arXiv Daily Digest - 2026-03-03

Layer-wise capacity in large language models is highly non-uniform: some layers contribute disproportionately to loss reduction while others are near-redundant. Existing methods for exploiting this non-uniformity, such as influence-function-based layer scoring, produce sensitivity estimates but offer no principled mechanism for translating them into allocation or pruning decisions under hardware constraints. We address this gap with a unified, curvature-aware framework grounded in the Minimum Description Length (MDL) principle. Our central quantity is the curvature-adjusted layer gain $ζ_k^2 = g_k^\top \widetilde{H}_{kk}^{-1} g_k$, which we show equals twice the maximal second-order reduction in empirical risk achievable by updating layer $k$ alone, and which strictly dominates gradient-norm-based scores by incorporating local curvature. Normalizing these gains into layer quality scores $q_k$, we formulate two convex MDL programs: a capacity allocation program that distributes expert slots or LoRA rank preferentially to high-curvature layers under diminishing returns, and a pruning program that concentrates sparsity on low-gain layers while protecting high-gain layers from degradation. Both programs admit unique closed-form solutions parameterized by a single dual variable, computable in $O(K \log 1/\varepsilon)$ via bisection. We prove an $O(δ^2)$ transfer regret bound showing that source-domain allocations remain near-optimal on target tasks when curvature scores drift by $δ$, with explicit constants tied to the condition number of the target program. Together, these results elevate layer-wise capacity optimization from an empirical heuristic to a theoretically grounded, computationally efficient framework with provable optimality and generalization guarantees.

Coarse-grained reconfigurable arrays (CGRAs) have attracted growing interest because they exhibit performance and energy efficiency competitive with ASICs while maintaining flexibility similar to FPGAs. These properties make CGRAs attractive in accelerator and other power-constrained system contexts. However, modern CGRA compilers aggressively pipeline for frequency and performance improvements, often violating hard power budgets. We empirically show that, in state-of-the-art CGRA compilers such as Cascade, post-place-and-route (post-PnR) pipelining increases power monotonically and ultimately exceeds fixed power caps across diverse workloads. In response, we introduce \emph{Capstone}, a power-aware extension of Cascade that integrates a fast, compiler-resident power model with a user-tunable controller that guides the bitstream selection process towards optimization targets. Capstone predicts per-iteration power directly inside the post-PnR compilation loop and selects one or a small set of PnR configurations such that at least one meets a user-specified power cap. Thus, we shift the objective from indiscriminately maximizing frequency to maximizing safe frequency under a discrete power cap. On a suite of kernels spanning fundamental dense and sparse applications, Capstone meets a power cap and minimizes remaining power headroom while preserving feasible performance. Our results indicate that cap-aware compilation is both necessary and practical, as the compiler can proactively land on cap-compliant points and expose predictable performance under power constraints.

The growing computational and memory demands of the Key-Value (KV) cache significantly limit the ability of Large Language Models (LLMs). While KV merging has emerged as a promising solution, existing methods that rely on empirical observations of KV asymmetry and gradient-based Hessian approximations lack a theoretical foundation and incur suboptimal compression and inference overhead. To bridge these gaps, we establish a theoretical framework that characterizes this asymmetry through the spectral energy distribution of projection weights, demonstrating that concentrated spectra in Query/Key weights induce feature homogeneity, whereas dispersed spectra in Value weights preserve heterogeneity. Then, we introduce KVSlimmer, an efficient algorithm that captures exact Hessian information through a mathematically exact formulation, and derives a closed-form solution utilizing only forward-pass variables, resulting in a gradient-free approach that is both memory- and time-efficient. Extensive experiments across various models and benchmarks demonstrate that KVSlimmer consistently outperforms SOTA methods. For instance, on Llama3.1-8B-Instruct, it improves the LongBench average score by 0.92 while reducing memory costs and latency by 29% and 28%, respectively.

Inspired by the human learning and memory system, particularly the interplay between the hippocampus and cerebral cortex, this study proposes a dual-learner framework comprising a fast learner and a meta learner to address continual Reinforcement Learning~(RL) problems. These two learners are coupled to perform distinct yet complementary roles: the fast learner focuses on knowledge transfer, while the meta learner ensures knowledge integration. In contrast to traditional multi-task RL approaches that share knowledge through average return maximization, our meta learner incrementally integrates new experiences by explicitly minimizing catastrophic forgetting, thereby supporting efficient cumulative knowledge transfer for the fast learner. To facilitate rapid adaptation in new environments, we introduce an adaptive meta warm-up mechanism that selectively harnesses past knowledge. We conduct experiments in various pixel-based and continuous control benchmarks, revealing the superior performance of continual learning for our proposed dual-learner approach relative to baseline methods. The code is released in https://github.com/datake/FAME.

Modern generative agents such as OpenClaw - an open-source, self-hosted personal assistant with a community skill ecosystem, are gaining attention and are used pervasively. However, the openness and rapid growth of these ecosystems often outpace systematic security evaluation. In this paper, we design, implement, and evaluate Clawdrain, a Trojanized skill that induces a multi-turn "Segmented Verification Protocol" via injected SKILL.md instructions and a companion script that returns PROGRESS/REPAIR/TERMINAL signals. We deploy Clawdrain in a production-like OpenClaw instance with real API billing and a production model (Gemini 2.5 Pro), and we measure 6-7x token amplification over a benign baseline, with a costly, failure configuration reaching approximately 9x. We observe a deployment-only phenomenon: the agent autonomously composes general-purpose tools (e.g., shell/Python) to route around brittle protocol steps, reducing amplification and altering attack dynamics. Finally, we identify production vectors enabled by OpenClaw's architecture, including SKILL.md prompt bloat, persistent tool-output pollution, cron/heartbeat frequency amplification, and behavioral instruction injection. Overall, we demonstrate that token-drain attacks remain feasible in real deployments, but their magnitude and observability are shaped by tool composition, recovery behavior, and interface design.

Large language models are increasingly used to produce runnable software. In practice, security is often addressed through a Detect--Repair--Verify (DRV) loop that detects issues, applies fixes, and verifies the result. This work studies such a workflow for project-level artifacts and addresses four gaps: L1, the lack of project-level benchmarks with executable function and security tests; L2, limited evidence on pipeline-level effectiveness beyond studying detection or repair alone; L3, unclear reliability of detection reports as repair guidance; and L4, uncertain repair trustworthiness and side effects under verification. A new benchmark dataset\footnote{https://github.com/Hahappyppy2024/EmpricalVDR} is introduced, consisting of runnable web-application projects paired with functional tests and targeted security tests, and supporting three prompt granularities at the project, requirement, and function level. The evaluation compares generation-only, single-pass DRV, and bounded iterative DRV variants under comparable budget constraints. Outcomes are measured by secure and correct yield using test-grounded verification, and intermediate artifacts are analyzed to assess report actionability and post-repair failure modes such as regressions, semantic drift, and newly introduced security issues.

Grading in large undergraduate STEM courses often yields minimal feedback due to heavy instructional workloads. We present a large-scale empirical study of AI grading on real, handwritten single-variable calculus work from UC Irvine. Using OCR-conditioned large language models with structured, rubric-guided prompting, our system produces scores and formative feedback for thousands of free-response quiz submissions from nearly 800 students. In a setting with no single ground-truth label, we evaluate performance against official teaching-assistant grades, student surveys, and independent human review, finding strong alignment with TA scoring and a large majority of AI-generated feedback rated as correct or acceptable across quizzes. Beyond calculus, this setting highlights core challenges in OCR-conditioned mathematical reasoning and partial-credit assessment. We analyze key failure modes, propose practical rubric- and prompt-design principles, and introduce a multi-perspective evaluation protocol for reliable, real-course deployment. Building on the dataset and evaluation framework developed here, we outline a standardized benchmark for AI grading of handwritten mathematics to support reproducible comparison and future research.

Smart contracts underpin high-value ecosystems such as decentralized finance (DeFi), yet recurring vulnerabilities continue to cause losses worth billions of dollars. Although numerous security analyzers that detect such flaws exist, real-world attacks remain frequent, raising the question of whether these tools are truly effective or simply under-used due to low developer trust. Prior benchmarks have evaluated analyzers on synthetic or vulnerable-only contract datasets, limiting their ability to measure false positives, false negatives, and usability factors that drive adoption. To close this gap, we present a mixed-methods study that combines large-scale benchmarking with practitioner insights. We evaluate six widely used analyzers (i.e., Confuzzius, Dlva, Mythril, Osiris, Oyente, and Slither) on 653 real-world smart contracts that cover three high-impact vulnerability classes from the OWASP Smart Contract Top Ten (i.e., reentrancy, suicidal contract termination, and integer arithmetic errors). Our results show substantial variation in accuracy (F1 = 31.2 to 94.6%), high false-positive rates (up to 32.6%), and runtimes exceeding 700 seconds per contract. We then survey 150 professional developers and auditors to understand how they use and perceive these tools. Our findings reveal that excessive false positives, vague explanations, and long analysis times are the main barriers to trust and adoption in practice. By linking measurable performance gaps to developer perceptions, we provide concrete recommendations for improving the precision, explainability, and usability of smart-contract security analyzers.

Large Language Models (LLMs) have recently exhibited remarkable reasoning capabilities, largely enabled by supervised fine-tuning (SFT)- and reinforcement learning (RL)-based post-training on high-quality reasoning data. However, reproducing and extending these capabilities in open and scalable settings is hindered by three fundamental data-centric challenges: (1) the cold-start problem, arising from the lack of seed datasets with detailed, long Chain-of-Thought (CoT) trajectories needed to initialize reasoning policies; (2) limited domain coverage, as most existing open-source reasoning datasets are concentrated in mathematics, with limited coverage of broader scientific disciplines; and (3) the annotation bottleneck, where the difficulty of frontier-level reasoning tasks makes reliable human annotation prohibitively expensive or infeasible. To address these challenges, we introduce CHIMERA, a compact synthetic reasoning dataset comprising 9K samples for generalizable cross-domain reasoning. CHIMERA is constructed with three key properties: (1) it provides rich, long CoT reasoning trajectories synthesized by state-of-the-art reasoning models; (2) it has broad and structured coverage, spanning 8 major scientific disciplines and over 1K fine-grained topics organized via a model-generated hierarchical taxonomy; and (3) it employs a fully automated, scalable evaluation pipeline that uses strong reasoning models to cross-validate both problem validity and answer correctness. We use CHIMERA to post-train a 4B Qwen3 model. Despite the dataset's modest size, the resulting model achieves strong performance on a suite of challenging reasoning benchmarks, including GPQA-Diamond, AIME 24/25/26, HMMT 25, and Humanity's Last Exam, approaching or matching the reasoning performance of substantially larger models such as DeepSeek-R1 and Qwen3-235B.

Despite exceptional predictive performance of Deep sequence models (DSMs), the main concern of their deployment centers around the lack of uncertainty awareness. In contrast, probabilistic models quantify the uncertainty associated with unobserved variables with rules of probability. Notably, Bayesian methods leverage Bayes' rule to express our belief of unobserved variables in a principled way. Since exact Bayesian inference is computationally infeasible at scale, approximate inference is required in practice. Two major bottlenecks of Bayesian methods, especially when applied in deep neural networks, are prior specification and approximation quality. In Chapter 3 & 4, we investigate how the architectures of DSMs themselves can be informative for the design of priors or approximations in probabilistic models. We first develop an approximate Bayesian inference method tailored to the Transformer based on the similarity between attention and sparse Gaussian process. Next, we exploit the long-range memory preservation capability of HiPPOs (High-order Polynomial Projection Operators) to construct an interdomain inducing point for Gaussian process, which successfully memorizes the history in online learning. In addition to the progress of DSMs in predictive tasks, sequential generative models consisting of a sequence of latent variables are popularized in the domain of deep generative models. Inspired by the explicit self-supervised signals for these latent variables in diffusion models, in Chapter 5, we explore the possibility of improving other generative models with self-supervision for their sequential latent states, and investigate desired probabilistic structures over them. Overall, this thesis leverages inductive biases in DSMs to design probabilistic inference or structure, which bridges the gap between DSMs and probabilistic models, leading to mutually reinforced improvement.

LLMs increasingly excel on AI benchmarks, but doing so does not guarantee validity for downstream tasks. This study evaluates the performance of leading foundation models (FMs, i.e., generative pre-trained base LLMs) with out-of-distribution (OOD) tasks of the teaching and learning of schoolchildren. Across all FMs, inter-model behaviors on disparate tasks correlate higher than they do with expert human behaviors on target tasks. These biases shared across LLMs are poorly aligned with downstream measures of teaching quality and often \textit{negatively aligned with learning outcomes}. Further, we find multi-model ensembles, both unanimous model voting and expert-weighting by benchmark performance, further exacerbate misalignment with learning. We measure that 50\% of the variation in misalignment error is shared across foundation models, suggesting that common pretraining accounts for much of the misalignment in these tasks. We demonstrate methods for robustly measuring alignment of complex tasks and provide unique insights into both educational applications of foundation models and to understanding limitations of models.

Discrete diffusion and flow matching models capture complex, non-additive and non-autoregressive structure in high-dimensional objective landscapes through parallel, iterative refinement. However, their implicit generative nature precludes direct integration with principled variational frameworks for online black-box optimisation, such as variational search distributions (VSD) and conditioning by adaptive sampling (CbAS). We introduce Active Flow Matching (AFM), which reformulates variational objectives to operate on conditional endpoint distributions along the flow, enabling gradient-based steering of flow models toward high-fitness regions while preserving the rigour of VSD and CbAS. We derive forward and reverse Kullback-Leibler (KL) variants using self-normalised importance sampling. Across a suite of online protein and small molecule design tasks, forward-KL AFM consistently performs competitively compared to state-of-the-art baselines, demonstrating effective exploration-exploitation under tight experimental budgets.

Large language models (LLMs) have demonstrated significant reasoning capabilities in scientific discovery but struggle to bridge the gap to physical execution in wet-labs. In these irreversible environments, probabilistic hallucinations are not merely incorrect, but also cause equipment damage or experimental failure. To address this, we propose \textbf{BioProAgent}, a neuro-symbolic framework that anchors probabilistic planning in a deterministic Finite State Machine (FSM). We introduce a State-Augmented Planning mechanism that enforces a rigorous \textit{Design-Verify-Rectify} workflow, ensuring hardware compliance before execution. Furthermore, we address the context bottleneck inherent in complex device schemas by \textit{Semantic Symbol Grounding}, reducing token consumption by $\sim$6$\times$ through symbolic abstraction. In the extended BioProBench benchmark, BioProAgent achieves 95.6\% physical compliance (compared to 21.0\% for ReAct), demonstrating that neuro-symbolic constraints are essential for reliable autonomy in irreversible physical environments. \footnote{Code at https://github.com/YuyangSunshine/bioproagent and project at https://yuyangsunshine.github.io/BioPro-Project/}

With the increasing demand for step-wise, cross-modal, and knowledge-grounded reasoning, multimodal large language models (MLLMs) are evolving beyond the traditional fixed retrieve-then-generate paradigm toward more sophisticated agentic multimodal retrieval-augmented generation (MM-RAG). Existing benchmarks, however, mainly focus on simplified QA with short retrieval chains, leaving adaptive planning and multimodal reasoning underexplored. We present MC-Search, the first benchmark for agentic MM-RAG with long, step-wise annotated reasoning chains spanning five representative reasoning structures. Each example specifies sub-questions, retrieval modalities, supporting facts, and intermediate answers, with fidelity ensured by HAVE (Hop-wise Attribution and Verification of Evidence), resulting in 3,333 high-quality examples averaging 3.7 hops. Beyond answer accuracy, MC-Search introduces new process-level metrics for reasoning quality, stepwise retrieval and planning accuracy. By developing a unified agentic MM-RAG pipeline, we benchmark six leading MLLMs and reveal systematic issues such as over- and under-retrieval and modality-misaligned planning. Finally, we introduce Search-Align, a process-supervised fine-tuning framework leveraging verified reasoning chains, showing that our data not only enables faithful evaluation but also improves planning and retrieval fidelity in open-source MLLMs.

Existing point cloud completion methods struggle to balance high-quality reconstruction with computational efficiency. To address this, we propose PPC-MT, a novel parallel framework for point cloud completion leveraging a hybrid Mamba-Transformer architecture. Our approach introduces an innovative parallel completion strategy guided by Principal Component Analysis (PCA), which imposes a geometrically meaningful structure on unordered point clouds, transforming them into ordered sets and decomposing them into multiple subsets. These subsets are reconstructed in parallel using a multi-head reconstructor. This structured parallel synthesis paradigm significantly enhances the uniformity of point distribution and detail fidelity, while preserving computational efficiency. By integrating Mamba's linear complexity for efficient feature extraction during encoding with the Transformer's capability to model fine-grained multi-sequence relationships during decoding, PPC-MT effectively balances efficiency and reconstruction accuracy. Extensive quantitative and qualitative experiments on benchmark datasets, including PCN, ShapeNet-55/34, and KITTI, demonstrate that PPC-MT outperforms state-of-the-art methods across multiple metrics, validating the efficacy of our proposed framework.

Machine learning based network intrusion detection systems are vulnerable to adversarial attacks that degrade classification performance under both gradient-based and distribution shift threat models. Existing defenses typically apply uniform detection strategies, which may not account for heterogeneous attack characteristics. This paper proposes an attack-aware multi-stage defense framework that learns attack-specific detection strategies through a weighted combination of ensemble disagreement, predictive uncertainty, and distributional anomaly signals. Empirical analysis across seven adversarial attack types reveals distinct detection signatures, enabling a two-stage adaptive detection mechanism. Experimental evaluation on a benchmark intrusion detection dataset indicates that the proposed system attains 94.2% area under the receiver operating characteristic curve and improves classification accuracy by 4.5 percentage points and F1-score by 9.0 points over adversarially trained ensembles. Under adaptive white-box attacks with full architectural knowledge, the system appears to maintain 94.4% accuracy with a 4.2% attack success rate, though this evaluation is limited to two adaptive variants and does not constitute a formal robustness guarantee. Cross-dataset validation further suggests that defense effectiveness depends on baseline classifier competence and may vary with feature dimensionality. These results suggest that attack-specific optimization combined with multi-signal integration can provide a practical approach to improving adversarial robustness in machine learning-based intrusion detection systems.

With recent development of artificial intelligence, it is more common to adopt AI agents in economic activities. This paper explores the economic actions of agents, including human agents and AI agents, in an economic game of trading products/services, and the equilibria in this economy involving multiple agents. We derive a range of equilibrium results and their corresponding conditions using a Markov chain stationary distribution based model. One distinct feature of our model is that we consider the long-term utility generated by economic activities instead of their short-term benefits. For the model consisting of two agents, we fully characterize all the possible economic equilibria and conditions. Interestingly, we show that unless each agent can at least double (not merely increase) its marginal utility by purchasing the other agent's products/services, purchasing the other agent's products/services will not happen in any economic equilibrium. We further extend our results to three and more agents, where we characterize more economic equilibria. We find that in some equilibria, the ``more powerful'' AI agents contribute zero utility to ``less capable'' agents.

Predicting physicochemical properties across chemical space is vital for chemical engineering, drug discovery, and materials science. Current molecular foundation models lack thermodynamic consistency, while domain-informed approaches are limited to single properties and small datasets. We introduce MultiPUFFIN, a domain-constrained multimodal foundation model addressing both limitations simultaneously. MultiPUFFIN features: (i) an encoder fusing SMILES, graphs, and 3D geometries via gated cross-modal attention, alongside experimental condition and descriptor encoders; (ii) prediction heads embedding established correlations (e.g., Wagner, Andrade, van't Hoff, and Shomate equations) as inductive biases to ensure thermodynamic consistency; and (iii) a two-stage multi-task training strategy.Extending prior frameworks, MultiPUFFIN predicts nine thermophysical properties simultaneously. It is trained on a multi-source dataset of 37,968 unique molecules (40,904 rows). With roughly 35 million parameters, MultiPUFFIN achieves a mean $R^2 = 0.716$ on a challenging scaffold-split test set of 8,877 molecules. Compared to ChemBERTa-2 (pre-trained on 77 million molecules), MultiPUFFIN outperforms the fine-tuned baseline across all nine properties despite using 2000x fewer training molecules. Advantages are strikingly apparent for temperature-dependent properties, where ChemBERTa-2 lacks the architectural capacity to incorporate thermodynamic conditions.These results demonstrate that multimodal encoding and domain-informed biases substantially reduce data and compute requirements compared to brute-force pre-training. Furthermore, MultiPUFFIN handles missing modalities and recovers meaningful thermodynamic parameters without explicit supervision. Systematic ablation studies confirm the property-specific benefits of these domain-informed prediction heads.

Systems based on Large Language Models (LLMs) have become formidable tools for automating research and software production. However, their governance remains a challenge when technical requirements demand absolute consistency, auditability, and predictable control over cost and latency. Recent literature highlights two phenomena that aggravate this scenario: the stochastic variance inherent in the model's judgment (often treated as "systemic noise") and the substantial degradation of context utilization in long inputs, with critical losses when decisive information is diluted in the middle of the prompt. This article proposes PARCER as an engineering response to these limitations. The framework acts as a declarative "operational contract" in YAML, transforming unstructured interactions into versioned and executable artifacts. PARCER imposes strict governance structured into seven operational phases, introducing decision hygiene practices inspired by legal judgments to mitigate noise, adaptive token budgeting, formalized recovery routes (fallbacks) for context preservation, and systemic observability via OpenTelemetry. The objective of this work is to present the conceptual and technical architecture of PARCER, positioning it as a necessary transition from simple "prompt engineering" to "context engineering with governable governance".

Counterfactual learning has become promising for understanding and modeling causality in complex and dynamic systems. This paper presents a novel method for counterfactual learning in the context of multivariate time series analysis and forecast. The primary objective is to uncover hidden causal relationships and identify potential interventions to achieve desired outcomes. The proposed methodology integrates genetic algorithms and rigorous causality tests to infer and validate counterfactual dependencies within temporal sequences. More specifically, we employ Granger causality to enhance the reliability of identified causal relationships, rigorously assessing their statistical significance. Then, genetic algorithms, in conjunction with quantile regression, are used to exploit these intricate causal relationships to project future scenarios. The synergy between genetic algorithms and causality tests ensures a thorough exploration of the temporal dynamics present in the data, revealing hidden dependencies and enabling the projection of outcomes under hypothetical interventions. We evaluate the performance of our algorithm on real-world data, showcasing its ability to handle complex causal relationships, revealing meaningful counterfactual insights, and allowing for the prediction of outcomes under hypothetical interventions.

Recommendation Systems are effective in managing the ever-increasing amount of multimodal data available today and help users discover interesting new items. These systems can handle various media types such as images, text, audio, and video data, and this has made it possible to handle content-based recommendation utilizing features extracted from items while also incorporating user preferences. Graph Neural Network (GNN)-based recommendation systems are a special class of recommendation systems that can handle relationships between items and users, making them particularly attractive for content-based recommendations. Their popularity also stems from the fact that they use advanced machine learning techniques, such as deep learning on graph-structured data, to exploit user-to-item interactions. The nodes in the graph can access higher-order neighbor information along with state-of-the-art vision-language models for processing multimodal content, and there are well-designed algorithms for embedding, message passing, and propagation. In this work, we present the design of a GNN-based recommendation system on a novel data set collected from field research. Designed for an endangered performing art form, the recommendation system uses multimodal content (text and image data) to suggest similar paintings for viewing and purchase. To the best of our knowledge, there is no recommendation system designed for narrative scroll paintings -- our work therefore serves several purposes, including art conservation, a data storage system for endangered art objects, and a state-of-the-art recommendation system that leverages both the novel characteristics of the data and preferences of the user population interested in narrative scroll paintings.

Retrieval-Augmented Generation (RAG) grounds Large Language Models (LLMs) to mitigate factual hallucinations. Recent paradigms shift from static pipelines to Modular and Agentic RAG frameworks, granting models autonomy for multi-hop reasoning or self-correction. However, current reflective RAG heavily relies on massive LLMs as universal evaluators. In high-throughput systems, executing complete forward passes for billion-parameter models merely for binary routing introduces severe computational redundancy. Furthermore, in autonomous agent scenarios, inaccurate retrieval causes models to expend excessive tokens on spurious reasoning and redundant tool calls, inflating Time-to-First-Token (TTFT) and costs. We propose Tiny-Critic RAG, decoupling evaluation by deploying a parameter-efficient Small Language Model (SLM) via Low-Rank Adaptation (LoRA). Acting as a deterministic gatekeeper, Tiny-Critic employs constrained decoding and non-thinking inference modes for ultra-low latency binary routing. Evaluations on noise-injected datasets demonstrate Tiny-Critic RAG achieves routing accuracy comparable to GPT-4o-mini while reducing latency by an order of magnitude, establishing a highly cost-effective paradigm for agent deployment.

Biomedical multimodal assistants have the potential to unify radiology, pathology, and clinical-text reasoning, yet a critical deployment gap remains: top-performing systems are either closed-source or computationally prohibitive, precluding the on-premises deployment required for patient privacy and PHI compliance. We introduce MEDGPT-OSS, an open-weight, 20B-parameter generalist vision-language model designed to facilitate open research in clinical AI. Rather than relying on architectural complexity, MEDGPT-OSS pairs the GPT-oss language backbone with a visual front-end via a optimized, three-stage training curriculum. By progressively domain-adapting these modules through rigorous data curation and long-context multimodal alignment, we demonstrate that a 20B model can bridge the capacity gap. It successfully outperforms larger open medical models on out-of-distribution (OOD) multimodal reasoning and complex text-only clinical tasks. By unifying diverse modalities under a single instruction-following interface, MEDGPT-OSS maintains a parameter-efficient footprint fully compatible with commodity GPUs. We release the complete training recipe, open-weight checkpoints, and a rigorous evaluation harness to serve as a verifiable foundation for privacy-preserving, institution-specific clinical AI research.

Nested named entity recognition identifies entities contained within other entities, but requires expensive multi-level annotation. While flat NER corpora exist abundantly, nested resources remain scarce. We investigate whether models can learn nested structure from flat annotations alone, evaluating four approaches: string inclusions (substring matching), entity corruption (pseudo-nested data), flat neutralization (reducing false negative signal), and a hybrid fine-tuned + LLM pipeline. On NEREL, a Russian benchmark with 29 entity types where 21% of entities are nested, our best combined method achieves 26.37% inner F1, closing 40% of the gap to full nested supervision. Code is available at https://github.com/fulstock/Learning-from-Flat-Annotations.

Safe deployment of Large Language Model (LLM) agents in autonomous settings requires reliable oversight mechanisms. A central challenge is detecting scheming, where agents covertly pursue misaligned goals. One approach to mitigating such risks is LLM-based monitoring: using language models to examine agent behaviors for suspicious actions. We study constitutional black-box monitors: prompted classifiers that detect scheming using only externally observable inputs and outputs, optimized on synthetic data generated from natural-language behavior specifications. We introduce two pipelines for generating synthetic agent trajectories, STRIDE (iterative refinement) and Gloom (agent-environment simulation), from which we generate 1,000 samples each. We optimize frontier LLM monitors on these datasets via prompt sweeps, human refinement, and automated prompt optimization, and evaluate performance on 7,500 held-out trajectories from ControlArena, a suite of grounded environments where agents operate in more realistic contexts. Our results demonstrate that monitors selected purely on synthetic data can generalize to more realistic environments, capturing a meaningful scheming signal. However, we find that performance saturates quickly in our setting, with simple prompt sweeps matching the results of more extensive optimization. Pushing beyond this limit yields no further improvements and instead leads to overfitting.

We develop a discrete gauge-theoretic framework for superposition in large language models (LLMs) that replaces the single-global-dictionary premise with a sheaf-theoretic atlas of local semantic charts. Contexts are clustered into a stratified context complex; each chart carries a local feature space and a local information-geometric metric (Fisher/Gauss--Newton) identifying predictively consequential feature interactions. This yields a Fisher-weighted interference energy and three measurable obstructions to global interpretability: (O1) local jamming (active load exceeds Fisher bandwidth), (O2) proxy shearing (mismatch between geometric transport and a fixed correspondence proxy), and (O3) nontrivial holonomy (path-dependent transport around loops). We prove and instantiate four results on a frozen open LLM (Llama~3.2~3B Instruct) using WikiText-103, a C4-derived English web-text subset, and \texttt{the-stack-smol}. (A) After constructive gauge fixing on a spanning tree, each chord residual equals the holonomy of its fundamental cycle, making holonomy computable and gauge-invariant. (B) Shearing lower-bounds a data-dependent transfer mismatch energy, turning $D_{\mathrm{shear}}$ into an unavoidable failure bound. (C) We obtain non-vacuous certified jamming/interference bounds with high coverage and zero violations across seeds/hyperparameters. (D) Bootstrap and sample-size experiments show stable estimation of $D_{\mathrm{shear}}$ and $D_{\mathrm{hol}}$, with improved concentration on well-conditioned subsystems.

Machine unlearning aims to remove the influence of specific training data from pre-trained models without retraining from scratch, and is increasingly important for large language models (LLMs) due to safety, privacy, and legal concerns. Although prior work primarily evaluates unlearning in static, single-turn settings, forgetting robustness under realistic interactive use remains underexplored. In this paper, we study whether unlearning remains stable in interactive environments by examining two common interaction patterns: self-correction and dialogue-conditioned querying. We find that knowledge appearing forgotten in static evaluation can often be recovered through interaction. Although stronger unlearning improves apparent robustness, it often results in behavioral rigidity rather than genuine knowledge erasure. Our findings suggest that static evaluation may overestimate real-world effectiveness and highlight the need for ensuring stable forgetting under interactive settings.

As Large Language Model (LLM) agents increasingly execute complex, autonomous software engineering tasks, developers rely on natural language Agent Instructions (e.g., AGENTS.md) to enforce project-specific coding conventions, tooling, and architectural boundaries. However, these instructions are passive text. Agents frequently deviate from them due to context limitations or conflicting legacy code, a phenomenon we term Context Drift. Because agents operate without real-time human supervision, these silent violations rapidly compound into technical debt. To bridge this gap, we introduce ContextCov, a framework that transforms passive Agent Instructions into active, executable guardrails. ContextCov extracts natural language constraints and synthesizes enforcement checks across three domains: static AST analysis for code patterns, runtime shell shims that intercept prohibited commands, and architectural validators for structural and semantic constraints. Evaluations on 723 open-source repositories demonstrate that ContextCov successfully extracts over 46,000 executable checks with 99.997% syntax validity, providing a necessary automated compliance layer for safe, agent-driven development. Source code and evaluation results are available at https://anonymous.4open.science/r/ContextCov-4510/.

Context: Static code analysis (SCA) tools play a vital role in software development, reducing the cost and time required for code reviews. However, high false-positive and false-negative rates are reported for the best tools in the community. Accordingly, studies often aim to develop datasets for learning SCA warning patterns to reduce false results. These datasets are meant to possess high-quality and high-volume in covering the full range of faults/rules that typically result in false warnings and be compliant with established coding standards. However, existing studies have not utilised such datasets or identified the breadth of rules that are prone to false positives and their compliance to coding standards. Objectives: We analysed code from Stack Overflow and Apache Tomcat to capture variations in code length and style in detecting false-positive warnings from best-performing tools PMD and SonarQube, addressing this gap. Method: In deriving false-positive warnings, outcomes from the tools were labelled using established coding standards. Deeper analyses were then conducted to identify the rules that are prone to false-positives, reasons for these, and agreement/gaps between SCA rules and established standards. Results: Among our main outcomes, we observe that only a few SCA rules generate false positives, ranging from 4.64% to 18.45% across four datasets. Additionally, eliminating rules that contradict established standards significantly reduce the false-positive rate. Additionally, our findings reveal discrepancies between tools and established standards. Conclusion: Given the evidence established in this study, we recommend further investigations into gaps between tools and established standards, including the use of machine learning approaches to annotate larger datasets.

This paper surveys recent developments at the intersection of operator learning, statistical learning theory, and approximation theory. First, it reviews error bounds for empirical risk minimization with a focus on holomorphic operators and neural network approximations. Next, it illustrates fundamental performance limits in terms of sample size by adopting a minimax perspective and considering various notions of regularity beyond holomorphy. The paper ends with a discussion on the interplay between these two perspectives and related open questions.

Work introduces a hierarchical binary tree-based reduction that replaces standard self-attention. The core idea is to use a recursive Gated Linear Unit merge operation, achieving O(n) total merge operations O(log n) parallel depth O(n d^2) total work and O(n) space complexity. In these experiments, the model significantly outperforms standard Transformers in both convergence speed and accuracy on long-range structural dependencies, specifically where hierarchical inductive bias is critical.

In machine learning, curation is used to select the most valuable data for improving both model accuracy and computational efficiency. Recently, curation has also been explored as a solution for private machine learning: rather than training directly on sensitive data, which is known to leak information through model predictions, the private data is used only to guide the selection of useful public data. The resulting model is then trained solely on curated public data. It is tempting to assume that such a model is privacy-preserving because it has never seen the private data. Yet, we show that without further protection, curation pipelines can still leak private information. Specifically, we introduce novel attacks against popular curation methods, targeting every major step: the computation of curation scores, the selection of the curated subset, and the final trained model. We demonstrate that each stage reveals information about the private dataset and that even models trained exclusively on curated public data leak membership information about the private data that guided curation. These findings highlight the previously overlooked inherent privacy risks of data curation and show that privacy assessment must extend beyond the training procedure to include the data selection process. Our differentially private adaptations of curation methods effectively mitigate leakage, indicating that formal privacy guarantees for curation are a promising direction.

A major challenge for world models in multi-agent systems is to understand interdependent agent dynamics, predict interactive multi-agent trajectories, and plan over long horizons with collective awareness, without centralized supervision or explicit communication. In this paper, MetaMind, a general and cognitive world model for multi-agent systems that leverages a novel meta-theory of mind (Meta-ToM) framework, is proposed. Through MetaMind, each agent learns not only to predict and plan over its own beliefs, but also to inversely reason goals and beliefs from its own behavior trajectories. This self-reflective, bidirectional inference loop enables each agent to learn a metacognitive ability in a self-supervised manner. Then, MetaMind is shown to generalize the metacognitive ability from first-person to third-person through analogical reasoning. Thus, in multi-agent systems, each agent with MetaMind can actively reason about goals and beliefs of other agents from limited, observable behavior trajectories in a zero-shot manner, and then adapt to emergent collective intention without an explicit communication mechanism. Extended simulation results on diverse multi-agent tasks demonstrate that MetaMind can achieve superior task performance and outperform baselines in few-shot multi-agent generalization.

The proliferation of neural radiance field (NeRF) research requires significant efforts to reimplement papers before building upon them. We introduce NERFIFY, a multi-agent framework that reliably converts NeRF research papers into trainable Nerfstudio plugins, in contrast to generic paper-to-code methods and frontier models like GPT-5 that usually fail to produce runnable code. NERFIFY achieves domain-specific executability through six key innovations: (1) Context-free grammar (CFG): LLM synthesis is constrained by Nerfstudio formalized as a CFG, ensuring generated code satisfies architectural invariants. (2) Graph-of-Thought code synthesis: Specialized multi-file-agents generate repositories in topological dependency order, validating contracts and errors at each node. (3) Compositional citation recovery: Agents automatically retrieve and integrate components (samplers, encoders, proposal networks) from citation graphs of references. (4) Visual feedback: Artifacts are diagnosed through PSNR-minima ROI analysis, cross-view geometric validation, and VLM-guided patching to iteratively improve quality. (5) Knowledge enhancement: Beyond reproduction, methods can be improved with novel optimizations. (6) Benchmarking: An evaluation framework is designed for NeRF paper-to-code synthesis across 30 diverse papers. On papers without public implementations, NERFIFY achieves visual quality matching expert human code (+/-0.5 dB PSNR, +/-0.2 SSIM) while reducing implementation time from weeks to minutes. NERFIFY demonstrates that a domain-aware design enables code translation for complex vision papers, potentiating accelerated and democratized reproducible research. Code, data and implementations will be publicly released.

We study an identification problem in multi-armed bandits. In each round a learner selects one of $K$ arms and observes its reward, with the goal of eventually identifying an arm that will perform best at a {\it future} time. In adversarial environments, however, past performance may offer little information about the future, raising the question of whether meaningful identification is possible at all. In this work, we introduce \emph{lookahead identification}, a task in which the goal of the learner is to select a future prediction window and commit in advance to an arm whose average reward over that window is within $\varepsilon$ of optimal. Our analysis characterizes both the achievable accuracy of lookahead identification and the memory resources required to obtain it. From an accuracy standpoint, for any horizon $T$ we give an algorithm achieving $\varepsilon = O\bigl(1/\sqrt{\log T}\bigr)$ over $Ω(\sqrt{T})$ prediction windows. This demonstrates that, perhaps surprisingly, identification is possible in adversarial settings, despite significant lack of information. We also prove a near-matching lower bound showing that $\varepsilon = Ω\bigl(1/\log T\bigr)$ is unavoidable. We then turn to investigate the role of memory in our problem, first proving that any algorithm achieving nontrivial accuracy requires $Ω(K)$ bits of memory. Under a natural \emph{local sparsity} condition, we show that the same accuracy guarantees can be achieved using only poly-logarithmic memory.

Language agents increasingly act as web-enabled systems that search, browse, and synthesize information from diverse sources. However, these sources can include unreliable or adversarial content, and the robustness of agents to adversarial ranking - where misleading information appears prominently in search results - remains poorly understood. Existing benchmarks evaluate functional navigation or static factuality but cannot causally isolate this vulnerability, and current mitigation strategies for retrieval-augmented generation remain largely untested under such conditions. We introduce Synthetic Web Benchmark, a procedurally generated environment comprising thousands of hyperlinked articles with ground-truth labels for credibility and factuality, process-level interaction traces, and contamination filtering to eliminate training-data leakage. By injecting a single high-plausibility misinformation article into a controllable search rank, we measure the causal effect of adversarial exposure in six frontier models. The results reveal catastrophic failures: accuracy collapses despite unlimited access to truthful sources, with minimal search escalation and severe miscalibration. These findings expose fundamental limitations in how current frontier models handle conflicting information, with immediate implications for deployment in high-stakes domains. Our benchmark enables systematic analysis of these failure modes and provides a controlled testbed for evaluating mitigation strategies under adversarial ranking - a gap in current research. This work establishes a reproducible baseline for developing search-robust and epistemically humble agents capable of resisting manipulation in high-stakes domains.

Template-based segmentation, a widely used paradigm in medical imaging, propagates anatomical labels via deformable registration from a labeled atlas to a target image, and is often used to compute volumetric biomarkers for downstream decision-making. While conformal prediction (CP) provides finite-sample valid intervals for scalar metrics, existing segmentation-based uncertainty quantification (UQ) approaches either rely on learned model features, often unavailable in classic template-based pipelines, or treat the registration process as a black box, resulting in overly conservative intervals when applied directly in output space. We introduce ConVOLT, a CP framework that achieves efficient volumetric UQ by conditioning calibration on properties of the estimated deformation field from template-based segmentation. ConVOLT calibrates a learned volumetric scaling factor from deformation space features. We evaluate ConVOLT on template-based segmentation tasks involving global, regional, and label volumetry across multiple datasets and registration methods. ConVOLT achieves target coverage while producing substantially tighter intervals than output-space conformal baselines. Our work paves way to exploit the registration process for efficient UQ in medical imaging pipelines.

Fluid-solid interaction (FSI) problems are fundamental in many scientific and engineering applications, yet effectively capturing the highly nonlinear two-way interactions remains a significant challenge. Most existing deep learning methods are limited to simplified one-way FSI scenarios, often assuming rigid and static solid to reduce complexity. Even in two-way setups, prevailing approaches struggle to capture dynamic, heterogeneous interactions due to the lack of cross-domain awareness. In this paper, we introduce \textbf{Fisale}, a data-driven framework for handling complex two-way \textbf{FSI} problems. It is inspired by classical numerical methods, namely the Arbitrary Lagrangian-Eulerian (\textbf{ALE}) method and the partitioned coupling algorithm. Fisale explicitly models the coupling interface as a distinct component and leverages multiscale latent ALE grids to provide unified, geometry-aware embeddings across domains. A partitioned coupling module (PCM) further decomposes the problem into structured substeps, enabling progressive modeling of nonlinear interdependencies. Compared to existing models, Fisale introduces a more flexible framework that iteratively handles complex dynamics of solid, fluid and their coupling interface on a unified representation, and enables scalable learning of complex two-way FSI behaviors. Experimentally, Fisale excels in three reality-related challenging FSI scenarios, covering 2D, 3D and various tasks. The code is available at \href{https://github.com/therontau0054/Fisale}.

Local journalism is vital in democratic societies where it informs people about local issues like, school board elections, small businesses, local health services, etc. But mounting economic pressures have made it increasingly difficult for local news stations to report these issues, underscoring the need to identify the salient geographical locations covered in local news (geo-foci). In response, we propose a novel geo-foci model for labeling US local news articles with the geographic locations (i.e., the names of counties, cities, states, countries) central to their subject matter. First, we manually labeled US local news articles from all 50 states with four administrative division labels (local, state, national, and international) corresponding to their geo-foci, and none for articles without a geographic focus. Second, we extracted and disambiguated geographic locations from them using Large Language Models (LLMs), since local news often contains ambiguous geographic entities (e.g., Paris, Texas vs. Paris, France). LLMs outperformed all eight geographic entity disambiguation methods we evaluated. Third, we engineered a rich set of spatial-semantic features capturing the prominence, frequency, and contextual positions of geographic entities. Using these features, we trained a classifier to accurately (F1: 0.86) detect the geographic foci of US local news articles. Our model could be applied to assess shifts from local to national narratives, and more broadly, enable researchers to better study local media.

Understanding how large-scale functional brain networks reorganize during cognitive decline remains a central challenge in neuroimaging. While recent self-supervised models have shown promise for learning representations from resting-state fMRI, their internal mechanisms are difficult to interpret, limiting mechanistic insight. We propose BrainInterNet, a network-aware self-supervised framework based on masked reconstruction with cross-attention that explicitly models inter-network dependencies in rs-fMRI. By selectively masking predefined functional networks and reconstructing them from remaining context, our approach enables direct quantification of network predictability and interpretable analysis of cross-network interactions. We train BrainInterNet on multi-cohort fMRI data (from the ABCD, HCP Development, HCP Young Adults, and HCP Aging datasets) and evaluate on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, in total comprising 5,582 recordings. Our method reveals systematic alterations in the brain's network interactions under AD, including in the default mode, limbic, and attention networks. In parallel, the learned representations support accurate Alzheimer's-spectrum classification and yield a compact summary marker that tracks disease severity longitudinally. Together, these results demonstrate that network-guided masked modeling with cross-attention provides an interpretable and effective framework for characterizing functional reorganization in neurodegeneration.

Autonomous navigation under uncertainty requires solving partially observable Markov decision processes (POMDPs) for planning and assigning sensor measurements to tracked targets--a task known as multi-target data association (MTDA). Both problems become computationally demanding at scale: belief conditioning costs $\mathcal{O}(P(e)^{-1})$ per node under rare evidence, while MTDA is NP-hard. Quantum amplitude amplification can quadratically reduce the belief-update query cost to $\mathcal{O}(P(e)^{-1/2})$, while QUBO reformulations expose MTDA to quantum and quantum-inspired optimisation heuristics. We present QANTIS, a modular platform that integrates quantum belief update (Grover amplitude amplification and BIQAE), QUBO-based data association via FPC-QAOA, and composable error mitigation, and we report a 45-experiment hardware study on three IBM Heron backends. On hardware, a single Grover iterate applied to a Tiger belief oracle amplifies a rare observation probability from $0.179$ to $0.907$ ($5.1\times$; ISA 18) while preserving the Bayesian posterior (Hellinger $0.0015$), increasing usable-shot yield from 1,463 to 7,429. We interpret this as a hardware validation of the quadratic query-complexity mechanism at $k=1$ with posterior preservation, rather than a wall-clock advantage claim. We further demonstrate, to our knowledge, the first closed-loop hybrid quantum-classical Tiger POMDP on superconducting hardware ($T=8$, max Hellinger below $0.015$), and empirically characterise NISQ feasibility boundaries: ZNE-based error mitigation is beneficial below ISA $\approx 100$ and harmful above ISA $\gtrsim 1{,}000$; FPC-QAOA is meaningful at $\leq 15$ QUBO variables (ISA $\lesssim 450$). These results characterise practical operating regimes on current superconducting hardware rather than wall-clock quantum advantage at today's problem scales.

Score-based generative models (SGMs) aim at generating samples from a target distribution by approximating the reverse-time dynamics of a stochastic differential equation. Despite their strong empirical performance, classical samplers initialized from a Gaussian distribution require a long time horizon noising typically inducing a large number of discretization steps and high computational cost. In this work, we present a Kullback-Leibler convergence analysis of Variance Exploding diffusion samplers that highlights the critical role of the backward process initialization. Based on this result, we propose a theoretically grounded sampling strategy that learns the reverse-time initialization, directly minimizing the initialization error. The resulting procedure is independent of the specific score training procedure, network architecture, and discretization scheme. Experiments on toy distributions and benchmark datasets demonstrate competitive or improved generative quality while using significantly fewer sampling steps.

A black hole is a malicious node in a graph that destroys any resource entering it without leaving a trace. In the Black Hole Search (BHS) problem with mobile agents, at least one agent must survive and terminate after locating the black hole. Recently, BHS has been studied on 1-bounded 1-interval connected dynamic graphs \cite{BHS_gen}, where a footprint graph exists and at most one edge may disappear per round while connectivity is preserved. Under this model, \cite{BHS_gen} presents an algorithm for the rooted initial configuration, where all agents start from a single node, and proves that at least $2δ_{BH}+1$ agents are necessary in the scattered initial configuration, where agents are arbitrarily placed and $δ_{BH}$ denotes the degree of the black hole. We present an algorithm that solves BHS in the scattered setting using $2δ_{BH}+17$ agents, matching asymptotically the rooted algorithm of \cite{BHS_gen} under the same assumptions. We further investigate the Eventual Black Hole Search (\textsc{Ebhs}) problem, where the black hole may appear at any node and at any time during execution, destroying all agents located there upon its emergence; however, it cannot appear at the home base in round 0, where all agents are initially co-located, and once created, it remains permanently active. While \textsc{Ebhs} has been studied on static rings \cite{Bonnet25}, we extend it to arbitrary static graphs and provide a solution using the minimum number of agents. For rings, our algorithm is optimal in both the number of agents and the running time, and it does not require knowledge of global parameters or additional model assumptions.

We introduce an algorithm that performs a one-directional mesh overset of a parallel forest of octrees with another distributed mesh of unrelated partition. The forest mesh consists of several adaptively refined octrees. Individual smooth mappings for every tree allow to represent a broad range of geometric domains. The other mesh is generic and defines a distributed set of query points, e.g. stemming from a quadrature rule applied to each cell. We face the problem of finding data for all queries in the remote forest. The forest is partitioned according to its natural Morton ordering. Thus, the partition boundaries can be encoded globally with one Morton index per process, which allows for precise, communication-free searching of the queries in the partition geometry. This is necessary to organize non-blocking communication of the queries to the relevant processes only. In a subsequent local search of the forest, we process the incoming queries and return the data of interest to each query's origin. The algorithm can be generalized, for example to load balancing of the overset, or adaptive refinement of the meshes around their intersection area. In 2D and 3D example scenarios we demonstrate the algorithm's performance and scalability to 12,288 processes.

Precision medicine promises to transform health care by offering individualised treatments that dramatically improve clinical outcomes. A necessary prerequisite is to identify subgroups of patients who respond differently to different therapies. Current approaches are limited to static measures of treatment success, neglecting the repeated measures found in most clinical trials. Our approach combines the concept of partly conditional modelling with treatment effect estimation based on the Virtual Twins method. The resulting time-specific responses to treatment are characterised using survLIME, an extension of Local Interpretable Model-agnostic Explanations (LIME) to survival data. Performance was evaluated using synthetic data and applied to clinical trials examining the effectiveness of panitumumab to treat metastatic colorectal cancer. An area under the receiver operating characteristic curve (AUC) of 0.77 for identifying fixed responders was achieved in a 1000 patient simulation. When considering dynamic responders, partly conditional modelling increased the AUC from 0.597 to 0.685. Applying the approach to colorectal cancer trials found genetic mutations, sites of metastasis, and ethnicity as important factors for response to treatment. Our approach can accommodate a dynamic response to treatment while potentially providing better performance than existing methods in instances of a fixed response to treatment. When applied to clinical data we attain results consistent with the literature.

Stroke is a major cause of death and disability worldwide. Accurate outcome and evolution prediction has the potential to revolutionize stroke care by individualizing clinical decision-making leading to better outcomes. However, despite a plethora of attempts and the rich data provided by neuroimaging, modelling the ultimate fate of brain tissue remains a challenging task. In this work, we apply recent ideas in the field of diffusion probabilistic models to generate a self-supervised semantically meaningful stroke representation from Computed Tomography (CT) images. We then improve this representation by extending the method to accommodate longitudinal images and the time from stroke onset. The effectiveness of our approach is evaluated on a dataset consisting of 5,824 CT images from 3,573 patients across two medical centers with minimal labels. Comparative experiments show that our method achieves the best performance for predicting next-day severity and functional outcome at discharge.

Handwritten character classification in the Bengali script is a significant challenge due to the complexity and variability of the characters. The models commonly used for classification are often computationally expensive and data-hungry, making them unsuitable for resource-limited languages such as Bengali. In this experiment, we propose a novel, efficient, and lightweight Vision Transformer model that effectively classifies Bengali handwritten basic characters and digits, addressing several shortcomings of traditional methods. The proposed solution utilizes a deep convolutional neural network (DCNN) in a more simplified manner compared to traditional DCNN architectures, with the aim of reducing computational burden. With only 0.65 million parameters, a model size of 0.62 MB, and 0.16 GFLOPs, our model, BornoViT, is significantly lighter than current state-of-the-art models, making it more suitable for resource-limited environments, which is essential for Bengali handwritten character classification. BornoViT was evaluated on the BanglaLekha Isolated dataset, achieving an accuracy of 95.77%, and demonstrating superior efficiency compared to existing state-of-the-art approaches. Furthermore, the model was evaluated on our self-collected dataset, Bornomala, consisting of approximately 222 samples from different age groups, where it achieved an accuracy of 91.51%.

This paper introduces General Proximal Flow Networks (GPFNs), a generalization of Bayesian Flow Networks that broadens the class of admissible belief-update operators. In Bayesian Flow Networks, each update step is a Bayesian posterior update, which is equivalent to a proximal step with respect to the Kullback-Leibler divergence. GPFNs replace this fixed choice with an arbitrary divergence or distance function, such as the Wasserstein distance, yielding a unified proximal-operator framework for iterative generative modeling. The corresponding training and sampling procedures are derived, establishing a formal link to proximal optimization and recovering the standard BFN update as a special case. Empirical evaluations confirm that adapting the divergence to the underlying data geometry yields measurable improvements in generation quality, highlighting the practical benefits of this broader framework.

Speech is a natural means of conveying emotions, making it an effective method for understanding and representing human feelings. Reliable speech emotion recognition (SER) is central to applications in human-computer interaction, healthcare, education, and customer service. However, most SER methods depend on heavy backbone models or hand-crafted features that fail to balance accuracy and efficiency, particularly for low-resource languages like Bangla. In this work, we present SpectroFusion-ViT, a lightweight SER framework built utilizing EfficientViT-b0, a compact Vision Transformer architecture equipped with self-attention to capture long-range temporal and spectral patterns. The model contains only 2.04M parameters and requires 0.1 GFLOPs, enabling deployment in resource-constrained settings without compromising accuracy. Our pipeline first performs preprocessing and augmentation on raw audio, then extracts Chroma and Mel-frequency cepstral coefficient (MFCC) features. These representations are fused into a complementary time-frequency descriptor that preserves both fine-grained spectral detail and broader harmonic structure. Using transfer learning, EfficientViT-b0 is fine-tuned for multi-class emotion classification. We evaluate the system on two benchmark Bangla emotional speech datasets, SUBESCO and BanglaSER, which vary in speaker diversity, recording conditions, and acoustic characteristics. The proposed approach achieves 92.56% accuracy on SUBESCO and 82.19% on BanglaSER, surpassing existing state-of-the-art methods. These findings demonstrate that lightweight transformer architectures can deliver robust SER performance while remaining computationally efficient for real-world deployment.

Accurate Remaining Useful Life (RUL) prediction is a key requirement for effective Prognostics and Health Management (PHM) in safety-critical systems such as aero-engines. Existing deep learning approaches, particularly LSTM-based models, often struggle to generalize across varying operating conditions and are sensitive to noise in multivariate sensor data. To address these challenges, we propose a novel Bidirectional Residual Corrected LSTM (Bi-cLSTM) model for robust RUL estimation. The proposed architecture combines bidirectional temporal modeling with an adaptive residual correction mechanism to iteratively refine sequence representations. In addition, we introduce a condition-aware preprocessing pipeline incorporating regime-based normalization, feature selection, and exponential smoothing to improve robustness under complex operating environments. Extensive experiments on all four subsets of the NASA C-MAPSS dataset demonstrate that the proposed Bi-cLSTM consistently outperforms LSTM-based baselines and achieves competitive state-of-the-art performance, particularly in challenging multi-condition scenarios. These results highlight the effectiveness of combining bidirectional temporal learning with residual correction for reliable RUL prediction.

In this work, we propose a new deep learning model for Genomic Prediction (GP), which involves correlating genotypic data with phenotypic. The genotypes are typically fed as a sequence of characters to the 1D-Convolution Neural Network layer of the underlying deep learning model. Inspired by earlier work that represented genotype as a 2D-image for genotype-phenotype classification, we extend this idea to GP, which is a regression task. We use a ResNet-18 as the underlying architecture, and term this model as ResGene-2D. Although the 2D-image representation captures biological interactions well, it requires all the layers of the model to do so. This limits training efficiency. Thus, as seen in the earlier work that proposed a 2D-image representation, our ResGene-2D performs almost the same as other models (3% improvement). To overcome this, we propose a novel idea of converting the 2D-image into a 3D/ tensor and feed this to the ResNet-18 architecture, and term this model as ResGene-T. We evaluate our proposed models on three crop species having ten phenotypic traits and compare it with seven most popular models (two statistical, two machine learning, and three deep learning). ResGene-T performs the best among all these seven methods (gains from 14.51% to 41.51%).

For a long period of time, Adam has served as the ubiquitous default choice for training deep neural networks. Recently, many new optimizers have been introduced, out of which Muon has perhaps gained the highest popularity due to its superior training speed. While many papers set out to validate the benefits of Muon, our paper investigates the potential downsides stemming from the mechanism driving this speedup. We explore the biases induced when optimizing with Muon, providing theoretical analysis and its consequences to the learning trajectories and solutions learned. While the theory does provide justification for the benefits Muon brings, it also guides our intuition when coming up with a couple of examples where Muon-optimized models have disadvantages. The core problem we emphasize is that Muon optimization removes a simplicity bias that is naturally preserved by older, more thoroughly studied methods like Stochastic Gradient Descent (SGD). We take first steps toward understanding consequences this may have: Muon might struggle to uncover common underlying structure across tasks, and be more prone to fitting spurious features. More broadly, this paper should serve as a reminder: when developing new optimizers, it is essential to consider the biases they introduce, as these biases can fundamentally change a model's behavior -- for better or for worse.

The development of deep learning methods for magnetic resonance spectroscopy (MRS) is often hindered by limited availability of large, high-quality training datasets. While physics-based simulations are commonly used to mitigate this limitation, accurately modeling all in-vivo signal components remains challenging. In this work, we propose a data-driven framework for synthesizing in-vivo MRS data using a variational autoencoder (VAE) trained exclusively on measured single-voxel spectroscopy data. The model learns a low-dimensional latent representation of complex-valued spectra and enables generation of new samples through latent-space sampling and interpolation. The generative performance of the proposed approach is evaluated using a comprehensive set of complementary analyses, including reconstruction quality, feature-level similarity using low-dimensional embeddings, application-based signal quality metrics, and metabolite quantification agreement. The results demonstrate that the VAE accurately reconstructs dominant spectral patterns and generates synthetic spectra that occupy the same feature space as in-vivo data. In an example application targeting GABA-edited spectroscopy, augmenting limited subsets of transients with synthetic spectra improves signal quality metrics such as signal-to-noise ratio, linewidth, and shape scores. However, the results also reveal limitations of the generative approach, including under-representation of stochastic noise and reduced accuracy in absolute metabolite quantification, particularly for applications sensitive to concentration estimates. These findings highlight both potential and limitations of data-driven MRS synthesis. Beyond the proposed model, this study introduces a structured evaluation framework for generative MRS methods, emphasizing the importance of application-aware validation when synthetic data are used for downstream analysis.

Deep reinforcement learning (RL) has been applied extensively to solve complex decision-making problems. In many real-world scenarios, tasks often have several conflicting objectives and may require multiple agents to cooperate, which are the multi-objective multi-agent decision-making problems. However, only few works have been conducted on this intersection. Existing approaches are limited to separate fields and can only handle multi-agent decision-making with a single objective, or multi-objective decision-making with a single agent. In this paper, we propose MO-MIX to solve the multi-objective multi-agent reinforcement learning (MOMARL) problem. Our approach is based on the centralized training with decentralized execution (CTDE) framework. A weight vector representing preference over the objectives is fed into the decentralized agent network as a condition for local action-value function estimation, while a mixing network with parallel architecture is used to estimate the joint action-value function. In addition, an exploration guide approach is applied to improve the uniformity of the final non-dominated solutions. Experiments demonstrate that the proposed method can effectively solve the multi-objective multi-agent cooperative decision-making problem and generate an approximation of the Pareto set. Our approach not only significantly outperforms the baseline method in all four kinds of evaluation metrics, but also requires less computational cost.

We present Qwen3-Coder-Next, an open-weight language model specialized for coding agents. Qwen3-Coder-Next is an 80-billion-parameter model that activates only 3 billion parameters during inference, enabling strong coding capability with efficient inference. In this work, we explore how far strong training recipes can push the capability limits of models with small parameter footprints. To achieve this, we perform agentic training through large-scale synthesis of verifiable coding tasks paired with executable environments, allowing learning directly from environment feedback via mid-training and reinforcement learning. Across agent-centric benchmarks including SWE-Bench and Terminal-Bench, Qwen3-Coder-Next achieves competitive performance relative to its active parameter count. We release both base and instruction-tuned open-weight versions to support research and real-world coding agent development.

Runtime monitoring checks, during execution, whether a partial signal produced by a hybrid system satisfies its specification. Signal First-Order Logic (SFO) offers expressive real-time specifications over such signals, but currently comes only with Boolean semantics and has no tool support. We provide the first robustness-based quantitative semantics for SFO, enabling the expression and evaluation of rich real-time properties beyond the scope of existing formalisms such as Signal Temporal Logic. To enable online monitoring, we identify a past-time fragment of SFO and give a pastification procedure that transforms bounded-response SFO formulas into equisatisfiable formulas in this fragment. We then develop an efficient runtime monitoring algorithm for this past-time fragment and evaluate its performance on a set of benchmarks, demonstrating the practicality and effectiveness of our approach. To the best of our knowledge, this is the first publicly available prototype for online quantitative monitoring of full SFO.

Effectively searching time-series data is essential for system analysis, but existing methods often require expert-designed similarity criteria or rely on global, series-level descriptions. We study language-driven segment retrieval: given a natural language query, the goal is to retrieve relevant local segments from large time-series repositories. We build large-scale segment--caption training data by applying TV2-based segmentation to LOTSA windows and generating segment descriptions with GPT-5.2, and then train a Conformer-based contrastive retriever in a shared text--time-series embedding space. On a held-out test split, we evaluate single-positive retrieval together with caption-side consistency (SBERT and VLM-as-a-judge) under multiple candidate pool sizes. Across all settings, LaSTR outperforms random and CLIP baselines, yielding improved ranking quality and stronger semantic agreement between retrieved segments and query intent.

Large language model alignment via reinforcement learning depends critically on reward function quality. However, static, domain-specific reward models are often costly to train and exhibit poor generalization in out-of-distribution scenarios encountered during RL iterations. We present RLAR (Reinforcement Learning from Agent Rewards), an agent-driven framework that dynamically assigns tailored reward functions to individual queries. Specifically, RLAR transforms reward acquisition into a dynamic tool synthesis and invocation task. It leverages LLM agents to autonomously retrieve optimal reward models from the Internet and synthesize programmatic verifiers through code generation. This allows the reward system to self-evolve with the shifting data distributions during training. Experimental results demonstrate that RLAR yields consistent performance gains ranging from 10 to 60 across mathematics, coding, translation, and dialogue tasks. On RewardBench-V2, RLAR significantly outperforms static baselines and approaches the performance upper bound, demonstrating superior generalization through dynamic reward orchestration. The data and code are available on this link: https://github.com/ZhuoerFeng/RLAR.

Fine-tuning Multimodal Large Language Models (MLLMs) with parameter-efficient methods like Low-Rank Adaptation (LoRA) is crucial for task adaptation. However, imbalanced training dynamics across modalities often lead to suboptimal accuracy due to negative interference, a challenge typically addressed with inefficient heuristic methods such as manually tuning separate learning rates. To overcome this, we introduce MARS (Multimodal Adaptive Rank Search), an approach to discover optimal rank pairs that balance training dynamics while maximizing performance. Our key innovation, a proposed framework of dual scaling laws, enables this search: one law models module-specific convergence time to prune the search space to candidates with aligned dynamics, while the other predicts final task performance to select the optimal pair from the pruned set. By re-purposing the LoRA rank as a controller for modality-specific convergence speed, MARS outperforms baseline methods and provides a robust, automated strategy for optimizing MLLM fine-tuning.

Real-world tool-using agents operate over long-horizon workflows with recurring structure and diverse demands, where effective behavior requires not only invoking atomic tools but also abstracting, and reusing higher-level tool compositions. However, existing benchmarks mainly measure instance-level success under static tool sets, offering limited insight into agents' ability to acquire such reusable skills. We address this gap by introducing SkillCraft, a benchmark explicitly stress-test agent ability to form and reuse higher-level tool compositions, where we call Skills. SkillCraft features realistic, highly compositional tool-use scenarios with difficulty scaled along both quantitative and structural dimensions, designed to elicit skill abstraction and cross-task reuse. We further propose a lightweight evaluation protocol that enables agents to auto-compose atomic tools into executable Skills, cache and reuse them inside and across tasks, thereby improving efficiency while accumulating a persistent library of reusable skills. Evaluating state-of-the-art agents on SkillCraft, we observe substantial efficiency gains, with token usage reduced by up to 80% by skill saving and reuse. Moreover, success rate strongly correlates with tool composition ability at test time, underscoring compositional skill acquisition as a core capability.

We study computationally and statistically efficient reinforcement learning under the linear $Q^π$ realizability assumption, where any policy's $Q$-function is linear in a given state-action feature representation. Prior methods in this setting are either computationally intractable, or require (local) access to a simulator. In this paper, we propose a computationally efficient online RL algorithm, named Frozen Policy Iteration, under the linear $Q^π$ realizability setting that works for Markov Decision Processes (MDPs) with stochastic initial states, stochastic rewards and deterministic transitions. Our algorithm achieves a regret bound of $\widetilde{O}(\sqrt{d^2H^6T})$, where $d$ is the dimensionality of the feature space, $H$ is the horizon length, and $T$ is the total number of episodes. Our regret bound is optimal for linear (contextual) bandits which is a special case of our setting with $H = 1$. Existing policy iteration algorithms under the same setting heavily rely on repeatedly sampling the same state by access to the simulator, which is not implementable in the online setting with stochastic initial states studied in this paper. In contrast, our new algorithm circumvents this limitation by strategically using only high-confidence part of the trajectory data and freezing the policy for well-explored states, which ensures that all data used by our algorithm remains effectively on-policy during the whole course of learning. We further demonstrate the versatility of our approach by extending it to the Uniform-PAC setting and to function classes with bounded eluder dimension.

Backdoor attacks pose a critical threat to the security of deep neural networks, yet existing efforts on universal backdoors often rely on visually salient patterns, making them easier to detect and less practical at scale. In this work, we introduce a novel imperceptible universal backdoor attack that simultaneously controls all target classes with minimal poisoning while preserving stealth. Our key idea is to leverage graph convolutional networks (GCNs) to model inter-class relationships and generate class-specific perturbations that are both effective and visually invisible. The proposed framework optimizes a dual-objective loss that balances stealthiness (measured by perceptual similarity metrics such as PSNR) and attack success rate (ASR), enabling scalable, multi-target backdoor injection. Extensive experiments on ImageNet-1K with ResNet architectures demonstrate that our method achieves high ASR (up to 91.3%) under poisoning rates as low as 0.16%, while maintaining benign accuracy and evading state-of-the-art defenses. These results highlight the emerging risks of invisible universal backdoors and call for more robust detection and mitigation strategies.

This paper presents a controlled empirical study of biologically motivated local learning for handwritten digit recognition. We evaluate an STDP-inspired competitive proxy and a practical hybrid benchmark built on the same spiking population encoder. The proxy is motivated by leaky integrate-and-fire E/I circuit models with three-factor delayed reward modulation. The hybrid update is local in pre x post rates but uses supervised labels and no timing-based credit assignment. On sklearn digits, fixed-seed evaluation shows classical pixel baselines from 98.06 to 98.22% accuracy, while local spike-based models reach 86.39 +/- 4.75% (hybrid default) and 87.17 +/- 3.74% (STDP-style competitive proxy). Ablations identify normalization and reward-shaping settings as the strongest observed levers, with a best hybrid ablation of 95.52 +/- 1.11%. A network-free synthetic temporal benchmark supports the same timing-versus-rate interpretation under matched local-update training. A descriptive 2x2 analysis further shows reward-shaping effects can reverse sign across stabilization regimes, so reward-shaping conclusions should be reported jointly with normalization settings.

