ exhibits a striking reversal of this behavior. Using edge- and surface-sensitive spectroscopy, we observe pronounced in-gap ABS at surfaces perpendicular to the out-of-plane direction, whereas in-plane edges exhibit a reduced intensity of the in-gap spectral features. We show that this anomalous anisotropy can arise from the inter-orbital character of the superconducting pairing. Both even- and odd-parity inter-orbital pairing channels naturally generate robust surface ABS while suppressing planar edge modes and can also provide a mechanism for the appearance of a horizontal line node. Supported by \textit{ab initio} and model calculations, including Sr$_2$RuO$_4$/Ag interface reconstructions, our results highlight the possible role of inter-orbital correlations in shaping the spectroscopic response and provide constraints on the structure of the superconducting order parameter in Sr$_2$RuO$_4$.

In recent years, techniques of intense THz laser have enabled us to experimentally observe nonlinear spin dynamics in antiferromagnets since the elementary excitations such as magnons reside on a THz to GHz range in antiferromagnets and THz laser thus can directly excite them. We numerically and theoretically investigate THz-laser or GHz-wave driven harmonic generations in typical ordered phases of antiferromagnets: Néel, canted and weak ferromagnetic phases. The radiation waves (harmonic generations) are created by the incident-wave driven magnon dynamics. We point out that magnetic orders and phase transitions can change the spectra of harmonic generations, differently from those of metallic, semiconductor, or atomic-gas systems without (spontaneous) symmetry breakings. We consider both the magnon harmonic generation driven by standard single-color laser and that by two-color laser in the antiferromagnets, and find several dynamical symmetries and the corresponding selection rules of the harmonic generations. These results indicate that the magnon harmonic generation spectra provide new information about symmetry or symmetry breaking of antiferromagnets.

Chiral liquid crystals and chiral magnets host a wide variety of topological solitons described by closely related continuum theories, namely the Frank-Oseen and Dzyaloshinskii models. Exploiting this correspondence, we develop a unified description of cholesteric fingers in confined liquid crystals and their magnetic counterparts. Within a continuum framework including bulk and surface anisotropies, we analyze the topology, structure, interactions, and collective states of the two main finger types, CF-1 and CF-2. We show that cholesteric fingers are composite chiral solitons built from merons. CF-2 corresponds to a bimeron with unit topological charge, while CF-1 is a topologically trivial composite of two merons with identical vorticities. From a homotopic viewpoint these textures correspond to skyrmions and droplets. Strong homeotropic anchoring induces confinement effects that reshape the meron structure and redistribute topological charge across the film thickness. Isolated fingers in the homogeneous state interact repulsively and behave as particle-like objects. Periodic phases emerge when the energy of an isolated finger becomes negative, leading to nucleation-type transitions with a diverging lattice period. Degenerate finger types allow mixed periodic sequences, analogous to stacking polytypes. In a conical background, interactions become attractive due to overlap of distortion regions. Film thickness controls stability and structure: at small thickness solitons collapse, while at large thickness bimerons exhibit bistability between surface-stabilized and bulk-like states.

The local H theorem is shown to hold for the Enskog equation with a modified Enskog factor proposed by the authors [Phys. Rev. E 111, 065108 (2025)]. This is a stronger statement than the global one in the same paper and has been obtained along the lines of Mareschal et al. [Phys. Rev. Lett. 52, 1169-1172 (1984)] for the modified (or revised) Enskog equation. Furthermore, it is shown that the local H theorem also holds for the corresponding Enskog-Vlasov equation.

Friction forces are conventionally modeled via semiclassical theories that associate energy dissipation with newtonian motion on corrugated interface potentials. This consolidated approach is challenged at the nanoscale by observation of nearly unimpeded water flow in narrow carbon nanotubes (CNTs), in spite of nonvanishing energy corrugations. Here we go beyond the standard newtonian perspective, adopting a quantum mechanical description of 4 He flow through narrow CNTs. Building upon our Bloch-wave dynamics [Phys. Rev. Lett. 131, 206301 (2023)] we explore realistic flow conditions, including non-negligible interface interactions, finite temperatures, and imperfect CNTs. At T = 0 K we found that 4 He waves can propagate through ideally periodic, corrugated interface potentials with no friction: below a critical velocity regulated by interface corrugations, energy loss by emission of plasmon and phonon quanta is forbidden. Introducing realistic impurities/defects one still finds very large mean free paths that can exceed the micrometer scale, while thermal phonons and plasmons yield even lower scattering rates. This establishes the unexpected emergence of ballistic wavelike transport in narrow CNTs within realistic nanoscale devices, and demonstrates the intrinsic quantumness of nanoscale interfaces.

In a recent work [Phys. Rev. E 109, L042102 (2024)], interesting dimensional crossovers [from two- to one-dimensional (2D to 1D) scaling] were found in the growth of Kardar-Parisi-Zhang (KPZ) interfaces on rectangular substrates, with lateral sizes $L_y > L_x$. Here, we extend this study to other universality classes for interface growth -- specifically, the Edwards-Wilkinson (EW), the Mullins-Herring (MH), and the Villain-Lai Das Sarma (VLDS) classes. From extensive simulations, we demonstrate that, in all systems with sufficiently large aspect ratio $\mathcal{R}=L_y/L_x$, the roughness $W$ scales with time $t$ in the growth regime as $W \sim t^{β_{\text{2D}}}$ for $t \ll t_c$ and $W \sim t^{β_{\text{1D}}}$ for $t \gg t_c$, where $t_c \sim L_x^{z_{2\text{D}}}$ in most cases. For the VLDS class, this crossover is also observed in the height distribution (HD), which approaches its characteristic probability density function for the 2D case at short times ($t \ll t_c$) and then crosses over to the asymptotic 1D HD. Dimensional crossovers are also found in the steady state regime, both in the roughness scaling as well as in the VLDS HD, which interpolate between the 2D and 1D ones as $\mathcal{R}$ increases. The particular case $L_x = L_y^δ$, with $0 < δ< 1$, is also discussed in detail and reveals interesting features of the investigated systems. For instance, there exist a `special' exponent $δ^* = z_{1\text{D}}/z_{2\text{D}}$ such that the temporal crossover cannot be observed for $δ> δ^*$. Moreover, this leads the saturation roughness to display a nonuniversal scaling: $W_s \sim L_y^Λ$, with $Λ= (1-δ) α_{1\text{D}} + δα_{2\text{D}}$.

Non-reciprocal interactions in active matter gives rise to a multitude of fascinating phenomena among which are collective oscillatory states without intrinsic particle chirality and active turbulence. Here we show that in a paradigmatic model for non-reciprocal flocking, the two species Vicsek model, these two states coexist: chiral order for small flocks, and extensive spatiotemporal chaos for large flocks, both separated by a finite wavelength instability whose scale is set by the rotation radius of the chiral orbits. For system sizes larger than this length scale extensive spatiotemporal chaos unfolds, as manifested by an extensive number of Floquet exponents for the unstable chiral state, a positive Lyapunov exponent, a finite correlation and chaotic length and a broad energy spectrum. Our results suggest that complex, turbulent behavior is a generic possibility in any system where particles or fields interact asymmetrically and may have significant implications for understanding how non-reciprocal interactions could drive chaotic, fluid-like behavior in active matter.

We introduce a systematic protocol for constructing quantum Hilbert-space-fragmented Hamiltonians, whose Krylov-sector structure, unlike in classically fragmented models, can be fully resolved only in an entangled basis. The protocol takes as input a classically fragmented model and uses a Rokhsar-Kivelson type construction to promote it to a quantum fragmented model. Notably, the procedure also works with non-fragmented inputs (such as Ising models). We explain how the Krylov sectors of the resulting quantum fragmented model may be labeled and counted in one dimension, and outline experimentally accessible verification of quantum fragmentation, assuming the ability to prepare specific initial states and perform tomography on reduced density matrices. We further analyze the entanglement structure of the entangled basis underlying quantum fragmentation, which sharply distinguishes it from both classical fragmentation and the trivial "fragmentation" of generic Hamiltonians in their eigenbasis. We also extend the construction to higher dimensions, with an explicit proof of principle example in two dimensions. We expect these results to open a new route to the systematic exploration of quantum fragmentation.

Correlated electron systems that host multiple electronic orders offer routes to multifunctional quantum materials, but strong competition between these orders often prevents their coexistence. Here we show that nanoscale, metastable intercalated heterostructures can stabilize a rare combination of long-range magnetism and a commensurate charge density wave (C-CDW) order in a single material. We synthesize a two-dimensional (2D) metastable crystal, T/H-Fe$_x$TaS2, which comprises an endotaxial polytype heterostructure of 1T-TaS$_2$ and H-TaS$_2$ with Fe intercalated in the van der Waals interfaces. In T/H-Fe$_x$TaS2, Fe intercalants provide localized spins that support ferromagnetism, while 1T layers host a robust commensurate charge density wave (C-CDW) that persists to room temperature. In these intercalated heterostructures, Fe content simultaneously tunes ordering of spin and charge degrees of freedom, positioning topochemically-prepared intercalated endotaxial heterostructures as a route to stabilize and control competing quantum phases in 2D materials.

Periodic boundary conditions are not always used in the study of disordered systems, but it can be advantageous to apply them to mimick thermodynamically large systems. In this case, polarization and its cumulants can not be obtained directly, but through the tools of the modern theory of polarization. This theory casts the polarization in crystalline systems as a geometric phase, rather than an operator expectation value. We develop disorder averaging techniques within the context of this theory which can calculate the variance of the polarization, its higher order moments, and the excess kurtosis (or Binder cumulant). We also derive an indicator of delocalization based on the degeneracy as a function of boundary conditions. We apply the computational techniques to two model systems. To test localization, we use a one-dimensional disordered model which is fully Anderson localized. Our calculations verify this. We also apply our techniques to the one dimensional de Moura-Lyra model, developed to study power law correlated (controlled by a parameter, $α$) disorder. While this model is a pathological one, our method is validated. We also point out the significance of pairwise degeneracies found in the parameter range, $α>2$ and near the band center (or near half filling), where the model was conjectured to exhibit a mobility edge.

We investigate a Hamiltonian formulation of vortex interactions on a doubly periodic inviscid fluid domain, based on an exact interaction expressed in terms of the Schottky-Klein prime function and its q-representation. The two-vortex problem is reduced to a single complex degree of freedom, from which explicit expressions for the orbital rotation frequency and dipole translation velocity are obtained and verified against simulations. Building on this framework, we derive a small-cluster expansion that reveals a universal decomposition of the dynamics into planar interactions, isotropic torus corrections, and geometry-induced anisotropic modes. At leading order, the collective dynamics admits a closed description in terms of a single complex quadrupole moment: its real part governs the corrections to the rotation rate, while its imaginary part controls the slow breathing of the cluster. These predictions are quantitatively confirmed by direct numerical simulations, establishing a reduced description of vortex clusters on the flat torus and compact fluid domains.

We revisit a learning-induced tricritical point, at which three phases with strong, weak, and broken $Z_2$ symmetry meet, in the phase diagram of a deformed toric code wavefunction subjected to weak measurements. This setting is exactly dual to a classical Bayesian inference phase diagram of the $2D$ classical Ising model. Here we demonstrate that this tricritical point lies on a distinct $\textit{higher Nishimori line}$, which has an emergent gauge-invariant formulation, just like the ordinary Nishimori line but with a higher replica symmetry as a replica stat-mech model in the replica number $R\rightarrow2$ limit, where disorder is averaged according to the Born rule. As such, the learning tricritical point is in fact a $\textit{higher Nishimori critical point}$. Using this identification, we obtain a number of $\textit{exact results}$ at this $\textit{higher}$ Nishimori critical point; e.g., we show that the power-law exponent of the Edwards-Anderson correlation function is exactly equal to that of the spin correlation function at the unmeasured Ising critical point and verify this in numerical simulations. Using the tools of the proof of a $c$-effective theorem [arXiv:2507.07959], we show that the Casimir effective central charge $c_{\text{eff}}$ $\textit{decreases}$ under renormalization group (RG) flow from the $\textit{higher}$ Nishimori critical point to the unmeasured $2D$ Ising critical point, and is thus greater than $1/2$. This is corroborated by extensive numerical simulations finding $c_{\text{eff}} = 0.522(1)$. The analytical result also explains, with a physically motivated assumption, the numerically observed increase of the Casimir effective central charge under the RG flow from the ordinary Nishimori critical point to the clean Ising critical point in the random-bond Ising model. We also discuss $\textit{higher}$ Nishimori criticality in general dimensions $D>1$.

Nonequilibrium response theory is a fundamental framework for understanding how physical systems respond to perturbations. Recently, a mutual linearity has been discovered for Markov jump processes using linear algebra analysis. This mutual linearity states that two observables are linearly dependent on each other in the long-time limit when the transition rate of a single edge is altered. It has also been extended to non-stationary cases for current observables. In this work, we provide a trajectory-based derivation of mutual linearity utilizing the trajectory-level linear response theory. The trajectory approach allows us to generalize the mutual linearity to non-stationary relaxation dynamics for state observables and counting observables. Our results shed light on the fundamental response properties far from equilibrium and the trajectory-level origin of mutual linearity. Our trajectory-based approach makes it possible to generalize the mutual linearity to a broader class of systems, including diffusion processes and open quantum systems.

The Peierls-Boltzmann transport equation (PBE), which governs non-equilibrium phonon transport, suffers from the curse of dimensionality due to its high-dimensional phase space including both real and modal spaces. We explore the use of matrix product states (MPS) for numerical simulation of the PBE. We show that an MPS configuration based on scattering events combined with a dimensionless form of the solution can drastically increase the locality of correlations between tensors in the MPS representation, enhancing its effectiveness in dimension reduction. We further examine the effects of index ordering in an MPS and find that the highest locality is achieved when tensor chains associated with both real and modal spaces are connected from the coarsest grid to each other in the center of the MPS. Using this optimal configuration and a solver inspired by the density matrix renormalization group, we solve the PBE discretized by a finite volume method (FVM). The solution is obtained for crystalline silicon across ballistic, quasi-ballistic, and diffusive transport regimes. An MPS truncated to the compression ratio of $10^{-3}$ suffices to reproduce reference solutions with high fidelity. The computational cost scales sublinearly with the number of grid points in both real and modal spaces, achieving roughly an order of magnitude reduction in computational time compared to the FVM with sparse matrix operation.

We consider a sequence of random Hamiltonians $H_n(h,σ)=\sum^n_{i=1}h_i(σ_i-m)$, and study the asymptotic ($n\to \infty$) distribution of the energy levels $(H_n(h,σ))_{σ\in \{-1,1\}^n}$, where $h_1,h_2,\cdots$ are i.i.d. random variables. We show that, when $e^{O(n)}$ configurations are sampled at random, the corresponding collection of energy levels converges in distribution to a Poisson point process with exponential intensity measure. This establishes the Random Energy Model (REM) universality for the present model. Our results strengthen earlier works on local REM universality by characterizing the distribution of $O(1)-$order fluctuations of $H_n$. In addition, we improve upon the REM universality by dilution studied by Ben Arous, Gayrard, Kuptsov by allowing an exponentially large number $e^{O(n)}$ of sampled configurations, instead of $e^{o(\sqrt{n})}$. Finally, we derive the asymptotic distribution of the Gibbs weight.

The nonlinear Hall effect (NHE) is a recently discovered member of the Hall effect family in which the Hall voltage shows a nonlinear behavior when a transverse electric field is applied. While the NHE does not require broken time-reversal symmetry, such as that induced by a magnetic field, it requires broken inversion symmetry, which limits the range of suitable systems and potential applications. Here, we demonstrate an ultrafast NHE in centrosymmetric black phosphorus through dynamical symmetry breaking using femtosecond light pulses. We provide a detailed microscopic picture of excited carrier dynamics and induced fields using momentum-resolved photoemission spectroscopy combined with \textit{ab-initio} calculations. The ultrafast NHE is observed exclusively for the light polarization aligned with the armchair high-symmetry direction and persists over 300 fs, which opens new possibilities for selective and ultrafast light-to-current conversions.

D. B. Brückner et al. [Phys. Rev. Lett. 125, 058103 (2020)] have described a novel method for inferring the dynamics of systems governed by an underdamped Langevin equation in the presence of measurement noise. While this is a significant achievement, the paper also presents a number of significant errors. These are explained and corrected in this note.

Bound states are dissipation-resilient states that may emerge when quantum systems are strongly coupled to reservoirs with band gaps. We analyze an exactly solvable bosonic model for bound state existence and reproduce these results by a weak-coupling treatment of a supersystem composed of the original system and multiple reaction coordinates, which are individually representing small energy intervals of the reservoir spectral function. Within the perturbative supersystem treatment, the bound state stability results from its energy being inside the band gap. We discuss cases of multiple band gaps and also show that already in presence of weak interactions the bound state's lifetime is finite -- but can be increased by raising the system-reservoir coupling strength.

This note reports the following observation: the finite-volume expectation value of the spin operator (the one-point function) between the $\mathbb{Z}_2$-even and odd ground states in the critical periodic Ising chain, when continued as a complex-analytic function of the system length $N$ through the Borel resummation of its large-$N$ expansion, has a natural boundary of analyticity along the negative real axis. The singular behavior near the negative real axis, after an exponential map, is the same as that of a Lambert-type series for the odd-divisor-squared sum near the unit circle $|z|=1$. The same divisor sum also governs the strengths of the Borel discontinuities of the one-point function's factorially-divergent large-$N$ asymptotics. We also report the all-order large-$N$ asymptotics of the leg function for the finite-volume spin-operator form factor, and the similarities to certain known quantities in the literature.

We introduce a new method that exploits the combination of the Heat Kernel (HK) and Background Field Method to compute gauge-invariant and gauge parameter-independent quantities such as the effective potential, anomalous dimensions, and renormalization group equations. In contrast to currently employed techniques, these results are obtained exclusively from the dynamics of the background fields, without relying on supplementary input from, e.g., traditional diagrammatic calculations. This is achieved by a consistent treatment of open and closed derivatives in the HK expansions. In this way, we compute the standard quantities such as $β$ functions and their gauge-parameter independence when background fields are on-shell. We demonstrate this formalism for instructive examples such as Scalar QED and Yukawa theory. Full results for the bosonic part of the Standard Model provide further validation of our approach.

We show that propagating switching fronts mediate directional state transmission and polarity selection in a passive many-body suspension. In confined trains of slipper-shaped deformable particles in Poiseuille flow, this behavior originates from directionally biased switching between neighboring particles: owing to the fore-aft asymmetry of the slipper, an upstream particle drives switching of its downstream neighbor more effectively than in the reverse direction. A local transition from an opposite-sign pair to a same-sign pair therefore launches a streamwise front that relays the inclination sign from particle to particle. A minimal coarse-grained model with local bistability and directional coupling captures front propagation and arrest. In periodic trains, the fronts coarsen into a uniformly polarized state, whereas in long open trains they arrest and leave persistent polarized domains. Our results point to local bistability and directional coupling as a route to collective polarization in passive many-body systems.

Numerous problems in development, regeneration, and disease involve simultaneous evolution of both spatial organization and the internal state of the constituents in addition to local interactions and crowding. This motivates us to study a minimal model for interacting populations evolving in coupled spatial and internal-state coordinates. We focus on a specific transition of particular biological interest: the reorganization of dense collectives from compact or arrested states toward boundary-led peeling and migration. In our formulation, each particle carries a spatial position and a scalar internal state, and interacts through finite-range forces. Mobilities are defined on both spatial and internal-state coordinates with a density dependence, and are asymmetrically cross-coupled. We derive update equations for stochastic dynamics in the overdamped limit and perform numerical simulations. We find that mobility in internal-state coordinates alone provides an independent control axis for large-scale spatial reorganization. In particular, increasing the baseline internal-state diffusivity and tuning its density dependence drives a transition from arrested aggregates to a peeling-like regime with broad spatial excursions, strong outward radial bias, and edge-localized activity, while the baseline positional diffusivity is held fixed. The transition is accompanied by correlated broadening of spatial and internal-state displacements, systematic reorganization of radial density and density-curvature profiles, and a pronounced dependence on system size, consistent with the idea that growing aggregates can cross into a boundary-dominated migratory state. These results establish the utility of our approach and motivate a broader framework aimed at modeling state change, spatial redistribution, and neighborhood structure within a common formalism.

We study strongly correlated superconductivity in magic-angle twisted bilayer graphene (MATBG) using variational Gutzwiller wavefunction where the Gutzwiller projector $\hat{P}_{R}$ is allowed to break charge U(1) symmetry to accommodate superconducting (SC) order. The ground state energy is evaluated via the Gutzwiller Approximation applied to an 8-band model consisting of correlated f-orbitals and uncorrelated c-orbitals, with interactions including onsite Coulomb repulsion $U$, phonon-mediated anti-Hund's coupling $\hat{H}_{J_A}$, and intra-orbital Hund's coupling $\hat{H}_{J_H}$. At filling $ν=2.5$, we map out the phase diagram as a function of $U$ and $J_A$, finding a dome-shaped Fermi liquid (FL) phase that separates a weakly correlated BCS-like SC (BCS-SC) at small $U$ from a strongly correlated SC (SC-SC) at large $U$. A nematic SC state, stabilized over a large region of the phase diagram including the realistic parameter regime of MATBG, acquires a nodal gap structure with V-shaped density of states at large $U$ via interaction-driven SC gap reconstruction. In the SC-SC regime, the off-diagonal (charge-U(1)-breaking) components of $\hat{P}_{R}$ strongly suppress $f$-orbital charge fluctuations while maintaining finite pairing order and a sizeable quasiparticle weight $Z$, distinguishing it from a conventional Mott insulator. We further identify a novel small Fermi liquid (sFL) state with effective Fermi surface volume $=ν+2$. Interestingly, in the intermediate- ($U \lesssim 40$ meV) and large-$U$ ($U \gtrsim 40$ meV) regimes, the conventional FL and the sFL are the lowest-energy normal phases, respectively, potentially serve as the parent states of the SC-SC phase. These results illuminate the interplay between strong correlations and unconventional pairing in MATBG, and establish a versatile Gutzwiller framework applicable to other strongly correlated superconductors.

What dynamical quantity is actually controlled by higher-order interactions in chaotic oscillator networks remains unclear. In amplitude-active systems, chaos is often interpreted through coherence, yet coherence is not the quantity that governs instability. In this work, we study a minimal globally coupled quartet of nonisochronous Stuart-Landau oscillators with pairwise and symmetric three-body interactions. The pairwise baseline already supports a connected chaotic branch, and higher-order coupling reconstructs rather than creates this irregular dynamics. We show that chaos is organized not by phase coherence but by effective-frequency shear: higher-order coupling regulates amplitude heterogeneity, which nonisochronicity converts into shear, and shear controls how chaos is expressed under higher-order coupling. The Lyapunov response collapses onto a reduced shear-based description, revealing an indirect control pathway. These results establish that higher-order interactions control chaos only indirectly, by regulating an amplitude-shear mechanism rather than acting directly on synchrony.

The interplay between classical chaos and quantum tunneling is examined in driven nonlinear systems, with emphasis on how semi classical phase space structures influence purely quantum transport phenomena. We show that, in the presence of external driving and stochastic perturbations, tunneling rates acquire an activated form determined by effective classical barriers, providing a transparent link between chaotic dynamics and quantum tunneling. Within this framework, chaos assisted tunneling and coherent destruction of tunneling emerge as closely related manifestations of the same underlying phase space restructuring and interference effects induced by driving. The results offer a unified perspective on tunneling control in non integrable systems and remain relevant for modern studies of driven quantum dynamics and decoherence resistant transport.

Understanding how large complex networks achieve synchronization is a problem of fundamental interest, and is typically studied in the asymptotic steady-state regime. In contrast, this study investigates how higher-order interactions affect the time required to reach steady-state synchronization in a complex dynamical system. To this end, an analytical expression for the first synchronization time is derived using the Ott-Antonsen ansatz on a Kuramoto oscillator network with higher-order interactions. Subsequent numerics reveal that increasing coupling strengths accelerates the transition to synchronization, whereas increasing the interaction order produces non-monotonic behavior. In particular, the inclusion of triadic interactions accelerates synchronization, whereas further incorporating higher-order interactions progressively delays convergence to the steady state, in some regimes even falling below the pairwise case.

We provide necessary and sufficient conditions for maps that satisfy associative-like conditions on families of n-ary magmas to be pentagon maps. We obtain parametric-pentagon maps and we propose a procedure that generates families of multicomponent pentagon and entwining pentagon maps from a given pentagon map.

The soliton-like coherent structure (SCS), which has been verified to exist in both transitional and turbulent boundary layers1-4, still poses a challenge in the understanding of its formation and behavior. In our previous study (Niu et al.5), the SCS was also found to exist in the transitional wake flow behind a sphere. In present study, the formation and evolution of the SCS is further investigated at four Reynolds numbers by numerical simulation. The results show that at the early stage of the turbulence transition, the SCS appears as a form of wave packet during the Tollmien-Schlichting (T-S) wave stage. With the increase of the Reynolds number, the SCS reaches its maximum amplitude downstream where the velocity discontinuity occurs. This position is located after the breakdown of the T-S wave and the three-dimensional structure is formed. Then, the SCS conserves its shape and amplitude over a long distance downstream. The relationships among the SCS, the spikes, the vortex structures, and the high-shear layers are further analyzed. It is found that the SCS in the wake flow has similarities to the phenomena observed in boundary layer flows during the turbulent transition. The vortex structures and high-shear layers mostly wrap around the border of the SCS. The vortex structure is considered to be as a consequence of the development of the SCS rather than its cause.

For piecewise-smooth ordinary differential equations, the occurrence of a Hopf bifurcation on a switching surface is known as a boundary Hopf bifurcation. Boundary Hopf bifurcations are codimension-two, so occur at points in two-parameter bifurcation diagrams. From any such point there issues a curve of grazing bifurcations, where the limit cycle born in the Hopf bifurcation hits the switching surface. For Filippov systems, these are usually grazing-sliding bifurcations whose local dynamics are dictated by piecewise-linear maps. In general, these maps have many independent parameters and extraordinarily rich dynamical behaviour. We show that for three-dimensional Filippov systems only a two-parameter family of piecewise-linear maps is relevant, because sliding motion induces a loss of dimension, and the stability of the limit cycle is degenerate at the Hopf bifurcation. We derive explicit formulas for the two parameters in terms of quantities associated with the boundary Hopf bifurcation, and perform a comprehensive numerical analysis to characterise the attractor of the family, which may be chaotic. The results are illustrated with a pedagogical example, a pest control model, and a model of a food chain with threshold-based harvesting. To evaluate the parameters, we use a formula for the linear term of the discontinuity map associated with grazing-sliding bifurcations. In this paper we present a new, simpler derivation of this formula for $n$-dimensional systems based on displacements from a virtual counterpart.

The study of dynamical systems has long focused on the characterization of their asymptotic dynamics such as fixed points, limit cycles and other types of attractors and how these invariant sets change their properties as systems parameters change. More recently, however, the importance of transient dynamics, especially of long transients and sequential transitions between them, has been increasingly recognized in various fields including ecology, neuroscience and cell biology. Among several possible origins of long transients, ghost attractors have received particular attention due to interesting dynamical properties in non-autonomous settings, new theoretical developments, and an increasing number of systems that empirically show dynamics consistent with ghost attractors. Despite this growing interest in transient dynamics generally and ghost attractors in particular, there are significantly fewer theoretical concepts and software tools available to researchers to classify and characterize their underlying mechanisms compared to asymptotic dynamics. To address this gap, we generalize saddle-nodes to account for higher-dimensional center manifolds and provide a definition for their ghost attractors. We then introduce algorithms to specifically identify and characterize ghost attractors and their composite structures such as ghost channels and ghost cycles and show how these concepts and algorithms can be used to gain new insights into the transient dynamics of a wide range of system models focusing on living systems, allowing, e.g., to describe bifurcations of ghosts. The algorithms are implemented in Python and available as PyGhostID, a user-friendly open-source software package.

Fresnel zone plates (FZPs), with patterns of $1\,μ$m resolution, allow the formation of clean, diffraction-limited foci -- but have a static phase profile. Spatial light modulators (SLMs) allow dynamic control of spatial beam intensity and phase -- but are bulky and currently limited to roughly $10\,μ$m pixel sizes and $1\,$Mega-pixel formats. Here, we present a new `best-of-both' kind of FZP, scalable to large area rings currently incompatible with direct SLM generation. It is equivalent to a plano-convex donut lens, whereby light's local intensity and global phase at the FZP map directly onto the image plane. The same FZP under different SLM illumination can generate: rings and arcs, double-rings, phase windings and ring lattices (or dynamic combinations thereof). The smooth and adaptable near-field waveguide this enables will be ideal for Sagnac interferometry with ultracold atoms.

Efficient, low-loss, and versatile optical modulators are a critical ingredient for practical integrated photonic systems. Modulators based on micro-electromechanical systems (MEMS) have unique advantages over more traditional thermal, electro-optic, or plasma dispersion modulators. In this work, we show that evanescent MEMS modulators (in which a dielectric slab is mechanically inserted into a waveguide's evanescent field) can exhibit anomalously dispersive modulation. That is, despite positive modulation of a waveguide mode's effective index, the modulator brings about a negative change in group index. We experimentally demonstrate these unique capabilities using a novel MEMS actuator design. The new theory and results here reveal that evanescent MEMS modulators possess a capability for control of wavelength dispersion not accessible to nearly any other type of modulator. These new capabilities may enable on-chip integration of systems for various optical applications, including broadband switching, photonic true time delay, pulse shaping, or phase matching of nonlinear processes.

Thermally activated delayed fluorescence (TADF) is a promising route towards high-efficiency, metal-free organic light-emitting diodes (OLEDs). However, the characterization of TADF kinetics in solid-state thin films is often complicated by pronounced multiexponential photoluminescence decays that prevent standard biexponential modeling. In this work, we introduce the 'Gamma-Fit' method, a streamlined analytical framework based on the gamma distribution that accounts for the continuous distribution of decay rates inherent in disordered molecular ensembles. By treating the decay as a result of conformational and kinetic heterogeneity, we accurately extract kinetic parameters for the benchmark emitters 4CzIPN and 5CzBN, as well as a series of novel diphenylamine (DPA)-based systems. Our results reveal that accounting for the local environment in thin films remains an important part in determining OLED efficiency. The experimental findings are complemented by a semiclassical Marcus-like computational approach. We evaluate the reliability of this conventional single-conformation rate calculation method and highlight the presence of conformational ensembles and multiple RISC-active triplet states as important factors for accurately describing the transition kinetics.

Electro-optic (E/O) modulators are crucial for optical communication but face a trade-off between modulation bandwidth and efficiency. A small footprint could reduce the capacitance and increase the bandwidth, however, this usually results in a low modulation efficiency. Here, we present an integrated E/O modulator that simultaneously achieves wideband large bandwidth and high modu- lation efficiency operation by embedding a partially overlapped double-layer graphene on a compact silicon slot waveguide microring resonator. At 1550 nm, the graphene-silicon slot-waveguide demon- strates a high phase modulation efficiency of Vπ L = 220 V μm, and the corresponding microring modulator has a large bandwidth of over 70 GHz, a compact active length of 10 μm, and an optical modulation amplitude (OMA) of -1.97 dBm under a 3-V voltage swing. The modulator operates at a data rate of 50 Gbit/s with an open eye diagram under a 2-V Vpp RF drive voltage. The graphene modulator operation is broadband, and we also characterize its performance at 2 μm wavelength band. At 2 μm wavelength band, the microring modulator has a large bandwidth of over 20 GHz, an OMA of -3.36 dBm under a 6-V voltage swing, and an open eye diagram at 20 Gbit/s with a 2-V Vpp RF drive voltage. The difference in performance is caused by the bandwidth limit of the 2 μm wavelength band measurement setup. The broadband, large bandwidth, compact, highly effi- cient, and energy efficient graphene E/O modulator has the potential to enable large-scale graphene photonic integrated circuits, facilitating a broad range of applications such as optical interconnects, optical neural networks, and programmable photonic circuits.

Physics-informed neural networks (PINNs) have emerged as a flexible framework for solving partial differential equations, but their performance on interface problems remains challenging because continuity and flux conditions are typically imposed through soft penalty terms. The standard soft-constraint formulation leads to imperfect interface enforcement and degraded accuracy near interfaces. We introduce two ansatz-based hard-constrained PINN formulations for interface problems that embed the interface physics into the solution representation and thereby decouple interface enforcement from PDE residual minimization. The first, termed the windowing approach, constructs the trial space from compactly supported windowed subnetworks so that interface continuity and flux balance are satisfied by design. The second, called the buffer approach, augments unrestricted subnetworks with auxiliary buffer functions that enforce boundary and interface constraints at discrete points through a lightweight correction. We study these formulations on one- and two-dimensional elliptic interface benchmarks and compare them with soft-constrained baselines. In one-dimensional problems, hard constraints consistently improve interface fidelity and remove the need for loss-weight tuning; the windowing approach attains very high accuracy (as low as $O(10^{-9})$) on simple structured cases, whereas the buffer approach remains accurate ($\sim O(10^{-5})$) across a wider range of source terms and interface configurations. In two dimensions, the buffer formulation is shown to be more robust because it enforces constraints through a discrete buffer correction, as the windowing construction becomes more sensitive to overlap and corner effects and over-constrains the problem. This positions the buffer method as a straightforward and geometrically flexible approach to complex interface problems.

Photonic neural networks promise ultrafast inference, yet most architectures rely on linear optical meshes with electronic nonlinearities, reintroducing optical-electrical-optical bottlenecks. Here we introduce small-scale photonic Kolmogorov-Arnold networks (SSP-KANs) implemented entirely with standard telecommunications components. Each network edge employs a trainable nonlinear module composed of a Mach-Zehnder interferometer, semiconductor optical amplifier, and variable optical attenuators, providing a four-parameter transfer function derived from gain saturation and interferometric mixing. Despite this constrained expressivity, SSP-KANs comprising only a few optical modules achieve strong nonlinear inference performance across classification, regression, and image recognition tasks, approaching software baselines with significantly fewer parameters. A four-module network achieves 98.4\% accuracy on nonlinear classification benchmarks inaccessible to linear models. Performance remains robust under realistic hardware impairments, maintaining high accuracy down to 6-bit input resolution and 14 dB signal-to-noise ratio. By using a fully differentiable physics model for end-to-end optimisation of optical parameters, this work establishes a practical pathway from simulation to experimental demonstration of photonic KANs using commodity telecom hardware.

This study investigates the influence of different types of non-pharmaceutical interventions (NPIs) on epidemic progression using SIR compartmental models. We analyze the optimization of two distinct targets: the final epidemic size and the infection peak, particularly how they respond to variations in the initiation time of the NPIs. We derive analytical approximations for the critical points of the infection curve of the standard mean-field SIR model with NPIs, and for the epidemic size, enabling a systematic comparison. The analytical results reveal the existence of six different allowed scenarios for the evolution of the epidemic with a single NPI. Furthermore, by employing degree-based mean-field network models, we distinguish between NPIs that decrease the transmission rate (individual and environmental measures) and those that reduce social contacts (lock down measures). We find that, when assuming equal effects on the reproductive number, the former are more efficient in reducing the final epidemic size. Meanwhile, the effectivities of both types of NPIs differ in reducing primary and secondary peaks. The results for all models consistently confirm that minimizing the infection peak requires earlier implementation of the NPI than minimizing the epidemic size, offering new insights for strategic public health timing.

