arXiv Daily Digest - 2026-01-30

Peer review is the main mechanism by which the software engineering community assesses the quality of scientific results. However, the rapid growth of paper submissions in software engineering venues has outpaced the availability of qualified reviewers, creating a growing imbalance that risks constraining and negatively impacting the long-term growth of the Software Engineering (SE) research community. Our vision of the Future of the SE research landscape involves a more scalable, inclusive, and resilient peer review process that incorporates additional mechanisms for: 1) attracting and training newcomers to serve as high-quality reviewers, 2) incentivizing more community members to serve as peer reviewers, and 3) cautiously integrating AI tools to support a high-quality review process.

Conventional supervised climate downscaling struggles to generalize to Global Climate Models (GCMs) due to the lack of paired training data and inherent domain gaps relative to reanalysis. Meanwhile, current zero-shot methods suffer from physical inconsistencies and vanishing gradient issues under large scaling factors. We propose Zero-Shot Statistical Downscaling (ZSSD), a zero-shot framework that performs statistical downscaling without paired data during training. ZSSD leverages a Physics-Consistent Climate Prior learned from reanalysis data, conditioned on geophysical boundaries and temporal information to enforce physical validity. Furthermore, to enable robust inference across varying GCMs, we introduce Unified Coordinate Guidance. This strategy addresses the vanishing gradient problem in vanilla DPS and ensures consistency with large-scale fields. Results show that ZSSD significantly outperforms existing zero-shot baselines in 99th percentile errors and successfully reconstructs complex weather events, such as tropical cyclones, across heterogeneous GCMs.

General-purpose open-domain dense retrieval systems are usually trained with a large, eclectic mix of corpora and search tasks. How should these diverse corpora and tasks be sampled for training? Conventional approaches sample them uniformly, proportional to their instance population sizes, or depend on human-level expert supervision. It is well known that the training data sampling strategy can greatly impact model performance. However, how to find the optimal strategy has not been adequately studied in the context of embedding models. We propose Inf-DDS, a novel reinforcement learning driven sampling framework that adaptively reweighs training datasets guided by influence-based reward signals and is much more lightweight with respect to GPU consumption. Our technique iteratively refines the sampling policy, prioritizing datasets that maximize model performance on a target development set. We evaluate the efficacy of our sampling strategy on a wide range of text retrieval tasks, demonstrating strong improvements in retrieval performance and better adaptation compared to existing gradient-based sampling methods, while also being 1.5x to 4x cheaper in GPU compute. Our sampling strategy achieves a 5.03 absolute NDCG@10 improvement while training a multilingual bge-m3 model and an absolute NDCG@10 improvement of 0.94 while training all-MiniLM-L6-v2, even when starting from expert-assigned weights on a large pool of training datasets.

Serving Large Language Models (LLMs) under mixed workloads--short, latency-sensitive interactive queries alongside long, throughput-oriented batch requests--poses a fundamental scheduling challenge. Standard First-Come, First-Served (FCFS) policies suffer from severe head-of-line blocking, leading to high tail latency and underutilized hardware. We introduce EWSJF (Effective Workload-based Shortest Job First), an adaptive request-level scheduler that learns workload structure in real time to jointly improve fairness and throughput. EWSJF operates upstream of execution-level schedulers and integrates four components: (1) Refine-and-Prune, an unsupervised partitioning algorithm that discovers performance-homogeneous request groups; (2) Dynamic Queue Routing for assigning requests to these groups; (3) Density-Weighted Scoring, a context-aware prioritization function balancing urgency and fairness; and (4) Bayesian Meta-Optimization, which continuously tunes scoring and partitioning parameters based on live performance feedback. Implemented in vLLM, EWSJF improves end-to-end throughput by over 30% and reduces average Time-To-First-Token for short requests by up to 4x compared to FCFS. These results demonstrate that adaptive, learning-based request scheduling is a critical missing layer for efficient and responsive LLM serving. Implementation available at https://anonymous.4open.science/r/vllm_0110-32D8.

Legacy programming languages such as FORTRAN 77 still play a vital role in many industrial applications. Maintaining and modernizing these languages is challenging, especially when migrating to newer standards such as Fortran 2008. This is exacerbated in the presence of legacy proprietary extensions on such legacy languages, because their semantics are often based on old context (limits of legacy language, domain logic,...). This paper presents an approach for automatically migrating FORTRAN 77 with a proprietary extension, named Esope, to Fortran 2008. We introduce a tool that converts Esope source code to Fortran 2008. While supporting readability of the generated code, we want to maintain the level of abstraction provided by Esope. Our method uses model-driven engineering techniques, with transformations to generate a target model from which we export easy-to-read Fortran 2008 source code. We discuss the advantages, limitations, and maintainability considerations of our approach and provide insights into its scalability and adaptability to evolving requirements.

While Large Language Models (LLMs) excel in language-based agentic tasks, their applicability to unseen, nonlinguistic environments (e.g., symbolic or spatial tasks) remains limited. Previous work attributes this performance gap to the mismatch between the pretraining distribution and the testing distribution. In this work, we demonstrate the primary bottleneck is the prohibitive cost of exploration: mastering these tasks requires extensive trial-and-error, which is computationally unsustainable for parameter-heavy LLMs operating in a high dimensional semantic space. To address this, we propose SCOUT (Sub-Scale Collaboration On Unseen Tasks), a novel framework that decouples exploration from exploitation. We employ lightweight "scouts" (e.g., small MLPs) to probe environmental dynamics at a speed and scale far exceeding LLMs. The collected trajectories are utilized to bootstrap the LLM via Supervised Fine-Tuning (SFT), followed by multi-turn Reinforcement Learning (RL) to activate its latent world knowledge. Empirically, SCOUT enables a Qwen2.5-3B-Instruct model to achieve an average score of 0.86, significantly outperforming proprietary models, including Gemini-2.5-Pro (0.60), while saving about 60% GPU hours consumption.

Training large-scale neural networks requires solving nonconvex optimization where the choice of optimizer fundamentally determines both convergence behavior and computational efficiency. While adaptive methods like Adam have long dominated practice, the recently proposed Muon optimizer achieves superior performance through orthogonalized momentum updates that enforce isotropic geometry with uniform singular values. However, this strict isotropy discards potentially valuable curvature information encoded in gradient spectra, motivating optimization methods that balance geometric structure with adaptivity. We introduce FISMO (Fisher-Structured Momentum-Orthogonalized) optimizer, which generalizes isotropic updates to incorporate anisotropic curvature information through Fisher information geometry. By reformulating the optimizer update as a trust-region problem constrained by a Kronecker-factored Fisher metric, FISMO achieves structured preconditioning that adapts to local loss landscape geometry while maintaining computational tractability. We establish convergence guarantees for FISMO in stochastic nonconvex settings, proving an $\mathcal{O}(1/\sqrt{T})$ rate for the expected squared gradient norm with explicit characterization of variance reduction through mini-batching. Empirical evaluation on image classification and language modeling benchmarks demonstrates that FISMO achieves superior training efficiency and final performance compared to established baselines.