Data flow scheduling for high-throughput multibeam satellites is a challenging NP-hard combinatorial optimization problem. As the problem scales, traditional methods, such as Mixed-Integer Linear Programming and heuristic schedulers, often face a trade-off between solution quality and real-time feasibility. In this paper, we present a hybrid quantum-classical framework that improves scheduling efficiency by casting Multi-Beam Time-Frequency Slot Assignment (MB-TFSA) as a Quadratic Unconstrained Binary Optimization (QUBO) problem. We incorporate the throughput-maximization objective and operational constraints into a compact QUBO via parameter rescaling to keep the formulation tractable. To address optimization challenges in variational quantum algorithms, such as barren plateaus and rugged loss landscapes, we introduce a layer-wise training strategy that gradually increases circuit depth while iteratively refining the solution. We evaluate solution quality, runtime, and robustness on quantum hardware, and benchmark against classical and hybrid baselines using realistic, simulated satellite traffic workloads.

Large language models (LLMs) are increasingly used as reasoning engines in autonomous driving, yet their decision-making remains opaque. We propose to study their decision process through counterfactual explanations, which identify the minimal semantic changes to a scene description required to alter a driving plan. We introduce DRIV-EX, a method that leverages gradient-based optimization on continuous embeddings to identify the input shifts required to flip the model's decision. Crucially, to avoid the incoherent text typical of unconstrained continuous optimization, DRIV-EX uses these optimized embeddings solely as a semantic guide: they are used to bias a controlled decoding process that re-generates the original scene description. This approach effectively steers the generation toward the counterfactual target while guaranteeing the linguistic fluency, domain validity, and proximity to the original input, essential for interpretability. Evaluated using the LC-LLM planner on a textual transcription of the highD dataset, DRIV-EX generates valid, fluent counterfactuals more reliably than existing baselines. It successfully exposes latent biases and provides concrete insights to improve the robustness of LLM-based driving agents.

Explainability and transparent decision-making are essential for the safe deployment of autonomous driving systems. Scene captioning summarizes environmental conditions and risk factors in natural language, improving transparency, safety, and human--robot interaction. However, most existing approaches target structured urban scenarios; in off-road environments, they are vulnerable to single-modality degradations caused by rain, fog, snow, and darkness, and they lack a unified framework that jointly models structured scene captioning and path planning. To bridge this gap, we propose Wild-Drive, an efficient framework for off-road scene captioning and path planning. Wild-Drive adopts modern multimodal encoders and introduces a task-conditioned modality-routing bridge, MoRo-Former, to adaptively aggregate reliable information under degraded sensing. It then integrates an efficient large language model (LLM), together with a planning token and a gate recurrent unit (GRU) decoder, to generate structured captions and predict future trajectories. We also build the OR-C2P Benchmark, which covers structured off-road scene captioning and path planning under diverse sensor corruption conditions. Experiments on OR-C2P dataset and a self-collected dataset show that Wild-Drive outperforms prior LLM-based methods and remains more stable under degraded sensing. The code and benchmark will be publicly available at https://github.com/wangzihanggg/Wild-Drive.

The world is undergoing a major demographic shift as older adults become a rapidly growing share of the population, creating new challenges for driving safety. In car-dependent regions such as the United States, driving remains essential for independence, access to services, and social participation. At the same time, aging can introduce gradual changes in vision, attention, reaction time, and driving control that quietly reduce safety. Today's assessment methods rely largely on infrequent clinic visits or simple screening tools, offering only a brief snapshot and failing to reflect how an older adult actually drives on the road. Our work starts from the observation that everyday driving provides a continuous record of functional ability and captures how a driver responds to traffic, navigates complex roads, and manages routine behavior. Leveraging this insight, we propose AURA, an Artificial Intelligence of Things (AIoT) framework for continuous, real-world assessment of driving safety among older adults. AURA integrates richer in-vehicle sensing, multi-scale behavioral modeling, and context-aware analysis to extract detailed indicators of driving performance from routine trips. It organizes fine-grained actions into longer behavioral trajectories and separates age-related performance changes from situational factors such as traffic, road design, or weather. By integrating sensing, modeling, and interpretation within a privacy-preserving edge architecture, AURA provides a foundation for proactive, individualized support that helps older adults drive safely. This paper outlines the design principles, challenges, and research opportunities needed to build reliable, real-world monitoring systems that promote safer aging behind the wheel.

Large Language Models have evolved from single-round generators into long-horizon agents, capable of complex text synthesis scenarios. However, current evaluation frameworks lack the ability to assess the actual synthesis operations, such as outlining, drafting, and editing. Consequently, they fail to evaluate the actual and detailed capabilities of LLMs. To bridge this gap, we introduce RAVEL, an agentic framework that enables the LLM testers to autonomously plan and execute typical synthesis operations, including outlining, drafting, reviewing, and refining. Complementing this framework, we present C3EBench, a comprehensive benchmark comprising 1,258 samples derived from professional human writings. We utilize a "reverse-engineering" pipeline to isolate specific capabilities across four tasks: Cloze, Edit, Expand, and End-to-End. Through our analysis of 14 LLMs, we uncover that most LLMs struggle with tasks that demand contextual understanding under limited or under-specified instructions. By augmenting RAVEL with SOTA LLMs as operators, we find that such agentic text synthesis is dominated by the LLM's reasoning capability rather than raw generative capacity. Furthermore, we find that a strong reasoner can guide a weaker generator to yield higher-quality results, whereas the inverse does not hold. Our code and data are available at this link: https://github.com/ZhuoerFeng/RAVEL-Reasoning-Agents-Text-Eval.

State-of-the-art speech-to-text models typically employ Transformer-based encoders that model token dependencies via self-attention mechanisms. However, the quadratic complexity of self-attention in both memory and computation imposes significant constraints on scalability. In this work, we propose a novel token-mixing mechanism, the Polynomial Mixer (PoM), as a drop-in replacement for multi-head self-attention. PoM computes a polynomial representation of the input with linear complexity with respect to the input sequence length. We integrate PoM into a self-supervised speech representation learning framework based on BEST-RQ and evaluate its performance on downstream speech recognition tasks. Experimental results demonstrate that PoM achieves a competitive word error rate compared to full self-attention and other linear-complexity alternatives, offering an improved trade-off between performance and efficiency in time and memory.

Long-horizon agents face the challenge of growing context size during interaction with environment, which degrades the performance and stability. Existing methods typically introduce the external memory module and look up the relevant information from the stored memory, which prevents the model itself from proactively managing its memory content and aligning with the agent's overarching task objectives. To address these limitations, we propose the self-memory policy optimization algorithm (MemPO), which enables the agent (policy model) to autonomously summarize and manage their memory during interaction with environment. By improving the credit assignment mechanism based on memory effectiveness, the policy model can selectively retain crucial information, significantly reducing token consumption while preserving task performance. Extensive experiments and analyses confirm that MemPO achieves absolute F1 score gains of 25.98% over the base model and 7.1% over the previous SOTA baseline, while reducing token usage by 67.58% and 73.12%.

Existing mobile device control agents often perform poorly when solving complex tasks requiring long-horizon planning and precise operations, typically due to a lack of relevant task experience or unfamiliarity with skill execution. We propose K2-Agent, a hierarchical framework that models human-like cognition by separating and co-evolving declarative (knowing what) and procedural (knowing how) knowledge for planning and execution. K2-Agent's high level reasoner is bootstrapped from a single demonstration per task and runs a Summarize-Reflect-Locate-Revise (SRLR) loop to distill and iteratively refine task-level declarative knowledge through self-evolution. The low-level executor is trained with our curriculum-guided Group Relative Policy Optimization (C-GRPO), which (i) constructs a balanced sample pool using decoupled reward signals and (ii) employs dynamic demonstration injection to guide the model in autonomously generating successful trajectories for training. On the challenging AndroidWorld benchmark, K2-Agent achieves a 76.1% success rate using only raw screenshots and open-source backbones. Furthermore, K2-Agent shows powerful dual generalization: its high-level declarative knowledge transfers across diverse base models, while its low-level procedural skills achieve competitive performance on unseen tasks in ScreenSpot-v2 and Android-in-the-Wild (AitW).

Sustainability disclosure standards (e.g., GRI, SASB, TCFD, IFRS S2) are comprehensive yet lengthy, terminology-dense, and highly cross-referential, hindering structured analysis and downstream use. We present SSKG Hub (Sustainability Standards Knowledge Graph Hub), a research prototype and interactive web platform that transforms standards into auditable knowledge graphs (KGs) through an LLM-centered, expert-guided pipeline. The system integrates automatic standard identification, configurable chunking, standard-specific prompting, robust triple parsing, and provenance-aware Neo4j storage with fine-grained audit metadata. LLM extraction produces a provenance-linked Draft KG, which is reviewed, curated, and formally promoted to a Certified KG through meta-expert adjudication. A role-based governance framework covering read-only guest access, expert review and CRUD operations, meta-expert certification, and administrative oversight ensures traceability and accountability across draft and certified states. Beyond graph exploration and triple-level evidence tracing, SSKG Hub supports cross-KG fusion, KG-driven tasks, and dedicated modules for insights and curated resources. We validate the platform through a comprehensive expert-led KG review case study that demonstrates end-to-end curation and quality assurance. The web application is publicly available at www.sskg-hub.com.

Federated learning (FL) enables collaborative model training over distributed private data. However, sustaining open participation requires incentive mechanisms that compensate contributors for their resources and risks. Enabled by Web3 primitives, especially blockchains, recent FL proposals incorporate incentive mechanisms for open participation, yet most focus primarily on algorithmic design and overlook system-level challenges, including coordination efficiency, secure handling of model updates, and practical usability. We present FWeb3, a practical Web3-enabled FL framework for incentive-aware training in open environments. FWeb3 adopts a modular architecture that separates FL functions from Web3 support services, decoupling the off-chain training and data plane from on-chain settlement while preserving verifiable incentive execution. The framework supports pluggable aggregation and contribution evaluation methods and provides a browser-native DApp interface to lower the participation barrier. We evaluate FWeb3 in real-world settings and show that it supports end-to-end incentive-aware FL with transaction and data-transfer overheads of only 21.3% and 3.4% in WAN; FWeb3 also deploys from zero configuration in under 3 minutes and enables user onboarding in under 1 minute.

Real-world user requests to LLM agents are often underspecified. Agents must interact to acquire missing information and make correct downstream decisions. However, current multi-turn GRPO-based methods often rely on trajectory-level reward computation, which leads to credit assignment problems and insufficient advantage signals within rollout groups. A feasible approach is to identify valuable interaction turns at a fine granularity to drive more targeted learning. To address this, we introduce InfoPO (Information-Driven Policy Optimization), which frames multi-turn interaction as a process of active uncertainty reduction and computes an information-gain reward that credits turns whose feedback measurably changes the agent's subsequent action distribution compared to a masked-feedback counterfactual. It then combines this signal with task outcomes via an adaptive variance-gated fusion to identify information importance while maintaining task-oriented goal direction. Across diverse tasks, including intent clarification, collaborative coding, and tool-augmented decision making, InfoPO consistently outperforms prompting and multi-turn RL baselines. It also demonstrates robustness under user simulator shifts and generalizes effectively to environment-interactive tasks. Overall, InfoPO provides a principled and scalable mechanism for optimizing complex agent-user collaboration. Code is available at https://github.com/kfq20/InfoPO.

Dataset pruning has been widely studied for 2D images to remove redundancy and accelerate training, while particular pruning methods for 3D data remain largely unexplored. In this work, we study dataset pruning for 3D data, where its observed common long-tail class distribution nature make optimization under conventional evaluation metrics Overall Accuracy (OA) and Mean Accuracy (mAcc) inherently conflicting, and further make pruning particularly challenging. To address this, we formulate pruning as approximating the full-data expected risk with a weighted subset, which reveals two key errors: coverage error from insufficient representativeness and prior-mismatch bias from inconsistency between subset-induced class weights and target metrics. We propose representation-aware subset selection with per-class retention quotas for long-tail coverage, and prior-invariant teacher supervision using calibrated soft labels and embedding-geometry distillation. The retention quota also serves as a switch to control the OA-mAcc trade-off. Extensive experiments on 3D datasets show that our method can improve both metrics across multiple settings while adapting to different downstream preferences. Our code is available at https://github.com/XiaohanZhao123/3D-Dataset-Pruning.

Assessing the correctness of patches generated by Automated Program Repair (APR) is a major bottleneck. Manual validation is labor-intensive and limited: exact matching overlooks valid variants, while semantic inspection is subjective and hard to reproduce. Existing Automated Patch Correctness Assessment (APCA) often relies on opaque predictive models that treat each patch as novel, repeatedly re-assessing semantically redundant patches. Our analysis of a large corpus of tool-generated patches reveals a duality: about 39% of unique correct patches are syntactic clones, suggesting opportunities for automation, yet about 65% of bugs have multiple distinct correct fixes, making single-reference assessment insufficient. We present Historian, a framework that leverages Large Language Models to perform multi-reference comparisons against a knowledge base of historically validated patches, producing traceable, evidence-based verdicts while conservatively isolating novel cases as Unknown. In leave-one-tool-out evaluation, Historian achieves 95.0% coverage with 88.4% accuracy, reducing manual validation to 5% of patches. As an evidence-based pre-filter, enhancing the accuracy of standalone APCA tools by up to 21.8% and enabling a hybrid pipeline with 86.2% overall accuracy and 100% coverage. A longitudinal analysis of tool-generated patches (2020-2024) shows that redundancy in repair attempts is common, indicating that many patches repeatedly rediscover established ones and strengthening the sustainability of evidence-based APR assessment.

Large language models (LLMs) are increasingly adopted as the backbone of recommender systems. However, user-item interactions in real-world scenarios are non-stationary, making preference drift over time inevitable. Existing model update strategies mainly rely on global fine-tuning or pointwise editing, but they face two fundamental challenges: (i) imbalanced update granularity, where global updates perturb behaviors unrelated to the target while pointwise edits fail to capture broader preference shifts; (ii) unstable incremental updates, where repeated edits interfere with prior adaptations, leading to catastrophic forgetting and inconsistent recommendations. To address these issues, we propose Region-Aware Incremental Editing (RAIE), a plug-in framework that freezes the backbone model and performs region-level updates. RAIE first constructs semantically coherent preference regions via spherical k-means in the representation space. It then assigns incoming sequences to regions via confidence-aware gating and performs three localized edit operations - Update, Expand, and Add - to dynamically revise the affected region. Each region is equipped with a dedicated Low-Rank Adaptation (LoRA) module, which is trained only on the region's updated data. During inference, RAIE routes each user sequence to its corresponding region and activates the region-specific adapter for prediction. Experiments on two benchmark datasets under a time-sliced protocol that segments data into Set-up (S), Finetune (F), and Test (T) show that RAIE significantly outperforms state-of-the-art baselines while effectively mitigating forgetting. These results demonstrate that region-aware editing offers an accurate and scalable mechanism for continual adaptation in dynamic recommendation scenarios. Our code is available at https://github.com/fengaogao/RAIE.

We propose a retrodictive forecasting paradigm for time series: instead of predicting the future from the past, we identify the future that best explains the observed present via inverse MAP optimization over a Conditional Variational Autoencoder (CVAE). This conditioning is a statistical modeling choice for Bayesian inversion; it does not assert that future events cause past observations. The approach is theoretically grounded in an information-theoretic arrow-of-time measure: the symmetrized Kullback-Leibler divergence between forward and time-reversed trajectory ensembles provides both the conceptual rationale and an operational GO/NO-GO diagnostic for applicability. We implement the paradigm as MAP inference over an inverse CVAE with a learned RealNVP normalizing-flow prior and evaluate it on six time series cases: four synthetic processes with controlled temporal asymmetry and two ERA5 reanalysis datasets (wind speed and solar irradiance). The work makes four contributions: (i) a formal retrodictive inference formulation; (ii) an inverse CVAE architecture; (iii) a model-free irreversibility diagnostic; and (iv) a falsifiable validation protocol with four pre-specified predictions. All pre-specified predictions are empirically supported: the diagnostic correctly classifies all six cases; the learned flow prior improves over an isotropic Gaussian baseline on GO cases; the inverse MAP yields no spurious advantage on time-reversible dynamics; and on irreversible GO cases, it achieves competitive or superior RMSE relative to forward baselines, with a statistically significant 17.7% reduction over a forward MLP on ERA5 solar irradiance. These results provide a structured proof-of-concept that retrodictive forecasting can constitute a viable alternative to conventional forward prediction when statistical time-irreversibility is present and exploitable.

Multilingual falsehoods threaten information integrity worldwide, yet detection benchmarks remain confined to English or a few high-resource languages, leaving low-resource linguistic communities without robust defense tools. We introduce BLUFF, a comprehensive benchmark for detecting false and synthetic content, spanning 79 languages with over 202K samples, combining human-written fact-checked content (122K+ samples across 57 languages) and LLM-generated content (79K+ samples across 71 languages). BLUFF uniquely covers both high-resource "big-head" (20) and low-resource "long-tail" (59) languages, addressing critical gaps in multilingual research on detecting false and synthetic content. Our dataset features four content types (human-written, LLM-generated, LLM-translated, and hybrid human-LLM text), bidirectional translation (English$\leftrightarrow$X), 39 textual modification techniques (36 manipulation tactics for fake news, 3 AI-editing strategies for real news), and varying edit intensities generated using 19 diverse LLMs. We present AXL-CoI (Adversarial Cross-Lingual Agentic Chainof-Interactions), a novel multi-agentic framework for controlled fake/real news generation, paired with mPURIFY, a quality filtering pipeline ensuring dataset integrity. Experiments reveal state-of-theart detectors suffer up to 25.3% F1 degradation on low-resource versus high-resource languages. BLUFF provides the research community with a multilingual benchmark, extensive linguistic-oriented benchmark evaluation, comprehensive documentation, and opensource tools to advance equitable falsehood detection. Dataset and code are available at: https://jsl5710.github.io/BLUFF/

Semantic IDs (SIDs) are compact discrete representations derived from multimodal item features, serving as a unified abstraction for ID-based and generative recommendation. However, learning high-quality SIDs remains challenging due to two issues. (1) Collision problem: the quantized token space is prone to collisions, in which semantically distinct items are assigned identical or overly similar SID compositions, resulting in semantic entanglement. (2) Collision-signal heterogeneity: collisions are not uniformly harmful. Some reflect genuine conflicts between semantically unrelated items, while others stem from benign redundancy or systematic data effects. To address these challenges, we propose Qualification-Aware Semantic ID Learning (QuaSID), an end-to-end framework that learns collision-qualified SIDs by selectively repelling qualified conflict pairs and scaling the repulsion strength by collision severity. QuaSID consists of two mechanisms: Hamming-guided Margin Repulsion, which translates low-Hamming SID overlaps into explicit, severity-scaled geometric constraints on the encoder space; and Conflict-Aware Valid Pair Masking, which masks protocol-induced benign overlaps to denoise repulsion supervision. In addition, QuaSID incorporates a dual-tower contrastive objective to inject collaborative signals into tokenization. Experiments on public benchmarks and industrial data validate QuaSID. On public datasets, QuaSID consistently outperforms strong baselines, improving top-K ranking quality by 5.9% over the best baseline while increasing SID composition diversity. In an online A/B test on Kuaishou e-commerce with a 5% traffic split, QuaSID increases ranking GMV-S2 by 2.38% and improves completed orders on cold-start retrieval by up to 6.42%. Finally, we show that the proposed repulsion loss is plug-and-play and enhances a range of SID learning frameworks across datasets.

LiTS is a modular Python framework for LLM reasoning via tree search. It decomposes tree search into three reusable components (Policy, Transition, and RewardModel) that plug into algorithms like MCTS and BFS. A decorator-based registry enables domain experts to extend to new domains by registering components, and algorithmic researchers to implement custom search algorithms. We demonstrate composability on MATH500 (language reasoning), Crosswords (environment planning), and MapEval (tool use), showing that components and algorithms are orthogonal: components are reusable across algorithms within each task type, and algorithms work across all components and domains. We also report a mode-collapse finding: in infinite action spaces, LLM policy diversity (not reward quality) is the bottleneck for effective tree search. A demonstration video is available at https://youtu.be/nRGX43YrR3I. The package is released under the Apache 2.0 license at https://github.com/xinzhel/lits-llm, including installation instructions and runnable examples that enable users to reproduce the demonstrated workflows.

Large time series models (LTMs) have emerged as powerful tools for universal forecasting, yet they often struggle with the inherent diversity and nonstationarity of real-world time series data, leading to an unsatisfactory trade-off between forecasting accuracy and generalization. Rather than continually finetuning new LTM instances for each domain, we propose a data-centric framework, time-series adaptive transformation optimization (TATO), that enables a single frozen pre-trained LTM to adapt to diverse downstream domains through an optimally configured transformation pipeline. Specifically, TATO constructs three representative types of transformations, including context slicing, scale normalization, and outlier correction, to help LTMs better align with target domain characteristics. To ensure robustness, we incorporate carefully selected time series augmentations and a two-stage ranking mechanism that filters out pipelines underperforming on specific metrics. Extensive experiments on state-of-the-art LTMs and widely used datasets demonstrate that TATO consistently and significantly improves domain-adaptive forecasting performance, achieving a maximum reduction in MSE of 65.4\% and an average reduction of 13.6\%. Moreover, TATO is highly efficient, typically completing optimization in under 2 minutes, making it practical for real-world deployment. The source code is available at https://github.com/thulab/TATO.

Catastrophic forgetting, the tendency of neural networks to forget previously learned knowledge when learning new tasks, has been a major challenge in continual learning (CL). To tackle this challenge, CL methods have been proposed and shown to reduce forgetting. Furthermore, CL models deployed in mission-critical settings can benefit from uncertainty awareness by calibrating their predictions to reliably assess their confidences. However, existing uncertainty-aware continual learning methods suffer from high computational overhead and incompatibility with mainstream replay methods. To address this, we propose idempotent experience replay (IDER), a novel approach based on the idempotent property where repeated function applications yield the same output. Specifically, we first adapt the training loss to make model idempotent on current data streams. In addition, we introduce an idempotence distillation loss. We feed the output of the current model back into the old checkpoint and then minimize the distance between this reprocessed output and the original output of the current model. This yields a simple and effective new baseline for building reliable continual learners, which can be seamlessly integrated with other CL approaches. Extensive experiments on different CL benchmarks demonstrate that IDER consistently improves prediction reliability while simultaneously boosting accuracy and reducing forgetting. Our results suggest the potential of idempotence as a promising principle for deploying efficient and trustworthy continual learning systems in real-world applications.Our code is available at https://github.com/YutingLi0606/Idempotent-Continual-Learning.

Agentic systems augment large language models with external tools and iterative decision making, enabling complex tasks such as deep research, function calling, and coding. However, their long and intricate execution traces make failure diagnosis and root cause analysis extremely challenging. Manual inspection does not scale, while directly applying LLMs to raw traces is hindered by input length limits and unreliable reasoning. Focusing solely on final task outcomes further discards critical behavioral information required for accurate issue localization. To address these issues, we propose TraceSIR, a multi-agent framework for structured analysis and reporting of agentic execution traces. TraceSIR coordinates three specialized agents: (1) StructureAgent, which introduces a novel abstraction format, TraceFormat, to compress execution traces while preserving essential behavioral information; (2) InsightAgent, which performs fine-grained diagnosis including issue localization, root cause analysis, and optimization suggestions; (3) ReportAgent, which aggregates insights across task instances and generates comprehensive analysis reports. To evaluate TraceSIR, we construct TraceBench, covering three real-world agentic scenarios, and introduce ReportEval, an evaluation protocol for assessing the quality and usability of analysis reports aligned with industry needs. Experiments show that TraceSIR consistently produces coherent, informative, and actionable reports, significantly outperforming existing approaches across all evaluation dimensions. Our project and video are publicly available at https://github.com/SHU-XUN/TraceSIR.

Research in CDCR remains fragmented due to heterogeneous dataset formats, varying annotation standards, and the predominance of the CDCR definition as the event coreference resolution (ECR). To address these challenges, we introduce uCDCR, a unified dataset that consolidates diverse publicly available English CDCR corpora across various domains into a consistent format, which we analyze with standardized metrics and evaluation protocols. uCDCR incorporates both entity and event coreference, corrects known inconsistencies, and enriches datasets with missing attributes to facilitate reproducible research. We establish a cohesive framework for fair, interpretable, and cross-dataset analysis in CDCR and compare the datasets on their lexical properties, e.g., lexical composition of the annotated mentions, lexical diversity and ambiguity metrics, discuss the annotation rules and principles that lead to high lexical diversity, and examine how these metrics influence performance on the same-head-lemma baseline. Our dataset analysis shows that ECB+, the state-of-the-art benchmark for CDCR, has one of the lowest lexical diversities, and its CDCR complexity, measured by the same-head-lemma baseline, lies in the middle among all uCDCR datasets. Moreover, comparing document and mention distributions between ECB+ and uCDCR shows that using all uCDCR datasets for model training and evaluation will improve the generalizability of CDCR models. Finally, the almost identical performance on the same-head-lemma baseline, separately applied to events and entities, shows that resolving both types is a complex task and should not be steered toward ECR alone. The uCDCR dataset is available at https://huggingface.co/datasets/AnZhu/uCDCR, and the code for parsing, analyzing, and scoring the dataset is available at https://github.com/anastasia-zhukova/uCDCR.

The growing number of languages considered in multilingual NLP, including new datasets and tasks, poses challenges regarding properly and accurately reporting which languages are used and how. For example, datasets often use different language identifiers; some use BCP-47 (e.g. en_Latn), others use ISO 639-1 (en), and more linguistically oriented datasets use Glottocodes (stan1293). Mapping between identifiers is manageable for a few dozen languages, but becomes unscalable when dealing with thousands. We introduce QwanQwa, a light-weight Python toolkit for unified language metadata management. QQ integrates multiple language resources into a single interface, provides convenient normalization and mapping between language identifiers, and affords a graph-based structure that enables traversal across families, regions, writing systems, and other linguistic attributes. QQ serves both as (1) a simple "glue" library in multilingual NLP research to make working with many languages easier, and (2) as an intuitive way for exploring languages, such as finding related ones through shared scripts, regions or other metadata.

Multi-domain graph pre-training integrates knowledge from diverse domains to enhance performance in the target domains, which is crucial for building graph foundation models. Despite initial success, existing solutions often fall short of answering a fundamental question: how is knowledge integrated or transferred across domains? This theoretical limitation motivates us to rethink the consistency and transferability between model pre-training and domain adaptation. In this paper, we propose a fresh Riemannian geometry perspective, whose core idea is to merge any graph dataset into a unified, smooth Riemannian manifold, enabling a systematic understanding of knowledge integration and transfer. To achieve this, our key contribution is the theoretical establishment of neural manifold gluing, which first characterizes local geometry using an adaptive orthogonal frame and then "glues" the local pieces together into a coherent whole. Building on this theory, we present the GraphGlue framework, which supports batched pre-training with EMA prototyping and provides a transferability measure based on geometric consistence. Extensive experiments demonstrate its superior performance across diverse graph domains. Moreover, we empirically validated GraphGlue's geometric scaling law, showing that larger quantities of datasets improve model transferability by producing a smoother manifold. Codes are available at https://github.com/RiemannGraph/GraphGlue.

The rapid growth of biomedical literature and curated databases has made it increasingly difficult for researchers to systematically connect biomarker mechanisms to actionable drug combination hypotheses. We present AI Co-Scientist (CoDHy), an interactive, human-in-the-loop system for biomarker-guided drug combination hypothesis generation in cancer research. CoDHy integrates structured biomedical databases and unstructured literature evidence into a task-specific knowledge graph, which serves as the basis for graph-based reasoning and hypothesis construction. The system combines knowledge graph embeddings with agent-based reasoning to generate, validate, and rank candidate drug combinations, while explicitly grounding each hypothesis in retrievable evidence. Through a web-based interface, users can configure the scientific context, inspect intermediate results, and iteratively refine hypotheses, enabling transparent and researcher-steerable exploration rather than automated decision-making. We demonstrate CoDHy as a system for exploratory hypothesis generation and decision support in translational oncology, highlighting its design, interaction workflow, and practical use cases.

While music generation models have evolved to handle complex multimodal inputs mixing text, lyrics, and reference audio, evaluation mechanisms have lagged behind. In this paper, we bridge this critical gap by establishing a comprehensive ecosystem for music reward modeling under Compositional Multimodal Instruction (CMI), where the generated music may be conditioned on text descriptions, lyrics, and audio prompts. We first introduce CMI-Pref-Pseudo, a large-scale preference dataset comprising 110k pseudo-labeled samples, and CMI-Pref, a high-quality, human-annotated corpus tailored for fine-grained alignment tasks. To unify the evaluation landscape, we propose CMI-RewardBench, a unified benchmark that evaluates music reward models on heterogeneous samples across musicality, text-music alignment, and compositional instruction alignment. Leveraging these resources, we develop CMI reward models (CMI-RMs), a parameter-efficient reward model family capable of processing heterogeneous inputs. We evaluate their correlation with human judgments scores on musicality and alignment on CMI-Pref along with previous datasets. Further experiments demonstrate that CMI-RM not only correlates strongly with human judgments, but also enables effective inference-time scaling via top-k filtering. The necessary training data, benchmarks, and reward models are publicly available.

Student repetition in secondary education imposes significant resource burdens, particularly in resource-constrained contexts. Addressing this challenge, this study introduces a unified machine learning framework that simultaneously predicts pass/fail outcomes and continuous grades, a departure from prior research that treats classification and regression as separate tasks. Six models were evaluated: Logistic Regression, Decision Tree, and Random Forest for classification, and Linear Regression, Decision Tree Regressor, and Random Forest Regressor for regression, with hyperparameters optimized via exhaustive grid search. Using academic and demographic data from 4424 secondary school students, classification models achieved accuracies of up to 96%, while regression models attained a coefficient of determination of 0.70, surpassing baseline approaches. These results confirm the feasibility of early, data-driven identification of at-risk students and highlight the value of integrating dual-task prediction for more comprehensive insights. By enabling timely, personalized interventions, the framework offers a practical pathway to reducing grade repetition and optimizing resource allocation.

Multi-subject image generation requires seamlessly harmonizing multiple reference identities within a coherent scene. However, existing methods relying on rigid spatial masks or localized attention often struggle with the "stability-plasticity dilemma," particularly failing in tasks that require complex structural deformations, such as identity-preserving age transformation. To address this, we present IdGlow, a mask-free, progressive two-stage framework built upon Flow Matching diffusion models. In the supervised fine-tuning (SFT) stage, we introduce task-adaptive timestep scheduling aligned with diffusion generative dynamics: a linear decay schedule that progressively relaxes constraints for natural group composition, and a temporal gating mechanism that concentrates identity injection within a critical semantic window, successfully preserving adult facial semantics without overriding child-like anatomical structures. To resolve attribute leakage and semantic ambiguity without explicit layout inputs, we further integrate a badcase-driven Vision-Language Model (VLM) for precise, context-aware prompt synthesis. In the second stage, we design a Fine-Grained Group-Level Direct Preference Optimization (DPO) with a weighted margin formulation to simultaneously eliminate multi-subject artifacts, elevate texture harmony, and recalibrate identity fidelity towards real-world distributions. Extensive experiments on two challenging benchmarks -- direct multi-person fusion and age-transformed group generation -- demonstrate that IdGlow fundamentally mitigates the stability-plasticity conflict, achieving a superior Pareto balance between state-of-the-art facial fidelity and commercial-grade aesthetic quality.

Detecting out-of-distribution (OOD) graphs is crucial for ensuring the safety and reliability of Graph Neural Networks. In unsupervised graph-level OOD detection, models are typically trained using only in-distribution (ID) data, resulting in incomplete feature space characterization and weak decision boundaries. Although synthesizing outliers offers a promising solution, existing approaches rely on fixed, non-adaptive sampling heuristics (e.g., distance- or density-based), limiting their ability to explore informative OOD regions. We propose a Policy-Guided Outlier Synthesis (PGOS) framework that replaces static heuristics with a learned exploration strategy. Specifically, PGOS trains a reinforcement learning agent to navigate low-density regions in a structured latent space and sample representations that most effectively refine the OOD decision boundary. These representations are then decoded into high-quality pseudo-OOD graphs to improve detector robustness. Extensive experiments demonstrate that PGOS achieves state-of-the-art performance on multiple graph OOD and anomaly detection benchmarks.

AI code agents excel at isolated tasks yet struggle with complex, multi-file software engineering requiring understanding of how dozens of modules relate. We hypothesize these failures stem from inability to construct, maintain, and update coherent architectural beliefs during codebase exploration. We introduce Theory of Code Space (ToCS), a benchmark that evaluates this capability by placing agents in procedurally generated codebases under partial observability, requiring them to build structured belief states over module dependencies, cross-cutting invariants, and design intent. The framework features: (1) a procedural codebase generator producing medium-complexity Python projects with four typed edge categories reflecting different discovery methods -- from syntactic imports to config-driven dynamic wiring -- with planted architectural constraints and verified ground truth; (2) a partial observability harness where agents explore under a budget; and (3) periodic belief probing via structured JSON, producing a time-series of architectural understanding. We decompose the Active-Passive Gap from spatial reasoning benchmarks into selection and decision components, and introduce Architectural Constraint Discovery as a code-specific evaluation dimension. Preliminary experiments with four rule-based baselines and five frontier LLM agents from three providers validate discriminative power: methods span a wide performance range (F1 from 0.129 to 0.646), LLM agents discover semantic edge types invisible to all baselines, yet weaker models score below simple heuristics -- revealing that belief externalization, faithfully serializing internal understanding into structured JSON, is itself a non-trivial capability and a first-order confounder in belief-probing benchmarks. Open-source toolkit: https://github.com/che-shr-cat/tocs

Hypergraphs are the natural description of higher-order interactions among objects, widely applied in social network analysis, cross-modal retrieval, etc. Hypergraph Neural Networks (HGNNs) have become the dominant solution for learning on hypergraphs. Traditional HGNNs are extended from message passing graph neural networks, following the homophily assumption, and thus struggle with the prevalent heterophilic hypergraphs that call for long-range dependence modeling. In this paper, we achieve heterophily-agnostic message passing through the lens of Riemannian geometry. The key insight lies in the connection between oversquashing and hypergraph bottleneck within the framework of Riemannian manifold heat flow. Building on this, we propose the novel idea of locally adapting the bottlenecks of different subhypergraphs. The core innovation of the proposed mechanism is the design of an adaptive local (heat) exchanger. Specifically, it captures the rich long-range dependencies via the Robin condition, and preserves the representation distinguishability via source terms, thereby enabling heterophily-agnostic message passing with theoretical guarantees. Based on this theoretical foundation, we present a novel Heat-Exchanger with Adaptive Locality for Hypergraph Neural Network (HealHGNN), designed as a node-hyperedge bidirectional systems with linear complexity in the number of nodes and hyperedges. Extensive experiments on both homophilic and heterophilic cases show that HealHGNN achieves the state-of-the-art performance.

Vision-Language-Action (VLA) models achieve over 95% success on standard benchmarks. However, through systematic experiments, we find that current state-of-the-art VLA models largely ignore language instructions. Prior work lacks: (1) systematic semantic perturbation diagnostics, (2) a benchmark that forces language understanding by design, and (3) linguistically diverse training data. This paper constructs the LangGap benchmark, based on a four-dimensional semantic perturbation method -- varying instruction semantics while keeping the tabletop layout fixed -- revealing language understanding deficits in π0.5. Existing benchmarks like LIBERO assign only one task per layout, underutilizing available objects and target locations; LangGap fully diversifies pick-and-place tasks under identical layouts, forcing models to truly understand language. Experiments show that targeted data augmentation can partially close the language gap -- success rate improves from 0% to 90% with single-task training, and 0% to 28% with multi-task training. However, as semantic diversity of extended tasks increases, model learning capacity proves severely insufficient; even trained tasks perform poorly. This reveals a fundamental challenge for VLA models in understanding diverse language instructions -- precisely the long-term value of LangGap.

As artificial intelligence (AI) is increasingly deployed across domains, ensuring fairness has become a core challenge. However, the field faces a "Tower of Babel'' dilemma: fairness metrics abound, yet their underlying philosophical assumptions often conflict, hindering unified paradigms-particularly in unified Multimodal Large Language Models (UMLLMs), where biases propagate systemically across tasks. To address this, we introduce the IRIS Benchmark, to our knowledge the first benchmark designed to synchronously evaluate the fairness of both understanding and generation tasks in UMLLMs. Enabled by our demographic classifier, ARES, and four supporting large-scale datasets, the benchmark is designed to normalize and aggregate arbitrary metrics into a high-dimensional "fairness space'', integrating 60 granular metrics across three dimensions-Ideal Fairness, Real-world Fidelity, and Bias Inertia & Steerability (IRIS). Through this benchmark, our evaluation of leading UMLLMs uncovers systemic phenomena such as the "generation gap'', individual inconsistencies like "personality splits'', and the "counter-stereotype reward'', while offering diagnostics to guide the optimization of their fairness capabilities. With its novel and extensible framework, the IRIS benchmark is capable of integrating evolving fairness metrics, ultimately helping to resolve the "Tower of Babel'' impasse. Project Page: https://iris-benchmark-web.vercel.app/

Visual autoregressive (VAR) models have recently emerged as a promising alternative for image generation, offering stable training, non-iterative inference, and high-fidelity synthesis through next-scale prediction. This encourages the exploration of VAR for image super-resolution (ISR), yet its application remains underexplored and faces two critical challenges: locality-biased attention, which fragments spatial structures, and residual-only supervision, which accumulates errors across scales, severely compromises global consistency of reconstructed images. To address these issues, we propose AlignVAR, a globally consistent visual autoregressive framework tailored for ISR, featuring two key components: (1) Spatial Consistency Autoregression (SCA), which applies an adaptive mask to reweight attention toward structurally correlated regions, thereby mitigating excessive locality and enhancing long-range dependencies; and (2) Hierarchical Consistency Constraint (HCC), which augments residual learning with full reconstruction supervision at each scale, exposing accumulated deviations early and stabilizing the coarse-to-fine refinement process. Extensive experiments demonstrate that AlignVAR consistently enhances structural coherence and perceptual fidelity over existing generative methods, while delivering over 10x faster inference with nearly 50% fewer parameters than leading diffusion-based approaches, establishing a new paradigm for efficient ISR.

Efficient representation learning is essential for optimal information storage and classification. However, it is frequently overlooked in artificial neural networks (ANNs). This neglect results in networks that can become overparameterized by factors of up to 13, increasing redundancy and energy consumption. As the demand for large language models (LLMs) and their scale increase, these issues are further highlighted, raising significant ethical and environmental concerns. We analyze our previously developed biologically inspired learning rule using information-theoretic concepts, evaluating its efficiency on the MNIST classification task. The proposed rule, which emulates the brain's structural plasticity, naturally prevents overparameterization by optimizing synaptic usage and retaining only the essential number of synapses. Furthermore, it outperforms backpropagation (BP) in terms of efficiency and storage capacity. It also eliminates the need for pre-optimization of network architecture, enhances adaptability, and reflects the brain's ability to reserve 'space' for new memories. This approach advances scalable and energy-efficient AI and provides a promising framework for developing brain-inspired models that optimize resource allocation and adaptability.

Evaluating machine unlearning remains challenging, as existing methods typically require retraining reference models or performing membership inference attacks, both of which rely on prior access to training configuration or supervision labels, making them impractical in realistic scenarios. Motivated by the fact that most unlearning algorithms remove a small, random subset of the training data, we propose a subset-level evaluation framework based on statistical independence. Specifically, we design a tailored use of the Hilbert-Schmidt Independence Criterion to assess whether the model outputs on a given subset exhibit statistical dependence, without requiring model retraining or auxiliary classifiers. Our method provides a simple, standalone evaluation procedure that aligns with unlearning workflows. Extensive experiments demonstrate that our approach reliably distinguishes in-training from out-of-training subsets and clearly differentiates unlearning effectiveness, even when existing evaluations fall short.

Recent advances in video generation have opened new avenues for macroscopic simulation of complex dynamic systems, but their application to microscopic phenomena remains largely unexplored. Microscale simulation holds great promise for biomedical applications such as drug discovery, organ-on-chip systems, and disease mechanism studies, while also showing potential in education and interactive visualization. In this work, we introduce MicroWorldBench, a multi-level rubric-based benchmark for microscale simulation tasks. MicroWorldBench enables systematic, rubric-based evaluation through 459 unique expert-annotated criteria spanning multiple microscale simulation task (e.g., organ-level processes, cellular dynamics, and subcellular molecular interactions) and evaluation dimensions (e.g., scientific fidelity, visual quality, instruction following). MicroWorldBench reveals that current SOTA video generation models fail in microscale simulation, showing violations of physical laws, temporal inconsistency, and misalignment with expert criteria. To address these limitations, we construct MicroSim-10K, a high-quality, expert-verified simulation dataset. Leveraging this dataset, we train MicroVerse, a video generation model tailored for microscale simulation. MicroVerse can accurately reproduce complex microscale mechanism. Our work first introduce the concept of Micro-World Simulation and present a proof of concept, paving the way for applications in biology, education, and scientific visualization. Our work demonstrates the potential of educational microscale simulations of biological mechanisms. Our data and code are publicly available at https://github.com/FreedomIntelligence/MicroVerse

While Large Language Models (LLMs) have demonstrated proficiency in Deep Research or Wide Search, their capacity to solve highly complex questions-those requiring long-horizon planning, massive evidence gathering, and synthesis across heterogeneous sources-remains largely unexplored. We introduce Super Research, a task for complex autonomous research tasks that integrates (i) structured decomposition into a research plan, (ii) super wide retrieval for diverse perspectives, and (iii) super deep investigation to resolve uncertainties through iterative queries. To evaluate this capability, we curated a benchmark of 300 expert-written questions across diverse domains, each requiring up to 100+ retrieval steps and 1,000+ web pages to reconcile conflicting evidence. Super Research produces verifiable reports with fine-grained citations and intermediate artifacts (e.g., outlines and tables) to ensure traceable reasoning. Furthermore, we present a graph-anchored auditing protocol that evaluates Super Research along five dimensions: Coverage, Logical Consistency, Report Utility, Objectivity and Citation Health. While super-complex questions may be infrequent in standard applications, Super Research serves as a critical ceiling evaluation and stress test for LLM capabilities. A model's proficiency within Super Research acts as a powerful proxy for its general research competence; success here suggests the robustness necessary to navigate nearly any subordinate research task. Leaderboard is available at: https://cnsdqd-dyb.github.io/Super-Research-Benchmark/

Federated Learning (FL) is a popular distributed learning paradigm to break down data silo. Traditional FL approaches largely rely on gradient-based updates, facing significant issues about heterogeneity, scalability, convergence, and overhead, etc. Recently, some analytic-learning-based work has attempted to handle these issues by eliminating gradient-based updates via analytical (i.e., closed-form) solutions. Despite achieving superior invariance to data heterogeneity, these approaches are fundamentally limited by their single-layer linear model with a frozen pre-trained backbone. As a result, they can only achieve suboptimal performance due to their lack of representation learning capabilities. In this paper, to enable representable analytic models while preserving the ideal invariance to data heterogeneity for FL, we propose our Deep Analytic Federated Learning approach, named DeepAFL. Drawing inspiration from the great success of ResNet in gradient-based learning, we design gradient-free residual blocks in our DeepAFL with analytical solutions. We introduce an efficient layer-wise protocol for training our deep analytic models layer by layer in FL through least squares. Both theoretical analyses and empirical evaluations validate our DeepAFL's superior performance with its dual advantages in heterogeneity invariance and representation learning, outperforming state-of-the-art baselines by up to 5.68%-8.42% across three benchmark datasets.

Long chain-of-thought~(CoT) has become a dominant paradigm for enhancing the reasoning capability of large reasoning models~(LRMs); however, the performance gains often come with a substantial increase in reasoning budget. Recent studies show that existing CoT paradigms tend to induce systematic overthinking, unnecessarily coupling reasoning capability with reasoning cost. Most prior approaches reduce token usage through post hoc techniques such as token compression, truncation, or length penalties, without explicitly addressing the core mechanisms of reasoning. We propose \textbf{Draft-Thinking}, which guides models to first learn a concise \textit{draft-style} reasoning structure that retains only the critical reasoning steps. Through a \textit{progressive curriculum learning}, the model stably internalizes this efficient reasoning pattern as its capability scales. Moreover, Draft-Thinking introduces adaptive prompting, which elevates reasoning depth to a flexible, model-selectable behavior. Extensive experiments demonstrate that Draft-Thinking substantially reduces reasoning budget while largely preserving reasoning performance; for example, on MATH500, it achieves an 82.6\% reduction in reasoning budget at the cost of only a 2.6\% performance drop.

Symbolic music generation is a challenging task in multimedia generation, involving long sequences with hierarchical temporal structures, long-range dependencies, and fine-grained local details. Though recent diffusion-based models produce high quality generations, they tend to suffer from high training and inference costs with long symbolic sequences due to iterative denoising and sequence-length-related costs. To deal with such problem, we put forth a diffusing strategy named SMDIM to combine efficient global structure construction and light local refinement. SMDIM uses structured state space models to capture long range musical context at near linear cost, and selectively refines local musical details via a hybrid refinement scheme. Experiments performed on a wide range of symbolic music datasets which encompass various Western classical music, popular music and traditional folk music show that the SMDIM model outperforms the other state-of-the-art approaches on both the generation quality and the computational efficiency, and it has robust generalization to underexplored musical styles. These results show that SMDIM offers a principled solution for long-sequence symbolic music generation, including associated attributes that accompany the sequences. We provide a project webpage with audio examples and supplementary materials at https://3328702107.github.io/smdim-music/.

Progress in software-engineering agents is increasingly constrained by the scarcity of executable, scalable, and realistic data for training and evaluation. This scarcity stems from three fundamental challenges in existing pipelines: environments are brittle and difficult to reproduce across languages; synthesizing realistic, system-level bugs at scale is computationally expensive; and existing data predominantly consists of short-horizon repairs, failing to capture long-horizon competencies like architectural consistency. We introduce \textbf{SWE-Hub}, an end-to-end system that operationalizes the data factory abstraction by unifying environment automation, scalable synthesis, and diverse task generation into a coherent production stack. At its foundation, the \textbf{Env Agent} establishes a shared execution substrate by automatically converting raw repository snapshots into reproducible, multi-language container environments with standardized interfaces. Built upon this substrate, \textbf{SWE-Scale} engine addresses the need for high-throughput generation, combining cross-language code analysis with cluster-scale validation to synthesize massive volumes of localized bug-fix instances. \textbf{Bug Agent} generates high-fidelity repair tasks by synthesizing system-level regressions involving cross-module dependencies, paired with user-like issue reports that describe observable symptoms rather than root causes. Finally, \textbf{SWE-Architect} expands the task scope from repair to creation by translating natural-language requirements into repository-scale build-a-repo tasks. By integrating these components, SWE-Hub establishes a unified production pipeline capable of continuously delivering executable tasks across the entire software engineering lifecycle.

Adapting pretrained multi-modal models to evolving test-time distributions, known as multi-modal test-time adaptation, presents a significant challenge. Existing methods frequently encounter negative transfer in the unbiased modality and catastrophic forgetting in the biased modality. To address these challenges, we propose Decoupling Adaptation for Stability and Plasticity (DASP), a novel diagnose-then-mitigate framework. Our analysis reveals a critical discrepancy within the unified latent space: the biased modality exhibits substantially higher interdimensional redundancy (i.e., strong correlations across feature dimensions) compared to the unbiased modality. Leveraging this insight, DASP identifies the biased modality and implements an asymmetric adaptation strategy. This strategy employs a decoupled architecture where each modality-specific adapter is divided into stable and plastic components. The asymmetric mechanism works as follows: for the biased modality, which requires plasticity, the plastic component is activated and updated to capture domain-specific information, while the stable component remains fixed. Conversely, for the unbiased modality, which requires stability, the plastic component is bypassed, and the stable component is updated using KL regularization to prevent negative transfer. This asymmetric design enables the model to adapt flexibly to new domains while preserving generalizable knowledge. Comprehensive evaluations on diverse multi-modal benchmarks demonstrate that DASP significantly outperforms state-of-the-art methods.

Large language models (LLMs) achieve remarkable performance on diverse downstream and domain-specific tasks via parameter-efficient fine-tuning (PEFT). However, existing PEFT methods, particularly MoE-LoRA architectures, suffer from limited parameter efficiency and coarse-grained adaptation due to the proliferation of LoRA experts and instance-level routing. To address these issues, we propose Core Space Mixture of LoRA (\textbf{CoMoL}), a novel MoE-LoRA framework that incorporates expert diversity, parameter efficiency, and fine-grained adaptation. Specifically, CoMoL introduces two key components: core space experts and core space routing. Core space experts store each expert in a compact core matrix, preserving diversity while controlling parameter growth. Core space routing dynamically selects and activates the appropriate core experts for each token, enabling fine-grained, input-adaptive routing. Activated core experts are then merged via a soft-merging strategy into a single core expert, which is combined with a shared LoRA to form a specialized LoRA module. Besides, the routing network is projected into the same low-rank space as the LoRA matrices, further reducing parameter overhead without compromising expressiveness. Extensive experiments demonstrate that CoMoL retains the adaptability of MoE-LoRA architectures while achieving parameter efficiency comparable to standard LoRA, consistently outperforming existing methods across multiple tasks.

TopoEdge is a topology-grounded, edge-deployable framework for end-to-end software-defined networking (SDN) configuration generation and repair, motivated by the brittleness of configuration artefacts under topology variation and by strict operational constraints on latency, privacy, and on-site execution. TopoEdge represents each target topology as a router-level graph and embeds it using a contrastively trained graph neural network (GNN), enabling nearest-neighbour retrieval of a verified reference configuration paired with an executable Python driver (a Topotest/pytest test script that orchestrates the emulated network and checks protocol assertions). The target topology, retrieved reference topology, and reference driver are assembled into a topology-grounded retrieval-augmented generation context (TopoRAG), which grounds a distributed, execution-centric generate--verify--repair loop coordinated by a central controller and realised by three role-specialised agents: (i) a Planning agent that produces a topology-consistent configuration plan and a per-device skeleton; (ii) a Generation agent that materialises executable configuration artefacts, including device configurations and the driver; and (iii) a Verification agent that runs the FRRouting Topotest/pytest harness, compresses failures into a compact trace, and emits localised patch directives for iterative repair.

Molecule representation learning is crucial for understanding and predicting molecular properties. However, conventional atom-centric models, which treat chemical bonds merely as pairwise interactions, often overlook complex bond-level phenomena like resonance and stereoselectivity. This oversight limits their predictive accuracy for nuanced chemical behaviors. To address this limitation, we introduce \textbf{DeMol}, a dual-graph framework whose architecture is motivated by a rigorous information-theoretic analysis demonstrating the information gain from a bond-centric perspective. DeMol explicitly models molecules through parallel atom-centric and bond-centric channels. These are synergistically fused by multi-scale Double-Helix Blocks designed to learn intricate atom-atom, atom-bond, and bond-bond interactions. The framework's geometric consistency is further enhanced by a regularization term based on covalent radii to enforce chemically plausible structures. Comprehensive evaluations on diverse benchmarks, including PCQM4Mv2, OC20 IS2RE, QM9, and MoleculeNet, show that DeMol establishes a new state-of-the-art, outperforming existing methods. These results confirm the superiority of explicitly modelling bond information and interactions, paving the way for more robust and accurate molecular machine learning.

Neural contextual bandits are vulnerable to adversarial attacks, where subtle perturbations to rewards, actions, or contexts induce suboptimal decisions. We introduce AdvBandit, a black-box adaptive attack that formulates context poisoning as a continuous-armed bandit problem, enabling the attacker to jointly learn and exploit the victim's evolving policy. The attacker requires no access to the victim's internal parameters, reward function, or gradient information; instead, it constructs a surrogate model using a maximum-entropy inverse reinforcement learning module from observed context-action pairs and optimizes perturbations against this surrogate using projected gradient descent. An upper confidence bound-aware Gaussian process guides arm selection. An attack-budget control mechanism is also introduced to limit detection risk and overhead. We provide theoretical guarantees, including sublinear attacker regret and lower bounds on victim regret linear in the number of attacks. Experiments on three real-world datasets (Yelp, MovieLens, and Disin) against various victim contextual bandits demonstrate that our attack model achieves higher cumulative victim regret than state-of-the-art baselines.

Multimodal Large Language Models (MLLMs) have achieved remarkable performance but remain vulnerable to jailbreak attacks that can induce harmful content and undermine their secure deployment. Previous studies have shown that introducing additional inference steps, which disrupt security attention, can make MLLMs more susceptible to being misled into generating malicious content. However, these methods rely on single-image masking or isolated visual cues, which only modestly extend reasoning paths and thus achieve limited effectiveness, particularly against strongly aligned commercial closed-source models. To address this problem, in this paper, we propose Multi-Image Dispersion and Semantic Reconstruction (MIDAS), a multimodal jailbreak framework that decomposes harmful semantics into risk-bearing subunits, disperses them across multiple visual clues, and leverages cross-image reasoning to gradually reconstruct the malicious intent, thereby bypassing existing safety mechanisms. The proposed MIDAS enforces longer and more structured multi-image chained reasoning, substantially increases the model's reliance on visual cues while delaying the exposure of malicious semantics and significantly reducing the model's security attention, thereby improving the performance of jailbreak against advanced MLLMs. Extensive experiments across different datasets and MLLMs demonstrate that the proposed MIDAS outperforms state-of-the-art jailbreak attacks for MLLMs and achieves an average attack success rate of 81.46% across 4 closed-source MLLMs. Our code is available at this [link](https://github.com/Winnie-Lian/MIDAS).

The Transformer-based Whisper model has achieved state-of-the-art performance in Automatic Speech Recognition (ASR). However, its Multi-Head Attention (MHA) mechanism results in significant GPU memory consumption due to the linearly growing Key-Value (KV) cache usage, which is problematic for many applications especially with long-form audio. To address this, we introduce Whisper-MLA, a novel architecture that incorporates Multi-Head Latent Attention (MLA) into the Whisper model. Specifically, we adapt MLA for Whisper's absolute positional embeddings and systematically investigate its application across encoder self-attention, decoder self-attention, and cross-attention modules. Empirical results indicate that applying MLA exclusively to decoder self-attention yields the desired balance between performance and memory efficiency. Our proposed approach allows conversion of a pretrained Whisper model to Whisper-MLA with minimal fine-tuning. Extensive experiments on the LibriSpeech benchmark validate the effectiveness of this conversion, demonstrating that Whisper-MLA reduces the KV cache size by up to 87.5% while maintaining competitive accuracy.

In operating rooms (OR), world-scale multi-view 3D tracking supports downstream applications such as surgeon behavior recognition, where physically meaningful quantities such as distances and motion statistics must be measured in meters. However, real clinical deployments rarely satisfy the geometric prerequisites for stable multi-view fusion and tracking: camera calibration and RGB-D registration are always unreliable, leading to cross-view geometric inconsistency that produces "ghosting" during fusion and degrades 3D trajectories in a shared OR coordinate frame. To address this, we introduce Geometry OR Tracker, a two-stage pipeline that first rectifies imprecise calibration into a scaleconsistent and geometrically consistent camera setup with a single global scale via a Multi-view Metric Geometry Rectification module, and then performs Occlusion-Robust 3D Point Tracking directly in the unified OR world frame. On the MM-OR benchmark, improved geometric consistency translates into tracking gains: our rectification front-end reduces cross-view depth disagreement by more than 30$\times$ compared to raw calibration. Ablation studies further demonstrate the relationship between calibration quality and tracking accuracy, showing that improved geometric consistency yields stronger world-frame tracking.

Evaluating persona-aligned empathy in LLM-based dialogue agents remains challenging. User states are latent, feedback is sparse and difficult to verify in situ, and seemingly supportive turns can still accumulate into trajectories that drift from persona-specific needs. We introduce EMPA, a process-oriented framework that evaluates persona-aligned support as sustained intervention rather than isolated replies. EMPA distills real interactions into controllable, psychologically grounded scenarios, couples them with an open-ended multi-agent sandbox that exposes strategic adaptation and failure modes, and scores trajectories in a latent psychological space by directional alignment, cumulative impact, and stability. The resulting signals and metrics support reproducible comparison and optimization of long-horizon empathic behavior, and they extend to other agent settings shaped by latent dynamics and weak, hard-to-verify feedback.

GPU architectural simulation is orders of magnitude slower than native execution, necessitating workload sampling for practical speedups. Existing methods rely on hand-crafted features with limited expressiveness, yielding either aggressive sampling with high errors or conservative sampling with constrained speedups. To address these issues, we propose GCL-Sampler, a sampling framework that leverages Relational Graph Convolutional Networks with contrastive learning to automatically discover high-dimensional kernel similarities from trace graphs. By encoding instruction sequences and data dependencies into graph embeddings, GCL-Sampler captures rich structural and semantic properties of program execution, enabling both high fidelity and substantial speedup. Evaluations on extensive benchmarks show that GCL-Sampler achieves 258.94x average speedup against full workload with 0.37% error, outperforming state-of-the-art methods, PKA (129.23x, 20.90%), Sieve (94.90x, 4.10%) and STEM+ROOT (56.57x, 0.38%).

Using Multimodal Large Language Models (MLLMs) as judges to achieve precise and consistent evaluations has gradually become an emerging paradigm across various domains. Evaluating the capability and reliability of MLLM-as-a-judge systems is therefore essential for ensuring trustworthy assessment. Existing judge benchmarks categorize samples by task types but fail to capture the fundamental judgment capabilities required for reliable evaluation. In this work, we introduce M-JudgeBench, a ten-dimensional capability-oriented benchmark designed to comprehensively assess the judgment abilities of MLLMs. Our benchmark decomposes evaluation into pairwise Chain-of-Thought (CoT) comparison, length bias avoidance, and process error detection tasks, jointly covering ten fine-grained subtasks. This design enables diagnosis of model reliability across reasoning styles, response lengths, and cross-model variations. Systematic evaluation uncovers the systematic weaknesses in existing MLLM-as-a-judge systems. To address this issue, we further propose Judge-MCTS, a data construction framework generating pairwise reasoning trajectories with various correctness and length. Using Judge-MCTS, we construct an MCTS-augmented dataset and train M-Judger, a series of strong judge models. Extensive experiments demonstrate the superiority of M-Judger on existing judge benchmarks as well as M-JudgeBench. Overall, our work establishes a more principled foundation for evaluating MLLM-as-a-judge through M-JudgeBench and Judge-MCTS framework, paving the way for future research on judge model evaluation and capability-driven judge training.

Generative foundation models are increasingly scaled in both width and depth, posing significant challenges for stable feature learning and reliable hyperparameter (HP) transfer across model sizes. While maximal update parameterization ($μ$P) has provided a principled solution to both problems for width scaling, existing extensions to the joint width-depth scaling regime remain fragmented, architecture- and optimizer-specific, and often rely on technically involved theories. In this work, we develop a simple and unified spectral framework for $μ$P under joint width-depth scaling. Considering residual networks of varying block depths, we first introduce a spectral $μ$P condition that precisely characterizes how the norms of weights and their per-step updates should scale with width and depth, unifying previously disparate $μ$P formulations as special cases. Building on this condition, we then derive a general recipe for implementing $μ$P across a broad class of optimizers by mapping the spectral constraints to concrete HP parameterizations. This approach not only recovers existing $μ$P formulations (e.g., for SGD and AdamW) but also naturally extends to a wider range of optimizers. Finally, experiments on GPT-2 style language models demonstrate that the proposed spectral $μ$P condition preserves stable feature learning and enables robust HP transfer under width-depth scaling.

The evolution of Large Language Models (LLMs) from static instruction-followers to autonomous agents necessitates operating within complex, stateful environments to achieve precise state-transition objectives. However, this paradigm is bottlenecked by data scarcity, as existing tool-centric reverse-synthesis pipelines fail to capture the rigorous logic of real-world applications. We introduce \textbf{LOGIGEN}, a logic-driven framework that synthesizes verifiable training data based on three core pillars: \textbf{Hard-Compiled Policy Grounding}, \textbf{Logic-Driven Forward Synthesis}, and \textbf{Deterministic State Verification}. Specifically, a Triple-Agent Orchestration is employed: the \textbf{Architect} compiles natural-language policy into database constraints to enforce hard rules; the \textbf{Set Designer} initializes boundary-adjacent states to trigger critical policy conflicts; and the \textbf{Explorer} searches this environment to discover causal solution paths. This framework yields a dataset of 20,000 complex tasks across 8 domains, where validity is strictly guaranteed by checking exact state equivalence. Furthermore, we propose a verification-based training protocol where Supervised Fine-Tuning (SFT) on verifiable trajectories establishes compliance with hard-compiled policy, while Reinforcement Learning (RL) guided by dense state-rewards refines long-horizon goal achievement. On $τ^2$-Bench, LOGIGEN-32B(RL) achieves a \textbf{79.5\% success rate}, substantially outperforming the base model (40.7\%). These results demonstrate that logic-driven synthesis combined with verification-based training effectively constructs the causally valid trajectories needed for next-generation agents.

Large language models (LLMs) have become essential tools in software development, widely used for requirements engineering, code generation and review tasks. Software engineers often rely on LLMs to verify if code implementation satisfy task requirements, thereby ensuring code robustness and accuracy. However, it remains unclear whether LLMs can reliably determine code against the given task descriptions, which is usually in a form of natural language specifications. In this paper, we uncover a systematic failure of LLMs in matching code to natural language requirements. Specifically, with widely adopted benchmarks and unified prompts design, we demonstrate that LLMs frequently misclassify correct code implementation as non-compliant or defective. Surprisingly, we find that more detailed prompt design, particularly with those requiring explanations and proposed corrections, leads to higher misjudgment rates, highlighting critical reliability issues for LLM-based code assistants. We further analyze the mechanisms driving these failures and evaluate the reliability of rationale-required judgments. Building on these findings, we propose a Fix-guided Verification Filter that treats the model proposed fix as executable counterfactual evidence, and validates the original and revised implementations using benchmark tests and spec-constrained augmented tests. Our results expose previously under-explored limitations in LLM-based code review capabilities, and provide practical guidance for integrating LLM-based reviewers with safeguards in automated review and development pipelines.

Learned indexes are a class of index data structures that enable fast search by approximating the cumulative distribution function (CDF) using machine learning models (Kraska et al., SIGMOD'18). However, recent studies have shown that learned indexes are vulnerable to poisoning attacks, where injecting a small number of poison keys into the training data can significantly degrade model accuracy and reduce index performance (Kornaropoulos et al., SIGMOD'22). In this work, we provide a rigorous theoretical analysis of poisoning attacks targeting linear regression models over CDFs, one of the most basic regression models and a core component in many learned indexes. Our main contributions are as follows: (i) We present a theoretical proof characterizing the optimal single-point poisoning attack and show that the existing method yields the optimal attack. (ii) We show that in multi-point attacks, the existing greedy approach is not always optimal, and we rigorously derive the key properties that an optimal attack should satisfy. (iii) We propose a method to compute an upper bound of the multi-point poisoning attack's impact and empirically demonstrate that the loss under the greedy approach is often close to this bound. Our study deepens the theoretical understanding of attack strategies against linear regression models on CDFs and provides a foundation for the theoretical evaluation of attacks and defenses on learned indexes.

Autonomous agents are increasingly entrusted with complex, long-horizon tasks, ranging from mathematical reasoning to software generation. While agentic workflows facilitate these tasks by decomposing them into multi-step reasoning chains, reliability degrades significantly as the sequence lengthens. Specifically, minor interpretation errors in natural-language instructions tend to compound silently across steps. We term this failure mode accumulated semantic ambiguity. Existing approaches to mitigate this often lack runtime adaptivity, relying instead on static exploration budgets, reactive error recovery, or single-path execution that ignores uncertainty entirely. We formalize the multi-step reasoning process as a Noisy MDP and propose DenoiseFlow, a closed-loop framework that performs progressive denoising through three coordinated stages: (1)Sensing estimates per-step semantic uncertainty; (2)Regulating adaptively allocates computation by routing between fast single-path execution and parallel exploration based on estimated risk; and (3)Correcting performs targeted recovery via influence-based root-cause localization. Online self-calibration continuously aligns decision boundaries with verifier feedback, requiring no ground-truth labels. Experiments on six benchmarks spanning mathematical reasoning, code generation, and multi-hop QA show that DenoiseFlow achieves the highest accuracy on every benchmark (83.3% average, +1.3% over the strongest baseline) while reducing cost by 40--56% through adaptive branching. Detailed ablation studies further confirm framework-level's robustness and generality. Code is available at https://anonymous.4open.science/r/DenoiseFlow-21D3/.