Achieving device-to-device reproducibility is a critical bottleneck for scalable photonic integrated circuits, as subtle variations in bottom-up epitaxial growth and fabrication severely limit yield. We present a machine learning workflow for III-V multi-quantum well microring lasers that first optimizes growth and geometry parameters via multi-objective Bayesian optimization, then leverages variational autoencoders (VAEs) to attribute residual device-to-device variability to its underlying sources. By explicitly targeting threshold variance alongside absolute performance, we demonstrate 100% lasing yield across all designs. The optimized multi-quantum well microring laser fields achieved a median lasing threshold of $16~μ\mathrm{J}\,\mathrm{cm}^{-2}\,\mathrm{pulse}^{-1}$, a $73\%$ reduction in threshold variance relative to the previously reported best values, and a median emission wavelength of $1333~\mathrm{nm}$, in the telecommunications O-band. Furthermore, to diagnose residual performance dispersion under nominally identical conditions, VAEs were used to isolate the key components of device morphology that impact performance. This analysis successfully decoupled geometric from material disorder, quantitatively linking previously unmeasured morphological variations to population-level threshold fluctuations. This data-driven workflow bridges the gap between fundamental epitaxy and reliable manufacturing, establishing a generalizable blueprint for designing and yield-optimizing complex, non-linear optoelectronic devices.

We extend the beat-wave framework for laser harmonic generation - where spectra form regular lattices in Fourier space - to the nonlinear response of isotropic chiral media driven by locally chiral light. We represent the enantio-sensitive response of the medium by a chiral zero-frequency (DC) mode derived from the transverse spin density induced by structured or focused fields. Beating between this DC mode and the driving electromagnetic modes yields alternating chiral and achiral contributions on a regular harmonic lattice. We derive a general criterion for when chiral and achiral pathways overlap at the same harmonic and generate enantio-sensitive interference that survives spatial or angular integration (global chirality), versus when enantio-sensitivity remains confined to spatially varying patterns (local chirality). We apply the criterion to published configurations of synthetic chiral light, including OAM-carrying bicircular fields and crossed multicolour beams, and show that it reproduces and clarifies their reported global-chirality and beam-bending regimes.

We develop a comprehensive analysis of the Kirkwood-Dirac distributions in classical optics, revealing their deep connection with optical coherence as fundamental concept in optics. From their very definition, the Kirkwood-Dirac distributions emerge as generalized mutual coherence functions involving two different bases instead of just one. This perspective provides a unified interpretation of the so-called anomalous values, that are complex and negative values, as direct manifestations of coherence. We show that this interpretation consistently applies across all field variables considered in this work, including polarization, interference and wave propagation. Furthermore, we propose diverse methods of experimental determination of these distributions based on interference, in full agreement with their coherence-based interpretation.

Chirality plays a crucial role in a broad range of processes including light-matter interactions in physics, chemistry and biology, which opens up new applications in nanophotonics, quantum technologies and photochemistry. Quantum tunnelling provides a promising mechanism for light generation at the nanoscale, however the realisation of chiral light emission has remained elusive. Here, by integrating tunnel junctions with chiral plasmonic nanohelicoids, we achieve nanoscale generation of chiral light at a single-particle level. The tunnelling-driven resonant excitation of chiral dipolar modes of the nanohelicoids results in emission of a vortex light beam possessing both spin angular momentum with handedness selectivity of over 0.8 and its orbital counterpart, equal in magnitude and opposite in sign. The developed approach offers a new means for sculpturing photon spin generation at the nanoscale, highlighting its potential for next-generation optical components in display and AR/VR applications, as well as quantum information processing and photochemistry.

This paper presents SPIROS (Streamlined, Precise, Intuitive, and Rapid Optical Simulator), a dedicated optical simulation tool developed for the design and analysis of particle physics detectors. Unlike general-purpose frameworks such as GEANT4, SPIROS offers a lightweight simulation engine and a user-friendly interface optimized for optical processes, including scintillation, Cherenkov emission, and photon transport with reflection, refraction, scattering, absorption, and detection. Detector geometries can be directly imported from 3D CAD models, and all configurations including materials, surfaces, sources, and sensors are specified via a single human-readable input file. Validation against GEANT4 shows excellent agreement in photon generation and propagation behaviors, while benchmark tests demonstrate that SPIROS runs more than two times faster for typical detector configurations. The software has already been applied to multiple neutrino experiments, including T2K, NINJA, and AXEL, for detector design, performance studies, and optimization. SPIROS is open-source and freely available at https://github.com/tkikawa/spiros.

Compact, low-noise coherent light sources in the visible remain challenging due to limited gain platforms and inefficient pumping. We report a new route to visible microlasing based on direct, one-photon blue pumping and an amorphous fluoride gain material platform. Dysprosium doped fluoride microspheres are fabricated via plasma-torch-induced, pressureless amorphization of single crystals, enabling compositions beyond conventional glass-forming limits while ensuring ultrasmooth morphology, low phonon energy, and homogeneous dopant distribution. We demonstrate the first fiber-coupled whispering-gallery-mode lasing from an amorphous fluoride microsphere in the yellow (573 nm), with an ultralow threshold of $190 μ$W despite spin-forbidden Dy$^{3+}$ transitions. Lasing is evidenced by characteristic light-light curve indicating a low spontaneous emission factor, narrow-linewidth emission, and relaxation oscillations yielding a loaded quality factor of $Q = 3.5 \times 10^6$. This platform is readily extendable to other rare-earth emitters, enabling entire visible spectral coverage beyond the limitations of upconversion pumping, with prospects for color-tunable and white-light emission. Finally, fiber-based amplification of the WGM signal demonstrates a pathway toward compact, fiber-integrated visible microlasers with controllable noise and linewidth.

Whispering-gallery-mode (WGM) microsphere resonators have emerged as a versatile platform across various photonic applications. Despite significant progress, their performance at short wavelengths is fundamentally limited by scattering-induced optical losses that restrict achievable quality factors (Q-factor). Although surface roughness has long been recognised as the leading cause of these losses, its physical origin has remained unclear, with current understanding attributing it to unavoidable fabrication imperfections. Here, we show that thermally excited capillary waves are the fundamental source of scattering losses in microsphere cavities. Using high-resolution atomic force microscopy (AFM) combined with rigorous statistical analysis, we quantitatively identify the characteristic signatures of frozen capillary fluctuations at the sub-nanometre level. The experimentally extracted roughness parameters show close agreement with theoretical predictions based on capillary wave theory. These findings fundamentally revise the prevailing interpretation of surface scattering losses and establish thermodynamic fluctuations, rather than fabrication defects, as the limiting roughness mechanism. By identifying frozen capillary waves as the limiting factor, this work opens new pathways for engineering ultra-high-Q microsphere resonators through fabrication management strategies, particularly for visible- and ultraviolet-photonic applications where scattering losses are most severe.

The rapid advancement of deep learning is reshaping the hardware design landscape toward AI tasks, posing fundamental challenges for HPC workloads such as atomistic simulation. Here we present SMC-AI, a general algorithmic framework that extends the SMC-X method for efficient canonical Monte Carlo simulation on AI accelerators, including GPUs and NPUs, while maintaining extreme scalability. The implementation of SMC-AI on an NPU cluster reaches unprecedented performance, achieving MC simulation of 4 trillion atoms on 4096 NPU dies. This represents the largest ML-accelerated atomistic simulation reported, delivering 32X system size and 1.3X throughput than previous records, with a relatively small computational budget. Excellent strong and weak scaling efficiency are reached for both the NPU and GPU implementation. By decoupling ML models from simulation, SMC-AI creates an abstraction that facilitates integration and porting of diverse ML models, laying a foundation for the future development of scalable scientific software.

The Majoron is a hypothetical (pseudo) Nambu-Goldstone boson arising from the spontaneous breaking of a global lepton number symmetry, and is known as a candidate for dark matter in our Universe. In this paper, we investigate the possibility of probing the Majoron dark matter with a linear optical cavity used in the interferometric gravitational wave detectors. We consider a scenario in which the Majoron dark matter couples to photons through a QED anomaly, leading to an oscillatory photon birefringence induced by the coherent dark matter background. The anomaly coefficient is fixed by requiring the model to simultaneously reproduce the electroweak Higgs scale and a typical right-handed Majorana neutrino mass scale, and the resulting dark matter-photon coupling naturally falls within the sensitivity range of optical interferometers. By incorporating additional optics to extract the birefringence signal, we find that ground-based laser interferometers such as Advanced LIGO, KAGRA, as well as future detectors, can probe a region of the parameter space of Majoron dark matter.

We develop a theoretical description of quantum wave mixing (QWM) in a cascaded waveguide-QED system of two superconducting qubits, where the probe is driven by an external coherent tone and by the resonance fluorescence of a strongly driven source qubit. Starting from the field correlation functions of the source emission, we derive an effective master-equation treatment for the probe and identify the regime in which the incident fluorescence is characterized by anomalous correlations. When the coherent Rayleigh component of the source spectrum is suppressed, the probe equations of motion become equivalent to those for a qubit driven by a coherent tone and broadband squeezed light. This equivalence implies a selection rule for the peaks of the QWM spectrum, with a strong suppression of sidebands associated with processes involving an odd number of photons taken from the source field. Numerical simulations of the full cascaded two-qubit model for different ratios of radiative decay rates unambiguously confirm the participation of correlated photon pairs in QWM processes. The current research illustrates that the analysis of peak amplitudes can be used to probe photon statistics in the incident nonclassical field.

The control of reflection and refraction at interfaces using engineered media is central to numerous optical technologies, with negative refraction and the suppression of backscattering representing two prominent research frontiers. In this work, we demonstrate that an effective negative refraction accompanied by an absence of backscattering can be realized at a moving spatiotemporal interface when temporal and spatial reflections occur concurrently. While such spatiotemporal co-reflection is prohibited in one-dimensional linear dispersive media, we show that it becomes permissible under oblique incidence within a specific range of traveling-wave modulation velocities. Leveraging this mechanism, we propose a spatiotemporal flat lens capable of nonreciprocal electromagnetic wave focusing. These findings provide a framework for developing advanced spatiotemporal metamaterials and time-varying metasurfaces.

Classical topological descriptors used in topological data analysis (TDA) are invariant under permutations of spatial axes and therefore cannot represent the loading direction, which is essential for modeling anisotropic mechanical response. Here, this limitation is addressed by introducing a \emph{direction-aware TDA framework} in which the compression axis is explicitly embedded into filtration functions used to compute both persistent homology and Euler characteristic profile descriptors. Across multiple porous-material datasets spanning a broad range of structural anisotropy, direction-aware descriptors yield higher predictive accuracy than their direction-agnostic counterparts, with performance gains that increase systematically with anisotropy. Notably, direction-aware descriptors remain competitive and often improve $R^2$ even for nominally isotropic ensembles, indicating sensitivity to mechanically relevant directional organization beyond bulk anisotropy measures. When used as inputs to gradient-boosted tree models, the proposed descriptors approach the accuracy of convolutional neural networks trained directly on voxelized structures while retaining a compact, transferable representation. The study considers multiple datasets spanning weak to strong anisotropy, enabling systematic validation of direction-aware topology across regimes. Overall, the results establish direction-aware TDA as a general route for linking porous structure to direction-dependent elastic properties and motivate its use in anisotropic materials modeling problems where a preferred direction naturally arises.

Multi-agent reinforcement learning serves as an effective tool for studying strategy adaptation in evolutionary games. Although prior work has integrated Q-learning with reputation mechanisms to promote cooperation, most existing algorithms adopt fixed exploration rates and overlook the influence of social context on exploratory behavior. In practice, individuals may adjust their willingness to explore based on their reputation and perceived social standing. To address this, we propose a Q-learning model that couples exploration rates with local reputation differences and incorporates asymmetric, state-dependent reputation updates. Our results show that each mechanism independently promotes cooperation, and their combination yields a reinforcing effect. The joint mechanism enhances cooperation by making ``high reputation--low exploration, low reputation--high exploration'', while adjusting reputation updates to amplify cooperative gains at low status and defection penalties at high status. This study thus offers insights into how social evaluation can shape learning behavior in complex environments.

Convolutional Neural Networks (CNNs) excel at extracting local features hierarchically, but their performance in capturing complex correlations hinges heavily on deep architectures, which are usually computationally demanding and difficult to interpret. To address these issues, we propose a physically-guided shallow model: tensor-augmented CNN (TACNN), which replaces conventional convolution kernels with generic tensors to enhance representational capacity. This choice is motivated by the fact that an order-$N$ tensor naturally encodes an arbitrary quantum superposition state in the Hilbert space of dimension $d^N$, where $d$ is the local physical dimension, thus offering substantially richer expressivity. Furthermore, in our design the convolution output of each layer becomes a multilinear form capable of capturing high-order feature correlations, thereby equipping a shallow multilayer architecture with an expressive power competitive to that of deep CNNs. On the Fashion-MNIST benchmark, TACNN demonstrates clear advantages over conventional CNNs, achieving remarkable accuracies with only a few layers. In particular, a TACNN with only two convolution layers attains a test accuracy of 93.7$\%$, surpassing or matching considerably deeper models such as VGG-16 (93.5$\%$) and GoogLeNet (93.7$\%$). These findings highlight TACNN as a promising framework that strengthens model expressivity while preserving architectural simplicity, paving the way towards more interpretable and efficient deep learning models.

Integrated photonics is a powerful platform for exploring Hermitian and non-Hermitian physics. Beyond device geometry, controlling how resonators are driven is crucial to access and tailor their modes. Coherent excitation via multiple input ports (interferometric excitation) enables such control, but its accurate description requires extending standard temporal coupled-mode theory to include interference between excitation pathways. Experimental realizations have so far been limited by phase-unstable, off-chip interferometers. Here we implement a fully integrated, phase-stable interferometric excitation scheme for an infinity-loop-microresonator, an established structure operating on an exceptional surface, and use it to test the extended theory. Phase-resolved measurements in the linear and thermo-optic nonlinear regimes show that the relative phase between inputs governs the intracavity energy distribution, enabling up to a twofold increase of the circulating power compared to single-port excitation. This integrated platform enables reproducible studies of phase-dependent effects and coherent-control schemes in non-Hermitian photonic devices.

Autonomous electrolyte discovery demands a computational engine that satisfies a critical trilemma: it must be fast enough for high-throughput screening, accurate enough for quantitative property prediction, and calibratable enough for online refinement. Classical empirical force fields (FFs) are fast but rely heavily on error cancellation, while standard machine learning interatomic potentials (MLIPs) are computationally expensive, lack rigorous long-range physics, and resist gradient-based calibration. In this Perspective, we highlight that differentiable hybrid FFs resolve this trilemma by fusing physically motivated functional forms with neural-network short-range corrections. Grounded in Energy Decomposition Analysis (EDA), state-of-the-art models such as PhyNEO-Electrolyte and ByteFF-Pol achieve zero-shot generalization to bulk phases, delivering throughputs on the order of tens of ns/day (up to $\sim$50 ns/day, depending on model complexity) for 10,000-atom systems. Crucially, their physical skeletons provide a well-conditioned parameter space for differentiable molecular dynamics (dMD). This enables a dual-calibration paradigm: bottom-up \textit{ab initio} parameterization combined with top-down fine-tuning from macroscopic experimental observables. We propose that this architecture meets the requirements of a ``ChemRobot-ready'' digital twin by integrating physics-grounded simulation with experimentally calibratable refinement, thereby enabling closed-loop autonomous electrolyte discovery.

In moiré superlattices, the band flatness governs the degree of wave localization, which is central to harnessing emergent phenomena and designing functional meta-devices. While research has focused on the magic conditions such as magic angle and magic distance for optimal flatness, a fundamental understanding of how flatness changes with the supercell size has remained elusive. Here, we establish a universal scaling between band flatness and supercell size. Theoretically, by recognizing the statistical equivalence between structural perturbations in moiré superlattices and disordered systems, we introduce the Thouless number to evaluate the strength of moiré localization. This approach allows us to establish a scaling theory for the evolution of band flatness with the supercell size, from which an analytical expression is derived. Our full-wave simulations with one-dimensional and two-dimensional moiré superlattices show excellent agreement with the theoretical prediction. Our work reveals a general scaling law for moiré band flatness, offering a new perspective for understanding and designing moiré-based resonant systems.

We propose and analyse a quantum thermal field-effect transistor (qtFET) composed of left-qubit, middle-qutrit, and right-qubit subsystems. In this architecture, the left qubit is coupled to the middle qutrit, which in turn interacts with the right qubit. Each subsystem interacts independently with its respective baths. The middle subsystem serves as a modulator. We have shown that the qtFET exhibits functionality analogous to that of a conventional electronic field-effect transistor (eFET). The left, right, and middle subsystems of the qtFET correspond to the drain, source, and gate of an eFET in a common gate configuration, respectively. Our results show that the qtFET can precisely modulate thermal currents, highlighting its potential as a fundamental building block for quantum thermal devices and amplifiers in emerging quantum technologies.

Space, time, and additional dimensions spawn remarkable complexity in optics. We encourage pairing black-box simulation and design tools with a complementary tool: understanding.

Background Double Kissing (DK) Crush is a two-stent technique for complex coronary bifurcation lesions, yet the biomechanical influence of plaque on its performance remains poorly understood. This study developed a computational biomechanical model of the DK-Crush procedure to quantify how plaque presence and composition affect procedural outcomes and the performance of Xience Sierra and Orsiro stents. Methods A population-representative coronary bifurcation was modelled with no plaque, lipid plaque, and fibrous plaque. The complete DK-Crush sequence was simulated using finite element analysis for both stent platforms. Mechanical outcomes included arterial wall stress, malapposition, side branch ostium clearance, and residual stenosis. Post-deployment hemodynamics was assessed using pulsatile computational fluid dynamics, quantifying high shear rate volume and lumen area exposed to low time-averaged endothelial shear stress (TAESS). Results Plaque presence and stiffness reduced lumen restoration, increased arterial wall stress, led to larger high shear rate regions and, for fibrous plaque, increased exposure to low TAESS. Malapposition and ostial clearance depended mainly on stent design. Plaque also altered the relative performance of the two platforms, revealing differences not observed in plaque-free models. Conclusions Plaque characteristics substantially affect DK-Crush biomechanics and modify stent behaviour. Incorporating plaque is therefore essential for realistic computational evaluation of bifurcation stenting.

Ghost imaging was first demonstrated with entangled photon pairs and well-known for its peculiar properties. The signal beam that illuminates the object possesses no spatial resolution, whereas the reference beam, which never interacts with the object, is spatially resolved. Either beam alone cannot retrieve the image, which can only be obtained when the signal and reference beams are correlated. Here we will report a ghost imaging experiment with even more peculiar properties, in which the image can be reconstructed when no photon interacts with the object or even no photon in neither signal nor reference beam. All the photons interacted with the object are discarded. Only the time bins with zero photon are employed to retrieve the image, a process referred to as "ghost imaging with zero photons" hereafter. The reason why ghost image can be retrieved with zero photons is jointly determined by photon-number projection measurement and photon statistics of thermal light. The results are helpful to resolve the debate on the physics of ghost imaging and understand the relation between quantum and classical correlations.

Understanding interphase mass transfer is essential for a variety of applications in porous media, ranging from groundwater remediation to geologic energy storage. While X-ray micro-Computed Tomography (microCT) provides critical in situ observations, analyzing mass transfer requires models and workflows compatible with the limited spatial and temporal resolution. Current literature presents three analytical frameworks for evaluating interphase mass transfer using microCT data: the Slice-Averaged Concentration (SAC) approach, the Non-Classified per-Cluster (NPC) approach, and the Classified per-Cluster (CPC) approach. This study evaluates the results of all three approaches across four sets of time-lapse tomography sequences that observe hydrogen dissolution at varying solvent injection rates. To mitigate biases arising from dissolution-driven cluster remobilization, we introduce a volume-ratio filtering technique to all workflows to ensure that estimates more accurately reflect true mass transfer events. Our analysis finds that all three analytical approaches estimate average mass transfer coefficients within one order of magnitude of one another at the same solvent injection rate. However, the similarity between the estimates of each approach diverges when approximating more complex phenomena, such as aqueous solute concentration profiles. Ultimately, the utility of one approach over another is determined by the desired level of system detail, at the cost of the computational resources required to achieve it. Higher phenomenological resolution requires greater computational processing and refinement due to increased sensitivity to measurement and processing noise, as well as outlier events. We anticipate that the findings will provide a framework for researchers to match analytical approaches to their available computational resources and desired level of physical detail.

The manipulation and sorting of chiral nanoparticles are of fundamental importance in multidisciplinary fields ranging from biochemistry to nanophotonics. In this study, we propose a novel and controllable chirality sorting mechanism for continuous particle separation using focused topological optical quasiparticles. Specifically, we investigate the sorting dynamics driven by tight-focused optical skyrmions and bimerons consisting of tailored spatial modes. By highly focusing free-space topological structure light fields, we generate intricate non-paraxial focal fields with tailored intensity and topological polarization textures. The sorting dynamics are systematically evaluated under the dipole approximation for fused silica nanoparticles. Our analytical calculation demonstrate that optical forces exert opposite directional pushes on particles of opposite chiralities, enabling highly efficient spatial separation. Notably, we demonstrate that this sorting process is controllable; by tuning the topological charges, the sorting distance can be flexibly tailored and expanded. The dynamic sorting process in customized topological structures introduces a promising new paradigm for tunable, wide-range chirality sorting of micro- and nano-particles.

Temporal hypergraphs capture time-resolved group interactions among nodes. Empirical data support that time-stamped group interactions show bursty event sequences and non-trivial temporal correlations. In the present study, we introduce node-driven temporal hypergraph models in which each node stochastically alternates between low- and high-activity states, and a hyperedge produces time-stamped events with a probability that depends on the number of high-state nodes in the hyperedge. For two event-generation rules, we analytically derive interevent time distributions and autocorrelation functions of event sequences, both for hyperedges and nodes. Despite Markovian node-state dynamics, the induced event processes become mixtures of Poissonian, short-tailed components, resulting in longer-tailed interevent time distributions and slowly decaying autocorrelation. The theory further shows the dependence of these features on the size of hyperedge, which largely agrees with various empirical data. We expect our models to provide a simple, interpretable framework for connecting individual-level activity fluctuations to the timing patterns observed in real group interactions.

Light in the mode of a Fabry-Pérot cavity heats the mirror surfaces via optical absorption, causing thermoelastic deformation of the mirror substrates, which in turn dictates the shape of the mode. We develop an analytical model which predicts that this effect limits the maximum focal intensity of the mode. Using two near-concentric Fabry-Pérot cavities -- one with 4.5-fold higher mirror absorption than the other -- we measure the thermoelastic properties of the cavity mirrors and demonstrate that it is possible to achieve at least 70% of this predicted limit (in the high-absorption cavity), and that the predicted limit is 2.9 TW/cm^2 (in the low-absorption cavity).

This paper introduces an optically controlled 4H-SiC MOSFET designed to avoid the gate-oxide interface unreliability and electromagnetic interference (EMI) susceptibility inherent in conventional voltage-driven devices. By replacing the conventional gate electrode with a semi-transparent optical window, the device enables direct modulation of channel conductivity through ultraviolet illumination. Electrical and optical characterization demonstrates that under an optical power density above 0.1 W/cm^2, the device achieves an on/off current ratio exceeding 10^6 between illuminated and dark states. Notably, at an optical power density of 0.031 W/cm^2, the photogenerated current density exceeds that obtained under a gate bias of 15 V in magnitude. Energy band analysis confirms that the optical switching mechanism operates through direct photogenerated carrier generation and transport, fundamentally differing from conventional gate voltage control and thus circumventing interface-trap and EMI-related limitations. Dynamic measurements further reveal fast switching capability, with a rise time of 1.44 ns. These results validate the feasibility of optically driven switching in SiC-based devices and highlight their potential for high-speed logic applications.

Bound states in the continuum (BICs) have enabled lasers with rich momentum-space textures. However, the output patterns of quasi-BIC lasers remain largely static and confined to a few geometries. Here, a reconfigurable momentum-space vectorial laser was proposed based on two-dimensional photonic crystal. By selectively exciting quasi-BIC modes, we identify the geometric asymmetry factors favoring single BIC, dual-BIC, and radiative mode with BIC operation. This approach yields vectorial lasing with characteristic patterns lasing in momentum space of bidirectional double lobes (BDL), radially polarized ring with BDL, azimuthally polarized ring with BDL, and linearly polarized spot with BDL. Importantly, reversible switching between a single donut and a donut with BDL was achieved in the same device by varying the pump energy density. Our work establishes a compact, versatile platform for reconfigurable vectorial lasers, with potential applications in tunable optical tweezers, super-resolution imaging, and on-chip optical interconnects.

Offering terahertz of bandwidths and femtosecond timescales, ultrafast optics is enabling both the study of fundamental quantum optical phenomena and the advancement of quantum-enhanced applications. However, unlocking the full potential of ultrafast quantum optics requires accessing the temporal characteristics of ultrashort quantum pulses across ultrabroad bandwidths. This is particularly important in the near-infrared and visible range of the optical spectrum, which, unlike the terahertz and long-wave infrared, has remained beyond the reach of current techniques. Here, we break this barrier by translating frequency-resolved optical gating (FROG), a widely used technique for ultrafast classical pulse characterization, to the quantum regime. We show how such a quantum FROG can measure complex temporal modes and sub-optical-cycle quadrature covariances in the near-infrared, enabling complete characterization of microscopic Gaussian states. We experimentally use the quantum-FROG to report the measurement of quadrature correlations, complex temporal modes, and squeezing levels of multimode ultrafast squeezed vacuum states generated on a nanophotonic chip. We access multimode squeezing levels of a femtosecond quantum pulse approaching 7 dB and demonstrate FROG-based measurement bandwidths exceeding 100 THz. Quantum FROG enables measurement of previously inaccessible quantum features of ultrashort pulses at the sub-optical-cycle regime and highlights a practical path to accessing terahertz of bandwidths in quantum optics for applications in computing, sensing, and imaging.

The quantum dynamics of quasiperiodic systems display a rich variety of physical behaviors due to the combination of rotational symmetry that is mathematically forbidden in periodic systems, and long-range order despite the lack of translation symmetry. New experimental probes into these dynamics with a quantum simulator, consisting of ultracold atoms in an optical lattice potential, will yield new insights into the physics of quasiperiodic systems. This potential is imbued with the flexibility, tunability, and purity of the individual laser beams that constitute it, allowing for exquisite control over a rich system. Programmable dynamic control over the lattice beam phases opens up an even richer space of achievable systems via Floquet engineering. We thus describe an experimental scheme for creating a programmable, dynamic, two-dimensional (2D) quasiperiodic optical lattice with heavily suppressed phase noise. We observe suppression of phase noise for frequency components up to 5 kHz, and report phase noise suppression of over 70 dB over the DC-60 Hz frequency band. We further demonstrate a phase modulation bandwidth of 350 kHz. This scheme allows for full translational and phasonic control of the lattice, including changes to the rotational symmetry of the potential, at speeds exceeding the lattice recoil velocity, which paves a path towards direct observation and control of quantum dynamics in quasicrystals.

We introduce a CUDA-Accelerated Timestepper for Alpha Particles Using Local Tricubics (CATAPULT) for use in Monte Carlo calculations of alpha particle confinement in stellarators. Our GPU implementation is significantly faster than existing parallelized CPU implementations, and handles both equilibrium magnetic fields and Shear Alfven Waves. We test our implementation on several example stellarators to exhibit both the speed and correctness of our code. The source code is included in the firm3d Python package.

Due to the inhomogeneous polarisation across the beam profile, cylindrical vector beams interact with optically active media in a complex manner. Here, we analyse evolution of polarisation of cylindrical vector beams propagating in an isotropic optically-active medium. After identifying polarisation normal modes of three-dimensional electromagnetic fields, we predict periodic inter-conversion between azimuthally- and radially-polarised modes of the beams accompanied by rotation of the transverse optical spin and pulsing field during the propagation. Theory and simulations are validated by experimental observations. The observed effects maybe important for imaging in biological chiral media, enhanced chiral sensing and enantioselective spectroscopy, nonlinear optics in chiral media, and generally enhanced spin-orbit coupling and nanoscale vector field engineering.

Free-space optical communications (FSOC) are a key enabling technology for future high-capacity space-based networks. Particularly, the backbone of global communication relies on intersatellite optical links. In a previous study, the authors proposed a method to mitigate the impact of transmitter pointing jitter by using a superposition of orthogonally polarized Gaussian and higher-order Laguerre-Gaussian (LG) beams. In this study, we experimentally characterize the proposed system using a spiral phase plate (SPP) to generate higher-order annular beams. We demonstrate that such superpositions can be reliably generated in a realistic optical setup, quantify the associated beam-shaping errors and losses, and assess their impact on intersatellite optical communication performance. It is found that the proposed beam-shaping approach can still yield power savings on the order of 20% compared to a conventional Gaussian beam under the considered conditions.

We report experimental results on second-harmonic generation in a z-cut quartz crystal under conditions of free-space quasi-phase matching in a multi-pass cell. In a 62-pass configuration, an efficiency of 0.027% or 1.4x10-4 %/MW/cm2 was achieved, delivering 1 uJ of the second harmonic at 3.7 mJ pump pulse. This corresponds to an enhancement factor of more than 1000 in conversion efficiency as compared to a single pass. The generated second-harmonic beam demonstrates high beam quality M2=1.1 and linear polarization. The scaling of the output power with the number of passes is in good agreement with the calculated values. Further increasing the pump intensity, number of passes, and amount of plates opens the way to scaling the conversion efficiency to values on the order of tens of percent.

Magnetic dipole (MD) transitions are intrinsically weak and highly sensitive to emitter orientation and position, making their controlled enhancement at optical frequencies particularly challenging. Here we show that structural symmetry provides a powerful route to robust magnetic light-matter interactions. We systematically investigate plasmonic core-satellite resonators composed of N metallic nanoparticles arranged on a dielectric core. We evaluate their performance using a unified figure of merit that accounts for magnetic Purcell enhancement, electric dipole suppression, quantum efficiency, and robustness to emitter orientation and fabrication tolerances. We find that the optimal structures correspond to the highest-symmetry geometries, which naturally produce spatially homogeneous and orientation-independent magnetic Purcell enhancement. In particular, the dodecapod configuration yields strong magnetic emission with Purcell factors approaching 250, high radiative efficiency, and suppressed electric dipole contributions. Quasinormal-mode and complex mode-volume analysis reveal that symmetry enforces uniform magnetic modal confinement within the core, explaining both the enhancement and its robustness. These results establish symmetry as a guiding principle for designing nanophotonic resonators with controlled multipolar light-matter interactions and provide a practical route toward bright and selective magnetic dipole emitters.

Commercial fusion energy requires frameworks to assess both the scientific and economic viability of a wide variety of fusion concepts. Inspired by the Lawson criterion's ability to universally describe fusion energy gain, a generalized framework is developed to determine the economic gain of fusion power plants. The model exploits temporal equilibrium, and engineering and cost parameters normalized to the energy capture surface. The derived criteria for economic gain are therefore independent of the power plant's absolute power, impartial to the particulars of its fusion technology, and can be applied to any fusion confinement concept. The derivation of the economic gain factor, $Q_{econ}$, results in nonlinear equations with ten controlling normalized design parameters ranging from fusion power density and surface component lifetime to energy fluence, price of energy, and component efficiency and cost. These ten controlling parameters are varied over a wide range to provide high-level insights in design, finance and operational tradeoffs that improve the prospects for economically viable fusion energy.

Recent cosmological data reveal tension between parameters inferred from measurements of the cosmic microwave background (CMB), baryon acoustic oscillations (BAO), and supernovae (SN) under $Λ$CDM. Typical dynamical dark energy parameterizations (such as $w_0w_a$) that seek to jointly resolve these tensions have an equation of state parameter that crosses into the phantom regime, leading to potential instabilities for physical models. We show that the BAO (early-time) and SN (late-time) sides of the tension can instead be treated independently. Early dark energy (EDE) can reduce the tension between CMB-BAO data by changing the calibration of the sound horizon at the drag epoch $r_d$, with a $Δχ^2 = -{9.4}$ relative to $Λ$CDM, raising $H_0$ to 70.87 $\rm km s^{-1}Mpc^{-1}$. EDE alone cannot bring consistency between CMB, BAO, and SN data, but combining with a thawing-quintessence component of dark energy reduces tensions between the three datasets, with $Δχ^2=-12.6$ relative to $Λ$CDM without a phantom component, vs. $Δχ^2=-15.8$ for $w_0 w_a$ with one. We consider different SN datasets, using the most recent DES Dovekie catalog as our default while assessing differences with the original DESY5 and Pantheon+ catalogs. While the significance of adding thawing quintessence changes, the EDE solution to the CMB-BAO tension remains nearly unaffected. Moreover, though we do not include direct Hubble constant measurements in these $Δχ^2$ values, the EDE solution reduces the Hubble tension with the Local Distance Network value from $7σ$ in $Λ$CDM to $2-3σ$ depending on the SN dataset, nominally the equivalent of an extra $Δχ^2 \sim -40$ or more.

We study the non-Gaussian features of the gravitational wave (GW) background generated by a population of inspiraling supermassive black hole (SMBH) binaries. We show that the SMBH GW amplitude distribution (GWAD) features a universal heavy power-law tail $\propto A^{-4}$, while the low-amplitude tail depends on the SMBH merger rate and the energy-loss mechanisms of the binaries. The distribution of the induced timing residuals inherits this heavy tail. As a result, the ensemble averaged statistical moments of order three and higher diverge, limiting their usefulness as measures of non-Gaussianity, and the GW background from SMBH binaries exhibits the single loud source principle, according to which the strongest signals are more likely to be caused by a small number of loud sources. We confirm that the variance-averaged Gaussian approximation accurately describes the timing residual statistics. This approximation justifies a factored likelihood structure that combines standard Gaussian-process PTA posteriors with the non-Gaussian population prior, enabling consistent incorporation of non-Gaussian effects into SMBH model inference. We provide a fast and flexible Python implementation to compute the distribution of timing residuals from a given SMBH merger rate or GWAD.

We analyse the effects of a non-standard evolution of the Universe during the reheating epoch on the spectrum of scalar-induced gravitational waves (SIGWs) accounting for the presence of primordial non-Gaussianity. We show that given values of $w$ and $c_s^2$ leave characteristic features in the spectrum which can be detectable by third generation interferometers like LISA. In addition, we argue that the specific reheating dynamics can suppress or even enhance the spectrum, with crucial consequences for its detectability. We perform a Fisher forecast for different values of $w$ and different scans to assess the detectability of the signal when different values of the amplitude and central frequency are considered.

The cosmological gravitational wave background provides a powerful window on parity-violating physics at energies far beyond the reach of terrestrial experiments. However, any co-located planar detector network is insensitive to isotropic circular polarization, independent of its relative orien- tation. In this letter, we show that this no-go result can be evaded by a new class of co-located 3D interferometer designs, which we call Pyramid, whose non-coplanar configuration geometrically isolates chirality. This new design is a natural extension of the third generation of gravitational wave detectors. The coplanar correlation channel is blind to circular polarization, whereas the co-located non-coplanar channel is insensitive to the unpolarized background and acquires a response only in the presence of nonzero net helicity. Pyramid interferometers therefore furnish a unique probe of cosmological gravitational-wave chirality, opening a realistic terrestrial pathway to test parity violation and fundamental symmetry breaking in the early Universe.