Sepsis remains one of the most complex and heterogeneous syndromes in intensive care, characterized by diverse physiological trajectories and variable responses to treatment. While deep learning models perform well in the early prediction of sepsis, they often lack interpretability and ignore latent patient sub-phenotypes. In this work, we propose a machine learning framework by opening up a new avenue for addressing this issue: a relational approach. Temporal data from electronic medical records (EMRs) are viewed as multivariate patient logs and represented in a relational data schema. Then, a propositionalisation technique (based on classic aggregation/selection functions from the field of relational data) is applied to construct interpretable features to "flatten" the data. Finally, the flattened data is classified using a selective naive Bayesian classifier. Experimental validation demonstrates the relevance of the suggested approach as well as its extreme interpretability. The interpretation is fourfold: univariate, global, local, and counterfactual.

Contrastive Decoding (CD) enhances the generation quality of large language models (LLMs) but incurs significant additional computational overhead due to the need for an auxiliary model. Existing internal self-contrastive decoding methods, such as Decoding by Contrasting Layers (DoLa), focus on discrepancies across different layers, which are notably unstable on small-scale models. In this work, based on the observation that LLMs exhibit local preferences, we propose a novel contrastive guidance strategy along the temporal dimension, namely Temporal Guidance (TeGu). Our method ingeniously leverages Multi-Token Prediction (MTP) to construct weaker amateur predictions for model self-contrast. To standardize the implementation of this mechanism, we further introduce a lightweight Conditional MTP Projector (cMTPP), which avoids maintaining multiple independent networks as required by other MTP modules. Across various model series and benchmarks, TeGu achieves significant performance improvements while maintaining low additional memory consumption and computational overhead.

Individual agents in multi-agent (MA) systems often lack robustness, tending to blindly conform to misleading peers. We show this weakness stems from both sycophancy and inadequate ability to evaluate peer reliability. To address this, we first formalize the learning problem of history-aware reference, introducing the historical interactions of peers as additional input, so that agents can estimate peer reliability and learn from trustworthy peers when uncertain. This shifts the task from evaluating peer reasoning quality to estimating peer reliability based on interaction history. We then develop Epistemic Context Learning (ECL): a reasoning framework that conditions predictions on explicitly-built peer profiles from history. We further optimize ECL by reinforcement learning using auxiliary rewards. Our experiments reveal that our ECL enables small models like Qwen 3-4B to outperform a history-agnostic baseline 8x its size (Qwen 3-30B) by accurately identifying reliable peers. ECL also boosts frontier models to near-perfect (100%) performance. We show that ECL generalizes well to various MA configurations and we find that trust is modeled well by LLMs, revealing a strong correlation in trust modeling accuracy and final answer quality.

Adam has been at the core of large-scale training for almost a decade, yet a simple empirical fact remains unaccounted for: both validation scores and the qualitative behaviour of the training runs improve when the momentum parameters satisfy $β_{1}=β_{2}$. Some recent studies have reported this pattern, but there is still no explanation for why this choice helps. We show that this choice is closely tied to a structural property that we refer to as \textit{gradient scale invariance}. We formalize this notion and prove that Adam becomes gradient scale invariant of first order if and only if $β_{1}=β_{2}$. This perspective places the balanced regime of Adam in direct alignment with the design principles underlying several recent optimizers that explicitly enforce scale-robust updates. The theory is supported by experiments across vision and language tasks, and across different architectural families, in which rescaling the gradient has a markedly smoother effect on the update when $β_{1}=β_{2}$. Overall, our results offer a coherent explanation for an open question in the behavior of Adam and provide a simple principle that helps guide the design of future optimizers.

Evaluation of Image Quality Assessment (IQA) models has long been dominated by global correlation metrics, such as Pearson Linear Correlation Coefficient (PLCC) and Spearman Rank-Order Correlation Coefficient (SRCC). While widely adopted, these metrics reduce performance to a single scalar, failing to capture how ranking consistency varies across the local quality spectrum. For example, two IQA models may achieve identical SRCC values, yet one ranks high-quality images (related to high Mean Opinion Score, MOS) more reliably, while the other better discriminates image pairs with small quality/MOS differences (related to $|Δ$MOS$|$). Such complementary behaviors are invisible under global metrics. Moreover, SRCC and PLCC are sensitive to test-sample quality distributions, yielding unstable comparisons across test sets. To address these limitations, we propose \textbf{Granularity-Modulated Correlation (GMC)}, which provides a structured, fine-grained analysis of IQA performance. GMC includes: (1) a \textbf{Granularity Modulator} that applies Gaussian-weighted correlations conditioned on absolute MOS values and pairwise MOS differences ($|Δ$MOS$|$) to examine local performance variations, and (2) a \textbf{Distribution Regulator} that regularizes correlations to mitigate biases from non-uniform quality distributions. The resulting \textbf{correlation surface} maps correlation values as a joint function of MOS and $|Δ$MOS$|$, providing a 3D representation of IQA performance. Experiments on standard benchmarks show that GMC reveals performance characteristics invisible to scalar metrics, offering a more informative and reliable paradigm for analyzing, comparing, and deploying IQA models. Codes are available at https://github.com/Dniaaa/GMC.

Computing-in-Memory (CIM) accelerators are a promising solution for accelerating Machine Learning (ML) workloads, as they perform Matrix-Vector Multiplications (MVMs) on crossbar arrays directly in memory. Although the bit widths of the crossbar inputs and cells are very limited, most CIM compilers do not support quantization below 8 bit. As a result, a single MVM requires many compute cycles, and weights cannot be efficiently stored in a single crossbar cell. To address this problem, we propose a mixed-precision training and compilation framework for CIM architectures. The biggest challenge is the massive search space, that makes it difficult to find good quantization parameters. This is why we introduce a reinforcement learning-based strategy to find suitable quantization configurations that balance latency and accuracy. In the best case, our approach achieves up to a 2.48x speedup over existing state-of-the-art solutions, with an accuracy loss of only 0.086 %.