Sampling from unnormalized densities using diffusion models has emerged as a powerful paradigm. However, while recent approaches that use least-squares `matching' objectives have improved scalability, they often necessitate significant trade-offs, such as restricting prior distributions or relying on unstable optimization schemes. By generalizing these methods as special forms of fixed-point iterations rooted in Nelson's relation, we develop a new method that addresses these limitations, called Bridge Matching Sampler (BMS). Our approach enables learning a stochastic transport map between arbitrary prior and target distributions with a single, scalable, and stable objective. Furthermore, we introduce a damped variant of this iteration that incorporates a regularization term to mitigate mode collapse and further stabilize training. Empirically, we demonstrate that our method enables sampling at unprecedented scales while preserving mode diversity, achieving state-of-the-art results on complex synthetic densities and high-dimensional molecular benchmarks.

Image captioning models are encoder-decoder architectures trained on large-scale image-text datasets, making them susceptible to adversarial attacks. We present CaptionFool, a novel universal (input-agnostic) adversarial attack against state-of-the-art transformer-based captioning models. By modifying only 7 out of 577 image patches (approximately 1.2% of the image), our attack achieves 94-96% success rate in generating arbitrary target captions, including offensive content. We further demonstrate that CaptionFool can generate "slang" terms specifically designed to evade existing content moderation filters. Our findings expose critical vulnerabilities in deployed vision-language models and underscore the urgent need for robust defenses against such attacks. Warning: This paper contains model outputs which are offensive in nature.

Mechanistic circuit discovery is notoriously sensitive to arbitrary analyst choices, especially pruning thresholds and feature dictionaries, often yielding brittle "one-shot" explanations with no principled notion of uncertainty. We reframe circuit discovery as an uncertainty-quantification problem over these analytic degrees of freedom. Our method, CIRCUS, constructs an ensemble of attribution graphs by pruning a single raw attribution run under multiple configurations, assigns each edge a stability score (the fraction of configurations that retain it), and extracts a strict-consensus circuit consisting only of edges that appear in all views. This produces a threshold-robust "core" circuit while explicitly surfacing contingent alternatives and enabling rejection of low-agreement structure. CIRCUS requires no retraining and adds negligible overhead, since it aggregates structure across already-computed pruned graphs. On Gemma-2-2B and Llama-3.2-1B, strict consensus circuits are ~40x smaller than the union of all configurations while retaining comparable influence-flow explanatory power, and they outperform a same-edge-budget baseline (union pruned to match the consensus size). We further validate causal relevance with activation patching, where consensus-identified nodes consistently beat matched non-consensus controls (p=0.0004). Overall, CIRCUS provides a practical, uncertainty-aware framework for reporting trustworthy, auditable mechanistic circuits with an explicit core/contingent/noise decomposition.

Tropical cyclone (TC) forecasting is critical for disaster warning and emergency response. Deep learning methods address computational challenges but often neglect physical relationships between TC attributes, resulting in predictions lacking physical consistency. To address this, we propose Phys-Diff, a physics-inspired latent diffusion model that disentangles latent features into task-specific components (trajectory, pressure, wind speed) and employs cross-task attention to introduce prior physics-inspired inductive biases, thereby embedding physically consistent dependencies among TC attributes. Phys-Diff integrates multimodal data including historical cyclone attributes, ERA5 reanalysis data, and FengWu forecast fields via a Transformer encoder-decoder architecture, further enhancing forecasting performance. Experiments demonstrate state-of-the-art performance on global and regional datasets.

The SWE-Bench Verified leaderboard is approaching saturation, with the top system achieving 78.80%. However, we show that this performance is inflated. Our re-evaluation reveals that one in five "solved" patches from the top-30 agents are semantically incorrect, passing only because weak test suites fail to expose their errors. We present SWE-ABS, an adversarial framework that strengthens test suites through a two-stage pipeline: (1) coverage-driven augmentation using program slicing to target untested code regions, and (2) mutation-driven adversarial testing that synthesizes plausible but incorrect patches to expose semantic blind spots. On SWE-Bench Verified (500 instances), SWE-ABS strengthens 50.2% of instances, a 25.1x improvement over prior work, and rejects 19.71% of previously passing patches. As a result, the top agent's score decreases from 78.80% to 62.20%, leading to significant leaderboard reshuffling, with the previous top-ranked agent dropping to fifth place.

Machine Learning often involves various imprecise labels, leading to diverse weakly supervised settings. While recent methods aim for universal handling, they usually suffer from complex manual pre-work, ignore the relationships between associated labels, or are unable to batch process due to computational design flaws, resulting in long running times. To address these limitations, we propose a novel general framework that efficiently infers latent true label distributions across various weak supervisions. Our key idea is to express the label brute-force search process as a probabilistic transition of label variables, compressing diverse weakly supervised DFS tree structures into a shared Bayesian network. From this, we derived a latent probability calculation algorithm based on generalized belief propagation and proposed two joint acceleration strategies: 1) introducing a low-rank assumption to approximate the transition matrix, reducing time complexity; 2) designing an end-to-end state evolution module to learn batch-scale transition matrices, facilitating multi-category batch processing. In addition, the equivalence of our method with the EM algorithm in most scenarios is further demonstrated. Extensive experiments show that our method achieves SOTA results under most weakly supervised settings, and achieves up to hundreds of times faster acceleration in running time compared to other general methods.

Visual Question Answering systems face reliability issues due to hallucinations, where models generate answers misaligned with visual input or factual knowledge. While Retrieval Augmented Generation frameworks mitigate this issue by incorporating external knowledge, static retrieval often introduces irrelevant or conflicting content, particularly in visual RAG settings where visually similar but semantically incorrect evidence may be retrieved. To address this, we propose Multimodal Adaptive RAG (MMA-RAG), which dynamically assesses the confidence in the internal knowledge of the model to decide whether to incorporate the retrieved external information into the generation process. Central to MMA-RAG is a decision classifier trained through a layer-wise analysis, which leverages joint internal visual and textual representations to guide the use of reverse image retrieval. Experiments demonstrated that the model achieves a significant improvement in response performance in three VQA datasets. Meanwhile, ablation studies highlighted the importance of internal representations in adaptive retrieval decisions. In general, the experimental results demonstrated that MMA-RAG effectively balances external knowledge utilization and inference robustness in diverse multimodal scenarios.

Multimodal large language models (MLLMs) project visual tokens into the embedding space of language models, yet the internal structuring and processing of visual semantics remain poorly understood. In this work, we introduce a two-fold analytical framework featuring a novel probing tool, $\textbf{EmbedLens}$, to conduct a fine-grained analysis. We uncover a pronounced semantic sparsity at the input level: visual tokens consistently partition into sink, dead, and alive categories. Remarkably, only the alive tokens, comprising $\approx60\%$ of the total input, carry image-specific meaning. Furthermore, using a targeted patch-compression benchmark, we demonstrate that these alive tokens already encode rich, fine-grained cues (e.g., objects, colors, and OCR) prior to entering the LLM. Internal visual computations (such as visual attention and feed-forward networks) are redundant for most standard tasks. For the small subset of highly vision-centric tasks that actually benefit from internal processing, we reveal that alive tokens naturally align with intermediate LLM layers rather than the initial embedding space, indicating that shallow-layer processing is unnecessary and that direct mid-layer injection is both sufficient. Ultimately, our findings provide a unified mechanistic view of visual token processing, paving the way for more efficient and interpretable MLLM architectures through selective token pruning, minimized visual computation, and mid-layer injection. The code is released at: https://github.com/EIT-NLP/EmbedLens.

The increasing adoption of UAVs equipped with advanced sensors and GPU-accelerated edge computing has enabled real-time AI-driven applications in domains such as precision agriculture, wildfire monitoring, and environmental conservation. However, the integrated design and orchestration of navigation, sensing, and analytics, together with seamless real-time coordination across drone, edge, and cloud resources, remains a significant challenge. To address these challenges, we propose AeroDaaS, a service-oriented framework that abstracts UAV-based sensing complexities and provides a Drone-as-a-Service (DaaS) model for intelligent decision-making. AeroDaaS offers modular service primitives for on-demand UAV sensing, navigation and analytics as composable microservices, ensuring cross-platform compatibility and scalability across heterogeneous UAV and edge-cloud infrastructures. AeroDaaS also supports plug-and-play scheduling modules, including Waypoint and Analytics schedulers, which enable trajectory optimization and real-time coordination of inference workloads. We implement and evaluate AeroDaaS for six real-world DaaS applications, of which two are evaluated in real-world scenarios and four in simulation. AeroDaaS requires less than 40 lines of code for the applications and has minimal platform overhead of less than 20 ms per frame and about 1 GB memory usage on Orin Nano and a AMD RTX 3090 GPU workstation. These results are promising for AeroDaaS as an efficient, flexible and scalable UAV programming framework for autonomous aerial analytics.

Early-stage users in a new scenario intensify cold-start challenges, yet prior works often address only parts of the problem through model architecture. Launching a new user experience to replace an established product involves sparse behavioral signals, low-engagement cohorts, and unstable model performance. We argue that effective recommendations require the synergistic integration of feature engineering, model architecture, and stable model updating. We propose Trinity, a framework embodying this principle. Trinity extracts valuable information from existing scenarios while ensuring predictive effectiveness and accuracy in the new scenario. In this paper, we showcase Trinity applied to a billion-user Microsoft product transition. Both offline and online experiments demonstrate that our framework achieves substantial improvements in addressing the combined challenge of new users in new scenarios.

The integration of large language models (LLMs) into wireless networks has sparked growing interest in building autonomous AI agents for wireless tasks. However, existing approaches rely heavily on manually crafted prompts and static agentic workflows, a process that is labor-intensive, unscalable, and often suboptimal. In this paper, we propose WirelessAgent++, a framework that automates the design of agentic workflows for various wireless tasks. By treating each workflow as an executable code composed of modular operators, WirelessAgent++ casts agent design as a program search problem and solves it with a domain-adapted Monte Carlo Tree Search (MCTS) algorithm. Moreover, we establish WirelessBench, a standardized multi-dimensional benchmark suite comprising Wireless Communication Homework (WCHW), Network Slicing (WCNS), and Mobile Service Assurance (WCMSA), covering knowledge reasoning, code-augmented tool use, and multi-step decision-making. Experiments demonstrate that \wap{} autonomously discovers superior workflows, achieving test scores of $78.37\%$ (WCHW), $90.95\%$ (WCNS), and $97.07\%$ (WCMSA), with a total search cost below $\$ 5$ per task. Notably, our approach outperforms state-of-the-art prompting baselines by up to $31\%$ and general-purpose workflow optimizers by $11.1\%$, validating its effectiveness in generating robust, self-evolving wireless agents. The code is available at https://github.com/jwentong/WirelessAgent-R2.

Fine-tuning-as-a-service introduces a threat to Large Language Models' safety when service providers fine-tune their models on poisoned user-submitted datasets, a process known as harmful fine-tuning attacks. In this work, we show that by regularizing the gradient contribution of harmful samples encountered during fine-tuning, we can effectively mitigate the impact of harmful fine-tuning attacks. To this end, we introduce Antibody, a defense strategy that first ensures robust safety alignment for the model before fine-tuning, and then applies a safety-preservation learning algorithm during fine-tuning. Specifically, in the alignment stage before fine-tuning, we propose optimizing the model to be in a flat loss region with respect to harmful samples, which makes the safety alignment more resilient to subsequent harmful fine-tuning. Then, in the fine-tuning stage, we design a fine-tuning algorithm that applies a weighting scheme to all samples in each training batch to inhibit the model from learning from harmful samples while encouraging learning from benign samples. Experimental results demonstrate that Antibody successfully mitigates harmful fine-tuning attacks while boosting fine-tuning performance on the user-submitted dataset.

In this paper, we provide a theoretically grounded and computationally efficient alternative to SHAP. To this end, we study feature attribution through the lens of cooperative game theory by formulating a class of XAI--TU games. Building on this formulation, we investigate equal-surplus-type and proportional-allocation-type attribution rules and propose a low-cost attribution rule, ESENSC_rev2, constructed by combining two polynomial-time closed-form rules while ensuring the null-player property in the XAI--TU domain. Extensive experiments on tabular prediction tasks demonstrate that ESENSC_rev2 closely approximates exact SHAP while substantially improving scalability as the number of features increases. These empirical results indicate that equal-surplus-type attribution rules can achieve favorable trade-offs between computational cost and approximation accuracy in high-dimensional explainability settings. To provide theoretical foundations for these findings, we establish an axiomatic characterization showing that ESENSC_rev2 is uniquely determined by efficiency, the null-player axiom, a restricted differential marginality principle, an intermediate inessential-game property, and axioms that reduce computational requirements. Our results suggest that axiomatically justified and computationally efficient attribution rules can serve as practical and theoretically principled substitutes for SHAP-based approximations in modern explainability pipelines.

We introduce AI Runtime Infrastructure, a distinct execution-time layer that operates above the model and below the application, actively observing, reasoning over, and intervening in agent behavior to optimize task success, latency, token efficiency, reliability, and safety while the agent is running. Unlike model-level optimizations or passive logging systems, runtime infrastructure treats execution itself as an optimization surface, enabling adaptive memory management, failure detection, recovery, and policy enforcement over long-horizon agent workflows.

Per-scene optimization methods such as 3D Gaussian Splatting provide state-of-the-art novel view synthesis quality but extrapolate poorly to under-observed areas. Methods that leverage generative priors to correct artifacts in these areas hold promise but currently suffer from two shortcomings. The first is scalability, as existing methods use image diffusion models or bidirectional video models that are limited in the number of views they can generate in a single pass (and thus require a costly iterative distillation process for consistency). The second is quality itself, as generators used in prior work tend to produce outputs that are inconsistent with existing scene content and fail entirely in completely unobserved regions. To solve these, we propose a two-stage pipeline that leverages two key insights. First, we train a powerful bidirectional generative model with a novel opacity mixing strategy that encourages consistency with existing observations while retaining the model's ability to extrapolate novel content in unseen areas. Second, we distill it into a causal auto-regressive model that generates hundreds of frames in a single pass. This model can directly produce novel views or serve as pseudo-supervision to improve the underlying 3D representation in a simple and highly efficient manner. We evaluate our method extensively and demonstrate that it can generate plausible reconstructions in scenarios where existing approaches fail completely. When measured on commonly benchmarked datasets, we outperform existing all existing baselines by a wide margin, exceeding prior state-of-the-art methods by 1-3 dB PSNR.

Support matrix machine (SMM) is an emerging classification framework that directly handles matrix-structured observations, thereby avoiding the spatial correlations destroyed by vectorization. However, most existing SMM variants rely on convex or nonconvex surrogate loss functions, which may lead to high sensitivity to noise. To address this issue, we propose a novel Heaviside low-rank SMM model called HL-SMM, which leverages the Heaviside loss instead of the common hinge or ramp losses for robustness. Moreover, the low-rank constraint is adopted to accurately characterize the inherent global structure. In theory, we analyze the Karush-Kuhn-Tucker (KKT) points and rigorously prove the sufficient and necessary conditions. In algorithms, we develop an effective proximal alternating minimization (PAM) scheme, where all subproblems have closed-form solutions. Extensive experiments on benchmark datasets validate that the proposed HL-SMM achieves superior classification accuracy and robustness compared to state-of-the-art methods.

The rapid progress of Multimodal Large Language Models (MLLMs) marks a significant step toward artificial general intelligence, offering great potential for augmenting human capabilities. However, their ability to provide effective assistance in dynamic, real-world environments remains largely underexplored. Existing video benchmarks predominantly assess passive understanding through retrospective analysis or isolated perception tasks, failing to capture the interactive and adaptive nature of real-time user assistance. To bridge this gap, we introduce LifeEval, a multimodal benchmark designed to evaluate real-time, task-oriented human-AI collaboration in daily life from an egocentric perspective. LifeEval emphasizes three key aspects: task-oriented holistic evaluation, egocentric real-time perception from continuous first-person streams, and human-assistant collaborative interaction through natural dialogues. Constructed via a rigorous annotation pipeline, the benchmark comprises 4,075 high-quality question-answer pairs across 6 core capability dimensions. Extensive evaluations of 26 state-of-the-art MLLMs on LifeEval reveal substantial challenges in achieving timely, effective and adaptive interaction, highlighting essential directions for advancing human-centered interactive intelligence.

The README file serves as a critical source of information for gaining an overview and helping developers onboard to an Open Source Software (OSS) project. Yet, documentation issues persist; in particular, ``outdated'' documentation is perceived by developers as one of the most frequent and severe challenges with gaining project understanding. While previous studies have aimed to mitigate this problem, they typically either rely on highly-engineered solutions focused on specific code components or employ generative methods that are ineffective for incremental maintenance. In this study, we propose a lightweight Large Language Model (LLM)-driven approach to facilitate precise, localised README file updates within a human-in-the-loop workflow. Specifically, given a pull request (PR), our pipeline determines whether an update is necessary; if so, it identifies the precise locations where updates should be applied and provides a justification based on the triggering events. Our evaluation on 27,772 PRs across 714 popular repositories demonstrates high precision and utility. Furthermore, we performed a qualitative failure case analysis to provide deeper insights and directions for improvement. We also conducted a retrospective study on 20 sampled repositories, complemented by a case study with a developer of a large OSS project. These evaluations demonstrate that the tool effectively identifies overlooked PRs requiring README updates, thereby helping to mitigate the risk of outdated documentation. Finally, we provide concrete implications for practitioners and researchers, highlighting the need to further explore effective interaction patterns to incorporate documentation update tools into the OSS development workflow.

Adolescent pornography addiction requires early detection based on objective neurobiological biomarkers because self-report is prone to subjective bias due to social stigma. Conventional machine learning has not been able to model dynamic functional connectivity of the brain that fluctuates temporally during addictive stimulus exposure. This study proposes a state-of-the-art Dynamic Spatio-Temporal Graph Neural Network (DST-GNN) that integrates Phase Lag Index (PLI)-based Graph Attention Network (GAT) for spatial modeling and Bidirectional Gated Recurrent Unit (BiGRU) for temporal dynamics. The dataset consists of 14 adolescents (7 addicted, 7 healthy) with 19-channel EEG across 9 experimental conditions. Leave-One-Subject-Out Cross Validation (LOSO-CV) evaluation shows F1-Score of 71.00%$\pm$12.10% and recall of 85.71%, a 104% improvement compared to baseline. Ablation study confirms temporal contribution of 21% and PLI graph construction of 57%. Frontal-central regions (Fz, Cz, C3, C4) are identified as dominant biomarkers with Beta contribution of 58.9% and Hjorth of 31.2%, while Cz-T7 connectivity is consistent as a trait-level biomarker for objective screening.

Recent text-to-image (T2I) diffusion models achieve remarkable realism, yet faithful prompt-image alignment remains challenging, particularly for complex prompts with multiple objects, relations, and fine-grained attributes. Existing training-free inference-time scaling methods rely on fixed iteration budgets that cannot adapt to prompt difficulty, while reflection-tuned models require carefully curated reflection datasets and extensive joint fine-tuning of diffusion and vision-language models, often overfitting to reflection paths data and lacking transferability across models. We introduce RAISE (Requirement-Adaptive Self-Improving Evolution), a training-free, requirement-driven evolutionary framework for adaptive T2I generation. RAISE formulates image generation as a requirement-driven adaptive scaling process, evolving a population of candidates at inference time through a diverse set of refinement actions-including prompt rewriting, noise resampling, and instructional editing. Each generation is verified against a structured checklist of requirements, enabling the system to dynamically identify unsatisfied items and allocate further computation only where needed. This achieves adaptive test-time scaling that aligns computational effort with semantic query complexity. On GenEval and DrawBench, RAISE attains state-of-the-art alignment (0.94 overall GenEval) while incurring fewer generated samples (reduced by 30-40%) and VLM calls (reduced by 80%) than prior scaling and reflection-tuned baselines, demonstrating efficient, generalizable, and model-agnostic multi-round self-improvement. Code is available at https://github.com/LiyaoJiang1998/RAISE.

Developments in the machine learning voting domain have shown both promising results and risks. Trained models perform well on ballot classification tasks (> 99% accuracy) but are at risk from adversarial example attacks that cause misclassifications. In this paper, we analyze an attacker who seeks to deploy adversarial examples against machine learning ballot classifiers to compromise a U.S. election. We first derive a probabilistic framework for determining the number of adversarial example ballots that must be printed to flip an election, in terms of the probability of each candidate winning and the total number of ballots cast. Second, it is an open question as to which type of adversarial example is most effective when physically printed in the voting domain. We analyze six different types of adversarial example attacks: l_infinity-APGD, l2-APGD, l1-APGD, l0 PGD, l0 + l_infinity PGD, and l0 + sigma-map PGD. Our experiments include physical realizations of 144,000 adversarial examples through printing and scanning with four different machine learning models. We empirically demonstrate an analysis gap between the physical and digital domains, wherein attacks most effective in the digital domain (l2 and l_infinity) differ from those most effective in the physical domain (l1 and l2, depending on the model). By unifying a probabilistic election framework with digital and physical adversarial example evaluations, we move beyond prior close race analyses to explicitly quantify when and how adversarial ballot manipulation could alter outcomes.

Few-shot transfer has been revolutionized by stronger pre-trained models and improved adaptation algorithms.However, there lacks a unified, rigorous evaluation protocol that is both challenging and realistic for real-world usage. In this work, we establish FEWTRANS, a comprehensive benchmark containing 10 diverse datasets, and propose the Hyperparameter Ensemble (HPE) protocol to overcome the "validation set illusion" in data-scarce regimes. Our empirical findings demonstrate that the choice of pre-trained model is the dominant factor for performance, while many sophisticated transfer methods offer negligible practical advantages over a simple full-parameter fine-tuning baseline. To explain this surprising effectiveness, we provide an in-depth mechanistic analysis showing that full fine-tuning succeeds via distributed micro-adjustments and more flexible reshaping of high-level semantic presentations without suffering from overfitting. Additionally, we quantify the performance collapse of multimodal models in specialized domains as a result of linguistic rarity using adjusted Zipf frequency scores. By releasing FEWTRANS, we aim to provide a rigorous "ruler" to streamline reproducible advances in few-shot transfer learning research. We make the FEWTRANS benchmark publicly available at https://github.com/Frankluox/FewTrans.

Stencil computation constitutes a cornerstone of scientific computing, serving as a critical kernel in domains ranging from fluid dynamics to weather simulation. While stencil computations are conventionally regarded as memory-bound and thus unsuitable for compute-centric Tensor Cores, recent empirical studies have demonstrated significant speedups after applying Tensor Cores, forming an apparent contradiction. This paper resolves this contradiction by conducting a systematic performance analysis of stencil computations on Tensor Cores. We begin by revisiting the adaptation of stencils onto Tensor Cores, quantifying the computational redundancy introduced by the transformations required to satisfy hardware constraints. These metrics are subsequently integrated into an enhanced performance model that explicitly accounts for the arithmetic intensity shifts driven by temporal fusion. Guided by this formulation, we derive analytical criteria to determine the suitability of Tensor Cores for varying stencil workloads. By classifying operational regions, we identify the specific \textit{sweet spot} for effective acceleration and further demonstrate how Sparse Tensor Cores expand this profitable design space. Extensive evaluations on NVIDIA GPUs across SOTA implementations, including DRStencil, EBISU, ConvStencil, and SPIDER, validate our performance model and analytical criteria. These results demonstrate the effectiveness of our approach in guiding stencil performance optimization.

Browser-use agents are widely used for everyday tasks. They enable automated interaction with web pages through structured DOM based interfaces or vision language models operating on page screenshots. However, web pages often change between planning and execution, causing agents to execute actions based on stale assumptions. We view this temporal mismatch as a time of check to time of use (TOCTOU) vulnerability in browser-use agents. Dynamic or adversarial web content can exploit this window to induce unintended actions. We present a large scale empirical study of TOCTOU vulnerabilities in browser-use agents using a benchmark that spans synthesized and real world websites. Using this benchmark, we evaluate 10 popular open source agents and show that TOCTOU vulnerabilities are widespread. We design a lightweight mitigation based on pre-execution validation. It monitors DOM and layout changes during planning and validates the page state immediately before action execution. This approach reduces the risk of insecure execution and mitigates unintended side effects in browser-use agents.

This paper investigates the power control problem in wireless networks by repurposing pre-trained large language models (LLMs) as relational reasoning backbones. In hyper-connected interference environments, traditional optimization methods face high computational cost, while standard message passing neural networks suffer from aggregation bottlenecks that can obscure critical high-interference structures. In response, we propose PC-LLM, a physics-informed framework that augments a pre-trained Transformer with an interference-aware attention bias. The proposed bias tuning mechanism injects the physical channel gain matrix directly into the self-attention logits, enabling explicit fusion of wireless topology with pre-trained relational priors without retraining the backbone from scratch. Extensive experiments demonstrate that PC-LLM consistently outperforms both traditional optimization methods and state-of-the-art graph neural network baselines, while exhibiting exceptional zero-shot generalization to unseen environments. We further observe a structural-semantic decoupling phenomenon: Topology-relevant relational reasoning is concentrated in shallow layers, whereas deeper layers encode task-irrelevant semantic noise. Motivated by this finding, we develop a lightweight adaptation strategy that reduces model depth by 50\%, significantly lowering inference cost while preserving state-of-the-art spectral efficiency.

Agentic AI systems exhibit numerous crosscutting concerns -- security, observability, cost management, fault tolerance -- that are poorly modularized in current implementations, contributing to the high failure rate of AI projects in reaching production. The goals-to-aspects methodology proposed at RE 2004 demonstrated that aspects can be systematically discovered from i* goal models by identifying non-functional soft-goals that crosscut functional goals. This paper revisits and extends that methodology to the agentic AI domain. We present a pattern language of 12 reusable patterns organized across four NFR categories (security, reliability, observability, cost management), each mapping an i* goal model to a concrete aspect implementation using an AOP framework for Rust. Four patterns address agent-specific crosscutting concerns absent from traditional AOP literature: tool-scope sandboxing, prompt injection detection, token budget management, and action audit trails. We extend the V-graph model to capture how agent tasks simultaneously contribute to functional goals and non-functional soft-goals. We validate the pattern language through a case study analyzing an open-source autonomous agent framework, demonstrating how goal-driven aspect discovery systematically identifies and modularizes crosscutting concerns. The pattern language offers a principled approach for engineering reliable agentic AI systems through early identification of crosscutting concerns.

Operators of Earth observation satellites need justifications for scheduling decisions: why a request was selected, rejected, or what changes would make it schedulable. Existing approaches construct post-hoc reasoning layers independent of the optimizer, risking non-causal attributions, incomplete constraint conjunctions, and solver-path dependence. We take a faithfulness-first approach: every explanation is a certificate derived from the optimization model itself: minimal infeasible subsets for rejections, tight constraints and contrastive trade-offs for selections, and inverse solves for what-if queries. On a scheduling instance with structurally distinct constraint interactions, certificates achieve perfect soundness with respect to the solver's constraint model (15/15 cited-constraint checks), counterfactual validity (7/7), and stability (Jaccard = 1.0 across 28 seed-pairs), while a post-hoc baseline produces non-causal attributions in 29% of cases and misses constraint conjunctions in every multi-cause rejection. A scalability analysis up to 200 orders and 30 satellites confirms practical extraction times for operational batches.

The transition to agentic Root Cause Analysis (RCA) necessitates benchmarks that evaluate active reasoning rather than passive classification. However, current frameworks fail to reconcile ecological validity with reproducibility. We introduce Cloud-OpsBench, a large-scale benchmark that employs a State Snapshot Paradigm to construct a deterministic digital twin of the cloud, featuring 452 distinct fault cases across 40 root cause types spanning the full Kubernetes stack. Crucially, Cloud-OpsBench serves as an enabling infrastructure for next-generation SRE research: (1) As a Data Engine, it harvests high-quality reasoning trajectories to bootstrap Supervised Fine-Tuning (SFT) for Small Language Models; (2) As an Reinforcement Learning (RL) environment, it transforms high-risk operations into a safe low-latency sandbox for training policy optimization agents; and (3) As a Diagnostic Standard, its process-centric protocol uncovers architectural bottlenecks guiding the design of robust specialized multi-agent system for RCA.

Automated assessment of open-ended student responses is a critical capability for scaling personalized feedback in education. While large language models (LLMs) have shown promise in grading tasks via in-context learning (ICL), their reliability is heavily dependent on the selection of few-shot exemplars and the construction of high-quality rationales. Standard retrieval methods typically select examples based on semantic similarity, which often fails to capture subtle decision boundaries required for rubric adherence. Furthermore, manually crafting the expert rationales needed to guide these models can be a significant bottleneck. To address these limitations, we introduce GUIDE (Grading Using Iteratively Designed Exemplars), a framework that reframes exemplar selection and refinement in automated grading as a boundary-focused optimization problem. GUIDE operates on a continuous loop of selection and refinement, employing novel contrastive operators to identify "boundary pairs" that are semantically similar but possess different grades. We enhance exemplars by generating discriminative rationales that explicitly articulate why a response receives a specific score to the exclusion of adjacent grades. Extensive experiments across datasets in physics, chemistry, and pedagogical content knowledge demonstrate that GUIDE significantly outperforms standard retrieval baselines. By focusing the model's attention on the precise edges of rubric, our approach shows exceptionally robust gains on borderline cases and improved rubric adherence. GUIDE paves the way for trusted, scalable assessment systems that align closely with human pedagogical standards.

Orthopantomograms (OPGs) are the standard panoramic radiograph in dentistry, used for full-arch screening across multiple diagnostic tasks. While Vision Language Models (VLMs) now allow multi-task OPG analysis through natural language, they underperform task-specific models on most individual tasks. Agentic systems that orchestrate specialized tools offer a path to both versatility and accuracy, this approach remains unexplored in the field of dental imaging. To address this gap, we propose OPGAgent, a multi-tool agentic system for auditable OPG interpretation. OPGAgent coordinates specialized perception modules with a consensus mechanism through three components: (1) a Hierarchical Evidence Gathering module that decomposes OPG analysis into global, quadrant, and tooth-level phases with dynamically invoking tools, (2) a Specialized Toolbox encapsulating spatial, detection, utility, and expert zoos, and (3) a Consensus Subagent that resolves conflicts through anatomical constraints. We further propose OPG-Bench, a structured-report protocol based on (Location, Field, Value) triples derived from real clinical reports, which enables a comprehensive review of findings and hallucinations, extending beyond the limitations of VQA indicators. On our OPG-Bench and the public MMOral-OPG benchmark, OPGAgent outperforms current dental VLMs and medical agent frameworks across both structured-report and VQA evaluation. Code will be released upon acceptance.

Clinical decision-making requires synthesizing heterogeneous evidence, including patient histories, clinical guidelines, and trajectories of comparable cases. While large language models (LLMs) offer strong reasoning capabilities, they remain prone to hallucinations and struggle to integrate long, structured medical documents. We present MED-COPILOT, an interactive clinical decision-support system designed for clinicians and medical trainees, which combines guideline-grounded GraphRAG retrieval with hybrid semantic-keyword similar-patient retrieval to support transparent and evidence-aware clinical reasoning. The system builds a structured knowledge graph from WHO and NICE guidelines, applies community-level summarization for efficient retrieval, and maintains a 36,000-case similar-patient database derived from SOAP-normalized MIMIC-IV notes and Synthea-generated records. We evaluate our framework on clinical note completion and medical question answering, and demonstrate that it consistently outperforms parametric LLM baselines and standard RAG, improving both generation fidelity and clinical reasoning accuracy. The full system is available at https://huggingface.co/spaces/Cryo3978/Med_GraphRAG , enabling users to inspect retrieved evidence, visualize token-level similarity contributions, and conduct guided follow-up analysis. Our results demonstrate a practical and interpretable approach to integrating structured guideline knowledge with patient-level analogical evidence for clinical LLMs.

Generative Flow Networks (GFlowNets) enable fine-tuning large language models to approximate reward-proportional posteriors, but they remain prone to mode collapse, manifesting as prefix collapse and length bias. We attribute this to two factors: (i) weak credit assignment to early prefixes, and (ii) biased replay that induces a shifted, non-representative training flow distribution. We propose Rooted absorbed prefix Trajectory Balance RapTB, an objective that anchors subtrajectory supervision at the root and propagates terminal rewards to intermediate prefixes via absorbed suffix-based backups, providing dense prefix-level learning signals. To mitigate replay-induced distribution shift, we further introduce SubM, a submodular replay refresh strategy that promotes both high reward and diversity. Empirically, on tasks such as molecule generation with LLM using SMILES strings, RapTB combined with SubM consistently improves optimization performance and molecular diversity while preserving high validity.

The growing deployment of Internet of Things (IoT) devices in smart cities and industrial environments increases vulnerability to stealthy, multi-stage advanced persistent threats (APTs) that exploit wireless communication. Detection is challenging due to severe class imbalance in network traffic, which limits the effectiveness of traditional deep learning approaches and their lack of explainability in classification decisions. To address these challenges, this paper proposes a neurosymbolic architecture that integrates an optimized BERT model with logic tensor networks (LTN) for explainable APT detection in wireless IoT networks. The proposed method addresses the challenges of mobile IoT environments through efficient feature encoding that transforms network flow data into BERT-compatible sequences while preserving temporal dependencies critical for APT stage identification. Severe class imbalance is mitigated using focal loss, hierarchical classification that separates normal traffic detection from attack categorization, and adaptive sampling strategies. Evaluation on the SCVIC-APT2021 dataset demonstrates an operationally viable binary classification F1 score of 95.27% with a false positive rate of 0.14%, and a 76.75% macro F1 score for multi-class attack categorization. Furthermore, a novel explainability analysis statistically validates the importance of distinct network features. These results demonstrate that neurosymbolic learning enables high-performance, interpretable, and operationally viable APT detection for IoT network monitoring architectures.

What if text could be sculpted and refined like clay -- or cultivated and pruned like a plant? Texterial reimagines text as a material that users can grow, sculpt, and transform. Current generative-AI models enable rich text operations, yet rigid, linear interfaces often mask such capabilities. We explore how the text-as-material metaphor can reveal AI-enabled operations, reshape the writing process, and foster compelling user experiences. A formative study shows that users readily reason with text-as-material, informing a conceptual framework that explains how material metaphors shift mental models and bridge gulfs of envisioning, execution, and evaluation in LLM-mediated writing. We present the design and evaluation of two technical probes: Text as Clay, where users refine text through gestural sculpting, and Text as Plants, where ideas grow serendipitously over time. This work expands the design space of writing tools by treating text as a living, malleable medium.

Accurate and unambiguous guidelines are critical for large language model (LLM) based graders, yet manually crafting these prompts is often sub-optimal as LLMs can misinterpret expert guidelines or lack necessary domain specificity. Consequently, the field has moved toward automated prompt optimization to refine grading guidelines without the burden of manual trial and error. However, existing frameworks typically aggregate independent and unstructured error samples into a single update step, resulting in "rule dilution" where conflicting constraints weaken the model's grading logic. To address these limitations, we introduce Confusion-Aware Rubric Optimization (CARO), a novel framework that enhances accuracy and computational efficiency by structurally separating error signals. CARO leverages the confusion matrix to decompose monolithic error signals into distinct modes, allowing for the diagnosis and repair of specific misclassification patterns individually. By synthesizing targeted "fixing patches" for dominant error modes and employing a diversity-aware selection mechanism, the framework prevents guidance conflict and eliminates the need for resource-heavy nested refinement loops. Empirical evaluations on teacher education and STEM datasets demonstrate that CARO significantly outperforms existing SOTA methods. These results suggest that replacing mixed-error aggregation with surgical, mode-specific repair yields robust improvements in automated assessment scalability and precision.

In this paper, we address the problem of tactile sim-to-real policy transfer for contact-rich tasks. Existing methods primarily focus on vision-based sensors and emphasize image rendering quality while providing overly simplistic models of force and shear. Consequently, these models exhibit a large sim-to-real gap for many dexterous tasks. Here, we present HydroShear, a non-holonomic hydroelastic tactile simulator that advances the state-of-the-art by modeling: a) stick-slip transitions, b) path-dependent force and shear build up, and c) full SE(3) object-sensor interactions. HydroShear extends hydroelastic contact models using Signed Distance Functions (SDFs) to track the displacements of the on-surface points of an indenter during physical interaction with the sensor membrane. Our approach generates physics-based, computationally efficient force fields from arbitrary watertight geometries while remaining agnostic to the underlying physics engine. In experiments with GelSight Minis, HydroShear more faithfully reproduces real tactile shear compared to existing methods. This fidelity enables zero-shot sim-to-real transfer of reinforcement learning policies across four tasks: peg insertion, bin packing, book shelving for insertion, and drawer pulling for fine gripper control under slip. Our method achieves a 93% average success rate, outperforming policies trained on tactile images (34%) and alternative shear simulation methods (58%-61%).

Computer-Aided Design (CAD) generative modeling has a strong and long-term application in the industry. Recently, the parametric CAD sequence as the design logic of an object has been widely mined by sequence models. However, the industrial CAD models, especially in component objects, are fine-grained and complex, requiring a longer parametric CAD sequence to define. To address the problem, we introduce Mamba-CAD, a self-supervised generative modeling for complex CAD models in the industry, which can model on a longer parametric CAD sequence. Specifically, we first design an encoder-decoder framework based on a Mamba architecture and pair it with a CAD reconstruction task for pre-training to model the latent representation of CAD models; and then we utilize the learned representation to guide a generative adversarial network to produce the fake representation of CAD models, which would be finally recovered into parametric CAD sequences via the decoder of MambaCAD. To train Mamba-CAD, we further create a new dataset consisting of 77,078 CAD models with longer parametric CAD sequences. Comprehensive experiments are conducted to demonstrate the effectiveness of our model under various evaluation metrics, especially in the generation length of valid parametric CAD sequences. The code and dataset can be achieved from https://github.com/Sunny-Hack/Code-for-Mamba-CAD-AAAI-2025-.

The need for machine unlearning is critical for data privacy, yet existing methods often cause Knowledge Contamination by unintentionally damaging related knowledge. Such a degraded model performance after unlearning has been recently leveraged for new inference and backdoor attacks. Most studies design adversarial unlearning requests that require poisoning or duplicating training data. In this study, we introduce a new unlearning-induced attack model, namely indirect unlearning attack, which does not require data manipulation but exploits the consequence of knowledge contamination to perturb the model accuracy on security-critical predictions. To mitigate this attack, we introduce a theoretical framework that models neural networks as Neural Knowledge Systems. Based on this, we propose ROKA, a robust unlearning strategy centered on Neural Healing. Unlike conventional unlearning methods that only destroy information, ROKA constructively rebalances the model by nullifying the influence of forgotten data while strengthening its conceptual neighbors. To the best of our knowledge, our work is the first to provide a theoretical guarantee for knowledge preservation during unlearning. Evaluations on various large models, including vision transformers, multi-modal models, and large language models, show that ROKA effectively unlearns targets while preserving, or even enhancing, the accuracy of retained data, thereby mitigating the indirect unlearning attacks.

Industrial chip development is inherently iterative, favoring localized, intent-driven updates over rewriting RTL from scratch. Yet most LLM-Aided Hardware Design (LAD) work focuses on one-shot synthesis, leaving this workflow underexplored. To bridge this gap, we for the first time formalize $Δ$Spec-to-RTL localization, a multi-positive problem mapping natural language change requests ($Δ$Spec) to the affected Register Transfer Level (RTL) syntactic blocks. We propose RTLocating, an intent-aware RTL localization framework, featuring a dynamic router that adaptively fuses complementary views from a textual semantic encoder, a local structural encoder, and a global interaction and dependency encoder (GLIDE). To enable scalable supervision, we introduce EvoRTL-Bench, the first industrial-scale benchmark for intent-code alignment derived from OpenTitan's Git history, comprising 1,905 validated requests and 13,583 $Δ$Spec-RTL block pairs. On EvoRTL-Bench, RTLocating achieves 0.568 MRR and 15.08% R@1, outperforming the strongest baseline by +22.9% and +67.0%, respectively, establishing a new state-of-the-art for intent-driven localization in evolving hardware designs.

Executing multiple tasks simultaneously in medical image analysis, including segmentation, classification, detection, and regression, often introduces significant challenges regarding model generalizability and the optimization of shared feature representations. While Vision Foundation Models (VFMs) provide powerful general representations, full fine-tuning on limited medical data is prone to overfitting and incurs high computational costs. Moreover, existing parameter-efficient fine-tuning approaches typically adopt task-agnostic adaptation protocols, overlooking both task-specific mechanisms and the varying sensitivity of model layers during fine-tuning. In this work, we propose Task-Aware Prompting and Selective Layer Fine-Tuning (TAP-SLF), a unified framework for multi-task ultrasound image analysis. TAP-SLF incorporates task-aware soft prompts to encode task-specific priors into the input token sequence and applies LoRA to selected specific top layers of the encoder. This strategy updates only a small fraction of the VFM parameters while keeping the pre-trained backbone frozen. By combining task-aware prompts with selective high-layer fine-tuning, TAP-SLF enables efficient VFM adaptation to diverse medical tasks within a shared backbone. Results on the FMC_UIA 2026 Challenge test set, where TAP-SLF wins fifth place, combined with evaluations on the officially released training dataset using an 8:2 train-test split, demonstrate that task-aware prompting and selective layer tuning are effective strategies for efficient VFM adaptation.

We introduce a typology-aware diagnostic for multilingual masked language models that tests reliance on word order versus inflectional form. Using Universal Dependencies, we apply inference-time perturbations: full token scrambling, content-word scrambling with function words fixed, dependency-based head--dependent swaps, and sentence-level lemma substitution (+L), which lemmatizes both the context and the masked target label. We evaluate mBERT and XLM-R on English, Chinese, German, Spanish, and Russian. Full scrambling drives word-level reconstruction accuracy near zero in all languages; partial and head--dependent perturbations cause smaller but still large drops. +L has little effect in Chinese but substantially lowers accuracy in German/Spanish/Russian, and it does not mitigate the impact of scrambling. Top-5 word accuracy shows the same pattern: under full scrambling, the gold word rarely appears among the five highest-ranked reconstructions. We release code, sampling scripts, and balanced evaluation subsets; Turkish results under strict reconstruction are reported in the appendix.

A high-performing, general-purpose visual understanding model should map visual inputs to a taxonomic tree of labels, identify novel categories beyond the training set for which few or no publicly available images exist. Large Multimodal Models (LMMs) have achieved remarkable progress in fine-grained visual recognition (FGVR) for known categories. However, they remain limited in hierarchical visual recognition (HVR) that aims at predicting consistent label paths from coarse to fine categories, especially for novel categories. To tackle these challenges, we propose Taxonomy-Aware Representation Alignment (TARA), a simple yet effective strategy to inject taxonomic knowledge into LMMs. TARA leverages representations from biology foundation models (BFMs) that encode rich biological relationships through hierarchical contrastive learning. By aligning the intermediate representations of visual features with those of BFMs, LMMs are encouraged to extract discriminative visual cues well structured in the taxonomy tree. Additionally, we align the representations of the first answer token with the ground-truth label, flexibly bridging the gap between contextualized visual features and categories of varying granularity according to user intent. Experiments demonstrate that TARA consistently enhances LMMs' hierarchical consistency and leaf node accuracy, enabling reliable recognition of both known and novel categories within complex biological taxonomies. Code is available at https://github.com/PKU-ICST-MIPL/TARA_CVPR2026.

Construction-based neural routing solvers, typically composed of an encoder and a decoder, have emerged as a promising approach for solving vehicle routing problems. While recent studies suggest that shifting parameters from the encoder to the decoder enhances performance, most works restrict the decoder size to 1-3M parameters, leaving the effects of scaling largely unexplored. To address this gap, we conduct a systematic study comparing two distinct strategies: scaling depth versus scaling width. We synthesize these strategies to construct a suite of 12 model configurations, spanning a parameter range from 1M to ~150M, and extensively evaluate their scaling behaviors across three critical dimensions: parameter efficiency, data efficiency, and compute efficiency. Our empirical results reveal that parameter count is insufficient to accurately predict the model performance, highlighting the critical and distinct roles of model depth (layer count) and width (embedding dimension). Crucially, we demonstrate that scaling depth yields superior performance gains to scaling width. Based on these findings, we provide and experimentally validate a set of design principles for the efficient allocation of parameters and compute resources to enhance the model performance.

Collaborative problem solving and learning are shaped by who or what is on the team. As large language models (LLMs) increasingly function as collaborators rather than tools, a key question is whether AI teammates can be aligned to express personality in predictable ways that matter for interaction and learning. We investigate AI personality alignment through a three-lens evaluation framework spanning self-perception (standardized self-report), behavioral expression (team dialogue), and reflective expression (memory construction). We first administered the Big Five Inventory (BFI-44) to LLM-based teammates across four providers (GPT-4o, Claude-3.7 Sonnet, Gemini-2.5 Pro, Grok-3), 32 high/low trait configurations, and multiple prompting strategies. LLMs produced sharply differentiated Big Five profiles, but prompt semantic richness added little beyond simple trait assignment, while provider differences and baseline "default" personalities were substantial. Role framing also mattered: several models refused the assessment without context, yet complied when framed as a collaborative teammate. We then simulated AI participation in authentic team transcripts using high-trait personas and analyzed both generated utterances and structured long-term memories with LIWC-22. Personality signals in conversation were generally subtle and most detectable for Extraversion, whereas memory representations amplified trait-specific signals, especially for Neuroticism, Conscientiousness, and Agreeableness; Openness remained difficult to elicit robustly. Together, results suggest that AI personality is measurable but multi-layered and context-dependent, and that evaluating personality-aligned AI teammates requires attention to memory and system-level design, not conversation-only behavior.

The automatic generation of medical reports utilizing Multimodal Large Language Models (MLLMs) frequently encounters challenges related to factual instability, which may manifest as the omission of findings or the incorporation of inaccurate information, thereby constraining their applicability in clinical settings. Current methodologies typically produce reports based directly on image features, which inherently lack a definitive factual basis. In response to this limitation, we introduce Fact-Flow, an innovative framework that separates the process of visual fact identification from the generation of reports. This is achieved by initially predicting clinical findings from the image, which subsequently directs the MLLM to produce a report that is factually precise. A pivotal advancement of our approach is a pipeline that leverages a Large Language Model (LLM) to autonomously create a dataset of labeled medical findings, effectively eliminating the need for expensive manual annotation. Extensive experimental evaluations conducted on two disease-focused medical datasets validate the efficacy of our method, demonstrating a significant enhancement in factual accuracy compared to state-of-the-art models, while concurrently preserving high standards of text quality.

Activation steering promises to be an extremely parameter-efficient form of adaptation, but its effectiveness depends on critical design choices -- such as intervention location and parameterization -- that currently rely on empirical heuristics rather than a principled foundation. We establish a first-order equivalence between activation-space interventions and weight-space updates, deriving the conditions under which activation steering can replicate fine-tuning behavior. This equivalence yields a principled framework for steering design and identifies the post-block output as a theoretically-backed and highly expressive intervention site. We further explain why certain intervention locations outperform others and show that weight updates and activation updates play distinct, complementary functional roles. This analysis motivates a new approach -- joint adaptation -- that trains in both spaces simultaneously. Our post-block steering achieves accuracy within 0.2%-0.9%$ of full-parameter tuning, on average across tasks and models, while training only 0.04% of model parameters. It consistently outperforms prior activation steering methods such as ReFT and PEFT approaches including LoRA, while using significantly fewer parameters. Finally, we show that joint adaptation often surpasses the performance ceilings of weight and activation updates in isolation, introducing a new paradigm for efficient model adaptation.

Counterfactual medical image generation have emerged as a critical tool for enhancing AI-driven systems in medical domain by answering "what-if" questions. However, existing approaches face two fundamental limitations: First, they fail to prevent unintended modifications, resulting collateral changes in demographic attributes when only disease features should be affected. Second, they lack interpretability in their editing process, which significantly limits their utility in real-world medical applications. To address these limitations, we present InstructX2X, a novel interpretable local editing model for counterfactual medical image generation featuring Region-Specific Editing. This approach restricts modifications to specific regions, effectively preventing unintended changes while simultaneously providing a Guidance Map that offers inherently interpretable visual explanations of the editing process. Additionally, we introduce MIMIC-EDIT-INSTRUCTION, a dataset for counterfactual medical image generation derived from expert-verified medical VQA pairs. Through extensive experiments, InstructX2X achieve state-of-the-art performance across all major evaluation metrics. Our model successfully generates high-quality counterfactual chest X-ray images along with interpretable explanations.

In-vivo environments, magnetically actuated soft robots offer advantages such as wireless operation and precise control, showing promising potential for painless detection and therapeutic procedures. We developed a trileg magnetically driven soft robot (TMR) whose multi-legged design enables more flexible gaits and diverse motion patterns. For the silicone made of reconfigurable soft robots, its navigation ability can be separated into sequential motions, namely squatting, rotation, lifting a leg, walking and so on. Its motion and behavior depend on its bending shapes. To bridge motion type description and specific low-level voltage control, we introduced TMR-VLA, an end-to-end multi-modal system for a trileg magnetic soft robot capable of performing hybrid motion types, which is promising for developing a navigation ability by adapting its shape to language-constrained motion types. The TMR-VLA deploys embodied endoluminal localization ability from EndoVLA, and fuses sequential frames and natural language commands as input. Low-level voltage output is generated based on the current observation state and specific motion type description. The result shows the TMR-VLA can predict how the voltage applied to TMR will change the dynamics of a silicon-made soft robot. The TMR-VLA reached a 74% average success rate.

Beyond binary classification, learnability can become a logically fragile notion: in EMX, even the class of all finite subsets of $[0,1]$ is learnable in some models of ZFC and not in others. We argue the paradox is operational. The standard definitions quantify over arbitrary set-theoretic learners that implicitly assume non-operational resources (infinite precision, unphysical data access, and non-representable outputs). We introduce physics-aware learnability (PL), which defines the learnability relative to an explicit access model -- a family of admissible physical protocols. Finite-precision coarse-graining reduces continuum EMX to a countable problem, via an exact pushforward/pullback reduction that preserves the EMX objective, making the independence example provably learnable with explicit $(ε,δ)$ sample complexity. For quantum data, admissible learners are exactly POVMs on $d$ copies, turning sample size into copy complexity and yielding Helstrom(-type) lower bounds. For finite no-signaling and quantum models, PL feasibility becomes linear or semidefinite and is therefore decidable.

Recommender systems (RecSys) are increasingly emphasizing scaling, leveraging larger architectures and more interaction data to improve personalization. Yet, despite the optimizer's pivotal role in training, modern RecSys pipelines almost universally default to Adam/AdamW, with limited scrutiny of whether these choices are truly optimal for recommendation. In this work, we revisit optimizer design for scalable recommendation and introduce MuonRec, the first framework that brings the recently proposed Muon optimizer to RecSys training. Muon performs orthogonalized momentum updates for 2D weight matrices via Newton-Schulz iteration, promoting diverse update directions and improving optimization efficiency. We develop an open-source training recipe for recommendation models and evaluate it across both traditional sequential recommenders and modern generative recommenders. Extensive experiments demonstrate that MuonRec reduces converged training steps by an average of 32.4\% while simultaneously improving final ranking quality. Specifically, MuonRec yields consistent relative gains in NDCG@10, averaging 12.6\% across all settings, with particularly pronounced improvements in generative recommendation models. These results consistently outperform strong Adam/AdamW baselines, positioning Muon as a promising new optimizer standard for RecSys training. Our code is available.

Deep neural networks (DNNs) enable high performance across domains but remain vulnerable to adversarial perturbations, limiting their use in safety-critical settings. Here, we introduce two quantum-optimization-based models for robust verification that reduce the combinatorial burden of certification under bounded input perturbations. For piecewise-linear activations (e.g., ReLU and hardtanh), our first model yields an exact formulation that is sound and complete, enabling precise identification of adversarial examples. For general activations (including sigmoid and tanh), our second model constructs scalable over-approximations via piecewise-constant bounds and is asymptotically complete, with approximation error vanishing as the segmentation is refined. We further integrate Quantum Benders Decomposition with interval arithmetic to accelerate solving, and propose certificate-transfer bounds that relate robustness guarantees of pruned networks to those of the original model. Finally, a layerwise partitioning strategy supports a quantum--classical hybrid workflow by coupling subproblems across depth. Experiments on robustness benchmarks show high certification accuracy, indicating that quantum optimization can serve as a principled primitive for robustness guarantees in neural networks with complex activations.

In this study, a novel idea, Uncertainty Structure Estimation (USE), a lightweight, algorithm-agnostic procedure that emphasizes the often-overlooked role of unlabeled data quality is introduced for Semi-supervised learning (SSL). SSL has achieved impressive progress, but its reliability in deployment is limited by the quality of the unlabeled pool. In practice, unlabeled data are almost always contaminated by out-of-distribution (OOD) samples, where both near-OOD and far-OOD can negatively affect performance in different ways. We argue that the bottleneck does not lie in algorithmic design, but rather in the absence of principled mechanisms to assess and curate the quality of unlabeled data. The proposed USE trains a proxy model on the labeled set to compute entropy scores for unlabeled samples, and then derives a threshold, via statistical comparison against a reference distribution, that separates informative (structured) from uninformative (structureless) samples. This enables assessment as a preprocessing step, removing uninformative or harmful unlabeled data before SSL training begins. Through extensive experiments on imaging (CIFAR-100) and NLP (Yelp Review) data, it is evident that USE consistently improves accuracy and robustness under varying levels of OOD contamination. Thus, it can be concluded that the proposed approach reframes unlabeled data quality control as a structural assessment problem, and considers it as a necessary component for reliable and efficient SSL in realistic mixed-distribution environments.

Accurately solving time-dependent partial differential equations (PDEs) with neural networks remains challenging due to long-time error accumulation and the difficulty of enforcing general boundary conditions. We introduce TENG-BC, a high-precision neural PDE solver based on the Time-Evolving Natural Gradient, designed to perform under general boundary constraints. At each time step, TENG-BC performs a boundary-aware optimization that jointly enforces interior dynamics and boundary conditions, accommodating Dirichlet, Neumann, Robin, and mixed types within a unified framework. This formulation admits a natural-gradient interpretation, enabling stable time evolution without delicate penalty tuning. Across benchmarks over diffusion, transport, and nonlinear PDEs with various boundary conditions, TENG-BC achieves solver-level accuracy under comparable sampling budgets, outperforming conventional solvers and physics-informed neural network (PINN) baselines.

Meta-Reinforcement Learning (Meta-RL) commonly generalizes via smoothness in the task encoding. While this enables local generalization around each training task, it requires dense coverage of the task space and leaves richer task space structure untapped. In response, we develop a geometric perspective that endows the task space with a "hereditary geometry" induced by the inherent symmetries of the underlying system. Concretely, the agent reuses a policy learned at the train time by transforming states and actions through actions of a Lie group. This converts Meta-RL into symmetry discovery rather than smooth extrapolation, enabling the agent to generalize to wider regions of the task space. We show that when the task space is inherited from the symmetries of the underlying system, the task space embeds into a subgroup of those symmetries whose actions are linearizable, connected, and compact--properties that enable efficient learning and inference at the test time. To learn these structures, we develop a differential symmetry discovery method. This collapses functional invariance constraints and thereby improves numerical stability and sample efficiency over functional approaches. Empirically, on a two-dimensional navigation task, our method efficiently recovers the ground-truth symmetry and generalizes across the entire task space, while a common baseline generalizes only near training tasks.

Hearables are becoming ubiquitous, yet their sound controls remain blunt: users can either enable global noise suppression or focus on a single target sound. Real-world acoustic scenes, however, contain many simultaneous sources that users may want to adjust independently. We introduce Aurchestra, the first system to provide fine-grained, real-time soundscape control on resource-constrained hearables. Our system has two key components: (1) a dynamic interface that surfaces only active sound classes and (2) a real-time, on-device multi-output extraction network that generates separate streams for each selected class, achieving robust performance for upto 5 overlapping target sounds, and letting users mix their environment by customizing per-class volumes, much like an audio engineer mixes tracks. We optimize the model architecture for multiple compute-limited platforms and demonstrate real-time performance on 6 ms streaming audio chunks. Across real-world environments in previously unseen indoor and outdoor scenarios, our system enables expressive per-class sound control and achieves substantial improvements in target-class enhancement and interference suppression. Our results show that the world need not be heard as a single, undifferentiated stream: with Aurchestra, the soundscape becomes truly programmable.

Inverse problems constrained by partial differential equations are often ill-conditioned due to noisy and incomplete data or inherent non-uniqueness. A prominent example is full waveform inversion, which estimates Earth's subsurface properties by fitting seismic measurements subject to the wave equation, where ill-conditioning is inherent to noisy, band-limited, finite-aperture data and shadow zones. Casting the inverse problem into a Bayesian framework allows for a more comprehensive description of its solution, where instead of a single estimate, the posterior distribution characterizes non-uniqueness and can be sampled to quantify uncertainty. However, no clear procedure exists for translating hard physical constraints, such as the wave equation, into prior distributions amenable to existing sampling techniques. To address this, we perform posterior sampling in the dual space using an augmented Lagrangian formulation, which translates hard constraints into penalties amenable to sampling algorithms while ensuring their exact satisfaction. We achieve this by seamlessly integrating the alternating direction method of multipliers (ADMM) with Stein variational gradient descent (SVGD) -- a particle-based sampler -- where the constraint is relaxed at each iteration and multiplier updates progressively enforce satisfaction. This enables constrained posterior sampling while inheriting the favorable conditioning properties of dual-space solvers, where partial constraint relaxation allows productive updates even when the current model is far from the true solution. We validate the method on a stylized Rosenbrock conditional inference problem and on frequency-domain full waveform inversion for a Gaussian anomaly model and the Marmousi~II benchmark, demonstrating well-calibrated uncertainty estimates and posterior contraction with increasing data coverage.

Colluding language-model agents can hide coordination in messages that remain policy-compliant at the surface level. We present CLBC, a protocol where generation and admission are separated: a message is admitted to transcript state only if a small verifier accepts a proof-bound envelope under a pinned predicate $Π$. The predicate binds policy hash, public randomness schedule, transcript chaining, latent schema constraints, canonical metadata/tool fields, and deterministic rejection codes. We show how this protocol yields an upper bound on transcript leakage in terms of latent leakage plus explicit residual channels, derive adaptive composition guarantees, and state a semantic lower bound when policy-valid alternatives remain choosable. We report extensive empirically grounded evidence: aggregate evaluation satisfies all prespecified thresholds; strict lane decoder advantage is bounded at 0.0000 with MI proxy 0.0636; adaptive-colluder stress tests remain below attacker thresholds; and baseline separation shows large gaps between reject-by-default semantics and audit-only controls. We further quantify operational tradeoffs. Strict full-proof mode has median turn latency 27.53s (p95 28.08s), while sampled proving reduces non-proved-turn latency to 0.327ms. The central finding is that bottlenecks alone are insufficient: security claims depend on verifiable admission semantics that are online, deterministic, and fail-closed.

Full Waveform Inversion (FWI) is a highly nonlinear and ill-posed problem that aims to recover subsurface velocity maps from surface-recorded seismic waveforms data. Existing data-driven FWI typically uses small models, as available datasets have limited volume, geological diversity, and spatial extent, leading to substantial concerns about overfitting. Although they perform well on synthetic datasets, current methods fail to generalize to more realistic geological structures. In this work, we show that a model trained entirely on simulated and relatively simple data can generalize remarkably well to challenging and unseen geological benchmarks. We provide a working recipe that tames a billion-parameter model for FWI through coordinated scaling across three axes: model capacity, data diversity, and training strategy. Our model achieves state-of-the-art performance on OpenFWI and significantly narrows the generalization gap in data-driven FWI. Across six challenging geophysical benchmarks, including Marmousi, 2D SEG/EAGE Salt and Overthrust, 2004 BP, Sigsbee, and SEAM Phase I, it infers complex structures absent from the training set and delivers significant performance improvements (SSIM from 0.5844 to 0.7669). Overall, our results demonstrate that with an appropriate scaling strategy, large models trained on simple synthetic data can achieve substantial generalization to more complex and realistic geological structures.

\textit{NeuroHex} is a hexagonal coordinate system designed to support highly efficient world models and reference frames for online adaptive AI systems. Inspired by the hexadirectional firing structure of grid cells in the human brain, NeuroHex adopts a cubic isometric hexagonal coordinate formulation that provides full 60° rotational symmetry and low-cost translation, rotation and distance computation. We develop a mathematical framework that incorporates ring indexing, quantized angular encoding, and a hierarchical library of foundational, simple, and complex geometric shape primitives. These constructs allow low-overhead point-in-shape tests and spatial matching operations that are expensive in Cartesian coordinate systems. To support realistic settings, the NeuroHex framework can process OpenStreetMap (OSM) data sets using an OSM-to-NeuroHex (\textit{OSM2Hex}) conversion tool. The OSM2Hex spatial abstraction processing pipeline can achieve a reduction of 90-99\% in geometric complexity while maintaining the relevant spatial structure map for navigation. Our initial results, based on actual city and neighborhood scale data sets, demonstrate that NeuroHex offers a highly efficient substrate for building dynamic world models to enable adaptive spatial reasoning in autonomous AI systems with continuous online learning capability.

Offline learning of strategies takes data efficiency to its extreme by restricting algorithms to a fixed dataset of state-action trajectories. We consider the problem in a mixed-motive multiagent setting, where the goal is to solve a game under the offline learning constraint. We first frame this problem in terms of selecting among candidate equilibria. Since datasets may inform only a small fraction of game dynamics, it is generally infeasible in offline game-solving to even verify a proposed solution is a true equilibrium. Therefore, we consider the relative probability of low regret (i.e., closeness to equilibrium) across candidates based on the information available. Specifically, we extend Policy Space Response Oracles (PSRO), an online game-solving approach, by quantifying game dynamics uncertainty and modifying the RL objective to skew towards solutions more likely to have low regret in the true game. We further propose a novel meta-strategy solver, tailored for the offline setting, to guide strategy exploration in PSRO. Our incorporation of Conservatism principles from Offline reinforcement learning approaches for strategy Exploration gives our approach its name: COffeE-PSRO. Experiments demonstrate COffeE-PSRO's ability to extract lower-regret solutions than state-of-the-art offline approaches and reveal relationships between algorithmic components empirical game fidelity, and overall performance.

Consider an organization whose users send requests in natural language to an AI system that fulfills them by carrying out specific tasks. In this paper, we consider the problem of ensuring such user requests comply with a list of diverse policies determined by the organization with the purpose of guaranteeing the safe and reliable use of the AI system. We propose, to the best of our knowledge, the first benchmark consisting of annotated user requests of diverse compliance with respect to a list of policies. Our benchmark is related to industrial applications in the technology sector. We then use our benchmark to evaluate the performance of various LLM models on policy compliance assessment under different solution methods. We analyze the differences on performance metrics across the models and solution methods, showcasing the challenging nature of our problem.

In this study, we present a meat freshness classification framework from Red-Green-Blue (RGB) images that supports both packaged and unpackaged meat datasets. The system classifies four in-distribution (ID) meat classes and uses an out-of-distribution (OOD)-aware abstention mechanism that flags low-confidence samples as No Result. The pipeline combines U-Net-based segmentation with deep feature classifiers. Segmentation is used as a preprocessing step to isolate the meat region and reduce background, producing more consistent inputs for classification. The segmentation module achieved an Intersection over Union (IoU) of 75% and a Dice coefficient of 82%, producing standardized inputs for the classification stage. For classification, we benchmark five backbones: Residual Network-50 (ResNet-50), Vision Transformer-Base/16 (ViT-B/16), Swin Transformer-Tiny (Swin-T), EfficientNet-B0, and MobileNetV3-Small. We use nested 5x3 cross-validation (CV) for model selection and hyperparameter tuning. On the held-out ID test set, EfficientNet-B0 achieves the highest accuracy (98.10%), followed by ResNet-50 and MobileNetV3-Small (both 97.63%) and Swin-T (97.51%), while ViT-B/16 is lower (94.42%). We additionally evaluate OOD scoring and thresholding using standard OOD metrics and sensitivity analysis over the abstention threshold. Finally, we report on-device latency using TensorFlow Lite (TFLite) on a smartphone, highlighting practical accuracy-latency trade-offs for future deployment.

Large language models such as GPT and Llama are trained with a next-token prediction loss. In this work, we suggest that training language models to predict multiple future tokens at once results in higher sample efficiency. More specifically, at each position in the training corpus, we ask the model to predict the following n tokens using n independent output heads, operating on top of a shared model trunk. Considering multi-token prediction as an auxiliary training task, we measure improved downstream capabilities with no overhead in training time for both code and natural language models. The method is increasingly useful for larger model sizes and keeps its appeal when training for multiple epochs. Gains are especially pronounced on generative benchmarks like coding, where our models consistently outperform strong baselines by several percentage points. Our 13B parameter models solves 12 % more problems on HumanEval and 17 % more on MBPP than comparable next-token models. Experiments on small algorithmic tasks demonstrate that multi-token prediction is favorable for the development of induction heads and algorithmic reasoning capabilities. As an additional benefit, models trained with 4-token prediction are up to 3X times faster at inference, even with large batch sizes.

Distribution shifts in attack patterns within RPL-based IoT networks pose a critical threat to the reliability and security of large-scale connected systems. Intrusion Detection Systems (IDS) trained on static datasets often fail to generalize to unseen threats and suffer from catastrophic forgetting when updated with new attacks. Ensuring continual adaptability of IDS is therefore essential for maintaining robust IoT network defence. In this focused study, we formulate intrusion detection as a domain continual learning problem and propose a method-agnostic IDS framework that can integrate diverse continual learning strategies. We systematically benchmark five representative approaches across multiple domain-ordering sequences using a comprehensive multi-attack dataset comprising 48 domains. Results show that continual learning mitigates catastrophic forgetting while maintaining a balance between plasticity, stability, and efficiency, a crucial aspect for resource-constrained IoT environments. Among the methods, Replay-based approaches achieve the best overall performance, while Synaptic Intelligence (SI) delivers near-zero forgetting with high training efficiency, demonstrating strong potential for stable and sustainable IDS deployment in dynamic IoT networks.