We present a novel, JAX-powered implementation of a parametric component-separation method for CMB polarization data, explicitly designed to handle spatially varying foreground Spectral Energy Distributions (SEDs). The approach models this variation across the sky by grouping sets of pixels that share common foreground spectral parameters, scanning over thousands of such configurations to evaluate the trade-off between model complexity and residual systematic contamination. Built within the FURAX framework -- a JAX-powered environment for CMB data analysis -- our pipeline extends the fgbuster parametric formalism. It enables fully vectorized, GPU-accelerated evaluation of the spectral likelihood, map reconstruction, and diagnostic metrics across tens of thousands of pixel subset configurations, noise realizations, and sky regions. Our implementation achieves up to $\sim 100\times$ speed-up over the scipy TNC optimizer used in fgbuster when running on GPUs, as well as giving more robust results. When applied to LiteBIRD-like simulations with spatially varying foreground SEDs, our optimized K-means configuration reduces the 68% upper limit on the tensor-to-scalar ratio $r$ by $\approx 30\%$ relative to a fixed, previously derived multi-resolution configuration, while maintaining competitive statistical uncertainties.

Motivated by the recent Dark Energy Spectroscopic Instrument (DESI) DR2 preference for dynamical dark energy, we study interacting dark energy models in which a canonical quintessence field couples to cold dark matter through a field-dependent mass $m(φ)$. In such scenarios, the effective equation of state inferred under the assumption of non-interacting dark sectors, $w_{\rm eff}(z)$, can differ from the intrinsic scalar-field equation of state $w_φ(z)$, making an apparent phantom crossing $w_{\rm eff}<-1$ possible without introducing a phantom scalar. We show that a viable realization of this mechanism requires the scalar field to originate from a frozen phase deep in the radiation era, in order for the effective coupling to remain sufficiently suppressed before recombination to evade cosmic microwave background constraints, and for the late-time evolution to become strong enough to reproduce the apparent behavior of $w_{\rm eff}(z)$ preferred by DESI. We identify the general conditions that allow these requirements to be satisfied simultaneously, and present an illustrative phenomenological realization in which $w_{\rm eff}(z)$ evolves from $w_{\rm eff}\approx -1.2$ at $z \approx 1.0$ to $w_{\rm eff}\approx -0.9$ at $z\approx 0.4$. These conditions and requirements serve as a guide for designing future models of this kind which can safely navigate the phantom divide at $w=-1$ in an effective way without phantom fields.

This paper investigates the shadow and polarization images of a Kerr-Sen black hole illuminated by geometrically thick and optically thin accretion disks. We adopt two classes of accretion models, namely the phenomenological radiatively inefficient accretion flow (RIAF) model and the analytical ballistic approximation accretion flow (BAAF) model. Based on radiative transfer theory, we examine the effects of the spin parameter $a$, black hole charge $Q$, and observer inclination angle $θ$ on the shadow images. Both models show that, as the charge $Q$ increases, the photon rings and the central dark regions shrink simultaneously. Meanwhile, frame dragging gives rise to a pronounced brightness asymmetry, which becomes more significant with increasing $a$ and $θ$. The main difference between isotropic and anisotropic radiation is that, in the latter case, the higher order images are brighter in the upper and lower polar regions. For the BAAF model, because the conical approximation renders certain regions geometrically thinner, the spatial extent of the higher order images is narrower than that in the RIAF model, and the separation between the direct image and the higher order images is more distinct. In the polarization images, the spatial distribution of the polarization vector directions is mainly determined by gravitational lensing and frame dragging, whereas the intensity near the photon ring and the scale of the higher order images are significantly influenced by $Q$.

We combine Gaia DR3 5-parameter astrometry with calibrated radial velocities for 23 nearby (<1 kpc) young (<60 Myr) clusters, with membership lists from Cantat-Gaudin et al. (2020). We characterise the plane-of-sky structure of the clusters using Q-Parameter and Angular Dispersion Parameter (ADP) methods. We measure plane-of-sky expansion using several methods. We determine plane-of-sky orientations along which expansion is maximised. We also estimate expansion timescales and traceback ages and compare to isochronal ages. We then look for correlations between cluster properties and discuss sample-wide trends. We find that most young clusters are more smoothly structured in their centers where the rate of dynamical interactions is highest, while hierarchical structure can survive in the sparse outskirts for >10 Myr. We also find that the majority of nearby young clusters exhibit clear signatures of expansion in the plane-of-sky, which in many cases is significantly anisotropic, even at ages >30 Myr. We find evidence that older clusters tend to have directions of maximum expansion oriented closer to parallel with the Galactic plane. The high degree of spatial structure and significant expansion anisotropy imply that the majority of these young clusters have formed with significant spatial and kinematic substructure and not as dense, monolithic clusters. Kinematic ages estimated from expansion timescales and on-sky traceback are generally in good agreement with estimates inferred from stellar evolution models for clusters <10 Myr old. However, many clusters with older isochronal ages appear to have significantly younger kinematic ages. We discuss potential reasons for this discrepancy, including a prolonged embedded and/or gravitationally bound phase in the early stages of the clusters.

As statistical systems, galaxies exhibit a rich interplay between organized structure and stochastic fluctuations across a broad range of spatial scales. This duality motivates the need for quantitative frameworks capable of capturing their morphological complexity. The ordinal patterns framework, along with its associated statistical measures: permutation entropy ($H$), disequilibrium ($D_E$), statistical complexity ($C$), and ordinal network node entropy, has recently emerged as a powerful tool for analyzing such complexity in physical systems. We apply this framework in a multiwavelength, multiscale analysis of the galaxy NGC 628, utilizing observations in the near-ultraviolet, near-infrared, mid-infrared, and millimeter bands. Our results reveal a characteristic spatial scale of approximately 200 parsecs, marking the transition from small-scale structures influenced by star formation and stellar feedback to larger-scale morphology governed by the galaxy's dynamics. Furthermore, we find that the $C$ vs. $H$ trajectories for all wavelengths converge toward a common attractor curve, consistent with the behavior of isotropic Gaussian random fields. This convergence suggests a universal statistical behavior in galactic structure at large scales, despite the differing physical processes traced by each wavelength.

We study the atomic Hydrogen (HI) content of galaxies in the core of the Shapley Supercluster (SSC) at <z> ~ 0.048, using observations from the MeerKAT Galaxy Cluster Legacy Survey and optical data from the Shapley Supercluster Survey (ShaSS) project. Our sample comprises 169 galaxies with HI detections in the dynamically active region of Abell 3558 and SC1329-313. Following the literature, we classify galaxies into star-forming main sequence (SFMS), transition (TZ), and red sequence (RS) populations, and examine how the HI content varies across these populations. Galaxies on the SFMS exhibit an average HI gas fraction offset of 0.038 dex from the gas fraction main sequence, while TZ and RS populations show depleted HI fractions of -0.034 and -0.211 dex. HI depletion timescales span from 6 to 170 Gyr (SFMS-TZ-RS) confirming increasingly inefficient star formation with quenching. Scaling relations between HI mass and stellar mass in the SSC are generally consistent with field samples. The most direct signature of the dense environment of the SSC is the marked predominance of TZ galaxies, in contrast to what is observed in the field-dominated sample of xGASS, where the population is mostly composed of SFMS galaxies. Moreover, the SFMS and RS populations have similar size, again in contrast with field populations. These results suggest that galaxies in the SSC are undergoing environmental quenching through starvation or strangulation, rather than rapid gas stripping. Despite detectable HI reservoirs, many galaxies exhibit long depletion times, indicating reduced gas accretion and inefficient star formation.

Weak gravitational lensing is one of the primary cosmological probes, providing powerful constraints on the cosmological model. As Stage IV surveys are expected to deliver data of unprecedented precision, accurate modeling of weak gravitational lensing observables across both linear and non-linear scales becomes increasingly important. In this work, we investigate weak lensing in modified gravity (MG) models, extensions of the standard $Λ$CDM cosmology in which gravity deviates from general relativity, generally introducing modifications to the lensing equation. We parametrize these modifications through the common phenomenological function $Σ_\mathrm{mg}$ and apply ray-tracing to the density maps of N-body and hydrodynamical simulations. We model the time dependence of $Σ_\mathrm{mg}$ analytically, while we introduce a phenomenological scale dependence to represent the screening mechanisms by which MG models reduce to general relativity in high-density environments. Starting from the output of the FLAMINGO hydrodynamical simulations, we generate fully ray-traced convergence maps using our modified lensing model. We analyze how the parameters of our prescription affect the weak lensing convergence power spectrum and compare these effects to other known sources of variation, in particular cosmological parameters and baryonic feedback. We find that the modifications to the lensing equation deriving from the MG model produce non-negligible signatures in the convergence power spectrum and that, within extensions of the $Λ$CDM framework, these effects can be larger than those induced by baryonic physics. Our results indicate that modified lensing should become a standard ingredient of the analysis of modified gravity simulations.

Scattering phase functions (SPFs) derived from resolved scattered-light images of debris discs are widely used to infer dust grain properties, often via parametric forms such as the Henyey-Greenstein (HG) phase function. However, it remains unclear to what extent the inferred scattering behaviour reflects intrinsic dust properties rather than projection effects, disc geometry, or methodological choices. We test how reliably SPFs and HG asymmetry parameters can be recovered from scattered-light images and identify regimes where geometric and observational effects introduce significant biases. We use a physically motivated forward-modelling framework combining dust-scattering calculations, grain dynamics, and ray-tracing to generate synthetic total-intensity images. Since the intrinsic SPFs are known a priori, phase functions extracted from the images can be directly compared to the input scattering behaviour. We explore a grid of grain size distributions, disc inclinations, and opening angles, and fit two-component HG functions to evaluate how well the forward-scattering parameter $g_{1}$ traces grain properties. Even under idealised conditions with perfect knowledge of disc geometry, the recovered phase functions can differ substantially from the intrinsic SPFs. Limited scattering-angle coverage is the dominant effect: strong forward-scattering peaks at small angles are typically unobservable, leading to non-monotonic trends of apparent anisotropy with grain size. Projection effects, line-of-sight mixing, and SPF-extraction choices further modify the recovered phase functions, causing the fitted $g_{1}$ to depend strongly on viewing geometry and methodology. We conclude that SPFs and HG parameters derived from scattered-light images should be interpreted as effective, observation-dependent quantities rather than direct proxies for dust properties.

Electromagnetic cascades are initiated by TeV gamma rays propagating through the intergalactic medium (IGM), and they can be used to constrain the weak intergalactic magnetic field (IGMF) in cosmic voids. Primary TeV photons produce electrons and positrons through electromagnetic pair production, which can be deflected out of the line-of-sight to the observer by IGMF. In addition, electron-positron pairs can perturb the IGM, triggering plasma instabilities that can cool down the pairs before they upscatter cosmic background photons to GeV energies via inverse Compton (IC) scattering. We investigate the influence of plasma instabilities on the cascade spectrum by introducing a parameterized model for the instability using a publicly available Monte Carlo framework CRPropa. We use extended-emission observations within the field of view of the observer to constrain the IGMF in the presence of plasma instability cooling. Based on spectral observations of the blazar 1ES 0229+200 from Fermi-LAT, we find the best-fit photon spectrum including the plasma instability and IGMF parameters that reproduces the observational data for different observer field-of-view angles and obtain the IGMF constraint in cosmic voids. We find that plasma instabilities with a characteristic length scale of order $10^{2}~\text{kpc}$ reproduce the observed photon spectrum and imply an IGMF strength of order $10^{-17}~\text{G}$.

Understanding the kinematic behaviour of the Small Magellanic Cloud (SMC) remains a challenge addressed by many authors using diverse approaches. Over time, increasing observational evidence has accumulated for tidal perturbations induced by the Large Magellanic Cloud (LMC) on the SMC, especially in its outer regions. In this study, we adopt star clusters as kinematic tracers of the SMC. We analyse 36 clusters distributed across the galaxy's structural regions (Northern Bridge, Southern Bridge, Wing/Bridge, West Halo, Main Body and Counter-Bridge). From each cluster's proper motions, radial velocity and heliocentric distance we estimate Cartesian velocities \((V_x,\,V_y,\,V_z)\) in the SMC reference frame. We also compute the same velocity components under the assumption that the SMC behaves as a rotating disc. We then define the residual velocity \(ΔV\) for each cluster as the difference between the two velocities derived. Additionally, we perform a kinematic anisotropy analysis to characterise the distribution of kinetic energy across the SMC. We find that increasing values of \(ΔV\) correlate with increasing cluster distance from the SMC center, and that \(ΔV \approx 60\ \mathrm{km\,s^{-1}}\) it appears to be a lower limit that separates, in kinematic terms, the areas of tidal origin from those with the best behavior.

The stochastic gravitational wave background (SGWB) is conventionally treated as a passive relic of its astrophysical and cosmological sources, with negligible back-reaction on the matter content of the Universe. Here we show that this assumption needs to be modified once the SGWB and matter are treated as a dynamically coupled non-equilibrium system. Combining linearized general relativity with the fluctuation-dissipation theorem, we derive a generalized Langevin framework that drives the coupled system toward a dynamical equilibrium, which is characterized by a distinctive strain spectrum with a high-frequency cutoff $\mathcal{W}$, and a scale-dependent coupling parameter that screens gravity progressively for the most massive structures. Three findings support this framework. Fitting the equilibrium spectrum to the NANOGrav 15-year dataset yields a Bayes factor of $48\pm 3.8$ over the supermassive black hole binary baseline, achieved entirely within general relativity and the Standard Model. The PTA-calibrated screening mass scale $m_{c}\sim 10^{12}\text{--}10^{14}\,M_{\odot}$ overlaps, with no free cosmological parameter, the $Λ$CDM-derived linear-to-nonlinear transition mass $M_{\rm NL}$ of cosmic structure at $\sim 8\,h^{-1}\,\mathrm{Mpc}$. Most strikingly, promoting this concordance to a structural identification expresses $\mathcal{W}$ entirely in terms of $M_{\rm NL}$, and its inverse acquires a transparent physical reading as a coherence threshold for SGWB-matter coupling. $\mathcal{W}$ is thereby a derived quantity linking nanohertz gravitational-wave observables to the late-time cosmological sector. The framework makes distinctive scale-dependent predictions testable by forthcoming large-scale structure surveys and space-borne gravitational-wave observatories.

We present an analysis of the parallel peculiar velocity correlation function using data from the Cosmicflows-4 (CF4) survey. CF4 significantly extends the depth of the peculiar velocity measurements, mitigating the impact of observers on the cosmic variance. We examine the distribution of cosmic variance using different velocity correlation estimators. The combination of the large peculiar velocity uncertainties and the anisotropy distribution of the CF4 data across the northern and southern hemispheres results in substantial statistical uncertainties in the velocity correlation function. To address this, we test different weighing schemes in the velocity correlation function and implement a more accurate peculiar velocity estimator that reduces velocity uncertainties, consequently decreasing the statistical uncertainty. Using the CF4 group dataset, we derive a growth rate of $fσ_8=0.384^{+0.116}_{-0.194}$ and a local growth rate of $fσ_8=0.569^{+0.054}_{-0.06}$ through a Markov Chain Monte Carlo method.

The aim of this chapter, focused on relativistic mean-field models and part of the Encyclopedia of Nuclear Physics, is to provide an introductory, self-contained discussion accessible to a broad audience, including advanced undergraduate students. The chapter surveys the fundamental ideas, assumptions, and theoretical framework underlying relativistic mean-field models, and illustrates their wide range of applications across nuclear science. Particular emphasis is placed on the central role that these models play in the construction of equations of state for strongly interacting matter, as well as on the intimate connections between nuclear experiments, astrophysical observations, and theoretical modeling. In this context, relativistic mean-field theory is shown to provide a unified description of bulk nuclear properties and dense neutron-rich matter, enabling the interpretation of the remarkable structural and observational properties of neutron stars in the emerging era of multi-messenger astronomy.

Small carbon hydride cations, such as the methylidyne ion (CH$^+$), play an important role in the chemistry of the interstellar medium (ISM). They participate in gas-phase reaction networks leading to the formation of hydrocarbon species that act as precursors to more complex organic molecules. CH$^+$ is a highly reactive ion that is rapidly destroyed by H, H$_2$, and free electrons, making its excitation challenging to model. Its level populations depend not only on radiative and inelastic processes but also on chemical formation and destruction rates, a mechanism known as chemical pumping. We investigate this effect using a new set of ab initio state-resolved ro-vibrational (reactive and inelastic) collision data to model the observed CH$^+$ emission. Multiple rotational and ro-vibrational transitions of CH$^+$ detected toward the planetary nebula NGC 7027 are analyzed. The chemical structure of CH$^+$ is modeled with the CLOUDY code using updated reaction rates, providing the temperature and density structure across the nebula. A non-local thermodynamic equilibrium (NLTE) analysis is performed using CLOUDY and the single-zone RADEX code with a comprehensive set of spectroscopic and collisional data. In addition, chemical formation and destruction processes are implemented in RADEX and explored via Markov Chain Monte Carlo sampling. The CLOUDY model reproduces the observed CH$^+$ line fluxes within a factor of 1.3 on average. It indicates that rotational and ro-vibrational lines arise from physically distinct regions, primarily differing in temperature. RADEX models show that chemical pumping significantly enhances populations above ($\upsilon = 0, J = 1$), strongly increasing ro-vibrational emission, especially in the $\upsilon =2 \to 1$ band. Single-zone models remain limited, highlighting the need for full 1D modeling including all excitation processes.

We report the discovery of six faint millisecond pulsars (MSPs) in the globular clusters NGC 6517 and NGC 7078 (M15) using the Five-hundred-meter Aperture Spherical radio Telescope (FAST). These discoveries were enabled by stacking power spectra from multiple observations, a method that effectively boosts the signal-to-noise ratio of faint sources. In NGC 6517, we identified four new MSPs (NGC 6517S-V) with spin periods ranging from 3.68 to 6.02 ms and dispersion measures (DMs) between 182.45 and 182.85 pc cm^-3. In M15, two additional MSPs (M15M and M15N) were discovered, with spin periods of 4.83 and 9.28 ms, and DMs of 67.89 and 66.65 pc cm^-3, respectively. A phase-coherent timing solution has been obtained for M15M; however, sparse detection rates currently preclude phase-connected solutions for the remaining five pulsars. Current timing parameters suggest all six MSPs are isolated, which is consistent with the expected pulsar populations in core-collapsed globular clusters. Notably, pulsars M15N, NGC 6517U, and NGC 6517V eluded detection by standard frequency-domain searches (e.g., PRESTO-based) and the Fast Folding Algorithm, demonstrating that the stack search technique significantly enhances detection sensitivity to inherently faint pulsar signals.

Merging luminous galaxies are ideal laboratories to study some of the most extreme astrophysical phenomena. The local (z=0.1232) obscured quasar J1356+1026 has two nuclei, North and South (J1356N and J1356S), but despite numerous efforts, J1356S had not yet been confirmed as an AGN. Thanks to the superb sensitivity and spatial resolution of the MIRI/MRS instrument on board the JWST, we present new evidence suggesting that J1356S may indeed host an AGN with log L$_{\rm bol}=43.4\pm^{0.6}_{0.5} erg s^{-1}$. This is supported by the detection of strong coronal line emission at this location and by a spectral shape that differs from that of J1356N and those of the narrow-line region (NLR). Aided by the spatially resolved information of MIRI/MRS and VLT/SINFONI, we also find that the high ionization gas, traced by the coronal lines [Ne V]$14.3~μ$m and [Si VI]$1.963 μ$m, has an extension of ~13-15.5 kpc. This is likely a lower limit of the true extension, as suggested by the comparison with optical imaging from HST. {The extended [Ne V] emission can be accounted for by photoionization from the quasar in J1356N in a relatively low density environment, ranging from $\rm n_e\leq 2000-3800 cm^{-3}$ in J1356N and $\rm n_e\leq 600-1200 cm^{-3}$ in J1356S and the NLR, as measured from the [Ne V]$14.3μ$m and $24.3~μ$m lines.

Quasars exhibit stochastic variability across wavelengths, typically well-described by a Damped Random Walk (DRW). However, extreme luminosity changes, known as quasar flares, represent significant departures from this baseline and offer crucial insights into accretion disc dynamics and the fundamental physics of supermassive black hole fueling. While transient surveys have spurred interest in flare detection, a systematic search within the legacy SDSS Stripe 82 dataset -- containing 9,258 confirmed quasars -- has not yet been performed. The primary statistical challenge lies in distinguishing these rare events from ever-present intrinsic noise. To address this, we present FLARE (Flare detection via physics-informed Learning, Anomaly scoring, and Recognition Engine), a generalized three-stage framework for detecting flares present in quasar data. FLARE operates by modeling baseline DRW behavior, applying anomaly scoring to flag potential flares, and utilizing a recognition engine to verify candidates. For Stripe 82, we implement this framework using a physics-informed probabilistic Gated Recurrent Unit (GRU) for baseline modeling, Extreme Value Theory (EVT) for anomaly detection, and benchmarking various open-weight and proprietary Vision Language Models as recognition engines for final verification. Detection is executed on r-band light curves, with candidates cross-checked against g-band data to definitively rule out instrumental artifacts. Applying this pipeline, we successfully identify 27 quasars exhibiting distinct flaring activity.

A population of Galactic gamma-ray binaries is currently emerging due to ever increasing sensitivity of gamma-ray observatories. The detection of very high energy (VHE) photons with energies well above 10 TeV from a dozen of sources and the estimated power of those sources make them potentially interesting cosmic ray accelerators. Multi-wavelength observations of gamma-ray binaries revealed that most of them include a young massive star in pair with a relativistic companion, either a black hole or energetic pulsar. Fast stellar winds interacting with powerful relativistic outflows from pulsars or the black hole jets in microquasars are favorable sites for VHE particle acceleration. To estimate the expected number of gamma-ray binaries, we present results of population synthesis calculations of Galactic binaries in which a young massive OB- or Be-star is accompanied by a pulsar capable of producing a powerful relativistic outflow. The distributions over the binary eccentricities, orbital periods, Be-disk inclinations, and the pulsar braking energy losses are taken into account. Conditions for a binary to accelerate VHE particles, radiate and absorb the non-thermal photons that may reach the observer are discussed. We model the anisotropic structure of the zone of interaction of the relativistic pulsar wind with the strongly magnetized massive star's wind. The stellar winds with strong ($\sim$ Gauss) magnetic fields at $\sim$ AU distances colliding with powerful pulsar outflows are capable of accelerating particles up to PeV energies at some orbital configurations and phases. The strong magnetic field in the interaction region produces a highly anisotropic structure of the particle accelerator and emitter in the pulsar outflow. The anisotropic radiation pattern may affect the gamma-ray photon absorption and the number of the observed gamma-ray loud systems.

DustPedia and LVL are two samples representative of the local galaxy population, including in total 1011 unique objects of all morphological types, with a wide range of stellar masses ($M_*$) and star-formation activity, and a spectral coverage from the FUV to the FIR. The purpose of this work is to show that these samples cover two complementary ranges in $M_*$ and morphology, making them an ideal set for constraining the dominant processes in the evolution of the galactic dust content. Using the multiwavelength data provided by the two surveys, we fitted the galaxies' spectral energy distribution and estimated their physical properties, in particular the $M_*$, $sM_\mathrm{dust}=M_\mathrm{dust}/M_*$, and sSFR = SFR$/M_*$. By combining DustPedia and LVL, we highlight that the $\log_{10}(sM_{\rm dust})$-$\log_{10}(M_*)$ trend is not monotonic. Thanks to a large number of objects across a wide range of $M_*$, we have been able to fit two smoothly-joined linear correlations: a positive for $\log_{10}(M_{*}/$M$_\odot)\lesssim9.5$ (mainly LVL late spirals and irregulars), and a negative one for larger-mass, mainly DustPedia spirals (early types are distinct and more dispersed in the same mass regime). For $\log_{10}(M_{*}/$M$_\odot)>9.5$, we confirm a strong sM_{\rm dust}-sSFR correlation; dwarf galaxies, instead, lie below this trend, with a large scatter of $sM_{\rm dust}$, for -10.5<$\log_{10}$(sSFR/yr$^{-1}$)<-9.0. By using chemical evolution models we find that the observed $\log_{10}(sM_{\rm dust})$-$\log_{10}(M_{*})$ and $\log_{10}(sM_{\rm dust})$-$\log_{10}$(sSFR) trends can be interpreted mainly by variations in the initial gas mass budget and the galaxy ages, respectively. Low-mass Sm-Irr galaxies with low $sM_{\rm dust}$ and high sSFR can only be reproduced by the models by assuming high photofragmentation rate of large grains, and/or low grain-growth in clouds.

We investigate propagation of generic waves on thin planar domain walls effectively described by the scalar DBI model. We pay a particular attention to the possibility of caustic (shock) formation - the process, which may lead to intensive particle emission by domain walls. It is demonstrated that no singularities arise in DBI in 2D flat spacetime in the hyperbolic case, if one starts from smooth initial conditions. Technically, this happens because the same family characteristics of the relevant PDE remain parallel at all the times, albeit not being straight lines generically. Crucially, characteristic curves cease to be parallel beyond the simplified setup of DBI in 2D flat spacetime. In particular, this is shown to be the case in $D>2$ for spherical waves, in an expanding Universe, and in the case of a minimal deformation of DBI necessary for avoiding the domain wall problem in cosmology. However, we prove that DBI remains shock free in the hyperbolic case in all these physically relevant situations. This strongly suggests that caustics can form on planar domain walls only due to the loss of hyperbolicity, and they have a cusp profile. We demonstrate, how the non-trivial structure of DBI characteristics beyond the 2D flat spacetime setup uncovered in this work can significantly affect cusp formation.

The pulsar timing array (PTA) is a powerful technique for detecting nanohertz gravitational wave backgrounds (GWBs). However, conventional PTAs lack sensitivity to parity violation in the GWB. In this work, we propose a dipole pulsar timing array system (dPTA). By deriving the overlap reduction functions (ORFs) from the cross-correlation of timing signals, we find that this system exhibits sensitivity to chiral GWBs in the nanohertz regime. Furthermore, through numerical calculations of its sensitivity curves, we demonstrate that the dPTA extends the detectable frequency range of PTAs for GWBs from the nanohertz to the microhertz regime.

We investigate measures of distance and redshift in cosmological space-times that admit a shear-free foliation, which we henceforth refer to as `quasi-Newtonian'. Space expands isotropically in this description, and small-scale gravitational physics has a natural Newtonian limit, which makes it ideal for considering the physics of wide classes of cosmological models. By assuming that the energy-momentum tensor is dominated by rest-mass density, and that the 3-velocity of matter is small in the quasi-Newtonian frame, we derive fundamental results for kinematics and light propagation. Our results provide a new way of formulating general-relativistic cosmologies with non-perturbative structures in terms of quantities that can be understood from cosmological perturbation theory and post-Newtonian expansions, and allow us to quantify departures of observables from the predictions of Friedmann cosmology. It thereby provides a route to understanding inherently relativistic space-time structures, such as those that occur in Lemaître-Tolman-Bondi, Szekeres solutions, and Bianchi cosmologies in terms of Newtonian degrees of freedom. We illustrate our results using the degenerate Kasner solution as an example, and explain how our approach can be used to provide new insights into the current cosmological tensions.

X-ray spectra of accretion-powered X-ray pulsars can often be described using a power-law continuum with a high-energy cutoff, which might be further modified by additional spectral components. The Be X-ray binary system 4U 0115+63 is well known for having one of the highest numbers of detected harmonics of its cyclotron resonant scattering features (CRSFs), a pronounced spectral component known as the ''10 keV feature,'' and quasiperiodic oscillations (QPOs) with a period of about 500 s during outbursts. The changes in count rate by a factor of two during the approximately 500 s QPOs allow us to probe the variation in the spectral components with flux. We study the ''10 keV feature'' in emission, aiming to disentangle it from the broadband continuum and CRSFs and investigate its origin. We focus on the flux-dependent behavior of the CRSF and its harmonics, and particularly the contribution of the ''10 keV feature,'' as seen in the flux-resolved analysis of two NuSTAR observations of the 2015 outburst. Comparing the flux-resolved spectra of a given observation with the respective total dataset revealed a distinct change in overall spectral shape at the position of the ''10 keV feature'' but no comparable deviation at the energies of the harmonic CRSFs. The change associated with the ''10 keV feature'' does not seem to involve its centroid energy, which remains constant within a given observation. We find indications for an anticorrelation between the continuum flux and the ratio of the ''10 keV feature'' flux to the continuum flux within each observation. The analysis strengthens previous claims that the ''10 keV feature'' shows some independence from the remaining features. This result supports the interpretation that the ''10 keV feature'' has a different formation mechanism than the continuum emission, although its origin lies within the same physical environment.

We present the discovery of SN 2025mkn, a gravitationally lensed Type II supernova. First detected as a blue transient in ZTF, 0.83$^{\prime\prime}$ from a $z=0.42$ elliptical galaxy, follow-up SNIFS/UH2.2m and LRIS/Keck spectra revealed absorption lines at $z=1.371$. Later JWST NIRCam imaging shows that the bright transient is a close pair of point sources separated by $\sim 0.07^{\prime\prime}$, and a 30 times fainter counterimage opposite the lens, for which NIRSpec reveals strong H$α$ emission also at $z=1.371$. The light curves and spectra are consistent with the Type II supernova source being magnified $\gtrsim 100$ times, with $\sim 250$ required to reconcile its luminosity with that of nearby events such as SN 2023ixf. Lens models are consistent with such high magnifications, and always show that the faint image arrived first (undetected in earlier ZTF imaging), consistent with the later spectral phase of this fainter image. A fourth image is also predicted and possibly detected in the NIRSpec data. Light-curve-based time-delay measurements are not possible due to the first image being the faintest; however, the resolved NIRSpec spectra offer a future opportunity for time-delay cosmography through supernova phase measurements.

This work establishes oblique shocks in Massive Star Clusters (MSC) as a primary mechanism for accelerating cosmic rays (CR) up to the knee of the energy spectrum. We develop a model that incorporates the combined contribution of supernova and collective wind shocks, emphasizing the critical role of the shock obliquity angle in determining the maximum particle energy. We illustrate, within our model that oblique shocks can significantly enhance acceleration efficiency, allowing particles to reach multi-PeV energies in a rigidity-dependent manner. Our preferred model, which incorporates oblique shocks, reproduces the all-particle spectrum and composition observed by The Large High Altitude Air Shower Observatory (LHAASO), interpreting the knee as arising from a sequence of rigidity-dependent cutoffs. The model also predicts subdominant but detectable gamma-ray and neutrino emissions. This study provides an attempt at building a unified framework connecting MSC particle acceleration to the observed features of the cosmic-ray knee.

We calculate the gravitational wave signal from the collapse of a rotating 300 $M_\odot$ star at the upper end of the pair-instability regime. The large-scale asymmetries that develop during the collapse produce a strong signal in the deci-Hz range that has a characteristic shape which is likely amenable to a template-based search. The most ambitious designs for deci-Hz detectors could detect such signals out to distances of 200 Mpc, possibly at a rate of 0.5 per year.

The enigmatic origin of the diffuse neutrino background detected by IceCube in the energy range from TeV to PeV remains one of the central open problems in high-energy astrophysics, and this puzzle is further deepened by the recent evidence for a spectral break. Could this convex-spectrum background arise predominantly from the evolution of active galactic nuclei (AGNs)? In this work, we claim that the spectral break is naturally predicted when AGN evolution is taken into account, and the diffuse flux can be interpreted as the superposition of contributions from AGNs at different evolutionary phases. We develop a unified framework that incorporates AGN evolution, where cosmic rays (CRs) accelerated during the active phase subsequently diffuse and interact in the host galaxy after the central engine switches off, producing a long-lived hadronic afterglow. Adopting physically plausible parameters, our model successfully reproduces both the spectral features of the diffuse background and the observed neutrino emission from representative sources such as TXS 0506$+$056 and NGC 1068. Our results suggest that AGN host galaxies are more efficient CR reservoirs than previously expected. Moreover, the model favors a lepton-dominated scenario for most AGNs. This conclusion accounts for the relatively low detection rate of point-like sources by IceCube and underscores the need for next-generation neutrino telescopes with larger effective areas and higher sensitivity.

We report new results from a nulling study of PSR~B0751+32 (PSR J0754+3231), observed at 1250~MHz with the Five hundred meter Aperture Spherical radio Telescope (FAST). Our analysis confirms the presence of periodic nulling in this pulsar. Using the recently developed mixture model method, we obtained a nulling fraction (NF) of $35.1\% \pm 0.6\%$. Three independent approaches were employed to estimate the nulling periodicity, and the results reveal significant temporal evolution of the modulation both within individual observations and across different \textbf{observing} essions. The pulsar exhibits an asymmetric two-component mean pulse profile, with the leading component brighter and narrower than the trailing one. Pulse energy analysis shows that both components remain stable immediately after the onset of the burst state, but subsequently undergo a progressive decline, with the trailing component most severely affected prior to burst termination. Notably, no evidence of the previously reported subpulse drifting was detected in our data. Our results challenge previous models that ascribed periodic nulling to purely geometric effects.

The magnetar 1E 1841-045 exhibited a new active episode starting on August 20, 2024, marked by X-ray bursts and enhanced persistent emission. Using data from the Einstein Probe (EP), we report on the timing and spectral results following the onset of this outburst. The pulse profile displays a multi-peaked structure, with notable phase shifts in the secondary peak. Energy-resolved pulse profile analysis indicates a transition in the dominant peak of the pulse profile above 5.8 keV. The 0.5-10 keV X-ray spectrum is well-modeled by a combined blackbody and power-law (BB+PL) model, showing a $\sim 20\%$ flux increase following the outburst. Phase-resolved spectroscopy indicates a correlation between BB temperature and pulse profile intensity, along with spectral hardening at a specific pulse phase. The high spatial resolution of EP enables effective separation of the supernova remnant emission, which is crucial for measuring the intrinsic pulse emission of the source. These findings underscore the intricate relationship between magnetar outbursts, pulse profile evolution, and spectral characteristics.

The next nearby core-collapse supernova (SN) promises to yield a treasure of scientific information through multi-messenger astronomy. Early observations of the shock breakout (SBO) emissions are especially critical to understand the SN explosive mechanism as well as the properties of the progenitor star. Neutrino observatories are able to provide an early alert of a SN before the arrival of the SBO radiation. Super-Kamiokande (SK) has the unique capability to independently reconstruct an accurate SN pointing direction as part of its real-time monitoring system, ``SNWATCH.'' Recent upgrades to SK by adding gadolinium (Gd) to the detection volume have been accompanied by efforts to improve the speed and accuracy of SN direction reconstruction. A new, novel HEALPix-based approach (``HP-Fitter'') can calculate the SN direction from the reconstructed burst event directions in less than one second. As well, the previous maximum-likelihood direction fitter (``ML-Fitter'') was upgraded by incorporating event information from Gd neutron-capture as well as using the HP-Fitter for the initial fit parameters and from code refactoring and optimization. The improved ML-Fitter has better angular resolution but direction reconstruction time is $\mathcal{O}$(sec). Together with improvements in burst detection and event reconstruction times, SNWATCH is now able to generate an SN alert with pointing information in about 90 seconds. These upgrades have been implemented at SK and integrated into a new automated system to provide GCN notices.