Large Language Models (LLMs) are increasingly used for question answering over scientific research papers. Existing retrieval augmentation methods often rely on isolated text chunks or concepts, but overlook deeper semantic connections between papers. This impairs the LLM's comprehension of scientific literature, hindering the comprehensiveness and specificity of its responses. To address this, we propose Central Entity-Guided Graph Optimization for Community Detection (CE-GOCD), a method that augments LLMs' scientific question answering by explicitly modeling and leveraging semantic substructures within academic knowledge graphs. Our approach operates by: (1) leveraging paper titles as central entities for targeted subgraph retrieval, (2) enhancing implicit semantic discovery via subgraph pruning and completion, and (3) applying community detection to distill coherent paper groups with shared themes. We evaluated the proposed method on three NLP literature-based question-answering datasets, and the results demonstrate its superiority over other retrieval-augmented baseline approaches, confirming the effectiveness of our framework.

Prior-Data Fitted Networks (PFNs) enable efficient amortized inference but lack transparent access to their learned priors and kernels. This opacity hinders their use in downstream tasks, such as surrogate-based optimization, that require explicit covariance models. We introduce an interpretability-driven framework for amortized spectral discovery from pre-trained PFNs with decoupled attention. We perform a mechanistic analysis on a trained PFN that identifies attention latent output as the key intermediary, linking observed function data to spectral structure. Building on this insight, we propose decoder architectures that map PFN latents to explicit spectral density estimates and corresponding stationary kernels via Bochner's theorem. We study this pipeline in both single-realization and multi-realization regimes, contextualizing theoretical limits on spectral identifiability and proving consistency when multiple function samples are available. Empirically, the proposed decoders recover complex multi-peak spectral mixtures and produce explicit kernels that support Gaussian process regression with accuracy comparable to PFNs and optimization-based baselines, while requiring only a single forward pass. This yields orders-of-magnitude reductions in inference time compared to optimization-based baselines.

Deep time series models are vulnerable to noisy data ubiquitous in real-world applications. Existing robustness strategies either prune data or rely on costly prior quantification, failing to balance effectiveness and efficiency. In this paper, we introduce DropoutTS, a model-agnostic plugin that shifts the paradigm from "what" to learn to "how much" to learn. DropoutTS employs a Sample-Adaptive Dropout mechanism: leveraging spectral sparsity to efficiently quantify instance-level noise via reconstruction residuals, it dynamically calibrates model learning capacity by mapping noise to adaptive dropout rates - selectively suppressing spurious fluctuations while preserving fine-grained fidelity. Extensive experiments across diverse noise regimes and open benchmarks show DropoutTS consistently boosts superior backbones' performance, delivering advanced robustness with negligible parameter overhead and no architectural modifications. Our code is available at https://github.com/CityMind-Lab/DropoutTS.

Pretraining directly on web-scale corpora is the de facto paradigm for building language models. We study an alternative setting where the model is initially exposed to abstract structured data, as a means to ease the subsequent acquisition of rich semantic knowledge, much like humans learn simple logic and mathematics before higher reasoning. We specifically focus on procedural data, generated by formal languages and other simple algorithms, as such abstract data. We first diagnose the algorithmic skills that different forms of procedural data can improve, often significantly. For example, on context recall (Needle-in-a-haystack), the accuracy jumps from 10 to 98% when pretraining on Dyck sequences (balanced brackets). Second, we study how these gains are reflected in pretraining larger models (up to 1.3B). We find that front-loading as little as 0.1% procedural data significantly outperforms standard pretraining on natural language, code, and informal mathematics (C4, CodeParrot, and DeepMind-Math datasets). Notably, this procedural pretraining enables the models to reach the same loss value with only 55, 67, 86% of the original data. Third, we explore the mechanisms behind and find that procedural pretraining instils non-trivial structure in both attention and MLP layers. The former is particularly important for structured domains (e.g. code), and the latter for language. Finally, we lay a path for combining multiple forms of procedural data. Our results show that procedural pretraining is a simple, lightweight means to improving performance and accelerating language model pretraining, ultimately suggesting the promise of disentangling knowledge acquisition from reasoning in LLMs.

This paper summarizes and consolidates fusion power-plant costing work performed in support of ARPA-E from 2017 through 2024, and documents the evolution of the associated analysis framework from early capital-cost-focused studies to a standards-aligned, auditable costing capability. Early efforts applied ARIES-style cost-scaling relations to generate Nth-of-a-kind (NOAK) estimates and were calibrated through a pilot study with Bechtel and Decysive Systems to benchmark balance-of-plant (BOP) costs and validate plant-level reasonableness from an engineering, procurement, and construction (EPC) perspective. Subsequent work, informed by Lucid Catalyst studies of nuclear cost drivers, expanded the methodology to treat indirect costs explicitly and to evaluate cost-reduction pathways for non-fusion-island systems through design-for-cost practices, modularization, centralized manufacturing, and learning. As ARPA-E's fusion portfolio expanded, these methods were applied across BETHE and GAMOW concepts (and select ALPHA revisits), including enhanced treatment of tritium handling and plant integration supported by Princeton/PPPL expertise. In 2023 the capability was refactored to align with the IAEA-GEN-IV EMWG-EPRI code-of-accounts lineage, while key ARIES-derived scaling relations were replaced by bottom-up subsystem models for dominant fusion cost drivers (e.g., magnets, lasers, power supplies, and power-core components) coupled to physics-informed power balances and engineering-constrained radial builds. These developments were implemented in the spreadsheet-based Fusion Economics code (FECONs) and released as an open-source Python framework (pyFECONs), providing a transparent mapping from subsystem estimates to standardized accounts and a consistent computation of LCOE.

Sustainability reports are critical for ESG assessment, yet greenwashing and vague claims often undermine their reliability. Existing NLP models lack robustness to these practices, typically relying on surface-level patterns that generalize poorly. We propose a parameter-efficient framework that structures LLM latent spaces by combining contrastive learning with an ordinal ranking objective to capture graded distinctions between concrete actions and ambiguous claims. Our approach incorporates gated feature modulation to filter disclosure noise and utilizes MetaGradNorm to stabilize multi-objective optimization. Experiments in cross-category settings demonstrate superior robustness over standard baselines while revealing a trade-off between representational rigidity and generalization.

We introduce the (Wishart) projection mechanism, a randomized map of the form $S \mapsto M f(S)$ with $M \sim W_d(1/r I_d, r)$ and study its differential privacy properties. For vector-valued queries $f$, we prove non-asymptotic DP guarantees without any additive noise, showing that Wishart randomness alone can suffice. For matrix-valued queries, however, we establish a sharp negative result: in the noise-free setting, the mechanism is not DP, and we demonstrate its vulnerability by implementing a near perfect membership inference attack (AUC $> 0.99$). We then analyze a noisy variant and prove privacy amplification due to randomness and low rank projection, in both large- and small-rank regimes, yielding stronger privacy guarantees than additive noise alone. Finally, we show that LoRA-style updates are an instance of the matrix-valued mechanism, implying that LoRA is not inherently private despite its built-in randomness, but that low-rank fine-tuning can be more private than full fine-tuning at the same noise level. Preliminary experiments suggest that tighter accounting enables lower noise and improved accuracy in practice.