We propose KROM, a kernel-based reduced-order framework for fast solution of nonlinear partial differential equations. KROM formulates PDE solution as a minimum-norm (Gaussian-process) recovery problem in an RKHS, and accelerates the resulting kernel solves by sparsifying the precision matrix via sparse Cholesky factorization. A central ingredient is an empirical kernel constructed from a snapshot library of PDE solutions (generated under varying forcings, initial data, boundary data, or parameters). This snapshot-driven kernel adapts to problem-specific structure -- boundary behavior, oscillations, nonsmooth features, linear constraints, conservation and dissipation laws -- thereby reducing the dependence on hand-tuned stationary kernels. The resulting method yields an implicit reduced model: after sparsification, only a localized subset of effective degrees of freedom is used online. We report numerical results for semilinear elliptic equations, discontinuous-coefficient Darcy flow, viscous Burgers, Allen--Cahn, and two-dimensional Navier--Stokes, showing that empirical kernels can match or outperform Matérn baselines, especially in nonsmooth regimes. We also provide error bounds that separate discretization effects, snapshot-space approximation error, and sparse-Cholesky approximation error.

Linguistic insights may help make Large Language Model (LLM) training more efficient. We trained Meta's OPT model on the 100M word BabyLM dataset, and evaluated it on the BLiMP benchmark, which consists of 67 classes, each defined by sentence pairs that differ in a targeted syntactic or semantic rule violation. We tested the model's preference for grammatical over ungrammatical sentences across training iterations and grammatical types. In nearly one-third of the BLiMP classes, OPT fails to consistently assign a higher likelihood to grammatical sentences, even after extensive training. When it fails, it often establishes a clear (erroneous) separation of the likelihoods at an early stage of processing and sustains this to the end of our training phase. We hypothesize that this mis-categorization is costly because it creates entrenched biases that must, eventually, be reversed in order for the model to perform well. We probe this phenomenon using a mixture of qualitative (based on linguistic theory and the theory of Deep Learning) and quantitative (based on numerical testing) assessments. Our qualitative assessments indicate that only some BLiMP tests are meaningful guides. We conclude by articulating a hypothesis, the Bigram Hypothesis, which claims that the learning process will exhibit erroneous entrenchment if bigram statistics bias the model toward wrong distinctions early in training, and we describe a method (in progress) of testing the hypothesis on appropriately selected BLiMP classes.

In large-scale LLM pre-training systems with 100k+ GPUs, failures become the norm rather than the exception, and restart costs can dominate wall-clock training time. However, existing fault-tolerance mechanisms are largely unprepared for this restart-dominant regime. To address this challenge, we propose SPARe - Stacked Parallelism with Adaptive Reordering - a fault-tolerance framework that masks node failures during gradient synchronization by stacking redundant data shards across parallelism groups and adaptively reordering execution. SPARe achieves availability comparable to traditional replication while maintaining near-constant computation overhead of only 2~3x, even under high redundancy where traditional replication would require linearly inflating overhead. We derive closed-form expressions for endurable failure count and computation overhead, validate them via SimGrid-based discrete-event simulation, and jointly optimize redundancy and checkpointing to minimize time-to-train. At extreme scale with up to 600k GPUs, SPARe reduces time-to-train by 40~50% compared to traditional replication.

Multi-tenant AI inference platforms must balance resource utilization against service-level guarantees under variable demand. Conventional approaches fail to achieve this balance: dedicated endpoints strand capacity on idle models, while rate limits ignore the heterogeneous cost of inference requests. We introduce \emph{token pools}, a control-plane abstraction that represents inference capacity as explicit entitlements expressed in inference-native units (token throughput, KV cache, concurrency). Unlike rate limits, which govern request admission without regard to execution cost, token pools authorize both admission and autoscaling from the same capacity model, ensuring consistency between what is promised and what is provisioned. The abstraction captures burst modes across multiple dimensions invisible to conventional throttling. Dynamic per-entitlement limits on each burst dimension enable fine-grained control over resource consumption while permitting work-conserving backfill by low-priority traffic. The design supports priority-aware allocation, service tiers with differentiated guarantees, and debt-based fairness mechanisms, all without modifying the underlying inference runtime or cluster scheduler. In experiments on a Kubernetes cluster with vLLM backends, token pools maintain a bounded P99 latency for guaranteed workloads during overload by selectively throttling spot traffic, while a baseline without admission control experiences unbounded latency degradation across all workloads. A second experiment demonstrates debt-based fair-share convergence among elastic workloads with heterogeneous SLO requirements during capacity scarcity.

Listening to heart and lung sounds - auscultation - is one of the first and most fundamental steps in a clinical examination. Despite being fast and non-invasive, it demands years of experience to interpret subtle audio cues. Recent deep learning methods have made progress in automating cardiopulmonary sound analysis, yet most are restricted to simple classification and offer little clinical interpretability or decision support. We present StethoLM, the first audio-language model specialized for cardiopulmonary auscultation, capable of performing instruction-driven clinical tasks across the full spectrum of auscultation analysis. StethoLM integrates audio encoding with a medical language model backbone and is trained on StethoBench, a comprehensive benchmark comprising 77,027 instruction-response pairs synthesized from 16,125 labeled cardiopulmonary recordings spanning seven clinical task categories: binary classification, detection, reporting, reasoning, differential diagnosis, comparison, and location-based analysis. Through multi-stage training that combines supervised fine-tuning and direct preference optimization, StethoLM achieves substantial gains in performance and robustness on out-of-distribution data. Our work establishes a foundation for instruction-following AI systems in clinical auscultation.

Universal jamming grippers excel at grasping unknown objects due to their compliant bodies. Traditional tactile sensors can compromise this compliance, reducing grasping performance. We present acoustic sensing as a form of morphological sensing, where the gripper's soft body itself becomes the sensor. A speaker and microphone are placed inside the gripper cavity, away from the deformable membrane, fully preserving compliance. Sound propagates through the gripper and object, encoding object properties, which are then reconstructed via machine learning. Our sensor achieves high spatial resolution in sensing object size (2.6 mm error) and orientation (0.6 deg error), remains robust to external noise levels of 80 dBA, and discriminates object materials (up to 100% accuracy) and 16 everyday objects (85.6% accuracy). We validate the sensor in a realistic tactile object sorting task, achieving 53 minutes of uninterrupted grasping and sensing, confirming the preserved grasping performance. Finally, we demonstrate that disentangled acoustic representations can be learned, improving robustness to irrelevant acoustic variations.

The prevailing paradigm in artificial intelligence research equates progress with scale: larger models trained on broader datasets are presumed to yield superior capabilities. This assumption, while empirically productive for general-purpose applications, obscures a fundamental epistemological tension between breadth and depth of knowledge. We introduce the concept of \emph{Monotropic Artificial Intelligence} -- language models that deliberately sacrifice generality to achieve extraordinary precision within narrowly circumscribed domains. Drawing on the cognitive theory of monotropism developed to understand autistic cognition, we argue that intense specialization represents not a limitation but an alternative cognitive architecture with distinct advantages for safety-critical applications. We formalize the defining characteristics of monotropic models, contrast them with conventional polytropic architectures, and demonstrate their viability through Mini-Enedina, a 37.5-million-parameter model that achieves near-perfect performance on Timoshenko beam analysis while remaining deliberately incompetent outside its domain. Our framework challenges the implicit assumption that artificial general intelligence constitutes the sole legitimate aspiration of AI research, proposing instead a cognitive ecology in which specialized and generalist systems coexist complementarily.

Real-world scenarios increasingly require multiple embodied agents to collaborate in dynamic environments under embodied constraints, as many tasks exceed the capabilities of any single agent. Recent advances in large language models (LLMs) enable high-level cognitive coordination through reasoning, planning, and natural language communication. However, fine-grained analyses of how such collaboration emerges, unfolds, and contributes to task success in embodied multi-agent systems are difficult to conduct with existing benchmarks. In this paper, we introduce EmCoop, a benchmark framework for studying cooperation in LLM-based embodied multi-agent systems. Our framework separates a high-level cognitive layer from a low-level embodied interaction layer, allowing us to characterize agent cooperation through their interleaved dynamics over time. Given a cooperation-constrained embodied task, we propose generalizable, process-level metrics that diagnose collaboration quality and failure modes, beyond final task success. We instantiate our framework in two embodied environments that scale to arbitrary numbers of agents and support diverse communication topologies, and use these instantiations to demonstrate how EmCoop enables systematic analysis of cooperation dynamics across team sizes and task settings. The project web page can be found at: https://happyeureka.github.io/emcoop.

With the increase in Internet censorship globally, various circumvention tools have been designed and developed. However, the monetary cost of these tools deeply impacts both user choice and the sustainability of provider operations. Recent developments in censorship circumvention research attempted to achieve cost efficiency by utilizing Infrastructure-as-a-Service (IaaS) spot instances as bridges, but still incurred substantial expenses related to network connectivity and instance maintenance. In this work, we present CensorLess, a circumvention proxy built leveraging the unique benefits of a serverless platform. CensorLess comprises three components: a local proxy that handles client-side communication and ensures compliance with serverless functions' security restrictions, a function refresher that periodically regenerates bridges, and a live migration mechanism that maintains continuous connectivity. CensorLess inherits the serverless platform's cost efficiency, ephemerality, scalability, concurrency, and performance. Compared to existing low-cost, state-of-the-art circumvention techniques, CensorLess reduces costs by 97%, while simultaneously enabling robust censorship resistance by employing bridge rotation.

Data valuation methods quantify how individual training examples contribute to a model's behavior, and are increasingly used for dataset curation, auditing, and emerging data markets. As these techniques become operational, they raise serious privacy concerns: valuation scores can reveal whether a person's data was included in training, whether it was unusually influential, or what sensitive patterns exist in proprietary datasets. This motivates the study of privacy-preserving data valuation. However, privacy is fundamentally in tension with valuation utility under differential privacy (DP). DP requires outputs to be insensitive to any single record, while valuation methods are explicitly designed to measure per-record influence. As a result, naive privatization often destroys the fine-grained distinctions needed to rank or attribute value, particularly in heterogeneous datasets where rare examples exert outsized effects. In this work, we analyze the feasibility of DP-compatible data valuation. We identify the core algorithmic primitives across common valuation frameworks that induce prohibitive sensitivity, explaining why straightforward DP mechanisms fail. We further derive design principles for more privacy-amenable valuation procedures and empirically characterize how privacy constraints degrade ranking fidelity across representative methods and datasets. Our results clarify the limits of current approaches and provide a foundation for developing valuation methods that remain useful under rigorous privacy guarantees.

Transportation mode detection is an important topic within GeoAI and transportation research. In this study, we introduce SpeedTransformer, a novel Transformer-based model that relies solely on speed inputs to infer transportation modes from dense smartphone GPS trajectories. In benchmark experiments, SpeedTransformer outperformed traditional deep learning models, such as the Long Short-Term Memory (LSTM) network. Moreover, the model demonstrated strong flexibility in transfer learning, achieving high accuracy across geographical regions after fine-tuning with small datasets. Finally, we deployed the model in a real-world experiment, where it consistently outperformed baseline models under complex built environments and high data uncertainty. These findings suggest that Transformer architectures, when combined with dense GPS trajectories, hold substantial potential for advancing transportation mode detection and broader mobility-related research.

READMEs shape first impressions of software projects, yet what constitutes a good README varies across audiences and contexts. Research software needs reproducibility details, while open-source libraries might prioritize quick-start guides. Through a design probe, LintMe, we explore how linting can be used to improve READMEs given these diverse contexts, aiding style and content issues while preserving authorial agency. Users create context-specific checks using a lightweight DSL that uses a novel combination of programmatic operations (e.g., for broken links) with LLM-based content evaluation (e.g., for detecting jargon), yielding checks that would be challenging for prior linters. Through a user study (N=11), comparison with naive LLM usage, and an extensibility case study, we find that our design is approachable, flexible, and well matched with the needs of this domain. This work opens the door for linting more complex documentation and other culturally mediated text-based documents.

Classification using sparse oblique random forests provides guarantees on uncertainty and confidence while controlling for specific error types. However, they use more data and more compute than other tree ensembles because they create deep trees and need to sort or histogram linear combinations of data at runtime. We provide a method for dynamically switching between histograms and sorting to find the best split. We further optimize histogram construction using vector intrinsics. Evaluating this on large datasets, our optimizations speedup training by 1.7-2.5x compared to existing oblique forests and 1.5-2x compared to standard random forests. We also provide a GPU and hybrid CPU-GPU implementation.

As AI agents increasingly perform economic tasks on behalf of humans, a fundamental tension arises between agent autonomy and human control over financial assets. We present the Agent Economic Sovereignty Protocol (AESP), a layered protocol in which agents transact autonomously at machine speed on crypto-native infrastructure while remaining cryptographically bound to human-defined governance boundaries. AESP enforces the invariant that agents are economically capable but never economically sovereign through five mechanisms: (1) a deterministic eight-check policy engine with tiered escalation; (2) human-in-the-loop review with automatic, explicit, and biometric tiers; (3) EIP-712 dual-signed commitments with escrow; (4) HKDF-based context-isolated privacy with batched consolidation; and (5) an ACE-GF-based cryptographic substrate. We formalize two testable hypotheses on security coverage and latency overhead, and specify a complete evaluation methodology with baselines and ablation design. The protocol is implemented as an open-source TypeScript SDK (208 tests, ten modules) with interoperability via MCP and A2A.

This paper details the baseline model selection, fine-tuning process, evaluation methods, and the implications of deploying more accurate LLMs in healthcare settings. As large language models (LLMs) are increasingly employed to address diverse problems, including medical queries, concerns about their reliability have surfaced. A recent study by Long Island University highlighted that LLMs often perform poorly in medical contexts, potentially leading to harmful misguidance for users. To address this, our research focuses on fine-tuning the Llama 2 7B, a transformer-based, decoder-only model, using transcripts from real patient-doctor interactions. Our objective was to enhance the model's accuracy and precision in responding to medical queries. We fine-tuned the model using a supervised approach, emphasizing domain-specific nuances captured in the training data. In the best scenario, the model results should be reviewed and evaluated by real medical experts. Due to resource constraints, the performance of the fine-tuned model was evaluated using text similarity metrics. The fine-tuned model demonstrated significant improvements across all key dimensions except GPT-4's evaluation. The evaluations of ChatGPT4 are quite different from the quantitative results; here, we not only suggest, but also propose that the result should be evaluated by human medical experts.

While multimodal large language models offer a promising solution to the "black box" nature of health AI by generating interpretable reasoning traces, verifying the validity of these traces remains a critical challenge. Existing evaluation methods are either unscalable, relying on manual clinician review, or superficial, utilizing proxy metrics (e.g. QA) that fail to capture the semantic correctness of clinical logic. In this work, we introduce a reproducible framework for evaluating reasoning in ECG signals. We propose decomposing reasoning into two distinct, components: (i) Perception, the accurate identification of patterns within the raw signal, and (ii) Deduction, the logical application of domain knowledge to those patterns. To evaluate Perception, we employ an agentic framework that generates code to empirically verify the temporal structures described in the reasoning trace. To evaluate Deduction, we measure the alignment of the model's logic against a structured database of established clinical criteria in a retrieval-based approach. This dual-verification method enables the scalable assessment of "true" reasoning capabilities.

Software engineers use regular expressions (regexes) across a wide range of domains and tasks. To support regexes, software projects must integrate a regex engine, whether provided natively by the language runtime (e.g., Python's re) or included as an external dependency (e.g., PCRE). However, these engines may contain bugs and introduce vulnerabilities. A common strategy for testing regex engines involves differential testing -- comparing outputs across different implementations. However, this approach is concerning because regex syntax and semantics vary significantly between dialects (e.g., POSIX vs. PCRE). Fuzzing is also utilized to ease testing of feature-rich regex implementations to expose defects, but naive byte-level mutations generate syntactically invalid inputs that exercise only parsing logic, not matching internals. In this work, we describe our progress towards ReTest, a framework that systematically tests regular expression engines by combining grammar-aware fuzzing for high code coverage with metamorphic testing to generate dialect-independent test oracles. So far, we have surveyed testing practices across 22 regex engines, analyzed 1,007 regex engine bugs and 156 CVEs to characterize failure modes, and curated 16 metamorphic relations for regexes derived from Kleene algebra. Our preliminary evaluation on PCRE shows that ReTest achieves 3x higher edge coverage than existing fuzzing approaches and has identified three new memory safety defects. We conclude by describing our next steps toward our ultimate goal: helping regex engine developers identify bugs without depending on a consistent cross-implementation standard.

The increasingly popular agentic AI paradigm promises to harness the power of multiple, general-purpose large language model (LLM) agents to collaboratively complete complex tasks. While many agentic AI systems utilize predefined workflows or agent roles in order to reduce complexity, ideally these agents would be truly autonomous, able to achieve emergent collaboration even as the number of collaborating agents increases. Yet in practice, such unstructured interactions can lead to redundant work and cascading failures that are difficult to interpret or correct. In this work, we study multi-agent systems composed of general-purpose LLM agents that operate without predefined roles, control flow, or communication constraints, relying instead on emergent collaboration to solve problems. We introduce the Dynamic Interaction Graph (DIG), which captures emergent collaboration as a time-evolving causal network of agent activations and interactions. DIG makes emergent collaboration observable and explainable for the first time, enabling real-time identification, explanation, and correction of collaboration-induced error patterns directly from agents' collaboration paths. Thus, DIG fills a critical gap in understanding how general LLM agents solve problems together in truly agentic multi-agent systems. The project webpage can be found at: https://happyeureka.github.io/dig.

We test whether a geometric hallucination taxonomy -- classifying failures as center-drift (Type~1), wrong-well convergence (Type~2), or coverage gaps (Type~3) -- can distinguish hallucination types through controlled induction in GPT-2. Using a two-level statistical design with prompts ($N = 15$/group) as the unit of inference, we run each experiment 20 times with different generation seeds to quantify result stability. In static embeddings, Type~3 norm separation is robust (significant in 18/20 runs, Holm-corrected in 14/20, median $r = +0.61$). In contextual hidden states, the Type~3 norm effect direction is stable (19/20 runs) but underpowered at $N = 15$ (significant in 4/20, median $r = -0.28$). Types~1 and~2 do not separate in either space (${\leq}\,3/20$ runs). Token-level tests inflate significance by 4--16$\times$ through pseudoreplication -- a finding replicated across all 20 runs. The results establish coverage-gap hallucinations as the most geometrically distinctive failure mode, carried by magnitude rather than direction, and confirm the Type~1/2 non-separation as genuine at 124M parameters.

Chain-of-Thought (CoT) prompting is a widely used inference-time technique for improving reasoning, yet its gains are uneven across tasks. We analyze when and why CoT helps by modeling the step-wise reasoning trajectory as a Markov chain. Each intermediate step is a state and the dependence between steps is captured by a transition kernel. Our theory identifies transition alignment, whether instances share a common step-wise transition kernel, as the key determinant of CoT's effectiveness. When transitions are identical across steps, CoT reduces inference-time sample complexity: fewer context sample trajectories suffice to recover the final decision. In contrast, when transitions differ across steps, these gains can vanish. We further quantify how noise in intermediate steps modulates CoT's benefit. Beyond theory, we design synthetic benchmarks that isolate these factors to complement prior results on real-world tasks and to empirically validate our predictions.

Differentiable logic gate networks (DLGNs) learn compact, interpretable Boolean circuits via gradient-based training, but all existing variants are restricted to the 16 two-input binary gates. Extending DLGNs to Ternary Kleene $K_3$ logic and training DTLGNs where the UNKNOWN state enables principled abstention under uncertainty is desirable. However, the support set of potential gates per neuron explodes to $19{,}683$, making the established softmax-over-gates training approach intractable. We introduce Polynomial Surrogate Training (PST), which represents each ternary neuron as a degree-$(2,2)$ polynomial with 9 learnable coefficients (a $2{,}187\times$ parameter reduction) and prove that the gap between the trained network and its discretized logic circuit is bounded by a data-independent commitment loss that vanishes at convergence. Scaling experiments from 48K to 512K neurons on CIFAR-10 demonstrate that this hardening gap contracts with overparameterization. Ternary networks train $2$-$3\times$ faster than binary DLGNs and discover true ternary gates that are functionally diverse. On synthetic and tabular tasks we find that the UNKNOWN output acts as a Bayes-optimal uncertainty proxy, enabling selective prediction in which ternary circuits surpass binary accuracy once low-confidence predictions are filtered. More broadly, PST establishes a general polynomial-surrogate methodology whose parameterization cost grows only quadratically with logic valence, opening the door to many-valued differentiable logic.

Large reasoning models improve with more test-time computation, but often overthink, producing unnecessarily long chains-of-thought that raise cost without improving accuracy. Prior reinforcement learning approaches typically rely on a single outcome reward with trajectory-level length penalties, which cannot distinguish essential from redundant reasoning steps and therefore yield blunt compression. Although recent work incorporates step-level signals, such as offline pruning, supervised data construction, or verifier-based intermediate rewards, reasoning length is rarely treated as an explicit step-level optimization objective during RL. We propose Step-wise Adaptive Penalization (SWAP), a fine-grained framework that allocates length reduction across steps based on intrinsic contribution. We estimate step importance from the model's on-policy log-probability improvement toward the correct answer, then treat excess length as a penalty mass redistributed to penalize low-importance steps more heavily while preserving high-importance reasoning. We optimize with a unified outcome-process advantage within group-relative policy optimization. Extensive experiments demonstrate that SWAP reduces reasoning length by 64.3% on average while improving accuracy by 5.7% relative to the base model.

We introduce a scalable Gaussian process (GP) framework with deep product kernels for data-driven learning of parametrized spatio-temporal fields over fixed or parameter-dependent domains. The proposed framework learns a continuous representation, enabling predictions at arbitrary spatio-temporal coordinates, independent of the training data resolution. We leverage Kronecker matrix algebra to formulate a computationally efficient training procedure with complexity that scales nearly linearly with the total number of spatio-temporal grid points. A key feature of our approach is the efficient computation of the posterior variance at essentially the same computational cost as the posterior mean (exactly for Cartesian grids and via rigorous bounds for unstructured grids), thereby enabling scalable uncertainty quantification. Numerical studies on a range of benchmark problems demonstrate that the proposed method achieves accuracy competitive with operator learning methods such as Fourier neural operators and deep operator networks. On the one-dimensional unsteady Burgers' equation, our method surpasses the accuracy of projection-based reduced-order models. These results establish the proposed framework as an effective tool for data-driven surrogate modeling, particularly when uncertainty estimates are required for downstream tasks.

Evaluating AI agents in finance faces two key challenges: static benchmarks require costly expert annotation yet miss the dynamic decision-making central to real-world trading, while LLM-based judges introduce uncontrolled variance on domain-specific tasks. We introduce TraderBench, a benchmark that addresses both issues. It combines expert-verified static tasks (knowledge retrieval, analytical reasoning) with adversarial trading simulations scored purely on realized performance-Sharpe ratio, returns, and drawdown-eliminating judge variance entirely. The framework features two novel tracks: crypto trading with four progressive market-manipulation transforms, and options derivatives scoring across P&L accuracy, Greeks, and risk management. Trading scenarios can be refreshed with new market data to prevent benchmark contamination. Evaluating 13 models (8B open-source to frontier) on ~50 tasks, we find: (1) 8 of 13 models score ~33 on crypto with <1-point variation across adversarial conditions, exposing fixed non-adaptive strategies; (2) extended thinking helps retrieval (+26 points) but has zero impact on trading (+0.3 crypto, -0.1 options). These findings reveal that current agents lack genuine market adaptation, underscoring the need for performance-grounded evaluation in finance.

When large language models encounter conflicting information in context, which memories survive -- early or recent? We adapt classical interference paradigms from cognitive psychology to answer this question, testing 39 LLMs across diverse architectures and scales. Every model shows the same pattern: proactive interference (PI) dominates retroactive interference (RI) universally (Cohen's d = 1.73, p < 0.0001), meaning early encodings are protected at the cost of recent information -- the opposite of human memory, where RI typically dominates. Three findings indicate that RI and PI reflect separate memory mechanisms. RI and PI are uncorrelated (R^2 = 0.044), rejecting a unified "memory capacity." Model size predicts RI resistance (R^2 = 0.49) but not PI (R^2 = 0.06, n.s.) -- only RI is capacity-dependent. And error analysis reveals distinct failure modes: RI failures are passive retrieval failures (51%), while PI failures show active primacy intrusion (56%); both show <1% hallucination. These patterns parallel the consolidation-retrieval distinction in cognitive science, suggesting that transformer attention creates a primacy bias with direct implications for interference-heavy applications.

Misinformation spreading over the Internet poses a significant threat to both societies and individuals, necessitating robust and scalable fact-checking that relies on retrieving accurate and trustworthy evidence. Previous methods rely on semantic and social-contextual patterns learned from training data, which limits their generalization to new data distributions. Recently, Retrieval Augmented Generation (RAG) based methods have been proposed to utilize the reasoning capability of LLMs with retrieved grounding evidence documents. However, these methods largely rely on textual similarity for evidence retrieval and struggle to retrieve evidence that captures multi-hop semantic relations within rich document contents. These limitations lead to overlooking subtle factual correlations between the evidence and the claims to be fact-checked during evidence retrieval, thus causing inaccurate veracity predictions. To address these issues, we propose WKGFC, which exploits authorized open knowledge graph as a core resource of evidence. LLM-enabled retrieval is designed to assess the claims and retrieve the most relevant knowledge subgraphs, forming structured evidence for fact verification. To augment the knowledge graph evidence, we retrieve web contents for completion. The above process is implemented as an automatic Markov Decision Process (MDP): A reasoning LLM agent decides what actions to take according to the current evidence and the claims. To adapt the MDP for fact-checking, we use prompt optimization to fine-tune the agentic LLM.

Protein language models (pLMs) have recently gained significant attention for their ability to uncover relationships between sequence, structure, and function from evolutionary statistics, thereby accelerating therapeutic drug discovery. These models learn from large protein databases that are continuously updated by the biology community and whose dynamic nature motivates the application of continual learning, not only to keep up with the ever-growing data, but also as an opportunity to take advantage of the temporal meta-information that is created during this process. As a result, we introduce the Continual Pretraining of Protein Language Models (CoPeP) benchmark, a novel benchmark for evaluating continual learning approaches on pLMs. Specifically, we curate a sequence of protein datasets derived from the UniProt Knowledgebase spanning a decade and define metrics to assess pLM performance across 31 protein understanding tasks. We evaluate several methods from the continual learning literature, including replay, unlearning, and plasticity-based methods, some of which have never been applied to models and data of this scale. Our findings reveal that incorporating temporal meta-information improves perplexity by up to 7% even when compared to training on data from all tasks jointly. Moreover, even at scale, several continual learning methods outperform naive continual pretraining. The CoPeP benchmark offers an exciting opportunity to study these methods at scale in an impactful real-world application.

Learned field transformations may help address ubiquitous critical slowing down and signal-to-noise problems in lattice field theory. In the context of an annealed sequence of distributions, field transformations are defined by integrating flow fields that exactly solve a local transport problem. These proceedings discuss a new Monte Carlo approach to evaluating these flow fields, which can then be used directly in such contexts or as a means of generating unbiased training data for machine learning approaches. By defining the Monte Carlo estimator using coupled Langevin noise, the statistical noise in the required integrals is significantly mitigated. Demonstrations of the method include a U(1) transport problem and an SU(N) glueball correlator.

Modern engineering design platforms excel at discipline-specific tasks such as CAD, CAM, and CAE, but often lack native systems engineering frameworks. This creates a disconnect where system-level requirements and architectures are managed separately from detailed component design, hindering holistic development and increasing integration risks. To address this, we present the conceptual framework for the GenAI Workbench, a Model-Based Systems Engineering (MBSE) environment that integrates systems engineering principles into the designer's workflow. Built on an open-source PLM platform, it establishes a unified digital thread by linking semantic data from documents, physical B-rep geometry, and relational system graphs. The workbench facilitates an AI-assisted workflow where a designer can ingest source documents, from which the system automatically extracts requirements and uses vision-language models to generate an initial system architecture, such as a Design Structure Matrix (DSM). This paper presents the conceptual architecture, proposed methodology, and anticipated impact of this work-in-progress framework, which aims to foster a more integrated, data-driven, and informed engineering design methodology.

Quantum generative modeling is a rapidly evolving discipline at the intersection of quantum computing and machine learning. Contemporary quantum machine learning is generally limited to toy examples or heavily restricted datasets with few elements. This is not only due to the current limitations of available quantum hardware but also due to the absence of inductive biases arising from application-agnostic designs. Current quantum solutions must resort to tricks to scale down high-resolution images, such as relying heavily on dimensionality reduction or utilizing multiple quantum models for low-resolution image patches. Building on recent developments in classical image loading to quantum computers, we circumvent these limitations and train quantum Wasserstein GANs on the established classical MNIST and Fashion-MNIST datasets. Using the complete datasets, our system generates full-resolution images across all ten classes and establishes a new state-of-the-art performance with a single end-to-end quantum generator without tricks. As a proof-of-principle, we also demonstrate that our approach can be extended to color images, exemplified on the Street View House Numbers dataset. We analyze how the choice of variational circuit architecture introduces inductive biases, which crucially unlock this performance. Furthermore, enhanced noise input techniques enable highly diverse image generation while maintaining quality. Finally, we show promising results even under quantum shot noise conditions.

Scaling video generation from seconds to minutes faces a critical bottleneck: while short-video data is abundant and high-fidelity, coherent long-form data is scarce and limited to narrow domains. To address this, we propose a training paradigm where Mode Seeking meets Mean Seeking, decoupling local fidelity from long-term coherence based on a unified representation via a Decoupled Diffusion Transformer. Our approach utilizes a global Flow Matching head trained via supervised learning on long videos to capture narrative structure, while simultaneously employing a local Distribution Matching head that aligns sliding windows to a frozen short-video teacher via a mode-seeking reverse-KL divergence. This strategy enables the synthesis of minute-scale videos that learns long-range coherence and motions from limited long videos via supervised flow matching, while inheriting local realism by aligning every sliding-window segment of the student to a frozen short-video teacher, resulting in a few-step fast long video generator. Evaluations show that our method effectively closes the fidelity-horizon gap by jointly improving local sharpness, motion and long-range consistency. Project website: https://primecai.github.io/mmm/.

The fast-growing demands in using Large Language Models (LLMs) to tackle complex multi-step data science tasks create an emergent need for accurate benchmarking. There are two major gaps in existing benchmarks: (i) the lack of standardized, process-aware evaluation that captures instruction adherence and process fidelity, and (ii) the scarcity of accurately labeled training data. To bridge these gaps, we introduce DARE-bench, a benchmark designed for machine learning modeling and data science instruction following. Unlike many existing benchmarks that rely on human- or model-based judges, all tasks in DARE-bench have verifiable ground truth, ensuring objective and reproducible evaluation. To cover a broad range of tasks and support agentic tools, DARE-bench consists of 6,300 Kaggle-derived tasks and provides both large-scale training data and evaluation sets. Extensive evaluations show that even highly capable models such as gpt-o4-mini struggle to achieve good performance, especially in machine learning modeling tasks. Using DARE-bench training tasks for fine-tuning can substantially improve model performance. For example, supervised fine-tuning boosts Qwen3-32B's accuracy by 1.83x and reinforcement learning boosts Qwen3-4B's accuracy by more than 8x. These significant improvements verify the importance of DARE-bench both as an accurate evaluation benchmark and critical training data.

Multi-turn interactions with large language models typically retain the assistant's own past responses in the conversation history. In this work, we revisit this design choice by asking whether large language models benefit from conditioning on their own prior responses. Using in-the-wild, multi-turn conversations, we compare standard (full-context) prompting with a user-turn-only prompting approach that omits all previous assistant responses, across three open reasoning models and one state-of-the-art model. To our surprise, we find that removing prior assistant responses does not affect response quality on a large fraction of turns. Omitting assistant-side history can reduce cumulative context lengths by up to 10x. To explain this result, we find that multi-turn conversations consist of a substantial proportion (36.4%) of self-contained prompts, and that many follow-up prompts provide sufficient instruction to be answered using only the current user turn and prior user turns. When analyzing cases where user-turn-only prompting substantially outperforms full context, we identify instances of context pollution, in which models over-condition on their previous responses, introducing errors, hallucinations, or stylistic artifacts that propagate across turns. Motivated by these findings, we design a context-filtering approach that selectively omits assistant-side context. Our findings suggest that selectively omitting assistant history can improve response quality while reducing memory consumption.

GPU kernel optimization is fundamental to modern deep learning but remains a highly specialized task requiring deep hardware expertise. Despite strong performance in general programming, large language models (LLMs) remain uncompetitive with compiler-based systems such as torch.compile for CUDA kernel generation. Existing CUDA code generation approaches either rely on training-free refinement or fine-tune models within fixed multi-turn execution-feedback loops, but both paradigms fail to fundamentally improve the model's intrinsic CUDA optimization ability, resulting in limited performance gains. We present CUDA Agent, a large-scale agentic reinforcement learning system that develops CUDA kernel expertise through three components: a scalable data synthesis pipeline, a skill-augmented CUDA development environment with automated verification and profiling to provide reliable reward signals, and reinforcement learning algorithmic techniques enabling stable training. CUDA Agent achieves state-of-the-art results on KernelBench, delivering 100\%, 100\%, and 92\% faster rate over torch.compile on KernelBench Level-1, Level-2, and Level-3 splits, outperforming the strongest proprietary models such as Claude Opus 4.5 and Gemini 3 Pro by about 40\% on the hardest Level-3 setting.

Modern optimizers like Adam and Muon are central to training large language models, but their reliance on first- and second-order momenta introduces significant memory overhead, which constrains scalability and computational efficiency. In this work, we reframe the exponential moving average (EMA) used in these momenta as the training of a linear regressor via online gradient flow. Building on this equivalence, we introduce LoRA-Pre, a novel low-rank optimizer designed for efficient pre-training. Specifically, LoRA-Pre reduces the optimizer's memory footprint by decomposing the full momentum matrix into a compact low-rank subspace within the online linear learner, thereby maintaining optimization performance while improving memory efficiency. We empirically validate LoRA-Pre's efficacy by pre-training models from the Llama architecture family, scaling from 60M to 1B parameters. LoRA-Pre achieves the highest performance across all model sizes. Notably, LoRA-Pre demonstrates remarkable rank efficiency, achieving comparable or superior results using only 1/8 the rank of baseline methods. Beyond pre-training, we evaluate LoRA-Pre's effectiveness in fine-tuning scenarios. With the same rank, LoRA-Pre consistently outperforms all efficient fine-tuning baselines. Specifically, compared to standard LoRA, LoRA-Pre achieves substantial improvements of 3.14 points on Llama-3.1-8B and 6.17 points on Llama-2-7B, validating our approach's effectiveness across both pre-training and fine-tuning paradigms. Our code is publicly available at https://github.com/mrflogs/LoRA-Pre.

The Forensics Investigations Network in Digital Sciences (FINDS) Research Center of Excellence (CoE), funded by the U.S. Army Research Laboratory, advances Digital Forensic Engineering Education (DFEE) through an integrated research education framework for AI enabled cybersecurity workforce development. FINDS combines high performance computing (HPC), secure software engineering, adversarial analytics, and experiential learning to address emerging cyber and synthetic media threats. This paper introduces the Multidependency Capacity Building Skills Graph (MCBSG), a directed acyclic graph based model that encodes hierarchical and cross domain dependencies among competencies in AI-driven forensic programming, statistical inference, digital evidence processing, and threat detection. The MCBSG enables structured modeling of skill acquisition pathways and quantitative capacity assessment. Supervised machine learning methods, including entropy-based Decision Tree Classifiers and regression modeling, are applied to longitudinal multi cohort datasets capturing mentoring interactions, laboratory performance metrics, curriculum artifacts, and workshop participation. Feature importance analysis and cross validation identify key predictors of technical proficiency and research readiness. Three year statistical evaluation demonstrates significant gains in forensic programming accuracy, adversarial reasoning, and HPC-enabled investigative workflows. Results validate the MCBSG as a scalable, interpretable framework for data-driven, inclusive cybersecurity education aligned with national defense workforce priorities.

Medical coding translates clinical documentation into standardized codes for billing, research, and public health, but manual coding is time-consuming and error-prone. Existing automation efforts rely on small datasets that poorly represent real-world patient heterogeneity. We trained a language model on 5.8 million electronic health records from 1.8 million patients across nearly all specialties in Eastern Denmark (2006--2016) to predict ICD-10 codes from clinical notes, medications, and laboratory results. Evaluated on 270,000 held-out patients, the model achieved a micro F1 of 71.8% and a top-10 recall of 95.5%. Performance varied by specialty (F1: 53--91%), with higher scores in specialties with well-defined diagnostic criteria. Codes appearing predominantly as secondary diagnoses had markedly lower F1 scores. For three such codes (suicide-related behaviors, weight disorders, and hypertension), the model identified thousands of uncoded cases, of which 76-86% were confirmed valid upon manual review, suggesting systematic under-coding rather than model error. These findings suggest under-coding of secondary diagnoses in Eastern Denmark during this period, with potential implications for epidemiological research, public health surveillance, and understanding of multimorbidity. Similar time constraints and reimbursement structures in other healthcare systems suggest this may not be isolated to this dataset. The model can automate coding for approximately 50% of cases and provide accurate suggestions for most others, and may offer a practical solution to help capture missed secondary conditions.

Transformers have been established as the de-facto backbones for most recent advances in sequence modeling, mainly due to their growing memory capacity that scales with the context length. While plausible for retrieval tasks, it causes quadratic complexity and so has motivated recent studies to explore viable subquadratic recurrent alternatives. Despite showing promising preliminary results in diverse domains, such recurrent architectures underperform Transformers in recall-intensive tasks, often attributed to their fixed-size memory. In this paper, we introduce Memory Caching (MC), a simple yet effective technique that enhances recurrent models by caching checkpoints of their memory states (a.k.a. hidden states). Memory Caching allows the effective memory capacity of RNNs to grow with sequence length, offering a flexible trade-off that interpolates between the fixed memory (i.e., $O(L)$ complexity) of RNNs and the growing memory (i.e., $O(L^2)$ complexity) of Transformers. We propose four variants of MC, including gated aggregation and sparse selective mechanisms, and discuss their implications on both linear and deep memory modules. Our experimental results on language modeling, and long-context understanding tasks show that MC enhances the performance of recurrent models, supporting its effectiveness. The results of in-context recall tasks indicate that while Transformers achieve the best accuracy, our MC variants show competitive performance, close the gap with Transformers, and performs better than state-of-the-art recurrent models.

Identifiability in representation learning is commonly evaluated using standard metrics (e.g., MCC, DCI, R^2) on synthetic benchmarks with known ground-truth factors. These metrics are assumed to reflect recovery up to the equivalence class guaranteed by identifiability theory. We show that this assumption holds only under specific structural conditions: each metric implicitly encodes assumptions about both the data-generating process (DGP) and the encoder. When these assumptions are violated, metrics become misspecified and can produce systematic false positives and false negatives. Such failures occur both within classical identifiability regimes and in post-hoc settings where identifiability is most needed. We introduce a taxonomy separating DGP assumptions from encoder geometry, use it to characterise the validity domains of existing metrics, and release an evaluation suite for reproducible stress testing and comparison.

Many readers today struggle to assess the trustworthiness of online news because reliable reporting coexists with misinformation. The TREC 2025 DRAGUN (Detection, Retrieval, and Augmented Generation for Understanding News) Track provided a venue for researchers to develop and evaluate assistive RAG systems that support readers' news trustworthiness assessment by producing reader-oriented, well-attributed reports. As the organizers of the DRAGUN track, we describe the resources that we have newly developed to allow for the reuse of the track's tasks. The track had two tasks: (Task 1) Question Generation, producing 10 ranked investigative questions; and (Task 2, the main task) Report Generation, producing a 250-word report grounded in the MS MARCO V2.1 Segmented Corpus. As part of the track's evaluation, we had TREC assessors create importance-weighted rubrics of questions with expected short answers for 30 different news articles. These rubrics represent the information that assessors believe is important for readers to assess an article's trustworthiness. The assessors then used their rubrics to manually judge the participating teams' submitted runs. To make these tasks and their rubrics reusable, we have created an automated process to judge runs not part of the original assessing. We show that our AutoJudge ranks existing runs well compared to the TREC human-assessed evaluation (Kendall's $τ= 0.678$ for Task 1 and $τ= 0.872$ for Task 2). These resources enable both the evaluation of RAG systems for assistive news trustworthiness assessment and, with the human evaluation as a benchmark, research on improving automated RAG evaluation.

We propose a minimal agentic baseline that enables systematic comparison across different AI-based theorem prover architectures. This design implements the core features shared among state-of-the-art systems: iterative proof refinement, library search and context management. We evaluate our baseline using qualitatively different benchmarks and compare various popular models and design choices, and demonstrate competitive performance compared to state-of-the-art approaches, while using a significantly simpler architecture. Our results demonstrate consistent advantages of an iterative approach over multiple single-shot generations, especially in terms of sample efficiency and cost effectiveness. The implementation is released open-source as a candidate reference for future research and as an accessible prover for the community.

Processing-in-Memory (PIM) architectures enable computation directly within DRAM and help combat the memory wall problem. Bit-shifting is a fundamental operation that enables PIM applications such as shift-and-add multiplication, adders using carry propagation, and Galois field arithmetic used in cryptography algorithms like AES and Reed-Solomon error correction codes. Existing approaches to in-DRAM shifting require adding dedicated shifter circuits beneath the sense amplifiers to enable horizontal data movement across adjacent bitlines or vertical data layouts which store operand bits along a bitline to implement shifts as row-copy operations. In this paper, we propose a novel DRAM subarray design that enables in-DRAM bit-shifting for open-bitline architectures. In this new design, we built upon prior work that introduced a new type of cell used for row migration in asymmetric subarrays, called a "migration cell". We repurpose and extend the functionality by adding a row of migration cells at the top and bottom of each subarray which enables bidirectional bit-shifting within any given row. This new design maintains compatibility with standard DRAM operations. Unlike previous approaches to shifting, our design operates on horizontally-stored data, eliminating the need and overhead of data transposition, and our design leverages the existing cell structures, eliminating the need for additional complex logic and circuitry. We present an evaluation of our design that includes timing and energy analysis using NVMain, circuit-level validation of the in-DRAM shift operation using LTSPICE, and a VLSI layout implementation in Cadence Virtuoso.

Neural networks are hypothesized to implement interpretable causal mechanisms, yet verifying this requires finding a causal abstraction -- a simpler, high-level Structural Causal Model (SCM) faithful to the network under interventions. Discovering such abstractions is hard: it typically demands brute-force interchange interventions or retraining. We reframe the problem by viewing structured pruning as a search over approximate abstractions. Treating a trained network as a deterministic SCM, we derive an Interventional Risk objective whose second-order expansion yields closed-form criteria for replacing units with constants or folding them into neighbors. Under uniform curvature, our score reduces to activation variance, recovering variance-based pruning as a special case while clarifying when it fails. The resulting procedure efficiently extracts sparse, intervention-faithful abstractions from pretrained networks, which we validate via interchange interventions.

Compositional generalization, the ability to recognize familiar parts in novel contexts, is a defining property of intelligent systems. Although modern models are trained on massive datasets, they still cover only a tiny fraction of the combinatorial space of possible inputs, raising the question of what structure representations must have to support generalization to unseen combinations. We formalize three desiderata for compositional generalization under standard training (divisibility, transferability, stability) and show they impose necessary geometric constraints: representations must decompose linearly into per-concept components, and these components must be orthogonal across concepts. This provides theoretical grounding for the Linear Representation Hypothesis: the linear structure widely observed in neural representations is a necessary consequence of compositional generalization. We further derive dimension bounds linking the number of composable concepts to the embedding geometry. Empirically, we evaluate these predictions across modern vision models (CLIP, SigLIP, DINO) and find that representations exhibit partial linear factorization with low-rank, near-orthogonal per-concept factors, and that the degree of this structure correlates with compositional generalization on unseen combinations. As models continue to scale, these conditions predict the representational geometry they may converge to. Code is available at https://github.com/oshapio/necessary-compositionality.

In this article, bipartite ranking, a statistical learning problem involved in many applications and widely studied in the passive context, is approached in a much more general \textit{active setting} than the discrete one previously considered in the literature. While the latter assumes that the conditional distribution is piece wise constant, the framework we develop permits in contrast to deal with continuous conditional distributions, provided that they fulfill a Hölder smoothness constraint. We first show that a naive approach based on discretisation at a uniform level, fixed \textit{a priori} and consisting in applying next the active strategy designed for the discrete setting generally fails. Instead, we propose a novel algorithm, referred to as smooth-rank and designed for the continuous setting, which aims to minimise the distance between the ROC curve of the estimated ranking rule and the optimal one w.r.t. the $\sup$ norm. We show that, for a fixed confidence level $ε>0$ and probability $δ\in (0,1)$, smooth-rank is PAC$(ε,δ)$. In addition, we provide a problem dependent upper bound on the expected sampling time of smooth-rank and establish a problem dependent lower bound on the expected sampling time of any PAC$(ε,δ)$ algorithm. Beyond the theoretical analysis carried out, numerical results are presented, providing solid empirical evidence of the performance of the algorithm proposed, which compares favorably with alternative approaches.

Identifying the full landscape of small and medium-sized enterprises (SMEs) in specialized industry sectors is critical for supply-chain resilience, yet existing business databases suffer from substantial coverage gaps -- particularly for sub-tier suppliers and firms in emerging niche markets. We propose a \textbf{Web--Knowledge--Web (W$\to$K$\to$W)} pipeline that iteratively (1)~crawls domain-specific web sources to discover candidate supplier entities, (2)~extracts and consolidates structured knowledge into a heterogeneous knowledge graph, and (3)~uses the knowledge graph's topology and coverage signals to guide subsequent crawling toward under-represented regions of the supplier space. To quantify discovery completeness, we introduce a \textbf{coverage estimation framework} inspired by ecological species-richness estimators (Chao1, ACE) adapted for web-entity populations. Experiments on the semiconductor equipment manufacturing sector (NAICS 333242) demonstrate that the W$\to$K$\to$W pipeline achieves the highest precision (0.138) and F1 (0.118) among all methods using the same 213-page crawl budget, building a knowledge graph of 765 entities and 586 relations while reaching peak recall by iteration~3 with only 112 pages.

Accurate fault detection and localization in electrical distribution systems is crucial, especially with the increasing integration of distributed energy resources (DERs), which inject greater variability and complexity into grid operations. In this study, FaultXformer is proposed, a Transformer encoder-based architecture developed for automatic fault analysis using real-time current data obtained from phasor measurement unit (PMU). The approach utilizes time-series current data to initially extract rich temporal information in stage 1, which is crucial for identifying the fault type and precisely determining its location across multiple nodes. In Stage 2, these extracted features are processed to differentiate among distinct fault types and identify the respective fault location within the distribution system. Thus, this dual-stage transformer encoder pipeline enables high-fidelity representation learning, considerably boosting the performance of the work. The model was validated on a dataset generated from the IEEE 13-node test feeder, simulated with 20 separate fault locations and several DER integration scenarios, utilizing current measurements from four strategically located PMUs. To demonstrate robust performance evaluation, stratified 10-fold cross-validation is performed. FaultXformer achieved average accuracies of 98.76% in fault type classification and 98.92% in fault location identification across cross-validation, consistently surpassing conventional deep learning baselines convolutional neural network (CNN), recurrent neural network (RNN). long short-term memory (LSTM) by 1.70%, 34.95%, and 2.04% in classification accuracy and by 10.82%, 40.89%, and 6.27% in location accuracy, respectively. These results demonstrate the efficacy of the proposed model with significant DER penetration.

Batch effects arising from technical variations in histopathology staining protocols, scanners, and acquisition pipelines pose a persistent challenge for computational pathology, hindering cross-batch generalization and limiting reliable deployment of models across clinical sites. In this work, we introduce Latent Manifold Compaction (LMC), an unsupervised representation learning framework that performs image harmonization by learning batch-invariant embeddings from a single source dataset through explicit compaction of stain-induced latent manifolds. This allows LMC to generalize to target domain data unseen during training. Evaluated on three challenging public and in-house benchmarks, LMC substantially reduces batch-induced separations across multiple datasets and consistently outperforms state-of-the-art normalization methods in downstream cross-batch classification and detection tasks, enabling superior generalization.

We propose Chunk-wise Attention Transducer (CHAT), a novel extension to RNN-T models that processes audio in fixed-size chunks while employing cross-attention within each chunk. This hybrid approach maintains RNN-T's streaming capability while introducing controlled flexibility for local alignment modeling. CHAT significantly reduces the temporal dimension that RNN-T must handle, yielding substantial efficiency improvements: up to 46.2% reduction in peak training memory, up to 1.36X faster training, and up to 1.69X faster inference. Alongside these efficiency gains, CHAT achieves consistent accuracy improvements over RNN-T across multiple languages and tasks -- up to 6.3% relative WER reduction for speech recognition and up to 18.0% BLEU improvement for speech translation. The method proves particularly effective for speech translation, where RNN-T's strict monotonic alignment hurts performance. Our results demonstrate that the CHAT model offers a practical solution for deploying more capable streaming speech models without sacrificing real-time constraints.

Accurate forecasting of transportation dynamics is essential for urban mobility and infrastructure planning. Although recent work has achieved strong performance with deep learning models, these methods typically require dataset-specific training, architecture design and hyper-parameter tuning. This paper evaluates whether general-purpose time-series foundation models can serve as forecasters for transportation tasks by benchmarking the zero-shot performance of the state-of-the-art model, Chronos-2, across ten real-world datasets covering highway traffic volume and flow, urban traffic speed, bike-sharing demand, and electric vehicle charging station data. Under a consistent evaluation protocol, we find that, even without any task-specific fine-tuning, Chronos-2 delivers state-of-the-art or competitive accuracy across most datasets, frequently outperforming classical statistical baselines and specialized deep learning architectures, particularly at longer horizons. Beyond point forecasting, we evaluate its native probabilistic outputs using prediction-interval coverage and sharpness, demonstrating that Chronos-2 also provides useful uncertainty quantification without dataset-specific training. In general, this study supports the adoption of time-series foundation models as a key baseline for transportation forecasting research.

This paper studies how well Naturalistic Adversarial Patches (NAPs) transfer to a physical traffic sign setting when the detector is trained on a customized dataset for an autonomous vehicle (AV) environment. We construct a composite dataset, CompGTSRB (which is customized dataset for AV environment), by pasting traffic sign instances from the German Traffic Sign Recognition Benchmark (GTSRB) onto undistorted backgrounds captured from the target platform. CompGTSRB is used to train a YOLOv5 model and generate patches using a Generative Adversarial Network (GAN) with latent space optimization, following existing NAP methods. We carried out a series of experiments on our Quanser QCar testbed utilizing the front CSI camera provided in QCar. Across configurations, NAPs reduce the detector's STOP class confidence. Different configurations include distance, patch sizes, and patch placement. These results along with a detailed step-by-step methodology indicate the utility of CompGTSRB dataset and the proposed systematic physical protocols for credible patch evaluation. The research further motivate researching the defenses that address localized patch corruption in embedded perception pipelines.

LLM agents are increasingly used for code generation, but physics-based simulation poses a deeper challenge: natural-language descriptions of simulation models are inherently underspecified, and different admissible resolutions of implicit choices produce physically valid but scientifically distinct configurations. Without explicit detection and resolution of these ambiguities, neither the correctness of the result nor its reproducibility from the original description can be assured. This paper investigates agentic scientific simulation, where model construction is organized as an execution-grounded interpret-act-validate loop and the simulator serves as the authoritative arbiter of physical validity rather than merely a runtime. We present JutulGPT, a reference implementation built on the fully differentiable Julia-based reservoir simulator JutulDarcy. The agent combines structured retrieval of documentation and examples with code synthesis, static analysis, execution, and systematic interpretation of solver diagnostics. Underspecified modelling choices are detected explicitly and resolved either autonomously (with logged assumptions) or through targeted user queries. The results demonstrate that agent-mediated model construction can be grounded in simulator validation, while also revealing a structural limitation: choices resolved tacitly through simulator defaults are invisible to the assumption log and to any downstream representation. A secondary experiment with autonomous reconstruction of a reference model from progressively abstract textual descriptions shows that reconstruction variability exposes latent degrees of freedom in simulation descriptions and provides a practical methodology for auditing reproducibility. All code, prompts, and agent logs are publicly available.

Clustering is a central primitive in unsupervised learning, yet practice is dominated by heuristics whose outputs can be unstable and highly sensitive to representations, hyperparameters, and initialisation. Existing theoretical results are largely objective-specific and do not explain these behaviours at a unifying level. We formalise the common optimisation core underlying diverse clustering paradigms by defining the Universal Clustering Problem (UCP): the maximisation of a polynomial-time computable partition utility over a finite metric space. We prove the NP-hardness of UCP via two independent polynomial-time reductions from graph colouring and from exact cover by 3-sets (X3C). By mapping ten major paradigms -- including k-means, GMMs, DBSCAN, spectral clustering, and affinity propagation -- to the UCP framework, we demonstrate that each inherits this fundamental intractability. Our results provide a unified explanation for characteristic failure modes, such as local optima in alternating methods and greedy merge-order traps in hierarchical clustering. Finally, we show that clustering limitations reflect interacting computational and epistemic constraints, motivating a shift toward stability-aware objectives and interaction-driven formulations with explicit guarantees.

Objective: ML-based clinical risk prediction models are increasingly used to support decision-making in healthcare. While class-imbalance correction techniques are commonly applied to improve model performance in settings with rare outcomes, their impact on probabilistic calibration remains insufficiently understood. This study evaluated the effect of widely used resampling strategies on both discrimination and calibration across real-world clinical prediction tasks. Methods: Ten clinical datasets spanning diverse medical domains and including 605,842 patients were analyzed. Multiple machine-learning model families, including linear models and several non-linear approaches, were evaluated. Models were trained on the original data and under three commonly used 1:1 class-imbalance correction strategies (SMOTE, RUS, ROS). Performance was assessed on held-out data using discrimination and calibration metrics. Results: Across all datasets and model families, resampling had no positive impact on predictive performance. Changes in the Receiver Operating Characteristic Area Under Curve (ROC-AUC) relative to models trained on the original data were small and inconsistent (ROS: -0.002, p<0.05; RUS: -0.004, p>0.05; SMOTE: -0.01, p<0.05), with no resampling strategy demonstrating a systematic improvement. In contrast, resampling in general degraded the calibration performance. Models trained using imbalance correction exhibited higher Brier scores (0.029 to 0.080, p<0.05), reflecting poorer probabilistic accuracy, and marked deviations in calibration intercept and slope, indicating systematic distortions of predicted risk despite preserved rank-based performance. Conclusion: In a diverse set of real-world clinical prediction tasks, commonly used class-imbalance correction techniques did not provide generalizable improvements in discrimination and were associated with degraded calibration.

Advances in large reasoning models have shown strong performance on complex reasoning tasks by scaling test-time compute through extended reasoning. However, recent studies observe that in vision-dependent tasks, extended textual reasoning at inference time can degrade performance as models progressively lose attention to visual tokens and increasingly rely on textual priors alone. To address this, prior works use reinforcement learning (RL)-based fine-tuning to route visual tokens or employ refocusing mechanisms during reasoning. While effective, these methods are computationally expensive, requiring large-scale data generation and policy optimization. To leverage the benefits of test-time compute without additional RL fine-tuning, we propose VisRef, a visually grounded test-time scaling framework. Our key idea is to actively guide the reasoning process by re-injecting a coreset of visual tokens that are semantically relevant to the reasoning context while remaining diverse and globally representative of the image, enabling more grounded multi-modal reasoning. Experiments on three visual reasoning benchmarks with state-of-the-art multi-modal large reasoning models demonstrate that, under fixed test-time compute budgets, VisRef consistently outperforms existing test-time scaling approaches by up to 6.4%.

Existing visual reasoning benchmarks predominantly rely on natural language prompts, evaluate narrow reasoning modalities, or depend on subjective scoring procedures such as LLM-as-judge. We introduce the TACIT Benchmark, a programmatic visual reasoning benchmark comprising 10 tasks across 6 reasoning domains: spatial navigation, abstract pattern completion, causal simulation, logical constraint satisfaction, graph theory, and topology. The benchmark provides dual-track evaluation: a generative track in which models must produce solution images verified through deterministic computer-vision pipelines, and a discriminative track offering five-way multiple choice with structurally plausible near-miss distractors. Each distractor violates exactly one structural constraint, requiring models to reason about fine-grained visual differences rather than exploit superficial cues. Version 0.1.0 distributes 6,000 puzzles (108,000 PNG images across three resolutions) with fully deterministic seeded generation and reproducible verification. The dataset, generation code, and evaluation harness are released under the Apache 2.0 license on HuggingFace (DOI: 10.57967/hf/7904).

Generative models, particularly Diffusion Models (DM), have shown strong potential for Computed Tomography (CT) reconstruction serving as expressive priors for solving ill-posed inverse problems. However, diffusion-based reconstruction relies on Stochastic Differential Equations (SDEs) for forward diffusion and reverse denoising, where such stochasticity can interfere with repeated data consistency corrections in CT reconstruction. Since CT reconstruction is often time-critical in clinical and interventional scenarios, improving reconstruction efficiency is essential. In contrast, Flow Matching (FM) models sampling as a deterministic Ordinary Differential Equation (ODE), yielding smooth trajectories without stochastic noise injection. This deterministic formulation is naturally compatible with repeated data consistency operations. Furthermore, we observe that FM-predicted velocity fields exhibit strong correlations across adjacent steps. Motivated by this, we propose an FM-based CT reconstruction framework (FMCT) and an efficient variant (EFMCT) that reuses previously predicted velocity fields over consecutive steps to substantially reduce the number of Neural network Function Evaluations (NFEs), thereby improving inference efficiency. We provide theoretical analysis showing that the error introduced by velocity reuse is bounded when combined with data consistency operations. Extensive experiments demonstrate that FMCT/EFMCT achieve competitive reconstruction quality while significantly improving computational efficiency compared with diffusion-based methods. The codebase is open-sourced at https://github.com/EFMCT/EFMCT.

Magnetic Resonance (MR) imaging is a diagnostic tool used in modern medicine; however, its output can be affected by motion artefacts and may be limited by equipment. This research focuses on MRI image quality enhancement using two efficient Generative Adversarial Networks (GANs) models: SOUP-GAN and CSR-GAN. In both models, meaningful architectural modifications were introduced. The generator and discriminator of each were further deepened by adding convolutional layers and were enhanced in filter sizes as well. The LeakyReLU activation function was used to improve gradient flow, and hyperparameter tuning strategies were applied, including a reduced learning rate and an optimal batch size. Moreover, spectral normalisation was proposed to address mode collapse and improve training stability. The experiment shows that CSR-GAN has better performance in reconstructing the image with higher frequency details and reducing noise compared to other methods, with an optimised PSNR of 34.6 and SSIM of 0.89. However, SOUP-GAN performed the best in terms of delivering less noisy images with good structures, achieving a PSNR of 34.4 and SSIM of 0.83. The obtained results indicate that the proposed enhanced GAN model can be a useful tool for MR image quality improvement for subsequent better disease diagnostics.

We study the expected star discrepancy under a newly designed class of non-equal volume partitions. The main contributions are twofold. First, we establish a strong partition principle for the star discrepancy, showing that our newly designed non-equal volume partitions yield stratified sampling point sets with lower expected star discrepancy than classical jittered sampling. Specifically, we prove that $\mathbb{E}(D^{*}_{N}(Z)) < \mathbb{E}(D^{*}_{N}(Y))$, where $Y$ and $Z$ represent jittered sampling and our non-equal volume partition sampling, respectively. Second, we derive explicit upper bounds for the expected star discrepancy under our non-equal volume partition models, which improve upon existing bounds for jittered sampling. Our results provide a theoretical foundation for using non-equal volume partitions in high-dimensional numerical integration.

One of the common issues in clinical decision-making is the presence of uncertainty, which often arises due to ambiguity in radiology reports, which often reflect genuine diagnostic uncertainty or limitations of automated label extraction in various complex cases. Especially the case of multilabel datasets such as CheXpert, MIMIC-CXR, etc., which contain labels such as positive, negative, and uncertain. In clinical decision-making, the uncertain label plays a tricky role as the model should not be forced to provide a confident prediction in the absence of sufficient evidence. The ability of the model to say it does not understand whenever it is not confident is crucial, especially in the cases of clinical decision-making involving high risks. Here, we propose AdURA-Net, a geometry-driven adaptive uncertainty-aware framework for reliable thoracic disease classification. The key highlights of the proposed model are: a) Adaptive dilated convolution and multiscale deformable alignment coupled with the backbone Densenet architecture capturing the anatomical complexities of the medical images, and b) Dual Head Loss, which combines masked binary cross entropy with logit and a Dirichlet evidential learning objective.

The rapid evolution of sophisticated cyberattacks has strained modern Security Operations Centers (SOC), which traditionally rely on rule-based or signature-driven detection systems. These legacy frameworks often generate high volumes of technical alerts that lack organizational context, leading to analyst fatigue and delayed incident responses. This paper presents LiaisonAgent, an autonomous multi-agent system designed to bridge the gap between technical risk detection and business-level risk governance. Built upon the QWQ-32B large reasoning model, LiaisonAgent integrates specialized sub-agents, including human-computer interaction agents, comprehensive judgment agents, and automated disposal agents-to execute end-to-end investigation workflows. The system leverages a hybrid planning architecture that combines deterministic workflows for compliance with autonomous reasoning based on the ReAct paradigm to handle ambiguous operational scenarios. Experimental evaluations across diverse security contexts, such as large-scale data exfiltration and unauthorized account borrowing, achieve an end-to-end tool-calling success rate of 97.8% and a risk judgment accuracy of 95%. Furthermore, the system exhibits significant resilience against out-of-distribution noise and adversarial prompt injections, while achieving a 92.7% reduction in manual investigation overhead.

Axial piston pumps are indispensable power sources in high-stakes fluid power systems, including aerospace, marine, and heavy machinery applications. Their operational reliability is frequently compromised by compound faults that simultaneously affect multiple friction pairs. Conventional data-driven diagnosis methods suffer from severe data scarcity for compound faults and poor generalization across varying operating conditions. This paper proposes a novel multi-condition physics-data coupled digital twin calibration framework that explicitly resolves the fundamental uncertainty of pump outlet flow ripple. The framework comprises three synergistic stages: in-situ virtual high-frequency flow sensing on a dedicated rigid metallic segment, surrogate model-assisted calibration of the 3D CFD source model using physically estimated ripple amplitudes, and multi-objective inverse transient analysis for viscoelastic unsteady-friction pipeline parameter identification. Comprehensive experiments on a test rig demonstrate that the calibrated digital twin accurately reproduces both single-fault and two representative compound-fault. These results establish a high-fidelity synthetic fault-generation capability that directly enables robust zero-shot fault diagnosis under previously unseen operating regimes and fault combinations, thereby advancing predictive maintenance in complex hydraulic systems.

Token reduction is an effective way to accelerate long-video vision-language models (VLMs), but most existing methods are designed for dense Transformers and do not directly account for hybrid architectures that interleave attention with linear-time state-space blocks (e.g., Mamba). We study query-conditioned token reduction for hybrid video VLMs and analyze reduction behavior through two properties: layerwise sparsity (how many tokens capture query-relevant information) and importance stability (whether token-importance rankings persist across depth). Although token importance is sparse within each layer, the set of important tokens changes across layers, so aggressive early pruning is unreliable. Motivated by this, we propose a low-to-high progressive reduction schedule and a unified language-aware scoring mechanism for both attention and Mamba blocks (using an implicit-attention proxy for Mamba), enabling all-layer token reduction in hybrids. Under an aggressive compression setting (retaining 25% of visual tokens), our approach delivers substantial prefilling speedups (3.8--4.2x) with near-baseline accuracy at test time, and light finetuning under reduction further improves performance on long-context video benchmarks.

Deep Neural Networks (DNNs) have advanced applications in domains such as healthcare, autonomous systems, and scene understanding, yet the internal semantics of their hidden neurons remain poorly understood. Prior work introduced a Concept Induction-based framework for hidden neuron analysis and demonstrated its effectiveness on the ADE20K dataset. In this case study, we investigate whether the approach generalizes by applying it to the SUN2012 dataset, a large-scale scene recognition benchmark. Using the same workflow, we assign interpretable semantic labels to neurons and validate them through web-sourced images and statistical testing. Our findings confirm that the method transfers to SUN2012, showing its broader applicability.

The increasing reliance on cloud-hosted Large Language Models (LLMs) exposes sensitive client data, such as prompts and responses, to potential privacy breaches by service providers. Existing approaches fail to ensure privacy, maintain model performance, and preserve computational efficiency simultaneously. To address this challenge, we propose Talaria, a confidential inference framework that partitions the LLM pipeline to protect client data without compromising the cloud's model intellectual property or inference quality. Talaria executes sensitive, weight-independent operations within a client-controlled Confidential Virtual Machine (CVM) while offloading weight-dependent computations to the cloud GPUs. The interaction between these environments is secured by our Reversible Masked Outsourcing (ReMO) protocol, which uses a hybrid masking technique to reversibly obscure intermediate data before outsourcing computations. Extensive evaluations show that Talaria can defend against state-of-the-art token inference attacks, reducing token reconstruction accuracy from over 97.5% to an average of 1.34%, all while being a lossless mechanism that guarantees output identical to the original model without significantly decreasing efficiency and scalability. To the best of our knowledge, this is the first work that ensures clients' prompts and responses remain inaccessible to the cloud, while also preserving model privacy, performance, and efficiency.