Supernova remnants (SNRs) can strongly affect the chemical composition of the interstellar dust. In this paper we investigate to what degree the dust and ices are modified by observing four stars expected to be absorbed by a giant molecular cloud interacting with SNR W44, using medium-resolution spectroscopy in 2-5 $μ$m. Absorption from H2O ice around 3.0 $μ$m and aliphatic hydrocarbon dust around 3.4 $μ$m were detected towards two stars, while probable CO ice at 4.67 $μ$m towards one of them. Millimeter gas-phase CO J = 1-0 lines and three-dimensional dust extinction maps show that the dense molecular gas associated with W44 dominates (> 60%) the total interstellar extinction (A_K ~ 2.6) along these two sightlines. The H2O ice column densities are a factor of 1.5-3 lower than nearby MCs at similar extinctions, possibly because of the destruction of ice by shocks and cosmic rays (CRs) from W44, consistent with the low CO ice abundance relative to H2O (< 12%). One of the sightlines shows an unusually strong 3.4 $μ$m aliphatic hydrocarbon absorption. If the carriers are located in diffuse dust along the sightline, unrelated to W44, its strength is ~ 4 times larger than those typically observed for diffuse dust clouds. Alternatively, the carriers may be enhanced in the W44 environment. We discuss several possible explanations, including shock formation of aliphatic hydrocarbons in diffuse clouds associated with W44, contribution from aliphatic hydrocarbons in shocked and CR-bombarded molecular clouds, and changes in the extinction law due to the SNR interaction.

Data from gravitational-wave (GW) detectors often contains a high rate of non-Gaussian transient noise, known as glitches. The parameters estimated from GW signals coinciding with detector glitches are occasionally biased away from their true values. During the first part of the fourth LIGO-Virgo-KAGRA (LVK) observing run, 29% of GW candidates had overlapping or nearby glitches in one or more detectors. In the latter part of the fourth observation run, sensitivity improvements have increased the rates of GW detection. Consequently, scenarios in which GW signals and detector glitches overlap in time are more likely. In this study, we quantify shifts in inferred posterior distributions for short-duration compact binary coalescence GW signals interacting with common LIGO glitches as a function of time between the signal merger time and the glitch. We find statistically significant biases in parameter estimation for mass, spin, and sky position for "blip", "thunder", and "fast-scattering" glitches. Using these results, we provide estimates of what parameters are most affected by overlapping noise sources, as well as what constitutes a "safe" time separation between a gravitational wave signal and a glitch, without requiring glitch subtraction for unbiased source property estimation. We find that in a majority of cases, all parameters are susceptible to significant bias due to glitch interference. Additionally, we find that glitches that occur within the time prior of the GW signal cause more extreme biases than glitches outside of the time prior.

Globular clusters (GCs) are powerful tracers of galaxy assembly, frequently used to identify accreted substructure and reconstruct hierarchical merger histories. With advances in GC formation models and cosmological simulations, we can now better quantify the information about galaxy evolution encoded in present-day GCs. Here, we investigate how GC kinematics evolve over cosmic time and assess the extent to which GCs retain memory of the past of their host galaxy. Using a GC formation model applied to five Milky Way (MW) analogues from the Latte suite of the FIRE-2 simulations, we track the evolution of kinematic properties. At $z=0$, in-situ and ex-situ GCs exhibit substantial overlap in kinematic space, indicating that these populations are not clearly separable. We find that a subset of kinematic properties evolve in an ordered fashion across both in-situ and ex-situ populations, whereas others are dominated by stochastic variations. As a result, by the present day, most memory of the progenitor of an accreted GC is erased and only a few correlations persist. These correlations link progenitor halo mass to the total mass and number of a GC population, and the galactocentric distance of GC substructure to progenitor maximum circular velocity. These results highlight how both deterministic and stochastic processes driven by galaxy evolution shape GC kinematics and demonstrate the limits of reconstructing the assembly history of a galaxy from present-day GC orbits alone.

Galaxy environment plays a crucial role in quenching star formation in dwarf galaxies. In Milky Way (MW)-like environments, dwarf satellite quenching is primarily driven by ram pressure stripping (RPS), the direct removal of satellite gas by the host halo gas. Using a suite of 20-pc resolution hydrodynamical wind tunnel simulations, we constrain the satellite mass scale at which the stripping of a dwarf galaxy's interstellar medium (ISM) becomes inefficient in MW-like halos. The simulations include radiative cooling in a multiphase satellite ISM, star formation, and stellar feedback, and vary both satellite masses ($M_{\star}=10^{6.2}, 10^{6.8}, 10^{7.2}\ M_{\odot}$) and host halo gas densities along a first-infall and post-pericentric orbit. We find that the degree of ISM stripping in our dwarf galaxies is consistent with the analytical prediction by McCarthy et al. (2008). Star formation is rapidly quenched when RPS is effective, but can be mildly enhanced or temporarily quenched and subsequently reignited when RPS is incomplete. ISM stripping is efficient for satellites with $M_{\star} \lesssim 10^{7}\ M_{\odot}$ (or $M_{200} \lesssim 10^{10}\ M_{\odot}$) but highly inefficient above this scale. This transitional mass ($M_{\star} \approx 10^{7}\ M_{\odot}$) is 0.5-1 dex lower than that found in observations and cosmological simulations, suggesting that additional mechanisms are needed to quench more massive satellites, such as tidal stripping of the satellite dark matter or RPS from a clumpy gaseous halo.

We apply a general relativistic framework to static and rotating black hole solutions in Scalar-Tensor-Vector Gravity or modified gravity (MOG). Our results yield exact analytic, closed-form relations expressing the mass $M$, the MOG coupling parameter $α$, and the distance $D$ of a Schwarzschild-MOG black hole in terms of a minimal set of directly measurable elements of the accretion disk: the total frequency shift, the telescope aperture angle, and the redshift rapidity. The resulting expressions are derived for particles close to the midline and line of sight, where the redshift rapidity is treated as a relativistic invariant encoding the evolution of the frequency shift with respect to the emitter's proper time in MOG spacetime. We further extend the formalism to the rotating Kerr-MOG geometry and obtain corresponding relations that determine the rotation parameter $a$ jointly with $M$, $α$, and $D$ on the midline. In the rotating background, we introduced the redshift acceleration (general-relativistic version of jerk) to disentangle the spacetime parameters. Crucially, the explicit appearance of $α$ in these formulas enables direct empirical estimation of this parameter, thereby providing a means to test for departures from standard general relativity. The previous results obtained in the standard Schwarzschild/Kerr backgrounds are recovered in the limit $α\to 0$. The derived expressions are concise and suitable for incorporation into black hole parameter-estimation pipelines.

The central few parsecs of the Milky Way host a complex X-ray-emitting environment in which several extended plasma components are blended along the line of sight, complicating attempts to measure the intrinsic properties of individual components. In particular, the supernova remnant (SNR) Sgr A East is strongly confused with the stellar wind-fed plasma associated with Sagittarius A* and the surrounding nuclear environment. Here we apply Poissonian Generalized Morphological Component Analysis (pGMCA) to deep, stacked Chandra ACIS-I observations of the Galactic Center to disentangle these overlapping X-ray components. By comparing the separated X-ray components with multiwavelength data, we identify the location of the reflected shock in Sgr A East and construct spatially resolved maps of Fe and S/Ar/Ca emission. The Fe emission is centrally concentrated, consistent with the properties of mixed-morphology supernova remnants. Separating the SNR emission from the shocked wind plasma around Sgr A* allows us to recover uncontaminated SNR properties and improve the robustness of the derived parameters. Spectral modeling of the isolated Sgr A East component reveals a lower ionization age and a higher electron density than previously reported, indicating strong interaction with dense surrounding material.

We present a study of the multiphase gas structure and kinematics of the circumnuclear disk (CND) of NGC 5506, a nearby edge-on Seyfert galaxy, at a spatial resolution of $\sim20$ pc. Observations of [C I](1-0), CO(3-2), and HCO$^{+}$(4-3) obtained with the Atacama Large Millimeter/submillimeter Array reveal the CND dominated by rotational motion on scales of several hundred parsecs. No significant differences in geometrical thickness or velocity structure are found between [C I](1-0) and CO(3-2) across the CND, whereas HCO$^{+}$(4-3) emission is more concentrated toward the disk plane. The ratio of velocity dispersion to rotational velocity, a proxy for disk scale height-to-radius ratio, is high ($\gtrsim0.9$) in the central region ($\lesssim30$ pc) for both [C I](1-0) and CO(3-2), indicating geometrically thick structures in both tracers. Regions where the [C I](1-0)/CO(3-2) ratio exceeds the CND average are spatially correlated with the [O III]$λ$5007 bicone observed with the Hubble Space Telescope, suggesting that CO is preferentially dissociated by the AGN-driven biconical ionized outflow. The observed CND scale height and velocity dispersions traced by [C I](1-0) and CO(3-2) are consistent with a model in which supernova-driven turbulence provides the vertical support for the CND.

Gamma-ray bursts (GRBs) are among the most luminous and rapidly evolving transients in the Universe. While space-based instruments have extended GRB observations up to energies of $\sim$100 GeV, the detection of very-high-energy (VHE; $E>100$ GeV) emission from ground-based telescopes, especially during prompt or/and the early afterglow phase, remains challenging. These difficulties arise from the rapid temporal decay of GRB afterglows, strong attenuation by the extragalactic background light (EBL), observational latency, and the typical poor sky localization provided by MeV-detectors such as Fermi/GBM. In this work, we investigate the prospects for detecting TeV ($\sim$100 GeV--1 TeV) emission from poorly localized GRBs by adopting optimized follow-up strategies based on rapid tiling of large localization regions. We simulate a realistic population of GRBs informed by more than fifteen years of Fermi/GBM and Swift/XRT detections and recent progresses in the afterglow emission modeling. Using these simulations, we evaluate the detectability of GRB early afterglows by the next-generation Imaging Atmospheric Cherenkov Telescopes, equipped with larger field-of-view (FoV), as a function of latency, exposure time, and observational strategy. Our strategy can significantly enhance the detection rate; for instruments such as ASTRI and LACT, it increases by up to a factor of two compared to strategies limited to well-localized (Swift-like) events. For CTAO, our proposed approach provides up to four VHE detections per year.

We investigate two neighboring clusters in the Cygnus complex, Berkeley 86 and Berkeley 87, with a primary emphasis on the evaluation of extinction in the field of view towards and across the clusters. We also analyze their kinematic behavior in space and time to discern their possible common origin and relation to the Cyg~OB1 association. New CCD photometry in the Vilnius seven-color system, obtained down to V=19.0 mag in the fields of these two clusters, is used to classify stars in terms of spectral and luminosity classes and to determine the individual values of interstellar extinction. The probable cluster members are identified in a 5-parameter space based on Gaia DR3. The cluster ages and stellar masses are derived through the use of the HR diagrams. To obtain the 3D kinematics of the clusters and trace their orbits back in time, we combine the Gaia-based proper-motions and distances with radial velocities from the literature. The estimated cluster properties show that both clusters are almost equidistant (1.7 kpc) and nearly coeval, with average ages of 6.1$\pm$0.5 and 6.5$\pm$0.4 Myr, respectively, and age dispersion of 3 Myr. The nonuniformity of extinction is evident within each cluster, especially pronounced across the face of Berkeley 86 where the most-massive stars show substantial substructure. By extrapolating the observed mass function to a minimum stellar mass, we obtain cluster masses of 519 M(Sun) and 1551 M(Sun) for Berkeley 86 and 87, respectively. Although both clusters share very similar properties, their orbital paths show no indication that they had a common birthplace, however Berkeley 87 and its neighbor NGC 6913 are very likely to have been born in pair.

Changing-look transitions challenge our understanding of active galactic nuclei (AGN), exhibiting dramatic changes in broad-line emission and continuum flux on timescales of months to years. We present a detailed study of the spectroscopically confirmed changing-look AGN ZTF18abuamgo. Combining photometric survey data with spectroscopy spanning three epochs over 20 years, we identify a turn-on transition from a Type 1.5 to Type 1.2 AGN and estimate the timescale of this change to be as short as four years. Spectral analysis indicates that this transformation is driven by a rapid increase in accretion rate, with the Eddington ratio rising from $0.032 \pm 0.005$ in the dim state to $0.08 \pm 0.01$ in the bright state. For the first time in a changing-look AGN, we apply the Boltzmann plot method to the visible Balmer series emission, deriving broad line region electron temperatures of $11,800 \pm 900$ K and $11,900 \pm 2,400$ K in 2022 and 2024, respectively. Applying single-epoch black hole mass estimation to the brightening H$α$ emission, we find a mass of $(5.0 \pm 0.4) \times 10^7 M_\odot$. The consistency in this estimate across all spectroscopic epochs suggest that even highly variable broad lines in CL-AGN do not bias the results derived using this method. Our results demonstrate that objects like ZTF18abuamgo provide a unique laboratory to study extreme AGN variability, probe the physical conditions in the broad line region, and assess the limitations of widely used black hole mass estimation methods.

We present a concise multimessenger inference of the neutron-star core sound-speed profile using GW170817 and three \textit{NICER} mass--radius posteriors (PSR J0030$+$0451, PSR J0740$+$6620, and PSR J0437$-$4715). The main result is not only a preference for intermediate-density stiffening within smooth equation-of-state families, but a translation of that inference into the stellar masses that access the relevant density regime. In the baseline smooth peaked family, the posterior probability that $c_s^2 > 1/3$ at $3.5\,n_{\rm sat}$ is $85.4\,\%$, while equal-prior averaging over peaked, monotonic, piecewise, and transition-capable families gives a more conservative $79.0\,\%$. Posterior-resampled exact Tolman--Oppenheimer--Volkoff solutions show that the onset density of the inferred stiffening is typically reached near $1.6\,M_\odot$, whereas the peak region is accessed only near $2.1\,M_\odot$. A J0740-like $2.07\,M_\odot$ pulsar reaches the onset in $91\,\%$ of posterior draws but the peak in only $46\,\%$, showing that current data mainly constrain whether massive stars have entered the stiffening regime rather than traversed its full peak.

Magnetic cataclysmic variables provide a natural laboratory for studying how accretion interacts with compact-object magnetospheres and generates stochastic variability. We present an optical variability study of the intermediate-polar candidate EP240309a, an Einstein Probe X-ray transient, using BOOTES photometry, high-cadence TESS light curves, and a SOAR/Goodman optical spectrum. Previous studies found a white-dwarf spin period of 3.97 min (Pspin ~ 238 s) and an orbital period of Porb = 3.7614(4) h. Power spectral densities from the BOOTES data are consistent with single power laws with slopes alpha ~ 1.2-1.8, with no statistically significant evidence for a bend across the sampled frequency range. Using red-noise simulations and injection-recovery tests, we place one-sided constraints on any putative break frequency, which translate, under standard dynamical identifications, into an upper limit on the magnetospheric radius of Rm <= few x 10^10 cm for MWD = 0.8 Msun. In the TESS data, we detect a linear rms-flux relation on hour timescales in three high-cadence sectors, while two other sectors do not show a robust detection, indicating epoch-dependent rms-flux behavior. The SOAR spectrum shows Balmer and He II emission lines with FWHM about 1000-1600 km s^-1; under a Keplerian interpretation, these imply characteristic radii of r about (0.9-3.4) x 10^10 cm, broadly comparable to the timing-based constraints. Overall, the data provide conservative, order-of-magnitude radius constraints consistent with accretion onto a magnetic white dwarf, but they do not establish the detailed accretion geometry or exclude stream-fed or mixed accretion scenarios.

Gravitational-wave observations have revealed an excess of binary black hole mergers with primary masses near $\sim 35\,M_\odot$. We show that if this feature originates from dynamical formation in dense stellar systems, and if the pair-instability supernova truncates the first-generation black hole mass spectrum, then second-generation mergers inevitably produce a second peak near $\sim 70\,M_\odot$. This structure reflects the suppression of first-generation black holes above a characteristic mass and the accumulation of merger remnants near twice that scale. Its location is robust, whereas its amplitude depends strongly on cluster initial conditions. Using a large suite of cluster population-synthesis models, we show that current gravitational-wave data already constrain the birth properties of globular clusters, irrespective of their overall contribution to the observed population. If clusters dominate mergers above the pair-instability scale, these constraints tighten further and imply a minimum first-generation merger rate of $\mathcal{R}(m_1 \leq 50\,M_\odot) \geq 0.099\,\mathrm{Gpc}^{-3}\,\mathrm{yr}^{-1}$ ($99\%$ confidence). We further show that a drop or gap in the secondary black hole mass spectrum is not a robust signature of a cluster origin for high-mass mergers within the pair-instability mass gap. A confirmed excess near $\sim 70\,M_\odot$ would support a dynamical origin of the $\sim 35\,M_\odot$ feature and provide independent evidence for a pair-instability mass gap with a lower edge at $\lesssim 50M_\odot$.

Modeling subphotospheric shocks in a gamma-ray burst (GRB) is challenging due to the various timescales that must be resolved, and the fact that the same radiation dynamically mediates the shocks while forming the observed signal. Here, we present the first self-consistent radiation-hydrodynamic simulation of a subphotospheric internal collision, following the system from formation and propagation of forward and reverse radiation-mediated shocks all the way to photon decoupling and free streaming toward the observer. The simulation evolves the plasma and photon field with full Compton coupling, including the feedback on the hydrodynamic flow. As the ejecta expands and the optical depth decreases, both shocks broaden and the radiation field becomes highly non-thermal. Surprisingly, we find that the reverse shock remains completely radiation-mediated down to upstream optical depths of order a few $\times 10^{-1}$, which indicates that Compton coupling is important even in moderately optically thin regions. The photons undergo last scattering over a broad range of radii rather than at a single photospheric surface. The light curve shows a late, quasi-thermal post-cursor produced by photons that decouple upstream of the reverse shock, which could be searched for in observations. The emitted time-integrated spectrum is GRB-like, with a low-energy photon index $α\sim -1$ and a high-energy photon index $β\sim -2.5$. These results show how radiation-mediated shocks evolve close to the photosphere and how they shape the emitted photon field.

In dense neutrino gases, which exist for instance in supernovae, the flavour states of different neutrinos may become entangled with one another. The theoretical description of such systems may therefore call for simulations on a quantum computer. Existing quantum simulations of simple toy systems are not optimal in the sense that they do not fully exploit the symmetries of the system. Here, we propose a new class of qubit-efficient algorithms based on Dicke states and the $su(2)$ spin algebra. We demonstrate the excellent performance of these algorithms both on classical and on quantum hardware.

The sub-kpc scale gas structure encodes key information of giant molecular cloud (GMC) formation. Therefore, we aim for a quantitative description of molecular gas structure across 150-1000 pc using a sample of 8984 GMCs from 40 galaxies observed by PHANGS-ALMA. We homogenize our data to a fixed resolution of 150 pc and mass sensitivity of 2.5 M$_{\odot}$ pc$^{-2}$ to remove observational bias. We then calculate nearest neighbour distances, neighbour number density, and two-point correlation functions for the catalogued GMCs. When analysing the two-point correlation function, we generate several control samples that reflect different null hypotheses on large spatial scales. We stack integrated intensity CO emission profiles around the position of catalogued GMCs to probe the gas distribution on scales between the resolution limit and the typical GMC-GMC spacing. Our measurements of cloud spacing and number of neighbours show that GMC clustering follows the large-scale gas distribution. Once we account for this contribution, the peak excess clustering in the two-point correlation function drops from 1+$ω$ of 2.3 to 1.3, with the power-law slope flattened from -0.25 to 0. We show that the stacked CO intensity profiles around CO peaks can be recovered by the "GMC size" measured by CPROPS, with an additional 20% of the flux in an extended component beyond 500 pc. We find that our stacked profiles can be fit with a double Gaussian function plus a constant offset. The broad Gaussian component accounts for 70% of the over-density power above the constant offset, and is stronger around massive and gravitationally bound GMCs. Our results indicate that galactic structure regulates the GMC distribution in galaxy disks, and the formation of massive, gravitationally bound GMCs is related to strong local gas clustering.

We present a weak-lensing (WL) analysis of filamentary structures in the merging galaxy cluster Abell 2744 using wide-field Magellan/MegaCam imaging data. We employ two complementary techniques: standard matched-filter scans to identify global orientations, and a new stepwise 2D tracing method to reconstruct locally varying filament orientations. The matched-filter analysis detects coherent filamentary features in the northwest and east directions across both inner (1.0-2.2 Mpc) and outer (2.2-3.4 Mpc) annuli. However, while the northwest filament yields consistent constraints across both regions, parameter inference for the eastern structure remains unstable and radially inconsistent when restricted to global reference-point scans. We demonstrate that re-characterizing the eastern structure using the locally preferred elongation directions from our stepwise tracer significantly resolves these tensions, improving fit quality and bringing inner and outer constraints into agreement. Furthermore, the detected filaments align well with diffuse X-ray structures and previously identified merger axes, supporting their physical connection to the cluster's mass assembly. These results highlight that stepwise 2D tracing is essential for characterizing curved or complex filaments where global reference-point scans are insufficient.

Black hole impacts on accretion disks in galactic nuclei can power luminous transients, but predicting their observable signatures is challenging because the post-collision flow is highly time-dependent and inhomogeneous. We present a radiative post-processing framework for relativistic hydrodynamics simulations of black hole-disk collisions. Using physically motivated prescriptions for shock heating, optical depth via an eikonal solver, and photon escape fractions that account for advection trapping and diffusion, we predict light curves and spectral energy distributions over a range of disk densities and collision velocities. Our results indicate that the emission is dominated by the long-lived, highly super-Eddington accretion flow onto the secondary black hole, rather than by cooling of the unbound ejecta. In the parameter range explored, the luminosity can reach several times the Eddington luminosity of the secondary, and the emission is generically dominated by soft X-rays. We find that lower velocity collisions produce brighter flares, while the disk surface density mainly controls spectral evolution: low-density disks typically produce keV-peaked flares with weak spectral evolution, whereas high-density disks show softer early emission and late-time hardening. A depletion-time estimate calibrated to our results suggests characteristic durations of hours to days for intermediate-mass secondaries, and yields $t_{\rm flare} \propto P_{\rm QPE}$. We discuss implications for QPE-like transients and for the SMBH-binary candidate OJ 287.

The tip of the red giant branch (TRGB) provides a key standard candle for extragalactic distance measurements and for refining the Hubble constant. We test its robustness by quantifying how metallicity, $α$-element enhancement, age, and initial helium abundance modulate the TRGB luminosity, using synthetic composite color--magnitude diagrams in the $I$ and $F814W$ bands. We find that metallicity and $α$-element enhancement are the primary drivers of TRGB variation, while age introduces only a modest effect and helium abundance is negligible. At fixed age and helium content, increasing the mean metallicity by 0.5 dex or the $α$-element enhancement by 0.3 dex produces the well-known systematic dimming of 0.046 and 0.050 mag, respectively, in $M_I^{\rm TRGB}$, and of 0.093 and 0.044 mag, respectively, in $M_{F814W}^{\rm TRGB}$. By comparison, changes in age of 3~Gyr and in initial helium abundance of 0.10 yield minor luminosity shifts, with average changes of 0.031 and 0.009~mag, respectively, in $M_I^{\rm TRGB}$, and of 0.035 and 0.027 mag, respectively, in $M_{F814W}^{\rm TRGB}$, substantially smaller than those caused by variations in metallicity or $α$-element enhancement. For mixed stellar populations under typical stellar-halo metallicity conditions, the net variation in $M_I^{\rm TRGB}$ arising from each combination of the $α$-element enhancement, age, and initial helium abundance remains below 0.028~mag, well within reported systematic uncertainties. Together, these results reaffirm the TRGB as a highly robust distance indicator and support its continued use as an independent anchor for precision cosmology in the era of the Hubble-tension debate.

Recent JWST observations revealed two massive ($M_{\star} \gtrsim 10^{11}\,M_{\odot}$), unexpectedly large spiral galaxies at $z \sim 3$, both in overdense environments. We focus on one of these, ADF22.1 at $z = 3.09$, which hosts an active galactic nucleus (AGN), exploiting its extended [CII] emission ($\sim$30 kpc in diameter) with high-resolution observations from the Atacama Large Millimetre Array and JWST. We find a flat outer rotation curve reaching $\sim$530 km s$^{-1}$, and perform, for the first time for a system of this type, a rotation-curve decomposition. We infer a dark-matter halo mass of $\log(M_{200}/M_{\odot})=12.9^{+0.4}_{-0.3}$, a baryon-to-halo mass ratio of $0.4^{+0.6}_{-0.3}$ in units of the cosmological baryon fraction, and a ratio between the baryonic and dark-matter halo specific angular momentum of $1.0^{+0.7}_{-0.5}$. Comparing these quantities with those of local galaxies, we find that ADF22.1 is indistinguishable from $z=0$ giant discs, pointing to the inefficiency of AGN feedback in halting disc growth. Using the Mapping Nearby Galaxies at Apache Point Observatory survey, we identify potential $z=0$ descendants of ADF22.1, suggesting it will evolve into an extreme (in either mass or angular momentum) early-type galaxy. Finally, we argue that cold-stream accretion, invoked to explain disc formation at $z > 1$, cannot simultaneously account for its size, dynamical properties, high specific angular momentum, and baryon-to-halo mass ratio. Instead, sustained accretion from the hot circumgalactic medium, either via spontaneous or fountain-driven condensation, offers a more physically plausible formation pathway.

Strong gravitational lensing offers a unique probe of dark matter (DM) on sub-galactic scales, where the abundance and distribution of low-mass halos are highly sensitive to the underlying properties of DM particles. In this work, we forecast LSST's sensitivity to DM substructure in galaxy-galaxy strong lenses using simulated samples and neural ratio estimators (NREs). Our simulations include both subhalos within the main deflector and line-of-sight (LOS) halos, with halo masses down to $\sim 10^7 M_\odot$ under the expected LSST ten-year survey imaging quality. We show that the constraining power on halo mass function (HMF) parameters improves significantly with sample size. Analyses based on a few hundred lenses yield broad posteriors comparable with other probes like the Ly-$α$ forest. By contrast, when combining 2500 lenses, $\approx 74\%$ and $\approx 36\%$ of the prior volume considered can be excluded at the $3σ$ and $5σ$ levels respectively, enabling statistically significant exclusions of non-$Λ$CDM scenarios. We further demonstrate that the sensitivity arises not only from the high-mass end of the HMF but also from low-mass halos: masking halos below $\log (m_{\rm halo}/M_\odot) \leq 7.5$ induces a measurable shift in the inferred posteriors. Finally, we find that LOS halos contribute significantly to the constraining power, with increasing importance of LOS halos at higher redshifts. While this analysis assumes perfect knowledge of the data-generating process and cannot be directly applied to data analysis, it quantifies constraints achievable with LSST alone and motivates the development of robust inference methods for real survey data.

MAXI J1744$-$294, likely a low-mass X-ray binary system, is a Galactic-center transient source, detected at radio and X-ray wavelengths, located approximately $19''$ southeast of Sgr A*. We report the first detection of its variable linear polarization in four epochs spanning 2025 Apr 04--09. The normalized 33 and 43 GHz Stokes parameters $q$ and $u$ over the four epochs imply a common Faraday rotation screen with a rotation measure RM $=-63\,606^{+844}_{-861}$ radians m$^{-2}$, the third largest RM detected within the Galaxy. The RM is consistent with that of the Galactic center magnetar PSR J1745$-$2900, giving the first direct evidence that MAXI J1744 lies within the Galactic center region, is bound to Sgr A*, and therefore, is part of the nuclear star cluster. The uniformity in the Galactic center Faraday screen suggests that Sgr A*'s $\approx-10^5$ rad m$^{-2}$ RM is intrinsic rather than originating from an unrelated line-of-sight source. On 2025 Apr 06, we detected a secondary polarized component with an additional RM $\approx-6000$ rad m$^{-2}$, which was not seen at any other epoch. Assuming this secondary component primarily cools by synchrotron radiation, the implied local magnetic field strength is $\sim$15--30 gauss. In the context of a jetted X-ray binary progenitor, the additional RM screen and magnetic field strength are explainable with a short-lived knot in a putative jet.

The Cosmological Collider (CC) provides a unique opportunity to probe the particle spectrum and fundamental interactions at extremely high energies. Massive particles, via their decay into inflaton quanta, can induce a non-analytic, oscillatory, primordial non-Gaussianity (NG), including the bispectrum. At tree level, three classes of such processes contribute to the bispectrum: 'single exchange', 'double exchange', and 'triple exchange', depending on the number of massive particle propagators. We provide a unified evaluation of all three diagrams and derive the explicit shape functions for the bispectrum, valid across the entire kinematic space. We perform a search for these three processes with the Planck data, finding no evidence for NG. We also consider simple extensions of the minimal scenario that can counter the exponential suppression of the non-analytic signature, and produce on-shell particles with masses $M\gg H$, the Hubble scale during inflation. In particular, we focus on the 'scalar chemical potential' mechanism and extend our previous search to a wider range of chemical potential ($ω$) and $M$, finding global 1.5$σ$ evidence for non-zero NG for the parameter space $ω- M \simeq 3H$.

Gravitational effective theories are essential for characterizing the space of deviations from General Relativity (GR). Testing these theories against fundamental principles, such as causality and unitarity, can yield constraints on the corresponding parameters. In this paper, we perform such an analysis on the very interesting dynamical Chern-Simons (dCS) gravity. This is a parity violating correction to GR wherein a new scalar field couples to the Pontryagin density $^*R\,R$. It has generated significant interest, including possible new gravitational wave shapes for LIGO/Virgo and new phenomena from cosmic inflation. In this work, we begin by deriving the dispersion relation and wave packet speed on top of a gravitational wave background in dCS gravity. This alters the corresponding Shapiro time delay (which we compute to second order), potentially giving superluminality. Causality then demands a bound on the dCS coupling constant, which we find to be moderately sharper, but compatible with, standard estimates. We then examine a UV completion in the form of a set of $N$ fermions with a (pseudo) Yukawa coupling. By imposing perturbativity and a gravitational species bound, we find that the dCS coupling constant is constrained significantly more, depending on the choice of scale of the species bound. We also identify higher order operators generated from the UV completion. Overall, we find that any dCS corrections to gravitational dynamics should likely be very small on macroscopic systems of observational interest, such as in late-time merging black holes.

The Lorentz force exerted by a primordial magnetic field (PMF) on the coupled baryon-dark matter system may enhance total matter power at small scales after recombination. In the post-reionization (post-EoR) era, a weakly scale-dependent PMF of sub-nG strength is thus expected to influence the Lyman-$α$ (Ly$α$) power spectrum, the 21 cm power spectrum, and the Ly$α$-21 cm cross-spectrum at scales $k\gtrsim 1\:h/\textrm{Mpc}$. We investigate the prospects of constraining the PMF sector via these three cosmological observables, by employing SNR estimation and Fisher forecast on the PMF amplitude $B_0$ and spectral index $n_{\rm B}$, for a next-generation DESI-like spectroscopic survey and two upcoming 21 cm facilities, namely SKA1-Mid and PUMA. Our results indicate the possibility of constraining both PMF parameters with $\lesssim10\%$ relative errors through the uncontaminated 21 cm auto-spectrum as well as the Ly$α$-21 cm cross-spectrum probed with the DESI-like+SKA1-Mid combination. Indicatively, the Ly$α$-21 cm cross-correlation via DESI-like+SKA1-Mid is predicted to constrain a fiducial scenario $B_0=0.8$ nG and $n_{\rm B}=-2.9$ with $1σ$ errors $ΔB_0\approx 0.07$ nG and $Δn_{\rm B}\approx0.02$. The DESI-like+PUMA setup is predicted to fare relatively worse due to its restriction to larger scales, resulting in comparatively one order of magnitude relaxed error bounds for similar fiducials. Since the Ly$α$-21 cm cross-signal is expected to be largely insensitive to foreground contamination (unlike the 21 cm auto-spectrum), it may serve as an optimal foreground-immune post-EoR probe to constrain a weakly scale-dependent sub-nG PMF via future DESI-like+SKA1-Mid observations.

Non-thermal fermions can be produced non-perturbatively in the early universe during coherent oscillations of a scalar field. We explore fermion production in $λφ^{4}$ inflation through this mechanism and analyze the momentum spectrum of the fermions produced, which depends on a coupling parameter $q$. For $q \gtrsim 0.01$, the main contribution to the total number density comes from an approximately half-filled Fermi sphere as a result of non-adiabaticity. For $q\lesssim 0.01$, we find that the major contributions instead come from resonance peaks at higher momentum values. We find a simple relation to predict the momentum values corresponding to resonance peaks for any $q$. We also obtain analytic power-law approximations for the total number density of fermions and find that it is proportional to $q^{1/2}$ for $q\lesssim 0.01$ and proportional to $q^{3/4}$ for $q\gtrsim 10$. If fermions produced by this mechanism make up the entirety of dark matter, we estimate lower bounds on their mass.

The formation of late-type galaxies has traditionally been described via two pathways: one producing pressure-supported classical bulges, the other rotationally supported pseudo-bulges. Early studies relied on photometric decompositions assuming an exponential disk extrapolated inwards. Recent high-resolution observations, however, reveal a far more complex landscape in disk galaxy centres. We investigated the morphology of central stellar components in intermediate-to-massive spiral galaxies, focusing on disentangling cold, warm, and hot orbital contributions, critically reassessing the standard approach of extrapolating the exponential disk profile inwards. We developed GLANCE (Galactic archaeoLogy via chronochemicAl and dyNamiCal modElling), a tool for photometric, chronochemical, and dynamical galaxy analysis, applied to 8 high-resolution MUSE galaxies to derive stellar population properties and decompose orbits into cold, warm, hot, and counter-rotating (CR) components. We uncovered remarkable structural diversity in the dynamically cold central component: one galaxy displays an exponential profile throughout, while the majority exhibit either a pronounced central drop resembling a doughnut-shaped structure or a compact inner disk significantly steeper than the outer disk. Most galaxies hosting nuclear disks are classifiable as classical bulges - hot, old, red, high bulge-to-total ratio - contrasting with galaxies showing a central cold-component deficit. Beyond the bulge, cold plus warm orbit contributions remain below the total, indicating non-negligible hot or CR orbits with Sérsic indexes consistently above unity. These results highlight the composite nature of disk galaxy centres and the need for decomposition methods that avoid extrapolating the outer disk inwards, requiring large IFS samples across a broad mass range, complemented by simulations such as IllustrisTNG50.

This proceedings contribution provides an overview of Loop Blow-up Inflation and updates its observational predictions and their comparison with the latest CMB and BAO data from combined analyses of SPT, Planck, ACT, and BICEP/Keck, as well as ACT DR6 constraints on extra dark radiation. It is based on work originally published in arXiv:2403.04831, carried out in collaboration with L. Brunelli, M. Cicoli, A. Hebecker, and R. Kuespert, and presented at the 2025 Workshop on Quantum Gravity and Strings. We focus on string loop corrections to the Kähler potential, long regarded as a potential threat to blow-up inflation in the Large Volume Scenario. We argue that these corrections, previously assumed avoidable, are in fact generically present and qualitatively alter the original non-perturbative picture: they invalidate slow-roll near the minimum and instead generate a new slow-roll regime at larger field values, where the scalar potential transitions from an exponential to a power-law plateau. This leads to modified inflationary dynamics and distinct cosmological predictions, including an increased tensor-to-scalar ratio. We confront all three SM-location scenarios with the latest constraints on $n_s$, $r$, and $ΔN_{\rm eff}$. The tighter bound on extra dark radiation requires an updated Giudice-Masiero coefficient in Scenario III, yielding revised predictions presented here for the first time. All scenarios remain consistent with recent observations, with the ACT+DESI combination yielding near-perfect agreement in $n_s$ for vanishing extra dark radiation at $0.03σ$ deviation. We also comment on subleading loop corrections, which improve robustness by reducing the field value required for slow roll. These results highlight that string loop effects, rather than being merely detrimental, can play a constructive role in realising viable inflation in string compactifications.