Behavior cloning (BC) is a practical offline imitation learning method, but it often fails when expert demonstrations are limited. Recent works have introduced a class of architectures named predictive inverse dynamics models (PIDM) that combine a future state predictor with an inverse dynamics model (IDM). While PIDM often outperforms BC, the reasons behind its benefits remain unclear. In this paper, we provide a theoretical explanation: PIDM introduces a bias-variance tradeoff. While predicting the future state introduces bias, conditioning the IDM on the prediction can significantly reduce variance. We establish conditions on the state predictor bias for PIDM to achieve lower prediction error and higher sample efficiency than BC, with the gap widening when additional data sources are available. We validate the theoretical insights empirically in 2D navigation tasks, where BC requires up to five times (three times on average) more demonstrations than PIDM to reach comparable performance; and in a complex 3D environment in a modern video game with high-dimensional visual inputs and stochastic transitions, where BC requires over 66\% more samples than PIDM.

Character image animation aims to synthesize high-fidelity videos by transferring motion from a driving sequence to a static reference image. Despite recent advancements, existing methods suffer from two fundamental challenges: (1) suboptimal motion injection strategies that lead to a trade-off between identity preservation and motion consistency, manifesting as a "see-saw", and (2) an over-reliance on explicit pose priors (e.g., skeletons), which inadequately capture intricate dynamics and hinder generalization to arbitrary, non-humanoid characters. To address these challenges, we present DreamActor-M2, a universal animation framework that reimagines motion conditioning as an in-context learning problem. Our approach follows a two-stage paradigm. First, we bridge the input modality gap by fusing reference appearance and motion cues into a unified latent space, enabling the model to jointly reason about spatial identity and temporal dynamics by leveraging the generative prior of foundational models. Second, we introduce a self-bootstrapped data synthesis pipeline that curates pseudo cross-identity training pairs, facilitating a seamless transition from pose-dependent control to direct, end-to-end RGB-driven animation. This strategy significantly enhances generalization across diverse characters and motion scenarios. To facilitate comprehensive evaluation, we further introduce AW Bench, a versatile benchmark encompassing a wide spectrum of characters types and motion scenarios. Extensive experiments demonstrate that DreamActor-M2 achieves state-of-the-art performance, delivering superior visual fidelity and robust cross-domain generalization. Project Page: https://grisoon.github.io/DreamActor-M2/

The evolution of Large Language Model (LLM) agents towards System~2 reasoning, characterized by deliberative, high-precision problem-solving, requires maintaining rigorous logical integrity over extended horizons. However, prevalent memory preprocessing paradigms suffer from destructive de-contextualization. By compressing complex sequential dependencies into pre-defined structures (e.g., embeddings or graphs), these methods sever the contextual integrity essential for deep reasoning. To address this, we propose E-mem, a framework shifting from Memory Preprocessing to Episodic Context Reconstruction. Inspired by biological engrams, E-mem employs a heterogeneous hierarchical architecture where multiple assistant agents maintain uncompressed memory contexts, while a central master agent orchestrates global planning. Unlike passive retrieval, our mechanism empowers assistants to locally reason within activated segments, extracting context-aware evidence before aggregation. Evaluations on the LoCoMo benchmark demonstrate that E-mem achieves over 54\% F1, surpassing the state-of-the-art GAM by 7.75\%, while reducing token cost by over 70\%.

Cloth manipulation is a ubiquitous task in everyday life, but it remains an open challenge for robotics. The difficulties in developing cloth manipulation policies are attributed to the high-dimensional state space, complex dynamics, and high propensity to self-occlusion exhibited by fabrics. As analytical methods have not been able to provide robust and general manipulation policies, reinforcement learning (RL) is considered a promising approach to these problems. However, to address the large state space and complex dynamics, data-based methods usually rely on large models and long training times. The resulting computational cost significantly hampers the development and adoption of these methods. Additionally, due to the challenge of robust state estimation, garment manipulation policies often adopt an end-to-end learning approach with workspace images as input. While this approach enables a conceptually straightforward sim-to-real transfer via real-world fine-tuning, it also incurs a significant computational cost by training agents on a highly lossy representation of the environment state. This paper questions this common design choice by exploring an efficient and modular approach to RL for cloth manipulation. We show that, through careful design choices, model size and training time can be significantly reduced when learning in simulation. Furthermore, we demonstrate how the resulting simulation-trained model can be transferred to the real world. We evaluate our approach on the SoftGym benchmark and achieve significant performance improvements over available baselines on our task, while using a substantially smaller model.

Large Language Models (LLMs) have shown remarkable performance on complex reasoning tasks, especially when equipped with long chain-of-thought (CoT) reasoning. However, eliciting long CoT typically requires large-scale reinforcement learning (RL) training, while often leading to overthinking with redundant intermediate steps. To improve learning and reasoning efficiency, while preserving or even enhancing performance, we propose TACLer, a model-tailored curriculum reinforcement learning framework that gradually increases the complexity of the data based on the model's proficiency in multi-stage RL training. TACLer features two core components: (i) tailored curriculum learning that determines what knowledge the model lacks and needs to learn in progressive stages; (ii) a hybrid Thinking/NoThinking reasoning paradigm that balances accuracy and efficiency by enabling or disabling the Thinking mode. Our experiments show that TACLer yields a twofold advantage in learning and reasoning: (i) it reduces computational cost, cutting training compute by over 50% compared to long thinking models and reducing inference token usage by over 42% relative to the base model; and (ii) it improves accuracy by over 9% on the base model, consistently outperforming state-of-the-art Nothinking and Thinking baselines across four math datasets with complex problems.

Attention patterns play a crucial role in both training and inference of large language models (LLMs). Prior works have identified individual patterns such as retrieval heads, sink heads, and diagonal traces, yet these observations remain fragmented and lack a unifying explanation. To bridge this gap, we introduce \textbf{Temporal Attention Pattern Predictability Analysis (TAPPA), a unifying framework that explains diverse attention patterns by analyzing their underlying mathematical formulations} from a temporally continuous perspective. TAPPA both deepens the understanding of attention behavior and guides inference acceleration approaches. Specifically, TAPPA characterizes attention patterns as predictable patterns with clear regularities and unpredictable patterns that appear effectively random. Our analysis further reveals that this distinction can be explained by the degree of query self-similarity along the temporal dimension. Focusing on the predictable patterns, we further provide a detailed mathematical analysis of three representative cases through the joint effect of queries, keys, and Rotary Positional Embeddings (RoPE). We validate TAPPA by applying its insights to KV cache compression and LLM pruning tasks. Across these tasks, a simple metric motivated by TAPPA consistently improves performance over baseline methods. The code is available at https://github.com/MIRALab-USTC/LLM-TAPPA.