The rapid proliferation of agentic AI skill ecosystems -- exemplified by OpenClaw (228,000 GitHub stars) and Anthropic Agent Skills (75,600 stars) -- has introduced a critical supply chain attack surface. The ClawHavoc campaign (January-February 2026) infiltrated over 1,200 malicious skills into the OpenClaw marketplace, while MalTool catalogued 6,487 malicious tools that evade conventional detection. In response, twelve reactive security tools emerged, yet all rely on heuristic methods that provide no formal guarantees. We present SkillFortify, the first formal analysis framework for agent skill supply chains, with six contributions: (1) the DY-Skill attacker model, a Dolev-Yao adaptation to the five-phase skill lifecycle with a maximality proof; (2) a sound static analysis framework grounded in abstract interpretation; (3) capability-based sandboxing with a confinement proof; (4) an Agent Dependency Graph with SAT-based resolution and lockfile semantics; (5) a trust score algebra with formal monotonicity; and (6) SkillFortifyBench, a 540-skill benchmark. SkillFortify achieves 96.95% F1 (95% CI: [95.1%, 98.4%]) with 100% precision and 0% false positive rate on 540 skills, while SAT-based resolution handles 1,000-node graphs in under 100 ms.

The rise of text-to-video generation models has raised growing concerns over content authenticity, copyright protection, and malicious misuse. Watermarking serves as an effective mechanism for regulating such AI-generated content, where high fidelity and strong robustness are particularly critical. Recent generative image watermarking methods provide a promising foundation by leveraging watermark information and pseudo-random keys to control the initial sampling noise, enabling lossless embedding. However, directly extending these techniques to videos introduces two key limitations: Existing designs implicitly rely on strict alignment between video frames and frame-dependent pseudo-random binary sequences used for watermark encryption. Once this alignment is disrupted, subsequent watermark extraction becomes unreliable; and Video-specific distortions, such as inter-frame compression, significantly degrade watermark reliability. To address these issues, we propose SKeDA, a generative watermarking framework tailored for text-to-video diffusion models. SKeDA consists of two components: (1) Shuffle-Key-based Distribution-preserving Sampling (SKe) employs a single base pseudo-random binary sequence for watermark encryption and derives frame-level encryption sequences through permutation. This design transforms watermark extraction from synchronization-sensitive sequence decoding into permutation-tolerant set-level aggregation, substantially improving robustness against frame reordering and loss; and (2) Differential Attention (DA), which computes inter-frame differences and dynamically adjusts attention weights during extraction, enhancing robustness against temporal distortions. Extensive experiments demonstrate that SKeDA preserves high video generation quality and watermark robustness.

In healthcare, predictive models increasingly inform patient-level decisions, yet little attention is paid to the variability in individual risk estimates and its impact on treatment decisions. For overparameterized models, now standard in machine learning, a substantial source of variability often goes undetected. Even when the data and model architecture are held fixed, randomness introduced by optimization and initialization can lead to materially different risk estimates for the same patient. This problem is largely obscured by standard evaluation practices, which rely on aggregate performance metrics (e.g., log-loss, accuracy) that are agnostic to individual-level stability. As a result, models with indistinguishable aggregate performance can nonetheless exhibit substantial procedural arbitrariness, which can undermine clinical trust. We propose an evaluation framework that quantifies individual-level prediction instability by using two complementary diagnostics: empirical prediction interval width (ePIW), which captures variability in continuous risk estimates, and empirical decision flip rate (eDFR), which measures instability in threshold-based clinical decisions. We apply these diagnostics to simulated data and GUSTO-I clinical dataset. Across observed settings, we find that for flexible machine-learning models, randomness arising solely from optimization and initialization can induce individual-level variability comparable to that produced by resampling the entire training dataset. Neural networks exhibit substantially greater instability in individual risk predictions compared to logistic regression models. Risk estimate instability near clinically relevant decision thresholds can alter treatment recommendations. These findings that stability diagnostics should be incorporated into routine model validation for assessing clinical reliability.

Continual Learning (CL) requires models to sequentially adapt to new tasks without forgetting old knowledge. Recently, Low-Rank Adaptation (LoRA), a representative Parameter-Efficient Fine-Tuning (PEFT) method, has gained increasing attention in CL. Several LoRA-based CL methods reduce interference across tasks by separating their update spaces, typically building the new space from the estimated null space of past tasks. However, they (i) overlook task-shared directions, which suppresses knowledge transfer, and (ii) fail to capture truly effective task-specific directions since these ``null bases" of old tasks can remain nearly inactive for new task under correlated tasks. To address this, we study LoRA learning capability from a projection energy perspective, and propose Low-rank Decomposition and Adaptation (LoDA). It performs a task-driven decomposition to build general and truly task-specific LoRA subspaces by solving two energy-based objectives, decoupling directions for knowledge sharing and isolation. LoDA fixes LoRA down-projections on two subspaces and learns robust up-projections via a Gradient-Aligned Optimization (GAO) approach. After each task, before integrating the LoRA updates into the backbone, LoDA derives a closed-form recalibration for the general update, approximating a feature-level joint optimum along this task-shared direction. Experiments indicate that LoDA outperforms existing CL methods.

Polysomnography (PSG) provides the gold standard for sleep assessment but suffers from substantial heterogeneity across recording devices and cohorts. There have been growing efforts to build general-purpose foundation models (FMs) for sleep physiology, but lack an in-depth understanding of the pre-training process and scaling patterns that lead to more generalizable sleep FMs. To fill this gap, we curate a massive corpus of 166,500 hours of sleep recordings from nine public sources and establish SleepBench, a comprehensive, fully open-source benchmark. Leveraging SleepBench, we systematically evaluate four families of self-supervised pre-training objectives and uncover three critical findings: (1) existing FMs fail to generalize to missing channels at inference; (2) channel-invariant feature learning is essential for pre-training; and (3) scaling sample size, model capacity, and multi-source data mixture consistently improves downstream performance.With an enhanced pre-training and scaling recipe, we introduce OSF, a family of sleep FMs that achieves state-of-the-art performance across nine datasets on diverse sleep and disease prediction tasks. Further analysis of OSF also reveals intriguing properties in sample efficiency, hierarchical aggregation, and cross-dataset scaling.

Large Vision-Language Models (VLMs) have emerged as powerful engines for autonomous GUI agents, yet their deployment is severely constrained by the substantial memory footprint and latency of the Key-Value (KV) cache during long-horizon interactions. While existing cache compression methods have proven effective for LLMs, we empirically demonstrate that they suffer from suboptimal performance in GUI scenarios due to a fundamental misalignment: unlike general visual tasks where attention sparsity varies across layers, GUI attention patterns exhibit uniform high-sparsity across all transformer layers. Motivated by this insight, we propose ST-Lite, a training-free KV cache compression framework tailored for efficient GUI agents that explicitly addresses the dynamic spatio-trajectory dependencies within GUI data streams. ST-Lite introduces a novel dual-branch scoring policy incorporating Component-centric Spatial Saliency (CSS) and Trajectory-aware Semantic Gating (TSG). Specifically, CSS preserves the structural integrity of interactive UI elements by evaluating local neighborhood saliency, while TSG mitigates historical redundancy by dynamically filtering visually repetitive KV pairs within the interaction trajectory. Extensive evaluations demonstrate that with only a 10-20% cache budget, ST-Lite achieves a 2.45x decoding acceleration while maintaining comparable or even superior performance compared to full-cache baselines, offering a scalable solution for resource-constrained GUI agents.

To integrate seamlessly into real-world software engineering, Code Agents must evolve from passive instruction followers into proactive collaborative partners. However, current evaluation paradigms predominantly reward "guessing" user intent under ideal conditions, neglecting the agent's ability to align with users through dialogue--a critical trait for collaborative intelligence. In this work, we propose a paradigm shift in evaluation to drive this transition. We introduce ClarEval, a framework designed to assess an agent's "Collaborative Quotient" by simulating the inherent ambiguity of human communication. By systematically injecting three types of realistic ambiguity (missing goals, premises, and ambiguous terminology) into standard tasks, we force agents to step out of their "generator" role and engage in requirement elicitation. To quantify this capability, we propose a metric suite led by Average Turns to Clarify (ATC) and Key Question Coverage (KQC), which measure not just the correctness of the generated code, but the efficiency and precision of the collaboration. Our experiments on eleven state-of-the-art agents reveal a stark reality: while models like GPT-5-Coder excel at coding, they often lack the strategic communication skills required for efficient partnership. ClarEval thus serves as a crucial roadmap for bridging the gap between strong coders and capable collaborators.The code is available at https://github.com/JialinLi13/ClarEval

Financial systems run nonstop and must stay reliable even during cyber incidents. Modern attacks move across many services (apps, APIs, identity, payment rails), so defenders must make a sequence of actions under time pressure. Most security tools still use fixed rules or static playbooks, which can be slow to adapt when the attacker changes behavior. Reinforcement learning (RL) is a good fit for sequential decisions, but much of the RL-in-finance literature targets trading and does not model real cyber response limits such as action cost, service disruption, and defender coordination across many assets. This paper proposes RLShield, a practical multi-agent RL pipeline for financial cyber defense. We model the enterprise attack surface as a Markov decision process (MDP) where states summarize alerts, asset exposure, and service health, and actions represent real response steps (e.g., isolate a host, rotate credentials, ratelimit an API, block an account, or trigger recovery). RLShield learns coordinated policies across multiple agents (assets or service groups) and optimizes a risk-sensitive objective that balances containment speed, business disruption, and response cost. We also include a game-aware evaluation that tests policies against adaptive attackers and reports operational outcomes, not only reward. Experiments show that RLShield reduces time-to-containment and residual exposure while keeping disruption within a fixed response budget, outperforming static rule baselines and single-agent RL under the same constraints. These results suggest that multi-agent, cost-aware RL can provide a deployable layer for automated response in financial security operations.

Intrusion detection in IoT and industrial networks requires models that can detect rare attacks at low false-positive rates while remaining reliable under evolving traffic and limited labels. Existing IDS solutions often report strong in-distribution accuracy, but they may degrade when evaluated on future traffic, unseen (zero-day) attack families, or adversarial feature manipulations, and many systems provide limited evidence to support analyst triage. To address these gaps, we propose ThreatFormer- IDS, a Transformer-based sequence modeling framework that converts flow records into time-ordered windows and learns contextual representations for robust intrusion screening. The method combines (i) weighted supervised learning for imbalanced detection, (ii) masked self-supervised learning to improve representation stability under drift and sparse labels, (iii) PGDbased adversarial training with scale-normalized perturbations to strengthen resilience against feature-level evasion, and (iv) Integrated Gradients attribution to highlight influential time steps and features for each alert. On the ToN IoT benchmark with chronological evaluation, ThreatFormer-IDS achieves AUCROC 0.994, AUC-PR 0.956, and Recall@1%FPR 0.910, outperforming strong tree-based and sequence baselines. Under a zero-day protocol with held-out attack families, it maintains superior generalization (AUC-PR 0.721, Recall@1%FPR 0.783). Robustness tests further show slower degradation in AUCPR as the adversarial budget increases, confirming improved stability under bounded perturbations. Overall, ThreatFormer- IDS provides a unified, deployment-oriented IDS pipeline that balances detection quality, zero-day behavior, robustness, and explainability.

Bird image segmentation remains a challenging task in computer vision due to extreme pose diversity, complex plumage patterns, and variable lighting conditions. This paper presents a dual-pipeline framework for binary bird image segmentation leveraging 2025 foundation models. We introduce two operating modes built upon Segment Anything Model 2.1 (SAM 2.1) as a shared frozen backbone: (1) a zero-shot pipeline using Grounding DINO 1.5 to detect birds via the text prompt "bird" before prompting SAM 2.1 with bounding boxes requiring no labelled bird data; and (2) a supervised pipeline that fine-tunes YOLOv11 on the CUB-200-2011 dataset for high-precision detection, again prompting SAM 2.1 for pixel-level masks. The segmentation model is never retrained for new species or domains. On CUB-200-2011 (11,788 images, 200 species), the supervised pipeline achieves IoU 0.912, Dice 0.954, and F1 0.953 outperforming all prior baselines including SegFormer-B2 (IoU 0.842) by +7.0 percentage points. The zero-shot pipeline achieves IoU 0.831 using only a text prompt, the first such result reported on this benchmark. We demonstrate that prompt-based foundation model pipelines outperform task specific end-to-end trained segmentation networks, while requiring only lightweight detector fine-tuning (~1 hour) for domain adaptation. Complete PyTorch implementation, dataset preparation scripts, and trained weights are publicly available.

Context: Test case prioritization (TCP) is a technique widely used by software development organizations to accelerate regression testing. Objectives: We aim to systematize existing TCP knowledge and to propose and empirically evaluate a new TCP approach. Methods: We conduct a snowballing review (SR) on TCP, implement a~comprehensive platform for TCP research (TCPFramework), analyze existing evaluation metrics and propose two new ones (\rAPFDc{} and ATR), and develop a~family of ensemble TCP methods called approach combinators. Results: The SR helped identify 324 studies related to TCP. The techniques proposed in our study were evaluated on the RTPTorrent dataset, consistently outperforming their base approaches across the majority of subject programs, and achieving performance comparable to the current state of the art for heuristical algorithms (in terms of \rAPFDc{}, NTR, and ATR), while using a distinct approach. Conclusions: The proposed methods can be used efficiently for TCP, reducing the time spent on regression testing by up to 2.7\%. Approach combinators offer significant potential for improvements in future TCP research, due to their composability.

Cross-robot policy learning -- training a single policy to perform well across multiple embodiments -- remains a central challenge in robot learning. Transformer-based policies, such as vision-language-action (VLA) models, are typically embodiment-agnostic and must infer kinematic structure purely from observations, which can reduce robustness across embodiments and even limit performance within a single embodiment. We propose an embodiment-aware transformer policy that injects morphology via three mechanisms: (1) kinematic tokens that factorize actions across joints and compress time through per-joint temporal chunking; (2) a topology-aware attention bias that encodes kinematic topology as an inductive bias in self-attention, encouraging message passing along kinematic edges; and (3) joint-attribute conditioning that augments topology with per-joint descriptors to capture semantics beyond connectivity. Across a range of embodiments, this structured integration consistently improves performance over a vanilla pi0.5 VLA baseline, indicating improved robustness both within an embodiment and across embodiments.

A recent report on "Learning the natural history of human disease with generative transformers" created an opportunity to assess the engineering challenge of delivering user-facing Generative AI applications in privacy-sensitive domains. The application of these models, particularly for personalized healthcare tasks like predicting individual morbidity risk, is typically constrained by data privacy concerns. This project was accordingly designed as an in-browser model deployment exercise (an "App") testing the architectural boundaries of client-side inference generation (no downloads or installations). We relied exclusively on the documentation provided in the reference report to develop the model, specifically testing the "R" component of the FAIR data principles: Findability, Accessibility, Interoperability, and Reusability. The successful model deployment, leveraging ONNX and a custom JavaScript SDK, establishes a secure, high-performance architectural blueprint for the future of private generative AI in medicine.

Generative modeling of neural network parameters is often tied to architectures because standard parameter representations rely on known weight-matrix dimensions. Generation is further complicated by permutation symmetries that allow networks to model similar input-output functions while having widely different, unaligned parameterizations. In this work, we introduce Neural Network Diffusion Transformers (NNiTs), which generate weights in a width-agnostic manner by tokenizing weight matrices into patches and modeling them as locally structured fields. We establish that Graph HyperNetworks (GHNs) with a convolutional neural network (CNN) decoder structurally align the weight space, creating the local correlation necessary for patch-based processing. Focusing on MLPs, where permutation symmetry is especially apparent, NNiT generates fully functional networks across a range of architectures. Our approach jointly models discrete architecture tokens and continuous weight patches within a single sequence model. On ManiSkill3 robotics tasks, NNiT achieves >85% success on architecture topologies unseen during training, while baseline approaches fail to generalize.

Process attestation systems verify that a continuous physical process, such as human authorship, actually occurred, rather than merely checking system state. These systems face a fundamental dependability challenge: the evidence collection infrastructure must remain available and tamper-resistant even when the attesting party controls the platform. Trusted Execution Environments (TEEs) provide hardware-enforced isolation that can address this challenge, but their integration with continuous process attestation introduces novel resilience requirements not addressed by existing frameworks. We present the first architecture for continuous process attestation evidence collection inside TEEs, providing hardware-backed tamper resistance against trust-inverted adversaries with graduated input assurance from software-channel integrity (Tier 1) through hardware-bound input (Tier 3). We develop a Markov-chain dependability model quantifying Evidence Chain Availability (ECA), Mean Time Between Evidence Gaps (MTBEG), and Recovery Time Objectives (RTO). We introduce a resilient evidence chain protocol maintaining chain integrity across TEE crashes, network partitions, and enclave migration. Our security analysis derives formal bounds under combined threat models including trust inversion and TEE side channels, parameterized by a conjectural side-channel leakage bound esc that requires empirical validation. Evaluation on Intel SGX demonstrates under 25% per-checkpoint CPU overhead (<0.3% of the 30 s checkpoint interval), >99.5% Evidence Chain Availability (ECA) (the fraction of session time with active evidence collection) in Monte Carlo simulation under Poisson failure models, and sealed-state recovery under 200 ms.

The proliferation of AI-generated text has intensified the need for reliable authorship verification, yet current output-based methods are increasingly unreliable. We observe that the ordinary typing interface captures rich cognitive signatures, measurable patterns in keystroke timing that reflect the planning, translating, and revising stages of genuine composition. Drawing on large-scale keystroke datasets comprising over 136 million events, we define the Cognitive Load Correlation (CLC) and show it distinguishes genuine composition from mechanical transcription. We present a non-intrusive verification framework that operates within existing writing interfaces, collecting only timing metadata to preserve privacy. Our analytical evaluation estimates 85 to 95 percent discrimination accuracy under stated assumptions, while limiting biometric leakage via evidence quantization. We analyze the adversarial robustness of cognitive signatures, showing they resist timing-forgery attacks that defeat motor-level authentication because the cognitive channel is entangled with semantic content. We conclude that reframing authorship verification as a human-computer interaction problem provides a privacy-preserving alternative to invasive surveillance.

Shared micromobility services such as e-scooters and bikes have become an integral part of urban transportation, yet their efficiency critically depends on effective vehicle rebalancing. Existing methods either optimize for average demand patterns or employ robust optimization and reinforcement learning to handle predefined uncertainties. However, these approaches overlook emergent events (e.g., demand surges, vehicle outages, regulatory interventions) or sacrifice performance in normal conditions. We introduce AMPLIFY, an LLM-augmented policy adaptation framework for shared micromobility rebalancing. The framework combines a baseline rebalancing module with an LLM-based adaptation module that adjusts strategies in real time under emergent scenarios. The adaptation module ingests system context, demand predictions, and baseline strategies, and refines adjustments through self-reflection. Evaluations on real-world e-scooter data from Chicago show that our approach improves demand satisfaction and system revenue compared to baseline policies, highlighting the potential of LLM-driven adaptation as a flexible solution for managing uncertainty in micromobility systems.

A constant weight binary code consists of $n$-bit binary codewords, each with exactly $w$ bits equal to 1, such that any two codewords are at least Hamming distance $d$ apart. $A(n,d,w)$ is the maximum size of a constant weight binary code with parameters $n,d,w$. We establish improved lower bounds on $A(n,d,w)$ by constructing new larger codes, for 24 values of $(n,d,w)$ with $6 \leq d \leq 18$ and $18 \leq n \leq 35$. The improved lower bounds come from two strategies. The first is a tabu search that operates at the level of bit swaps. The second is a novel greedy heuristic that repeatedly chooses the candidate codeword that maximizes a randomly-scored histogram of distances to previously-added codewords. These strategies were proposed by CPro1, an automated protocol that generates, implements, and tests diverse strategies for combinatorial constructions.

Large-scale commercial search systems optimize for relevance to drive successful sessions that help users find what they are looking for. To maximize relevance, we leverage two complementary objectives: behavioral relevance (results users tend to click or download) and textual relevance (a result's semantic fit to the query). A persistent challenge is the scarcity of expert-provided textual relevance labels relative to abundant behavioral relevance labels. We first address this by systematically evaluating LLM configurations, finding that a specialized, fine-tuned model significantly outperforms a much larger pre-trained one in providing highly relevant labels. Using this optimal model as a force multiplier, we generate millions of textual relevance labels to overcome the data scarcity. We show that augmenting our production ranker with these textual relevance labels leads to a significant outward shift of the Pareto frontier: offline NDCG improves for behavioral relevance while simultaneously increasing for textual relevance. These offline gains were validated by a worldwide A/B test on the App Store ranker, which demonstrated a statistically significant +0.24% increase in conversion rate, with the most substantial performance gains occurring in tail queries, where the new textual relevance labels provide a robust signal in the absence of reliable behavioral relevance labels.

We describe our experience training Summer-22B, a video foundation model developed from scratch. This report documents the engineering challenges, design decisions, and lessons learned while scaling from raw footage collection to a functional model trained on approximately 50 million clips. We outline our approach combining metadata-driven dataset curation, multi-stage filtering, $μ$P parameterization, and hypersphere-constrained optimization. We developed the Lavender Data system for dataset management and adopted inference-aware architectural choices. We share observations on what worked in our setting: dataset engineering consumed the majority of effort, architectural variants showed smaller differences than we expected, and $μ$P hyperparameter transfer appeared effective even under geometric constraints. We hope this account proves useful to others undertaking similar projects.

Retrieval-augmented generation (RAG) has emerged as a powerful paradigm for enhancing multimodal large language models by grounding their responses in external, factual knowledge and thus mitigating hallucinations. However, the integration of externally sourced knowledge bases introduces a critical attack surface. Adversaries can inject malicious multimodal content capable of influencing both retrieval and downstream generation. In this work, we present MM-MEPA, a multimodal poisoning attack that targets the metadata components of image-text entries while leaving the associated visual content unaltered. By only manipulating the metadata, MM-MEPA can still steer multimodal retrieval and induce attacker-desired model responses. We evaluate the attack across multiple benchmark settings and demonstrate its severity. MM-MEPA achieves an attack success rate of up to 91\% consistently disrupting system behaviors across four retrievers and two multimodal generators. Additionally, we assess representative defense strategies and find them largely ineffective against this form of metadata-only poisoning. Our findings expose a critical vulnerability in multimodal RAG and underscore the urgent need for more robust, defense-aware retrieval and knowledge integration methods.

Multimodal Large Language Models (MLLMs) are shifting towards "Thinking with Images" by actively exploring image details. While effective, large-scale training is computationally expensive, which has spurred growing interest in lightweight, training-free solutions. However, existing training-free methods suffer from two flaws: perceptual redundancy from indiscriminate cropping, which adds overhead and noise; and a drift between semantic intent and spatial attention, which prevents accurate localization of user-focused regions. To address these challenges, we propose AdaFocus, a novel training-free framework designed for adaptive visual reasoning. AdaFocus follows a two-stage pipeline: a confidence-based module decides when to crop, and a semantic-guided localization module determines where to crop. This enables adaptive visual reasoning without additional training. Experimentally, AdaFocus delivers substantial performance gains while achieving approximately 4.0\times speedup inference speedup than the SOTA method ZoomEyes, representing a significant advance in both accuracy and efficiency.

Craniofacial Superimposition is a forensic technique for identifying skeletal remains by comparing a post-mortem skull with ante-mortem facial photographs. A critical step in this process is Skull-Face Overlay (SFO). This stage involves aligning a 3D skull model with a 2D facial image, typically guided by cranial and facial landmarks' correspondence. However, its accuracy is undermined by individual variability in soft-tissue thickness, introducing significant uncertainty into the overlay. This paper introduces Lilium, an automated evolutionary method to enhance the accuracy and robustness of SFO. Lilium explicitly models soft-tissue variability using a 3D cone-based representation whose parameters are optimized via a Differential Evolution algorithm. The method enforces anatomical, morphological, and photographic plausibility through a combination of constraints: landmark matching, camera parameter consistency, head pose alignment, skull containment within facial boundaries, and region parallelism. This emulation of the usual forensic practitioners' approach leads Lilium to outperform the state-of-the-art method in terms of both accuracy and robustness.

Human intelligence naturally intertwines omni-modal perception -- spanning vision, audio, and language -- with complex reasoning and tool usage to interact with the world. However, current multi-modal LLMs are primarily confined to bi-modal interactions (e.g., vision-language), lacking the unified cognitive capabilities required for general AI assistants. To bridge this gap, we introduce OmniGAIA, a comprehensive benchmark designed to evaluate omni-modal agents on tasks necessitating deep reasoning and multi-turn tool execution across video, audio, and image modalities. Constructed via a novel omni-modal event graph approach, OmniGAIA synthesizes complex, multi-hop queries derived from real-world data that require cross-modal reasoning and external tool integration. Furthermore, we propose OmniAtlas, a native omni-modal foundation agent under tool-integrated reasoning paradigm with active omni-modal perception. Trained on trajectories synthesized via a hindsight-guided tree exploration strategy and OmniDPO for fine-grained error correction, OmniAtlas effectively enhances the tool-use capabilities of existing open-source models. This work marks a step towards next-generation native omni-modal AI assistants for real-world scenarios.

Recent advances in generative AI have demonstrated remarkable ability to produce high-quality content. However, these models often exhibit "Paradox of Simplicity": while they can render intricate landscapes, they often fail at simple, deterministic tasks. To address this, we formalize Obedience as the ability to align with instructions and establish a hierarchical grading system ranging from basic semantic alignment to pixel-level systemic precision, which provides a unified paradigm for incorporating and categorizing existing literature. Then, we conduct case studies to identify common obedience gaps, revealing how generative priors often override logical constraints. To evaluate high-level obedience, we present VIOLIN (VIsual Obedience Level-4 EvaluatIoN), the first benchmark focused on pure color generation across six variants. Extensive experiments on SOTA models reveal fundamental obedience limitations and further exploratory insights. By establishing this framework, we aim to draw more attention on AI Obedience and encourage deeper exploration to bridge this gap.

Deepfake speech utterances can be forged by replacing one or more words in a bona fide utterance with semantically different words synthesized with speech-generative models. While a dedicated synthetic word detector could be developed, we developed a cost-effective method that fine-tunes a pre-trained Whisper model to detect synthetic words while transcribing the input utterance via next-token prediction. We further investigate using partially vocoded utterances as the fine-tuning data, thus reducing the cost of data collection. Our experiments demonstrate that, on in-domain test data, the fine-tuned Whisper yields low synthetic-word detection error rates and transcription error rates. On out-of-domain test data with synthetic words produced with unseen speech-generative models, the fine-tuned Whisper remains on par with a dedicated ResNet-based detection model; however, the overall performance degradation calls for strategies to improve its generalization capability.

We introduce Reverse CAPTCHA, an evaluation framework that tests whether large language models follow invisible Unicode-encoded instructions embedded in otherwise normal-looking text. Unlike traditional CAPTCHAs that distinguish humans from machines, our benchmark exploits a capability gap: models can perceive Unicode control characters that are invisible to human readers. We evaluate five models from two providers across two encoding schemes (zero-width binary and Unicode Tags), four hint levels, two payload framings, and with tool use enabled or disabled. Across 8,308 model outputs, we find that tool use dramatically amplifies compliance (Cohen's h up to 1.37, a large effect), that models exhibit provider-specific encoding preferences (OpenAI models decode zero-width binary; Anthropic models prefer Unicode Tags), and that explicit decoding instructions increase compliance by up to 95 percentage points within a single model and encoding. All pairwise model differences are statistically significant (p < 0.05, Bonferroni-corrected). These results highlight an underexplored attack surface for prompt injection via invisible Unicode payloads.

The binary segmentation of whiteboard strokes is hindered by extreme class imbalance, caused by stroke pixels that constitute only $1.79%$ of the image on average, and in addition, the thin-stroke subset averages $1.14% \pm 0.41%$ in the foreground. Standard region metrics (F1, IoU) can mask thin-stroke failures because the vast majority of the background dominates the score. In contrast, adding boundary-aware metrics and a thin-subset equity analysis changes how loss functions rank and exposes hidden trade-offs. We contribute an evaluation protocol that jointly examines region metrics, boundary metrics (BF1, B-IoU), a core/thin-subset equity analysis, and per-image robustness statistics (median, IQR, worst-case) under seeded, multi-run training with non-parametric significance testing. Five losses -- cross-entropy, focal, Dice, Dice+focal, and Tversky -- are trained three times each on a DeepLabV3-MobileNetV3 model and evaluated on 12 held-out images split into core and thin subsets. Overlap-based losses improve F1 by more than 20 points over cross-entropy ($0.663$ vs $0.438$, $p < 0.001$). In addition, the boundary metrics confirm that the gain extends to the precision of the contour. Adaptive thresholding and Sauvola binarization at native resolution achieve a higher mean F1 ($0.787$ for Sauvola) but with substantially worse worst-case performance (F1 $= 0.452$ vs $0.565$ for Tversky), exposing a consistency-accuracy trade-off: classical baselines lead on mean F1 while the learned model delivers higher worst-case reliability. Doubling training resolution further increases F1 by 12.7 points.

Long-running agentic tasks, such as deep research, require multi-hop reasoning over information distributed across multiple webpages and documents. In such tasks, the LLM context is dominated by tokens from external retrieval, causing memory usage to grow rapidly and limiting decode performance. While several KV cache compression techniques exist for long-context inputs, we find that existing heuristics fail to support multi-step reasoning models effectively. We address this challenge with SideQuest -- a novel approach that leverages the Large Reasoning Model (LRM) itself to perform KV cache compression by reasoning about the usefulness of tokens in its context. To prevent the tokens associated with this management process from polluting the model's memory, we frame KV cache compression as an auxiliary task executed in parallel to the main reasoning task. Our evaluations, using a model trained with just 215 samples, show that SideQuest reduces peak token usage by up to 65% on agentic tasks with minimal degradation in accuracy, outperforming heuristic-based KV cache compression techniques.

Early ophthalmic screening in low-resource and remote settings is constrained by access to specialized equipment and trained practitioners. We present SKINOPATHY AI, a smartphone-first web application that delivers five complementary, explainable screening modules entirely through commodity mobile hardware: (1) redness quantification via LAB a* color-space normalization; (2) blink-rate estimation using MediaPipe FaceMesh Eye Aspect Ratio (EAR) with adaptive thresholding; (3) pupil light reflex characterization through Pupil-to-Iris Ratio (PIR) time-series analysis; (4) scleral color indexing foricterus and anemia proxies via LAB/HSV statistics; and (5) iris-landmark-calibrated lesion encroachment measurement with millimeter-scale estimates and longitudinal trend tracking. The system is implemented as a React/FastAPI stack with OpenCV and MediaPipe, MongoDB-backed session persistence, and PDF report generation. All algorithms are fully deterministic, privacy-preserving, and designed for non-diagnostic consumer triage. We detail system architecture, algorithm design, evaluation methodology, clinical context, and ethical boundaries of the platform. SKINOPATHY AI demonstrates that multi-signal ophthalmic screening is feasible on unmodified smartphones without cloud-based AI inference, providing a foundation for future clinically validated mobile ophthalmoscopy tools.

A mathematical concept is identified and analyzed that is implicit in the 2012 paper Turing Incomputable Computation, presented at the Alan Turing Centenary Conference (Turing-100, Manchester). The concept, called dynamic level sets, is distinct from mathematical concepts in the standard literature on dynamical systems, topology, and computability theory. A new mathematical object is explained and why it may have escaped prior characterizations, including the classical result of de Leeuw, Moore, Shannon, and Shapiro that probabilistic Turing machines (with bias $p$ where $p$ is Turing computable) compute no more than deterministic ones. A key mechanism underlying the concept is the Principle of Self-Modifiability, whereby the physical realization of an invariant logical level set is reconfigured at each computational step by an incomputable physical process.

Electrical manipulation of magnetization via current-induced spin orbit torques offers a promising route toward nonvolatile and energy efficient spintronic devices. In this work, we present a comprehensive investigation of SOTs in Py/FePS$_3$ bilayer devices, where Py/FePS$_3$ is a layered van der Waals antiferromagnetic insulator. Using low frequency harmonic Hall measurements, we quantify both field like and damping like torque components and examine their dependence on temperature. We find that interfacing Py with Py/FePS$_3$ leads to a pronounced enhancement of the field-like torque efficiency compared to Py reference devices, while the damping-like torque remains largely unaffected. Strikingly, the field like torque efficiency exhibits a strong temperature dependence, including a clear sign reversal upon cooling. This behavior occurs despite negligible charge current flow through the Py/FePS$_3$ layer, indicating that the observed torque modulation arises from interfacial effects rather than bulk transport. The close correlation between the temperature evolution of the field like torque and the antiferromagnetic ordering of Py/FePS$_3$ highlights the active role of antiferromagnetic insulators in controlling spin orbit torque symmetry and efficiency, and suggests new pathways for torque engineering in magnetic heterostructures.

Twisted moiré Dirac systems enable powerful miniband engineering but are largely fixed once the twist angle is set, whereas unidirectional (1D) electrostatic superlattices offer continuous control of Dirac anisotropy; yet robust single-particle gaps at charge neutrality are generally difficult to obtain in either setting. Here we show that combining the two into a hybrid moiré--1D superlattice provides a gate-defined configuration space that hosts both gap-opening resonances and strongly anisotropic gapless regimes. Using full-wave continuum miniband calculations for twisted bilayer graphene, we map the charge-neutrality-point (CNP) gap versus the 1D wavevector $\mathbf G_{\rm 1D}$ and identify a Dirac--Dirac resonance condition. At resonance, a single-particle CNP gap emerges from a parity--chirality selection rule for the resonant inter-cone coupling, which can be electrically reprogrammed by layer-asymmetric modulation that switches the relative chirality and the active mass channel. The insulating phase persists within a finite near-resonant window, providing quantitative fabrication tolerances, while off-resonant settings remain gapless but enable strong suppression of the transverse Dirac velocity and continuous anisotropic band renormalization. Hybrid moiré--1D superlattices thus provide a practical route to programmable Dirac minibands and electrically selectable mass channels in coupled Dirac systems.

The objective of this thesis is to advance the understanding of complex physical phenomena through the lens of statistical physics. Specifically, it addresses two fundamental questions: What types of interactions can induce buckling of slender structures when their temperature is increased? And, how can we devise an optimal strategy for locating a hidden target? The thesis is divided into two distinct parts, both employing mesoscopic descriptions -- neither fully microscopic nor fully macroscopic -- to capture the essential interactions and behaviours that qualitatively govern the phenomena under investigation. In the first part, we examine the buckling behavior of low-dimensional materials under thermal load. To this end, we develop a comprehensive model that characterises the system using a minimal setup for mimicking: (i) elastic and electronic degrees of freedom, and (ii) coupling between the elastic and the electronic modes. In the second part, we investigate stochastic resetting processes as a means to formulate efficient search strategies. We explore various resetting mechanisms to understand how to optimise the search performance in real scenarios, where: (i) resetting involves a finite cost, and (ii) the target location is only partially known.

We consider stochastic lattice gases with stationary product weights and a polynomial perturbation vanishing with the system size that leads to condensation. If the density of particles exceeds a critical value the system phase separates into a bulk with homogeneous distribution of particles and a condensed phase. Depending on parameter values, the latter consists of a single macroscopic cluster or a diverging number of independent clusters on a smaller scale. We establish the condensation transition via the equivalence of ensembles and the main novelty is a derivation of the cluster size distribution using size-biased sampling, generalizing previous work on zero-range and inclusion processes. Simulations of zero-range processes illustrate our theoretical results on the condensate scale and size distribution.

We revisit the machine-learning (ML) approach to the universal density functional $F[\mathbf{n}]$ of the one-dimensional Hubbard model with a site-dependent random potential $\mathbf{v}=\{v_{i}\}$. We generate exact ground-state data via exact diagonalization for a periodic chain with $L=8$ in the paramagnetic sector $(N_\uparrow,N_\downarrow)=(2,2)$, with site electron densities $n_{i} = n_{i\uparrow}=n_{i\downarrow}$. The resulting density-potential dataset is analyzed. Using principal component analysis of the joint feature space $(\mathbf n,\mathbf v)$, we identify the intrinsic low-dimensional structure of the data. Then, we restricted the study of the dataset with an energy-based filtering criterion to concentrate the data around weakly perturbed energy values with zero potential. A compact one-dimensional convolutional neural network is trained to learn the universal functional considering the lattice periodicity through unilateral wrapping and enforce the lattice symmetries by data augmentation (translations and mirror reflections), achieving near-exact predictions of $F[\mathbf n]$. Finally, we address the fact that accurate functional values do not necessarily imply accurate functional derivatives. By augmenting training with a variational consistency term that constrains the Euler-Lagrange relation between $\partial F/\partial n_i$ and the gauge-fixed potential we reconstruct the external potentials from automatic differentiation. These results clarify the roles of dataset geometry, symmetry, gauge fixing, and derivative-based constraints in learning physically consistent density functionals.

We study the self-diffusiophoresis of a spherical chemically active particle near a planar, impermeable wall, with a focus on the influence of particle orientation on propulsion. We analyze a Janus particle with asymmetric surface chemical activity, consisting of a small inert region within a catalytically active cap. While numerical simulations have been used to study such particles, they encounter difficulties resolving the flow and transport in the extreme near-wall regime due to geometric confinement and steep solute concentration gradients. We address this limitation through an asymptotic analysis in the near-contact limit, where the gap between the particle and the wall is narrow. In particular, we consider the distinguished limit in which the inert region is asymptotically comparable in size to the lubrication region. We analyze an axisymmetric configuration in which the inert face is oriented parallel to the wall and extend the analysis to slightly tilted orientations. We find that the capsize determines whether a tilted particle rotates back toward the axisymmetric state or continues to reorient, thereby characterizing its rotational stability in the near-contact regime.

Multi-component alloys offer broad tunability for addressing challenges in materials science, but their vast configurational space makes their surface chemistry highly sensitive to operating conditions, for example through adsorption and segregation. Here, we study Pd-Au-Cu alloy surfaces in H$_2$ and CO environments motivated by their use in H technologies, in particular plasmonic H$_2$ sensing, where alloying can mitigate limitations intrinsic to Pd such as hysteresis and CO poisoning. Modeling multicomponent surfaces with multiple adsorbate species under realistic conditions is challenging. To this end, we establish an accurate and efficient framework that combines machine-learned interatomic potentials trained on density functional theory data to generate training data for cluster expansions with effectively no limitations on training set size. By constructing continuous surface phase diagrams for H-CO coadsorption we find that coadsorption under operating conditions is governed primarily by the H coverage during annealing. Au-rich surfaces, formed under H-poor conditions, suppress both CO and H adsorption, while H-rich conditions yield Pd-rich surfaces that maintain higher H coverages compared to Pd at relevant CO partial pressures, indicating improved CO poisoning resistance. This effect is insensitive to relative amounts of Au and Cu, despite experimental evidence of the mitigating effect of specifically Cu on CO poisoning. Kinetic barriers for dilute alloy surfaces indicate that absorption pathways near Au are highly unfavorable, while Cu leave the energetics unchanged compared to pure Pd. This finding suggests that Cu in the surface region provides viable pathways to shuttle H into the material when Pd-dominated paths are blocked by CO.

We demonstrate a method to determine energy level degeneracies using non-equilibrium electronic transport through voltage-biased quantum dots. We establish the general validity of this approach using single and double quantum dots in bilayer graphene and GaAs. Unlike established methods based on entropy measurements or time-resolved tunneling statistics, our approach achieves comparable precision without requiring calibrated electron heating or real-time charge detection. We resolve the predicted symmetric shell structure in bilayer graphene quantum dots, including a singlet ground state at half filling and the ground state degeneracies of the first 13 carriers. Extending the method to double quantum dots, we observe degeneracy doubling associated with bonding and antibonding orbitals for a single carrier and a fourfold degeneracy for two carriers, previously inaccessible with existing techniques. These results establish a conceptually general and experimentally straightforward approach for probing energy level degeneracies in complex quantum systems.

Dimensionality of a physical system, conventionally an invariant geometric characteristic, fundamentally governs the universality class of phase transitions and the landscape of emergent collective phenomena. In low-dimensional or layered high-temperature superconductors, the macroscopic phase coherence of superconducting orders is typically confined in two dimensions, underscoring the critical role of phase fluctuations in determining the overall phase diagrams. Here, we strategically enhance the phase fluctuations by fabricating periodically arranged nano-holes in the infinite-layer nickelate superconducting films, effectively constructing Josephson junction arrays. In the nano-patterned films, the weakening of macroscopic phase coherence drives a two-stage superconducting transition towards an anomalous metallic ground state with saturated resistance. The emergence of charge-2e quantum oscillations manifests the coherence across the array, while an anomalous zero-field magnetoresistance peak signifies the extreme quantum phase fluctuations persisting to ultralow temperatures. Remarkably, with quantum fluctuations enhanced synergistically by nano-patterning and magnetic fields, an anomalous reversal of superconducting anisotropy is observed in Nd-nickelates, where in-plane critical fields fall below out-of-plane values. The evolution of anisotropy may unmask an internal exchange-Zeeman field coupled to the collective electronic states. Our results unveil how superconductivity evolves in response to phase fluctuations, establishing nano-patterning as a powerful paradigm to uncover hidden intertwined orders in strongly correlated systems.

Magnetoelectric coupling is vital for exploring fundamental science and driving the development of high-density memory and energy-efficient spintronic devices. Altermagnets, which merge the benefits of ferromagnets and antiferromagnets, pave the way for unprecedented magnetoelectric coupling effects. However, the spin splitting in altermagnets is robustly protected by spin space group symmetry, posing a significant challenge for external manipulation. Here, we propose to utilize the coupling between the layer degree of freedom and the altermagnet to achieve an altermagnetic multiferroic with strong magnetoelectric coupling. In the GaSe-VPSe3-GaSe sandwiched structure, the magnetic order can be switched between altermagnetic and conventional antiferromagnetic by controllably breaking and restoring the combined spatial inversion and time-reversal symmetry using sliding ferroelectricity. Moreover, our systematic investigation of all pathways revealed that the transition from a ferroelectric CB stacking, through an antiferroelectric CC stacking, to a ferroelectric BC stacking is the most favorable, with an energy barrier of only 50.13 meV/f.u.. More importantly, we reveal that the microscopic mechanism of the magnetic phase transition stems from the interlayer covalent bonding of Se-Se or Se-P atomic pairs at the interface. Our findings unveil a new form of magnetoelectric coupling and lay the groundwork for designing miniature information processing and multiferroic memory devices based on altermagnetism.

We present a first-principles molecular dynamics study of an excess electron in condensed phase models of solvated DNA bases. Calculations on increasingly large microsolvated clusters taken from liquid phase simulations show that adiabatic electron affinities increase systematically upon solvation, as for optimized gas-phase geometries. Dynamical simulations after vertical attachment indicate that the excess electron, which is initially found delocalized, localizes around the nucleobases within a 15 fs time scale. This transition requires small rearrangements in the geometry of the bases.

We investigate the orientational ordering and close-packing behavior of hard Gaussian overlap (HGO) particles, which are confined into a quasi-one-dimensional (q1D) channel. In the channel, particles are allowed to move along the channel and to rotate in three dimensions. Using the transfer operator method, we show that oblate particles align with their short axes along the channel, while prolate particles favor planar alignment perpendicular to the axis of the channel. While perfect orientational ordering develops in the fluid of oblate particles, the ordering is just partial in the fluid of prolate ones even at the close-packing density. The pressure ratio of freely rotating and parallel particles (P / P_parallel), which is an effective marker of structural changes, exhibits a single peak for oblate particles and no peak for prolate ones with increasing density. The close-packing behavior is characterized by exponents for the divergence of pressure (P ~ alpha P_parallel), the decay of orientational fluctuations (<(theta_p - theta)^2> ~ P^beta), and the behavior of the orientational correlation length (xi ~ P^gamma). The obtained values are beta = -1 and gamma = 0 for both oblate and prolate particles, while alpha = 2 for oblate and alpha = 1.5 for prolate particles. Moreover, prolate particles belong to the universality class of hard superellipses, where the combinations alpha + beta = 1/2 and beta + gamma = -1 hold exactly for any k > 1 (S. Mizani et al., Phys. Rev. E 111, 064121 (2025)). However, oblate particles do not belong to this universal class because alpha + beta = 1.

Chip-scale nonlinear optics enables strong light-matter interactions within compact devices, serving as a fundamental platform for multifunctional integrated photonics from classical optical signal processing to quantum information technologies. Transition metal dichalcogenide (TMDC) waveguides have recently emerged as a highly promising platform owing to their giant material nonlinearity and extended interaction lengths. To date, however, research has predominantly focused on conversion efficiency, leaving the mechanisms governing the polarization state of nonlinear signal largely unexplored. Here, we establish a comprehensive framework for engineering the polarization of second-harmonic generation (SHG) in 3R-MoS$_2$ waveguides. By synergizing polarization-resolved measurements with theoretical modeling, we reveal that the SHG polarization is determined by guided-mode interactions constrained by waveguide geometry and crystal symmetry, and further reshaped during propagation. We demonstrate that thickness-dependent guided-mode confinement and in-plane crystal symmetry provide robust, static control over SHG polarization, while propagation length offers a dynamic tuning knob for continuously tailoring the nonlinear output. Our findings provide a deterministic approach for on-chip polarization engineering, opening opportunities for reconfigurable nonlinear light sources and quantum photonic circuits.

The emergence of a long-range magnetic order in the atomically thin, two-dimensional (2D) limit has long remained a fundamental question in condensed matter physics. The advent of exfoliable van der Waals (vdW) materials, particularly transition-metal phosphorus trisulfides (T MPS3; T M = Fe, Ni, and Mn), provided the first experimental access to this regime and established a foundational platform for investigating 2D magnetism. The 2016 experimental demonstrations of intrinsic magnetism in monolayer FePS3 provided a platform to test key aspects of 2D Ising criticality in the true 2D limit. It was followed by a rapid growth resulting in a wealth of emergent phenomena arising from the interplay of low-dimensional magnetism and quantum materials. We begin this review with the historical development of vdW antiferromagnets and highlight the key physical insights gained over the past decade. We finish with emerging opportunities in which vdW antiferromagnets can serve as versatile platforms for exploring low-dimensional magnetism and its interplay with other quantum degrees of freedom.

We theoretically study nonequilibrium superconductivity in voltage-biased normal metal-superconductor-normal metal (NSN) junctions, focusing on effects of lead-coupling asymmetry and impurity scattering. Using the Keldysh Green's function technique, we extend the thermal-equilibrium mean-field BCS theory to the case where the system is out of equilibrium, to analyze superconducting properties in the nonequilibrium steady state. We find that, in close analogy with the thermal-equilibrium case, the inhomogeneous nonequilibrium Fulde-Ferrell-Larkin-Ovchinnikov (NFFLO) state induced by nonequilibrium electron distributions is highly sensitive to impurity scattering, whereas the uniform nonequilibrium BCS (NBCS) state remains robust against nonmagnetic impurities. Moreover, lead-coupling asymmetry is also found to suppress the NFFLO phase and to split the NBCS phase into two distinct regimes, characterized by the presence or absence of a chemical-potential imbalance between quasiparticles and the condensate. We identify a phase transition or a crossover between these two NBCS states, as well as parameter regimes exhibiting bistability. Our results provide a unified microscopic understanding of nonequilibrium superconductivity in NSN junctions under experimentally relevant conditions and are expected to provide a theoretical framework applicable to a broad class of nonequilibrium superconducting hybrid structures.

We study the energy expenditure and structural correlations in semi-dilute to concentrated suspensions of squirmers using active fast Stokesian dynamics simulations. Specifically, we simulate apolar active suspensions of squirmers, or 'shakers,' and show that shear enhances the total dissipation but reduces the relative viscosity for both puller- and pusher-type shakers. At low shear rates where activity dominates, pushers dissipate more energy than pullers, and more so at higher volume fractions, in contrast to bacterial suspensions displaying a 'superfluid' transition. At high shear rates where shear dominates, pullers and pushers behave effectively as passive spheres, generating negative normal stress differences due to shear-induced collision. Remarkably, the rate-dependent rheological responses are accompanied by unusual microstructural signatures of an enhanced nematic order and anisotropic pair correlation, both of which contribute to a higher viscosity under shear. Further simulations of self-propelled, neutral squirmers exhibit similar but weaker shear-thinning, highlighting the importance of activity over motility, underpinned by hydrodynamic interactions. Overall, our results elucidate the interplay of internal activity and external flow on the dissipation and microstructure in sheared active suspensions of squirmers.

In one-dimensional diffusive processes with discrete steps characterized by geometrically decaying magnitudes, the usual Gaussian broadening familiar from Brownian motion is replaced by bounded probability distributions over particle positions that are characterized by multi-scale fractal structures. In this work, we study random walks with shrinking steps (known as Bernoulli convolutions), focusing on their behavior in the vicinity of the dyadic contraction ratio 1/2. Our analytical and numerical results show that the coarse-grained Shannon entropy of particle distributions induced by Bernoulli convolutions exhibits a local maximum at the dyadic ratio, arising from the competition between diffusive spreading, which increases entropy, and emergent fine structure, which tends to decrease it. This entropy maximum is a general property of systems driven by non-Gaussian discrete noise, whose dynamics near stable fixed points can be viewed as an autoregressive process - an approximation that is mathematically equivalent to unbiased random walks with shrinking steps. We discuss potential implications of Bernoulli convolution dynamics for protocell self-replication and vesicle proliferation, establishing a link between our information-theoretic approach and biophysical models of cell division.

Nonbenzenoid carbon frameworks expand low-dimensional material design via controlled asymmetry. Here, we show the experimentally realized 4-5-6-8 carbon nanoribbon establishes a topology-driven paradigm for multiproperty engineering, not just a graphene variant. Using hybrid DFT, tight-binding, and molecular dynamics in a multiscale framework, we demonstrate the symmetry-broken lattice stabilizes hierarchical bonds within standard energy ranges. This geometry produces a robust semiconducting state (hybrid gap >1 eV) and enables strain as a controllable modulation parameter. A tight-binding Hamiltonian fitted only at equilibrium accurately captures strain-dependent band evolution, proving the essential physics is topology-dominated. Mechanical analysis reveals high stiffness with fracture governed by the largest polygons, showing asymmetry redistributes stress without compromising integrity. Intrinsic phonon scattering suppresses thermal conductance, enabling favorable thermoelectric performance without extrinsic disorder. Optical response confirms non-equivalent ring connectivity reorganizes interband transitions, promoting strong visible absorption and efficient photocarrier generation. These results position topology as a governing parameter coupling elasticity, electronics, thermal transport, and optics, establishing the 4-5-6-8 nanoribbon as a unified platform for predictive design of multifunctional carbon materials.

While the generalized Langevin equation (GLE) is a powerful tool to understand the behavior of complex dissipative systems, driving by external fields renders standard GLE construction workflows invalid. Filtering approaches that separate fluctuations from the non-equilibrium response can sometimes circumvent the need for a non-equilibrium formalism when the residual fluctuations are homoskedastic, stationary, and preferably Gaussian. Here, we introduce the temperature time series from Boulder, Colorado, as representative of the more general and complex case where the filtered time series remains non-Gaussian, non-stationary, and heteroskedastic. With this example, we develop a protocol to build an accurate and efficient low-dimensional description of the weather fluctuations. Our protocol classifies the weather data based on the position in the annual cycle, and introduces local homoskedasticity as a metric to identify seasons of likely stationarity. Within these seasons, we build pseudo-equilibrium models. Leveraging state-based generalized master equation modelling as an alternative to the GLE, we resolve difficulties like non-Gaussianity and position dependence of the memory (friction) kernel. Our data-driven approach accurately reproduces the evolving fluctuations of the Boulder temperature time series, illustrating the feasibility of our method as a general tool to describe driven, dissipative systems.

Quantum magic resources, or nonstabilizerness, are a central ingredient for universal quantum computation. In noisy many-body systems, the interplay between these resources and errors leads to sharp magic phase transitions. However, the microscopic mechanism behind these critical phenomena is still an open question, especially since early empirical evidence showed conflicting results regarding their universality classes. In this work, we provide a comprehensive picture of magic phase transitions for the class of encoding-decoding quantum circuits to resolve these ambiguities. We analytically show that the behavior of magic resources is fundamentally dictated by the chosen measurement protocol. When we fix, or post-select, the class of measurement syndromes, the magic transition inherits the universal features of the error-resilience phase transition in the circuits. Interestingly, this clean transition survives even for fully random Haar encoders showing that it is a consequence of initial's state retrieval, and not an artifact of the Clifford encoders. On the other hand, if we consider realistic Born-rule sampling, the intrinsic statistical fluctuations of a given syndrome measurement act as a relevant perturbation. This brings in strong finite-size drifts and an apparent multifractality, which end up altering the critical behavior of the system. We reveal that magic phase transitions are actually direct manifestations of error-resilience thresholds, rather than independent critical phenomena, reconciling conflicting observations from the earlier literature. Ultimately, our framework clarifies how the quantum computational power can survive, or be irreversibly destroyed, due to the competition between scrambling, measurements, and errors.

We present a joint experimental and theoretical study of a ratchet implemented in arra ys of evanescently coupled plasmonic waveguides with tailored losses. In this setup the time-periodic dissipation is the only active mechanism and notably, we find better rectified transport and lower losses in the transmitted signal with increased local dissipation. Using Floquet theory, we uncover a driving regime that allows efficient directional tr ansport for suitable driving frequencies and loss rates, which are linked to linear qu asienergy bands with minimal losses. These regions are separated from non-resonant beh avior by sharp transitions with characteristic exceptional points in the spectrum. Direct experimental observation of the Floquet-dissipative ratchet effect using a comb ination of real- and Fourier-space leakage radiation microscopy is provided.

Thermalization in driven open quantum systems is often described as ordinary thermodynamics supplemented by additional dissipation that depends on how the system is driven. We show that when relaxation is treated consistently at the generator level within Gaussian GKLS dynamics, thermodynamics itself acquires an intrinsically dynamical state space. For a frequency-modulated harmonic oscillator coupled to a thermal bath, detailed balance selects a relaxing quadratic frame characterized by an emergent frequency $ω_I(t)$. This coordinate obeys an Onsager-type relaxation equation with a positive kinetic coefficient set by the bath coupling. As a consequence, the first law acquires an additional generalized work term, and eliminating auxiliary variables yields an internal energy of the form $E=E(ω_{I},\dotω_{I})$, or equivalently $E=E(S,\dot S)$ during thermalization. The rate dependence originates from detailed-balance relaxation rather than external driving protocols. Our results predict a measurable energy shift proportional to $\dotω_{I}/α$, providing a direct signature that thermalization enlarges thermodynamic state space.

We extend the semi-classical analysis of the double-well potential to a quartic system featuring four degenerate minima. Utilizing the Feynman path integral in imaginary time, we identify longitudinal, transverse, and diagonal instanton configurations that mediate tunneling between minima. The zero mode is handled by transforming to a rotating frame whose origin lies on the classically determined path. By generalizing the dilute instanton gas approximation to account for these distinct pathways, we derive the coherent Rabi-type oscillations and the energy splittings of the four lowest-lying states. These semi-classical results are validated against high-precision numerical diagonalization, showing excellent agreement in the deep semi-classical limit. We further identify a critical coupling regime where the discrete $D_4$ symmetry undergoes a `melting' transition into a continuous $O(2)$ rotational symmetry, signaling a fundamental breakdown of the localized instanton description.

A colony of the queenless ant species, \emph{Pristomyrmex punctatus}, can broadly be seen as consisting of small-body sized worker ants and relatively larger body-sized cheater ants. Hence, in the presence of inter-colony migration, a set of constituent colonies act as a metapopulation exclusively composed of cooperators and defectors. Such a set-up facilitates an evolutionary game-theoretic replication-selection model of population dynamics of the ants in a metapopulation. Using the model, we analytically probe the effects of territoriality induced hostility. Such hostility in the ant-metapopulation proves to be crucial in preventing the tragedy of the commons, specifically, the workforce, a social good formed by cooperation. This mechanism applies to any metapopulation -- not necessarily the ants -- composed of cooperators and defectors where inter-population migration occurs asymmetrically, i.e., cooperators and defectors migrate at different rates. Furthermore, our model validates that there is evolutionary benefit behind the queenless ants' behavior of showing more hostility towards the immigrants from nearby colonies than those from the far-off ones. In order to calibrate our model's parameters, we have extensively used the data available on the queenless ant species, \emph{Pristomyrmex punctatus}

In many real-world systems, partial synchronization is the dominant dynamical regime and, in systems such as the brain, is often accompanied by collective oscillations in which multiple overlapping modes interact to produce complex rhythmic activity. Here, we investigate duplex networks with reactive interlayer links, where full synchronization cannot be achieved. We show that when interlayer frequency differences between mirror nodes are uniformly distributed with sufficient width, the network self-organizes into collective macroscopic oscillations composed of multiple interacting modes. By linking macroscopic phase transitions to microscopic directed information transfer between nodes, we uncover the mechanisms underlying the emergence of these multimodal dynamics.

Understanding the dynamic behavior of microstructures formed under fusion conditions is critical for designing high-performance structural materials for fusion reactors. Under fusion conditions, cavities of core-shell structures are formed due to the interaction between irradiation-induced vacancies and H and He atoms produced via transmutation. In this study, thermodynamic analysis and molecular dynamics simulations are combined to investigate the atomic-scale mechanisms and dynamic response of core-shell cavities formed in BCC-Fe under applied stress/strain fields. The thermodynamic analysis provides both the foundational reference for cavity structures under fusion neutron irradiation and the initial configurations for atomistic simulations. Building on this framework, atomic-scale simulations demonstrate that H and He play a decisive role in the stress-strain response and the evolution of elastic-plastic deformation within the cavities. In core-shell configurations, H atoms serve a function analogous to that in He-filled cavities, synergistically interacting with He to induce cavity deformation under mechanical loading.

Deuterium-tritium fusion reactions produce high-energy neutrons that can transmute materials into valuable isotopes. Over the next ten years, the cost of fusion neutrons is projected to decrease by roughly seven orders of magnitude. Most ($\sim$5 orders of magnitude) is technological overhang driven by the low availability of current experiments; the remaining $\sim$2 orders of magnitude require higher plasma gain and lower capital intensity. We introduce the levelized cost of a neutron (LCON), an economic metric analogous to the levelized cost of energy that gives the minimum neutron value for economic breakeven of a fusion system. LCON depends on plasma gain, capital intensity, availability, and neutron flux, and is offset by revenue from co-produced electricity, precious metals, and radioisotopes. The revenue per neutron spans at least ten orders of magnitude, from electricity and gold ($\sim$\$$10^{-20}$/neutron) to actinium-225 ($\sim$\$$10^{-10}$/neutron), defining a `neutron ladder': a staged, revenue-positive development pathway from current fusion devices to terawatt-scale power plants.

Academic publishing requires solving a collective coordination problem: among thousands of possible publication venues, which deserve a community's attention? A clear consensus helps scholars allocate attention, match submissions to appropriate outlets, and evaluate scholars for hiring and promotion. Yet preferences are not centrally coordinated--they emerge within each field over time. Here we ask whether all fields have arrived at similar solutions to this coordination problem, and whether preferences vary systematically with individual characteristics. Using an adaptive survey of 3,510 US tenure-track faculty yielding 163,002 pairwise comparisons across 8,044 venues, we show that fields occupy a wide spectrum of coordination. Economics, Chemistry, and Physics exhibit strong consensus, with respondents agreeing on elite venues and accurately predicting one another's choices. Computer Science and Engineering show fragmented preferences distributed across hundreds of outlets with minimal overlap. Within fields, preferences correlate with institutional prestige--faculty at elite institutions prefer higher-ranked venues--and with gender, as men prefer higher-ranked venues than women even after accounting for prestige and career stage. Scholars realize their personal preferences more successfully than their respective fields' consensus preferences, indicating that heterogeneity, not just selective hierarchy, shapes publishing outcomes. Journal Impact Factors explain only 64% of preference choices, systematically undervaluing what fields actually prefer. These results quantify how publication preferences vary across the structural diversity of academic fields.

Aeroacoustic emissions from intense evaporation are widely measured yet often treated as noisy byproducts and used mainly in empirical monitoring. Here, we show that airborne sound encodes physics-governed sub-millisecond fingerprints of vapor-jet dynamics in excessive vaporization, exemplified by vapor keyholes in laser metal processing. From first principles, we develop a vapor-jet-cavity oscillation framework and incorporate it into an aeroacoustic formulation, thereby coupling measured sound to transient cavity depth and oscillation frequency. Reconciled with synchronized multimodal in-situ data, airborne acoustics enable accurate tracking of vapor-cavity properties within tens to hundreds of microseconds. Combined with newly discovered correlations, cavity-jet-acoustic theory recasts the transition from steady, pore-free to pore-shedding vaporizations as a critical-frequency event. Aeroacoustic emissions thus become scalable, physics-guided, and cost-efficient probes of rapidly evolving liquid-vapor systems.

Refractory multi-principal element alloys (MPEAs) are key research focus for excellent high-temp properties and engineering potential. Deformation mechanisms/mechanical behaviors of quaternary NbTaTiZr MPEA under high strain rates/extreme temps remain unclear. We built a variable-composition ML potential for NbTaTiZr, combined with MD simulations to study effects of crystal orientation, strain rate, temp, composition on compressive mechanics. NbTaTiZr shows structural/mechanical anisotropy in compression [111] max yield strength, [110] min (prone to twinning), [100] via local disorder/dislocation slip (dominant 1/2<111> dislocations). At 10^10 s^-1, yield strength rises sharply, disordered structures increase; high strain rates suppress dislocations to promote disordering. Retains high strength at 2100 K. Higher Nb/Ta boosts yield strength, Ti/Zr reduce it. Reveals MPEA mechanical anisotropy and strain-rate-dependent disordering, guiding high-performance refractory alloy design.

DFT+X methods, such as DFT+U and DFT+DMFT, are important supplements to standard density functional theory when strong on-site Coulomb interactions are present. However, the involvement of external parameters in the underlying model Hamiltonian introduces intrinsic ambiguity when comparing the total energies obtained with different model parameters. This renders DFT+X approaches semi-empirical and severely hinders their capability to describe phase ordering and phase stability, especially when reliable experimental benchmarks are unavailable, such as under high pressure. In this work, we resolve this longstanding problem by proposing a general linear correction method that eliminates the ambiguous energy contributions introduced by the model Hamiltonian in DFT+X approaches, thereby enabling direct comparison of their energies calculated with different interaction parameters. The method is demonstrated and validated within the framework of DFT+U, an important member of the DFT+X family. It is then applied to important nuclear materials of uranium-based binaries U-M (M=Al, Ga, In) alloys. With this approach, we resolve the long-standing discrepancy between theoretical predictions and experimental observations of phase stability with unprecedented accuracy, and predict several previously unknown stable intermetallic compounds under high pressure. The broad applicability of the method is further confirmed by accurate predictions of formation enthalpies for diverse systems, including Np-Al, U-Si, and Cu-O binaries, the ternary MnSnAu compound, and oxygen adsorption on the Cu(111) surface. This work establishes linear-corrected DFT+U as a fully first-principles approach and validates the linear correction method as a robust and general scheme that can be readily extended to other DFT+X methods.

We present the synthesis of broadband multilayer metamaterial absorbers (MMA) based on MXenes, which are novel two-dimensional conductive materials with higher ohmic losses than copper. MXene resonator of different conductivity can be implemented at each layer or across the same layer, offering increased design flexibility. We examine the possibility of utilizing topology-optimized stereolithography (SLA) 3D-printed substrates as a complementary means for enhancing absorption. Combining MXenes with topology optimized 3D-printable structures paves the way for realizing MMAs of enhanced absorption.

We present a systematic, computationally efficient approach for synthesizing 3D-printable all-dielectric devices. Inverse-design optimization methods lead to devices of a continuous dielectric constant profile with complex and conformal shapes. However, stereolithography 3D printers have a limited range of materials; usually, only resin and air are available. As the size and complexity of the devices increase, performing simulations of the entire detailed manufacturable device becomes computationally challenging or even prohibitive. We introduce the LOCABINACONN methodology for transforming an optimized device of a continuous material profile to a manufacturable one while preserving performance as close as possible to the continuous case. The LOCABINACONN is a local and computationally efficient methodology where we identify suitable air/resin configurations that will substitute non-manufacturable material components without simulating the entire manufacturable device. This work paves the way for synthesizing optimized larger-scale 3D-printable devices in a computationally tractable manner.

Three-photon (3-P) fluorescence microscopy enables deep in vivo imaging with subcellular resolution, but its performance is fundamentally constrained by the maximum permissible laser power required to avoid tissue heating and photodamage. Under these power-limited conditions, fluorescence signal generation, image contrast, and achievable imaging depth are strongly affected by the illumination beam profile and aberration correction strategy. In this paper, we showed that using a fixed illumination beam size was suboptimal across different imaging depths. We further showed that conventional Zernike-based adaptive optics (AO) correction degrades under reduced Gaussian illumination beam sizes due to loss of modal orthogonality. This degradation results in slow convergence, unintended focal and field-of-view shifts, and excessive wavefront deformations. To overcome these limitations, we introduced a depth-adapted AO framework in which both the illumination beam profile and the aberration correction basis were dynamically matched to the imaging conditions. By combining depth-optimised beam underfilling with a bespoke set of illumination-matched aberration modes, we achieved faster and more stable AO convergence, enhanced fluorescence signal and image quality during deep in vivo multi-channel neuroimaging. Together, these results established a practical and robust AO-enabled three-photon microscopy strategy that maximised imaging performance under realistic power constraints.