We report our new lattice results for the sphaleron rate calculated within a thermal effective field theory of soft SU(N) gluons, where $N=2,3$, for a wide range of temperatures spanning from $0.6$-$10^{15}$ GeV at sufficiently large volumes. Comparing these results with sphaleron rates in a non-thermal SU(N) plasma where the infrared gluons are over-occupied, we estimate the typical thermalization time for these ultra-soft gluons during the early stages of reheating after inflation. We have also calculated the non-perturbative thermal axion production rate using lattice techniques which shows significant deviation from its perturbative estimate even at the electroweak scale.

We investigate observational tests that can be used to distinguish between broad classes of cosmological models. This is achieved using curvature-consistency tests of the Friedmann-Lemaître-Robertson-Walker (FLRW) models, which we investigate in two scenarios where they can be violated; (i) when the optical properties of the cosmology deviate from the expectations of FLRW, and also (ii) when the large-scale expansion of the cosmology is different from FLRW. We identify useful ways to determine the properties of these alternative scenarios in terms of the violation of the curvature-consistency tests, and propose a new null test that can be used to isolate cosmologies with non-FLRW observational relations. The characteristic signatures we find can be used, together with the results of recent and upcoming cosmological observations, to probe and/or rule out large classes of cosmological models. This becomes an increasingly important task as the number of proposals in the literature increases, as cosmologists attempt to explain tensions, anomalies, and the dark sector of the Universe. Our approach provides a clear route for telling apart these different proposals, and offers a new opportunity for using precision cosmological data to efficiently discriminate between cosmological models.

Little Red Dots (LRDs) have emerged as a key population linked to early black hole growth, yet photometric selections have predominantly targeted only the most extreme red systems, thereby shaping our current understanding of this new population of objects. In this work, we deliberately explore a broad range of optical redness while enforcing stringent compactness and visual inspection to ensure robustness and minimize contamination. Leveraging the depth and multiwavelength coverage of the JWST Advanced Deep Extragalactic Survey (JADES) data in the GOODS-North and GOODS-South fields, we construct the largest photometric census of LRDs to date in these fields, comprising 412 sources over $z\approx2\text{--}11$ across $\approx349.6$ arcmin$^2$. We show that classic extreme color cuts isolate only a minor fraction of this population ($\lesssim25\%$), while the majority of LRDs span a broader, largely unexplored parameter space. We quantify how selection strategies impact UV and optical luminosity functions and number density evolution, finding that current demographic trends of LRDs are strongly driven by selection biases and further limited by incomplete identification at both high and low redshift. Spectroscopically confirmed LRDs reveal a continuous range of spectral shapes, consistent with varying Active Galactic Nucleus (AGN) and host contributions in agreement with recent findings. Our results demonstrate that commonly adopted, purity-driven selections bias current demographic constraints toward the most extreme systems, potentially misrepresenting the diversity and evolution of the LRD population. Accounting for these selection effects is essential for interpreting LRDs and their role in early black hole growth.

Understanding the formation and evolution of late type galaxies (LTG) requires deep imaging for tracing the faintest stellar components in their outskirts. Despite their crucial role in the buildup of stellar mass, these low surface brightness (LSB) features remain largely unexplored due to observational limitations. The VST-SMASH is designed to fill this gap, providing deep, wide field optical imaging for a volume limited sample of nearby LTG, overlapping with the Euclid Wide Survey in the South. This paper aims to introduce the VST-SMASH survey and showcase its scientific potential through the analysis of IC 5332, a LTG observed in the g, r, and i bands. The main goal is to demonstrate the depth, quality, and diagnostic power of the dataset in tracing LSB features and structural components in galactic outskirts. We carried out detailed surface photometry of IC 5332 to extract radial surface brightness and color profiles down to LSB regime. We performed multicomponent Sersic decompositions and constructed stellar mass surface density profiles. We identified and characterized faint stellar streams, estimating their colors and comparing them with adjacent galactic regions. While the internal (1Reff) negative colour gradients can be explained by dissipative collapses and SN outflows, the color profiles at larger radii reveal a significant gradient toward redder colors, consistent with the presence of accreted populations in the outskirts. We also find bluer r - i, which could be explained by strong Ha emission. These findings support a scenario of ongoing stellar mass assembly through accretion and highlight the capability of VST-SMASH to uncover faint structures in nearby galaxies.(abridged)

We present a JWST/NIRSpec-IFU study of metallicity gradients in seven low-metallicity systems at $z=7.2-9.5$. The main sample spans stellar masses of $\rm \log(M_*/M_{\odot}) \sim 7.8-9.5$, star formation rates (SFRs) of $\rm \log(\text{SFR} / M_{\odot} \text{yr}^{-1}) \sim 0.5-2.5$, and gas-phase metallicities of $4\%-15 \%~Z_\odot$. Within our sample, we also identify three low-metallicity satellite galaxies associated with two of our sources, providing a rare view of early-epoch interactions. The three satellites exhibit even more primordial properties, with metallicity $3\% -4\% ~Z_\odot$ and low star-formation activity ($\rm \log(\text{SFR} / M_{\odot} \text{yr}^{-1}) \sim -0.5$ to $-0.9$). We find that our galaxies, and especially the satellites, are significantly offset from the local Fundamental Metallicity Relation (FMR), with deviations reaching $Δ\text{FMR} \approx -0.9$ dex. This indicates that these galaxies are likely experiencing strong accretion of pristine gas. Overall, we observe a large scatter in radial metallicity gradients, ranging from positive to negative with an average metallicity gradient of $\rm -0.02 \pm 0.04 \ dex \ kpc^{-1}$. Flat gradients are found in systems with confirmed satellites, suggesting that tidal interactions and mergers drive the radial mixing necessary to homogenise the interstellar medium. The (tentative) presence of an AGN in two of our sources suggests that strong feedback may also be responsible for the observed flat gradients. Conversely, the detection of a positive gradient in one source points toward a direct funnelling of metal-poor gas inflow into the central region of the galaxy. These results show that galaxies in the first billion years grow through diverse, episodic processes, suggesting that early evolution is characterised by structural variety rather than a single, predictable path.

Dust torus plays the key role in determining active galactic nuclei (AGN) observational appearance. Here, the scenario of accretion regulated central dust torus is tested for the first time in the individual changing-look AGN (CLAGN) SDSS J1011+5442. Through the dependence of broad H$α$ luminosity on continuum luminosity, the scenario of moving dust clouds can be ruled out in SDSS J1011+5442. Meanwhile, virial BH mass in the bright state is consistent with the M-sigma relation determined mass, indicating the virialization assumptions efficient in central BLRs. However, the virial BH mass determined in the dim state is 60 times smaller than the M-sigma relation determined value. The contrary properties of broad H$α$ in different states can be naturally explained by the scenario of accretion regulated dust torus. Below a critical Eddington ratio, opening angle of dust torus declines with increasing accretion rate, leading to only outer part of central BLRs for broad H$α$ with smaller line widths detected in the dim state but all the BLRs detected in the bright state. The results in this manuscript not only indicate properties of central dust torus having apparent effects on variability properties of CLAGN, but also indicate that studying CLAGN could provide further clues to check dynamical evolving models for dust torus in AGN.

Dust in the diffuse interstellar medium remains incompletely understood with regard to the structure, composition, size distribution, and alignment properties of the grains. Joint observations of reddening, starlight polarisation spectra, and polarised dust emission for individual sightlines provide essential constraints on such properties. We study a far-UV selected sample of 96 reddening curves, for which optical linear polarisation spectra were obtained with FORS at the VLT as part of the Large Interstellar Polarisation Survey (LIPS). Starlight polarisation spectra for 60 stars are presented in this work. These data are combined with Gaia distance estimates and Planck thermal dust emission. A three-component dust model is made publicly available. It consists of nanoparticles, amorphous grains, and micrometre-sized dust agglomerates, varying axial ratios, porosities, sizes, element abundances, and alignment efficiencies to match the observations. The diversity of reddening and polarisation spectra is well reproduced by prolate grains with typical axial ratios of two, a porosity of 10%, and high alignment efficiencies. Such efficiencies can be achieved with radiative torque alignment theory (RAT), but not with imperfect Davis-Greenstein (IDG) alignment, except when adjusting the magnetic-field orientation to maximise the polarisation. Micrometre-sized dust contributes wavelength-independent grey extinction in the optical, accounts for about one-third of the visual extinction, and carries one-third of the dust mass. A follow-up submillimetre survey with high-resolution polarimetry will further constrain grain shapes and alignment physics.

Extinction maps are essential for tracing interstellar dust and enabling accurate stellar population studies in galaxies. Here, a high-resolution extinction distribution of nearby galaxy M33 is constructed by fitting multiband color indexes of the individually resolved red giant branch (RGB) stars from the Panchromatic Hubble Andromeda Treasury: Triangulum Extended Region (PHATTER) survey. Achieving an angular resolution of approximately 6$^{\prime\prime}$ ($\sim$ 24.4 pc), the extinction map reveals the intricate and heterogeneous distribution of dust throughout the entire disk of M33, with distinct delineation of spiral arms, inter-arm regions, and compact dust clouds. In addition, it exhibits strong spatial correspondence with the distributions of total hydrogen, H I, and CO, underscoring the reliability of the extinction map for tracing both diffuse and dense components of the interstellar medium. The derived $V$-band extinction reaches up to 2.5 mag per pixel, with a mean value of about 1.05 mag. Beyond providing new insights into the dust structure of M33, the extinction map offers a robust foundation for accurate extinction corrections and will support future studies, including upcoming observations with the Chinese Space Station Telescope.

We examine the relationship between the mass of present-day central supermassive black holes (SMBHs, $M_{\rm BH}$), and the stellar mass ($M_{\star}$) and halo mass ($M_{200}$) of their host galaxies in the EAGLE simulation, and find that scatter about these relations correlates with both halo structure and galaxy morphology. EAGLE reproduces the observed $M_{\rm BH}$-$M_{\star}$ relation, including (qualitatively) its dependence on morphology: at fixed $M_{\star}$, disc-dominated galaxies host less massive SMBHs than ellipticals. We show that $M_{\rm BH}$ correlates with $M_{200}$, as expected if SMBHs are regulated by processes acting on the scale of the host dark matter halo, but exhibits a tighter correlation with the halo binding energy ($E_{\rm bind}$), signalling that this quantity, which encodes information about both halo mass and halo structure, is more fundamental to $M_{\rm BH}$. As with $M_{\rm BH}$-$M_{\star}$, scatter about the $M_{\rm BH}$-$E_{\rm bind}$ relation is strongly correlated with morphology. Gas in the central few parsecs of galaxies with present-day discs retains strong rotational support as the galaxy grows, inhibiting inward transport and precluding periods of rapid SMBH growth by gas accretion. In galaxies destined to be present-day ellipticals, however, this rotational support is disrupted, enabling gas to be funnelled onto the central SMBH, triggering rapid growth. Evolution of the mass fraction of stars formed ex-situ indicates that this disruption is caused by galaxy-galaxy interactions and mergers. Our findings corroborate the conclusion of recent studies, based on controlled simulations of an ~$L^{\star}$ galaxy, that prolonged secular galaxy evolution inhibits central SMBH growth.

We perform a comprehensive observational test of a canonical quintessence model driven by an exponential potential, motivated by its emergence in higher-dimensional theories, string-inspired scenarios, and modified gravity. Using a Markov Chain Monte Carlo framework, we constrain the model with the latest high-precision observational datasets including Cosmic Chronometers, Baryon Acoustic Oscillation, Pantheon+, and DES-SN5YR Type Ia Supernovae. The combined data significantly tighten the parameter bounds on (H0, Omega_m0, eta0, gamma) and yield predictions for the Hubble parameter H(z), the distance modulus mu(z), and the scaled comoving angular diameter distance that remain in excellent agreement with observations and closely follow the LCDM baseline. An information-theoretic model comparison using the Akaike Information Criterion shows that the exponential quintessence model remains statistically comparable with LCDM despite having additional parameters. The model successfully reproduces the transition from matter domination to late-time acceleration, maintains w_tot > -1, and provides an age of the universe consistent with Planck 2018. Statefinder diagnostics indicate trajectories approaching the LCDM fixed point with small deviations, and energy condition analysis confirms physical viability, with only the Strong Energy Condition violated at late times as required for acceleration.

We analyze three NuSTAR observations and two NICER observations of the transient accreting millisecond pulsar SAX J1808.4-3658 in the hard spectral state during its most recent outbursts in 2022 and 2025. The spectral analysis of the persistent emission shows that the continuum is well described by an absorbed thermal Comptonization model with a high plasma temperature of ~25-90 keV. A prominent iron emission line around 5-8 keV and a Compton hump around 15-30 keV have been detected from all NuSTAR observations, indicating the reflection of the hard X-ray photon from the accretion disk. We employ the relativistic reflection model relxillCP to describe the reflection phenomena. The spectral fit of three NuSTAR observations shows that the inner disk radius moves outward, the Comptonized thermal emission decreases in flux, the mass accretion rate decreases, and the disk becomes less ionized as we proceed from the 2022 to the 2025 observations. Reflection studies also reveal a moderate inclination of the source within ~30-50 degrees. During the 2025 September observation, the inner radius of the disk is significantly truncated (~23R_g), and the corresponding magnetospheric radius is comprehensively larger than the disk's co-rotation radius, suggesting a hint of the transition to the propeller regime. Although the disk is truncated at the larger radius, accreted material is still reaching the surface of the neutron star, which is confirmed through the detection of a Type-I X-ray burst during this NuSTAR observation. The spectral analysis of the burst suggests helium burning at a low ignition depth.

Blazars, the jet dominated class of AGN comprising flat spectrum radio quasars (FSRQs) and BL Lac objects (BL Lacs) are now increasingly identified as potential sources of high energy neutrinos. Such neutrino blazars are ideal targets to investigate the high energy emission processes and to understand their role as neutrino sources. We report results on four candidate neutrino blazars, PKS 0446+112, TXS 0506+056, PKS 1424$-$418 and PKS 1502+106. We carried out $γ$-ray spectral and timing analysis on three time periods that comprise a quiescent epoch, an epoch that corresponds to neutrino detection and a flaring epoch. We also carried out modeling of the broadband pectral energy distribution (SED) on those three epochs. We found that the $γ$-ray spectra of the BL Lac TXS 0506+056 can be adequately described by a power-law, while the spectra of the other three FSRQs require a log-parabola model. On shorter timescales, we observed flux variability with doubling/halving timescales of 4.70 hrs, 9.24 hrs, 30.76 hrs and 15.42 hrs for PKS 0446+112, TXS 0506+056, PKS 1424$-$418 and PKS 1502+106, respectively. The SEDs of most of the epochs for the sources are well explained by a leptonic scenario. However, the quiescent epoch of PKS 1502+106 and the neutrino-emission epoch of PKS 0446+112 required an additional hadronic component to reproduce the observed SEDs. Our analysis reveals a complex interplay of leptonic and hadronic processes. While certain neutrino-associated epochs align with a leptonic model, others necessitate a hadronic component to explain the emission features.

The approximately flat outer parts of spiral galaxy rotation curves are commonly interpreted as evidence for a discrepancy between the observed baryonic mass and the dynamical mass inferred from the measured orbital velocities. In most standard analyses, this discrepancy is quantified using $v^2(R)=GM(<R)/R$, which is exact only under spherical symmetry. However, spiral galaxies are flattened disk systems, for which mass exterior to the galactocentric radius under consideration can contribute non-negligibly to the gravitational field. We introduce the Lost and Found (LF) model, a geometrically consistent Newtonian framework based on direct full-disk gravitational integration and a parametrized representation of the disk surface density. In this approach, the gravitational field is computed without imposing spherical symmetry, and the disk mass distribution is represented by two exponential components with a smooth outer truncation. We apply the LF model to a heterogeneous sample of disk galaxies spanning a broad range of masses and radial extents. The model reproduces the main observed features of the rotation curves, including the inner rise and the approximately flat outer behavior, without explicitly invoking a dark matter halo or modifying Newtonian gravity. Across the sample, the LF-inferred mass scales nearly linearly with the conventional dynamical mass, with a characteristic reduction factor $η_{LF}$ ~ 0.67. These results indicate that part of the inferred mass discrepancy may arise from the geometric treatment of gravitation in disk galaxies, and motivate a reassessment of mass inference in non-spherical systems.

The orbital decay of binary systems is a critical process for understanding the evolution of massive binary black holes (MBBHs) and binary star formation. Performing high-resolution three-dimensional magnetohydrodynamic (MHD) simulations, we investigate a binary system that accretes gas from an infalling envelope analogous to the collapse of molecular clouds in the context of binary star formation. Our simulations reveal the presence of outflows/jets launched from both the circumstellar (mini) disks and the circumbinary disk (CBD). The magneto-rotational instability is also excited within the CBD. These magnetic processes efficiently extract orbital angular momentum from the binary and thus drive orbital decay, while a purely hydrodynamical model exhibits orbital expansion. The decay rate reaches $\sim 0.3-0.7\%$ per orbital period, depending on the initial magnetic field strength. By appropriately scaling these numerical results, we propose a new mechanism for MBBHs mergers within a Hubble time, overcoming the bottlenecks encountered at separations near the final parsec scales. Additionally, we present a formation scenario for close twin binary star systems, emphasizing the significant role of magnetic processes in driving orbital evolution across various astrophysical systems.

We present a low-redshift-agnostic compression of anisotropic baryon acoustic oscillation (BAO) distances to constrain the normalized dark-energy density evolution, $X(z)\equiv ρ_{\rm DE}(z)/ρ_{\rm DE}(0)$, above the lowest BAO redshift node $z_1$. Standard BAO summaries include the transverse comoving distance $D_{\rm M}/r_{\rm d}$, which depends on the integral of $H^{-1}(z)$ from $z=0$ to $z$ and therefore mixes the expansion history at $z<z_1$ with the higher-redshift signal. We instead replace the set $D_{\rm M}(z_i)/r_{\rm d}$ by adjacent increments $ΔD_{\rm M}(z_i,z_{i+1})/r_{\rm d}$ while retaining the radial distances $D_{\rm H}(z_i)/r_{\rm d}$. The mapping is linear, so the covariance propagates exactly. This compression intentionally removes one absolute transverse-distance mode, namely the additive contribution to $D_{\rm M}/r_{\rm d}$ below the first BAO node, and preserves the remaining information relevant to reconstructing the expansion history above $z_1$. Applied to DESI DR1 and DR2 anisotropic BAO measurements, the method yields almost uncorrelated constraints on piecewise-constant interval parameters $X_j$. In this sense, the compressed likelihood provides a conservative band-power-like estimate of dark-energy evolution: each interval is constrained mainly by BAO information from its own redshift range, while one nonlocal transverse mode and stronger global assumptions are deliberately projected out or marginalized over. Because our baseline analysis also marginalizes over bin-local matter-density and distance-scale parameters with broad external priors, the resulting $X_j$ constraints should be interpreted as a low-redshift-agnostic BAO baseline rather than as a fully prior-free reconstruction. All bins are consistent with $X=1$ within current uncertainties.

The discovery of the phase space spirals in the Solar neighborhood in Gaia Data Release 2 has prompted various attempts to understand their origin. A source of bending waves, which has been neglected as a cause of the phase spiral, is irregular gas inflow along the warp. We aim to study whether perturbations by the gas warp could induce phase spirals. Accounting for this additional formation scenario for phase spirals could improve our current understanding of the perturbation history of the Milky Way disc. We use two N-body + SPH (Smooth Particle Hydrodynamics) simulations of an isolated galaxy to search for, and study, warp-induced phase spirals. We study the emergence and propagation of the detected phase spirals using Fourier decomposition. We detect strong one-armed phase spirals in the warped simulation. These phase spirals are prevalent and persist over ~10 Gyr. The morphology of these phase spirals varies with location and evolves with time. In particular, the emergence rate of the phase spiral evolves with the gas inflow at the outer disc and the bending wave amplitude, indicating that these phase spirals are a record of warp-induced bending waves. We find that these phase spirals can reach amplitudes comparable to those in the Gaia DR3. We only detect weak and stochastically distributed phase spirals in an unwarped control simulation. We conclude that phase spirals can be induced by the irregular gas accretion along the warp. These phase spirals occur globally and are long-lived.

Millisecond magnetars, rapidly rotating neutron stars with ultra-strong magnetic fields, have long been proposed as central engines of gamma-ray bursts (GRBs). For GRBs produced by neutron star mergers, the survival of a long-lived magnetar remnant remains uncertain, as the merger remnant may rapidly collapse into a black hole. In GRB 230307A, multiwavelength observations together with a previously reported 909-Hz periodic signal consistent with millisecond spin in its prompt emission provide strong evidence that such a post-merger magnetar may power the burst. Here we report the discovery of a quasiperiodic modulation with a characteristic period of 4.5 s in the spectral evolution of GRB 230307A, detected consistently across multiple gamma-ray instruments. The modulation is manifested as a coherent, energy-dependent variation of the spectral shape, with the strongest signature in the evolution of the peak energy. Within the magnetar-engine framework, such a low-frequency modulation can be interpreted as a manifestation of large-scale periodic variations associated with the central engine. If interpreted in terms of free precession, the observed timescale implies a stellar ellipticity of $ε\gtrsim 2.4 \times 10^{-4}$, corresponding to an internal magnetic field strength of $B_t \gtrsim 1.6 \times 10^{16}$ G, alongside a dipole field of $B_p \approx 5.6 \times 10^{15}$ G inferred from the early X-ray emission. These results suggest that such systems may provide potential sources of post-merger gravitational waves (GWs), motivating targeted searches following GRB triggers.

Due to its brightness and proximity, the radio galaxy 3C 84 (optical counterpart NGC 1275 in the Perseus cluster) has been the target of extensive studies investigating the central parsec region of its active galactic nucleus. In 2003, its most recent active phase resulted in a plasma ejection visible in the southern jet, which presented a unique opportunity to study jet formation and evolution at high angular resolution with very long baseline interferometry (VLBI). We aim to study the morphology, evolution, and spectral properties of the restarted jet three years after the first ultra-high angular resolution observations with the RadioAstron space-VLBI satellite in September 2013. To study 3C 84, we used space-VLBI observations carried out in September 2016 at 22 GHz with a global VLBI network and the 10 m Spektr-R radio telescope in orbit as well as quasi-simultaneous multifrequency observations at 4.8, 8, 15, and 43 GHz from the Very Long Baseline Array, including the Effelsberg 100 m telescope. We present the 22 GHz RadioAstron image of 3C 84 from 2016, which reveals the source's central region at a 58 microarcsecond effective resolution. During the three years that elapsed between the first and second space-VLBI observations, the source underwent significant morphological changes. We confirm the existence of the limb-brightened jet and counter-jet reported earlier as well as a flip in the position of the hotspot discovered recently via VLBI monitoring at 43 GHz. Based on measuring the collimation profile, we find that it has evolved from being quasi-cylindrical to parabolic. This is most likely the result of the decreased pressure of the mini-cocoon, which was inflated by the jet and contains hot gas that cannot confine the jet efficiently as it propagates further away from the core. Finally, we also constrained the magnetic field strength in the core region and the hotspot.

Short and long-short gamma-ray bursts (GRBs) are widely believed to be powered by neutron star mergers. In this work, we calculate local rate of such GRBs and find a relatively high value of $\sim 786-2468~{\rm Gpc^{-3}~yr^{-1}}$ when including the very narrow collimation event GRB 061201. Considering that its redshift is not very reliable, after excluding this event, the rate is $\sim 195-666~{\rm Gpc^{-3}~yr^{-1}}$. We also calculate the electromagnetically (EM) bright neutron star merger rate inferred from the LIGO/Virgo/KAGRA observations up to the end of the first epoch of the O4 run, and derive a rate of $\sim 66-347~{\rm Gpc^{-3}~yr^{-1}}$. This rate is somewhat lower than the value obtained from the GRBs, even after excluding GRB 061201. The non-detection of any viable EM bright merger in the O4b and O4c observing runs favors an even lower rate, which starts to challenge the neutron star merger origin of the short and long-short GRBs and may suggest additional contribution from the mergers of other compact object (like the neutron star-white dwarf) binaries, as speculated initially by King et al. (2007) in interpreting the long-short event GRB 060614.

While collisionless plasmas are ubiquitously present near astrophysical compact objects, the impact that their composition has on the high-energy emission is presently unknown. We present the first investigation of particle-acceleration mechanisms in kinetic, special-relativistic turbulence, modeling electrons, positrons, and protons with realistic mass ratios. Under global charge neutrality, we introduce a positron fraction and cover regimes ranging from an electron-proton plasma over to pair-dominated plasmas. Using a novel generalized Ohm's law for multispecies relativistic plasmas, we analyze particle acceleration due to electric fields in reconnection events that self-consistently emerge from turbulence. We demonstrate, for the first time, that energization occurs at reconnection current sheets driven by the divergence of the relativistic pressure tensor, which locally aligns with the particle velocity and leads to an efficient energy transfer. The imbalance between electrons and positrons systematically favors electron acceleration, highlighting the necessity of realistic multispecies modeling to capture the nonthermal contributions in accretion flows and relativistic jets from black holes.

We report results from the simultaneous XRISM (183 ks) and NuSTAR (62 ks) observations of the Seyfert-2 galaxy NGC 4388. This AGN has the brightest Fe K$α$ line among Compton-thin, obscured sources. To model the reflection continuum and fluorescent lines, we employ an updated version of XCLUMPY and a broad line region model with a disk-like geometry. The profile of the neutral Fe-K fluorescent line is well described as the sum of three components convolved with Gaussians with FWHM values of $\sim 290\ \mathrm{km\ s^{-1}}$, $\sim 1470\ \mathrm{km\ s^{-1}}$, and $\sim 11100\ \mathrm{km\ s^{-1}}$. These line widths correspond to radii of 1.5 pc, 0.060 pc, and $1.0\times10^{-3}$ pc by assuming Keplerian motion, which we interpret as the dusty torus, its inner edge region, and the BLR, respectively. The data suggest that the Fe K$α$ BLR component is larger than that of H$α$ (FWHM of 4500 $\mathrm{km\ s^{-1}}$) in the polarized optical spectrum, implying that the velocity field of the BLR is dominated by that parallel to the equatorial plane. In addition, Fe XXVI Ly$α$ and Fe XXV absorption lines are detected, characterized by $\logξ \sim 3.50~\mathrm{erg\ cm\ s^{-1}}$, $\log{N_{\mathrm{H}}} \sim 22.1~\mathrm{cm^{-2}}$, $v_{\mathrm{out}} \sim 40\ \mathrm{km\ s^{-1}}$, and $σ_v \sim 160\ \mathrm{km\ s^{-1}}$. We infer that the absorber is gravitationally bound and is possibly associated with a failed wind, consistent with a radiation-driven fountain flow.

We provide a comprehensive analysis of GW190711_030756 and GW200114_020818, two of the most significant binary black hole merger candidates in the IAS catalog, with probabilities of astrophysical origin $p_{\rm astro}=0.99$ and $0.71$, respectively, and signal-to-noise ratios of approximately $10.0$ and $13.4$. We employ numerical relativity surrogate models to infer both the source properties and the remnant properties of these two candidates. We find that both GW190711_030756 and GW200114_020818 are asymmetric-mass binaries, with inferred mass ratios of $0.35^{+0.32}_{-0.15}$ and $\leq 0.20$. In addition, GW200114_020818 is inferred to have a source-frame total mass of approximately $220M_{\odot}$ and highly spinning black holes, with primary (secondary) dimensionless spin magnitudes of $0.96^{+0.03}_{-0.07}$ ($0.84^{+0.13}_{-0.34}$), closely resembling GW231123_135430. We further find that GW200114\_020818 has a confidently negative effective inspiral spin of $χ_{\rm eff}=-0.60^{+0.22}_{-0.13}$ and exhibits strong spin precession, characterized by an effective precession parameter of $χ_{\rm p}=0.60^{+0.21}_{-0.19}$. GW200114_020818 (when considered alongside GW231123_135430) points towards an emerging population of massive, rapidly spinning BBH mergers. While GW231123_135430 is consistent with mergers in globular clusters, producing systems like GW200114_020818 in such environments remains difficult even under hierarchical merger scenarios. The probability that the remnant black hole of GW190711_030756 (GW200114_020818) is retained in its host environment is $0.079$ ($0.0002$), $0.62$ ($0.965$), and $0.997$ ($1$) if the merger occurred in a globular cluster, a nuclear star cluster, or an elliptical galaxy, respectively.

We discover an intrinsic degeneracy in the semi-analytic two-spot model for parameter inference in thermal X-ray pulse-profile modeling. Although this degeneracy exists in our simplified model with two small circular hot spots and without Doppler effects, it causes the likelihood surface of parameter inference based on ST-U and more complex models with Doppler effects to have multi-modal structures. Consequently, the posterior surface may also exhibit multi-modal structures if there is insufficient prior knowledge of parameters. Because of this, the inferred value of the neutron star radius can be biased even by $30\%$. This finding also provides a promising way to explain the multi-modal structures discovered in the evaluation of recovery performance using synthetic pulse-profiles that mimic the PSR J0030+0451 pulse-profiles~\citep{Vinciguerra2023,Vinciguerra2024}. Our work may have profound implications for the reanalysis of NICER data and the analysis of upcoming eXTP data.

We present an end-to-end, iterative pipeline for efficient identification of strong galaxy--galaxy lensing systems, applied to the Euclid Q1 imaging data. Starting from VIS catalogues, we reject point sources, apply a magnitude cut (I$_E$ $\leq$ 24) on deflectors, and run a pixel-level artefact/noise filter to build 96 $\times$ 96 pix cutouts; VIS+NISP colour composites are constructed with a VIS-anchored luminance scheme that preserves VIS morphology and NISP colour contrast. A VIS-only seed classifier supplies clear positives and typical impostors, from which we curate a morphology-balanced negative set and augment scarce positives. Among the six CNNs studied initially, a modified VGG16 (GlobalAveragePooling + 256/128 dense layers with the last nine layers trainable) performs best; the training set grows from 27 seed lenses (augmented to 1809) plus 2000 negatives to a colour dataset of 30,686 images. After three rounds of iterative fine-tuning, human grading of the top 4000 candidates ranked by the final model yields 441 Grade A/B candidate lensing systems, including 311 overlapping with the existing Q1 strong-lens catalogue, and 130 additional A/B candidates (9 As and 121 Bs) not previously reported. Independently, the model recovers 740 out of 905 (81.8%) candidate Q1 lenses within its top 20,000 predictions, considering off-centred samples. Candidates span I$_E$ $\simeq$ 17--24 AB mag (median 21.3 AB mag) and are redder in Y$_E$--H$_E$ than the parent population, consistent with massive early-type deflectors. Each training iteration required a week for a small team, and the approach easily scales to future Euclid releases; future work will calibrate the selection function via lens injection, extend recall through uncertainty-aware active learning, explore multi-scale or attention-based neural networks with fast post-hoc vetters that incorporate lens models into the classification.

We present XRISM/Resolve high-resolution X-ray spectroscopy of the prototypical magnetic cataclysmic variable AM Herculis. All satellite lines of highly ionized Fe are fully resolved. Lighter element lines (Si, S, Ca) show 2 - 3 eV widths consistent with purely thermal broadening, while the broader 6 - 7 eV Fe lines require additional bulk Doppler broadening. Spin-phase-resolved modulations are clearly detected in the Fe XXV and Fe XXVI lines, with semi-amplitudes of $81.8\pm6$ km s$^{-1}$ and $132.5\pm9$ km s$^{-1}$, and mean velocities of $143.6\pm6$ km s$^{-1}$ and $225.6\pm8$ km s$^{-1}$, respectively. After removing these bulk Doppler shifts, we obtain intrinsic Doppler widths of $5.23_{-0.15}^{+0.16}$ eV for Fe XXV and $6.23_{-0.18}^{+0.19}$ eV for Fe XXVI, directly revealing gradients of bulk velocity and temperature in the cooling-flow plasma. We additionally examined the resonance anisotropy predicted by Terada et al. (1999, 2001): the equivalent widths of the Fe XXV and Fe XXVI resonance lines increase at the pole-on phase by factors of 1.30 - 1.35, in positive correlation with their oscillator strengths. Combining XRISM with simultaneous NuSTAR data and PSAC/MCVSPEC plasma models, we derive a self-consistent shock temperature of $24.0\pm0.1$ keV and shock velocity of $1,116\pm2$ km s$^{-1}$. Radiative transfer simulations of the resonance lines further constrain the shock density to about $(5 - 6)\times10^{15}$ cm$^{-3}$, providing a new density diagnostic for accretion columns. The resulting accretion column geometry has a height of 200 - 300 km and a radius of 200 - 400 km.

Gamma-ray bursts (GRBs) have traditionally been classified by their prompt emission duration and spectral hardness, with short GRBs (sGRB; $\lesssim2 \ \rm{s}$) originating from compact object mergers and long GRBs (LGRB; $\gtrsim2 \ \rm{s}$) from massive star core-collapse. Recent kilonova (KN) associations with long-duration GRBs have challenged this standard picture. We analyze the host galaxies of nine GRBs with associated kilonova candidates at $z<0.6$, including five sGRB-KNe and four LGRB-KNe. Using both parametric and non-parametric modeling of the host light distributions, we investigate the progenitor environments of these events and test whether their hosts show evidence for recent galaxy interactions that could favor dynamical formation channels or isolated pathways following merger-driven star formation episodes for neutron star binaries. We find that five of the nine hosts display tidal features that show they have likely undergone recent mergers, suggesting that merger-driven, dynamical formation pathways may contribute in some systems. We find no clear morphological distinction between sGRB-KN and LGRB-KN hosts as both populations span a wide range of morphologies, including ellipticals, spirals, and interacting systems with tidal features. Multi-Sérsic modeling of the host light profiles further shows that host-normalized offsets inferred from single-Sérsic fits can be overestimated when the transient is associated with a specific subcomponent of a complex host light profile. These results highlight the importance of decomposing host morphology into physically relevant components when interpreting GRB environments and galactocentric offsets.

We identify shell-like tidal structures in flattened haloes that appear stream-like under different projections. This projection dependence demonstrates how changes in the host halo directly impact the formation and classification of tidal debris, highlighting the challenges of relying solely on visual inspection. To address this, we employ our clustering-based classification framework to systematically categorise tidally disrupted satellites into stream-like and shell-like structures. Our host consists of a static three-component MW model with flattening introduced along the z-axis NFW dark halo. We consider three halo shapes: spherical q = 1, extremely oblate q = 0.5, and prolate q = 1.5. We evolve three subhalo types: a highly radial massive subhalo favouring shell formation, an eccentric orbit leading to stream formation, and an intermediate orbit. We first classify the tidal structures visually using face-on and edge-on density projections of the 3D position distribution. This reveals shell-like and stream-like formations across face-on projections, while edge-on views lead to contrary classifications in some cases. To resolve these ambiguities, we apply the classification method developed in our earlier work, analysing structures in ordered density, radial, and energy-angle space. We further investigate the spatial dispersion of stream-like structures and the rate at which core density reduces as the flattening parameter varies. Our results demonstrate that halo shape variations affect tidal debris formation and classification, as well as the spatial dispersion and core density evolution of streams. These findings offer new insights into the role of dark matter halo geometry in shaping tidal structure formation and its contribution to hierarchical galaxy formation and evolution.