Large language models (LLMs) excel across many tasks, yet inference is still dominated by strictly token-by-token autoregression. Existing acceleration methods largely patch this pipeline and miss core human-reading ingredients: content-adaptive foresight, chunk-structure-aware compute allocation, and train--test consistency for preview/skimming. We propose the \textbf{Fovea-Block-Skip Transformer} (FBS), which injects a causal, trainable loop into Transformers via Parafovea-Attention Window (PAW), Chunk-Head (CH), and Skip-Gate (SG). Across diverse benchmarks, FBS improves the quality-efficiency trade-off without increasing parameters, and ablations show the three modules are complementary.

Smart meter data is the foundation for planning and operating the distribution network. Unfortunately, such data are not always available due to privacy regulations. Meanwhile, the collected data may be corrupted due to sensor or transmission failure, or it may not have sufficient resolution for downstream tasks. A wide range of generative tasks is formulated to address these issues, including synthetic data generation, missing data imputation, and super-resolution. Despite the success of machine learning models on these tasks, dedicated models need to be designed and trained for each task, leading to redundancy and inefficiency. In this paper, by recognizing the powerful modeling capability of flow matching models, we propose a new approach to unify diverse smart meter data generative tasks with a single model trained for conditional generation. The proposed flow matching models are trained to generate challenging, high-dimensional time series data, specifically monthly smart meter data at a 15 min resolution. By viewing different generative tasks as distinct forms of partial data observations and injecting them into the generation process, we unify tasks such as imputation and super-resolution with a single model, eliminating the need for re-training. The data generated by our model not only are consistent with the given observations but also remain realistic, showing better performance against interpolation and other machine learning based baselines dedicated to the tasks.

We consider representation misdirection (RM), a class of LLM unlearning methods that achieves forgetting by manipulating the forget-representations, that is, latent representations of forget samples. Despite being important, the roles of target vectors used in RM, however, remain underexplored. Here, we approach and revisit RM through the lens of the linear representation hypothesis. Specifically, if one can somehow identify a one-dimensional representation corresponding to a high-level concept, the linear representation hypothesis enables linear operations on this concept vector within the forget-representation space. Under this view, we hypothesize that, beyond forgetting, machine unlearning elicits controllable side behaviors and stronger side capabilities corresponding to the high-level concept. Our hypothesis is empirically validated across a wide range of tasks, including behavioral control (e.g., controlling unlearned models' truth, sentiment, and refusal) and capability enhancement (e.g., improving unlearned models' in-context learning capability). Our findings reveal that this fairly attractive phenomenon could be either a hidden risk if misused or a mechanism that can be harnessed for developing models that require stronger capabilities and controllable behaviors.

Large Language Models (LLMs) increasingly support culturally sensitive decision making, yet often exhibit misalignment due to skewed pretraining data and the absence of structured value representations. Existing methods can steer outputs, but often lack demographic grounding and treat values as independent, unstructured signals, reducing consistency and interpretability. We propose OG-MAR, an Ontology-Guided Multi-Agent Reasoning framework. OG-MAR summarizes respondent-specific values from the World Values Survey (WVS) and constructs a global cultural ontology by eliciting relations over a fixed taxonomy via competency questions. At inference time, it retrieves ontology-consistent relations and demographically similar profiles to instantiate multiple value-persona agents, whose outputs are synthesized by a judgment agent that enforces ontology consistency and demographic proximity. Experiments on regional social-survey benchmarks across four LLM backbones show that OG-MAR improves cultural alignment and robustness over competitive baselines, while producing more transparent reasoning traces.

While reinforcement learning (RL) has empowered multi-turn reasoning agents with retrieval and tools, existing successes largely depend on extensive on-policy rollouts in high-cost, high-accuracy regimes. Under realistic resource constraints that cannot support large models or dense explorations, however, small language model agents fall into a low-cost, low-accuracy regime, where limited rollout budgets lead to sparse exploration, sparse credit assignment, and unstable training. In this work, we challenge this trade-off and show that small language models can achieve strong multi-hop reasoning under resource constraints. We introduce DAVID-GRPO, a budget-efficient RL framework that (i) stabilizes early learning with minimal supervision, (ii) assigns retrieval credit based on evidence recall, and (iii) improves exploration by resampling truncated near-miss trajectories. Evaluated on agents up to 1.5B parameters trained on only four RTX 3090 GPUs, DAVID-GRPO consistently outperforms prior RL methods designed for large-scale settings on six multi-hop QA benchmarks. These results show that with the right inductive biases, small agents can achieve low training cost with high accuracy.

Curriculum learning changes the order of pre-training data, but it remains unclear whether it changes the learning trajectory or mainly reorders exposure over a fixed trajectory. We train Pythia models (14M-410M parameters) for 300B tokens under three linguistically motivated curricula-Age-of-Acquisition, word frequency, and Verb Variation (VV)-and compare each against Random ordering; at 1B parameters we compare Random and VV. Across orderings, training follows a shared sequence of latent phases, while curricula mainly change within-phase data exposure. In smaller models (up to 160M parameters), Random ordering exhibits higher gradient noise and stronger late-training output-head spectral saturation, alongside lower final accuracy; curricula reduce both effects at matched compute. At larger scales, saturation differences are smaller and curriculum gains shrink. We formalize the link between difficulty pacing and optimization stability in an idealized analysis based on gradient-variance control, and our results point to a practical takeaway: curricula help by stabilizing within-phase optimization rather than by creating new phases.

Transformer-based deep neural networks (DNNs) affected by backdoor attacks and unfairness typically exhibit anomalous attention patterns, leading to over-attend to backdoor triggers or protected attributes. Existing neuron-editing mitigation strategies often struggle to handle such situation and most of them lack flexibility and tend to distort feature representations. Motivated by such over-attention phenomenon and software engineering paradigms such as delta debugging and hot patching, we propose AtPatch, a hot-fix method that dynamically redistributes attention maps during model inference. Specifically, for a given input, AtPatch first extracts the attention map from the model's inference process. Then, it uses a pre-trained detector to identify anomalous columns and replace them with unified benign attention. Then, AtPatch rescales other columns to mitigate the impact of over-attention. Finally, AtPatch returns the redistributed attention map to the model for continued inference. Notably, if the detector does not report any anomalous columns, AtPatch directly returns the original attention map to the model. Unlike existing techniques, AtPatch selectively redistributes the attention map, making it better at preserving the model's original functionality. Furthermore, AtPatch's on-the-fly nature allows it to work without modifying model parameters or retraining, making it better suited for deployed models. We conducted extensive experiments to validate AtPatch. Experimental results show that, compared to existing methods, AtPatch can more effectively mitigate backdoor attacks and unfairness while better preserving the model's original functionality.