Understanding the mechanisms that govern species coexistence and biodiversity represents a fundamental challenge in ecology. This study extends the classic rock-paper-scissors model by introducing a context-dependent higher-order interaction mechanism where intraspecific competition is dynamically regulated by local resource availability. Crucially, our quantitative analysis reveals that higher-order interactions significantly modulate the system's structural organization: Enhanced strength of higher-order interactions leads to a decrease in spatial wavelength, resulting in the formation of more compact species domains. However, this structural change makes the system more sensitive to mobility, shifting the extinction threshold to lower values. These findings highlight the dual role of resource-mediated higher-order regulation: it promotes local pattern formation but alters the system's resilience to dispersal, providing new theoretical perspectives for biodiversity conservation.

Chiral bound states in the continuum (BICs) are confined photonic modes with infinite quality factors and chiral response, offering significant potential for chiral optics. Although a novel type of spin-orbital-locking chiral BIC was recently predicted in magneto-optical (MO) photonic crystals (PhCs) that break time-reversal symmetry (TRS), its experimental realization has remained elusive. Here, we report the first experimental observation of such chiral BICs in self-biased MO PhCs, which operate without external magnetic fields. Moreover, we experimentally demonstrate that the chirality and the surrounding near-circular polarization of these chiral BICs can be switched simply by reversing the remanent magnetization, without any structural changes. Unlike conventional chiral BICs that preserve TRS, these magnetically induced chiral BICs exhibit exceptional robustness against structural imperfections and perturbations. This work represents a significant advancement in the topological photonics of chiral BICs, opening new pathways toward robust chiral optical devices.

Controlling the flow of light within complex and dynamic environments is essential for a wide range of applications, from deep-tissue imaging and optogenetics to precision phototherapy. Typically, such light flows are controlled using external optical systems requiring line-of-sight access or by embedded nanoparticle scatterers with limited directional control, underscoring the need for mobile photonic agents capable of actively delivering and steering light within complex media. Here, we present magneto-photonic metaparticles: mobile, magnetically actuated microstructures that integrate a magnetic core with a nanoimprinted photonic surface. This hybrid design merges the reconfigurability of photonic metasurfaces with the mobility of magnetic actuation, enabling programmable translation, rotation, and real-time beam steering in aqueous media. In a concept-proof demonstration, we realize polymeric metaparticles with embedded magnetic core and nanoimprinted surface that exhibit controlled locomotion and active, magnetically programmable beam steering. Our design approach further points to more sophisticated metaparticle designs with high-efficiency, polarization-insensitive light steering, compatible with a scalable, single-step nanoimprint process. The mobile magneto-photonic metaparticle platform combines metasurface-level optical control with magnetic mobility, offering a versatile and scalable approach for active photonic control in complex environments.

In a typical Rydberg mixer, modulated communication signals are detected using a radio frequency (RF) heterodyne technique. The mixer outputs an intermediate frequency (IF), which must be filtered and mixed down to baseband. In this work, we apply an RF-homodyne technique to demonstrate simultaneous detection and a direct, baseband readout of the in-phase (I) and quadrature (Q) components of standard communication signals using a dual ladder Rydberg receiver. We further show that the inherent polarization sensitivity of this receiver can be used to determine the signal's angle of arrival. We also compare the dual ladder system with a typical Rydberg mixer. The RF-heterodyne-based system's maximum detectable symbol rate is constrained by a signal amplitude which decays with the heterodyne field's detuning from the Rydberg-Rydberg atomic transition used to detect the signal, but the dual ladder design is not subject to this limitation. However, the dual ladder system is more sensitive to low-frequency noise. As a result, its performance is degraded relative to its conventional counterpart when subjected to pink noise. We show that once pink noise effects have been accounted for, both systems behave comparably.

High-brightness X-ray Free Electron Lasers (FELs) produce spatially and temporally coherent pulses on attosecond to femtosecond timescales, providing a transformative tool for discovery across biology, chemistry, physics, and materials science. However, most existing FELs are kilometer-scale facilities with billion-dollar construction costs and low repetition rates (about 100 Hz), which limits their accessibility and scientific throughput. This paper introduces a novel design for a compact, high-repetition-rate (MHz) EUV to 1 nm soft X-ray FEL with a footprint of less than 100 meters. This design is suitable for installation within university or research institution settings where space is limited. The facility leverages a multi-turn recirculating linear accelerator that integrates state-of-the-art superconducting accelerator technology with recent advances in diffraction-limited storage rings. We present the conceptual design and analyze the impact of incoherent and coherent synchrotron radiation, demonstrating that these effects are not limiting factors for achieving high-quality electron beams. Such a compact X-ray FEL facility would substantially reduce both construction and operational costs, greatly expanding access to these powerful research tools. Furthermore, the design provides a potential upgrade path to generating hard X-ray radiation by incorporating high accelerating gradient structures to further accelerate a portion of the MHz electron beam.

Collectives of entities, including groups of living systems and artificial swarms, self-organize to achieve common goals. Collective systems frequently employ a division of labor, wherein individuals take on different tasks or perform different amounts of work. However, the rules and mechanisms used by collectives to divide labor remain poorly understood. In this study, we investigate the methods used by biological collectives to complete tasks using experimental and theoretical approaches. We use social insects, which form remarkably integrated societies, as model systems to study division of labor. We specifically explore how workload inequality might arise by studying digging behavior in Solenopsis invicta fire ants. We introduce an experimental technique for estimating each ant's workload by tracking individual grain depositions during digging behavior. These experimental results suggest that workload distribution becomes more unequal for increasing group size. We then implement an agent-based cellular automata model which predicts experimental trends, and suggests that local decisions driven by crowding emergently account for the scaling of workload inequality. Finally, we examine experimental workload results and show that the number of ``active'' digging ants roughly scales with the square root of the total group size. This finding parallels scaling laws from other domains of social and natural science, such as Price's law, which suggest that a core group of individuals perform the majority of work. We introduce a simplified rate equation model which recovers the square root scaling via a quadratic failure rate. Together, these results provide a mechanistic explanation for the emergent workload scaling patterns in collectives.

This study presents an automated topology optimization framework for designing pentamode acoustic metamaterials. It provides precise control over the material effective acoustic properties while minimizing the shear modulus to achieve fluid-like behavior. The approach combines low-frequency homogenization for accurate property evaluation and the adjoint method for efficient sensitivity analysis. The Virtual Temperature Method (VTM) ensures structural connectivity and manufacturability, addressing the typical challenges of low-stiffness, high-mass-density microstructures. The framework is demonstrated through the design of a Lüneburg lens and an acoustic invisibility cloak for underwater applications. Acoustic-elastic simulations validate the performance of both the unit cells and the complete devices. This method eliminates the need for predefined geometries, offering a flexible, reliable, and scalable alternative to conventional parametric optimization. It provides a powerful tool for the automated design of complex, anisotropic, pentamode metamaterials.

We use micromagnetic simulations to demonstrate how machine learning can be applied to arrays of chiral magnonic resonators to build a magnonic full adder. The chiral magnonic resonators have form of nano-sized permalloy disks that nonlinearly scatter spin waves propagating in a YIG waveguide. The spin waves are injected from multiple outputs, and the dynamic stray magnetic field of the scattered spin waves is sampled in multiple locations to form several physical output signals. These signals are weighted and combined, either linearly or nonlinearly, to satisfy the logic output of a full adder. The process is known as training and forms the device's output layer. The full adder's performance is evaluated in terms of robustness to input and output noise for a given number of physical output signals and the form of the output layer. When the output layer is linear, as few as three physical signals per logical output are sufficient. When a multilayer perceptron neural network forms the output layer, the number of required output signals is reduced to one, and a nearly perfect classification accuracy is achieved when appropriate preprocessing and augmentation strategies are used.

We report on extensive experimental investigation of novel barium chalcogenides (BaGa4Se7, BaGa2GeSe6 and BaGa2GeS6) for generation of a strong narrowband terahertz (THz) radiation. The Ba-containing compounds demonstrate the large bandgap values, high optical damage threshold and wide transparency range from the visible to mid-infrared region, coupled with narrowband transparency windows in the THz frequency range. In these windows a THz generation via the second-order nonlinear process of optical rectification occurs under optical pumping by a 50-GW high-power Cr:forsterite laser leading to intrinsically narrowband THz pulses. The highest optical-to-THz conversion efficiency of 0.8^10(-5) is reached in the Z-cut BaGa2GeS6 crystal in the saturation regime and corresponds to an output energy of 20 nJ. Ba compounds provide narrow-bandwidth THz pulses at the frequencies in the range from 1.9 to 2.5 THz with a relative bandwidth of 3-4%, necessary for imaging and spectroscopy for security or medical applications.

We present mrfmsim, an open-source Python package that facilitates the design, simulation, and analysis of magnetic resonance force microscopy (MRFM) experiments. MRFM is a scanning-probe technique that detects magnetic resonance from nanoscale ensembles of nuclear or electron spins with a force sensor. Because MRFM experiments are complex and operate at sensitivity limits, numerical simulation is essential for designing experiments and estimating per-spin sensitivity and imaging resolution from measured signals. In this paper, we highlight the challenges of developing MRFM simulations and show that software designed to simulate specific experiments only in a rapidly evolving experimental field can yield erroneous results. The mrfmsim package addresses these challenges by supporting post-definition customization without rewriting the internal model and by employing a plugin system for extending functionality. We show that the package's modular, extendable, and readable architecture improves reproducibility and accelerates development.

One of the main goals of artificial life research is to recreate in artificial systems the trends for ever more complex and novel entities, interactions and processes that we see in Earth's biosphere, that is, to create open-ended systems. In this paper, we test for Tokyo type 1 open-ended evolution (OEE) of the Tree of Life Simulation (ToLSim), an artificial life software created by Lana Sinapayen. To do so, we conducted an experiment to measure evolutionary activity statistics. These require us to define the notion of components. Here, we define components as the agent's genes. The results show that ToLSim is capable of exhibiting unbounded total cumulative evolutionary activity. However, total and median normalized cumulative evolutionary activity appear bounded and new evolutionary activity is persistently null, suggesting that ToLSim is not open-ended. Further studies on ToLSim could repeat this experiment with individuals or even species, rather than genes, to test whether the present results are valid.

This paper addresses the estimation of a dynamic interaction network, a network of influence among individuals, under projected visual stimuli to quantify the influences of inter-individual interactions and external stimuli on collective behavior. Building upon our previously proposed network estimation model, which assumes a Boids-type model and employs a sparse regression framework to infer inter-individual influence networks from trajectory data, we extend the formulation by introducing a stimulus term. This enables the model to capture how individuals react to and propagate externally projected visual stimuli within the group. The resulting framework allows simultaneous estimation of inter-individual and stimulus-related interaction strengths. We also introduce entropy-based indices to capture the possible biases of individuals' influence. Our experiments with fish schools under projector-based visual stimuli demonstrate the effectiveness of the proposed indices in quantifying schooling behavior and identifying influential individuals within the group, serving as the basis for real-time, interpretable metrics of collective dynamics.

Generalization to novel visual conditions remains a central challenge for both human and machine vision, yet standard robustness metrics offer limited insight into how systems trade accuracy for robustness. We introduce a rate-distortion-theoretic framework that treats stimulus-response behavior as an effective communication channel, derives rate-distortion (RD) frontiers from confusion matrices, and summarizes each system with two interpretable geometric signatures - slope ($β$) and curvature ($κ$) - which capture the marginal cost and abruptness of accuracy-robustness trade-offs. Applying this framework to human psychophysics and 18 deep vision models under controlled image perturbations, we compare generalization geometry across model architectures and training regimes. We find that both biological and artificial systems follow a common lossy-compression principle but occupy systematically different regions of RD space. In particular, humans exhibit smoother, more flexible trade-offs, whereas modern deep networks operate in steeper and more brittle regimes even at matched accuracy. Across training regimes, robustness training induces systematic but dissociable shifts in beta/kappa, revealing cases where improved robustness or accuracy does not translate into more human-like generalization geometry. These results demonstrate that RD geometry provides a compact, model-agnostic lens for comparing generalization behavior across systems beyond standard accuracy-based metrics.

The contributions of model complexity, data volume, and feature modalities to knowledge graph-based drug repurposing remain poorly quantified under rigorous temporal validation. We constructed a pharmacology knowledge graph from ChEMBL 36 comprising 5,348 entities including 3,127 drugs, 1,156 proteins, and 1,065 indications. A strict temporal split was enforced with training data up to 2022 and testing data from 2023 to 2025, together with biologically verified hard negatives mined from failed assays and clinical trials. We benchmarked five knowledge graph embedding models and a standard graph neural network with 3.44 million parameters that incorporates drug chemical structure using a graph attention encoder and ESM-2 protein embeddings. Scaling experiments ranging from 0.78 to 9.75 million parameters and from 25 to 100 percent of the data, together with feature ablation studies, were used to isolate the contributions of model capacity, graph density, and node feature modalities. Removing the graph attention based drug structure encoder and retaining only topological embeddings combined with ESM-2 protein features improved drug protein PR-AUC from 0.5631 to 0.5785 while reducing VRAM usage from 5.30 GB to 353 MB. Replacing the drug encoder with Morgan fingerprints further degraded performance, indicating that explicit chemical structure representations can be detrimental for predicting pharmacological network interactions. Increasing model size beyond 2.44 million parameters yielded diminishing returns, whereas increasing training data consistently improved performance. External validation confirmed 6 of the top 14 novel predictions as established therapeutic indications. These results show that drug pharmacological behavior can be accurately predicted using target-centric information and drug network topology alone, without requiring explicit chemical structure representations.

Predicting the response of an observed system to a known input is a fruitful first step to accurately control the system's dynamics. Despite the recent advances in fully data-driven algorithms, the most interpretable way to reach this goal is through mechanistic mathematical modeling. Here, we leverage optimal control and propose a closed-loop iterative method to choose among a set of candidate models the one that most accurately predict an observed system. We assume that one has control over an input of the observed system and access to measurements of its response. Our approach is to identify the input control that maximally discriminates the response of the candidate models, allowing us to determine which model is best by comparing such responses with the observed data. We demonstrate our proposed framework in numerical simulations before applying it during an electrophysiology experiment, successfully discriminating between different models for photocurrents produced via opsin dynamics.

The capacity to precisely edit genomes has outpaced our ability to predict the consequences. A cell can be genetically perfect and therapeutically useless: edited exactly as intended, yet unstable, drifting toward unintended fates, or selected for properties that compromise safety. This paradox reflects a deeper gap in how we evaluate biological intervention. Current frameworks excel at measuring what was done to a cell but remain blind to what the cell has become. We argue that this blindness stems from treating cells as collections of independent variables rather than as dynamical systems occupying positions on high-dimensional state manifolds. Drawing on Waddington's epigenetic landscape, we propose geometric stability as a missing axis of evaluation: the directional coherence of cellular responses to perturbation. This metric distinguishes interventions that guide cells coherently toward stable states from those that scatter them across the state manifold. Validation across diverse perturbation datasets reveals that geometric stability captures regulatory architecture invisible to conventional metrics, discriminating pleiotropic master regulators from lineage-specific factors without prior biological annotation. As precision medicine increasingly relies on cellular reprogramming, the question shifts from ``did the intervention occur?'' to ``is the resulting state stable?'' Geometric stability provides a framework for answering.

Chemical space exploration underlies drug discovery, yet most generative models treat chemical space as a fixed, implicitly learned distribution, focusing on sampling molecules rather than deliberately designing the space itself. We introduce SpaceGFN, a generative framework that elevates chemical space to a programmable computational object: a controllable degree of freedom enabling explicit construction and adaptive traversal of structured molecular universes. SpaceGFN decouples space definition from exploration. Users specify building blocks and reaction rules to construct chemically and synthetically coherent spaces, while a GFlowNet performs efficient, property-biased sampling within them. In Discovery mode, we demonstrate programmable space design through two strategies. A pseudo-natural product space assembles natural product-like architectures. An evolution-inspired (Evo) space recombines endogenous metabolite fragments via enzyme-consistent transformations, introducing an evolutionary prior into chemical generation. This bias yields favorable shifts in predicted metabolic and toxicological profiles while preserving pharmacological diversity, supported by broad docking enrichment across therapeutic targets. In Editing mode, SpaceGFN enables reaction-consistent lead optimization through a curated toolkit of executable synthetic transformations, allowing local, synthesis-aware modification of existing compounds instead of unrestricted graph mutation. Across 96 drug targets, SpaceGFN achieves strong optimization performance while maintaining structural diversity under synthetic constraints. By integrating programmable chemical universe construction with flow-based exploration and reaction-level editing, SpaceGFN establishes a general paradigm for deliberate navigation of therapeutic chemical space.

Understanding how long-term stimulation reshapes neural circuits requires uncovering the rules of brain plasticity. While short-term synaptic modifications have been extensively characterized, the principles that drive circuit-level reorganization across hours to weeks remain unknown. Here, we formalize these principles as a latent dynamical law that governs how recurrent connectivity evolves under repeated interventions. To capture this law, we introduce the Stimulus-Evoked Evolution Recurrent dynamics (STEER) framework, a dual-timescale model that disentangles fast neural activity from slow plastic changes. STEER represents plasticity as low-dimensional latent coefficients evolving under a learnable recurrence, enabling testable inference of plasticity rules rather than absorbing them into black-box parameters. We validate STEER with four benchmarks: synthetic Lorenz systems with controlled parameter shifts, BCM-based networks with biologically grounded plasticity, a task learning setting with adaptively optimized external stimulation and longitudinal recordings from Parkinsonian rats receiving closed-loop DBS. Our results demonstrate that STEER recovers interpretable update equations, predicts network adaptation under unseen stimulation schedules, and supports the design of improved intervention protocols. By elevating long-term plasticity from a hidden confound to an identifiable dynamical object, STEER provides a data-driven foundation for both mechanistic insight and principled optimization of brain stimulation.

Computational pathology models that use digitized histopathology whole-slide images have the potential to become a cost-effective and scalable alternative to molecular assays for the prediction of genomic biomarkers, a key task in precision oncology. However, as new genomic biomarkers are discovered or quantified, large, labeled datasets must be prospectively collected to train new models. To address this challenge, we developed MARBLE, a multimodal contrastive pretraining strategy that integrates structured biomarker knowledge into representation learning of histopathology images. MARBLE aligns histopathology-derived representations with representations of genomic biomarkers generated by a large language model (LLM) and a protein language model (PLM). This biologically informed alignment enables data-efficient generalization to novel, out-of-distribution biomarkers. Using the MSK-IMPACT cohort of over 40,000 patients across multiple biomarker panel versions, we design experiments grounded in real-world data to demonstrate the value of our proposed approach.

The ribosome is responsible for protein synthesis in all cells, and is the largest energy consumer in the cell. We propose that the ribosome originated as a mutualistic symbiont of an RNA-dependent RNA polymerase ribozyme, supplying peptides that enhanced replication. As life transitioned from the RNA to the RNA-protein world, autonomous replicators became irreversibly addicted to the ribosome for producing replication proteins. Subsequent evolution is construed as a ribosomal takeover, whereby the ribosome evolved to consume most of the resources of the cell, while other cellular componentry ensured the propagation of the ribosome. Under this perspective, the ribosome is the ultimate biological selfish element.

Detecting weak signals buried in complex, non-Gaussian noise is a fundamental challenge in science and engineering, with applications ranging from radar systems and communications to industrial monitoring and gravitational wave detection. The Rao detector, a key concept in this domain, achieves asymptotically optimal performance as the number of measurements increases, but requires precise knowledge of the data's statistical properties, often relying on simplified noise models. We propose a hybrid framework that combines a lightweight neural network with the Rao detection framework to address this limitation. The neural network, trained on noise-only data, learns the optimal multivariate nonlinearity, transforming noisy data to enhance signal detectability. The newly introduced LRao detector then fully extracts the signal information, achieving asymptotically optimal performance even under challenging noise conditions. Validated on both simulated and real-world magnetic sensor data, our method significantly outperforms conventional approaches. By bridging data-driven techniques with model-based signal processing, it offers a robust and interpretable solution for signal detection across diverse applications.

Non-Line-of-Sight (NLoS) sensing and detection of low-observable (stealth) targets are challenging for conventional radar due to blockage and severe propagation loss. Intelligent Reflective Surface (IRS)-assisted radar can extend the field-of-view (FOV), but common architectures rely on the four-hop radar--IRS--target--IRS--radar link, whose attenuation limits estimation performance. This paper proposes an alternative architecture, that exploits the target-scattered component received at a spatially separated IRS and redirected back to a mono-static radar receiver. The geometry provides bi-static/multi-static-like diversity using a passive panel, while retaining a mono-static front-end and avoiding inter-node time synchronization concerns. We develop a signal model for the proposed configuration and recast it into a compact, parameterized form that is suitable for angle estimation. Using this reformulation, we derive the Fisher Information Matrix and the associated Cramér--Rao Lower Bounds (CRLB) for target azimuth and elevation angles with respect to the IRS. Numerical evaluations quantify the impact of various signal-model parameters on the achievable bounds. These results provide insights on the parameter-estimation limits within the FOV against SNR, snapshots and IRS elements.

Ray tracing has become a standard for accurate radio propagation modeling, but suffers from exponential computational complexity, as the number of candidate paths scales with the number of objects raised to the power of the interaction order. This bottleneck limits its use in large-scale or real-time applications, forcing traditional tools to rely on heuristics to reduce the number of path candidates at the cost of potentially reduced accuracy. To overcome this limitation, we propose a comprehensive machine-learning-assisted framework that replaces exhaustive path searching with intelligent sampling via Generative Flow Networks. Applying such generative models to this domain presents significant challenges, particularly sparse rewards due to the rarity of valid paths, which can lead to convergence failures and trivial solutions when evaluating high-order interactions in complex environments. To ensure robust learning and efficient exploration, our framework incorporates three key architectural components. First, we implement an \emph{experience replay buffer} to capture and retain rare valid paths. Second, we adopt a uniform exploratory policy to improve generalization and prevent the model from overfitting to simple geometries. Third, we apply a physics-based action masking strategy that filters out physically impossible paths before the model even considers them. As demonstrated in our experimental validation, the proposed model achieves substantial speedups over exhaustive search -- up to $10\times$ faster on GPU and $1000\times$ faster on CPU -- while maintaining high coverage accuracy and successfully uncovering complex propagation paths. The complete source code, tests, and tutorial are available at https://github.com/jeertmans/sampling-paths.

We propose a block least mean square (LMS) algorithm to monitor the longitudinal power profile of a fiber-optic link through receiver-based digital data from a coherent detector. Compared to the benchmark least squares (LS) method, the proposed algorithm does not require large matrix inversions or batch processing, thus allowing the received data to be processed in blocks of minimum size by an overlap-save algorithm, reducing complexity and latency. We propose an efficient implementation of the method with a stochastic gradient update leveraging a key computation in the frequency domain, offering computational savings over state-of-the-art monitoring techniques. We test the proposal in different scenarios by means of numerical simulations.

Inertial sensors can track object kinematics, however, unbounded drift from integrating noisy signals makes them impractical for MRI motion correction at millimeter resolution and minute-long scans. We introduce MR-Compass, which exploits the MRI system's static magnetic and gravitational fields to estimate 3-DOF orientation at 2 kHz directly, without integration, eliminating random-walk. The remaining 3-DOF translation is recovered via phase correlation from the MRI data. We experimentally validate the efficacy of the method retrospectively using a 3D radial koosh-ball sequence and prospectively using 2D EPI fMRI during large volunteer motions. MR-Compass followed by phase-correlation achieved a mean accuracy of 0.6$^o$ and 0.4 pixels across all experiments. Image quality improved when motion correction was applied in all volunteer scans for both retrospective and prospective correction cases. MR-Compass was effective in measuring head motion in the MRI scanner with high accuracy at unprecedented sample rates, and enabled both retrospective and prospective reconstruction to improve image quality by aligning the k-space data appropriately and by reducing the motion related artifacts.

Tagged MRI enables tracking internal tissue motion non-invasively. It encodes motion by modulating anatomy with periodic tags, which deform along with tissue. However, the entanglement between anatomy, tags and motion poses significant challenges for post-processing. The existence of tags and imaging blur hinders downstream tasks such as segmenting anatomy. Tag fading, due to T1-relaxation, disrupts the brightness constancy assumption for motion tracking. For decades, these challenges have been handled in isolation and sub-optimally. In contrast, we introduce a blind and nonlinear inverse framework for tagged MRI that, for the first time, unifies these tasks: anatomical image recovery, high-resolution cine image synthesis, and motion estimation. At its core, the synergy of MR physics and generative priors enables us to blindly estimate the unknown forward imaging models and high-resolution underlying anatomy, while simultaneously tracking 3D diffeomorphic Lagrangian motion over time. Experiments on tagged brain MRI demonstrate that our approach yields high-resolution anatomy images, cine images, and more accurate motion than specialized methods.

We propose a simple quantum algorithm for implementing the diffusion step of grid-based Bayesian filters. The method encodes the advected state density and the process noise density into quantum registers and realizes diffusion using a quantum Fourier transform--based adder. This avoids the explicit convolution required in classical implementations and the repeated coin-flip operations used in quantum random walk approaches. Numerical simulations using a gate-based quantum computing simulator confirm that the proposed approach reproduces the desired probability densities while requiring significantly fewer quantum gates and much shallower circuit depth.

Low-altitude network is a key enabler for extending coverage and recovering connectivity in 6G systems, especially when terrestrial infrastructure is unavailable. This paper studies a uncrewed aerial vehicle (UAV)-mounted rotatable intelligent reflecting surface (IRS) as a low-altitude reflector between a blocked base station (BS) and a ground terminal (GT). Unlike the conventional isotropic-element assumption, each IRS element is modeled with a hemispherical directive radiation pattern, whose boresight can be adjusted via element rotations. We formulate a new optimization problem that jointly designs IRS phase shifts, per-element rotation vectors, and UAV placement to maximize the received signal-to-noise ratio (SNR). Leveraging the problem structure, we derive closed-form solutions for phase alignment and element rotations, showing that the optimal boresight points are along the internal angular bisector between the BS-IRS and GT-IRS directions. With these closed forms, the design reduces to a placement optimization problem over a box-constrained airspace; we solve it using an efficient projected gradient algorithm with majorization-minimization update and a global Lipschitz constant. Numerical results demonstrate substantial SNR gains from directive elements and reveal a fundamental trade-off between directional gain and path loss, yielding useful insights into low-altitude deployment of UAV-mounted IRSs.

Machine learning (ML)-assisted outage-based resource allocation has recently emerged as an effective alternative to conventional scheduling methods in reliability-critical wireless systems. However, existing approaches are fundamentally limited to single-resource allocation, whereas modern and emerging systems increasingly require the simultaneous allocation of multiple resources to meet aggregate rate and reliability constraints. In this paper, we extend outage-based learning to the bulk resource allocation regime, where a user requires at least $D$ reliable resources from a pool of $R$ candidates. We first introduce a practical allocation policy, termed gate + top-$D$ allocation (GTBA), which combines threshold-based admission control with ranking-based selection. We then propose a novel ranking-aware bulk outage loss (RBOL) that provides a differentiable surrogate for the bulk outage event induced by GTBA, explicitly accounting for both gate failures and ranking errors near the selection boundary. An exact reliability analysis is developed, establishing a decomposition of bulk outage probability (BOP), identifying dominant failure mechanisms and deriving an oracle lower bound that characterizes the fundamental performance limit. Extensive simulations under balanced, light and heavy stress regimes demonstrate that RBOL consistently outperforms conventional pointwise losses and baselines, achieving substantial reductions in BOP and remaining significantly closer to the oracle bound across a wide range of operating conditions. These results confirm that set-level, ranking-aware training objectives are essential for reliable ML-assisted bulk resource allocation.

False discovery rate (FDR) control is a popular approach for maintaining the integrity of statistical analyses, especially in high-dimensional data settings, where multiple comparisons increase the risk of false positives. FDR control has been extensively researched for real-valued data. However, the complex data case, which is relevant for many signal processing applications, remains widely unexplored. We therefore present a fast and FDR-controlling variable selector for complex-valued high-dimensional data. The proposed Complex-Valued Terminating-Random Experiments (CT-Rex) selector controls a user-defined target FDR while maximizing the number of selected variables. This is achieved by optimally fusing the solutions of multiple early terminated complex-valued random experiments. We benchmark the performance in sparse complex regression simulation studies and showcase an example of FDR-controlled compressed-sensing-based single snapshot multi-source detection and direction of arrival (DOA) estimation. The proposed work applies to a wide range of research areas, such as DOA estimation, communications, mechanical engineering, and magnetic resonance imaging, bridging a critical gap in signal processing for complex-valued data.

The pinching-antenna system (PASS) has been proposed as a promising solution for mitigating line-of-sight (LoS) blockages by dynamically repositioning pinching antennas (PAs) along a dielectric waveguide. This paper develops a fairness-oriented downlink design for a non-orthogonal multiple access (NOMA)-enabled PASS, where the longitudinal placement of PAs and the NOMA power allocation coefficients are jointly optimized to maximize the minimum user signal-to-interference-plus-noise ratio (SINR) across all users under transmit power and waveguide constraints. A soft-blockage channel model incorporating waveguide attenuation and imperfect channel state information (CSI) is developed. To ensure the feasibility of successive interference cancellation under CSI uncertainty, a conservative SINR evaluation framework is proposed. The resulting non-convex max-min SINR optimization problem is efficiently solved using a tailored particle swarm optimization (PSO) algorithm. Numerical results demonstrate that the proposed design improves the minimum user SINR by approximately 7-10 dB compared with fixed-antenna systems and non-robust optimization baselines under moderate blockage and imperfect CSI.

This paper presents joint sampling frequency offset (SFO) estimation and compensation algorithms based on the Farrow structure. Unlike conventional approaches that treat estimation and compensation separately, the proposed framework exploits the interpolator structure to enable a low-complexity, fully time-domain solution applicable to arbitrary bandlimited signals, without imposing constraints on the waveform or requiring Fourier transform based processing. The estimation stage can operate on a real-valued component of a complex signal and supports the simultaneous estimation of SFO and sampling time offset, while being inherently robust to other synchronization impairments such as carrier frequency offset. The proposed estimation algorithms rely on two complementary methods, specifically, Newton's method and iterative least-squares formulation. The implementations of the estimators are presented and the overall computational complexity is analyzed, showing that the complexity scales only linearly with the number of samples employed. Numerical results for real and complex multisine and bandpass-filtered white noise signals demonstrate accurate estimation and effective compensation over a wide range of operating conditions, confirming the flexibility and efficiency of the proposed approach. Moreover, the influence of the Farrow structure approximation error on the SFO estimation accuracy is investigated.

The pinching-antenna systems (PASS) enable blockage mitigation in urban micro (UMi) networks through flexible antenna placement. However, the joint optimization of antenna positions and beamforming precoding is inherently nonconvex and becomes significantly more challenging under user mobility. To address this issue, we propose a bilevel optimization framework for dynamic antenna positioning and beamforming precoding design. In the outer level, a soft actor-critic (SAC) agent learns a continuous control policy for real-time antenna positioning, while in the inner level, zero-forcing (ZF) precoding is applied based on the instantaneous effective channel. Numerical results demonstrate that the proposed framework significantly improves spectral efficiency (SE) and enhances robustness against user mobility and random blockages.

Active noise control (ANC) is an effective approach to noise suppression, and the filtered-reference least mean square (FxLMS) algorithm is a widely adopted method in ANC systems, owing to its computational efficiency and stable performance. However, its convergence speed and noise reduction performance are highly dependent on the step size parameter. Common step-size algorithms-such as normalized and variable step-size variants-require additional computational resources and exhibit limited adaptability under varying environmental conditions. To address this challenge, a novel Monte Carlo gradient meta-learning (MCGM) approach is proposed herein to determine an appropriate step size, into which a forgetting factor is incorporated to mitigate the impact of initial zero effect. Compared to other algorithms, the proposed method imposes no additional computational burden on FxLMS operations. Numerical simulations involving real-world acoustic paths and noise signals further confirm its effectiveness and robustness.

This paper investigates the application of Index Modulation (IM) to Modulation on Conjugate-Reciprocal Zeros (MOCZ) to enhance spectral efficiency (SE) in short packet communications. The proposed IM-MOCZ scheme splits an $N$-bit message into two streams: $N-K$ bits select one of $2^{N-K}$ uniquely designed codebooks, while the remaining $K$ bits are transmitted with conventional MOCZ using the selected codebook. At the receiver, Root Finding Minimum Distance (RFMD) and Direct Zero-Testing (DiZeT) detectors evaluate all candidate codebooks and compute penalty metrics, with a majority-vote rule selecting the most confident codebook and recovering the transmitted message. The proposed IM-MOCZ provides higher SE gains than conventional MOCZ, and simulations demonstrate improved bit error rate (BER) performance for larger $K$ relative to $N$.

Passive long-wave infrared (LWIR) absorption-based ranging relies on atmospheric absorption to estimate distances to objects from their emitted thermal radiation. First demonstrated decades ago for objects much hotter than the air and recently extended to scenes with low temperature variations, this ranging has depended on reflected radiance being negligible. Downwelling radiance is especially problematic, sometimes causing large inaccuracies. In two new ranging methods, we use characteristic features from ozone absorption to estimate the contribution of reflected downwelling radiance. The quadspectral method gives a simple closed-form range estimate from four narrowband measurements, two at a water vapor absorption line and two at an ozone absorption line. The hyperspectral method uses a broader spectral range to improve accuracy while also providing estimates of temperature, emissivity profiles, and contributions of downwelling from a collection of zenith angles. Experimental results demonstrate improved ranging accuracy, in one case reducing error from over 100 m when reflected light is not modeled to 6.8 m with the quadspectral method and 1.2 m with the hyperspectral method.

Quantum frequency conversion (QFC) is essential for bridging the spectral gap between stationary qubits and low-loss optical communication channels. In this work, we demonstrate a short-wavelength-pumping QFC with the first-order quasi-phase matching period of 3.07 um on thin-film lithium niobate, converting ultraviolet photons to the telecom C-band. By constructing a theoretical model that correlates the normalized conversion efficiency with domain defects in the short-period phase-matched waveguide, we found the critical tolerance of domain defects along the waveguide should be $\le 2$ (excluding the ends). Based on this, we achieved a theoretical limit normalized conversion efficiency of 839%/(W*cm^2) for the fundamental guided mode through fabrication optimization. Furthermore, we propose a robust noise suppression strategy for short-wavelength pumping by utilizing the counter-tuning behaviors of difference-frequency generation and spontaneous parametric down-conversion. By combining these advances with ultra-narrowband filtering, we achieve a record-high external efficiency of 28.8% and an ultra-low noise of 35 counts per second. This high-performance QFC connecting ultraviolet and telecom bands satisfies the stringent requirements for long-lived remote ion-ion entanglement in scalable quantum networks [W.-Z. Liu et al., Nature (2026)].

After an exhaustive introduction highlighting the strengths and weaknesses of the non-local models proposed so far as ultraviolet completions of the Starobinsky theory, we propose a new nonlocal completion of a general $f(R)$ theory (in the Einstein's frame) suitable for driving inflation in the early universe consistently with observations. The nonlocal theory shares with $f(R)$ the same background solutions and the same equations of motion for perturbations at linear and nonlinear level. Therefore, the classical cosmological observables are not affected by the nonlocal operators needed for the quantum completion. Our construction applies to any local action written in the Einstein's frame, but we will provide the details only for two explicit examples: the Starobinsky model and a general $f(R)$ theory. The new model overcomes the incompatibility of renormalizability and stability present in the previous proposals. Since the nonlocal theory is at least super-renormalizable, but can also be finite depending on the details of the model, this work shows the consistency of the inflationary paradigm with a well-defined quantum theory of gravity at high energy. We could rephrase the latter statement saying that the success of the $f(R)$ theories in cosmology can be traced back to the existence of an ultraviolet completion that preserves all the classical features. The inflationary paradigm survives, or it is actually insensitive to quantum gravity, because it is an exact solution of quantum gravity, up to perturbative corrections.

Analogue studies represent an important tool in modern Physics. In particular, analogue gravity had a strong success in the recent years with the demonstrations of Hawking radiation and superradiance of analogue black holes in classical and quantum fluids. So far, the metric of the analogue black holes was mostly fixed by the conditions of the experiment, preventing the simulation of any significant evolution of their properties, such as the change of their mass, their spatial motion, gravitation attraction to other bodies, and, ultimately, black hole mergers. Polariton condensates represent a perfect setting for the analogue simulation of black hole evolution and mergers because of the velocity-dependent losses creating a convergent flow associated with each quantum vortex, which thus becomes an analogue black hole capable of spatial motion. We show that while two vortices are unable to form a common horizon, four or more vortices can exhibit a complete black hole merger, with the radius of the common horizon given by a simple geometrical law. We also discuss the difference between the horizon and the apparent horizon in these analogue black holes with quantized constituents.

We present $\mathcal{H}\mathtt{-EFTCAMB}$, the official successor to $\mathtt{EFTCAMB}$. The original $\mathtt{EFTCAMB}$ is designed as a consistent and numerically stable implementation of the effective field theory (EFT) of dark energy in the Einstein-Boltzmann code $\mathtt{CAMB}$. On top of this, $\mathcal{H}\mathtt{-EFTCAMB}$ introduces a new Horndeski module that supports computing cosmology for an arbitrary input covariant Horndeski Lagragian. $\mathcal{H}\mathtt{-EFTCAMB}$ supports both mapping the Horndeski theory to an EFT lagrangian to solve in the EFT framework as well as directly solving for the scalar field equations of motion derived from the covariant Lagrangian. The latter approach also works for the cases when the Horndeski field experiences turn-overs, e.g. oscillation, where the EFT approach breaks down. The Horndeski module has been validated by comparing internally with existing models in the original $\mathtt{EFTCAMB}$ and externally with $\mathtt{hi\_class}$. $\mathcal{H}\mathtt{-EFTCAMB}$ features a flexible Python wrapper that is seamlessly integrated into the widely utilized cosmological sampler $\mathtt{Cobaya}$. \heft~is publicly available and serves as a comprehensive tool for testing gravity against the precision data from current and next-generation surveys.

We demonstrate the novel effect of gain-induced spectral shifts in the type-II parametric down-conversion (PDC) process, which results in a transition from degenerate to non-degenerate PDC with increasing parametric gain. This effect, originating from the second-order dispersion terms, significantly alters the properties of PDC in the high-gain regime, where it leads to increased distinguishability of the generated photon pairs. The effect is established by evaluating a rigorous theoretical model, which is based on solving a system of coupled integro-differential equations for monochromatic operators. The widely used spatially-averaged approximate model fails to reproduce this important effect.

The exchange of quantum information among nodes in a quantum network is one of the main challenges in modern technologies. Superconducting waveguide QED networks hold great potential for realizing distributed quantum computation, where distinct nodes communicate via itinerant single photons. Yet, different frequencies among the nodes restrict their applicability and limit scalability. Here we derive the controls required to shape single photons arbitrarily detuned with respect to their natural frequency, allowing thus for an on-demand and deterministic exchange of quantum information among frequency detuned nodes. We provide a theoretical framework, analyzing the properties of the controls for typical photon shapes, identifying operation regimes amenable for experimental realization. We then show how these controls enable frequency-selective quantum state transfer among non-resonant and distant nodes of a realistic network. In addition, we also provide a simple extension for remote entanglement generation between these nodes. The suitability and high-fidelity of these protocols is supported by numerical simulations, highlighting the novel networking possibilities unlocked when shaping frequency-tunable single photons.

Hook errors are a major challenge in implementing logical operations with the surface code, because they can reduce the fault distance below the code distance. This motivates syndrome-extraction circuits that suppress hook-error effects for the stabilizer layouts that appear during logical operations. However, the existing methods either increase circuit depth or require simultaneous execution of measurements and CNOT gates, both of which introduce additional overheads and degrade the threshold. We propose the ZX interleaving syndrome extraction, which preserves the full fault distance $d$ for any surface-code layout with regular stabilizer tiles at minimum depth, i.e., four layers of CNOT gates, without requiring additional circuit depth or simultaneous execution of measurements and CNOT gates. The key idea is to interleave the Z and X stabilizer tiles so that hook-error edges in the decoding graph are shortened and effectively eliminated. Numerical simulations under uniform depolarizing noise for memory and lattice-surgery experiments confirm that the proposed method achieves a full fault distance of $d$, whereas the best existing minimum-depth approach achieves $d-1$. Since the full fault distance is achievable for any regular tiling layout of the surface code, the proposed method may serve as an indispensable technique for practical fault-tolerant quantum computation.

The realization of fault-tolerant quantum computation hinges on the ability to execute deep quantum circuits while maintaining gate fidelities consistently above error-correction thresholds. Although neutral-atom arrays have recently demonstrated high-fidelity two-qubit gates and early-stage logical quantum processors, sustaining such high performance across deep, repetitive circuits remains a formidable challenge due to cumulative motional heating and atom loss. Here we demonstrate a sustainable neutral-atom framework that overcomes these limitations by integrating a suite of hardware-efficient mid-circuit operations. We report a two-qubit controlled logic gate with a raw fidelity of 99.60(1)%, which is further increased to a fidelity of 99.81(1)% via non-destructive erasure detection. Crucially, by implementing in-circuit Raman sideband cooling and qubit re-initialization, we demonstrate that gate fidelities can be maintained at the ~99.8% level across multiple operational rounds without observable degradation. By actively managing the internal and motional entropy of the system mid-stream, our in-situ refreshable architecture provides a critical pathway for executing the repeated syndrome-extraction cycles required for large-scale, continuous quantum error correction.

The quantum phonon laser state is a vibrational state generated by phonon coherent amplification based on quantum mechanics. Its core is coherent excitation and manipulation of phonon quantum states by controlling phonon dynamics. This technology breaks classical limits of traditional phonon lasers, offering new methods for quantum information. Previous research on quantum phonon lasers focused on quantum van der Pol oscillators. As typical nonlinear quantum systems, they show significant value in trapped-ion systems. These breakthroughs extend nonlinear dynamics into the quantum domain and provide platforms for exploring quantum nonlinear phenomena. Although realized in two-ion systems, practical applications remain challenging. This paper explores how a single trapped ion generates quantum phonon laser states using a three-level model. By solving the quantum master equation numerically, steady-state characteristics are analyzed, focusing on quantum statistics including the Wigner function and second-order correlation function. An experimental scheme is proposed based on a single trapped 40Ca+ ion, using bichromatic blue-sideband and red-sideband lasers to generate quantum phonon laser states. By introducing the characteristic function of motional states, precise quantum state tomography is achieved. Additionally, a two-level model discusses the phonon laser threshold effect. However, the three-level model shows significantly different thresholds and more accurately describes the quantum phonon laser's physical mechanisms.

The qubit mapping problem is a challenge in quantum computing that is related to mapping logical qubits to the physical ones on the quantum computer. Due to the diversity of quantum computer topologies and circuits, numerous approaches solving this problem exist. Finding the best solution for specific combination of topology and circuit remains difficult and no unified framework currently exists for systematically evaluating and comparing qubit mapping algorithms across different cases. We present QuMeld, an open-source framework that is designed for solving this issue. The framework currently supports six qubit mapping algorithms, sixteen quantum computer topologies and multiple evaluation metrics. The modular design of the framework allows integration of new mapping algorithms, quantum circuits, hardware topologies, and evaluation metrics, ensuring extensibility and adaptability to future developments.

We identify an operational principle that singles out Projection-Valued Measures (PVMs) among general Positive Operator-Valued Measures (POVMs), bridging the modern quantum measurement theory and the traditional formulation based on projective measurements of physical observables. We reformulate Ozawa's intersubjectivity condition, which requires inter-observer agreement of the measurement outcomes, in a quantitative manner within the framework of generalized probabilistic theories. We prove that (i) a POVM is a PVM if and only if its every coarse-graining is intersubjective, and (ii) a system is classical if and only if intersubjectivity is preserved under any coarse-graining, establishing a complete characterization of the physical observables and the classical theory. Furthermore, measurements with intersubjectivity are sufficiently rich for the informational tasks of state tomography and state discrimination, testifying to its operational significance in quantum and beyond information processing.

We investigate how internal coupling influences the operation and performance of a quantum Otto cycle operating as the Gibbs-state limit cycle (GSLC), equilibrating limit cycle (ELC), and non-equilibrating limit cycle (NELC). We show that the internal coupling significantly broadens the operational regime of the cycle. In particular, in parameter regimes where the uncoupled Otto cycle fails to operate as any thermal machine, the coupled system can function as an engine or a refrigerator. For the GSLC, in which we assume that the system quickly equilibrates during the isochoric processes, the internal coupling not only shifts and enlarges the operational regime but also enhances the efficiency and the coefficient of performance (COP), allowing the performance to exceed the standard Otto bounds while remaining below the Carnot limit. For ELC and NELC, we validate the global approach of the Gorini--Kossakowski--Sudarshan--Lindblad (GKSL) master equation by comparison with the GSLC, and examine the NELC for finite interaction time and the ELC for infinite interaction time. Although the efficiency and COP of NELC are lower than those of ELC, shorter interaction times yield higher power output, consistent with the power--efficiency trade-off.

In this paper, we introduce new mass-type invariants for time-symmetric initial data in space-times obeying the Dominant Energy Condition. When the cosmological constant is positive, these invariants, unlike the total Hawking mass, turn out to be genuinely effective in providing new characterizations of de Sitter solution. From a theoretical standpoint, this opens a new perspective on how one might refine the rigidity statement originally proposed by Min-Oo in his well known conjecture, later refuted by the counterexamples of Brendle, Marques, and Neves.

We develop a general theory of Landau-Zener (LZ) tunneling in a two-level system with amplitude-dependent, sign-reversible nonlinear coupling, distinguishing it fundamentally from conventional on-site nonlinearity. Through a combination of analytical and phase-space analysis, we show that beyond a critical interaction strength, the nonlinear coupling fundamentally reshapes the adiabatic energy landscape, introducing a topological twisted and knotted structure. This structure leads to a complete breakdown of the standard exponential LZ formula, even in the adiabatic limit. Central to this anomalous behavior is the emergence of a black-hole-like fixed point, which acts as a universal attractor: upon traversing the critical region, all quantum trajectories converge to this fixed point, irreversibly erasing any memory of the initial state. From this fixed-point picture, we derive an exact analytical expression for the adiabatic tunneling probability, revealing a characteristic power-law dependence on both linear and nonlinear coupling strength. Our work establishes a paradigmatic framework for nonlinear-coupling-induced anomalous adiabaticity breaking and offers a universal mechanism for state control in driven quantum and wave systems.

Efficiently characterizing large quantum states and processes is a central yet notoriously challenging task in quantum information science, as conventional tomography methods typically require resources that grow exponentially with system size. Here, we introduce a provably efficient and structure-agnostic learning framework for noisy $n$-qubit quantum circuits under generic noise with arbitrary noise strength. We first develop a sample-efficient learning algorithm for unital noisy quantum states. Building on this result, we extend the framework to quantum process tomography, obtaining a unified protocol applicable to both unital and non-unital channels. The resulting approach is input-agnostic and does not rely on assumptions about specific input distributions. Our theoretical analysis shows that both state and process learning require only polynomially many samples and polynomial classical post-processing in the number of qubits, while achieving near-unit success probability over ensembles generated by local random circuits. Numerical simulations of two-dimensional Hamiltonian dynamics further demonstrate the accuracy and robustness of the approach, including for structured circuits beyond the random-circuit setting assumed in the theoretical analysis. These results provide a scalable and practically relevant route toward characterizing large-scale noisy quantum devices, addressing a key bottleneck in the development of quantum technologies.

It is shown that anisotropic matter black holes with two parameters $w$ and $K$ are identified as nonlinear electrodynamics (NED) black holes with power-index $s$ and charge term $ξ(s,q)$ by introducing a NED term. These NED black holes include constant scalar hair ($s=1$), charged quantum Oppenheimer-Snyder ($s=3/2$), and Einstein-Euler-Heisenberg ($s=2$) black holes derived from their known actions. Rotating NED black holes can be obtained from rotating anisotropic matter black holes when replacing $w$ and $K$ by $2s-1$ and $ξ(s,q)$. The extremal rotating NED black holes being the boundary between rotating charged NED black hole and naked singularity are derived as functions of the rotation parameter $a(q)$.

In this work, we analyze the Casimir effect associated with a massive, non-minimally coupled scalar field in static, spherically symmetric black hole spacetimes arising in bumblebee gravity. Three distinct solutions are considered, corresponding to different vacuum expectation value configurations of the Lorentz-violating vector field, including metric and \textit{metric-affine} scenarios. Finite-size effects are implemented through the Thermo Field Dynamics formalism by compactifying the radial direction, allowing the construction of renormalized vacuum expectation values of the energy-momentum tensor. Closed-form expressions for the Casimir energy and pressure are obtained in the massless limit as functions of the radial position of a spherical capacitor and the plate separation. Both observables depend explicitly on the bumblebee parameters and on the location of the apparatus relative to the horizon $R_0=2M$. In the weak-field regime, $r \gg R_0$, the standard flat-space behavior $E \propto -1/d^4$ is recovered. As $r \to R_0$, the Casimir energy vanishes while the radial pressure diverges. Inside the black hole, the interaction may alternate between attractive and repulsive regimes depending on the plate separation and on the Lorentz-violating couplings. A domain-dependent hierarchy among the three configurations emerges, with \textit{metric-affine} effects amplifying the interior vacuum energy, while configurations with simultaneous temporal and radial deformations dominate in the exterior region. Although all geometries share the same asymptotic Schwarzschild structure, their quantitative deviations become increasingly pronounced as the Lorentz-violating parameters grow.

We present a deterministic framework for preparing an arbitrary three-qubit pure state. To leverage entanglement structure in the state-preparation task, we classify three-qubit pure states into five types with respect to a $1|2$ bipartition. Given a target state specified by its amplitudes, we provide concrete criteria and concurrence-based tests that determine its type. For each type, we derive an explicit circuit template composed of elementary single-qubit rotations and CNOT gates, with gate parameters determined systematically from the Schmidt decomposition. The full construction is described step by step from the target amplitudes, with no procedural ambiguity. As an application, we further group frequently encountered three-qubit pure states in quantum information into four classes and provide an explicit circuit for each class. Compared with prior approaches, our circuits are designed for practical use: they admit a direct algorithmic instantiation, use only CNOT gates between adjacent qubits, and for certain classes achieve smaller gate counts and circuit depth.

Quantum error correction (QEC) promises to exponentially suppress qubit noise, but typically assumes spatially-uniform and temporally-constant noise rates. However, real quantum hardware exhibits variation in noise levels over time, which will be amplified by QEC if not addressed. To mitigate this drift in error rates, we leverage transient information readily available in surface code quantum error correction to predict logical error rates (LER) in real time. We infer a prediction model by sampling physical error rates from real hardware, and mapping detector fire rate (DFR), or parity of stabilizer measurements across QEC rounds, to LER. This allows for on-the-fly LER predictions without the typical characterization overhead required to determine LER. This method can easily be extended to other stabilizer codes. Importantly, we observe that this prediction should be accurate yet conservative (i.e. give an upper estimate) to enable appropriately fast responses to real-time physical error changes. That is, responses should be executed marginally ahead of time to allow for their execution to complete, and minimize time spent (ideally none) above intolerable error rates. More importantly, we pair this predictor with a scheme which remaps drifted logical qubits to fresh tiles in a patch-based architecture while their original tiles are recalibrated. Our results demonstrate DFR-based prediction to be an effective LER predictor, and remapping as a spatially efficient and timely mitigation response for small code distances, both of which are significant steps in furthering practical QEC.

We study single-field inflation in which the inflaton is identified with the lightest scalar (dilaton) excitation of a confining gauge theory. The inflaton potential is not postulated: it follows from the pure effective Gluodynamics Lagrangian tightly constrained by the trace anomaly and the associated infinite tower of Ward identities, yielding a Coleman--Weinberg form with a logarithmic term fixed by nonperturbative condensates. After coupling to gravity via a non-minimal interaction $ξ\,\varphi^2 R$, the Einstein-frame potential develops a plateau consistent with current CMB observables. In the large-$ξ$ limit the model approaches the standard plateau attractor, while the Migdal--Shifman(MS) logarithmic structure induces a controlled, testable deformation governed by $A/λ$ across the CMB window. We quantify the resulting shifts in $(n_s,r)$ and the running analytically and confirm them with numerical scans over $(ξ,λ,A,μ)$, making the departure from the attractor both microphysically motivated and observationally predictive.

We prove energy estimates for solutions to a tensorial system of coupled non-linear wave equations, in a way that is suitable to deal with the structure of the non-linearity that arises from the Einstein-Yang-Mills system in the Lorenz gauge as well as with other new different non-linearities. We establish suitable bounds on the $L^2$-norm of each component in a frame decomposition of the tensorial solutions, in way that does not involve all the other components of the tensor, which would allow us to decouple the higher order energy estimates for certain components from the other components. We achieve this partly by exploiting the tensorial structure of the coupled non-linear wave equations, where the background metric that is à priori unknown, is a perturbation of the Minkowski space-time in a certain fixed system of coordinates, and by exploiting the structure of the commutator term for the Lie derivatives of the solutions. These decoupled energy estimates for each component of the tensor in a frame, are new and motivated by a problem that we address in a subsequent paper to prove the exterior non-linear stability of the $(1+3)$-Minkowski space-time governed by a general class of perturbations, that includes the non-linearities that arise from the Einstein-Yang-Mills system in the Lorenz gauge as well as other new non-linearities, which have a different non-linear structure than the one treated by Lindblad-Rodnianski, for which their seminal $L^\infty$-estimate does not work to the best of our knowledge. The decoupled energy bounds on each component in a frame derived here allow us to replace the celebrated $L^\infty$-estimate of Lindblad-Rodnianski in a novel way that permits us to treat these new non-linear structures.

We construct stationary coherent states concentrated on Lissajous figures of the isotropic and anisotropic harmonic oscillators, the latter having coprime frequencies, by projecting products of ordinary coherent states (one coherent state for each degree of freedom) onto sets of degenerate states. By performing these projections, we are deriving our states from sets of coherent states that are known to follow the classical motion of a two-dimensional harmonic oscillator for arbitrary frequencies. We clarify the nature of any singularities present in the phase of the wavefunction for each of the states we derive, and we establish a rigorous connection between the laminar flow of probability current and the emergence of quantum interference. Through this analysis, we are able to provide a clear and quantifiable definition for a vortex state of the two-dimensional harmonic oscillator (2DHO). In an appendix, we show that our stationary states are true coherent states as they can be used to resolve the relevant identity operators (the above mentioned projection operators) on their respective degenerate subspaces. In the special case of the isotropic oscillator, the states obtained are the SU(2) coherent states, and we derive from our formalism the familiar resolution of unity for these states.

Collective radiance in quantum degenerate systems is shaped by the interplay of spatial confinement and exchange statistics. We investigate this interplay using a purely dissipative field theoretic quartic Lindblad master equation, which captures the nonlinear dynamics of the combined motional and electronic manifolds. Our framework maps the crossover between the permutational symmetry of the trap and the exchange symmetry of the particles, quantifying how bosonic enhancement and Pauli blocking dictate superradiant and subradiant scaling. We identify two distinct routes to distinguishable dynamics: thermal dilution of the initial state at high temperatures and the dynamical breakdown of collective order via recoil induced transport in soft traps. This analysis provides a benchmark for collective emission in quantum-degenerate atomic systems with coupled motional and internal dynamics, such as optical lattice clocks and spinor gases, when dissipation is engineered to control recoil and motional heating.

We present COMPOSER, a compile-once modular parametric oracle for similarity-encoded effective reduction of electronic-structure operators (e.g., Schrieffer-Wolff-type constructions). Low-rank factorizations compress Hamiltonians and anti-Hermitian generators into rank-one bilinear and projected-quadratic ladders with near-linear scaling at fixed thresholds; each ladder admits deterministic, number-conserving preparation and a block encoding using constant number of signal ancillas. A fixed PREP-SELECT-PREP template multiplexes these ladders, and one QSP polynomial performs the spectral transformation with degree set by operator norms. For a fixed orbital pool and qubit register, the two-qubit fabric is compiled once; geometry, active-space (mask) updates, and truncations are absorbed by re-dialed single-qubit rotations. We introduce a mask-aware similarity-sandwich effective-Hamiltonian construction and benchmark stability under low-rank and second-order-perturation-guided screening. COMPOSER is an execution architecture: algorithmic errors (block-encoding and QSP approximation) are tunable for any supplied parameters, while physical accuracy depends on how those parameters are obtained if not refined.

Laser-ranged satellites have demonstrated exceptional effectiveness in high-precision verification of General Relativity but also in the accurate inversion of geophysical parameters such as Earth tidal parameters. Due to the extremely weak frame-dragging signal, after utilizing the orbital symmetry of LARES 2 and LAGEOS satellites to cancel most of the nodal precession caused by Earth's oblateness, precise modeling of incompletely symmetric Earth tidal perturbation patterns becomes the core prerequisite for effectively extracting this signal. This study, based on Kaula's perturbation theory and Lagrange equations, investigates the perturbations of LARES 2 and LAGEOS satellite orbital nodes and inclinations caused by Earth tides. From the tidal perturbations computed from 402 earth tide constituents, 83 tidal perturbations with significant amplitudes were selected by threshold based on the RMS of overlap orbit differences of the two satellites, the asymmetric characteristics of tidal perturbations between the two satellites were quantitatively analyzed. The traditional amplitude threshold method for individual tidal perturbations is limited, as the coherent superposition of minor tidal constituents and frequency-orbit resonance lead to the total effect exceeding the threshold. These results provide important support for high-precision tests of General Relativity and the refinement of satellite orbital dynamics modeling.

A higher-order relativistic k.p model is developed to describe plasmon excitations in two-dimensional (2D) electronic systems with spin-orbit coupling (SOC) and magnetic-exchange interactions. Derived entirely from ab initio band structure, the model allows for a non-Rashba spin-momentum locking and enables a direct coupling of the exchange field to the real spin of electrons. Using the BiTeI trilayer (hexagonal C3v symmetry) and the Si-terminated surface state of TbRh2Si2 (cubic C4v symmetry) as prototypes, we show that the exchange field induces strong, symmetry-dependent modifications of the band structure and plasmon dispersion. In BiTeI, it breaks the sixfold symmetry and leads to anisotropic, nonreciprocal plasmon modes, while in TbRh2Si2 it suppresses the characteristic triple spin winding and alters the plasmon damping. The results reveal that the interplay between SOC and exchange magnetism enables magnetic control of collective charge excitations in 2D spin-orbit systems beyond the Rashba paradigm.

We establish the black-hole Meissner effect for extremal Kerr--Bertotti--Robinson (Kerr--BR) black holes, which are exact solutions of the Einstein--Maxwell equations describing a rotating black hole immersed in a uniform Bertotti--Robinson electromagnetic universe. Using the near-horizon framework of Bičák and Hejda, we prove that for a purely magnetic external BR field the horizon-threading magnetic flux vanishes in the static limit of the near-horizon geometry, i.e.\ as the twist parameter $k\to 0$ when $Ba\to 1^-$, thereby establishing the Meissner effect analytically. The proof relies on two exact identities that hold at extremality for all values of the external field: $Ω_x|_{r_e}=0$ and $Ω_r|_{r_e}=B^2 a$, both consequences of the double-root structure of the horizon function $Δ$. Together they force the azimuthal gauge potential $A_φ|_{r_e}$ to become independent of the polar angle in the static limit, reducing to a pure-gauge constant on the horizon $S^2$ and expelling all magnetic flux. The Kerr--BR result is contrasted with the Kerr--Melvin family, where the static limit occurs at a finite interior field strength, and with the Melvin--Kerr--Newman--Taub--NUT spacetime, where the NUT parameter is known to prevent expulsion. An independent geometric argument based on the logarithmic divergence of the proper throat length corroborates the result, and its implications for Blandford--Znajek jet suppression near extremality are discussed.

We extend the finite-distance Jacobi-metric Gauss-Bonnet framework of Li \textit{et al}. [10.1103/PhysRevD.101.124058] to massive test particles carrying intrinsic spin. At pole-dipole order, the Mathisson-Papapetrou-Dixon dynamics generically drives the spatial ray away from Jacobi geodesics, so the standard Gauss-Bonnet construction must be reformulated to accommodate a non-geodesic particle boundary. Working in the aligned-spin planar sector with the Tulczyjew-Dixon spin supplementary condition and retaining terms linear in the spin, we derive a spin-generalized deflection identity in which the spin dependence enters through a single additional boundary functional: the geodesic-curvature integral of the physical ray in the Jacobi manifold. We show that Li's circular-orbit boundary choice remains fully compatible with this generalization and continues to collapse the Gaussian-curvature surface term to an effective one-dimensional integral. We then provide an implementation-ready weak-field recipe that relates the required geodesic curvature directly to the MPD spin-curvature force, enabling systematic perturbative evaluation without introducing model-dependent definitions of asymptotic angles. As applications, we validate the Schwarzschild limit, including the expected linear-in-spin weak-field scaling, and compute leading spin corrections for Reissner-Nordström and Kottler (Schwarzschild-de Sitter) geometries with finite source and receiver distances. In Kottler, we show that the constant-curvature part of the cosmological constant does not generate a linear-in-spin MPD force under the Tulczyjew-Dixon condition; nevertheless, the finite-distance spin correction acquires an explicit $Λ$-dependence through the Jacobi-metric prefactor entering the Gauss-Bonnet boundary functional, in addition to the Weyl-driven (mass-sourced) contribution.

Hybrid quantum-classical neural networks (HQNNs) integrate quantum circuits with classical layers, each operating under fundamentally different computational paradigms, which makes hardware resource estimation challenging. The training of quantum circuits on real devices requires thousands of circuit executions, which is impractical on current NISQ devices. Therefore, most HQNNs are evaluated on classical simulators, with hardware cost approximated using floating-point operations (FLOPs). However, FLOPs and existing quantum resource estimation methods (e.g., gate counts) overlook key quantum hardware-specific factors such as gate durations, limited qubit connectivity, and noise, all of which ultimately determine the true cost and scalability of quantum circuits. In this paper, we propose an analytical quantum cost model that estimates quantum hardware resources using real backend calibration data, incorporating gate durations, routing overheads, and noise-induced sampling inefficiencies. To complement this, we develop a classical cost model that converts FLOPs into device-specific throughput, enabling a unified time-based representation of hardware resource cost for both subsystems of HQNNs. Building on these analytical models, we present Hyb-HANAS, a hardware-aware hybrid neural architecture search framework, which jointly optimizes accuracy, hardware cost, and parameter count using NSGA-II. Hyb-HANAS identifies Pareto-optimal trade-offs and cross-domain co-adaptation between classical and quantum components of HQNNs. Beyond NAS, the proposed analytical quantum cost model is broadly applicable to quantum hardware benchmarking, compiler evaluation, and training-time estimation of quantum circuits on NISQ devices.

We investigate Extended Geometric Trinity of Gravity at both classical and quantum cosmological levels using the minisuperspace approach. Adopting Noether symmetries to select viable models, we examine metric-affine theories of gravity, in particular the extensions of General Relativity, Teleparallel Equivalent General Relativity and Symmetric Teleparallel Equivalent General Relativity, and show that the equivalence among these different formulations can be restored by including in the Lagrangian the divergence terms that relate their respective geometric invariants to the Ricci scalar. Exact cosmological solutions are derived and compared in the different models.

Quantum computing shows substantial potential in accelerating simulations and alleviating memory bottlenecks in computational fluid dynamics (CFD), owing to its inherent properties of superposition and entanglement. The lattice Boltzmann method (LBM), being largely algebraic in nature, has inspired the development of various quantum LBMs. However, most existing approaches fix the relaxation time at $τ$ = 1, thereby confining a given mesh resolution to simulations at a single Reynolds number. Although our earlier quantum lattice kinetic scheme (LKS) lifted this restriction, it suffers from instability at high Reynolds numbers. To address this challenge, we propose a quantum fractional-step LBM (FS-LBM). In this framework, the predictor step is implemented on a quantum circuit using the standard LBM formulation, while the corrector step is performed classically. The relaxation time is retained at $τ$ = 1 to ensure seamless compatibility with existing quantum LBMs. Benchmark simulations of representative two- and three-dimensional incompressible isothermal and thermal flows demonstrate that the quantum FS-LBM achieves accuracy and convergence orders consistent with its classical counterpart, while significantly outperforming the quantum LKS in both precision and stability. Notably, this work presents the first quantum LBM simulation of three-dimensional incompressible thermal flows.