We utilize JWST/MIRI IFU observations from the Galaxy Activity, Torus and Outflow Survey (GATOS) to investigate the diverse range of ionized outflow rates of obscured AGN with similar bolometric luminosity and explore potential associations with AGN feedback. We explore spatial correlations between ionized emission potentially associated with fast shocks ([Fe II]5.34μm) and the excitation of H2. We further constrain our investigation to the inner 400 pc (the nuclear and circumnuclear regions r < 200 pc), and estimate the excitation temperature and column density of H2 assuming local thermodynamic equilibrium (LTE) and using the S(1) to S(8) rotational H2 emission lines visible to JWST/MIRI spectroscopy. We report the molecular gas temperature of the deprojected 400 pc nuclear region to correlate with the ionized mass outflow rate. We also observe the stronger degree of spatial correlation between [Fe II]5.34um emission and H2 gas temperature. We observe regions of enhanced [Fe II]5.34μm / [Ar II]6.99μm spatially coincident with the ionization cones of objects with higher ionized outflow rate and [Fe II]5.34μm / [Ar II]6.99μm in the deprojected 400 pc nuclear region to scale positively with both ionized outflow rate and estimated molecular gas temperature. We do not observe the estimated jet cavity power within the central 400 pc to strongly correlate with the ionized mass outflow rate or molecular gas temperature of the nuclear region. We take the preceding observations to suggest a higher degree of interaction between AGN outflows and the circumnuclear disk.

Close quasar pairs are rare products of galaxy mergers in which both supermassive black holes (SMBHs) are actively accreting, offering strong constraints on merger-driven active galactic nuclei (AGN) evolution. Identifying close quasar pairs at $z\gtrsim4$ is challenging due to the declining quasar number density in the early Universe. Here we report the confirmation of a close quasar pair at $z=5.7$, J2037--4537, utilizing high-resolution Atacama Large Millimeter/submillimeter Array (ALMA) observations. The quasar host galaxies exhibit tidal disturbed features in both the far-infrared continuum emission and the {\cii} line emission, ruling out the doubly-imaged lensed quasar scenario. The two quasar hosts are massive $(M_\text{dyn}\gtrsim10^{10}M_\odot)$ and star-forming (SFR $\gtrsim500 M_\odot~ \mathrm{yr^{-1}}$). The confirmation of J2037--4537 puts a lower limit on the quasar pair fraction at $5.5<z<6$, $F_\text{pair}>1.2\%$, which is much higher than the quasar pair fraction at $z\lesssim4$. J2037--4537 is expected to form a gravitationally-bound SMBH binary within $\lesssim2$ Gyr. The elevated quasar pair fraction at $z>5.5$, as indicated by J2037--4537, likely contributes to the high gravitational-wave background reported by recent Pulsar Timing Array experiments.

Magnetars are usually interpreted as highly magnetized neutron stars, yet a small subset of low spin-down sources has motivated alternative scenarios involving highly magnetized white dwarfs. We test whether the observed magnetar sample is consistent with a single neutron-star population or whether the data favor an additional compact-object channel. We combine exploratory machine-learning diagnostics with hierarchical Bayesian population modeling. First, we apply principal component analysis and K-means clustering in $(P,\dot{P},L_X)$ space, and then train a Random Forest classifier with leave-one-out cross-validation to identify the observables driving the empirical split. We subsequently construct a hierarchical Bayesian mixture model that links spin parameters to magnetic-field distributions through covariate-dependent mixing fractions. Posterior inference is performed with Hamiltonian Monte Carlo, and predictive performance is assessed with Pareto-smoothed importance sampling leave-one-out cross-validation. The exploratory analysis reveals a reproducible sub-structure: the Random Forest reaches $>95\%$ LOOCV accuracy, with $L_X$, $\dot{P}$, and $kT$ emerging as the dominant predictors. However, the Bayesian comparison shows no statistically significant preference for a two-population model. Instead, a few low spin-down sources receive intermediate posterior membership probabilities, indicating that they are better interpreted as transitional or outlying objects than as members of a clearly distinct class. Overall, current data do not require a separate white-dwarf magnetar population. The main result is therefore conservative but strong: the observed sample is adequately described by a predominantly neutron-star population, while still allowing physically interesting deviations in specific sources.

Star formation occurs within dense regions of giant molecular clouds (GMCs), however, exactly how gas collects and evolves to form individual stars and what role dense cores play remains unclear. We use the Lagrangian cell information in the STARFORGE simulation suite to track star-forming gas in three GMCs with varying magnetic field strengths. We find that, once a protostar forms, the lifetime of the unaccreted gas correlates with the final stellar mass, where low-mass stars ($M_*$ < 0.5 M$_\odot$) accrete for 0.5-0.6 Myr from a relatively local reservoir of gas, and high-mass stars ($M_*$ > 2 M$_\odot$) accrete over 3.3-4.7 Myr from a much larger volume. Although the protostellar accretion time increases weakly with magnetic field strength, the accreting gas radii, velocity dispersions, virial parameters, and magnetic energy ratios are largely insensitive to the global cloud properties. At the time of protostar formation, the unaccreted gas exhibits linewidth-size and mass-size relations characteristic of turbulently regulated, isothermal dense cores, following $σ_v \propto R^{1.0-1.1}$ and $M \propto R^{0.47-0.55}$, respectively. Low- and intermediate-mass stars undergo relatively continuous accretion and their accretion histories are well-fit by either isothermal sphere, turbulent core, or competitive accretion models, where no one model fits all masses. However, many high-mass stars experience intermittent accretion and their accretion histories are not well-fit by any of these models. While the distribution of accreting gas is more extended than typically-defined dense cores, the physical properties and structure of the star-forming gas resemble those of observed cores and are largely regulated by turbulence and feedback.

Neutron star mergers are multimessenger sources whose dynamics and signals depend critically on neutrinos and their flavor transformations. We investigate whether fast and collisional neutrino flavor instabilities (FFIs and CFIs) arise in a GW170817-like post-merger accretion disk, and how they develop and relax, by performing global and local classical and quantum-kinetic simulations that resolve anisotropies and inhomogeneities in the full six-dimensional phase space. In the accretion disk, the neutrino radiation field naturally develops electron-lepton-number crossings through the interplay between the more isotropic electron neutrino field and the more anisotropic electron antineutrino field. The neutrino field in the disk is also unstable to CFI, although on longer timescales than the FFI. Using local, multi-energy quantum-kinetic calculations at selected points, we find that the growth of unstable modes is well-predicted by a fully anisotropic linear stability analysis and the flavor transformation increases the heavy lepton neutrino fluxes. CFI likewise enhances heavy-flavor fluxes, shows significant impacts from the growth of multi-energy anisotropic modes, and breaks the symmetry of the heavy-flavor sector by raising the average energy of heavy-flavor antineutrinos above that of heavy-flavor neutrinos. However, the CFI remains subdominant to the FFI in most of the disk. In our global quantum-kinetic simulations with an attenuated Hamiltonian, flavor coherence develops primarily in the polar regions. Because the attenuation causes advection to outpace the growth of the instabilities, coherence and flavor conversion remain artificially suppressed within the disk. These results emphasize the resolution and scaling requirements for future global simulations that capture instability growth, saturation, and advection simultaneously.

Close-binary evolution couples Roche-lobe overflow (RLOF), common-envelope (CE) drag, stellar winds, magnetic braking, and gravitational-wave losses, exchanging mass and angular momentum while reshaping orbits and spins. We present interoperable effects in the REBOUNDx extension to REBOUND that embed these processes within high-accuracy N-body dynamics. The suite includes: a momentum-conserving RLOF operator with conservative and systemic channels and configurable specific-j loss; a CE drag model based on Mach-dependent dynamical friction with kick limiting; isotropic Reimers winds, Parker-type thermal winds, and Eddington-limited outflows powered by a parametric stellar-evolution module supplying mass-dependent R and L; magnetic braking via the Verbunt-Zwaan/Kawaler torque with a saturation-aware closed-form spin update; and post-Newtonian corrections 2PN point-mass and spin-spin; 2.5PN radiation reaction. Linear momentum is conserved for conservative transfer, a minimal corrective torque enforces angular-momentum consistency, and adaptive sub-stepping stabilizes evolution near contact. Inter-module flags coordinate wind/RLOF/CE activity. The unit-agnostic framework enables self-consistent, time-resolved studies of close binaries in isolated or dynamically rich settings. Multiple examples and comparisons against other codes are provided in the Appendix. The code is available at https://github.com/malidib/ReboundS .

Here we investigate a possible presence of a fifth force at the Galactic Center (GC), and its potential influence on the stellar orbits around the central supermassive black hole of our Galaxy. For this purpose we simulated the stellar orbits in a Yukawa gravity model that predicts the emergence of a fifth force, and fitted them into the observed orbit of S2 star around Sgr A* at the GC. The fitting was performed using Markov chain Monte Carlo method which enabled us to constrain the parameters of Yukawa interaction describing the strength $δ$ and the range $λ$ of a fifth force. We studied the following cases for a fifth force range $λ$, when it is: i) about a few hundred AU (i.e. deep inside the orbit of S2 star), ii) about a thousand AU (i.e. approximately the size of S2 star orbit), and iii) several thousand AU (i.e. much larger than the size of S2 star orbit). The obtained results showed that as the range $λ$ of a fifth force increases, its strength $δ$ also increases and relative error $Δδ/δ$ decreases. The resulting fifth-force strengths in all three cases are respectively: $δ\sim$ 0.005, 0.02 and 0.15. These results are consistent with the corresponding results of both our previous studies and those of other authors, regardless of the different Yukawa-like potentials used to model a fifth force. In addition, we also studied whether the possible small discrepancies from the prediction of General Relativity for the Schwarzschild precession of S2 star could be caused by a fifth force. For this purpose we used the $f_\mathrm{SP}$ parameter that was recently measured in the case of S2 star by GRAVITY Collaboration in 2020. We found that the obtained estimates in all three cases are compatible, within the error intervals, with the measured value of $f_\mathrm{SP} = 1.10\pm 0.19$.

Type Ia supernova (SN) magnitudes present correlations due to the fact that their peculiar velocities are sourced by the large-scale structure of the Universe. This effect can be used to constrain properties related to the distribution and growth of matter perturbations. We analyze both Pantheon+ and Dark Energy Survey (DES-Y5) SN catalogues in combination with CMB data from Planck PR4 to constrain $σ_8$ in $Λ$CDM, optionally including both curvature and a modified growth index $γ$. We show that SN and CMB datasets are highly complementary and capable of measuring $σ_8$, $γ$ and $Ω_k$ simultaneously. Using only SN, we find $σ_8 = 0.73 \pm 0.22$ ($0.87 \pm 0.31$) for Pantheon+ (DES-Y5) in the base flat $Λ$CDM model. Interestingly, allowing for free $γ$ and $Ω_k$, we find hints of positive curvature: $Ω_k = -0.011 \pm 0.006$ $(-0.014 \pm 0.005)$, which exclude flatness at 2.2$σ$ (3.0$σ$), for the combination of CMB with Pantheon+ (DES-Y5). Such hints do not degrade if we also include a modified amplitude of CMB lensing, parametrized by $A_L$. We find that $γ= 0.519^{+0.061}_{-0.099}$ ($0.461^{+0.085}_{-0.069}$), which are consistent with the predictions of General Relativity. In terms of $fσ_8(z)$, we find $fσ_8(0.024)=0.461^{+0.066}_{-0.035}$ ($fσ_8(0.038) = 0.498^{+0.045}_{-0.050}$) for CMB + Pantheon+ (DES-Y5). Finally, the strong degeneracy between all three $Ω_k$, $γ$ and $H_0$ results in a broader CMB $H_0$ posterior. However, if we include SH0ES $H_0$ data, which is in known strong tension with the CMB in flat $Λ$CDM, we find that the $H_0$ tension is recast in terms of a significantly negative curvature and suppressed growth of structures.

The structures and dynamics of the interstellar medium are governed by a combination of self-gravity, external gravity, and various sources of ordered and random motions on different spatial scales. This paper uses ALMA CO (3-2) observations at 0.1" $\approx$ 5 pc resolution to examine the scale dependence of molecular gas structure and dynamics in the central molecular zone (CMZ) of a nearby galaxy, NGC 3351. We use the dendrogram technique to characterize hierarchical molecular gas structures spanning two decades in spatial scales and measure their size, gas mass, and velocity dispersion. Their size-linewidth relation shows a power-law slope of 0.58, comparable to measurements for CMZs in other galaxies and suggestive of significant contribution from ordered motion on large scales. We further decompose the observed velocity dispersion in each gas structure into ordered versus random motions. The former appears stronger in gas structures at $\gtrsim$ 30 pc while the latter becomes more dominant at $\lesssim$ 30 pc. Modulo uncertainties with the CO-to-H$_2$ conversion factor, the estimated gravitational free-fall time is comparable to the crossing time of ordered motions for structures on all spatial scales, and both becomes longer than the crossing time of random motions at small, $\lesssim$ 10 pc scales. Our results highlight the varying sources and drivers of gas motions on different spatial scales in the CMZ of a Milky Way-like galaxy.

Semi-analytic models (SAMs) have been treating galaxy populations as dynamical systems for $\gtrsim50$ years, but their evolution equations remain poorly constrained. We introduce sapphire, a modular, automatically differentiable, GPU-accelerated SAM written from scratch in JAX. For the first time, we compute exact Jacobian matrices of our nonlinear differential equations and show that they have interpretable, non-random structures, using the Pandya et al. (2023) physical model as an initial example. Both local and global sensitivity analyses reveal that supernova energy loading is a key astrophysical parameter for galaxy evolution. We use gradient descent and Hamiltonian Monte Carlo (HMC) to perform comprehensive mock parameter recovery tests. These indicate that the z=0 stellar-to-halo-mass relation alone does not contain enough information to infer many astrophysical parameters. Using observations of star-forming galaxies from the MaNGA survey and the Behroozi et al. (2019) empirical model as one baseline, we derive multiple posteriors assuming different combinations of data, including z=0 interstellar medium gas fractions and metallicities. The inferred physical parameters suggest that galaxies self-regulate their star formation primarily through preventative rather than ejective feedback. Both Fisher and HMC forecasts demonstrate the potential of sapphire to enable precision inference for galaxy formation, but more work is needed to expand its library of models. We discuss how our unique blend of differentiability, massive GPU parallelization, numerical robustness and principled Bayesian methods sets the stage for hybrid physics-informed, data-driven discovery of galaxy formation astrophysics and cosmology. We make sapphire publicly available at https://github.com/virajpandya/sapphire.

We present the first implementation of an evolving dust grain size distribution (GSD) within a semi-analytic cosmological model (SAM) of galaxy evolution. This flexible model self-consistently accounts for stellar dust production, shattering, coagulation, accretion of gas-phase metals, and destruction in supernova-driven shocks and hot gas, successfully reproducing key observational constraints. The purpose of this paper is to present the key physical elements of this novel dust implementation in a SAM and to explore controlled numerical experiments to identify the mechanisms shaping the GSD and extinction law in galaxies. Our results show that the GSD evolves from a large-grain-dominated regime at high redshift to a flatter, MRN-like shape at low redshift. This transition occurs earlier for massive galaxies, at a characteristic metallicity determined by the galaxy depletion time. The resulting extinction curves show an increase of the UV/optical slope and a pronounced $2175$ A bump toward lower redshift, in good agreement with the extinction properties of the MW. Through numerical experiments, we find that once stars provide the initial reservoir of large grains, shattering and ISM accretion are the principal mechanisms driving the growth of small grains. When accretion is included, the model robustly reproduces the observed $z \approx 0$ dust masses, largely independent of the specific assumptions adopted for grain-size physics. The extinction properties of MW-like galaxies are also generally recovered, except in extreme cases, such as when grain velocities in turbulent media are assumed to be independent of grain size.

Discoveries of gamma-ray bursts (GRBs) have become commonplace in recent decades, totalling $\mathcal{O}(10^4)$ unique detections across various missions. However, there have been no confirmed discoveries of a gravitationally-lensed GRB, despite expected lensing rates of $\sim1$ in $10^{3}$. In light of this, we complete an archival search for lensed GRBs by cross-matching well-localised \emph{Swift}/XRT-detected bursts with a large all-sky sample of galaxy clusters as potential lenses. We find a total of 17 candidate lensed GRBs defined by a 2 arcminute search radius from a cluster in our sample. 14 of our candidates are either confirmed to be at higher redshifts than their cross-matched cluster, or are consistent with a higher redshift origin based on the Amati relation between $E_{p,i}$ and $E_{\rm iso}$ of GRBs, indicating they are, at some level, lensed by their nearby cluster. Using the Amati relation and the lens-GRB separation, we quantify the magnification experienced by each GRB. We find $μ< 10$ for all except for one candidate, GRB~071031, which is consistent with $μ> 10$, but is uncertain. Another candidate, GRB~050509B, does not have a directly measured redshift, but was previously assumed to be at the redshift of its nearby cluster, $z=0.225$. We produce a lens model of this cluster and show that GRB~050509B is consistent with $z>1$ and magnified by $μ\simeq2-6$. We present these findings in anticipation of future lensed GRB discoveries enabled by facilities such as the Vera C. Rubin Observatory in the coming years.

We present a parameterized ("toy") model for shock interaction and $γ$-ray emission in classical novae, in which a white dwarf envelope of mass $M_{\rm env}$ is removed over a timescale $τ$ (proportional to the nova speed class, $t_{2}$) in an outflow that accelerates on the same timescale to a terminal speed $v_{\rm f}$. Particle acceleration occurs at the reverse shock generated when the outflow collides with a thin, dense shell of slower material released earlier. Accelerated protons are then advected into the shell, where for typical ${ M_{\rm env}, τ, \text{and } v_{\rm f}}$ they radiate in the calorimetric limit, consistent with correlated optical and $γ$-ray emission seen in well-sampled novae. The maximum proton energy, set by a Hillas-like argument, scales with the thickness of the hot post-shock region. Recent work shows turbulent mixing of hot post-shock gas with cooler dense gas may limit this thickness to $\lesssim 10^{-4}$ of the shock radius, explaining low X-ray luminosities. Using this empirically motivated thickness, and assuming efficient magnetic amplification, we predict maximum proton energies $E_{\rm max} \sim 10$ GeV, consistent with $γ$-ray spectra of Fermi-detected novae near optical peak ($\sim τ$). However, as the shock and post-shock layer expand, $E_{\rm max}$ can grow to $\gtrsim 10$ TeV on timescales of a few $τ$, enabling potential detection by atmospheric Cherenkov telescopes. We encourage TeV follow-up of Fermi-detected novae weeks to months after the optical/GeV peak and quantify the most promising events.

Neutron stars, at their cores, are highly dense and, thus, are expected to have a number of exotic processes. This includes a possible phase transition to deconfined quark matter at the core, leading to a hybrid star. The quark matter is expected to additionally be color superconducting. The physics of superconductivity plays an important role in understanding the high density matter in the interiors of neutron/hybrid stars. At their high densities, additionally, both proton superconductivity and neutron superfluidity are expected. We study the effect of superconducting (quark/proton) matter, along with the internal magnetic field, leading to pressure anisotropy within hybrid stars. We aim to probe the effect of superconductivity, especially from color superconducting quarks, in hybrid star structure. We propose new phenomenological model anisotropy profiles within a one-dimensional framework. We model quark matter using the vector interaction enhanced Bag model, and hadron matter with the DD2 equation of state. A Maxwell construction joins both phases. We further investigate the possible observational signatures of these hybrid stars. These include mass enhancement and continuous gravitational waves, possibly arising from the anisotropy induced deformation, helping us further constrain our model and its physical parameters.

We use controlled N-body simulations to study the collisional exchange of energy between stars and dark matter in ultra-faint galaxies. We find that dynamical friction between stars and subsolar-mass dark matter particles results in the depletion of dark matter from the galaxies' centers, thereby transforming dark matter cusps into constant-density cores. The process is particularly effective in tidally limited galaxies with low stellar velocity dispersion. As high-mass stars sink toward the center of the dark matter halo, the dynamical-to-stellar mass ratio within the stellar half-light radius decreases monotonically. The stellar population of a dark matter-dominated galaxy is thereby compacted into a dense, baryon-dominated cluster, surrounded by a dark matter halo. Such a cluster would share the chemical composition of an ultra-faint galaxy, yet would be virtually dark matter-free within its half-light radius. We moreover find that the collisional cooling with dark matter particles provides an efficient pathway for the formation of stellar binaries in the contracting cluster. The contraction is eventually slowed down due to the decreasing central dark matter densities and the formation of stellar binaries. Our models highlight that the dynamical processes governing the faintest galaxies give rise to a rich phenomenology, blurring the line between the dynamics of globular clusters and galaxies.

Gas properties in the outer intracluster medium (ICM) are not well-constrained, as traditional probes lose sensitivity at Mpc distances. We show that the morphology of radio relics effectively encodes the power spectrum of the surrounding density fluctuations, and that they hence represent a new observational window. To demonstrate this, we use cosmologically motivated shock-tube simulations in which we systematically vary the coherence length, amplitude, and power-law slope of the upstream density power spectra. We then post-process our simulations with the cosmic ray electron spectral solver, Crest, thereby producing a suite of mock radio relics. We find that the downstream morphology of our simulated relics is independently sensitive to each of the aforementioned parameters. Specifically, we show that observed 'double strand' features can be formed by curved shock fronts in projection, and that the scale of these features maps directly to the fluctuation coherence length. Increasing the fluctuation amplitude, meanwhile, progressively lengthens the downstream extent of the relic, thus explaining why relics are observed to be broader than the idealised expectation. It also broadens the Mach number distribution across the shock, which simultaneously increases the integrated radio flux density and produces patchier emission. Finally, steepening the power-law slope makes 'double strand' features more likely, and additionally increases both the number of radio filaments oriented parallel to the shock front and their spacing. At higher Mach numbers, steepening the power-law slope can further lead to the production of curved radio filaments, which trace large eddies. We apply our analysis to the Toothbrush and Sausage relics, and find evidence for a typical fluctuation coherence length of ~500 kpc, a non-uniform amplitude, and power spectra that are steeper than a Kolmogorov-like scaling.

We present JWST/MIRI mid-infrared integral field spectroscopy combined with ALMA CO(2-1) observations of the post-starburst galaxy ESO 420-G13, hosting a low-luminosity AGN. The unprecedented spatial and spectral resolution of MIRI enables a detailed study of the molecular and ionised gas kinematics, excitation, and energetics in the nuclear kiloparsec, revealing the impact of AGN feedback in a system with modest radiative output. Despite its faint radio and X-ray emission ($L_{2-10keV} \sim 10^{40}$ erg/s), ESO 420-G13 exhibits powerful kinetic feedback in the form of massive molecular and ionised gas outflows, with a total kinetic power of $\sim 1.5 \times 10^{41}$ erg/s. This corresponds to a jet-ISM coupling efficiency of ~3.8%, within the range observed in more powerful AGN. The feedback is driven by a previously undetected compact jet, traced by collimated coronal-line and extended X-ray emission to >870 pc from the nucleus. The interaction is strongest ~370 pc north of the nucleus, where a fast ionised gas stream emerges perpendicular to the jet axis, coinciding with a bend in the jet direction. Enhanced velocity dispersion in warm H2 surrounds this gas stream, consistent with an expanding molecular bubble. Massive molecular outflows are detected at its edges; the blueshifted outflow is devoid of CO emission, likely due to CO destruction in shocks or by cosmic rays from the jet-ISM interaction. About 5% of the central molecular reservoir has already been expelled, and the remaining gas is turbulent and warm, suggesting an ongoing phase of AGN-driven feedback in this post-starburst galaxy. Our results highlight the enormous potential of mid-IR imaging spectroscopy to uncover jet-driven feedback in low-luminosity AGN. Without the spatially resolved MIRI diagnostics, the kinetic power of the AGN in ESO 420-G13 and its role in shaping the host galaxy ISM would have remained hidden.

We study the inflationary bispectrum generated by the tree-level exchange of a massive hidden-sector scalar during inflation. When the interaction between the inflaton and the hidden sector arises only from the leading boost-breaking operator of the Effective Field Theory (EFT) of inflation, the equilateral bispectrum for principal-series scalar exchange is known to be universally negative, independent of the sign of the coupling. We revisit this result within the full EFT operator basis. Using bootstrap methods, we construct the de Sitter-invariant seed four-point function and obtain the inflationary bispectrum via weight-shifting operators and a soft-limit procedure. While the equilateral bispectrum remains strictly negative when only the leading interaction is present, additional operators generate independent cubic structures whose contributions compete in the equilateral configuration. As a result, the sign of the bispectrum is no longer universal. We derive a critical ratio of interaction coefficients that separates regions of positive and negative equilateral bispectrum. We further study the effects of reduced sound speed $c_s<1$ and the exchange of multiple particles. In both cases, the critical ratio is modified, and for multi-particle exchange a positive equilateral bispectrum can arise even when the higher-order operator is subdominant. Our results show that the negativity of the equilateral bispectrum from massive exchange is not generic, but reflects a restricted operator structure in the EFT of inflation.

Astrophysical black holes appear well-represented by the Kerr metric, but this metric has the philosophical problem of a ring-like curvature singularity. We show that a phenomenological correction to the Kerr metric known as the Kerr-Hayward metric can eliminate the curvature singularity while preserving in detail many features of polarized black hole images now testable by the Event Horizon Telescope (EHT). To establish this, we produce new general relativistic magnetohydrodynamics (GRMHD) simulations of a magnetized plasma in a Kerr-Hayward spacetime, then we extend the EHT analysis framework to perform polarized radiative transfer in this spacetime. We detail our methodology for implementing this modified spacetime into an open-source pipeline. From fluid quantities such as the magnetic flux parameter and jet efficiency, to image quantities such as the polarization pattern and the photon ring structure, our results for the Kerr-Hayward metric appear functionally indistinguishable from the Kerr metric. Our study finds that under certain conditions, the singularity-free correction to the Kerr metric can yield observables that are effectively indistinguishable in EHT measurements.

When selecting a model to characterize an astrophysical population, it is crucial to assess whether that model fits the data and, if not, how it can be improved. To this end, posterior predictive checks (PPCs) are a widely-used statistical test of model fit when inferring gravitational-wave source populations. However, PPCs exhibit limitations when assessing single-event parameters with large measurement uncertainty, like the spin tilt angles of the binary black holes (BBHs) observable with the LIGO-Virgo-KAGRA (LVK) detectors. When single-event inference is prior-dominated, traditional PPCs fail to flag even very poor model fits. In this work, we assess the efficacy of various alternative PPCs on poorly-constrained parameters. We compare PPCs conducted on event- vs. data-level parameters (e.g. posterior samples vs. maximum likelihood points), and explore two additional event-level PPCs: partial predictive checks and split predictive checks. Independent of measurement uncertainty, we find that PPCs on maximum likelihood parameters are always more discerning of model misspecification than any event-level PPC. However, when investigating simulated GWTC-3.0-like catalogs, none of the alternative PPCs show significant improvement over those traditionally used, indicating that at that sensitivity, any limited information in the data about spin tilts is insufficient to diagnose model misspecification. Finally, we apply our suite of PPCs to the spin magnitude and tilt distributions inferred in the most recent LVK catalog, GWTC-4.0. We conclude that the Gaussian Component Spins model used therein under-predicts BBHs with large spin magnitudes and over-predicts those with perfectly anti-aligned tilts.

Too light primordial black holes evaporate and are therefore strongly constrained by various bounds, e.g. Cosmic Microwave Background distortion. However, if they are formed strongly clustered, the corresponding haloes may collapse in heavier black holes which may form the entirety of the dark matter of the universe. The indirect signal of such scenario is the production of a flat stochastic background of gravitational waves which is detectable by the Einstein Telescope.

In this paper we investigate the long range gravitational effect of curvature-scalar field non-minimal coupling, in the form of $ξR φ^2$, in the perturbative quantum gravity framework. Such coupling is most naturally motivated from the renormalisation of a scalar field theory with a quartic self interaction in a curved spacetime background. This coupling results in two scalar-$n$ graviton vertices which contain no explicit momenta of the scalar, qualitatively different from the usual, e.g. $κh^{μν}T_{μν}$-type minimal matter-graviton vertices. Assuming the dimensionless coupling parameter $ξ$ to be small, we compute the 2-2 scattering Feynman amplitudes between such scalars up to ${\cal O}(G^2 ξ)$. From the non-relativistic limit of these amplitudes, we compute the corresponding long range gravitational potential. There exists no tree level contribution $({\cal O}(ξG))$ here, and hence the one loop ${\cal O}(G^2 ξ)$ result is leading. Recently, the effect of a cosmological constant in such non-minimal interaction and the subsequent gravitational potential was computed. In this work we take the cosmological constant to be vanishing. The resulting potential is found to have $r^{-4}$ leading behaviour. We further extend these results for scalar-massive spin-1 and massive spin-1/2 scattering. Spin and polarisation dependence of the two body potential have been explicitly demonstrated. We discuss some possible physical implications of these results.

We investigate Lyman-alpha forest flux statistics in mixed fuzzy dark matter (FDM) and cold dark matter (CDM) cosmologies using the Fluctuating Gunn-Peterson Approximation (FGPA) applied to hybrid Schrödinger-Poisson and N-body simulations. We evolve the dark matter distribution from z = 120 to z = 2 for an axion mass ( m_22 = 0.01) and FDM fraction (f_A = 0.1), and compare two realizations with identical initial conditions: one evolved with a particle-only approximation and one with full wave--mechanical dynamics. We find that, despite near-degeneracy in the nonlinear matter power spectrum, the corresponding Ly α flux power spectra differ at the 10 percent level on intermediate scales. This discrepancy arises from a strong suppression of small-scale velocity power in the Schrödinger--Poisson evolution, which is not captured by N-body treatments with matched initial transfer functions. As a result, the flux statistics cannot be fully characterized by the matter power spectrum alone, but depend sensitively on the dynamical evolution of the velocity field. These results demonstrate that wave-mechanical effects in FDM leave distinct kinematic imprints in Ly α observables beyond those associated with initial-condition suppression. While our analysis is based on an idealized FGPA framework, it isolates a mechanism by which mixed dark matter models can break degeneracies present in standard structure-based probes, motivating further investigation with full hydrodynamical simulations.

We present cuRAMSES, a suite of advanced domain decomposition strategies and algorithmic optimizations for the ramses adaptive mesh refinement (AMR) code, designed to overcome the communication, memory, and solver bottlenecks inherent in massive cosmological simulations. The central innovation is a recursive k-section domain decomposition that replaces the traditional Hilbert curve ordering with a hierarchical spatial partitioning. This approach substitutes global all-to-all communications with neighbour-only point-to-point communications. By maintaining a constant number of communication partners regardless of the total rank count, it significantly improves strong scaling at high concurrency. To address critical memory constraints at scale, we introduce a Morton-key hash table for octree-neighbour lookup alongside on-demand array allocation, drastically reducing the per-rank memory footprint. Furthermore, a novel spatial hash-binning algorithm in box-type local domains accelerates supernova and AGN feedback routines by over two orders of magnitude (an about 260 times speedup). For hybrid architectures, an automatic CPU/GPU dispatch model with GPU-resident mesh data is implemented and benchmarked. The multigrid Poisson solver achieves a 1.7 times GPU speedup on H100 and A100 GPUs, although the Godunov solver is currently PCIe-bandwidth-limited. The net improvement is about 20 per cent on current PCIe-connected hardware, and a performance model predicts about 2 times on tightly coupled architectures such as the NVIDIA GH200. Additionally, a variable-Nrank restart capability enables flexible I/O workflows. Extensive diagnostics verify that all modifications preserve mass, momentum, and energy conservation, matching the reference Hilbert-ordering run to within 0.5 per cent in the total energy diagnostic.

Detection of parity violation on cosmological scales would have profound implications for fundamental physics. Motivated in part by recent measurements of parity-odd four-point correlation functions in BOSS and DESI luminous red galaxy samples, which probe parity violation in the scalar sector, we present the first measurement of parity-odd kurto spectra in spectroscopic galaxy survey data. We analyse two composite-field spectra, $\mathcal{P}_{2\times2}$ (vector--pseudo-vector) and $\mathcal{P}_{3\times1}$ (scalar--pseudo-scalar). Compared with parity-odd four-point correlation function analyses, the kurto-spectrum formalism performs physically motivated compression on the trispectrum into a substantially lower-dimensional data vector, allowing direct estimation of covariance matrices from mock catalogues and reducing sensitivity to covariance-modelling systematics. Using null-hypothesis $χ^2$ tests and cross-patch consistency checks, we find no evidence for a cosmological parity-violating signal in either survey. We examine the impact of the adopted mock catalogues and find that the high-fidelity mocks provide a better match to the data of both surveys than the approximate mocks. The DESI DR1 measurements exhibit a scatter smaller than that of BOSS DR12 by about a factor of four, consistent with the improved statistical precision expected from the higher tracer number density. Future DESI data releases, with larger volume and number density, together with larger suites of high-fidelity mocks, can enable significantly sharper tests of parity violation using kurto spectra.

We extended the radio K-band spectroscopic survey for organics in southern hemisphere dense cores by observing seven sources using NASA's Deep Space Network 70-m antenna in Canberra, Australia, over the frequency range of 18 to 25 GHz. Molecular column densities of NH$_3$, $c$-C$_3$H$_2$, HC$_3$N, HC$_5$N, CCS, C$_3$S, and $c$-C$_3$HD were derived for each source assuming LTE. The resulting column density ratios were compared with predictions of a state-of-the art astrochemical model to constrain the C/O ratio and chemical age of each source. Most cores have similar C/O ratios of $0.5 - 0.7$, much different from the best studied TMC-1 dense core characterized by a high C/O ratio of $\sim 1.4$. The chemical ages of the cores are also similar and fall between 0.6 and 5~Myr. The less dense cores tend to have the oldest chemical ages, as might be expected given that chemical timescales scale with density. Our results showcase the synergistic approach of combining radio observations using the DSS-43 antenna with state-of-the-art astrochemical models to study the chemical composition of southern hemisphere dense cores, enabling constraints on their C/O ratios and chemical ages, which remain largely unexplored.

PG~1553+113 is the blazar with the most-significantly detected periodic patter in its multiwavelength (MWL) emission, making it one of the most promising candidates for hosting a supermassive black hole binary. However, the presence of this periodic behavior in the X-ray band remains under debate, largely due to the lack of continuous monitoring. This has led to differing conclusions in previous studies. In addition, we aim to examine whether the recently identified linear long-term trends in the gamma-ray and optical bands also exist in the X-ray regime. Here, we evaluate the 2.1-year period in the X-ray light curve of PG 1553+113 using two dedicated monitoring campaigns with Swift-XRT and UVOT, guided by predictions of future oscillation phases. We also examine whether the long-term trend is present in X-rays, the potential periodic behavior of the X-ray power-law photon index, and its potential correlation to the X-ray flux. As a result, we find tentative evidence for a correlation between the predicted high-emission states in the gamma-ray band and those observed in the X-ray and UV bands. Therefore, we do not find a strong evidence of the same periodic pattern in X-ray. In addition, we find that the X-ray light curve is consistent with the presence of a long-term linear trend, in agreement with those previously reported in gamma-ray, optical, and radio. Overall, these results indicate that the X-ray emission is likely to share the same long-term behavior observed in the gamma-ray and optical bands. Nevertheless, the pronounced stochastic variability that characterizes the X-ray light curve limits our ability to draw firm conclusions regarding the presence of the periodic behavior.