Fine-Tuning-as-a-Service (FTaaS) facilitates the customization of Multimodal Large Language Models (MLLMs) but introduces critical backdoor risks via poisoned data. Existing defenses either rely on supervised signals or fail to generalize across diverse trigger types and modalities. In this work, we uncover a universal backdoor fingerprint-attention allocation divergence-where poisoned samples disrupt the balanced attention distribution across three functional components: system instructions, vision inputs, and user textual queries, regardless of trigger morphology. Motivated by this insight, we propose Tri-Component Attention Profiling (TCAP), an unsupervised defense framework to filter backdoor samples. TCAP decomposes cross-modal attention maps into the three components, identifies trigger-responsive attention heads via Gaussian Mixture Model (GMM) statistical profiling, and isolates poisoned samples through EM-based vote aggregation. Extensive experiments across diverse MLLM architectures and attack methods demonstrate that TCAP achieves consistently strong performance, establishing it as a robust and practical backdoor defense in MLLMs.

Model merging efficiently aggregates capabilities from multiple fine-tuned models into a single one, operating purely in parameter space without original data or expensive re-computation. Despite empirical successes, a unified theory for its effectiveness under heterogeneous finetuning hyperparameters (e.g., varying learning rates, batch sizes) remains missing. Moreover, the lack of hyperparameter transparency in open-source fine-tuned models makes it difficult to predict merged-model performance, leaving practitioners without guidance on how to fine-tune merge-friendly experts. To address those two challenges, we employ $L_2$-Stability theory under heterogeneous hyperparameter environments to analyze the generalization of the merged model $\boldsymbol{x}_{avg}$. This pioneering analysis yields two key contributions: (i) \textit{A unified theoretical framework} is provided to explain existing merging algorithms, revealing how they optimize specific terms in our bound, thus offering a strong theoretical foundation for empirical observations. (ii) \textit{Actionable recommendations} are proposed for practitioners to strategically fine-tune expert models, enabling the construction of merge-friendly models within the pretraining-to-finetuning pipeline. Extensive experiments on the ResNet/Vit family across 20/8 visual classification tasks, involving thousands of finetuning models, robustly confirm the impact of different hyperparameters on the generalization of $\boldsymbol{x}_{avg}$ predicted by our theoretical results.

Disentangled representation learning aims to map independent factors of variation to independent representation components. On one hand, purely unsupervised approaches have proven successful on fully disentangled synthetic data, but fail to recover semantic factors from real data without strong inductive biases. On the other hand, supervised approaches are unstable and hard to scale to large attribute sets because they rely on adversarial objectives or auxiliary classifiers. We introduce \textsc{XFactors}, a weakly-supervised VAE framework that disentangles and provides explicit control over a chosen set of factors. Building on the Disentangled Information Bottleneck perspective, we decompose the representation into a residual subspace $\mathcal{S}$ and factor-specific subspaces $\mathcal{T}_1,\ldots,\mathcal{T}_K$ and a residual subspace $\mathcal{S}$. Each target factor is encoded in its assigned $\mathcal{T}_i$ through contrastive supervision: an InfoNCE loss pulls together latents sharing the same factor value and pushes apart mismatched pairs. In parallel, KL regularization imposes a Gaussian structure on both $\mathcal{S}$ and the aggregated factor subspaces, organizing the geometry without additional supervision for non-targeted factors and avoiding adversarial training and classifiers. Across multiple datasets, with constant hyperparameters, \textsc{XFactors} achieves state-of-the-art disentanglement scores and yields consistent qualitative factor alignment in the corresponding subspaces, enabling controlled factor swapping via latent replacement. We further demonstrate that our method scales correctly with increasing latent capacity and evaluate it on the real-world dataset CelebA. Our code is available at \href{https://github.com/ICML26-anon/XFactors}{github.com/ICML26-anon/XFactors}.

Key--value (KV) caching enables fast autoregressive decoding but at long contexts becomes a dominant bottleneck in High Bandwidth Memory (HBM) capacity and bandwidth. A common mitigation is to compress cached keys and values by projecting per-head matrixes to a lower rank, storing only the projections in the HBM. However, existing post-training approaches typically fit these projections using SVD-style proxy objectives, which may poorly reflect end-to-end reconstruction after softmax, value mixing, and subsequent decoder-layer transformations. For these reasons, we introduce StiefAttention, a post-training KV-cache compression method that learns \emph{orthonormal} projection bases by directly minimizing \emph{decoder-layer output reconstruction error}. StiefAttention additionally precomputes, for each layer, an error-rank profile over candidate ranks, enabling flexible layer-wise rank allocation under a user-specified error budget. Noteworthy, on Llama3-8B under the same conditions, StiefAttention outperforms EigenAttention by $11.9$ points on C4 perplexity and $5.4\%$ on 0-shot MMLU accuracy at iso-compression, yielding lower relative error and higher cosine similarity with respect to the original decoder-layer outputs.

Test-Time Scaling enhances the reasoning capabilities of Large Language Models by allocating additional inference compute to broaden the exploration of the solution space. However, existing search strategies typically treat rollouts as disposable samples, where valuable intermediate insights are effectively discarded after each trial. This systemic memorylessness leads to massive computational redundancy, as models repeatedly re-derive discovered conclusions and revisit known dead ends across extensive attempts. To bridge this gap, we propose \textbf{Recycling Search Experience (RSE)}, a self-guided, training-free strategy that turns test-time search from a series of isolated trials into a cumulative process. By actively distilling raw trajectories into a shared experience bank, RSE enables positive recycling of intermediate conclusions to shortcut redundant derivations and negative recycling of failure patterns to prune encountered dead ends. Theoretically, we provide an analysis that formalizes the efficiency gains of RSE, validating its advantage over independent sampling in solving complex reasoning tasks. Empirically, extensive experiments on HMMT24, HMMT25, IMO-Bench, and HLE show that RSE consistently outperforms strong baselines with comparable computational cost, achieving state-of-the-art scaling efficiency.