One-time programs (OTPs) aim to let a user evaluate a program on a single input while revealing nothing else. Classical OTPs require hardware assumptions, and even with quantum information, OTPs for deterministic functionalities remain impossible due to gentle-measurement attacks (Broadbent, Gutoski and Stebila, 2013). While recent works achieve positive results for certain randomized functionalities, the fundamental limits and the strongest achievable security notions remain poorly understood. In this paper, we ask for a "best-possible" OTP that achieves the strongest one-time security achievable by any OTP construction. We first show that a generic best-possible one-time compiler cannot exist, even for classical randomized functionalities (assuming lossy encryption schemes exist). Given this impossibility, we introduce a natural subclass of one-time compilers called "testable one-time program" compilers, which output quantum states augmented with reflection oracles for these program states. We show that best-possible testable OTP compilers are achievable by (1) formulating a generalized Single-Effective-Query (SEQ) simulation security notion for quantum channels and show that SEQ security implies best-possible testable one-time security, and (2) constructing SEQ-secure OTPs for all quantum functionalities in the classical oracle model. This yields the first OTP for arbitrary quantum channels beyond classical randomized functionalities. Finally, we propose stateful quantum indistinguishability obfuscation (stateful quantum iO) -- quantum state obfuscation for stateful quantum programs. We show that (1) stateful quantum iO implies best-possible testable OTPs and (2) stateful quantum iO is also achievable in the classical oracle model. These results identify stateful quantum iO as a promising approach towards best-possible testable OTPs.

We present global, three-dimensional general relativistic magnetohydrodynamic simulations of accreting black holes that incorporate pair physics. Pairs are modeled as a passive scalar that maintains a constant temperature. For high accretion rate models, we observe a maximum pair fraction of $\sim \mathcal{O}(0.01)$, consistent with those inferred from some X-ray binaries, and identify a `pair void' extending to a few gravitational radii from the black hole. Pair fractions peak in the midplane just outside the plunging region and within a thin strip at the base of the corona. For moderate to high accretion rate models, pairs are near equilibrium close to the disk midplane, where the scattering optical depth is high and pair equilibrium timescales are short, and could be comparable to the Coulomb collision timescale. This suggests the possibility of a pair-regulated coronal temperature. In contrast, the upper corona and jets, where the scattering optical depth is relatively low and pair equilibrium timescales are long, are populated with pairs that may exceed their equilibrium value by orders of magnitude. These pairs are transported by advection from the disk, which dominates over local pair processes. This result highlights advection as a significant source of pair injection, which may be relevant for certain X-ray binaries exhibiting $γ$-ray signatures. The pair density along the magnetically dominated poles exceeds the Goldreich-Julian density in some models.

Two-dimensional noncommutative quantum mechanics (NCQM) is often formulated through linear transformations of represented position and momentum operators and through Darboux-type canonicalizations. We clarify the representation-theoretic meaning of such constructions at the level of kinematical symmetry groups and irreducible unitary representations. The standard NCQM commutators are naturally encoded by a step-two nilpotent Lie group $G_{\hbox{\tiny{NC}}}$ with three-dimensional center, whose irreducible sectors are labeled by central characters (equivalently, coadjoint-orbit labels), parametrized on the regular stratum by $(\hbar,\vartheta,B_{\mathrm{in}})$. In this language, ordinary two-dimensional quantum mechanics (QM) is the quotient (equivalently, inflation) sector $(\hbar,0,0)\subset \widehat{G_{\hbox{\tiny{NC}}}}$, the unitary dual of $G_{\hbox{\tiny{NC}}}$; i.e., it consists of those representations that factor through the central quotient $G_{\hbox{\tiny{NC}}}\twoheadrightarrow G_{\hbox{\tiny{WH}}}$, where $G_{\hbox{\tiny{WH}}}$ denotes the Weyl--Heisenberg group. We show that generalized Bopp-shift and Seiberg--Witten-type linear recombinations of represented operators, and the existence of an auxiliary quadruple satisfying the canonical commutation relations obtained by Darboux canonicalization, do not imply unitary equivalence between a fixed generic NCQM sector $(\hbar_{0},\vartheta_{0},B_{0})$ and the ordinary-quantum-mechanics sector $(\hbar_{0},0,0)$ of $\widehat{G_{\hbox{\tiny{NC}}}}$.

In this work, we explore physical systems which support not only multipartite interparticle entanglement, but also intraparticle entanglement between different degrees of freedom of the constituent particles and entanglement between different degrees of freedom of different particles, i.e., algebraic entanglement. We derive a simple method for calculating the algebraic entanglement entropy between two of the particles' degrees of freedom from collective states of the whole ensemble. Our procedure makes use of underlying symmetries in these systems, in particular permutation symmetry of the particle indices, and shows a connection between the algebraic entanglement entropy in these systems and the irreducible representations of Lie groups which describe the particles' degrees of freedom. Namely, we use the direct sum over irreducible representations to diagonalize the reduced density matrices in a block-by-block manner, then utilize the multiplicity of these irreducible representations to reproduce the results from an exponentially-scaled Hilbert space in only polynomial complexity. We use this to explore a variety of systems where the constituent particles support two degrees of freedom each with two levels, such as atoms with two electronic states and two momentum states. Notably, these systems may be exactly simulated in a polynomial-scaled Hilbert space, yet they support an algebraic entanglement entropy that grows linearly with the particle number which typically requires an exponentially-scaled Hilbert space.

We present a cross-cavity system in which steady-state superradiance is achieved using solely collective dissipative dynamics. Two cavities symmetrically couple an ensemble of four-level atoms by driving transitions between two electronic states and two motional states along perpendicular cavity axes. Both cavities operate in the bad-cavity regime: one cavity mediates collective atomic decay, while the other cavity, together with a coherent drive, mediates collective pumping via an off-resonant Raman transition. With this, we find steady-state superradiant states that possess nonclassical properties, such as super-Poissonian photon statistics. The system thus requires a beyond mean-field description, and so we develop an exact master equation simulation technique utilizing strong symmetries of the system's jump operators. Because superradiant decay is accompanied by a momentum impulse along the corresponding cavity axis, the system exhibits substantial hybrid entanglement between the atoms' spin and motional degrees of freedom at steady state. We also demonstrate that heralded measurements of the two cavity outputs prepare a state with significant particle-particle entanglement with prospects for quantum-enhanced acceleration sensing.

In Jackiw-Teitelboim gravity, the naive Schwarzian quantum mechanics leads to a continuous bulk spectrum, in apparent contradiction with the finite entropy of the black hole, which requires a discrete spectrum with level spacing of order $e^{-S_0}$. It was recently shown that restoring spectral discreteness with random statistics requires the introduction of a left confining potential that becomes relevant when the renormalized wormhole length reaches order $e^{S_0}$. In this work, we show how the known perturbative results of JT gravity are recovered within this modified framework. More importantly, we demonstrate that this modification has a direct dynamical consequence: it resolves the black-hole singularity. The confining potential generates a repulsive force at exponentially large wormhole length, preventing the indefinite growth that would otherwise lead to a singularity. We explain in detail how this turnaround arises and explore its implications for late-time bulk gravitational dynamics, the disappearance of horizons, and possible observational consequences.

We report on the fabrication and characterization of superconducting coplanar waveguide (CPW) resonators based on titanium nitride (TiN) thin films deposited on 200\,mm diameter high-resistivity Si(100) substrates. We systematically investigate how deposition conditions, dry-etch power and in-situ resist strip temperature affect morphology, superconducting properties and dielectric losses. By tuning reactive sputtering conditions, three distinct preferred crystal orientations - (111), (200), and mixed are achieved. Our results demonstrate that all films exhibiting similar minimal two-level system (TLS) losses, with TiN111 exhibit the lowest median TLS losses $\tildeδ_\mathrm{TLS}$, and greater robustness against reoxidation. The applied structuring process, in contrast, had a far greater influence on the TLS loss than the crystal orientation of the TiN film and, consequently, the intrinsic material properties of the superconducting layer. The lowest TLS losses for all TiN depositons were achieved with a low power etch and low temperature resist strip. An additional buffered oxide etch (BOE) treatment could remove high-loss interfacial oxides at the metal-air (MA) and substrate-air (SA) interface and recover the etch-induced TLS losses. Consequently, TiN resonators exhibiting $\tildeδ_\mathrm{TLS}$ values as low as $9.67 \times 10^{-7}$ were realized. The corresponding median low-power loss, $\tildeδ_\mathrm{LP}$, amounts to $11.04 \times 10^{-7}$, which translates to an internal quality factor approaching one million. These findings highlight the critical role of process induced oxide formation at the MA and SA interfaces in limiting the performance of TiN resonators and provide a scalable, low-loss process compatible with industry-grade 200 mm CMOS qubit fabrication workflows.

Cosmological time dilation (CTD) serves as a fundamental probe of cosmic expansion, historically verified through the characteristic (1+z) broadening of Type Ia supernova (SNe Ia) light curves. However, significant tensions arise when extending this test to other astrophysical regimes. While discrete, event-based transients such as Gamma-Ray Bursts (GRBs) exhibit large scatter in interred time-dilation signatures, analyses of stochastic variability in persistent sources, specifically Quasars (QSOs), frequently yield null results. I demonstrate that these discrepancies stem from a previously overlooked distinction between discrete geometric clocks and continuous thermal emission, presenting a resolution within the framework of Generalized Cosmological Time (GCT). The central premise relies on strictly distinguishing global coordinate time, characterized by a generalized lapse function, from the local proper time measured within gravitationally bound systems. We propose that the progenitors of transients, specifically SNe Ia and GRB central engines, are effectively shielded from background time evolution due to strong gravitational binding and environmental decoupling. Consequently, they act as standard clocks tracing pure geometric path dilation, obeying τ_{\rm obs} \propto (1+z)^{1+b/4}. Conversely, the lack of dilation in QSOs is derived as a consequence of observing persistent thermal accretion disks at fixed wavelengths, introducing an intrinsic selection effect (τ_{\rm intr} \propto (1+z)^{-2}) that masks the cosmological signal. This framework reconciles the diverse behaviors of transient and persistent sources without modifying local physical laws.

We develop a comprehensive axiomatic framework for quantum-inspired distance metrics on projective Hilbert spaces, providing a unified foundation that organizes and generalizes existing measures in quantum information theory. Starting from five fundamental axioms, projective invariance, unitary covariance, superposition sensitivity, entanglement awareness, and measurement contextuality, we show that any admissible distance depends solely on state overlap and establish the uniqueness of the Fubini-Study metric as the canonical geodesic distance. Our framework further yields a hierarchy of comparison results relating the Fubini-Study metric, Bures distance, Euclidean distance, measurement-based pseudometrics, and entanglement-sensitive distances. Key contributions include an entanglement-geometry complementarity principle, high-dimensional concentration bounds, and operational interpretations connecting distances to state discrimination and quantum metrology. This work places the geometry of quantum state spaces on a rigorous axiomatic footing, bridging abstract metric theory, information geometry, and operational quantum principles.

Randomness is intrinsic to quantum mechanics; the outcome of a measurement on a quantum state is a random variable. This feature has been applied to randomness certification, where one party must decide whether the data they receive is truly random. However, existing demonstrations are not black-box, to avoid falsely certifying deterministic data, assumptions must be made on how the data was generated. Here we demonstrate genuine randomness certification in the black-box setting -- one in which no deterministic adversary, even with unlimited computational power, will succeed in getting their data certified. We use it to provably generate random numbers using only measurements on single particle states and without a random seed.

Ultracold molecules are becoming an increasingly important technology for quantum simulation, computation, and sensing, but their state preparation in large, low-entropy arrays remains a key challenge. We propose to deterministically load single molecules into optical tweezer arrays or lattices from either thermal or degenerate gases, with a high probability of occupying the tweezer's motional ground state. Strong repulsion between microwave-shielded molecules prevents multiparticle occupancy. Our proposal represents a robust scheme for deterministic single molecule preparation directly in the motional ground state with expected fidelities exceeding 99 percent for small trap volumes and highly polar species. This method can be scaled to thousands of traps limited by the reservoir molecule number, opening the door to large, low-entropy polar molecule arrays for quantum computation, quantum simulation, and precision measurement.

Microfabricated surface-electrode traps are a scalable platform for trapped-ion quantum processors. Recent advances in fabrication techniques have enabled the design of increasingly complex multi-layer structures. Yet the control electrodes remain mostly unchanged and of rectangular shape. We systematically analyze asymmetric inner control electrode shapes for simultaneous axial and radial control in multi-layer surface traps, characterize and compare a selection of different shapes, and verify their capabilities in realistic use-case scenarios for ion transport and micromotion compensation. Eliminating the need for the commonly used additional outer control electrodes, asymmetric inner control electrodes increase the compactness and space efficiency of surface-electrode traps while concurrently reducing the number of control signals. The improved control voltage efficiency of using solely inner electrodes enables the device's entire direct-current (DC) supply to be provided by integrated Cryo-CMOS circuits, further enhancing the scalability of the processor.

We establish a universal relation between the coefficient $C_T$ of the energy momentum tensor two point function and the coefficient $c$ multiplying the term quadratic in the Weyl tensor in the Weyl anomaly of a generic even dimensional conformal field theory. Our first derivation combines long known holographic results for $C_T$ and for the Weyl anomaly in Einstein bulk gravity with a recently obtained Chern Gauss Bonnet formula for compact Einstein manifolds. This theorem isolates the Weyl squared contribution in the relation between the Euler density and the $Q$ curvature, allowing us to identify the relevant quadratic term unambiguously. We then provide a genuine CFT derivation based on the renormalization group running of the TT correlator with respect to the arbitrary but necessary mass scale $μ$. Several known examples are revisited to illustrate and validate the general result.

Constrained combinatorial optimization problems are challenging for quantum computing, particularly at utility-relevant scales and on near-term hardware. At the same time, these problems are of practical significance in industry; for example, the Unit Commitment (UC) problem in energy systems involves complex operational constraints. To address this challenge, we apply copula-QAOA (cop-QAOA), a hardware-efficient approach for constrained optimization to a single-period UC that can be reduced to a one-dimensional knapsack. Cop-QAOA biases the quantum state toward feasible solutions using constant-depth mixers and appropriately biased initial states. We implement our benchmark on problem instances that are confirmed to be hard for classical solvers such as Gurobi. Our results show that cop-QAOA often finds solutions better than a lazy greedy baseline and very close to, and in some instances surpasses, those obtained by Gurobi, with only a few QAOA rounds. This work presents the largest successful demonstration of the knapsack problem on IBM Quantum hardware using up to 150 qubits, and more generally, the largest demonstration of constrained combinatorial optimization where constraints are enforced via shallow mixers.

We investigate the scattering features of a non-Hermitian rectangular potential within the framework of space-fractional quantum mechanics. Using the Riesz fractional derivative, we analytically derive locus equations for spectral singularities (SSs) and their time-reversed counterparts, coherent perfect absorption (CPA), in a dimensionless complex-potential parameter space. This geometric locus formulation provides a transparent representation of the SS and CPA conditions and enables direct visualization of how fractional quantum dynamics modifies non-Hermitian scattering. We show that reducing the Lévy index $α$, which enhances nonlocal transport associated with Lévy-flight dynamics, systematically lowers the gain-loss strength required for the emergence of SSs and CPAs, while increasing the mode index further suppresses this threshold. In addition, for fixed potential parameters, we demonstrate that decreasing $α$ induces a blue shift of the SS energy, in direct agreement with earlier studies. From this perspective, the Lévy index $α$ emerges as a tunable control knob for SS-CPA settings in fractional non-Hermitian quantum systems. Beyond its quantum-mechanical setting, this study may find applications in fractional waveguides and metamaterials governed by fractional wave equations. This work also bridges the gap between non-Hermitian quantum mechanics and space-fractional quantum mechanics.

We quantify the impact of finite catalog size, or "catalog variance," on current gravitational-wave population analyses. The distribution of merging binary black holes is commonly reconstructed via hierarchical Bayesian inference, with uncertainties reported as credible intervals. Such intervals are conditioned on the specific realization of the observed events and are therefore themselves subject to variability arising from the finite size of the catalog. We estimate this "uncertainty on the uncertainty" using statistical bootstrapping applied to data segments containing both detected events and sensitivity injections. Applying this framework to GWTC-4, we find that the inferred population distributions exhibit substantially broader uncertainties than those obtained in a standard single-catalog analysis. In particular, the $\sim 35\,M_\odot$ peak in the primary-mass distribution is largely absorbed by statistical fluctuations once catalog variance is taken into account. Unlike other studies that rely on simulating catalogs by assuming an underlying population, this work provides the first data-driven assessment of the uncertainty intrinsic to the observed gravitational-wave catalog. Accounting for catalog variance is important for drawing robust astrophysical conclusions from gravitational-wave data, avoiding inferences driven by a particular finite realization rather than genuine population features.

The quantum switch is a basic network primitive that allows one to connect multiple nodes in a quantum network via a central node. We show that the same functionality can be achieved with a different geometry that does not rely on a powerful and large central unit, but instead utilizes evenly distributed resources. This approach is resilient against node failures. We provide a nested construction with logarithmically many qubits per node and a total of $O(n\log n)$ Bell pairs, in contrast to other distributed approaches based on pre-shared entanglement that scale as $O(n^2)$. The construction achieves fully flexible pairwise connectivity, where the shared resource state can be locally transformed into $n/2$ arbitrarily distributed Bell states. We also present a graph state variant with just one qubit per node, which allows one to generate $O(n/\log^2 n)$ Bell pairs.

Floquet codes define fault-tolerant protocols through periodic measurement sequences that drive a dynamically evolving stabilizer group. They provide a natural framework for hardware supporting two-qubit parity measurements but no unitary entangling gates. However, few known constructions achieve both high encoding rates and high thresholds. We close this gap by introducing Stairway codes, a family of high-rate Floquet protocols obtained by Floquetifying Abelian two-block group algebra codes, a class that includes the bivariate bicycle codes. By representing the static code as a foliated ZX-calculus network within a $(w{-}1)$-dimensional space-time lattice and rotating the time axis, we decompose its weight-$w$ stabilizers into a periodic sequence of pairwise measurements. This reduces the design of new codes within this family to the selection of favorable periodic boundary conditions. We identify instances with competitive parameters, analyze their distance under circuit-level noise, and demonstrate logical error rates surpassing those of other Floquet codes at comparable encoding rates. Remarkably, our construction requires fewer than 300 physical qubits to match the distance and encoding rate of semi-hyperbolic Floquet codes that use over 1300 qubits.

We investigate a quantum-inspired approach to supervised multi-class classification based on the \emph{Pretty Good Measurement} (PGM), viewed as an operator-valued decision rule derived from quantum state discrimination. The method associates each class with an encoded mixed state and performs classification through a single POVM construction, thus providing a genuinely multi-class strategy without reduction to pairwise or one-vs-rest schemes. In this perspective, classification is reformulated as the discrimination of a finite ensemble of class-dependent density operators, with performance governed by the geometry induced by the encoding map and by the overlap structure among classes. To assess the practical scope of this framework, we apply the PGM-based classifier to two biomedical radiomics case studies: histopathological subtyping of non-small-cell lung carcinoma (NSCLC) and prostate cancer (PCa) risk stratification. The evaluation is conducted under protocols aligned with previously reported radiomics studies, enabling direct comparison with established classical baselines. The results show that the PGM-based classifier is consistently competitive and, in several settings, improves upon standard methods. In particular, the method performs especially well in the NSCLC binary and three-class tasks, while remaining competitive in the four-class case, where increased class overlap yields a more demanding discrimination geometry. In the PCa study, the PGM classifier remains close to the strongest ensemble baseline and exhibits clinically relevant sensitivity--specificity trade-offs across feature-selection scenarios.

The development and operation of large-scale particle physics facilities rely not only on accelerators and detectors, but also on sustained, high-precision simulation infrastructure. Originating in Lund in the late 1970s and continuously developed in Sweden for nearly five decades, PYTHIA has evolved into one of the most widely used Monte Carlo event generators in high-energy physics. Today it functions as a facility-scale software infrastructure underpinning the physics programmes of major international experiments, including those at the Large Hadron Collider, and plays a central role in validation, tuning, and uncertainty evaluation. In this article, we present PYTHIA as a Swedish contribution to big science facilities. We outline its historical development, analyze its contemporary user base through citation and text-based studies, and map its integration across experimental frameworks, generator ecosystems, validation infrastructures, and emerging machine-learning workflows. These analyses show that PYTHIA We discuss the operational model and sustainability challenges associated with maintaining long-lived research software at facility scale. As particle physics moves toward the High-Luminosity LHC era and future facilities such as the EIC and FCC, continued investment in robust, interoperable simulation infrastructure remains essential. We discuss the operational model and sustainability challenges associated with maintaining long-lived research software at facility scale. As particle physics moves toward the High-Luminosity LHC era and future facilities such as the EIC and FCC, continued investment in robust, interoperable simulation infrastructure remains essential.

We study the Higgs branch and associated vertex operator algebra (VOA) of 4d $\mathcal{N}=2$ superconformal field theories (SCFTs) from the geometric engineering of IIB superstring on canonical threefold singularities. For terminal singularities, we explain how to derive the 4d Higgs branch from their small resolution. We also investigate singularities with compact 4-cycles in their crepant resolution, and discuss different ways to compute their Higgs branch. Using a symplectic duality argument, we propose the first examples of 4d $\mathcal{N}=2$ SCFTs with the E-type Kleinian singularities as their Higgs branches, and conjecture their associated VOA to be affine E-type W-algebra. Many new VOAs with no known W-algebra descriptions are found, with conjectured associated varieties. We investigate the singularities associated with lisse VOAs and propose predictions for the BPS quivers of $D_N^N[k]$ and $E_7^{14}[k]$ from the perspective of deformed singularities. We further analyze the structure of the Schur index using the Coulomb branch IR formula, derive the expressions for the Schur index corresponding to these two classes of singularities, and illustrate, in a general setting, how the Schur index is determined by the BPS quiver.

New decay modes of $B$-meson into a baryon and invisible dark antibaryon $Ψ$ are among the most distinctive signatures of the $B$-mesogenesis scenario. We concentrate on the proton mode and consider two versions of the underlying interaction of $Ψ$ with quarks, the so-called models $(d)$ and $(b)$. To estimate the width of the $B^+\to p Ψ$ decay, we obtain the $B^+\to p$ transition form factors, applying QCD light-cone sum rules (LCSRs) with $B$-meson distribution amplitudes with an accuracy up to twist-5, while interpolating the proton with a current. This method is independent of the previously applied one, which was based on the nucleon distribution amplitudes. We estimate the partial width of the $B^+\to p Ψ$ decay as a function of the dark antibaryon mass. Furthermore, we use the ratio of this width to the inclusive $B\to X_N Ψ$ width, the latter predicted earlier using the heavy quark expansion method. This ratio, which is independent of the effective coupling, when combined with the minimal inclusive branching fraction of $O(10^{-4})$, necessary for the feasibility of $B$-mesogenesis, yields lower limits on the $B^+\to p Ψ$ branching fraction. We confront these limits with the most recent upper bounds obtained from BaBar and Belle/Belle II searches for the decays of $B^+$-meson into a proton and missing energy. The comparison indicates that experimental upper bounds on the branching fraction of $B\to p Ψ$ at the level of $10^{-8}-10^{-7}$ are needed for a decisive probe of this invisible mode of $B$ decays.

The distribution of heat-producing elements (U, Th, K) within the Moon is critical for understanding its thermal evolution and formation history. Based on a refined lunar interior model, we calculate the geoneutrino fluxes at two representative detector locations that bracket the expected signal intensity. The maximum flux is found to be slightly lower than the corresponding predicted fluxes for the KamLAND site on Earth, while the minimum flux is approximately a factor of 8.63 lower than this maximum value. The angular distributions of geoneutrinos arriving at the two locations were further computed. Finally, we evaluate the detection prospects for lunar geoneutrinos using three reaction channels: inverse beta decay reaction, elastic scattering on electrons, and a novel radiochemical approach based on $\barν_e + ^3$He $\to e^+ + ^3$H. For each reaction, we calculate the expected event rates and briefly discuss the potential for measuring the total geoneutrino flux, as well as the relative contributions from U, Th, and K.

We study TeV-scale electromagnetic leptogenesis (EMLG) in an effective field theory (EFT) framework, starting from an ultraviolet (UV)-complete model in which integrating out heavy states generates the gauge-invariant dimension-six dipole operators $O_{NB}$ and $O_{NW}$. After electroweak symmetry breaking these operators induce effective dipole couplings of the right-handed neutrinos $N$ to $γ$, $Z$, and $W^{\pm}$, enabling decays, inverse decays, and scattering processes in the electroweak crossover window. Gauge invariance enforces a Higgs insertion and thus a parametric suppression of both the CP-odd source and the washout in the non-resonant, hierarchical regime, preventing reproduction of the observed baryon asymmetry. We therefore focus on the quasi-degenerate limit, in which the self-energy contribution to the CP asymmetry is resonantly enhanced and coherent flavor dynamics among the nearly degenerate $N$ becomes essential. Using the CTP/Schwinger--Keldysh formalism, we derive density-matrix quantum kinetic equations (QKEs) whose collision term, at leading order in the effective dipole couplings and SM gauge interactions, incorporates $1\leftrightarrow 2$ processes and the leading $ΔL=0$ scatterings in a unified and non-overcounted manner, without the need for real-intermediate-state (RIS) subtraction. Solving the resulting system numerically, we present a coherent EFT pipeline from UV matching and renormalization-group (RG) running to electroweak-scale effective dipole couplings, resonantly regulated CP sources, and the frozen-out baryon asymmetry. For both thermal and zero initial heavy-neutrino abundances, we find in the oscillation-motivated region that the undiluted freeze-out yield $Y_B^{\rm FO}$ can exceed $Y_B^{\rm obs}\simeq 8.7\times10^{-11}$ by several orders of magnitude.

We derive the covariant generalization of the Frank-Tamm formula describing the Cherenkov radiation by a charged particle moving uniformly with a speed faster than the local speed of light within a homogeneous dielectric medium. We use our result to derive the covariant Cherenkov radiation reaction force and obtain a four-force explicitly orthogonal to particle four-velocity consistent with a relativistic friction force. We present the photon emission spectrum that is dependent primarily on the dielectric properties of the medium. We hint at a possible use of this work to interpret an excess of soft photons seen in relativistic hadron collisions.

The precise measurement of the $e^+e^-\toπ^+π^-π^0$ cross section is performed in the center-of-mass energy range $E = 560$--1100 MeV using a data sample of 66 pb$^{-1}$ collected in the experiment with the SND detector at the VEPP-2000 $e^+e^-$ collider. The systematic uncertainty of the cross section measurement is 0.9\% at the maximum of the $ω$ resonance and 1.2\% at the maximum of the $φ$ resonance. The leading-order hadronic contribution to the muon magnetic anomaly calculated using the $e^+e^-\toπ^+π^-π^0$ cross section measured by SND from 0.62 to 1.975 GeV is $(42.96\pm0.06\pm 0.45)\times 10^{-10}$. From the fit to the cross section data with the vector meson dominance model, the parameters of the $ω$, $ρ$, and $φ$ resonances are obtained. The obtained values of ${\cal B}(ω\to e^+e^-) {\cal B}(ω\to 3π)$, mass and width of the $ω$ meson, and ${\cal B}(ρ\to 3π)$ have accuracy better than the current world average values.

Habiro lifted the Witten-Reshetikhin-Turaev invariant of an integer homology 3-sphere (a complex-valued function on the set of complex roots of unity) to an element of the Habiro ring. We lift the colored Jones polynomial of a knot, with Alexander polynomial $Δ(t)$, to the recently introduced Habiro ring of the étale map $\mathbb{Z}[t^{\pm 1}]\to \mathbb{Z}[t^{\pm 1},Δ(t)^{-1}]$ (with Frobenius lifts $t\mapsto t^p$ for all primes $p$). This implies the existence of a loop expansion at roots of unity (confirming a conjecture of Habiro), and a lift of power series invariants of Ohtsuki for 3-manifolds with Betti number 1 to a Habiro ring. Our results have natural extensions to the skein module of a knot complement, and they suggest a natural lift of the colored Jones polynomial colored by representations of a simple Lie (super) algebra.

We present new non-perturbative results about the meson spectrum and the low-energy constants of QCD in the 't Hooft large-$N$ limit, $N\to\infty$ with $N_{\scriptscriptstyle{\rm f}}/N\to 0$. These are obtained from lattice Monte Carlo simulations of the Twisted Eguchi-Kawai (TEK) model up to $N=841$. More precisely, we will discuss: our findings for the meson mass spectrum; the determination of the radial Regge trajectories in the $π$ and $ρ$ channels; the computation of the coefficients of the $1/N$ expansion of the chiral condensate, of the pion decay constant, and of the next-to-leading-order coupling $\bar{\ell}_4$, up to $\mathcal{O}(1/N^3)$ from the combination of TEK and standard finite-$N$ results.

Acceptance plenary talk for the 2025 Kenneth G.~Wilson Award for Excellence in Lattice Field Theory: For significant contributions to the understanding of topology in QCD, QCD-like, and large-$N_c$ gauge theories, including algorithmic developments to reduce topological freezing, studies of Dirac spectral properties, and axion phenomenology.

Effective field theories include contact-range interactions (or counterterms) for two reasons: representing the unknown short-range physics in a model independent manner and ensuring the cutoff independence of observables. Both are intertwined: cutoff independence alone (modulo truncation errors) already generates counterterms encoding physical information not present in the known long-range physics. Yet, there is also residual cutoff dependence, which is smaller than the uncertainties that are achievable within the effective field theory description and thus can be safely neglected in most settings. If one insists on exact cutoff independence though, new counterterms will be required, but they encode no new physical information and are thus what one could call redundant, or auxiliary, counterterms. It happens that auxiliary counterterms are still useful for solving certain inconsistencies that appear during renormalization or for improving the convergence of the effective field theory expansion. Examples of these use cases are discussed, including the interpretation of the improved actions or the relation between perturbative and non-perturbative renormalization.

We study the behavior of the cusp structures focusing on the isospin-breaking effects. The properties of the near-threshold exotic hadrons are encoded in the shapes of the cusp structures. In hadron scattering, it is often the case that the thresholds of isospin partner channels are located within a narrow energy region. To analyze the scattering in such systems, it is therefore essential to study the cusp structures that emerge at two closely separated thresholds with isospin symmetry breaking. In this study, we propose a practical representation of the scattering amplitude and show that the two neighboring cusp structures are related through the isospin symmetry.

We present the \textsc{LHEReader} program that converts an Les Houches Event file into a \Root file allowing one to subsequently analyze the file using \Root. We evaluate the performance of this conversion by simulating $pp\to jj$ and $e^+e^- \to ZH$. The application of this program is illustrated through the simulation of $pp\rightarrow ZH$, with the $Z\to\ell^+\ell^-$ and $H\to b\bar{b}$, analyzing the events at the parton level after hard-scatter event generation as well as after the stage of fast simulation using \MadAnalysis. The analysis is implemented in \Root demonstrating the use of this program. The program is available here: https://github.com/amanmdesai/LHEReader

Next-generation gravitational-wave detectors will make black-hole ringdown an increasingly sensitive probe of small departures from General Relativity in the strong-field regime. This motivates obtaining high-precision predictions of gravitational effective field theory, as spectral shifts can be quite small. Here we perform a focused stress test of the modified-Teukolsky framework by designing two null diagnostics. First, we consider an action with redundant operators that must produce zero first-order vacuum QNM shifts. Second, we exploit a Ricci-flat identity relating two physical cubic Riemann to test such a relation is satisfied by the ringdown spectra obtained. We compute the shifts using two independent numerical approaches: the eigenvalue-perturbation and generalized continued-fraction (Leaver-type) methods. Both null tests are passed across multiple multipoles and overtones, and the control-operator results agree in magnitude with the benchmark values reported in Ref. [1]. These validations support using the framework for obtaining accurate precitions for robust strong-field tests, with straightforward extensions to rotating backgrounds and coupling with matter fields.

Measuring neutrino mass ordering (NMO) poses a fundamental challenge in neutrino physics. To address this, the Jiangmen Underground Neutrino Observatory (JUNO) experiment is scheduled to commence data collection in late 2024, with the ambitious goal of determining the NMO at a 3-sigma confidence level within a span of 6 years. A key factor in achieving this is ensuring a high-quality energy resolution of positrons. However, the presence of residual $^{14}$C isotopes in the liquid scintillator introduces pile-up effects that can impact the positron energy resolution. Mitigating these pile-up effects requires the identification of pile-up events, which presents a significant challenge. The signal from $^{14}$C is considerably smaller compared to the positron signal, making its identification difficult. Additionally, the close event time and vertex between a positron and a $^{14}$C further compound the identification challenge. This contribution focuses on the application of deep learning models for the identification of $^{14}$C pile-up events. It encompasses a range of models, including convolution-based models and advanced transformer models. Through performance evaluation, it shows the deep learning-based methods is promising to identify the pile-up events.

Fine-tuning and naturalness, the sensitivity of low-energy observables to small changes in the fundamental parameters of a theory, are cornerstones of physics beyond the Standard Model. We propose a new measure of fine-tuning based on information theory. To each point in parameter space we associate a probability distribution over observables. Divergence measures encode the sensitivity of observables to model parameters and determine a Riemannian metric on parameter space. By Chentsov's theorem, the physically motivated metric is the Fisher information metric, up to scaling. We propose a rescaled fine-tuning matrix $\mathcal{F}_{ij}$ derived from the Fisher information matrix, whose non-zero eigenvalues serve as our measure of fine-tuning. When the number of observables exceeds the number of parameters, $\mathcal{F}_{ij}$ admits a natural geometric interpretation as the pullback of the Euclidean metric from observable space to the submanifold of admissible predictions, with large eigenvalues corresponding to highly stretched directions and indicative of fine-tuning. Our measure reproduces the familiar Barbieri--Giudice criterion as a special case, while generalising it to multiple correlated parameters. We illustrate its behaviour on dimensional transmutation, the Wilson--Fisher fixed point, a simple model of the hierarchy problem, and the electron Yukawa coupling, finding agreement with physical intuition in each case.

We investigate the one-dimensional \emph{Generalized Klein--Gordon Oscillator} (G-KGO) within Doubly Special Relativity (DSR) kinematics. The G-KGO extends the Klein--Gordon oscillator by replacing the usual linear non-minimal coupling with a general interaction function $f(x)$, leading to a factorized (SUSY-like) Schrödinger operator $H_-=\mathcal{A}^{\#}\mathcal{A}$ whose real spatial spectrum $\{ε_n\}$ can be ensured either by Hermiticity or, for complex $f(x)$, by $η$-pseudo-Hermiticity and/or $\mathcal{PT}$ symmetry with a consistent metric inner product \cite{Bender1998,Bender2007RPP,BBJ2002PRL,Mostafazadeh2002,Mostafazadeh2003,Mostafazadeh2010,ElGanainy2018NatPhys}. DSR is then implemented at the level of the energy reconstruction map $ε_n\mapsto E_{n,\pm}$, and we provide closed-form Magueijo--Smolin (MS) and Amelino--Camelia (AC) branches.

A brief historical overview of various modern approaches to the problem under consideration is given. It includes existing models based on a sum of different terms of the scattering amplitude with different signs and Regge-eikonal models based on the Born terms of the scattering amplitudes. An example of such a model is a new Regge-eikonal model is given, taking into account the generalized structure of nucleons (the HEGS model), which is based on the analyticity of the scattering amplitude. A unified quantitative description of various hadron reactions and a description of differential cross sections and the spin-correlation parameter for interactions were obtained. In the framework of the model, the existence of experimental data of elastic hadron scattering in the energy range of LHC and in a wide energy region $\sqrt{s}=3.6 -13000$ GeV was describe a quantitatively from a unified point of view. The predictions for $σ_{tot}(s)$ at superhigh energies are presented. The possible thin structure of differential cross sections at small angles of elastic nucleon-nucleon scattering at high energies is discussed.

The axion, a well-motivated hypothetical particle arising in extensions of the Standard Model, can be produced copiously within the hot, compact cores of white dwarf stars. The shape of the white dwarf luminosity function (WDLF) is a powerful tool for constraining theoretical particles that would imply an additional cooling channel in white dwarfs. In this work, and for the first time, we use the 100-parsec Gaia DR3 white dwarf sample and compare it with theoretical predictions. We have simulated synthetic populations of white dwarfs using a population synthesis code based on Monte Carlo techniques, incorporating realistic observational errors, and based on state-of-the-art white dwarf models that incorporate the anomalous cooling caused by the presence of axions. Axion bremsstrahlung emission rates were implemented using the latest theoretical calculations. We find that, for the brightest white dwarfs in the sample ($M_{\mathrm{Bol}} < 10$), the $χ^2$ statistic is largely insensitive to the assumed stellar formation rate (SFR), which is typically the dominant uncertainty in modeling the Galactic-disk WDLF. The resulting $χ^2$ analysis disfavors a sizable additional cooling contribution. This conclusion contrasts with earlier studies in which axion-electron couplings in the range $0.7 \times 10^{-13} < g_{ae} < 2.1 \times 10^{-13}$ provided mildly improved fits to the Galactic-disk WDLF. We attribute the discrepancy to simplifying assumptions in previous modeling and to the substantially improved observational quality of the 100-pc Gaia DR3 sample. We obtain the upper limit $g_{ae} < 1.68 \times 10^{-13}$ ($95\%$ C.L.), which is among the strongest available.

Resolving mass ordering is the important thing in the neutrino physics. In Ref.\cite{yt} the authors investigate the phenomenology of a unified neutrino mixing framework and reveal that the predicted sum of neutrino masses derived from an approximate $μ-τ$ reflection symmetric flavor neutrino mass matrix based on the unified neutrino mixing with an inverted mass ordering, is excluded from DESI2024 and Supernova Ia luminosity distance data. We note that in Ref.\cite{yt} an error is present in Eq.(20), {\it i.e.}, the expression for $ M_{23}^{μ-τ} $, a similar error also appears in Eq.(26) for $M_{13}^{μ-τ}$. That is, the condition that $M_{ee}$ and $M_{μτ}$ are real is not added. We add the condition and analyze its consequences. Using the newest data\cite{azz}, it is obtained that the predicted sm derived from the approximate $μ-τ$ reflection symmetric neutrino mass matrices $M_{23}$ for both NO and IO are allowed. We note also that their conclusion is invalid, as it is based on outdated data.

We clarify the structure of Yukawa couplings and mass matrices for matter fields in heterotic string theory on smooth Calabi-Yau threefolds with standard embedding. The topological structure of Calabi-Yau threefolds leads to interesting Yukawa textures that cannot be derived from group-theoretical symmetries, e.g., the so-called Weinberg texture in the case of two generations of matter fields. Furthermore, we find that a $U(2)$ flavor symmetry, which plays an important role in controlling higher-dimensional operators in the Standard Model effective field theory, emerges at specific loci in the moduli space of multi-Higgs fields. Small perturbations around these loci generate semi-realistic patterns of quark masses and mixings.

Jets are produced in the earliest stages of heavy-ion collisions, where they can interact with a medium that is not yet close to local equilibrium. Motivated by this, we generalize the usual jet transport coefficient $\hat q$ to a Lorentz-covariant diffusion tensor $\hat q^{μν}$ within a leading-order elastic (Boltzmann/Fokker--Planck) description of jet--medium interactions. The tensor formulation organizes medium effects in a frame-covariant way and reveals additional information beyond the standard scalar definition, including energy diffusion and off-diagonal components that encode correlations between energy and momentum exchange which are absent (or redundant) in equilibrium. We illustrate the formalism in (tree-level) massless $λ\varphi^4$ theory for isotropic but out-of-equilibrium states. For sufficiently large jet momentum, quantum statistical effects become subleading, so that the non-equilibrium evolution can be studied reliably in the classical (Boltzmann) limit. This allows us to solve the corresponding Boltzmann equation for the medium and determine the time dependence of $\hat q^{μν}$ as the system approaches equilibrium. We find that out-of-equilibrium corrections can either enhance or reduce jet momentum broadening, depending on the initial distribution function.

We present the quark weak-mixing component of a Froggatt--Nielsen program, with one flavon and three messengers, in which a single hierarchy parameter $B$ (with $ε\equiv 1/B$) and a rational-exponent ``$B$-lattice'' organize fermion Yukawa textures. Building on companion mass-fit work, we translate the lattice into sharp predictions for quark mixing. The four-magnitude parameterization serves as a practical interface between the flavon Yukawa textures and quark weak mixing observables, yielding coefficient-free ratio tests of the lattice structure.

The central objects in a quantum field theory are its n-point correlation functions and matrix elements. Their structure is determined by Lorentz invariance and leads to tensor decompositions whose Lorentz-invariant coefficient functions encode the physics of the process. For growing n, the complexity of these objects may increase considerably and make it challenging to deal with them. Here we give a pedagogical introduction to the topic and provide some tools to manage this complexity, and we will show how symmetries can be used as organizing principles.

We present the first amplitude analysis of the semileptonic decay $D^0\to K^-π^0 e^{+}ν_{e}$ by analyzing $e^+e^-$ annihilation data corresponding to an integrated luminosity of 20.3 fb$^{-1}$ collected at the center-of-mass energy of 3.773 GeV with the BESIII detector. A tiny $\mathcal{D}$-wave component of the $K^*_2(1430)^-$ accounting for $(0.16 \pm 0.05_{\rm stat} \pm 0.02_{\rm syst})\%$ of the $K^-π^0$ is observed for the first time with a significance of $7.9σ$ in addition to the dominant $\mathcal{P}$-wave component of $K^*(892)^-$ and the sub-dominant $K^-π^0$ $\mathcal{S}$-wave. The hadronic form factors of the $D^0 \to K^*(892)^-$ transition are measured precisely as $r_V=V(0)/A_1(0)=1.41 \pm 0.05_{\rm stat} \pm 0.01_{\rm syst}$ and $r_2=A_2(0)/A_1(0)=0.77 \pm 0.04_{\rm stat} \pm 0.02_{\rm syst}$. The branching fraction of $D^0\to K^*(892)^-e^+ν_e$ with $K^*(892)^-\to K^-π^0$ is measured to be $(7.403\pm0.061_{\rm stat.} \pm 0.048_{\rm syst.})\times10^{-3}$. Combining the measurements of the $D^0\to K^*(892)^-(K^*(892)^-\to K^-π^0)\ell^+ ν_\ell$, lepton flavor universality is tested by the ratio $\mathcal{R}_{\rm LFU}=\mathcal{B}(D^0\to K^*(892)^-μ^+ ν_μ)/\mathcal{B}(D^0\to K^*(892)^-e^+ν_e)=0.928\pm0.020_{\rm stat}\pm0.012_{\rm syst}$ with unprecedented precision; no violation is found. Furthermore, isospin symmetry in the decay $K^*(892) \to Kπ$ is tested by $\mathcal R_{K^{*-}} =\mathcal{B}(K^*(892)^-\to K^- π^0)/\mathcal{B}(K^*(892)^-\to K_S^0 π^-)= 1.09\pm0.02_{\rm stat}\pm0.02_{\rm syst}$ for the first time using the previous measurement of $D^0\to K^*(892)^-e^+ν_e$ with $K^*(892)^-\to K^0_Sπ^-$. Finally, the phase shift of the $Kπ$ \wv{S} is extracted in a model-independent way, which sheds light on the nature of the lightest strange scalar meson, the $K^*_0(700)$.

Next-generation neutrino oscillation experiments such as DUNE require percent-level knowledge of neutrino--nucleon interaction cross sections. The nucleon axial form factor $F_A(Q^2)$, parameterized by the axial mass $\MA$, is the dominant source of uncertainty in the quasi-elastic channel, and the parity-violating structure function $xF_3$ is poorly constrained on free nucleons. We propose using charged-current (CC) electron--proton scattering at the Electron-Ion Collider (EIC) to address both problems simultaneously. The measurement exploits three key features of the EIC: (1)~helicity-selective electron bunches provide \emph{in situ} electromagnetic background rejection; (2)~a longitudinally polarized proton target enables extraction of $F_A(Q^2)$ through the target-spin asymmetry $A_{UL}$; and (3)~the $y$-distribution leverage in CC deep inelastic scattering separates $F_2$ and $xF_3$ on a \emph{free proton}, without nuclear corrections. Using a Fisher-information analysis at $\sqrt{s} = 141\GeV$ with $500\fb^{-1}$ of integrated luminosity, we project the Cramér--Rao statistical floor of $δ\MA \approx 0.03\GeV$ (3\%). Incorporating first-order realistic detector effects: ZDC acceptance, $Q^2$ smearing (5\%), and background noise from helicity subtraction, the projected sensitivity is severely background-limited due to the small signal-to-background ratio ($S/B \approx 3 \times 10^{-4}$) in the elastic channel. Achieving competitive sensitivity ($δ\MA \approx 0.14\GeV$) would require $\sim\!10^{-7}$ background suppression, three orders of magnitude beyond current projections. The CC DIS $y$-distribution provides sub-percent extraction of $xF_3^{\Wminus}$ over $0.05 < x < 0.5$, representing the most robust electroweak measurement in the near term.

In low-background particle physics experiments, surface deposition of radon progeny presents a significant background challenge. To characterize this contamination, a high-sensitivity surface $α$-activity measurement system was developed, which employs a 3$\times$3 Si-PIN array operating in vacuum to perform $α$-spectroscopy on samples. The system was calibrated using Poly(Methyl MethAcrylate) (PMMA) plates exposed to a controlled high-radon atmosphere, achieving an energy resolution of 2.09 \% for 5.30~MeV $α$ particles and a one-day measurement sensitivity of 1.27~$μ$Bq/cm$^2$ for $^{210}$Po surface activity. Using this system and a self-built high radon concentration chamber, the deposition behavior of radon progeny on PMMA surfaces was investigated. Results indicate a non-monotonic dependence on exposure time, a significant enhancement of deposition with increasing negative surface electrostatic potential, and a strong modulation by ambient humidity. This paper details the apparatus design, calibration, and experimental study of radon progeny deposition dynamics on PMMA surfaces.

The singlet majoron $J$ is the pseudo-Nambu-Goldstone boson of a global, anomaly-free $U(1)_{B-L}$ symmetry whose spontaneous breaking generates Majorana masses for right-handed neutrinos. At tree level, the only direct coupling of $J$ to Standard Model fields is $Jνν\propto m_ν/f$ (where $m_ν$ denotes the light neutrino mass and $f$ the $B-L$ breaking scale). Couplings to charged fermions and gauge bosons, in contrast, arise only at loop level. Consequently, $J$ can be long-lived over wide regions of parameter space, motivating displaced-decay searches. We study majoron production and displaced decays at proton beam dump experiments, neutrino facilities, and LHC forward detectors (including DUNE, NA62, FASER/FASER2, MATHUSLA, and SHiP), and we quantify the resulting reach in the $(m_J,\,f)$ plane. We show that, for realistic seesaw-induced coupling textures, lepton-flavor-violating (LFV) $τ$ decays $τ\to \ell J$ ($\ell=e,μ$) dominate majoron production at these facilities and can extend sensitivity into the intermediate-mass window $m_J\simeq 0.2\text{-}1.7~\mathrm{GeV}$, complementary to supernova bounds at lower masses and to dedicated LFV searches at higher masses. We also identify physically consistent benchmark textures for the matrix $K=M_D M_D^\dagger/(vf)$ with $M_D$ denoting the Dirac mass matrix and $v$ the electroweak scale (including positive semidefinite ``anarchical'', single-flavor, and CP-violating cases) and map their impact on experimental reach.

The recent observation of coherent elastic neutrino-nucleus scattering from solar $^8$B neutrinos in dark matter direct detection experiments has inaugurated the \emph{neutrino fog} era, highlighting the extended potential of these experiments as precision neutrino observatories. Recent measurements by the XENONnT, PandaX-4T, and LUX-ZEPLIN experiments provide new opportunities to test Standard Model predictions and to probe physics beyond it, in complementarity with dedicated neutrino facilities. We perform a combined analysis of nuclear recoil data from these three facilities to extract information on the solar $^8$B neutrino flux normalization and on the weak mixing angle at low-momentum transfer. We further investigate the impact of new vector interactions on the solar neutrino event rate, deriving constraints on nonstandard neutrino interactions and on scenarios with light vector mediators. Our results demonstrate that dark matter detectors are rapidly becoming complementary to terrestrial neutrino experiments in probing neutrino interactions, and already set competitive bounds on both light and heavy vector mediators.

We report the design and functionality of the Murchison Widefield Array Particle Detector Array (MWA PDA), an array of eight particle scintillation detectors deployed to Inyarrimanha Ilgari Bundara, the Murchison Radio-astronomy Observatory (MRO). The purpose of the instrument is to identify cosmic ray extensive air showers (EAS) occurring over the core of the MWA radio telescope, and generate a trigger to allow radio data on the event to be captured and analysed. The system also acts as a pathfinder for a much larger instrument to be deployed in the core of the low-frequency component of the Square Kilometre Array, SKA-Low, by the SKA's ultra-high-energy particles science working group. Here, we describe the instrument and associated infrastructure, which has been verified to comply with the strict radio-frequency emissions requirements of the MRO, and was deployed in November 2024. We present calibration data, which demonstrates the ability of each detector to identify individual atmospheric muons at the expected rate, and we characterise the temperature dependence of the system. We describe a sample of 35,500 EAS identified using multi-detector coincidence over a 13-day period, and show how the detector data can be used to reconstruct the arrival directions and approximate energies of these events. We conclude that the PDA can reliably trigger on and reconstruct EAS contained within the $\sim 103 \times 90$ m$^2$ core region, arriving within 20$^{\circ}$ of zenith, at primary cosmic ray energies above $\sim 4$ PeV. We have also verified that the detector array can generate triggers, allowing the capture of radio data from the MWA correlator for offline analysis.

The transverse-momentum-dependent (TMD) factorization theorem for dijet production in deep-inelastic scattering is used here to make predictions of the gluon Sivers function. We revise the previously studied unpolarized case and develop the formalism for a transversely polarized target. We study the impact of TMD evolution in two different schemes and we use the current extractions of the evolution kernel at N$^3$LO to make predictions for the future Electron-Ion Collider (EIC). The results strongly depend on the TMD gluon distributions and their evolution kernel. Big values of the Sivers asymmetry at the EIC are predicted, between 5-50$\%$

For a long time studies of femtoscopic correlations have provided information about space-time characteristics of particle sources in high-energy collisions. Recently, the correlation functions have been also used to determine interaction parameters of correlated particles which is especially important for short-lived particles, for which scattering experiment are impossible. The abundance of experimental data and their high accuracy require an improved theoretical approach to femtoscopic correlations. We discuss relativistic effects and their role in detail. Since a general relativistic approach is currently unavailable, due to serious theoretical difficulties, the correlation functions must be computed in the center-of-mass frame where the correlated particles are mostly nonrelativistic. This requires transforming the source function to this frame, the consequences of which we discuss. Since the deutron formation has a similar physical origin to femtoscopic correlations, we also discuss relativistic effects in the former process. We illustrate our considerations with calculations of some correlation functions and the deuteron coalescence coefficient to demonstrate a magnitude of relativistic effects. We also argue that a discrepancy between the coalescence coefficient computed with the source radii inferred from the baryon-baryon correlation functions and the experimental data can be removed by taking into account the relativistic elongation of the source radius in the center of mass of correlated particles.

First-order cosmological phase transitions (PT) can take place in a dark sector at relatively late times between the big-bang nucleosynthesis and recombination epochs. Because bubble nucleation is stochastic, the PT completes at different times in different regions of the Universe. This fluctuation sources a curvature perturbation whose (dimensionless) power spectrum ${\cal P}_ζ(k)$ features a universal infrared tail, independent of the microscopic details of the PT. Even in the absence of any non-gravitational interaction between the dark sector and the Standard Model, these additional curvature perturbations at small scales impact a variety of observables. We derive new constraints on dark sector phase transitions using {\it Planck}, baryon acoustic oscillation (BAO), Lyman-$α$ observations, spectral distortion limits from FIRAS, constraints on early reionization, and the existence of ultra-faint dwarf galaxies.

A fixed-target program at the Electron-Ion Collider (EIC) would broaden the facility's scientific reach by providing key measurements for studies of cold nuclear matter (CNM), the QCD phase diagram, and nuclear reactions relevant for space radiation. Constraining CNM effects is essential for interpreting observables in proton-nucleus ($p+A$) and nucleus-nucleus ($A+A$) collisions, yet these effects are poorly understood at low center-of-mass energies. In particular, the range $\sqrt{s}\approx10$--20 GeV, where several CNM effects may compete at similar scales, has not been explored with high statistical precision. Mapping the QCD phase diagram similarly requires high-statistics $A+A$ data with broad rapidity and transverse-momentum coverage to probe the onset of deconfinement and the possible location of the QCD critical point (CP). Currently, such data are limited at 4.5 GeV $< \sqrt{s_{NN}} < 7.7$ GeV A fixed-target program at the EIC would fill these gaps, providing CNM baselines and complementary data for QCD CP studies. By delivering high luminosity and systematic measurements across a broad range of nuclear targets, the program would link $p+A$ and $A+A$ systems at the same center-of-mass energies, enabling a unified, quantitative description of cold QCD matter and clarifying the interpretation of QGP signatures in $A+A$ collisions at low energies where comparable $p+A$ data are lacking. Finally, the program would offer a unique opportunity to measure nuclear cross sections critical for improving cosmic-ray models, including studies of space-radiation protection for both autonomous spacecraft and long-duration human spaceflight.

The possibility of a photon collider extension to the beam dump of the $17.5$ GeV European XFEL has already been discussed before as the first high energy collider of its sort. It would not only be the first proof of concept and test of a photon collider but would also be a collider without competition in the region of $E_{γγ}=5-12$ GeV for photon-photon collision. In this range, $b\bar{b}$ and $c\bar{c}$ resonances, tetraquarks and mesonic molecules can be observed. Furthermore, some BSM processes can also be reached in this range. In this paper we want to discuss the possibility of observing ALPs in the process of light-by-light scattering at such a collider. We will use a simplified description of the Compton backscattering process to get a first look at cross sections for the Standard Model light-by-light scattering and the extension including ALPs. Furthermore, we extend this to the full beam dynamics included prediction, discuss all effects that are important when working with a photon collider and show that the photon collider with energy $E_{γγ}<12$ GeV would offer an extended physics reach compared to current limits.

We study a 3-Higgs Doublet Model (3HDM) with an imposed Z3 symmetry, allowing for two Inert scalar doublets and one active Higgs doublet. The WIMP dark matter candidates correspond to two mass-degenerate states, H1 and A1, which possess opposite CP quantum numbers and can reproduce the correct relic density simultaneously with all theoretical and experimental constraints. We use state-of-the-art machine learning algorithms to probe the parameter space of the model by employing an Evolutionary Strategy augmented with Novelty Reward. We consider two situations: a limit for the dark matter mixing angle θ that closes a gauge annihilation channel that would deplete the DM relic density, and the general case without imposing this limit. For both scenarios, we find viable dark matter candidates within two separate mass regimes, ranging from 50 GeV < mDM < mW and 380 < mDM < 1000 GeV. Moreover, we find it is possible to fulfill all the existing constraints while still obtaining values for the dark matter-higgs coupling of order O(0.1). Exploring the model outside the θ = π/4 limit proved to be an extremely challenging task, as there are regions which seem easy to explore when projected on a 2D plane, yet may be completely disconnected on the hypersurface supporting the valid points, given its non-convex and multi-dimensional nature. We consider new methods of prototype selection to seed new exploration runs, which allow for efficient global scans over the parameter space.

The dark photon is a new gauge boson that naturally arises in many beyond the Standard Model theoretical models, featuring interactions that resemble quantum electrodynamics. Due to this feature, it is often considered the portal between dark and visible sectors. For this reason, it has become the target of many experimental searches worldwide. In this work, we propose a search for dark photons based on the Inverse Compton scattering, $γe^- \rightarrow A^\prime e^-$, to be conducted at electron accelerators. In this setup, photons from a laser source would impinge on the accelerated electron beam, producing a dark photon in the final state. We propose an experimental setup to take advantage of the photon counting technique, and we derive the projected sensitivity by considering the energy of the incident photon to be about 1 eV and an electron beam of 3 GeV. We show that this experimental setup could cover an unexplored region of parameter space and constitute a promising probe for dark sectors in the future.

We consider phase transitions out of a general topological phase in $2+1$ dimensions. We assume that the transition is triggered by a single Abelian anyon, which becomes light near the transition and whose worldlines proliferate after the transition. (This proliferation is often referred to as ``condensation.'') We describe the transition using a continuum field theory obtained by coupling the corresponding topological quantum field theory (TQFT) to a single complex scalar field associated with this anyon. With these assumptions, we find the most general relativistic field theory for such a transition. Even though for a given TQFT and a choice of anyon, there are infinitely many such field theories, the transition theory depends on only a single additional integer parameter. We analyze all these theories, their global symmetries, and their phases. In generic cases, the theory after the transition can be related to the original one via an Abelian hierarchy construction. In special cases, the theory after the transition is gapless, and with a particular deformation, it is related to the original TQFT by gauging an anomaly-free one-form global symmetry. We also explore the enrichment of this setup by a global U(1) symmetry. In some cases, enriching the original TQFT is incompatible with the full transition theory. Finally, we demonstrate our construction with many specific examples.

We use the EFT developed in \cite{Mehtar-Tani:2025xxd}, to understand the interference driven phenomenon of color decoherence in inclusive jet production in a dense nuclear medium such as Nuclei or Quark Gluon Plasma. Using the factorization formula in \cite{Mehtar-Tani:2024smp,Mehtar-Tani:2025xxd}, expressed as a series of multi-sub-jet operators, we define and calculate the contribution of the two sub-jet effective operator. This explicitly reveals the emergent angular scale $θ_c$ that controls color decoherence as well an intricate renormalization group structure for the factorized functions. We show that for a jet of radius R in a medium of size L characterized by a jet quenching parameter $\hat q$, both the LPM effect and color decoherence are controlled by a single dimensionless parameter $\sqrt{\hat q L}LR$ and therefore are equally important for phenomenology. This paper shows how interference driven emergent effects can be included in a factorized framework for computing jet observables in heavy ion collisions.

A compelling production mechanism for QCD axion dark matter is from the scaling dynamics of early universe axion strings. We show that in DFSZ-like models containing tree-level interactions between fermions and the axion, friction between the thermal bath and the axion string drastically changes the behavior of the axion string network for lower $f_a$ values. Friction delays the onset of scaling and increases the energy density of axions. Once the effects of friction are included, we argue that in addition to the standard value of $m_a \sim$ meV, $m_a \sim 0.1$ eV also reproduces the dark matter energy density.

The Swampland Cobordism Conjecture predicts the existence of end-of-the-world branes for every consistent Quantum Gravity theory, and domain walls connecting the Landscape. A perturbative string worldsheet description of these objects is only expected to exist when certain worldsheet invariants are vanishing or coincide across the domain wall. In this paper, we observe that many of these worldsheet obstructions can be evaded by allowing non-compact string worldsheets as part of the bordism. Using these ideas, we provide a worldsheet QFT (flowing to a critical CFT under RG flow) that connects the worldsheets of 0A and 0B string theory, as well as those of IIA and IIB string theories. The non-compact character of these interpolations means that the description of the actual domain walls develop strongly coupled regions described by a linear dilaton background, where the worldsheet description breaks down. As a result, the IIA/IIB domain wall requires a strongly coupled region, in agreement with effective field theory considerations. Our constructions are heavily inspired by supercritical string theory, but are logically independent from it. We also analyze the fate of IIA/IIB NS5 branes as they cross the domain wall between these two theories.

We revisit the strong CP problem from the viewpoint of the infrared structure of non--Abelian gauge theories. In Yang--Mills theory, motion between topologically inequivalent vacua may be described in terms of a compact collective coordinate associated with the Chern--Simons number. Implementing an adiabatic separation between slow topological modes and fast gluonic fluctuations leads to a reduced Born--Oppenheimer Hamiltonian governing the infrared dynamics. We show that the physical parameter entering this reduced Hamiltonian is not the bare vacuum angle $θ$, but an effective holonomy $θ_{\rm eff}$ that includes a Berry phase induced by the fast gluonic sector. The induced holonomy becomes a self--consistent response function of the infrared dressing, leading to a nonperturbative renormalization group flow for $θ_{\rm eff}$. This infrared flow admits CP--invariant fixed points toward which the effective vacuum angle is dynamically driven in the infrared limit. In this framework, CP violation is not forbidden by the fundamental theory but becomes dynamically suppressed along the infrared flow generated by adiabatic dressing. The strong CP problem is thus realized as a nonperturbative infrared relaxation mechanism governed by the Berry response of the fast gluonic sector, without the introduction of additional dynamical fields.

Analysis of binary-lens microlensing events typically requires intensive computation because of the multimodal and complex posterior distributions. With the recent development of the JAX-based differentiable binary-lensing modeling package microlux, we present an analysis of two microlensing events with planet/brown-dwarf candidates, KMT-2025-BLG-1314 and KMT-2025-BLG-1392. Both events exhibit the "Close/Wide" degeneracy, and KMT-2025-BLG-1314 suffers from the "Planet/Binary" degeneracy and a recently recognized "Point/Finite" degeneracy among the planetary solutions. For KMT-2025-BLG-1314, the binary mass ratio is $\log q \sim -3.5$ for the planetary solutions and $\log q > -1.5$ for the binary solutions, while for KMT-2025-BLG-1392, we find $\log q \sim -1.3$. We show that for the analysis of KMT-2025-BLG-1314, Hamiltonian Monte Carlo (HMC), enabled by microlux, provides robust parameter inference and outperforms traditional Markov chain Monte Carlo (MCMC) methods in the presence of bimodal posteriors.

Stellar mass, size, and star formation rate (SFR) are fundamental properties that encode the structural and evolutionary states of galaxies. Observations reveal a mass-SFR-size relation whereby galaxies become more compact both above and below the ridge of the star-forming main sequence (SFMS), linking galaxy structure to star formation activity. We investigate this relation by comparing galaxies from two cosmological hydrodynamical simulations, EAGLE and TNG100, with observational samples from SDSS and CANDELS over three redshift intervals (0 < z < 0.2, 0.5 < z < 1.5, and 1.5 < z < 2.5). Both simulations reproduce the observed trend that galaxy sizes decrease with increasing offset away from the SFMS. This trend, however, weakens and is not detected in the observational sample at 1.5 < z < 2.5, likely due to increased measurement uncertainties. In contrast, the trend persists in both simulations up to z = 2.5. Across all redshifts, EAGLE predicts a stronger size dependence on SFMS offset than observed, whereas TNG100 exhibits a weaker dependence. We discuss how this mass-SFR-size relation can be understood in terms of different time variability in star formation rate across the SFMS.

Recent JWST surveys have revealed two puzzling high-redshift phenomena: (1) an unexpectedly large abundance of flattened, disk-like galaxies at z > 3, and (2) a rare population of compact, extremely red sources at z ~ 4-9 ("Little Red Dots," LRDs) that often show V-shaped SEDs and very broad Balmer lines. These findings lack a consensus interpretation and have motivated models ranging from dusty starbursts to obscured AGN and more exotic scenarios. We propose that both phenomena are linked by a simple geometric consequence of a third clue: mounting evidence from structural modeling and axis-ratio statistics indicates that many low-mass galaxies at z > 3 are intrinsically prolate (cigar-like), not oblate rotation-supported disks. In this picture, a substantial fraction of the flattened, disk-like morphologies reported at z > 3 arise from side-on and intermediate-angle projections of prolate systems, while the rare near end-on views appear extremely compact and high-surface-brightness, and are preferentially reddened by the maximal line-of-sight column, naturally matching key elements of LRD selection. The expected fraction of near end-on systems, $P_{\text{end}} = 1 - \cos(θ_{\text{max}})$, is ~1-3% for $θ_{\text{max}}$ ~ 10°-15°, consistent with LRD demographics in wide JWST fields. This orientation-based framework does not exclude AGN or starburst activity; rather, it explains LRD rarity as an orientation effect and provides a natural route to the large columns of gas/dust and scattering depths inferred in recent dense-gas and electron-scattering interpretations of LRD spectra, without fine-tuned new physics. The model makes falsifiable predictions to validate or rule out this geometric interpretation.

We investigate the spectro-polarimetric properties of the long-duration GRB~220107A, which exhibited two distinct emission episodes separated by a 40 s quiescent gap, to test whether such multi-episode bursts show evidence for evolution in their underlying radiation mechanisms. We analyzed prompt emission data from AstroSat/CZTI, Fermi/GBM, and Konus-Wind, performing spectro-polarimetric analysis for each emission episode. The time-integrated polarization analysis shows no significant detection (PF$ < 38 \%$, $2σ$). Time-resolved analysis reveals clear spectral evolution between the two episodes, with episode 1 exhibiting a hard low-energy photon index and episode 2 showing substantial spectral softening ($α\sim -0.72$). Regarding polarization: Episode 1 shows a low polarization upper limit (< 52\%), consistent with expectations for photospheric emission dominated by quasi-thermal Comptonization in a baryon-rich outflow. Episode 2 also shows overall low polarization (PF$ < 55 \%$, $2σ$), though sliding-window analysis yields a marginally elevated signal (PF$= 70 \pm 30\%$, BF = 2.8) between T0+76 to T0+88 s. The robust spectral softening between episodes could arise from sub-photospheric dissipation, optically thin synchrotron radiation in small-scale magnetic fields, or if the tentative polarization enhancement proves intrinsic, it would favor synchrotron emission in large-scale ordered magnetic fields. The spectral evolution of GRB 220107A, combined with our polarimetric constraints, demonstrates the diagnostic potential of time-resolved spectro-polarimetry for constraining GRB prompt emission physics. We present GRB 220107A as a test case illustrating how future higher sensitivity observations could discriminate between competing emission models for multi-episode bursts. Our results emphasize both the promise and current limitations of prompt phase polarimetry.