The first detection of supernova burst neutrinos was achieved through the observation of SN1987A, almost four decades ago. However, neutrinos produced during the burning stages of a star prior to core collapse are yet to be detected. Detection of pre-supernova neutrinos could provide an early warning of an imminent supernova and allow the scientific community time to focus their resources on the observation and study of such an event leading to better understanding of these rare phenomena. Integrating pre-supernova models into a neutrino event generator would help to provide a unified framework for studying these neutrinos in current and next generation detectors. sntools is a neutrino event generator for supernova burst neutrinos, originally developed to study supernova model discrimination with Hyper-Kamiokande. Work to add support for pre-supernova event generation to sntools is presented, detailing the adaptations and additions to the code, with emphasis on how time binning can be optimised for a robust simulation, and also detailing the status of the validation process. The current status and capabilities of the package will be explained alongside plans for any further work and the intended use for the new functionality within the Hyper-Kamiokande Collaboration.

In this short comment, I discuss the relationship between the results presented in arXiv:2512.01403 and those previously published in Phys.~Rev.~D~101,~083011~(2020). The 2020 study provides a full Monte Carlo simulation of cosmic-ray interactions with the solar atmosphere using the FLUKA code, including realistic solar-atmosphere models, PFSS/Parker/BIFROST magnetic-field configurations, and predictions for gamma rays, electrons, positrons, neutrons, and neutrinos. Given the substantial scientific overlap -- particularly in the modelling of hadronic interactions, magnetic-field effects, cascade development, and comparison with Fermi-LAT observations -- a direct comparison is relevant to assess consistency and complementarity. Here I summarize the main points of agreement, highlight differences in modeling assumptions, and outline how the two approaches can jointly contribute to understanding high-energy emission from the solar disk.

We investigate several phenomenological dark energy parameterizations using a joint analysis of late-time cosmological observations, including cosmic-chromatometer measurements of the Hubble parameter, DESI DR2 baryon acoustic oscillation data, and the Pantheon+ Type Ia supernova sample. Our results show that allowing for a time-varying dark energy equation of state significantly improves the overall fit compared to $Λ$CDM. The present-day equation-of-state parameter departs from the standard cosmological constant value. In contrast, the evolution parameter in two-parameter models tends to be negative, indicating a possible time dependence of dark energy. However, the constraints on the evolution remain moderate, and current data cannot clearly distinguish the specific functional form of dark energy. Model comparison using information criteria suggests that dynamical dark energy models are favored over $Λ$CDM, with the most straightforward one-parameter extension emerging as the most parsimonious scenario. These findings indicate a mild preference for dark energy evolution, though future high-precision observations will be required for definitive conclusions.

The $Λ$CDM cosmological model faces increasingly significant and robust tensions among independent cosmological probes, prompting renewed scrutiny of its foundational assumptions. While General Relativity and the nature of dark energy are now routinely tested with cosmological surveys, less progress has been made testing the space-time geometry at the largest scales, and in particular testing the assumption that observables (distances, redshifs, expansion of space, etc.) on the largest scales are described by a single Friedmann-Lemaître-Robertson-Walker (FLRW) metric. In order to enable such tests, we introduce a model-independent framework that combines successive derivatives of the angular diameter distance, $d_A(z)$, with the line-of-sight expansion rate, $\mathcal{H}(z)$, to expose the physical content of well-known FLRW consistency relations. This allows us to perform diagnostic tests of the large-scale geometry, that are free of assumptions about dark energy and the theory of gravity on large scales. In addition, we derive a new nonparametric estimator for the cosmic density field that is independent of the Friedmann equations. This enables qualitatively new, observationally accessible tests of the FLRW framework and provides a stringent, model-independent diagnostic for departures from standard cosmology using current and forthcoming distance and expansion rate measurements.

Cluster catalogs selected by optical properties are subject to selection biases, primarily arising from unresolved systems along the line of sight. These biases affect key observables for cluster cosmology, such as the lensing signal and clustering statistics. In this work, we present a fully predictive forward analytical model to quantify the impact of optical-selection bias due to projection effects on cluster density profiles. This is achieved by introducing a scale-dependent parametrization of the optical cluster bias, whose small- and large-scale behaviour is set by the amplitude of projection effects, and by expressing the two-halo component of the density profile in terms of the contributions from off-axis halos along the line of sight. As a case study, we consider a DES Y3-like cluster catalog and validate our model against simulated samples. Our model successfully captures the dependence of the two-halo component on richness boosts induced by projections, as well as its evolution with richness and redshift. It also recovers the overall bias in the projected density profile relative to a randomly selected sample with the same mass distribution. The framework presented here provides a consistent methodology for modeling the impact of line-of-sight structures on the observed richness and density profiles of optically selected clusters, directly linking selection biases to the underlying cosmology and survey specifications.

The standard $Λ$CDM cosmological model faces increasing tensions between key observations, motivating tests that probe its underlying assumptions. In a companion letter, we present a model-independent framework that combines derivatives of the angular diameter distance, $d_A(z)$, and the line-of-sight expansion rate, $\mathcal{H}(z)$, to clarify the physical content of FLRW consistency relations and to construct a general-spacetime estimator of the cosmic density field. Here, we apply these tests to data, introducing a non-parametric reconstruction method based on symbolic regression combined with bootstrapping to provide data-driven uncertainty estimates. Using supernova and BAO data, we reconstruct $d_A$, $\mathcal{H}$, and their derivatives, enabling model-independent evaluation of FLRW relations and recovery of the sky-averaged density field over $z \in [0.38, \sim 2]$. Current data are too sparse to tightly constrain $\mathcal{H}(z)$, and the reconstructed density is consistent with both Planck and SH0ES $Λ$CDM. Reconstructed FLRW consistency tests show mild to moderate deviations from FLRW expectations at the $\sim 2$-$4σ$ level, although their significance depends on data selection and reconstruction stability. If these indicated deviations from an FLRW geometry are real, it would signify that most of the cosmological solutions considered for solving the cosmological tensions (evolving/interacting dark energy, new types of matter/energy, modified gravity, etc., within the FLRW framework) are ruled out. These preliminary indications highlight the importance of future, denser distance and expansion rate measurements, as well as further work toward standardizing uncertainty estimation for symbolic regression reconstructions.

Axion-like particles (ALPs) produced via the Primakoff process in the cores of Galactic core-collapse supernovae (SNe) could convert into MeV-energy gamma-rays through interactions with the Milky Way's magnetic field. To evaluate the detection prospects for such signals, we perform sensitivity projections for next-generation MeV telescopes by combining hypothetical instrument responses with realistic background estimates. Our analysis incorporates detailed simulations of the expected ALP flux from nearby SNe, the energy-dependent conversion probability in Galactic magnetic fields, and the telescope's angular/energy resolution based on advanced detector designs. Background components are modeled using data from current MeV missions and extrapolated to future sensitivity regimes. Our simulations demonstrate that next-generation telescopes with improved effective areas and energy resolution could achieve sensitivity to photon-ALP couplings as low as gagamma approx 1.61 x 10^-13 GeV^-1 for ALP masses ma < 10^-9 eV in Galactic Center. These results indicate that future MeV missions will probe unexplored regions of ALP parameter space, with conservative estimates suggesting they could constrain gagamma values two orders of magnitude below current astrophysical limits. Such observations would provide the most stringent tests to date for axion-like particles as a dark matter candidate in the ultra-light mass regime.

Mergers of neutron stars are believed to be one of the primary sites for the synthesis of the universe's heavy elements via the rapid neutron capture process. AT2017gfo, the kilonova following GW170817 provided the first direct spectroscopic evidence of the $r$-process happening in the universe. A prominent line feature near $1\,μ$m in its spectrum was attributed to strontium -- a claim that has been independently recovered by several teams. However, in recent years it has been debated whether the feature arises instead from helium. Here, we present non--local thermodynamic equilibrium (NLTE) radiative transfer modelling of the observed kilonova spectra, including detailed radiation-matter interaction physics for both strontium and helium. We make use of freshly calculated strontium atomic data for e$^-$ impact collisions, photoionization, and recombination processes. Our strontium model self-consistently reproduces the temporal evolution of the $1\,μ$m feature at early times, with its absence at $0.92\,$days to its clear emergence at $1.17\,$days. This transition mimics LTE, because at early epochs ($t\lesssim 1.5\,$days) the radiation field dominates the ionization state of the ejecta over thermal and non-thermal electron collisions. We further test if helium can form the feature under the same plasma conditions. The helium mass required at $1.17\,$days is comparable to the total ejecta mass, while a few percent by mass of helium suffices at 4.4 days. On the other hand, the strength of the strontium lines decrease with time, and may require a radially stratified abundance to consistently produce the feature. We conclude that strontium is required to explain the onset of the feature at early times, but helium can contribute to, or even dominate the feature at later epochs.

Calculations of the lowest valence π* as well as the 3s and higher energy 3pσ Rydberg excited states of the CO2 molecule are carried out using density functionals with variational optimization of the orbitals, an approach involving relatively little computational effort. Five functionals with varying degree of exchange are used in combination with real or complex-valued orbitals that are optimized by finding saddle points on the electronic energy surface corresponding to the excited states. When the PBE functional is used in combination with complex orbitals, the calculated excitation energy is found to be within 0.3 eV of multireference configuration interaction reference values, and the results are further improved with hybrid functionals. In contrast, linear-response time-dependent density functional theory calculations give errors up to 1.9 eV for the most diffuse 3pσ excitation and exhibit stronger dependence on both the excitation character and the functional used. Calculated C-O dissociation curves using the PBE functional and the orbital-optimized approach compare remarkably well with the reported multireference configuration interaction and equation-of-motion coupled-cluster singles and doubles calculations. Thanks to the low computational cost, these results demonstrate that orbital-optimized density functional calculations can be a promising route for modelling photorelaxation in condensed-phase CO2, for example in the context of interstellar cosmic-ray radiation driven process involving high-energy Rydberg states.

Gravitational waves from intermediate-mass black-hole (IMBH) binaries is a probe of strong-field gravity and black-hole evolution. Detection of IMBH is challenging because of their typically low frequency where the seismic noise, radiation pressure noise, and thermal noise dominate. The Cryogenic sub-Hz cROss torsion bar detector with quantum NOn-demolition Speed meter (CHRONOS) has been proposed to reach a strain sensitivity of $10^{-18} {\rm Hz}^{-1/2}$ at 2 Hz. It aims to detect GW from IMBH mergers with the mass of $\mathcal{O}(10^4)$ M$_{\odot}$ and to explore stochastic gravitational background of $Ω_{\rm GW} \sim 2\times 10^{-3}$ at 2 Hz. We present the overview of the CHRONOS hardware which is designed to integrate key techniques for improving low frequency sensitivity; torsion bar, speed meter, and cryogenic mirror. As a demonstration of the interferometer operation, we also report the commissioning status of a Michelson interferometer in National Central University in Taiwan which has been assembled as a partial component of CHRONOS.

The material ejected from a binary neutron star merger produces "kilonova," a radioactively powered emission at ultraviolet, optical, and infrared wavelengths. The early-phase spectra of the kilonova AT2017gfo, following the gravitational wave event GW170817, exhibit a strong absorption feature around $1\,\mathrm{μm}$. Helium (He) and strontium (Sr) have been proposed as the candidate elements contributing to this feature. However, due to the lack of consistent modeling including these two elements simultaneously, the exact contributions of each element to this feature remain unclear. In this study, we develop non-local thermodynamic equilibrium ionization models for He and Sr that take into account ionization by high-energy electrons, and estimate the abundances of each element required to reproduce the observed feature. Our modeling indicates that about $1\, \%$ of He or $1\mathrm{-}10\, \%$ of Sr in mass fraction are present in the ejecta moving at $v \sim 0.15 \, c$. This Sr mass fraction nicely agrees with the mass fraction in the solar $r$-process abundance. Based on comparison with nucleosynthesis calculations, our constraints suggest that $r$-process nucleosynthesis in GW170817 occurs at relatively low electron fraction ($Y_{\rm e} \lesssim 0.35$) and low entropy ($s \lesssim 30 \ k_B/\, \mathrm{nucleon}$) conditions. Interestingly, for $Y_{\rm e}$ $\lesssim 0.15$, the observed feature is reproduced by He with a mass fraction expected from $α$ decays of trans-Pb nuclei, which gives an indirect signature for the production of elements beyond the third $r$-process peak.

Dust in Early-Type galaxies (ETGs) may originate from internal or external sources. In this paper we study the stellar populations of particularly dusty ETGs to search for evidence of the dust's origin. Using the Southern African Large Telescope (SALT), we obtained long-slit optical spectra within the effective radius (R_e), along the major axis of 15 nearby ETGs, selected from the GAMA and Herschel-ATLAS surveys for their high levels of interstellar dust. Using full spectrum fitting and Lick index fitting we analysed their major axis kinematics and stellar population characteristics. We used stellar population models from the newly developed sMILES library and from the empirical MILES library. Kinematic results show that most of our sample of dusty ETGs are rotationally supported and there are no detectable kinematic discontinuities. 12 of our sample of 15 dusty ETGs show evidence of young/intermediate age stellar population components suggesting ongoing/recent star formation. Using simulations, we show that these recent ($\approx$1~Gyr) populations are not artefacts of the fitting process or data. As a check with a control sample we use stacked SDSS spectra and find that dusty ETGs show a component with intermediate age, whereas non-dusty ETGs do not. Age, metallicity and $α$-element abundance ratio increase with increasing central velocity dispersion in the SALT spectra, as seen in previous studies of ETGs, but with larger scatter in our sample. Given our stellar population findings, we discuss formation scenarios that might cause or rule out a high dust/molecular gas content.

Galactic outflows are predicted to be ubiquitous in low-mass galaxies, but observational evidence is lacking. Both a low signal-to-noise and a low spectral resolution can severely hamper the detection of galactic outflows, especially in small galaxies that have intrinsically narrow spectral lines. We obtained new, medium-high resolution (FWHM$_\mathrm{inst}\sim50-110$~km/s) optical spectra of 52 local star forming galaxies ($0.01 < z < 0.03$) with stellar masses $10^{8.5}<M_*/[M_{\odot}]<10^{10}$, using the TNG/DOLORES and NTT/EFOSC2 instruments. Our parent sample consists of SDSS galaxies with available heterodyne single-dish molecular (i.e., CO) line data. The targets of this study are selected among those that, based on the comparison between CO line widths, SDSS spectral resolution, and corresponding SDSS-based H$α$ line widths, have a high chance of being unresolved by SDSS spectroscopy. Our new, spectra reveal overall narrower H$α$ and [OIII]$\lambda5007$ lines, with signs of asymmetries and broad wings that are absent in the SDSS spectra of the same galaxies. This confirms that SDSS spectroscopy does not resolve the narrow emission lines of low-M$_*$ galaxies, which hinders the detection of outflows. We identify outflow signatures in $\sim30\%$ of our targets based on the H$α$ line spectra. Assuming a typical bi-conical outflow geometry, this detection rate is consistent with theoretical predictions of ubiquitous outflows in the low-mass regime. The outflow incidence is enhanced ($\sim60\%$) for galaxies with above average star formation rates for the sample (SFR $>10^{-0.74}~\mathrm{M_{\odot}/yr}$). We estimate ionized gas mass outflow rates ranging from $\sim0.1-50\times10^{-3}~\mathrm{M_{\odot}/yr}$ (mean $\sim20\times10^{-3}~\mathrm{M_{\odot}/yr}$) and corresponding mass loading factors between 0.03 and 0.14 (mean $\sim0.07$) for the sample.

The Kerr black hole spacetime is symmetric with respect to a well-defined equatorial plane. When such a symmetry is broken, for instance, by some putative effects beyond general relativity, the Keplerian circular orbits around the black hole are distorted vertically away from the equatorial plane by an amount depending on the orbital radius. As a result, the Keplerian thin disk acquires a curved surface. In this work, we extend such results to thick tori configurations by considering non-self-gravitating Polish doughnut models. We show that due to the equatorial asymmetry of the spacetime, the centers and the cusps of tori are distorted away from the original equatorial plane toward the same direction as that experienced by the stable Keplerian orbits, and the entire tori configurations are twisted toward that direction as well. The shape of the distorted tori is demonstrated explicitly using a constant specific angular momentum profile $\ell(r,y)=\ell_0$ of the disk fluid. However, the result also applies to non-constant profiles of $\ell(r,y)$ generically in the sense that any asymmetric profile of $\ell(r,y)$ that attempts to produce a symmetric tori configuration either turns out to be ill-defined near the equatorial plane or suffers from fine-tuning issues.

We investigate hierarchical mergers among subhalos within a $Λ$CDM simulation using the HBT+ subhalo finder. Unlike previous methods, HBT+ tracks subhalo evolution across hierarchy levels, identifying the coalescence of subhalo cores in phase-space as a ''sinking" event. This coalescence marks a distinct stalled phase in orbital decay, providing a physically motivated and natural definition of a resolved merger. Our main findings include: 1) Over 90% of sinking events occur between adjacent subhalo levels, while cross-level pathways arise from tidal stripping, group infall, and numerical constraints. 2) Resolved mergers are predominantly major mergers (mass ratios > 1:10), while the occurrence of minor mergers decreases with the dynamical age of the host halo. 3) Although deep-level subhalos have low mass ratios relative to the host halo, their high mass ratios relative to direct parents significantly boost merger statistics. Consequently, the satellite-satellite merger rate can rival or exceed the central-satellite rate at lower mass thresholds. 4) Satellite-satellite mergers are spatially biased toward the outer regions of the host, suggesting that the central tidal field suppresses their orbital decay. 5) A bidirectional sinking detection recovers 32% more sinking events than the original algorithm, revealing that child-dispersion-driven mergers are dominated by tidal heating at the final stage of sinking, while parent-dispersion-driven and doubly-identified events proceed primarily via orbital decay. Altogether, these results reveal a complex landscape of hierarchical satellite mergers that deviate from the self-similarity of host halo mergers, due to additional physical processes including dynamical friction and the scale-dependent halo growth history.

Extremely high-peaked BL Lac objects - also named extreme blazars - are among the most energetic and persistent extragalactic accelerators in the Universe, defined by a synchrotron emission peaking above $10^{17}$ Hz in X-rays. Such emission is then reprocessed and produces radiation extending deeply into very-high-energy (VHE, energy E>100 GeV) gamma rays. Observations in this energy band - optimally investigated by the Imaging Air-Shower Cherenkov telescopes - are crucial for probing the physical processes that drive their extreme behavior. This study extends our investigation of extreme blazars in the VHE gamma-ray range, providing a second new mini-catalog of sources observed by the MAGIC telescopes. We report on the monitoring of seven targets between 2017 and 2025, including four newly observed sources and three that have been part of long-term observation campaigns, for a total of approximately 338 hours of observations. The analysis of MAGIC data reveals two new VHE detections of extreme blazars, along with three additional sources showing hints of VHE emission. Joint observations of MAGIC and the first Large-Sized Telescope (LST-1) also confirmed a new VHE extreme blazar. Our results are complemented by simultaneous multiwavelength observations in other energy bands, including optical-UV, X-rays, and high-energy gamma rays (100 MeV<E<100 GeV). We confirm typical behavior of extreme blazars, such as a modest variability and a ``harder-when-brighter'' trend in X-rays across the sample. This new set increases the population of extreme blazars and their broadband analysis confirms the physical properties of these extreme sources.

Subhalo abundance matching (SHAM) is a commonly used framework for modeling the galaxy-halo connection. Yet, its standard implementation has difficulty reproducing the observed galaxy clustering with high accuracy (e.g., $χ^2/\mathrm{dof} \approx 1$). To overcome this issue, we propose a novel CS-SHAM framework, in which central and satellite galaxies are independently matched to main and satellite subhalos in simulations. Within this scheme, we introduce three free parameters to explicitly characterize the satellite fraction, $f_{\mathrm{sat}}$, as a function of stellar mass or absolute magnitude. To evaluate the performance of CS-SHAM, we apply it to two sets of mock galaxy catalogs built with the conventional SHAM method but using different subhalo mass proxies, $M_{\mathrm{peak}}$ and $V_{\mathrm{peak}}$, as well as two additional galaxy samples generated from a SAM and from TNG-300. We demonstrate that CS-SHAM reliably reproduces galaxy clustering whether $M_{\mathrm{peak}}$ or $V_{\mathrm{peak}}$ is used as the subhalo mass proxy. We also find that the models are unable to place robust constraints on $f_{\mathrm{sat}}$ if different mass proxies are employed. Indeed, within the CS-SHAM framework the halo occupation distribution (HOD) and conditional luminosity or stellar mass function (CLF/CSMF) are accurately recovered. Furthermore, we demonstrate for the first time that galaxy clustering constrains the HOD and CLF/CSMF primarily for relatively massive halos. Because the halo bias is nearly constant for low-mass halos, galaxy clustering is generally not very sensitive to the satellite population residing in these low-mass systems.

We consider a modular-invariant extension of the Standard Model. Assuming that the modulus is the inflaton, the CP-violating phases of the Yukawa couplings evolve during inflation. This dynamics favours a Higgs condensate, so that Standard Model fermions mediate a one-loop cosmological collider signal enhanced by chemical potentials. Next-generation experiments can probe sub-Planckian values of the modulus decay constant. We provide precise expressions for Dirac fermions with chemical potentials in de Sitter.

Neutral atomic hydrogen (HI) reservoirs typically extend far beyond the inner star-forming regions of galaxies, and global HI measurements, which mix these distinct environments, limit our understanding of the gas-star formation cycle. In particular, global HI depletion times combine gas and star formation from different physical scales, contributing to long measured timescales (5-9 Gyr) and large scatter compared to molecular gas. Using 841 gas-rich galaxies from the Widefield ASKAP L-band Legacy All-sky Blind Survey (WALLABY) pilot observations, we investigate how HI depletion time and its scaling relations change when HI and star formation are both confined to the stellar disc (R25, the isophotal radius at 25 mag arcsec-2 in i-band). We find that depletion times within this region are on average 1.4 Gyr shorter than global values, though some remain very long, indicating that a substantial fraction of HI remains inactive for star formation. HI depletion times anti-correlate strongly with stellar surface density, and this trend becomes even tighter within the stellar disc. The Kennicutt-Schmidt relation further reveals an almost constant HI depletion time at fixed stellar surface density, similar to the behaviour seen for molecular gas, suggesting that HI and star formation are regulated by conditions that enable HI-to-H2 conversion, traced by stellar surface density. Beyond the stellar disc, HI depletion times are on average almost 10 Gyr longer than within R25, confirming extremely inefficient star formation in low-density outer regions. These results highlight the critical role of spatial location and local conditions for HI to serve as a fuel for star formation.

Determining the astrophysical origin of binary black holes and whether merger remnants are retained in their birth environments is essential for understanding hierarchical mergers and the growth of intermediate-mass black holes. We identified the gravitational-wave (GW) events most consistent with dynamical formation and assessed whether their merger remnants are retained in globular clusters, nuclear star clusters, or galactic potentials. We considered the 84 events consistent with binary-black-hole (BBH) mergers from the first part of the fourth observing run (O4a) of the LIGO-Virgo-KAGRA (LVK) GW detector network, and 3 selected events from the second part (O4b). We compared parameter-estimation posteriors with synthetic population models for field and cluster binaries using Bayes factors, accounting for the relative abundances of these formation channels in the local Universe. We computed recoil-velocity posteriors for all events using the IMRPhenomXPNR waveform model, which incorporates multipole asymmetries. We identified five events showing preference for a dynamical origin, including the most massive O4a event GW231123_135430, while excluding the high-spinning O4b event GW241011_233834. Typical recoil velocities of analyzed events are of order a few hundred km/s, with extended high-velocity tails. These kicks suggest that merger remnants are likely ejected from typical globular clusters, while retention in nuclear star clusters remains possible but not guaranteed. Our results disfavour efficient hierarchical growth in globular clusters, whereas nuclear star clusters remain viable environments for repeated mergers. Although results depend on the adopted astrophysical population models, this analysis highlights the importance of improved and larger population models, as well as higher-quality detections enabled by future developments in GW detectors.

We report the discovery of an intervening 21\,cm absorption line at z = 0.882 towards the z = 1.284 quasar PKS 0405-385, identified in the First Large Absorption Survey in HI (FLASH). This quasar once displayed the most rapid known intraday variability at radio frequencies, from which it earned the title of `the smallest radio quasar'. Although its size was revised upwards soon after based on updated scattering theory, PKS 0405-385 remains an important probe of Galactic plasma, and now also of intervening gas discovered through HI absorption. We present new long-slit spectroscopy spanning both PKS 0405-385 and the candidate host of the intervening HI gas. We identify MgII and FeII absorption lines in this spectrum consistent with the redshift of the intervening HI, as well as two additional, independent metal-line systems at z = 0.907 and z = 0.966, but we cannot accurately pinpoint the host(s) of this intervening gas in current data. We revisit the radio variability of PKS 0405-385 in light of advances in scintillation theory, as well as extended monitoring with the Australia Telescope Compact Array and the Australian SKA Pathfinder, and find a revised linear size >0.3 pc, but no new evidence of repeating intraday variability.

We study the equation of state of neutron-star matter with color-spin molecular dynamics. The calculation includes the internal color and spin degrees of freedom and their time evolution. The matter composition, including strangeness under beta equilibrium, is determined by energy minimization. We find two main trends. First, within the present CSMD framework and under the adopted clustering criterion along the stable neutron-star branch, isolated quark-like configurations do not appear; instead, color-magnetic interactions favor the self-consistent formation of multi-quark clusters. Within the same criterion, the cluster-size distribution is concentrated at quark numbers that are multiples of three, corresponding to integer baryon numbers. Second, the interaction between strange and light quarks has a strong impact on neutron-star radii. This suggests that future radius measurements may help constrain flavor-sector interactions, including those involving strangeness.

Extragalactic jets exhibit a wide range of propagation orientations relative to the host galaxy's principal axis. This study investigate the spatiotemporal evolution of jets as a function of their propagation direction within their triaxial hosts-introducing varying degrees of environmental hindrance-and as a function of internal jet properties (while maintaining identical jet power)-introducing varying collimation and thrust. Observational data on extended radio sources are re-analyzed to identify key traits arising from variations in jet orientation and intrinsic properties. These findings are then systematically tested using a suite of 3D RMHD simulations. When a jet propagates along host's major axis (path of maximal environmental resistance), it produces an X-shaped morphology with secondary lobe aligns along the minor axis, co-evolving actively alongside the active jet. At intermediate angles to the major axis, the jet morphology transitions into a double-boomerang structure with notably curved lobes. Such lobes are interestingly regenerative through both backflow and jet precession mechanisms, making it difficult to disentangle their origin. Jets propagating along the minor axis (path of minimal resistance) exhibit faster propagation, forming classical double-lobed sources. With increased thrust and improved collimation (keeping jet power constant), these jets advance even more rapidly, potentially evolving into giant radio galaxy candidates. Counterexample sources that deviate from these traits were also modeled. The spatial variation of internal turbulence shows significant fluctuations below 1 kpc, with stronger magnetic fields further suppressing these irregularities. Magnetic field plays a key role in the radiative appearance of these sources, modulating features like missing or one-sided (wing) lobe emission, filamentary structures, and warmspot versus hotspot formation.

The Hubble parameters measured by the DESI DR2 BAO observations show a significant discrepancy from the prediction of the standard cosmological model. This discrepancy, together with the long-discussed Hubble tension, may originate from large-scale inhomogeneities in the matter distribution. This interpretation is motivated by infrared galaxy surveys, which suggest that our galaxy resides within the $\sim300$ Mpc under-dense region known as the KBC void. In this study, we apply a linear order relation -- relating the horizon-scale Hubble parameter inferred from CMB observations and the local-scale Hubble parameter -- to the Pantheon+ Type Ia supernovae and the DESI DR2 BAO data. We show that a simple inhomogeneous cosmological model consisting of eight top-hat shells can consistently explain the Hubble parameters inferred from both observations. Based on the matter-density distribution, we also briefly discuss its possible impact on cosmological observables, including the magnitude--redshift relation, the kinematic Sunyaev--Zel'dovich effect, and the integrated Sachs--Wolfe effect.

The average pulse shape of a pulsar is typically stable over decadal timescales, enabling estimation of pulse times of arrival to better than a small fraction of the pulse width using matched filtering techniques. However, in North American Nanohertz Observatory for Gravitational Waves (NANOGrav) observations of PSR J1713+0747, three discrete timing events that depart from the prevailing timing model have been seen in the last 20 yr. All three correspond to morphological changes in pulse shape. Using principal component analysis, we analyze the pulse profiles of nine NANOGrav pulsars, including seven with profiles from the 15 yr dataset and two with additional profiles from the forthcoming 20 yr dataset. We recover the three known pulse shape change events in PSR J1713+0747 and another previously known event in PSR J1643$-$1224. We implement a ranking metric for candidate events and address four highly ranked candidates in this nine-pulsar sample. We also recover known slow pulse shape variations in PSR J1643$-$1224, PSR J1903+0327, and PSR B1937+21 and report an unexpected recurrence after ~10 yr of one such variation in PSR B1937+21.

Rotation can significantly affect neutron-star (NS) properties, but accurate modeling of rapidly rotating NSs requires solving a two-dimensional, axially symmetric system, making traditional calculations too expensive for inference analyses that demand a large amount of model evaluations. We develop a causal convolutional neural networks that preserve the chronological-like dependence of NS properties on the equation of state (EoS) and rapidly reconstruct observables for static, Keplerian, and rotating configurations. Using \texttt{RNS}, we generate a dataset of NS observables and use it to train our networks. We validate our networks with three representative EoS (SFHo, SLy4, and DD2) and find that the they accurately reproduce the \texttt{RNS} results. The trained networks evaluate NS configurations for a single EoS in $\sim 50$ms, providing a substantial speedup over typical \texttt{RNS} runtimes of $\sim 30$ min and enabling efficient inference analyses involving rapidly rotating NSs.

Binary systems with ultrashort-period planetary-mass companions are expected to radiate continuous gravitational waves (GWs). However, earlier studies found that the detectability of such systems by the Laser Interferometer Space Antenna (LISA) is unlikely. In this study, we investigate the detectability of GWs from planetary-mass companions orbiting pulsars (PSRs) or white dwarfs (WDs) whose fundamental parameters, essential for calculating GW properties, have been measured. We compare the GW signals from our sample with the sensitivity curves of space-based GW detectors. We find that fourteen sources achieve a signal-to-noise ratio (\(\text{S/N}\)) of \(\gtrsim 5\) within four years of observations. Among these, three sources have PSR primaries (2S 0918-549 b, 4U 0513-40 b, and 4U 1543-62), and eleven systems possess WD primaries (BW Scl b, CP Eri b, CR Boo b, EF Eri b, GP Com b, GW Lib b, SDSS J0926+3624 b, SDSS J1507+5230 b, SMSS J1606-1000 b, SRGeJ0453 b, and WZ Sge b). We note that their detectability is less probable with near-term missions such as LISA, TianQin, and Taiji. Nevertheless, they could be detected by more advanced, future-generation observatories, such as the Deci-hertz Interferometer Gravitational wave Observatory (DECIGO) and the Big Bang Observer (BBO). This offers the potential to investigate the formation and evolution of ultrashort-period planetary-mass companions around compact stars through joint GW and electromagnetic surveys.

An accurate polarization calibration is essential for a spectral data analysis and Zeeman splitting measurements. Two anomalies challenge our understanding of OH megamasers in IRAS 02524+2046: an unexplained 1667/1665 MHz flux-ratio deviation, and complex Stokes V signatures. Well-calibrated sensitive polarization observations are required to understand them. We develop a polarization calibration solution for the L-band 19-beam receiver installed on the Five-hundred-meter aperture spherical radio telescope (FAST) to achieve a high calibration accuracy and thus enable accurate measurements of the OH megamaser properties in IRAS 02524+2046. We determined the Mueller matrix solution for spectral observations across the 1050-1450 MHz frequency range with an accuracy of about 0.01%-0.08% for circular polarization. We then applied it to FAST observational data of IRAS 02524+2046. Our results show narrower emission line components in the OH megamasers than previously reported, which are indistinguishable in the total power spectrum, but are detected in the circular polarization spectrum. The 1667 MHz OH megamaser emissions probably span a wide velocity range from ~54750 to ~53580 km/s, indicating greater complexity than previously recognized. Our fit of the total power and circular polarization spectra for IRAS 02524+2046 revealed ten line components with significant Zeeman splitting (>3sigma), indicating in situ magnetic fields with a strength of approximately -24.5 mG to +20.6 mG, most of which (8/10) have positive values.

We construct a catalog of mid-infrared (MIR) variable sources using the multi-epoch 3.6 (W1) and 4.5 $μ$m (W2) dataset from the Near-Earth Object Wide-field Infrared Survey Explorer (NEOWISE) at the north and south ecliptic poles (NEP and SEP). The catalog provides well-sampled light curves that cover areas within a radius of 5 degrees from the poles, which are frequently observed by current and forthcoming missions. By carefully processing the NEOWISE data to secure reliable photometric measurements, we identified 2764 and 27581 variables in the NEP and SEP, respectively, using the probability deviating from the non-variable and the correlation coefficient between W1 and W2. Cross-correlation with various complementary datasets reveals that, in the NEP, variability is dominated by active galactic nuclei, whereas stellar objects are more common in the SEP due to its proximity to the Large Magellanic Cloud. In particular, proper motion measurements from Gaia and MIR color-color diagrams are ideal for narrowing down the physical origin of the MIR variable sources. We identify three MIR transients in the NEP. Interestingly, all coincide with obscured QSOs, suggesting a physical connection between transient events and circumnuclear obscuration. Finally, we discuss the potential applications of our catalog in synergy with existing and future time-domain surveys.

Integrated-light star cluster catalogues in external galaxies are subject to complex, often poorly-characterised selection effects that can bias inferred cluster demographics and introduce significant uncertainties, limiting the physical parameter space accessible to analysis. To mitigate this problem, here we introduce the Cluster Completeness Correction Calculator (C-4): a new software tool to quantify and predict these effects in both physical and photometric parameter spaces. C-4 adds artificial star clusters to observed galaxy images, processes these images through the same detection and filtering steps used to construct the original cluster catalogue, and then trains multilayer perceptron neural networks to learn the resulting selection function. The trained neural networks provide continuous, differentiable completeness functions that can be used for direct completeness corrections or incorporated into forward models. We present a pilot application of C-4 to NGC~628, demonstrating that the learned selection operator is highly accurate and successfully captures the strongly non-separable dependence of completeness on mass, age, and extinction. Applying the completeness correction to NGC 628 extends the range of cluster demographic analyses by roughly an order of magnitude in both mass and age, and removes artificial flattening in the observed cluster mass and age distributions. These results establish neural-network-based completeness modelling as a powerful and general approach for recovering intrinsic cluster populations, and provide a scalable framework for modelling high-dimensional selection functions in resolved stellar population studies.