While end-to-end self-supervised learning with backpropagation (global BP-SSL) has become central for training modern AI systems, theories of local self-supervised learning (local-SSL) have struggled to build functional representations in deep neural networks. To establish a link between global and local rules, we first develop a theory for deep linear networks: we identify conditions for local-SSL algorithms (like Forward-forward or CLAPP) to implement exactly the same weight update as a global BP-SSL. Starting from the theoretical insights, we then develop novel variants of local-SSL algorithms to approximate global BP-SSL in deep non-linear convolutional neural networks. Variants that improve the similarity between gradient updates of local-SSL with those of global BP-SSL also show better performance on image datasets (CIFAR-10, STL-10, and Tiny ImageNet). The best local-SSL rule with the CLAPP loss function matches the performance of a comparable global BP-SSL with InfoNCE or CPC-like loss functions, and improves upon state-of-the-art for local SSL on these benchmarks.

Large language models (LLMs) demonstrate impressive capabilities across diverse tasks but raise concerns about privacy, copyright, and harmful materials. Existing LLM unlearning methods rarely consider the continual and high-volume nature of real-world deletion requests, which can cause utility degradation and catastrophic forgetting as requests accumulate. To address this challenge, we introduce \fit, a framework for continual unlearning that handles large numbers of deletion requests while maintaining robustness against both catastrophic forgetting and post-unlearning recovery. \fit mitigates degradation through rigorous data \underline{F}iltering, \underline{I}mportance-aware updates, and \underline{T}argeted layer attribution, enabling stable performance across long sequences of unlearning operations and achieving a favorable balance between forgetting effectiveness and utility retention. To support realistic evaluation, we present \textbf{PCH}, a benchmark covering \textbf{P}ersonal information, \textbf{C}opyright, and \textbf{H}armful content in sequential deletion scenarios, along with two symmetric metrics, Forget Degree (F.D.) and Retain Utility (R.U.), which jointly assess forgetting quality and utility preservation. Extensive experiments on four open-source LLMs with hundreds of deletion requests show that \fit achieves the strongest trade-off between F.D. and R.U., surpasses existing methods on MMLU, CommonsenseQA, and GSM8K, and remains resistant against both relearning and quantization recovery attacks.

Deep learning has emerged as a promising paradigm for spatio-temporal modeling of fluid dynamics. However, existing approaches often suffer from limited generalization to unseen flow conditions and typically require retraining when applied to new scenarios. In this paper, we present LLM4Fluid, a spatio-temporal prediction framework that leverages Large Language Models (LLMs) as generalizable neural solvers for fluid dynamics. The framework first compresses high-dimensional flow fields into a compact latent space via reduced-order modeling enhanced with a physics-informed disentanglement mechanism, effectively mitigating spatial feature entanglement while preserving essential flow structures. A pretrained LLM then serves as a temporal processor, autoregressively predicting the dynamics of physical sequences with time series prompts. To bridge the modality gap between prompts and physical sequences, which can otherwise degrade prediction accuracy, we propose a dedicated modality alignment strategy that resolves representational mismatch and stabilizes long-term prediction. Extensive experiments across diverse flow scenarios demonstrate that LLM4Fluid functions as a robust and generalizable neural solver without retraining, achieving state-of-the-art accuracy while exhibiting powerful zero-shot and in-context learning capabilities. Code and datasets are publicly available at https://github.com/qisongxiao/LLM4Fluid.

While language progresses through a sequence of semantic states, the underlying dynamics of this progression remain elusive. Here, we treat the semantic progression of written text as a stochastic trajectory in a high-dimensional state space. We utilize Allan deviation, a tool from precision metrology, to analyze the stability of meaning by treating ordered sentence embeddings as a displacement signal. Our analysis reveals two distinct dynamical regimes: short-time power-law scaling, which differentiates creative literature from technical texts, and a long-time crossover to a stability-limited noise floor. We find that while large language models successfully mimic the local scaling statistics of human text, they exhibit a systematic reduction in their stability horizon. These results establish semantic coherence as a measurable physical property, offering a framework to differentiate the nuanced dynamics of human cognition from the patterns generated by algorithmic models.

Multimodal learning aims to integrate complementary information from heterogeneous modalities, yet strong optimization alone does not guaranty well-structured representations. Even under carefully balanced training schemes, multimodal models often exhibit geometric pathologies, including intra-modal representation collapse and sample-level cross-modal inconsistency, which degrade both unimodal robustness and multimodal fusion. We identify representation geometry as a missing control axis in multimodal learning and propose \regName, a lightweight geometry-aware regularization framework. \regName enforces two complementary constraints on intermediate embeddings: an intra-modal dispersive regularization that promotes representation diversity, and an inter-modal anchoring regularization that bounds sample-level cross-modal drift without rigid alignment. The proposed regularizer is plug-and-play, requires no architectural modifications, and is compatible with various training paradigms. Extensive experiments across multiple multimodal benchmarks demonstrate consistent improvements in both multimodal and unimodal performance, showing that explicitly regulating representation geometry effectively mitigates modality trade-offs.

Many reinforcement learning (RL) problems admit multiple terminal solutions of comparable quality, where the goal is not to identify a single optimum but to represent a diverse set of high-quality outcomes. Nevertheless, policies trained by standard expected return maximization routinely collapse onto a small subset of outcomes, a phenomenon commonly attributed to insufficient exploration or weak regularization. We show that this explanation is incomplete: outcome level mode collapse is a structural consequence of the expected-return objective itself. Under idealized learning dynamics, the log-probability ratio between any two outcomes evolves linearly in their reward difference, implying exponential ratio divergence and inevitable collapse independent of the exploration strategy, entropy regularization, or optimization algorithm. We identify the source of this pathology as the probability multiplier inside the expectation and propose a minimal correction: inverse probability scaling, which removes outcome-frequency amplification from the learning signal, fundamentally changes the learning dynamics, and provably yields reward-proportional terminal distributions, preventing collapse in multimodal settings. We instantiate this principle in Group Relative Policy Optimization (GRPO) as a drop-in modification, IPS-GRPO, requiring no auxiliary models or architectural changes. Across different reasoning and molecular generation tasks, IPS-GRPO consistently reduces outcome-level mode collapse while matching or exceeding baseline performance, suggesting that correcting the objective rather than adding exploration heuristics is key to reliable multimodal policy optimization.