Braking of pulsars, or the law of their spin deceleration, is a manifestation of the combination of various processes occurring in the magnetospheres of neutron stars and their internal structural dynamics. The interaction of superfluid neutron vortex lines with superconducting magnetic flux tubes in the neutron star core plays a significant role in rotational evolution and magnetic field evolution through spin glitches and magnetic flux expulsion. In this study, the effect of this interaction on the temporal variation of the braking index of pulsars is investigated. The variation of the deviation from the $n=3$ value predicted by the generally accepted magnetic dipole formula with respect to physical parameters such as the magnetic field and the age of the neutron star has been elaborated, and applications have been made to pulsars with reliably measured braking indices.

We study the role of domain walls and their relics in the very early Universe within the framework of the mode-matching technique. Domain walls formed during the spontaneous breaking of discrete symmetries are modelled as a short lived contribution to the background energy density, leading to a controlled deviation from standard slow-roll inflation. Solving the modified Friedmann equation, we obtain a smooth, time-dependent Hubble parameter that asymptotically approaches the standard inflationary value. We analyze the evolution of scalar perturbations using the gauge-invariant Mukhanov variable and perform mode matching across the transition between the domain wall affected phase and standard inflation. We find that only modes exiting the horizon near the transition experience a modified evolution, while modes far from the transition remain unaffected. As a result, the primordial scalar power spectrum exhibits localized, scale-dependent deviations from scale invariance, while preserving the overall success of inflation. These results demonstrate that unstable domain-wall relics can leave subtle imprints on primordial fluctuations, potentially providing a probe of early-Universe phase transitions.

This paper presents the CosmicWeb-21cm array, a novel radio interferometer designed to overcome the key challenges in 21 cm cosmology. Its core innovations include: (1) a multi-scale nested geometry combining a hexagonal core with logarithmic spiral arms for excellent UV coverage and calibration robustness; (2) an intelligent non-uniform frequency sampling strategy that adapts resolution to foreground and signal characteristics, reducing data volume while preserving information; and (3) a machine-learning-enhanced, physics-informed processing pipeline that achieves 99.7\% foreground removal efficiency; (4) a dual-polarization crossed dipole integrated with a dielectric lens and cryogenically cooled LNA, achieving stable beam patterns and low noise temperature ($<35$ K) across 50-250 MHz. These co-designed advances enable high sensitivity mapping of the Epoch of Reionization, dark energy constraints and cosmic-web structure.

High-redshift Lyman continuum emitters (LCEs) are crucial for understanding how galaxies ionize the neutral hydrogen in the epoch of reionization. However, detected LCEs at $z > 4$ are quite rare. Here we report an LCE candidate at $z = 4.444$, dubbed LCEz4-M1, which is the highest-redshift LCE to date with LyC detections confirmed in two independent data sets. The redshift is determined from the $Lyα$ emission line detected in the VLT/MUSE spectrum. The LyC signal is detected independently in the Hubble Space Telescope (HST) F435W image and the VLT/MUSE spectrum at significances of $\simeq 3.7~σ$ and $\simeq 2.8 - 3.0~σ$, respectively. The centroid of the LyC emission is closely aligned with the rest-frame optical continuum traced by James Webb Space Telescope (JWST) imaging, with an offset of $\simeq 0''06$ (0.4 kpc in physical scale). Based on HST/ACS F435W photometry and MUSE spectroscopy, we infer LyC escape fractions of $f_{esc}(F435W) = 0.38_{-0.15}^{+0.25}$ and $f_{esc}(MUSE) = 0.33_{-0.13}^{+0.22}$. Using the combined JWST and MUSE data set, we characterize the physical properties and morphology of LCEz4-M1. The galaxy is compact and lies in the starburst regime, with a high star formation rate surface density of $Σ_{\rm SFR} \simeq 7~M_{\odot}~{\rm yr}^{-1}~{\rm kpc}^{-2}$, consistent with conditions that can drive strong feedback and outflows. The feedback may generate low-column-density pathways in the interstellar medium that facilitate LyC escape. While we find no clear evidence for an ongoing major merger, the presence of a faint companion ($\sim 0''05$) detected in the F277W band suggests a potential minor interaction. This is also consistent with LCEz4-M1 residing in an overdense environment, where elevated interaction rates and dynamical perturbations are expected.

Little red dots (LRDs) draw extensive attention because of their unique observational characteristics and apparent overabundance in the early Universe, raising new insights into early black hole formation and growth. Early studies show that LRDs exhibit weak variability in broad-band photometry and emission-line fluxes, suggesting a preference for super-Eddington accretion or disfavouring an AGN origin. However, the cadence of the current data, and therefore, the resulting light curves for LRDs, is limited, preventing us from placing strong constraints on their variability. Based on Zwicky Transient Facility (ZTF) light curves with a baseline of $\sim6$ years, we here study the optical variability of seven previously reported local analogs of LRDs at $z \sim 0.3$, offering an insight into LRDs from a low-redshift sample. Three out of seven local analogs show excess variances on all three bands of their light curves, and two of them can be fitted with the damping random walk model, supporting their AGN origins for the variability. The remaining sources show weak variance in at least one band, but no detectable variability at the current sensitivity level, exhibiting $\rm SF_\infty$ upper limits consistent with estimates from high-redshift (high-$z$) LRDs. Their non-detection of variability is likely due to the large photometric uncertainty. As an implication, by simulating long baseline light curves with the variability amplitude of local analogs and adopting JWST observation cadence, we investigate the limitation of the variability amplitude estimate for LRDs. Our mock observations imply that the current constraints on LRDs' variability are probably underestimated. This underestimation might be induced by the short temporal baseline of observations, as well as the intrinsic scatter of the empirical $M_{\rm BH}-τ$ relation.

We investigate the infrared (IR) properties of a sample of local star-forming galaxies using the WISE data. Focusing on the 20 confirmed strong Lyman continuum (LyC) leakers ($f_\mathrm{esc}>5\%$) included in this sample, we find that the IR detection separates these strong LyC leakers into two populations. The IR-undetected strong leakers in our sample exhibit high [OIII]5007/[OII]3725,3727 (O32) ratios, blue UV slopes, and low stellar masses, consistent with the classical density-bound scenario where the entire ISM is highly ionized. However, IR-bright strong leakers display unexpectedly low O32 ratios while maintaining a substantial escape fraction of $f_{\text{esc}} \sim 12\%$. They also have higher stellar masses and redder UV slopes, similar to weak or non-leakers, along with low nebular extinction, suggesting that the LyC photons are likely to escape through low-column-density channels in a high-density, ionization-bound clumpy medium. Morphologically, these two populations echo the compact starburst and merger-driven LyC leakers observed at $z \sim 3$, indicating that both escape pathways coexist across cosmic time. Our results demonstrate that efficient LyC escape is not limited to the lowest-mass, dust-poor dwarfs but can also occur in more massive, dusty environments, highlighting the potential contribution from dusty systems and posing new questions about the actual dominance of different galaxy populations during the Epoch of Reionization.

Compton-thick active galactic nuclei (CT-AGNs), defined by column density $\mathrm{N_H} \geqslant 1.5 \times 10^{24} \ \mathrm{cm}^{-2}$, are so heavily absorbed that their X-ray emission is often feeble, even undetectable by X-ray instruments. Nevertheless, their radiation is expected to be a substantial contributor to the cosmic X-ray background (CXB), predicting that CT-AGNs should comprise at least $\sim$30\% of the total AGN population. In the Cosmological Evolution Survey (COSMOS), the identified CT-AGN fraction falls far below theoretical expectations, indicating that a substantial population of CT-AGNs is hidden due to their low photon counts or their flux below the current flux limits of X-ray instruments. This work focuses on identifying CT-AGNs hidden in mid-infrared (MIR)-selected AGNs. First, we selected a sample of 1,104 MIR-selected AGNs that were covered but individually undetected by X-ray. Next, we reduced the X-ray data in the COSMOS and analyzed multiwavelength data in our sample to derive the key physical parameters required for CT-AGN identification. Using MIR diagnostics, we first find out 7 to 23 CT-AGN candidates. Their subsequent X-ray stacking analysis reveals a clear detection at $>3σ$ significance in the soft band and only $>1σ$ significance in the hard band. We fit the stacked soft- and hard-band fluxes with a physical model and confirm that these sources are absorbed by Compton-thick material. However, CT-AGNs constitute only 2.1\% (23/1104) of our sample, significantly below the fraction predicted by CXB synthesis models, indicating that a considerable population of CT-AGNs remains missed by our selection. A comparison of host-galaxy properties between CT-AGNs and non-CT-AGNs reveals no significant differences.

Time-domain photometric monitoring is essential for characterizing cometary evolution, particularly for rare interstellar objects with limited observing opportunities. We aimed to characterize the pre-perihelion photometric behavior and dust activity of the interstellar comet 3I/ATLAS, and to test the capability of the Black Hole Target and Observation Manager (BHTOM) platform and telescope network for coordinated high-cadence non-sidereal observations. We obtained 70 days of time-series photometry of 3I/ATLAS from 2025 July 4 - September 11 using 16 telescopes and 1554 images. The data were processed and calibrated with the BHTOM pipeline. High-cadence, multi-band imaging was used to measure the rotation period and color evolution, while the dust activity was quantified via Afp measurements. We present a pre-perihelion light curve of 3I/ATLAS from Rh = 3.18 - 2.19 au, which exhibited a steady increase of ~3 magnitudes with no evidence of anomalous behavior. We measured a rotation period of P_rot = 15.98 +/- 0.08 h. The relative dust production increased from A(0)fp ~600 - 1100 cm, and the upper limit on the dust mass-loss rate increased from \leq 217 kg/s to \leq 328 kg/s. We measured an activity index of n = -1.24 +/- 0.02, consistent with a well-developed dust coma. The colors were statistically non-changing, with only a weak, non-significant tendency for 3I/ATLAS to become bluer at 3.5 > Rh > 2.2 au.

While the size-mass relation provides insight into the structural evolution of galaxies, the data available and methods employed have hindered our ability to study a detailed and comprehensive description of this key relation across cosmic history. The first paper in this series presents a morphology catalog based on 20 band JWST data in the field of Abell 2744. In this paper we utilize this catalog to measure the size-mass relation from $0.5<z<8$ and $0.5<z<3$ for star-forming and quiescent galaxies respectively. We perform a global fit to our sample using B-splines to flexibly model the redshift evolution which enforces smooth evolution and can account for all observational uncertainties. Symbolic regression is used to derive simple and portable expressions that describe the redshift evolution of the size-mass relation. Analyzing the size evolution of star-forming galaxies in the context of previous work at $z\sim0$ and $z>10$, we discuss three distinct phases: Rapid growth at $z>5$, growth that mimics dark matter halos at $5< z <1$ and a late plateau at $0.5<z<1$. For quiescent galaxies we confirm previous findings that the size-mass relation flattens at $\log\ M_*/M_\odot < 10$, which inverts at $z>1$. Our results imply that quiescent galaxies are smaller than their star-forming counterparts only at around $\log M_*/M_\odot = 10$; the two populations have similar sizes at lower and higher masses.

We present ultra-high-resolution (0.03"~230 pc) Atacama Large Millimeter/submillimeter Array (ALMA) observations of the hyperluminous dust-obscured galaxy W2305-0039 at z=3.111, targeting the CO J=7-6 and J=11-10 lines. The CO(11-10) emission is extremely compact and exhibits anomalously high excitation relative to CO(7-6) within the central <500 pc. X-ray-dominated region models successfully reproduce this excitation, providing strong evidence for intense X-ray irradiation by a deeply obscured active galactic nucleus (AGN), while photodissociation-region models fail to match the observed ratio. Forward modeling of the nuclear CO(11-10) position-velocity diagram yields a dynamical black-hole mass of log(M$_{\rm BH}$/M$_{\odot}$) = 8.3$^{+0.7}_{-0.6}$ and an intrinsic gas velocity dispersion of $277~^{+16}_{-14}$ km s$^{-1}$. Combined with the AGN luminosity from infrared spectral energy distribution decomposition, these measurements imply a highly super-Eddington accretion state with $λ~_{\rm Edd}~\gtrsim 4$. Our results provide dynamical evidence that the most rapid phases of black-hole growth can occur within a compact, heavily obscured nuclear region. Extending ALMA beyond its current 16 km maximum baselines will be essential for pushing such dynamical measurements to tens-of-parsec scales and resolving the black-hole sphere of influence in massive galaxies at $z \gtrsim 6$.

Understanding the formation pathway for close-companion protostars is central to unraveling the processes that govern stellar multiplicity and very early star formation. We analyze a large sample of 51 Class 0/I close-companion protostellar systems, of which 38 show detectable outflows, yielding 42 measured outflows used in our analysis. We use ALMA observations of 11 systems in Perseus and 40 systems in Orion. These companions formed either directly at these small scales ($\lesssim 500$ au separations) via disk fragmentation or at larger scales ($> 1000$ au separations) via turbulent fragmentation followed by inward migration. Because of differences in formation mechanism, the former is expected to have preferentially aligned disks and outflows, whereas the latter is expected to show no preferred alignment. The relative prevalence of these formation pathways remains uncertain, yet it is critical to forming a comprehensive picture of star formation. We examine the distribution of position angles of companion protostars relative to the position angles of their molecular outflows. The outflow, as traced by $^{12}$CO ($J=2\rightarrow1$), is a useful proxy for the angular momentum of the system, expected to be orthogonal to the binary orbital plane. We use a simple model to account for random sampling of inclination and orbital phase in each system, finding that the observations are consistent with a distribution in which the outflows are preferentially orthogonal to the companions. Based on this analysis, we suggest disk fragmentation is the dominant formation pathway for close-companion protostellar systems.

The disagreement between early and late Universe electromagnetic measurements of the Hubble constant, $H_0$, known as the Hubble tension, highlights the need for independent and complementary probes. Gravitational-wave events have recently emerged as such a probe for constraining cosmological parameters. $H_{0}$ inference using these events relies on sky localization and luminosity distance estimates, both of which can be significantly improved for strongly lensed events with appropriate lens modeling. In this context, we propose utilizing strong lensing of dark sirens, gravitational-wave events without identified electromagnetic counterparts, in combination with strong lensing of galaxies as a novel method for measuring $H_0$. The constant is inferred from the luminosity distances of these lensed dark sirens and the redshifts of their host galaxies, combining information from individual events to obtain statistically stronger constraints when multiple events are available. We adopt a simulated galaxy catalog, \texttt{MICECATv2}, as the basis for simulating strong lensing of galaxies and to provide the redshift information of host galaxy candidates required to infer $H_0$. We also examine the impact of galaxy catalog incompleteness on the resulting $H_0$ inference. Our results demonstrate that using only 8 strongly lensed dark sirens, analyzed with a dedicated galaxy-galaxy lensing catalog, can improve the precision of $H_{0}$ by roughly 50\% compared to 250 unlensed events.

Context. This paper presents a detailed description of the Decametre Pulsar and Transient Survey of the Northern Sky that was carried out in 2012-2017 using the world's largest radio telescope at decametre wavelengths - UTR-2 in Ukraine. This extensive survey covers the northern sky from declination -10 to +80 deg , with a temporal resolution of 8 ms, and explores dispersion measures up to 30 pc/cc. Aims. The major advantage of the decametre wavelength range is a comparatively wide band, in which the dispersive delay due to the interstellar plasma reaches hundreds of seconds, giving us the opportunity to determine the dispersion measure with a very high accuracy. This enables us to discover new transients, while avoiding data contamination from numerous weak signals of a different nature. Methods. The drift-scan survey in 5-beam mode of UTR-2 was carried out at night time. To cover the entire sky along the right ascension, the duration of the sessions was more than 12 hours at a time close to the autumn and spring equinoxes (to obtain the same conditions for the interference situation). 90 degrees along the declination were covered by five beams in 40 days (each equinox). Results. We discovered 380 individual transient signals with dispersion measures significantly differ from those of known sources. We determined the parameters of each single transient signal. We show that they cannot be explained by ionospheric scintillations. Repeated observations have shown that some detected transient signals are repetitive and are thus likely to originate from pulsars or rotating radio transients. Key words. Stars: neutron - pulsars: general - Methods: data analysis - Methods: observational - Astronomical databases: Surveys

We study outflow from an unmagnetized, shocked accretion disk around a non-rotating super-massive black hole using multidimensional hydrodynamics simulation with radiative cooling. We aim to investigate whether such shocked accretion flow can launch sustained collimated bipolar outflow reaching out to thousands of gravitational radii even in the absence of magnetic field and if yes, what terminal velocity can they achieve. We present the results of a few simulations of geometrically thick accretion flow with increasing specific angular momentum on a vertically elongated cylindrical domain. We show that bipolar outflow from a region very close to the black hole is originating and propagating vertically out to our simulation domain boundary at around 2651 Schwarzschild radius. The outflow attains a terminal velocity with a maximum value found to be 0.14c and the outflow rate depends on the angular momentum value of the accreting material. We also compute the self-Comptonized bremsstrahlung spectra for all the disk-jet runs.

The 21-cm signal from neutral hydrogen serves as a critical tool for unraveling the astrophysical processes that shaped cosmic dawn and the epoch of reionization. We explore the usually overlooked impact of re-scattering of 21-cm photons during and after reionization, similarly to cosmic microwave background photons. This scattering affects the observed brightness temperature by mixing the original signal with light scattered into the line of sight from other regions, effectively at the mean 21-cm brightness temperature. This gives a small but significant effect. We show that it attenuates the fluctuations in a 21-cm image by $4-7\%$, while reducing the 21-cm power spectrum by a scale-independent $7-13\%$ during cosmic dawn and reionization. Incorporating this correction is vital for precisely comparing theoretical predictions with observations from experiments such as NenuFAR, LOFAR, and HERA, and the upcoming Square Kilometre Array.

We report on a giant stellar flare from the RS CVn-type binary UX Arietis, detected with the Monitor of All-sky X-ray Image (MAXI) and followed by a 12-day optical spectroscopic campaign using the 3.8~m Seimei Telescope. The flare released $5 \times 10^{37}$~erg in X-rays (0.1--100~keV) and $(2$--$6) \times 10^{36}$~erg in the H$α$ line, placing it among the most energetic events of its kind. The H$α$ light curve showed sinusoidal modulation atop an exponential decay, consistent with reappearance of the flaring region due to binary rotation. At orbital phase 0, when the primary star is farthest from the observer, 40\% of the H$α$ flux was obscured, while at phase 0.5 the full emission was visible. This suggests the H$α$ emitting region is located at a relatively low latitude and is comparable in size to the stellar disk. Radial velocity modulation implies that the region lies at $\sim19\,R_{\odot}$ from the system's rotation axis, farther out than the stellar limb at $14.4\,R_{\odot}$. Photometric monitoring with the Chuo-university Astronomical Telescope revealed a large low-latitude starspot covering $\sim25\%$ of the surface. These findings are consistent with a scenario in which the flare occurred above the starspot, and the H$α$-emitting plasma was magnetically confined in a loop extending at least $5\,R_{\odot}$ above the stellar surface. From the MAXI data and assuming a radiatively cooling plasma, the electron density and volume are estimated to be $10^{10}$~cm$^{-3}$ and $1 \times 10^{35}$~cm$^3$, respectively. If cubic in shape, this corresponds to $7\,R_{\odot}$, consistent with the H$α$ region height. These results provide direct constraints on the geometry of the plasma and its spatial relationship with the starspot in one of the most energetic stellar flares ever observed.

Wind accretion in binary systems is commonly described using the Bondi-Hoyle-Lyttleton (BHL) formalism. However, its standard implementation fails in the slow-wind regime, where the wind velocity of the donor star ($v_\mathrm{w}$) is comparable to or smaller than the orbital velocity of the accretor ($v_\mathrm{o}$). Tejeda & Toalá recently proposed a geometrical correction to the BHL formalism that accounts for the wind aberration caused by the binary's orbital motion, which tilts the accretion cylinder and reduces its effective cross-section. Here we present a suite of smoothed particle hydrodynamic simulations performed with PHANTOM to test wind accretion in binary systems operating in this slow-wind regime. We explore circular configurations and directly measure mass accretion efficiencies from the simulations. Our results confirm that the standard BHL prescription systematically overestimates accretion rates for $v_\mathrm{w}/v_\mathrm{o} < 1$, while the geometrically corrected model reproduces the simulated efficiencies with remarkable accuracy. A key finding is that the velocity relevant for accretion estimates is not the value derived from the unperturbed stellar wind, but the local gas velocity measured upstream of the accretor. The gravitational potential of the accretor perturbs the flow, altering the effective relative velocity and modifying the accretion efficiency, particularly for compact orbits. These results provide strong numerical support for the geometrically corrected framework and establish a physically motivated basis for modeling wind-fed accretion in interacting binaries, including symbiotic systems.

We investigate the velocity field derived from HI measurements of the irregular galaxy ESO 358-60 using the Velocity Ring Model (VRM) method. This technique, which assumes a coplanar disk, allows us to reconstruct coarse-grained maps of both radial and tangential velocity components from the observed line-of-sight velocity field. Such maps reveal that tangential motions dominate the inner regions, while radial motions become increasingly significant toward the outskirts. This kinematic behavior contrasts with that inferred from the Tilted Ring Model (TRM), which suggests that radial motions are more prominent in the intermediate disk and negligible in the outskirts and detects a pronounced warp of approximately $20^\circ$, with the inner disk nearly edge-on and the outer regions inclined by approximately $60^\circ$. In contrast, the VRM analysis finds that the disk exhibits a bar-like structure in its central regions. This interpretation is further supported by the intensity and velocity dispersion maps. To test the origin of the TRM-derived warp, we construct a toy model based on the TRM results and analyze it with the VRM technique, finding evidence that the warp is likely an artifact arising from the TRM's assumptions. Finally, we estimate the galaxy's mass using both the standard dark matter halo model and a dark matter disk (DMD) model, where all mass lies in the disk plane. The DMD yields a total mass approximately three times lower and provides a slightly better fit to the rotation curve.

We present chemical abundance measurements for 7,302 red giant branch stars within the half-light radius (~5') of $ω$ Centauri ($ω$ Cen), derived from MUSE spectra using the neural network model DD-Payne. DD-Payne effectively identifies spectral features of C, N, and O for [Fe/H]>-1.0 dex; Mg for [Fe/H]>-1.5 dex; and Na, Ca, and Ba for all metallicities. By combining these measurements with previous high-resolution studies, we create the most comprehensive picture of $ω$ Cen's rich chemical evolutionary history. For the first time, we map elemental variations across the entire chromosome diagram, which is widely used to identify multiple populations. We analyze the median chemical abundance trends as functions of age and metallicity for different subpopulations. The DD-Payne measurements of [C/Fe], [N/Fe], and [O/Fe] extend literature trends to higher metallicities and show continuous abundance-metallicity relations, with [(C+N+O)/Fe] increasing steadily with [Fe/H]. [Ca/Fe] and the s-process element [Ba/Fe] also increase with metallicity across all populations. For [Ba/Fe], the chemically enhanced (P2) populations are more enriched than primordial (P1) and the intermediate (Im) populations. Furthermore, [N/Fe] correlates strongly with stellar age while [Ca/Fe] and [Ba/Fe] exhibits a weaker age dependence. Using these abundance-metallicity-age relations, we evaluate different formation scenarios of $ω$ Cen proposed in the literature. Our study demonstrates that combining MUSE with machine learning enables large-sample stellar abundance measurements in crowded cluster cores, overcoming the limitations of fiber-fed spectroscopy for studying multiple stellar populations and their evolutionary histories.

The master student at Lomonosov University in Moscow E.P. Velikhov formulated 1959 the theory of magnetorotational instability, which dominates current astrophysics. A meteoric career made him later the science, nuclear and disarmament advisor to Gorbachev and Yeltsin. This article describes the interactions between Velikhov and the PROMISE team from Potsdam and Dresden-Rossendorf when it came to experimentally testing the theory in the laboratory in the 2000s. At an MHD conference in Catania, Sicily, he offered the public the use of small, transportable Russian nuclear power plants anywhere in the world until the fusion machines currently under development had finally solved the energy problem.

We present data from the ALMA Central Molecular Zone Exploration Survey (ACES) Large Program, which provides broad spectral-line and 3 mm continuum coverage of the Central Molecular Zone (CMZ) at a spatial resolution of 0.1 pc. The survey delivers homogeneous, wide-field mosaics that enable direct comparisons of the physical and chemical conditions across diverse environments in the Galactic center. In this data release paper, we present the CS(2-1), SO(2_3-1_2), CH3CHO(5_1,4-4_1,3), HC3N(11-10), and H40a lines observed simultaneously within two broad spectral windows. These lines reveal pronounced spatial and chemical variations across the CMZ, tracing distinct components of molecular gas, shock-affected regions, and ionized structures. The high angular resolution and multi-line capability of the ACES dataset make it a powerful resource for future studies of gas dynamics, star formation activity, and the physical connection between the CMZ and Sgr A*.

The accretion-ejection activities of black holes play a vital role in shaping the Universe. Bright and recurrent black hole X-ray binaries are ideal objects for studying accretion physics across a wide range of accretion rates, providing insights into the understanding of their supermassive counterparts. This short review summarizes X-ray techniques capable of measuring accretion geometry, our current understanding, and open questions. In particular, X-ray spectroscopic studies indicate that the accretion disk can extend close to the innermost stable circular orbit in the bright hard state. Some hints of disk-corona-jet connections are also discussed.

Cosmic explosions play a critical role in a broad range of astrophysical fields. Although considerable progress has been made to understand the explosive engines and their progenitors, many of the details are not well understood. One of the most powerful electromagnetic probes of the explosive mechanism and the stellar progenitor is the first burst of photons emitted from this blastwave as it exits the stellar photosphere, known as shock breakout (SBO). Our understanding of SBO has evolved considerably in the past decade. Shock heating as the blastwave propagates through the star and circumstellar material can drastically alter this emission producing a much broader range of potential SBO signals than that predicted by standard analytical approaches. Here we present a semi-analytic approach to model this diverse SBO emission, focused on thermal Bremsstrahlung radiation, which more accurately captures the complexities in Nature over previous treatments. We calculate a range of signals for a range of supernova and gamma-ray burst types. Our models demonstrate how we can use these signals to place constraints on the nature of the explosive engines and better understand the role SBO can play in prompt gamma-ray bursts. We study the implications of these results to historic observations, Einstein Probe transients, and in the context of proposed missions. We find that stripped envelope events can be detected serendipitously with survey telescopes, but type Ia and II SBO detections require fast-pointing X-ray observations in response to early warning alerts from gravitational wave or neutrino detectors.

The dynamical effects on the stellar motion produced by the Galactic central bar and the spiral arms perturbations are investigated separately and compared. The stars from the Gaia DR3 catalog are selected in the region of observable completeness, which we estimate as $\sim$1 kpc from the Sun. We apply the 2D model of the Galactic potential consisting of three axisymmetric components, the disk, the bulge, and the dark matter halo, and two non-axisymmetric components, the central bar and the spiral arms. The stellar dynamics is studied using analytical and numerical techniques, such as Hamiltonian topology analysis, the construction of dynamical maps on the representative planes, dynamic spectra, and Poincaré sections. We identify the main dynamical features in the solar neighborhood (SNd), the corotation (CR) and Lindblad resonances (LRs). By assuming that the main moving groups (MGs) in the SNd originate from the resonances, we compare their locations, structures, and intensities with the theoretical predictions and provide a description of the process involved in the formation of the MGs. In addition, we explore parametric planes by adjusting the values of the pattern rotation speed $Ω_{p}$ with the positions of the MGs, for both the spiral arms and bar models, and conclude that the spiral arms model shows better results when compared to those of the bar, under the hypothesis of the dynamical origin of MGs.

We present a detailed analysis of the trajectory of the interstellar comet 3I/ATLAS, focusing on its non-gravitational acceleration (NGA) parameters and their uncertainties. Orbital solutions are computed with models that implement symmetric, time-offset, and asymmetric radial dependence of the outgassing law relative to perihelion. We assess solution robustness through multiple data-selection strategies and comparison with independent determinations. The radial and normal NGA components (A_1 and A_3) are broadly consistent across all configurations, whereas the transverse component (A_2) is more sensitive to data selection, parameter correlations, and orbital phase coverage. Models with asymmetric radial decay slopes marginally improve the fit, but they also introduce additional degeneracies, contributing to systematic uncertainties. The magnitude of the NGA scaled to 1 au constrains the nucleus size of 3I. While our total acceleration estimates agree well with that of JPL's Small-Body Database solution, inclusion of systematic modeling effects implies a significantly larger uncertainty in the inferred radius of 3I.

Galaxy clusters imprint a distinctive signature on the cosmic microwave background through the thermal Sunyaev-Zel'dovich (SZ) effect, which enables to study the intracluster plasma distribution and makes them powerful cosmological probes. We present the first Sardinia Radio Telescope (SRT) detection of the SZ effect in the galaxy cluster MACS J1752+4440 at 18.6 GHz, with a resolution of 0.9'. We detected a decrement in brightness toward the cluster centre, which we attributed to the thermal SZ effect. We modelled the signal using a spherically symmetric $β$ model for the electron density distribution and we employed a Bayesian retrieval to estimate the core radius, central electron density, and $β$ parameter of the cluster. We found values consistent with expectations for a galaxy cluster of the mass of MACS J1752+4440: a core radius of ($160 \pm 30$) kpc, a central electron density of ($2.5^{+0.7}_{-0.5} \cdot 10^{-3}$) cm$^{-3}$, and $β$=$0.6\pm0.1$. The mean Compton-$y$ parameter within a radius of 3.5' is $(2.6 \pm 0.3) \cdot 10^{-5}$, higher than the value reported by Planck, which is coherent considering the different resolution of the instruments and the modelling adopted. This work demonstrates the potential of the SRT to detect the onset of the SZ decrement at low frequencies, providing higher angular resolution than current all-sky surveys and enabling an improved reconstruction of the SZ decrement profile and the plasma distribution in the intracluster medium.

Compact binaries with orbital periods shorter than about 7 days show an absence of transiting planets, a feature known as the ``circumbinary planet desert". The physical mechanism behind this desert remains unclear. We investigate its origin by simulating the long-term dynamics of multi-planet circumbinary systems with evolving inner binaries. Our simulations are based on the single-averaged secular equations that average only over the binary orbital period and fully incorporate planet-planet interactions. When an eccentric binary decays via tides, an outer planet can be captured into resonance advection in eccentricity, a state in which its apsidal precession locks with that of the binary, driving extreme eccentricity growth. While such growth can occur in a binary-single planet system, the parameter space is limited and may not necessarily induce instability. In a multi-planet system, however, the excited orbit inevitably crosses those of its neighbors, which triggers violent planet-planet scatterings and produces collisions or ejections. Crucially, these mutual gravitational interactions amplify the ``localized" instability of a single planet into a system-wide chain reaction, drastically reshaping the orbital architecture and potentially clearing out the inner regions of planetary systems. Our results suggest that the resonance-induced instability provides a natural explanation for the observed circumbinary planet desert.

The redshifted 21-cm signal is a unique probe of the early universe, particularly the Epoch of Reionization (EoR). While the 21-cm power spectrum has been the primary statistic for parameter inference, it fails to capture the non-Gaussian information in the signal, motivating the use of higher-order statistics such as the bispectrum. We perform a rigorous cross-simulation validation to infer the mean neutral hydrogen fraction ($\bar{x}_{HI}$) by training a Bayesian neural network on 21cmFAST simulations and applying it to mock observations generated by ReionYuga code. Our analysis spans six redshifts and includes realistic SKA system noise and cosmic variance, calculated from 50 statistically independent realizations. We find that the bispectrum adds useful information, but the improvement is moderate, with constraints tightened by $\sim 1.4\times$ the power-spectrum only case. The cross-simulation analysis also identifies a persistent systematic discrepancy between inferred and true values that often exceeds the statistical uncertainties, implying that modeling uncertainty remains the dominant limitation. Our results, therefore, indicate that the highly stringent constraints obtained in same-code validation studies may be overly optimistic, and mitigating cross-model systematics is crucial for robust parameter inference in the SKA era.

In view of the current and increasing evidence of a running Hubble constant, we investigate its redshift dependence within the flat $Λ$CDM framework using a 20-bin analysis of the Master SNe~Ia Sample \citep{2025JHEAp..4800405D}, considering cases with and without very low-redshift data. For each case, we obtain best-fitting values of $H_0$ and $Ω_{m0}$, and employ both logarithmic \citep{2025arXiv250902636L} and power-law \citep{2021ApJ...912..150D,2022Galax..10...24D,2025JHEAp..4800405D} parameterizations. The two parameterizations are consistent over the redshift range considered and coincide for low redshifts. To assess their behavior at earlier epochs, we extrapolate both forms to the Cosmic Microwave Background radiation (CMB) era ($z\simeq1100$), Big Bang Nucleosynthesis (BBN, $z\sim10^{9}$), and inflationary scales ($z\sim10^{20}$). The reconstructed Hubble constant remains nearly indistinguishable up to the CMB scale, diverges at the few-to-ten percent level around BBN, and differs more substantially when extrapolated to inflationary redshifts. A qualitative distinction emerges at very-high redshift: the logarithmic form predicts a vanishing of $\mathcal{H}_0^{\mathrm{Log}}(z)$ at finite $z$, while the power-law form, $\mathcal{H}_0^{\mathrm{PL}}(z)$, approaches zero asymptotically as $z \rightarrow \infty$. In future studies, independent high-redshift observations and extensions beyond $Λ$CDM, such as $f(R)$ modified gravity, could allow a comparative study of the two parameterizations beyond the SNe~Ia regime and their high-$z$ physical implications.

Using the TNG50 simulation, we determine observationally motivated metrics that can distinguish quenching galaxies from star forming galaxies for $M_{*} \geqslant 10^{9.5}~M_{\odot}$, based on the spatial distribution of their stellar populations. Quenching galaxies are not fully quenched but have low levels of ongoing star formation that decreases over time. The morphological metrics consider the concentration of star formation, size of the star forming disk, and characteristic radii that trace sharp truncations of star formation. These metrics can separate simulated quenching galaxies based on morphology into populations where star formation is suppressed inside-out and outside-in. Inside-out quenched galaxies are more likely to be the most massive galaxy within their halo in the field, while outside-in quenched galaxies are satellites residing in dense environments and begin quenching ${\sim} 1~\text{Gyr}$ after being accreted. Outside-in quenched galaxies typically take ${\sim} 1.5~\text{Gyr}$ to quench, and inside-out quenched galaxies can take up to ${\sim} 3.5~\text{Gyr}$, where the duration of quenching is a function of stellar mass. We find that each population of quenched galaxy experiences evolution of their morphological metrics, where the different quenched populations reside in unique locations in parameter space. Galaxies in the later stages of quenching are more easily distinguished than those in the early stages, when compared to star forming galaxies. In addition, inside-out quenched galaxies can be distinguished compared to outside-in quenched galaxies, and the progress through the quenching episode can be estimated for both populations. These results have broad implications for distinguishing quenching galaxies in large galaxy surveys.

We investigate the role of the accretion history in shaping the depletion radius of dark matter halos using a large cosmological N-body simulation. We show that the inner depletion radius, rescaled by the virial radius, depends strongly on the recent mass accretion rate (MAR) measured over a dynamical timescale, while exhibiting only weak dependence on halo mass. While this dependence mirrors that of the splashback radius and the two radii are tightly correlated, the depletion radius exhibits a more nuanced response to the detailed accretion mode. Specifically, we find that the dependence on MAR steepens at lower redshifts, aligning with self-similar spherical collapse models yet contrasting with the behavior of the splashback radius. This redshift dependence is largely driven by dynamic events, as it diminishes significantly when halos undergoing recent major mergers are excluded. Furthermore, we identify a dichotomy in the drivers of the depletion radius. For slowly accreting halos, the MAR is the primary dependence, whereas for rapidly accreting halos, other properties (shape, spin, concentration, and formation time of the central subhalo) related to the anisotropic or perturbed accretion mode also play a significant role. These results establish the depletion radius as a sensitive physical probe of the detailed accretion history of dark matter halos, complementary to the splashback radius.

Quantifying the natal kick distribution of pulsars is essential for understanding supernova physics and binary evolution, yet measurements are historically limited by the lack of radial velocity data. Most previous studies rely on transverse velocities under the assumption of spatial isotropy. In this work, we reconstruct the intrinsic three-dimensional (3D) velocity distribution for a curated sample of 18 pulsars by explicitly incorporating the observational constraint of spin-velocity alignment. Using a hierarchical Bayesian framework that accounts for measurement uncertainties, we compare nine candidate velocity distribution models. We find that a Gamma distribution provides an adequate description of the inferred 3D velocities; however, the modest Bayes factor (1.65 relative to a single Maxwellian) indicates that the current data lack sufficient resolving power to discriminate decisively among the models considered. The Gamma model is characterized by a peak velocity of $237^{+67}_{-84}$ km s$^{-1}$. The reconstructed 3D velocities under alignment are systematically lower than those inferred under isotropy, indicating that projection effects can bias individual kick estimates high, while leaving the overall population scale largely unchanged within uncertainties. A complementary analysis of 465 pulsars with transverse velocity estimates favors a Log-Normal distribution for the full sample, while isolated young pulsars remain consistent with a Gamma-like profile. Our results underscore the importance of geometric assumptions in population inference and highlight the need for larger samples with improved distance and spin-axis measurements to place tighter constraints on natal kick physics.

We present spectrally-resolved structural parameter measurements of 29,608 sources from the legacy lensing field of Abell 2744, quantifying global structures from observed $0.7 μm - 4.8 μm$ and spanning rest-frame UV to NIR at $R\sim15$. These measurements are made on imaging mosaics mainly from the UNCOVER/MegaScience survey, including 20 JWST NIRCam broad and medium bands. We perform single-component Sérsic fitting to these galaxies using \texttt{pysersic}, a Bayesian structural fitting tool, to infer their structural parameters and associated random uncertainties from the posterior distributions. Through various quality evaluation criteria, we infer robust structural parameters among $> 90\%$ of the selected $\rm SNR>10$ sources. For each galaxy with reliable sizes in at least two bands and a high-quality redshift, we fit its observed size as a function of wavelength and infer rest-frame UV, optical, and near-infrared sizes where applicable. By performing injection-recovery tests on simulated galaxy cutouts in selected bands, we establish that our structural parameter measurements achieve fractional error $< 10 -20\%$ above $\rm SNR>10$. With this paper, all raw structural measurements and fitted rest-frame sizes are quality-flagged, cataloged, and released to the community. Finally, we demonstrate that this catalog enables the structural study of galaxies over an unprecedentedly wide parameter space of redshift ($0.3<z<8$), stellar mass ($\rm 10^{7}\, M_{\odot}<M_{*} <10^{11.5}\, M_{\odot}$), and rest-frame optical size ($\rm 100 \,pc<R_{e}<10\,kpc$), after correcting for lensing magnification.

Identifying the host galaxies of fast radio bursts (FRBs), and comparing their redshifts and dispersion measures, has unlocked a new probe of the cosmological distribution of ionised gas. However the necessary optical observations to identify FRB hosts, and measure their redshifts, are becoming increasingly onerous as the detection rate of precisely localised FRBs increases. Here we analyse the ability of the Legacy Survey of Space and Time (LSST), being conducted by the Vera C. Rubin Observatory, to identify FRB host galaxies, and the utility of LSST photometric redshifts for FRB cosmology. By combining a model of FRB host galaxy r-band magnitudes, $m_r$, with predictions for the FRB z-DM distribution, we create a method to predict the $m_r(z)$ distribution for the host galaxies of FRBs detected by radio surveys. We then predict these distributions for the coherent modes of the Australian Square Kilometre Array Pathfinder (ASKAP) and MeerKAT. We find that even a single visit with Rubin will be able to identify 65% of FRB host galaxies detected by ASKAP's coherent upgrade, `CRACO'; while the final 10 year co-added images will identify 81% of those from MeerKAT's tied array beams. We also simulate the impact of using photometric redshifts for a simplified analysis to determine $H_0$, finding that estimated photo-z errors result in a decreased precision of only 7% on $H_0$ for ASKAP's CRACO system. The impact of missing dim FRB hosts, which are likely at higher redshifts, is more significant, and might degrade sensitivity to $H_0$ by 47%, or 62% when combined with photo-z errors. All told, Rubin's LSST will be an incredibly powerful survey for facilitating FRB cosmology, although supplemental observations may be useful for particularly dim and distant host galaxies.

The Askaryan Radio Array (ARA) is a neutrino experiment at the South Pole, designed to detect radio-frequency emissions produced by interactions of ultra-high energy (UHE) neutrinos with the Antarctic ice. The array consists of five autonomous stations, each equipped with deep in-ice antennas sensitive to both vertically and horizontally polarized radio signals. With nearly 30 station-years of livetime accumulated, ARA is now conducting its first comprehensive array-wide search for diffuse UHE neutrinos. This analysis is expected to deliver the most stringent constraints from any in-ice radio-based detector up to 1~ZeV and is capable of probing flux levels suggested by KM3NeT around 220~PeV. The results from this analysis marks a critical step toward establishing scalable techniques for next-generation detectors.

The small-scale properties of circumgalactic gas in ordinary galaxies drive its bulk properties: the mass loading of cold neutral gas in galactic outflows affects their bulk momentum; gas cooling processes on small scales affect the spatial distribution of gas in the cool (T~$10^4$K) circumgalactic medium (CGM). However, hydrodynamical simulations have yet to resolve the CGM on such small scales. Spectroscopy remains our primary probe of the small-scale CGM, with which sub-parsec scales are challenging to resolve. Fast radio bursts (FRBs)--microsecond to millisecond duration radio pulses--are temporally broadened ("scattered") by gradients in the electron density transverse to the line of sight, often generated by fluctuations on the smallest spatial scales. This makes FRB scattering a powerful, complementary, and scalable probe of the small-scale CGM. We show that the distribution of scattering timescales introduced by density fluctuations within a single, foreground halo--the tau distribution function, or TDF--is sensitive to the small-scale spatial morphology of the gas. The TDF is readily measurable and is analogous to areal covering factors reported in quasar absorption statistics. We compute the TDF in two regimes: scattering from a turbulent, volume-filling medium ("volumetric scattering") distributed along the line of sight; and scattering from discrete structures localized along the line of sight ("intermittent scattering"). Within these regimes, the TDF is sensitive to whether the cool gas comprises primarily spherical, filamentary (1D), or sheet-like (2D) structures. This work sets the stage for upcoming observations which will use hundreds of sight-lines through nearby halos to probe the small-scale CGM, and points out a novel science case for FRB detectors like MeerKAT, Parkes, FAST, and the DSA-2000, which are exquisitely sensitive over a narrow field of view.

[...] We observed XTE J1810-197 between 29 September 2022 and 14 July 2023 with the radio telescopes at the Argentine Institute of Radioastronomy (IAR). We searched for single pulses in time series at a DM range of 100-400 pc cm-3 , with a threshold in signal-to-noise ratio (S/N) of 8. [...] We found 249 giant pulses at a DM mean value of 178.8$\pm$0.1 pc cm-3 . We measured peak flux densities up to 119 Jy, and fluences up to 58 Jy ms. We fitted a power law distribution to the flux density, obtaining an index of -4.0$\pm$0.3. We observed a maximum rate of approximately 15 pulses per hour on 20 February 2023, followed by an abrupt disappearance of transient radio emission, indicating a transition to a less active state. The brightest single pulses are limited to a $\sim$2$\%$ of the rotational phase and have similar fluence values to the reported intermediate FRB-like bursts of SGR 1935+2154. No significant X-ray activity in the MAXI data was detected during the radio observing period. This is the first study of single radio pulses of a magnetar using IAR data, showing the potential of the upgraded telescopes for investigating the transient radio sky. The properties of the single pulses detected here show the magnetar transient nature and capability to emit high-luminosity pulses. We compared the detected emission to FRB-like bursts and single pulses emitted by SGR 1935+2154. Even though the mechanism producing all the events should be coherent, the luminosity of the events, features on the dynamic spectra, and the difference between being phase confined or not, indicate that XTE J1810-197 presents GP emission, while SGR 1935+2154 only shows normal single pulses or FRB bursts. This could indicate that the conditions for producing each type of event differ.

We have conducted a search for radio pulsars in six Galactic stellar binary systems having unseen primary stars. All six systems have estimated primary masses in the range that could be consistent with neutron stars. We used the Green Bank Telescope at a center frequency of 350 MHz to search for dispersed periodicities and single pulses across a range of possible dispersion measures (DMs) and binary accelerations. No astrophysical signals were detected in our search. The estimated 400-MHz luminosity upper limits from the search are comparable to or smaller than the lowest values observed for almost all the known Galactic binary pulsars having cataloged 400 MHz radio luminosities. This implies that the systems we observed either do not harbor radio-emitting pulsars, contain pulsars that do not beam in our direction, or contain pulsars with luminosities that are significantly lower than this subset of the known Galactic binary pulsar population.

The cold gas content of post-starburst galaxies (PSBs) provides important insight into the mechanisms that drive rapid quenching, but a multiphase assessment of both the atomic and molecular gas in PSBs does not yet exist. We introduce the Ensemble of Multiphase Baryons Evolving in Rapidly-quenching Systems, or EMBERS, a homogeneously selected, nearly mass- and redshift-complete survey of the global atomic (HI) and molecular gas (H2) in PSBs, observed with the Five Hundred-metre Aperture Spherical Telescope (FAST) and the Institut de radioastronomie millimetrique (IRAM) 30m telescope. We present new CO(1-0) observations for 52 PSBs with the IRAM 30m, which, combined with 9 archival observations, gives a total H2 sample of 61, of which 58/61 have ancillary HI measurements. We detect CO(1-0) in 34/61 galaxies, corresponding to molecular gas fractions (fH2 = MH2/M*) ranging from two to 250 per cent. By comparing with a stellar-mass matched star-forming (SF) control sample from xCOLD GASS, we find that PSBs on average are 0.3-0.6 dex depleted in H2. However, considering both HI and H2, individual PSBs host diverse gas reservoirs ranging from gas-rich in both phases, elevated in one phase, or gas-poor, the latter of which is common at lower stellar mass. The existence of gas-normal and gas-depleted PSBs in both phases suggests that some PSBs may rejuvenate their star formation, but the rapid shutdown of star formation in others is likely terminal. Despite this diversity, the majority of EMBERS PSBs are gas-poor compared to SF controls, with the typical PSB hosting gas reservoirs intermediate to those found in star-forming and quenched galaxies.

As part of its commissioning, the Vera C. Rubin Observatory observed several fields repeatedly for a month with ComCam, an instrument that uses the same hardware as the LSST camera but covers a smaller field of view. We photometrically classify AT 2024ahzi, a transient discovered by ComCam, as a Type IIP supernova (SN IIP) using both ComCam and DECam photometry. We find that the duration, luminosity, and color of AT 2024ahzi's photometric plateau are all consistent with those from a large sample of SNe II. By comparing its multi-band light curves to SN II models and analytic relations, we place constraints on the SN progenitor, explosion dynamics, and circumstellar environment. We argue that the progenitor has an extended density profile indistinguishable from a slowly accelerating CSM. We discuss how a similar workflow can identify and characterize future Rubin SNe II.

We use controlled $N$-body experiments to study the dynamical evolution of dwarf spheroidal galaxies (dSphs) embedded in dark-matter (DM) haloes containing a large population of dark subhaloes. We show that stellar orbits subject to stochastic force fluctuations irreversibly gain energy and expand toward a dynamical attractor characterized by a stellar half-light radius $r_{\rm half} \approx r_{\rm max}$ and a velocity dispersion $σ\approx 0.5\,v_{\rm max}$, where $v_{\rm max}$ is the peak circular velocity of the host halo at radius $r_{\rm max}$. This state is reached both in isolation and under tidal stripping, although tidal mass loss significantly accelerates the evolution. Assuming that the Milky Way (MW) dSphs have reached this state, we find that the inferred halo masses collapse onto narrow sequences as a function of $r_{\rm half}$. Under this assumption, MW satellites with $r_{\rm half} \lesssim 1\,\mathrm{kpc}$ follow the tidal tracks of cuspy haloes, while larger systems deviate in a manner consistent with cored DM profiles. Moreover, the mass--luminosity relation follows the slope expected from abundance matching, but with halo masses systematically lowered from their peak values at fixed luminosity. These results suggest that the structural diversity of dSphs is largely an evolutionary outcome driven by internal heating and tides, rather than by the conditions of star formation. This framework predicts that isolated, early-quenched dSphs should have systematically larger sizes than satellites, a prediction testable with upcoming surveys.

We present new Hubble Space Telescope (HST) imaging of ionised filaments in the brightest group galaxy NGC 5044. These filaments extend several kiloparsecs and have widths of $\sim$50--120 pc, with some as narrow as those in cluster cores and others broader, reflecting the lower confining pressure in groups. Filament width ($W$) scales with ambient pressure ($P$) as $W \propto P^{-0.4}$. Combining HST, ALMA, and MUSE data, we measure column densities and magnetic field strengths. Equipartition fields decline from $\sim$40 $μ$G at the centre to $\sim$20 $μ$G at 5 kpc, about 2--3 times weaker than in clusters. Dynamical stability requires stronger radial fields ($\sim$10$^2$ $μ$G), consistent with simulations and magnetic draping, though such high values exceed Faraday Rotation Measure limits. Turbulence and cosmic rays also contribute support. Group and cluster filaments are stable against gravitational collapse, and ultraviolet imaging reveals no star formation in NGC 5044 ($<$10$^{-3}$ M$_\odot$ yr$^{-1}$). NGC 5044 hosts an ionised gas core within its Bondi radius with $n_e \propto r^{-1}$ and filling factor $f \gtrsim 3 \times 10^{-3}$, that is connected to the extended filaments, suggesting a channel for gas inflow toward the black hole. Group and cluster filaments likely share a common origin, with magnetic fields and AGN feedback preserving their structure. Ambient pressure and dust survival regulate molecular gas formation. Lower-pressure groups favour broader, more diffuse filaments with sporadic molecular clumps and weaker dust shielding, whereas higher-pressure clusters host narrower strands with stronger molecular-ionised gas alignment. We predict that (i) filament width scales with ambient pressure, (ii) filament-coincident Faraday rotation structures emerge at $\leq 0.1$ kpc resolution, and (iii) molecular/ionised gas co-spatiality is weaker in groups than in clusters.

Thermal dust continuum polarimetry is a powerful indirect probe of magnetic field geometry in dense molecular clouds while at the same time providing information on the alignment of dust grains with the magnetic field. The leading theory of grain alignment, Radiative Torque Alignment (RAT), has been successful in explaining a variety of observations, including the loss of polarization fraction toward high column densities. One prediction of RAT is that an increase in grain alignment efficiency should be observed in the environments surrounding protostars, due to radiation from the embedded source. However, observational confirmation of this prediction remains scarce. In this study, we sought to test the theoretical prediction of enhanced grain alignment near protostars in the high-mass star forming region DR21 using 214 $μm$ SOFIA/HAWC+ observations. We investigated the correlation of the polarization fraction of dust emission, $p$, and the polarization angular dispersion, $S$, with respect to total intensity. We also probed intrinsic dust polarization properties using the product $S\times p$ as a proxy. We detected significant polarization fractions even at the highest intensities, where strong depolarization is typically expected. The polarization fraction-intensity trend flattens at $I > 1.6^{+0.3}_{-0.3} \times 10^4$ MJy/sr ($N_{H_2}$ $\sim 2\times 10^{23}$ ${cm}^{-2}$). We compared the observed trends with predictions from an analytical model of a centrally heated envelope surrounding an embedded luminous protostar. The predictions from the simple model agree well with the observed trends. Our results provide strong support for enhancement of grain alignment by local radiation from embedded sources.

On the scale of the cosmic horizon, signatures that are unique to general relativity are concealed within the statistics of the large scale distribution of galaxies. These were thought to be beyond the reach of all but the most ambitious galaxy surveys, as they are substantially suppressed relative to standard redshift-space distortions. We show that the detectability of these higher-order relativistic effects can be dramatically enhanced by a sampling strategy that splits a galaxy catalogue into faint and bright subsamples and then combines their auto-bispectra. For current surveys such as DESI, this implies that this new signal will be detectable for the first time using our new strategy.

We measure the star formation history (SFH) of Local Group dwarf galaxy WLM using wide-area ($\sim4$ half-light radii) ground-based NIR imaging of bright ($M_{J}<-4.9$ mag) AGB stars. From our NIR CMD of 825 stars, we find that our recovered SFH is in excellent agreement with literature SFHs of WLM measured from much deeper CMDs ($M_{F090W}\sim+4.3$ mag) based on JWST imaging. We find good agreement in the qualitative shape of the SFHs as well as quantitative metrics such as the timescales for which 50% and 90% of the stellar mass formed with $τ_{50, {\rm AGB}}=5.16_{-0.50}^{+2.07}$ Gyr ago and $τ_{90, {\rm AGB}}=1.33_{-0.09}^{+0.11}$ Gyr ago versus $τ_{50, {\rm JWST}}=5.29_{-0.28}^{+0.34}$ Gyr ago and $τ_{90, {\rm JWST}}=1.42_{-0.01}^{+0.16}$ Gyr ago. The coarser precision of the AGB star-based values is driven by the low number of AGB stars. We also recover an age gradient that is in good agreement with the age gradient measured from JWST data, where we find the outer regions of WLM are on average older than the inner regions. We derive an age-metallicity relation (AMR) from the AGB star CMD fitting that is similar to the JWST-based AMR and is consistent with other reported metallicities in WLM (i.e., spectroscopy, RR Lyrae). From our wide-area AGB star map and SFH, we identify a stellar over-density ($M_*\sim2.0\times10^6~M_{\odot}$, $r_h\sim340$ pc) in WLM's northwestern outer disk. The over-density's SFH shows a burst of star formation $\sim8$ Gyr ago and its spatial location is near a known warp in WLM's H I. Despite WLM having long been considered an isolated galaxy, the mass, size, and age of this over-density are highly suggestive of an accreted dwarf galaxy. Overall, our findings illustrate the power of NIR observations of AGB stars for efficiently and accurately measuring SFHs and for identifying and characterizing substructures in nearby galaxies.

Dwarf galaxies are ideal laboratories to probe the interplay between galaxy formation and the growth of black holes (BHs) in the early Universe. Mounting observational evidence reveals the presence of BHs in low-mass galaxies across cosmic time, with $\textit{JWST}$ uncovering a likely population of $\textit{overmassive}$ BHs at $2 \lesssim z \lesssim 11$. Simulations struggle to reproduce this high-redshift regime, motivating revisions to models of BH accretion and feedback from active galactic nuclei (AGN). To address this, we present high-resolution cosmological zoom-in simulations of a dwarf galaxy based on FABLE physics, introducing novel sink-based BH accretion models and relaxing the fiducial assumption of strong supernova feedback. BHs accrete more efficiently in the sink-based runs compared to the `traditional' Bondi-based counterparts, with AGN feedback leading to early, rapid quenching maintained by fast, hot and metal-enriched outflows. These outflows pollute the outer circumgalactic medium, yielding flat metallicity gradients down to $z=0$. We further assess the performance of two widely used virial estimators and find significant departures from the true dynamical mass, especially during the high-redshift dwarf assembly. Since our galaxy is dark-matter-dominated at all times and radii, BH growth, tied to the baryon cycle, shows no clear correlation with global dynamical properties. Efficient AGN feedback, produced by overmassive BHs relative to extrapolated local $M_\bullet - M_\star$ relations, indicates that dormant BHs residing in local, quenched dwarfs might be the relics of some of the high-redshift $\textit{JWST}$ BHs.

Hot subdwarf variables (sdBVs) that display large-amplitude ($>$1%), short-period variability, as a result of radial-mode pulsations, have recently become objects of interest as they show unique properties among the sdBV classes. Since the discovery of objects such as Balloon 090100001 and CS 1246, twelve more have been discovered in the Zwicky Transient Facility (ZTF) survey that display similar characteristics. However, due to lack of broad spectroscopic investigations, it remains unclear whether these objects constitute a distinct class of radial-mode dominant sdBVs that share common atmospheric and fundamental properties. Here we aim to spectroscopically define these peculiar sdBVs as a population. We collected low-resolution spectroscopy on a sample of sdBVs discovered in the ZTF survey, including time-series observations. We fitted the spectra to a grid of theoretical models to determine their mean effective temperature, surface gravity and helium abundance and any corresponding variability. We then use these properties to estimate the mass, radius and luminosity using a spectral energy distribution fitting method. We show that the resulting properties are similar to the radial-mode dominant sdBVs, Balloon 090100001 and CS 1246, and that they are distinguishable from other similar radial-mode pulsators, such as blue large-amplitude pulsators. We find that these stars, on average, have mean effective temperatures of 28,300 K and surface gravity measurements of $\log\,g=5.56$, with changes in these parameters on the order of 1000 K and 0.10 dex, respectively. The location of these stars on the $T_{\textrm{eff}}$ -- $\log\,g$ plane places them on the boundary region between the low-amplitude, multi-periodic V361 Hya and V1093 Her stars, where the hybrid DW Lyn pulsators lie. The masses and radii of the majority of the sdBVs in our sample align with canonical-mass sdB properties.

We present the results of a multi-wavelength study of a population of X-ray bright ($\rm log(F_{0.2-12 \ keV})>-12.5$), non-$γ$-ray detected high and extreme high synchrotron peak (HSP, EHSP; $\rm log(ν_{\rm peak,\ Hz})>16$) BL Lacs to $i$) put stronger constraints on the synchrotron peak location and shape and $ii$) model their expected behaviour in the very high-energy band. First, we performed an X-ray spectral analysis, using XMM-Newton, Chandra, Swift-XRT, and eROSITA data, and fitting the spectra using both a power law and a log parabola model. Out of 78 sources in the initial sample, 17 were best described by a log parabola model, a result that supports a scenario where the synchrotron peak falls in the X-ray band. Among these 17 sources, we further selected the 10 objects dominated by the jet emission, with no significant contamination of the host galaxy. We performed a $γ$-ray analysis of \lat\ data for these objects, obtaining upper limits providing information on their flux in the 100 MeV - 300 GeV energy range. We then modelled the broadband SED of these objects with JetSeT using two models: one assuming a log parabola for the electron distribution and the other one with a broken power law electron distribution, using parameters consistent with those describing the emission of the prototypical EHSP 1ES 0229+200. We found the models to be generally consistent with the available multi-wavelength detections and upper limits. Furthermore, they confirmed that a subsample of sources could display relevant emission in the TeV energy range, even potentially reaching the threshold for detectability by the Cherenkov Telescope Array Observatory.

Recent observations revealed that distant galaxies have bursty star formation histories, regulated by stellar or active galactic nuclei (AGN) feedback and gas inflows. According to theoretical models, feedback preferentially removes metal-rich gas, while subsequent starbursts are triggered by mergers and newly-accreted gas that is generally less enriched than the galaxy interstellar medium (ISM). Therefore, gas-phase metallicity holds key insights into the baryonic processes shaping early galaxies. We present the first NIRSpec/IFU study of spatially resolved ISM properties in the MACS0647-JD system ($z=10.17$). The system consists of two stellar components detected in NIRSpec/IFU and NIRCam photometry. The main component ($\log (M_{\ast}/M_{\odot}) =7.77 \pm 0.09$; $12+ \log(O/H)=7.89 \pm 0.11$) is more massive and significantly more metal-rich compared to its companion ($\log (M_{\ast}/M_{\odot}) =7.42 \pm 0.07$; $12+ \log(O/H)=7.47 \pm 0.14$), suggesting an older stellar population and a prolonged chemical enrichment history. We find that the H$γ$ line emission centroid is offset by $0.1^{\prime \prime}$ (150 pc in the source plane) from the stellar continuum centroid; the latter coincides with the location of the main stellar component. This offset provides possible evidence of a merger-driven starburst in this system. By comparing the spatial distribution of the metallicity, velocity dispersion, and the burstiness of star formation history, we infer the presence of turbulent, metal-poor gas outside the stellar components. This metal-poor, dynamically unstable gas is likely responsible for the increase in the recent star formation in the north-east region of the system.

We report the results of a radial velocity monitoring program in the Hyades region, carried out at the Center for Astrophysics over a period of more than 45 yr. Nearly 12,000 spectra were gathered for 625 stars brighter than $V \approx 14.5$, of which 55% are members or possible members of the cluster. New or updated spectroscopic orbital solutions are presented for more than 100 members and non-members, including several triple systems. In a few cases we incorporate available astrometry. The frequency of binaries in the Hyades with periods up to $10^4$ days is determined to be $40 \pm 5$%, after corrections for incompleteness. This is marginally higher than in other open clusters. The orbital period and eccentricity distributions are found to be similar to those of solar-type binaries in the field. The mass ratio distribution is essentially flat, or slightly rising toward mass ratios of unity. We revisit the determination of the tidal circularization period, obtaining a longer $P_{\rm circ}$ value of $5.9 \pm 1.1$ days compared to the previous estimate of 3.2 days, still somewhat short of the value expected if most or all of the action of tides happens during the pre-main-sequence phase. We estimate a line-of-sight velocity dispersion of $0.21 \pm 0.05$ km s$^{-1}$ within 5.5 pc of the cluster center (approximately the half-mass radius) and a larger dispersion beyond that distance. Our velocity measurements are accurate enough to clearly reveal the signatures of gravitational redshift and convective blueshift among the dwarfs and giants in the Hyades.

The last evolutionary stages of massive black hole binaries prior to coalescence is dominated by the emission of gravitational waves, which will be probed by the future Laser Interferometer Space Antenna. If gas is present around the two black holes, however, the associated electromagnetic emission can provide additional information about the binary properties and location before the merger event. For this reason, a proper characterisation of the electromagnetic emission during these phases is of fundamental importance, and requires a detailed description of the gas dynamics close to the event horizon of the two black holes, only achievable via numerical simulations. Within this context, we present the implementation of the Superposed Kerr-Schild dynamic metric in the relativistic scheme in the meshless code GIZMO. Our code can now simulate black hole binaries approaching merger with high computational efficiency and accuracy, taking into account relativistic effects on the gas. To validate our implementation, we perform two tests. First, we explore the case of a relativistic Bondi flow around a binary, finding very good agreement with numerical relativity simulations. Then we explore the case of an inviscid relativistic circumbinary disc, comparing our results with a similar simulation run assuming Newtonian gravity. In this second case, we find moderate differences in the mass accretion rate and in the inflow dynamics, which suggest that the presence of a non-Keplerian potential and of apsidal precession in the orbiting gas trajectories may produce stronger shocks and boost angular momentum transport in the disc. Our work highlights the importance of accounting for relativistic corrections in accretion disc simulations around black hole binaries approaching merger, even at scales much larger than those currently probed by numerical relativity simulations.

We provide a complete census of the baryons in early galaxies to investigate the phases in which gas and stars reside, their corresponding budgets, depletion times, and stellar return fraction as a function of redshift and stellar age. We use the ColdSIM hydrodynamical time-dependent non-equilibrium chemistry simulations and perform a detailed analysis of the cold, warm, hot, and stellar phases for both bound structures (galaxies/CGM) and the diffuse IGM. We investigate in depth the cold HI and H2 components, explicitly computed in our simulations, and their relations with host mass, SFR, metallicity and depletion times. We also provide observational insights and discuss the implications for stellar mass functions, PopIII star formation and changes in the IMF. We find that cosmic gas prior to reionisation is mostly cold, while at later epochs the warm phase becomes dominant due to enhanced star formation activity and increasing UV reionising radiation. Stellar return fractions at these times are ~0.15-0.20, a factor of two lower than the values usually adopted. Cold, warm, and hot gas masses as well as HI and H2 components show increasing trends with mass and SFR, while depletion times decrease down to 0.01-0.1 Gyr with a weak metallicity dependence. The resulting star formation efficiency remains at the level of a few per cent and gas-to-star fractions decline with mass, influenced by local feedback and environment. Our findings are consistent with ALMA, VLA and IRAM surveys at later epochs, including ALFALFA, xCOLDGASS, GASS, xGASS, EDGE-CALIFA, PHIBBS, and ASPECS. Gas phases are quantitatively related to the underlying stellar populations and can be used to infer unknown quantities. In the appendix we provide fit functions describing the trends of the stellar return fraction, the main sequence, phase mass relations, gas-to-star fractions and depletion times.

We investigate the origins of quasi-periodic eruptions (QPEs) in galactic nuclei using global three-dimensional meshless finite-mass (MFM) simulations. By modeling stellar and black-hole impactors traversing accretion disks under various inclinations and surface densities, we evaluate their consistency with the observed properties of QPEs. Stellar impacts produce highly asymmetric bipolar ejecta with forward outbursts dominating by over an order of magnitude in energy and luminosity due to the star blocking downstream flow and creating a low-density wake. This shock-compression mechanism often renders backward events unobservable, implying one detectable burst per orbit, and challenging the standard assumption of two bursts. It also fails to explain alternating long-short recurrence patterns and places several sources near or within twice the tidal disruption radius for solar-mass stars, raising severe stability concerns. Whereas a stellar-mass black hole (sBH) gravitationally focuses and heats disk gas extending from its Bondi radius $R_{\rm B}$ to its Hill radius $R_{\rm H}$ during an impact, yielding nearly symmetric ejecta with mild contrasts. This gravitational-drag mechanism generates higher energy budgets at low inclinations due to enhanced mass accumulation. We suggest an ad hoc effective interaction radius $ R_{\rm eff} \simeq 0.5\, R_{\rm B}^{1/3} R_{\rm H}^{2/3} $ to quantify this trend. Our semi-analytical model confirms that sBH-disk collisions can power the full QPE energy range ($10^{44}$-$10^{48}$ erg), naturally accounting for periodicity, asymmetry, durations and diversity.