We present a framework for cosmological model selection using Neural Networks (NNs) trained directly on simulated Cosmic Microwave Background (CMB) temperature and polarisation maps. By operating at the map level rather than on compressed angular power spectra, our approach retains the full spatial information of temperature and polarisation anisotropies, enabling the identification of subtle signatures of primordial features beyond the standard $Λ$CDM model. We describe the generation of Planck-like CMB maps, and the hybrid architecture that combines principal component analysis and neural networks, optimised for classification tasks. To understand how the classifier reaches its decisions, we apply Shapley Additive exPlanations (SHAP) as a post-hoc interpretability tool, identifying which regions of the sky and which scales contribute most to the distinction between $Λ$CDM and feature models. This work serves as a follow-up to previous analyses at the level of summary statistics and as a proof-of-concept for using interpretable machine learning to uncover higher-order information in CMB data, with the potential to enhance the detection of nontrivial inflationary signals and improve cosmological model discrimination. Results for model classification performance, calibration, and interpretability are presented as a placeholder for the full analysis. In addition, we introduce the Open Science project, providing public access to the full pipeline for simulation, training, and interpretability of CMB map-based neural networks.

Cepheids are key distance indicators and benchmarks for stellar evolution, yet most of them are members of binary or multiple systems. While spectroscopic surveys and Gaia proper-motion anomalies reveal a high binary fraction, the population of resolved companions remains poorly characterised. We aim to search for and characterise visual companions to bright Galactic Cepheids using high-contrast imaging and to derive quantitative limits on undetected companions to constrain the architecture of Cepheid multiple systems. We observed 47 Cepheids with SPHERE using the ZIMPOL instrument in classical imaging mode and the V, R, and I filters. The data were obtained in pupil-stabilised mode and analysed using PCA-based imaging technique. For detected companions, we injected negative fake companions in a Monte Carlo approach to measure the relative astrometry. For non-detections, synthetic companions were injected to compute 5sigma contrast curves as a function of separation. We detected companions with a signal-to-noise ratio of > 5 for 8 Cepheids, corresponding to about 17% of the sample. Our SPHERE imaging confirms previously known visual companions with improved astrometry and reveals new wide components for AP Pup, T Vel, and TX Del) at projected separations of 0.16-0.9". For the remaining Cepheids, we derived typical maximum contrasts of 10, 11, and 12mag at 0.25", 0.5", and > 1, respectively. For a sub-set of targets, these limits ruled out main sequence companions more massive than late-K dwarfs beyond 0.5". Our SPHERE survey provides the first homogeneous set of high-contrast optical constraints on wide companions of Galactic Cepheids. The low detection rate of visual companions compared to the high overall binary fraction implies that most companions inferred from radial velocities and Gaia astrometry are either closer than 20mas or significantly fainter than the limits reached here.

We consider the generic injection of radiation (both dark and electromagnetic) during the epoch between big bang nucleosynthesis (BBN) and recombination. The contribution of the additional radiation to the number of effective neutrinos may be quite small in this scenario, since dark radiation and electromagnetic radiation provide contributions of opposite sign. However, the injection of electromagnetic radiation dilutes the baryon-to-entropy ratio, which is measured both at BBN and at recombination. As a result, this scenario is expected to be tightly constrained. Indeed, performing a numerical study, we find that the allowed amount of extra radiation may be no more than $\sim 25\%$ greater than in the case where it is assumed to be entirely dark radiation.

We examine several observable optical properties of a Skyrmion black hole (BH), focusing on the photon sphere, BH shadow, and photon trajectories. The Skyrme term, along with other geometric parameters of the spacetime, determines the photon sphere location and shapes the resulting BH shadow. Parameter variations produce observable departures from standard BH geometries, offering potential signatures of nonlinear field effects. We also analyze the sparsity of Hawking radiation and the associated energy emission spectra, showing how these quantities respond to the Skyrme coupling and background parameters. Our findings illuminate the connection between nonlinear field contributions and BH optics, with implications for observational and theoretical studies of modified gravity scenarios.

We explore the lowest mass limit that can be placed on the halo mass function in CDM using 28 strong gravitational lenses. For this purpose, we study an extreme model in which the halo mass function and mass-concentration relation follow CDM, with a sharp cutoff at some mass scale, $m_{\rm{low}}$. Lensing provides a unique window into this quantity as it does not depend on the presence of baryons in dark matter halos and also allows the detection of low mass halos at cosmological distances, both in the lens galaxies and along the line-of-sight. Our model incorporates the effects of tidal stripping of subhalos, leading to the presence of many subhalos below a given model cutoff scale. We place an upper limit on the low-mass cutoff of the halo mass function of $m_{\rm{low}}<10^{8.3}$ M$_\odot$ at 10:1 odds using a prior for the normalization of the subhalo mass function from the semi-analytic model {\tt galacticus} and $m_{\rm{low}}<10^{8.2}$ M$_\odot$ at 10:1 odds using a prior from $N$-body simulations. These limits are comparable to, or stronger than, existing constraints based on Milky Way satellite galaxies. Based on these results, we forecast more than an order of magnitude improvement with a sample of 200 quadruply imaged quasar lenses. This number represents a small subset of the thousands that are anticipated to be discovered by Rubin, Euclid, and Roman. Furthermore, with this larger sample of lenses we expect to directly constrain the normalization of the subhalo mass function, thereby eliminating a major source of uncertainty in our current measurements.

We present the first stellar elemental abundance study for two very low-mass stars, similar in mass to TRAPPIST-1, in the $\sim5-10$\,Myr-old Upper-Sco association. Their mid-infrared JWST/MIRI spectra, like those of many very low-mass stars, are hydrocarbon-rich, indicating C/O ratios greater than unity in the inner disk gas inside their snowlines. By fitting synthetic spectra to high-resolution APOGEE near-infrared stellar spectra, we show that, unlike their inner disks, both stars have solar C/O ratios. Their Fe, C, O, Mg, and Ca abundances are likewise consistent with solar values, placing them within the Galactic thin-disk population, as expected for nearby star-forming regions. This contrast between stellar and inner disk C/O ratios provides the first direct evidence that the inner disk's supersolar values are not inherited from the natal cloud but arise from disk processes. If these enhanced C/O ratios are primarily driven by inward drift of icy pebbles, there are major implications for disk evolution and planet formation, which we also discuss.

We constrain local primordial non-Gaussianity by a combined analysis of auto and cross-correlations of DESI DR1 tracers, leveraging LRGs and QSOs as well as ELGs between $0.8<z<3.1$. By cross-validating the signal across different clustering tracers within the same redshift range, we evaluate potential systematics in the $f^\mathrm{loc}_\mathrm{NL}$ measurements, capitalizing on the reduced susceptibility of cross-correlations to non-common systematics. We find that the cross-correlation between LRG and quasars can robustly improve the DESI DR1 $f^\mathrm{loc}_\mathrm{NL}$ constraints, by $\sim9\%$ to a measurement of $f^\mathrm{loc}_\mathrm{NL}=2.1_{-8.3}^{+8.8}$ at 68\% confidence. On the other hand, we do not find a clear improvement when including the DESI DR1 ELG sample. Mock tests predict an additional $\sim8\%$ gain with statistical scatter, and the lack of improvement in the data remains consistent with this expectation. This project serves as an exploratory analysis of DESI ELG clustering for $f^\mathrm{loc}_\mathrm{NL}$ through its cross-correlation in preparation for future DESI data analyses.

We derive and showcase a novel approach to approximating Fourier transforms in higher dimensions, focusing specifically on the case of 2D radially concentrated ('ring-like') functions. We first reduce the problem to that of evaluating the Hankel transforms of each angular mode of the image and then use our focal expansion to approximate these remaining Hankel transforms. Our method provides a single approximation that remains accurate from small to large spatial frequencies, bridging regimes where moment-based or large-frequency asymptotic expansions are individually reliable. We explore a series of examples, showing that the leading focal term provides an accurate global approximation for a broad range of functions. We demonstrate the utility of this approximation by examining the interferometric response for toy models of a black hole's "photon ring," highlighting the application to experiments designed to measure this feature such as the Black Hole Explorer.

SN2023ixf, a Type II supernova (SN) showing early signs of interaction with circumstellar material (CSM), has been observed with unprecedented detail across the electromagnetic spectrum since shock breakout. Here, we present nonlocal thermodynamic equilibrium time-dependent radiative-transfer calculations of its photospheric-phase evolution (i.e., ~20 to ~120d), and for the first time encompassing from the ultraviolet (UV) to the infrared (IR). The explosion of a 15Msun progenitor star, evolved with enhanced mass loss during the red-supergiant phase, yielding an ejecta of 7-8Msun, a kinetic energy of 1.2x10^51 erg, and a 56Ni mass of 0.05Msun, yields a satisfactory match to the photospheric-phase duration and brightness. Prolonged interaction with a decreasing CSM density is required to match a number of salient features of SN2023ixf during the photospheric phase, including the persistent UV continuum and line fluxes, the optical brightness and line profiles (in particular Halpha), as well as the IR flux (interaction boosts the free-free emission at long wavelengths). The presence of a cold dense shell (CDS), which is hard to infer at early times when the CDS and photosphere lie at similar velocities, becomes evident at later times and more so in the IR - we find no evidence for material faster than the CDS at ~8000km/s. Exploratory two-dimensional radiative-transfer calculations based on axially symmetric CSM or ejecta suggest that asymmetry can produce a diversity of profile shapes, with absorption troughs exhibiting a flat bottom or notches at any Doppler velocity. We emphasize the complexity of UV spectra influenced by complex metal-line blanketing at these phases. We document the sensitivity of model results to the adopted clumping in the CDS, though the largest offset is obtained here in the unlikely case of a smooth CDS.

Isolated star-forming galaxies require inflows of fresh gas from the surrounding medium to sustain episodes of star formation over time. However, there are very few direct detections of accretion onto external galaxies. Studies in absorption can only observe along limited sightlines, while those in emission can have difficulty distinguishing inflowing gas in the foreground of the galactic disk from similarly Doppler-shifted outflowing gas in the background. We explore the possibility of using the Balmer decrement (H$α$/H$β$) in low-inclination systems as a diagnostic for disentangling the flow geometry in disk-like galaxies. We leverage mock spatial-spectral observations of an isolated Milky Way-mass galaxy simulated using the radiation-hydrodynamics code AREPO-RT and post-processed with the Monte Carlo radiative transfer code COLT. We find that gas components located in front of the disk exhibit systematically lower Balmer decrements than gas embedded in or behind the disk, with a mean front-back offset of $Δ(\text{H}α/\text{H}β) \approx -0.14$. The ability to differentiate between the disk and far-side components is limited by the extremely clumpy, multiphase dust distribution along the line of sight introducing substantial scatter. Overall, the results provide a useful observational diagnostic of inflow and outflow in dusty face-on galaxies.

On 2025 August 18, the LIGO-Virgo-KAGRA collaboration reported S250818k, a sub-threshold gravitational-wave (GW) candidate consistent with a binary neutron star (NS) merger potentially involving a sub-solar-mass NS. Optical follow-up by the Zwicky Transient Facility identified AT2025ulz, a transient temporally coincident with the GW trigger that initially resembled a kilonova but was later classified as a young stripped-envelope Type IIb supernova (SN), dubbed SN 2025ulz. A key question is whether SN 2025ulz harbors fast, possibly collimated, non-thermal ejecta indicative of a central engine, as invoked in "superkilonova" scenarios linking sub-solar-mass NSs to accretion-disk fragmentation or core fission. We present early-to-late-time multi-band radio observations of SN 2025ulz obtained with the Karl G. Jansky Very Large Array as part of the JAGWAR program, complemented by observations with the upgraded Giant Metrewave Radio Telescope and MeerKAT. We detect a faint but significant radio counterpart to SN 2025ulz at 6-10 GHz. The data are consistent with non-thermal emission from SN ejecta interacting with circumstellar material, favoring a compact progenitor and relatively fast ejecta akin to those of Type cIIb SNe. Our data are also consistent with emission from an off-axis jet peaking at about 50-100 days after the GW trigger. Overall, our radio detection is compatible with a superkilonova scenario and would motivate future systematic multi-wavelength follow-up of core-collapse events coincident with sub-solar NS GW candidates, should the association between S250818k and SN 2025ulz be supported by offline GW analyses.

I show that early-time (pre-recombination) solutions to the Hubble tension are generically expected to increase the preferred baryon density $ω_b$. This puts these models in tension with Big Bang Nucleosynthesis (BBN), as measurements of primordial deuterium constrain $ω_b$ at percent level. I show that existing analyses are in tension with the BBN determination of $ω_b$, and that including a likelihood component for primordial deuterium deters two representative models from recovering a high $H_0$, and leads to worse fits to CMB, BAO, supernova, and BBN data than $Λ$CDM.

The dispersion measures (DMs) of fast radio bursts (FRBs) encode the total ionized-gas column densities along their sightlines. Most observed FRBs originate at distances where the cosmological principle applies. Thus, variations in the DM distribution of FRBs observed in different regions on the sky trace local sources of anisotropy, such as the warm ionized medium and circum-galactic medium of the Milky Way, and local large-scale structure. We present a map of extragalactic DM variations across the Northern sky using a few thousand FRBs from the second \chime{} catalog. We detect a $\gtrsim 4σ$ excess of $\sim$150 pc cm$^{-3}$ above the global mean, extended over $\sim$30$^\circ$ scales and centered near $α\approx$ $12^{\rm h}$, $δ\approx$ $55^\circ$. This excess, termed Wall 1, is robust to variations in sample selection and jackknife resampling, and cannot be explained by Galactic-disk DM-model uncertainties. The excess is likely too large to correspond to anisotropy in the Milky Way halo. The signal spatially coincides with the Ursa Major supercluster and associated large-scale structures. A secondary, more tentative Wall 2 near $α\approx 2^{\rm h}$, $δ\approx$ $45^\circ$ is spatially coincident with the Perseus-Pisces supercluster. Although the spatial coincidences suggest that the Walls may correspond to baryons in the local large-scale structure, the probability of chance coincidence is likely too high ($\sim10-20\%$) to claim confident associations. These results highlight the potential of using FRB DMs to detect baryon overdensities associated with local large-scale structure, and have important implications for near-field baryon mapping and FRB cosmology.

The classical Penrose process and the collisional Penrose processes involve particles decaying or interacting very close to a spinning black hole, in which some particles acquire negative energy and fall into the black hole while others acquire that energy and can leave the system. Both processes involve an extreme form of frame dragging, i.e. the spinning black hole drags spacetime with it, and the spacetime ejects some of the particles with a large energy gain, similar to a projectile in a slingshot. Such extreme forms of frame dragging had long been believed to be unobservable as the efficiency for a black hole energizing particles in this way is very low. Here we report a new observational signature of this extreme sort of frame dragging. In rapidly spinning black holes in X-ray binaries, processes similar to collisional Penrose processes, but slightly less extreme, can give rise to a new spectral component with distinct spectral and polarimetric properties. Observations of this new spectral component with current or future broadband X-ray polarimeters will open a new window into testing Einstein's theory of gravity close to the edge of a black hole and can be used to measure the black hole spin.

We compute the stochastic gravitational wave background sourced at second order by a spectator scalar field $χ$ coupled to the inflaton $φ$ through a portal interaction $σφ^2χ^2$ and with quartic self-interaction $λ_χχ^4/4!$. In the large portal coupling regime ($σ/λ\gg 1$, with $λ$ the inflaton normalization), parametric resonance during reheating amplifies the spectator power spectrum by many orders of magnitude near the resonance band until Hartree backreaction from the quartic coupling detunes the instability, while the large inflationary effective mass suppresses superhorizon power and ensures compatibility with CMB isocurvature bounds. We focus on the direct field-gradient source $\partial_aχ\,\partial_bχ$ in the second-order Einstein equations and derive a master formula that factorizes into a spectral integral over the frozen, vacuum-subtracted spectator spectrum and a time integral encoding the post-inflationary expansion history. For our benchmark reheating history we obtain analytic scaling relations, including a peak amplitude $Ω_{\rm GW}\propto T_{\rm reh}^{8/3}$, strong dependence on the portal strength, and weak sensitivity to $m_χ$. We validate the framework against nonlinear lattice simulations, demonstrating complementarity: the Hartree treatment captures superhorizon evolution inaccessible to the lattice, while the lattice resolves rescattering and fragmentation near the spectral peak. For $σ/λ\simeq 10^4$ and $T_{\rm reh}=2 \times 10^{14}\,\mathrm{GeV}$, the signal reaches $Ω_{\rm GW}h^2\sim 10^{-11}$ at $f\sim10^{7}$-$10^{8}\,\mathrm{Hz}$. Increasing $λ_χ$ at fixed $σ$ has a non-monotonic effect: small values enhance the signal via rescattering, whereas larger values suppress it by detuning the resonance.

The first multimessenger discovery of a binary neutron star (BNS) merger, GW170817, proved that such mergers can source short gamma-ray bursts (SGRBs) and produce \rprocess elements. The initial merger rate from this single event in the first two observing runs of the LIGO-Virgo observatory network, $110$--$3840\,\mathrm{Gpc}^{-3}\,\mathrm{yr}^{-1}$, was found to be broadly consistent with the SGRB rate, the Milky Way (MW) r-process mass, and the Galactic population of double neutron star (DNS) systems that will merge in a Hubble time. However, only one additional BNS merger has been detected since, and the BNS merger rate has been consistently revised downwards with the past few gravitational wave (GW) catalog updates. Analyzing GW data from the latest catalog GWTC-4, we find a total BNS merger rate of $28$--$300\,\mathrm{Gpc}^{-3}\,\mathrm{yr}^{-1}$ (consistent with the most recently published values from LIGO-Virgo-KAGRA) consisting of $53^{+176}_{-49}\,\mathrm{Gpc}^{-3}\mathrm{yr}^{-1}$ in GW170817-like $\sim(1.3,1.3)\,M_\odot$ BNSs (90\% credibility). In light of this updated GW rate, we revisit the consistency of the BNS merger rate with SGRBs, r-process and Galactic DNSs. In all cases, there is an emerging tension with the BNS (and EM-bright neutron star--black hole, NSBH) merger rate. Comparing to a BNS merger rate of $100\,\mathrm{Gpc}^{-3}\mathrm{yr}^{-1}$, the cosmological SGRB rate is a factor of 3.6--18 higher, the r-process rate is a factor of 0.9--4.1 higher, and the rate inferred from Galactic DNSs is a factor of 2.3--5.1 higher than the BNS rate. We discuss how various uncertainties in the inferred rates either alleviate or exacerbate this tension, which point to the various physical processes that can be constrained by such rate comparisons.

Diffuse accretion flows near a supermassive black hole are fundamentally weakly collisional. In such weakly collisional plasmas, the ion and electron distribution functions can deviate significantly from thermal equilibrium, and particle kinetic effects can influence large-scale fluid motion by driving pressure anisotropy, heat conduction, and plasma instabilities. Modeling these plasma effects in highly relativistic flows could be important for interpreting horizon-scale observations of black hole images. In this paper, we present a theoretical framework for understanding weakly collisional plasmas in general relativity by deriving the relativistic drift kinetic equations from the Vlasov-Maxwell equations. We present the evolution equations for the moments of the gyroaveraged distribution function and introduce a new analytic Landau fluid closure to capture anisotropic heat flow in relativistic plasmas. Unlike standard (collisional) general relativistic magnetohydrodynamics or extended magnetohydrodynamics, our model does not rely on strong collisions to enforce thermal equilibrium and consistently incorporates Landau damping in a fluid closure. The model introduced in this work provides a complementary approach to fully kinetic simulations in understanding weakly collisional effects in low-luminosity relativistic black hole accretion disks.

The Crab pulsar (PSR B0531+21) provides an unusually rich test bed for statistical studies of high-energy photon-counting data, owing to its extreme brightness and the contrasting behavior of its main pulse (MP) and interpulse (IP) components. Using 78.8 ks of Neutron star Interior Composition Explorer (NICER; Gendreau and Arzoumanian 2017) data-over two million individual X-ray pulses- we construct the single-pulse photon-count distributions of the MP and IP at keV energies. We find that the IP is well described by the Skellam distribution expected for the difference of two Poisson processes, providing a rare, high-statistics empirical demonstration of Skellam behavior in an astrophysical photon-counting context. The MP also shows pulse-by-pulse variability best described by a Skellam framework when compared to Gaussian alternatives, but exhibits a significant excess variance driven by high-count events. When photon counts are summed over successive pulses, this excess averages out and the MP distribution becomes consistent with Skellam expectations, indicating that the enhanced variability does not persist across rotations. We further search for short-lag (memory) correlations between successive X-ray pulses and find no statistically significant lag-1 correlation. Although giant radio pulses occur in the MP phase window, their contribution is insufficient to account for the observed excess variability. Together, these results highlight a clear statistical distinction between the MP and IP and underscore the importance of using statistically appropriate models for high-energy photon-counting analyses. The distributional fits and memory limits reported here provide quantitative constraints on pulsar emission models and illustrate the broader utility of Skellam-based approaches.

Gravitational wave (GW) and short Gamma Ray Burst (sGRB) observations provide us with complementary views of compact object mergers. The paucity of binary neutron star merger (BNS) detections in the latest LIGO/Virgo/KAGRA (LVK) observing run raises the question of whether the GW merger rates are sufficient to explain the observed sGRB rate with compact object mergers alone. We investigate this connection using the latest merger rate constraints from the fourth LVK observing run (O4) and published estimates of the local sGRB rate density. For an observed sGRB rate density of $ \sim 1-7~\mathrm{Gpc^{-3}\,yr^{-1}}$, if $>55\%$ of BNS mergers can successfully launch a jet, we find that the current LVK BNS merger rate can be reconciled with a sGRB merger population containing a significant fraction of relatively wide jets with core half-opening angles $θ_j \geq 10^\circ$. Meanwhile, a narrow jet population ($θ_j \sim 6^\circ$) can only be matched with the O4 neutron star merger rate estimates for an observed sGRB rate density of $\lesssim 1~\mathrm{Gpc^{-3}\,yr^{-1}}$, which is broadly consistent with several of the latest available estimates. We also find that neutron star-black hole mergers (NSBH) are expected to be a subdominant component of the sGRB population compared to BNS mergers, and they cannot help reconcile some of the highest available sGRB rate ($ >7~\mathrm{Gpc^{-3}\,yr^{-1}}$) with the GW rate estimates. However, they can still substantially contribute to the sGRB population, comprising $\sim 6-16\%$ of it for an observed sGRB rate density of $\sim 1-3~\mathrm{Gpc^{-3}\,yr^{-1}}$. Overall, our results indicate that present GW and sGRB observations remain broadly consistent with BNS mergers as the main progenitors of sGRBs.

We present a comprehensive analysis of the radial dark-matter (DM) density profiles of cosmic filaments in the hydrodynamical simulation TNG50. The cosmic web is extracted from high-resolution density grids at redshifts $z =$ 0, 0.5, 1, 2 and 3 using the DisPerSE algorithm. We show that the filament spine locations returned directly by DisPerSE do not accurately reflect the true density ridges. To address this issue, we introduce a "shrinking-cylinder" re-centering algorithm, which significantly increases the inferred central densities and restores the inner power-law behavior of the profiles. When the radial coordinate is scaled by the virial radii of the terminal nodes, the filament density profiles exhibit a nearly universal form that depends only weakly on redshift, node mass, and filament length. This result suggests that cosmic filaments, much like dark-matter halos, obey a form of structural self-similarity once an appropriate characteristic scale is introduced. By repeating the measurement using only smoothly distributed, unbound DM particles, we find that the apparent central cusp of the full profile is primarily produced by low-mass halos embedded along the filament spines, while the smooth component develops a flat core within $R/R_{\rm vir}\lesssim0.1$. The redshift evolution of this smooth component further suggests a transition from predominantly smooth filamentary accretion at high redshift to increasingly clumpy accretion at late times. Finally, we show that the universal filament profile is accurately described by a generalized triple-power-law model.

To measure precise distances beyond the Magellanic Clouds and determine an accurate value of the Hubble constant, eclipsing binary systems composed of early-type stars can play a crucial role. However, it is fundamental to first obtain a reliable empirical surface brightness-color relation (SBCR) for the hottest possible stars. Based on our previous study of six detached eclipsing binaries composed of O- and B-type stars in the Large Magellanic Cloud, we calibrated the SBCR using 12 stars with $V-K_{s} < -0.6$ mag. We found a significant difference between O-type and B-type stars in SBCRs, which are clearly separated in mass. The relation based on B-type stars is consistent with the relation for redder stars from the literature. This allowed us to provide a combined relation valid for stars less massive than $\sim 16\,M_\odot$ in the wide color range $-0.9 < V-K_{s} < 2.1$ mag, with $σ= 0.025$ mag. Such a relation can provide extragalactic distances precise to as high as $\sim$1.2% given the sufficient quality and number of target objects. The relation for O-type stars ($σ= 0.055$ mag) remains uncertain due to its strong dependence on the method used to determine reddening and requires further study. However, we tested it on the only known eclipsing system in M33, and obtained distance modulus DM=$24.90 \pm 0.17 $ mag, which perfectly agrees with the published distance to the system.

A majority of the active galactic nuclei (AGN) in the local Universe are classified as low-luminosity AGN (LLAGN), having bolometric luminosities $\lesssim 10^{42} \ \mathrm{erg \ s^{-1}}$. Although high-energy gamma-ray emission is predicted from both the jets and disks of LLAGN, to date only four have been detected by the Fermi Large Area Telescope (Fermi-LAT). In this work, we therefore conduct a comprehensive study of all the LLAGN from the Palomar spectroscopic survey of bright, northern galaxies, including both subthreshold and detected gamma-ray sources, using 14.4 years of LAT data. Our analysis results in a new detection of one LLAGN, as well as a detection of the subthreshold population using a stacking technique. We find that the signal from the subthreshold sample is consistent with being dominated by star-formation activity, although a contribution from compact jets or a mixed contribution from jetted and non-jetted systems is also feasible. On the other hand, the individually detected LLAGN are likely dominated by jet emission. We perform detailed spectral modeling for a subset of these sources and find that the gamma-ray signal can be explained by synchrotron self-Compton radiation, if the inner jet emission region is weakly magnetized with its total energy density being strongly particle dominated, and only slowly moving. With this work we also publicly release our Python-based stacking library for analyzing subthreshold source populations with the LAT, based on a proven technique used in numerous studies.

Massive quiescent galaxies have been identified out to $z\sim7$ in early JWST data in a substantial excess ($\rm \gtrsim 1\,dex$ at $z>4$) of number densities from most theoretical predictions. We investigate whether the number densities implied by the star formation histories of quiescent galaxies at $2<z<5$ are consistent with the observed number density evolution of that population since $z>7$. For this work, we rely on stellar population synthesis modeling of JWST NIRCam photometry (from CEERS and PRIMER) and NIRSpec/PRISM spectra of massive ($\rm M_{*} > 10^{10.5}M_{\odot}$) quiescent galaxies in the RUBIES survey. We infer their star-formation histories through Bayesian spectro-photometric fitting with Prospector, exploring the sensitivity of our results to stellar libraries and SFH priors. For each source, we compute a timescale over which it would be identified as quiescent -- leveraging the recent and most robust SFH timescale -- and deduce the number density of the quiescent population at previous epochs. These reconstructed number densities are then compared to existing observational constraints, including a new measurement from the PANORAMIC pure parallel survey, whose wide-area and independent sightlines reduce sensitivity to cosmic variance. We find striking agreement between reconstructed and observed number densities up to $z\sim7$, a self-consistency that lends credence to stellar population synthesis modeling of distant quiescent galaxies. Furthermore, by connecting the recent ($\rm \sim 1\,Gyr$) star-formation histories and number densities of quiescent galaxies and their implied progenitors, we reinforce the known tension between observations and model predictions at $3<z<7$.

Stellar collapse models predict that some stars more massive than $\sim$15$M_\odot$ may collapse directly to a black hole, sometimes with a weak optical transient, a phenomenon known as a failed supernova. Detecting such events is challenging, but searches of vanishing stars have found two promising candidates, N6946-BH1 and M31-2014-DS1. We re-analyze the JWST data of N6946-BH1 to characterize the remnant emission of the object and its surrounding sources. We found four near-infrared stellar neighbors not related to the mid-infrared emission of the candidate. The SED of N6946-BH1 is well modeled by a $\sim$10$^{4.7}L_\odot$ source obscured by a silicate dust shell with a maximum grain size of $\sim$3 $μ$m and producing negligible emission at $\lesssim$2 $μ$m. We model the progenitor and remnant emission of four Galactic and seven extragalactic stellar mergers to compare their properties with those of failed supernova candidates. We found that the merger remnants are 10-100 times more luminous than their progenitors at these late phases while the remnants of failed supernovae are $\sim$10 times dimmer than their progenitors. Asymmetric (disky) dust distributions cannot explain the factor of $\sim$100 difference in the ratios of the progenitor and remnant luminosities.

Accretion disks in active galactic nuclei (AGN) are promising sites for mergers of stellar-mass black holes (BHs) detectable via gravitational waves (GWs). These environments facilitate both in-situ formation and dynamical capture of compact objects, and their subsequent mergers. The uncertain origin of GW events detected by LIGO, Virgo and KAGRA motivates searching for accompanying electromagnetic (EM) signatures. Here we investigate post-merger EM flares associated with jets launched from merger remnants, as well as from the shocked ambient gas as the jet breaks out of the disk. We find that jet breakout produces luminous gamma-ray emission, detectable with MeV-band telescopes. Cooling emission from a shocked circum-BH minidisk, winds and background AGN-disk peaks in the UV and optical, with durations ranging from about an hour to a month, and can be identified through year-long monitoring of $\sim10^3$ AGNs with luminosities ranging from $\sim 10^{44}$ to $\sim 10^{45}~{\rm erg~s^{-1}}$. With a single set of parameters, this post-merger jet model produces gamma-ray, hard X-ray and optical flares similar to those claimed to be associated with GW events. Furthermore, by incorporating a transition from a high- to low-angular-momentum accretion state after the merger, the model avoids excessive BH growth, alleviating tensions with hyper-Eddington accretion scenarios.

We measure the number density and field-to-field variance of massive quiescent galaxies at $z\sim3$ - 8 using the JWST/NIRCam pure-parallel imaging survey PANORAMIC together with archival observations, covering an area of 0.28 deg$^2$ ($\sim1000$ arcmin$^2$) in at least six filters. We identify quiescent galaxy candidates at $z\gtrsim3$ with $M_\ast \gtrsim 10^{10}\,M_\odot$, comprising 101 galaxies in a gold sample of high-confidence candidates and 137 in a more inclusive silver sample. We measure their evolving comoving number density, finding $(1.5$ vs. $3.1)\times10^{-5}\,\mathrm{Mpc}^{-3}$ at $z=3$ - 4 for the gold and silver samples, respectively, and a decline by more than a factor of 20 by $z\sim6$. Comparisons with empirical models and cosmological simulations show that widely used frameworks underpredict the abundance of massive quiescent galaxies at $z\gtrsim4$ by $\gtrsim1$ dex, indicating that current implementations of early star formation, feedback, and quenching do not produce enough early quenched systems. With 34 independent sightlines, we present the first direct empirical measurement of field-to-field variance for quiescent galaxies at $z>3$, finding a high cosmic variance of $σ_{\rm CV}\approx0.7\pm0.3$. This exceeds predictions from abundance-matched mock catalogs, suggesting that early quiescent galaxies are more strongly clustered, and more likely to be found near one another or in more biased regions, than expected in current galaxy-formation models. Any successful model for the emergence of early massive quiescent galaxies must reproduce both their abundance evolution and their imprint on the large-scale distribution.

The study of collisionless shocks and their role in cosmic-ray acceleration has gained increasing importance through both observations and simulations. Accurately modeling the shock transition region, where particle injection occurs, requires a proper description of the microinstabilities governing its structure. In high-Mach-number shocks, such as those associated with supernova remnants, the ion Weibel instability is believed to provide the dominant dissipation mechanism. In this work, we investigate the ion Weibel instability driven by counterstreaming beams in the presence of an external perpendicular magnetic field. We employ hybrid simulations, in which ions are treated kinetically while electrons are modeled as a charge-neutralizing fluid. Although hybrid models are widely employed to study collisionless shocks, the resolution requirements needed to accurately capture ion-scale instabilities remain poorly understood. We address this issue by developing a linear theory of the ion Weibel instability tailored to the massless electron assumption of hybrid models and validating it with one- and two-dimensional simulations over a wide range of Alfvénic Mach numbers. We show that hybrid simulations can reliably reproduce the growth, saturation, and polarization of Weibel-generated magnetic fields in weakly magnetized regimes, provided that the relevant ion-scale modes are properly resolved. From the scaling of the dominant mode, we derive a minimum spatial resolution required as a function of Alfvénic Mach number. We also demonstrate that excessive resolution introduces unphysical small-scale whistler modes inherent to the massless-electron approximation. We validate the analysis by comparing the results with full particle-in-cell simulations. Together, these results provide practical guidance for hybrid simulations of collisionless shocks and beam-driven plasma systems.

Weakly interacting massive particles provide a well-motivated framework for dark matter, naturally reproducing the observed relic abundance through thermal freeze-out. A recent claim of an indirect-detection signal from the Galactic halo, consistent with dark matter annihilation in the mass range 400--800 GeV, motivates a reexamination of minimal models that can account for such a signal while remaining consistent with existing constraints. In this paper, we analyze the simplest extensions of the Standard Model capable of explaining the signal. We show that electroweak doublet dark matter with Higgs-portal interactions provides a natural and economical explanation. The model predicts annihilation predominantly into longitudinal gauge bosons with characteristic branching fractions and allows for inelastic dark matter with a mass splitting of order 100 keV, intriguingly consistent with a recent direct-detection anomaly. Possible enhancements of the present-day annihilation rate relative to the thermal value are also discussed, including a simple extension with a light scalar field, whose mass can be chosen such that the enhancement is suppressed in dwarf galaxies.

We propose a novel concept of astrophysical mirroring in the schwarzschild framework, which emerges as a direct consequence of gravitational lensing effects occurring in the immediate vicinity of extremely dense massive objects within spacetime. Through rigorous theoretical calculations and numerical ray-tracing analysis, we demonstrate that sufficiently compact astrophysical objects possess the capability to induce such extreme curvature in spacetime that the resulting gravitational field can bend light rays to extraordinary degrees, creating what we term a "reflection image" or mirror-like appearance of the source in distant regions of space. We discuss the theoretical framework as well as the observational consequences of this phenomenon.

Supercritical gas filaments in molecular clouds host the dense cores in which new stars form. The mechanisms governing their formation and subsequent gas accretion remain poorly understood. In this study, we conduct a statistical analysis of a large sample of sub-parsec supercritical filaments using H13COp J=1-0 data from the ALMA Three-millimeter Observations of Massive Star-forming regions (ATOMS) Survey. We identified velocity-coherent filaments in position-position-velocity (PPV) space and systematically examined velocity gradients both along and perpendicular to their skeletons. Our analysis uncovers a remarkable result: at scales of ~ 0.1-1 pc, the local velocity gradients within these supercritical filaments show no preferred alignment with the filament skeletons and exhibit no correlation with the local gravitational field. This random orientation suggests the presence of chaotic gas motions deep inside these dense structures. These findings may indicate that turbulence-rather than gravity-dominates gas dynamics and structural evolution at small scales, even in regions on the verge of star formation, challenging the paradigm of gravity-dominated structure formation within molecular clouds. This scenario should be further tested by more state-of-the-art simulations. This study offers key observational insights into the roles of turbulence and gravity in establishing the initial conditions for star formation.