Multimodal Large Language Models (MLLMs) are a major focus of recent AI research. However, most prior work focuses on static image understanding, while their ability to process sequential audio-video data remains underexplored. This gap highlights the need for a high-quality benchmark to systematically evaluate MLLM performance in a real-world setting. We introduce SONIC-O1, a comprehensive, fully human-verified benchmark spanning 13 real-world conversational domains with 4,958 annotations and demographic metadata. SONIC-O1 evaluates MLLMs on key tasks, including open-ended summarization, multiple-choice question (MCQ) answering, and temporal localization with supporting rationales (reasoning). Experiments on closed- and open-source models reveal limitations. While the performance gap in MCQ accuracy between two model families is relatively small, we observe a substantial 22.6% performance difference in temporal localization between the best performing closed-source and open-source models. Performance further degrades across demographic groups, indicating persistent disparities in model behavior. Overall, SONIC-O1 provides an open evaluation suite for temporally grounded and socially robust multimodal understanding. We release SONIC-O1 for reproducibility and research: Project page: https://vectorinstitute.github.io/sonic-o1/ Dataset: https://huggingface.co/datasets/vector-institute/sonic-o1 Github: https://github.com/vectorinstitute/sonic-o1 Leaderboard: https://huggingface.co/spaces/vector-institute/sonic-o1-leaderboard

Subword tokenization methods, such as Byte-Pair Encoding (BPE), significantly impact the performance and efficiency of large language models (LLMs). The standard approach involves training a general-purpose tokenizer that uniformly processes all textual data during both training and inference. However, the use of a generic set of tokens can incur inefficiencies when applying the model to specific domains or languages. To address this limitation, we propose a post-training adaptation strategy that selectively replaces low-utility tokens with more relevant ones based on their frequency in an adaptation corpus. Our algorithm identifies the token inventory that most effectively encodes the adaptation corpus for a given target vocabulary size. Extensive experiments on generation and classification tasks across multiple languages demonstrate that our adapted tokenizers compress test corpora more effectively than baselines using the same vocabulary size. This method serves as a lightweight adaptation mechanism, akin to a vocabulary fine-tuning process, enabling optimized tokenization for specific domains or tasks. Our code and data are available at https://github.com/vijini/Adapt-BPE.git.

Bridging the gap between data-rich training regimes and observation-sparse deployment conditions remains a central challenge in spatiotemporal field reconstruction, particularly when target domains exhibit distributional shifts, heterogeneous structure, and multi-scale dynamics absent from available training data. We present SENDAI, a hierarchical Sparse-measurement, EfficieNt Data AssImilation Framework that reconstructs full spatial states from hyper sparse sensor observations by combining simulation-derived priors with learned discrepancy corrections. We demonstrate the performance on satellite remote sensing, reconstructing MODIS (Moderate Resolution Imaging Spectroradiometer) derived vegetation index fields across six globally distributed sites. Using seasonal periods as a proxy for domain shift, the framework consistently outperforms established baselines that require substantially denser observations -- SENDAI achieves a maximum SSIM improvement of 185% over traditional baselines and a 36% improvement over recent high-frequency-based methods. These gains are particularly pronounced for landscapes with sharp boundaries and sub-seasonal dynamics; more importantly, the framework effectively preserves diagnostically relevant structures -- such as field topologies, land cover discontinuities, and spatial gradients. By yielding corrections that are more structurally and spectrally separable, the reconstructed fields are better suited for downstream inference of indirectly observed variables. The results therefore highlight a lightweight and operationally viable framework for sparse-measurement reconstruction that is applicable to physically grounded inference, resource-limited deployment, and real-time monitor and control.

Vision-Language Models (VLMs) are typically deterministic in nature and lack intrinsic mechanisms to quantify epistemic uncertainty, which reflects the model's lack of knowledge or ignorance of its own representations. We theoretically motivate negative log-density of an embedding as a proxy for the epistemic uncertainty, where low-density regions signify model ignorance. The proposed method REPVLM computes the probability density on the hyperspherical manifold of the VLM embeddings using Riemannian Flow Matching. We empirically demonstrate that REPVLM achieves near-perfect correlation between uncertainty and prediction error, significantly outperforming existing baselines. Beyond classification, we also demonstrate that the model also provides a scalable metric for out-of-distribution detection and automated data curation.

Clustering tabular data is a fundamental yet challenging problem due to heterogeneous feature types, diverse data-generating mechanisms, and the absence of transferable inductive biases across datasets. Prior-fitted networks (PFNs) have recently demonstrated strong generalization in supervised tabular learning by amortizing Bayesian inference under a broad synthetic prior. Extending this paradigm to clustering is nontrivial: clustering is unsupervised, admits a combinatorial and permutation-invariant output space, and requires inferring the number of clusters. We introduce TabClustPFN, a prior-fitted network for tabular data clustering that performs amortized Bayesian inference over both cluster assignments and cluster cardinality. Pretrained on synthetic datasets drawn from a flexible clustering prior, TabClustPFN clusters unseen datasets in a single forward pass, without dataset-specific retraining or hyperparameter tuning. The model naturally handles heterogeneous numerical and categorical features and adapts to a wide range of clustering structures. Experiments on synthetic data and curated real-world tabular benchmarks show that TabClustPFN outperforms classical, deep, and amortized clustering baselines, while exhibiting strong robustness in out-of-the-box exploratory settings. Code is available at https://github.com/Tianqi-Zhao/TabClustPFN.

Tool-augmented large language models have advanced from single-turn question answering to deep research workflows that iteratively plan queries, invoke external tools, and synthesize information to address complex information needs. Evaluating such workflows presents a fundamental challenge: reliance on live APIs introduces non-determinism, as tool invocations may yield different results across runs due to temporal drift, rate limiting, and evolving backend states. This variance undermines reproducibility and invalidates cross-system comparisons. We present ScholarGym, a simulation environment for reproducible evaluation of deep research workflows on academic literature. The environment decouples workflow components into query planning, tool invocation, and relevance assessment, enabling fine-grained analysis of each stage under controlled conditions. Built on a static corpus of 570K papers with deterministic retrieval, ScholarGym provides 2,536 queries with expert-annotated ground truth. Experiments across diverse backbone models reveal how reasoning capabilities, planning strategies, and selection mechanisms interact over iterative refinement.

Deep networks learn internal representations whose geometry--how features bend, rotate, and evolve--affects both generalization and robustness. Existing similarity measures such as CKA or SVCCA capture pointwise overlap between activation sets, but miss how representations change along input paths. Two models may appear nearly identical under these metrics yet respond very differently to perturbations or adversarial stress. We introduce representation holonomy, a gauge-invariant statistic that measures this path dependence. Conceptually, holonomy quantifies the "twist" accumulated when features are parallel-transported around a small loop in input space: flat representations yield zero holonomy, while nonzero values reveal hidden curvature. Our estimator fixes gauge through global whitening, aligns neighborhoods using shared subspaces and rotation-only Procrustes, and embeds the result back to the full feature space. We prove invariance to orthogonal (and affine, post-whitening) transformations, establish a linear null for affine layers, and show that holonomy vanishes at small radii. Empirically, holonomy increases with loop radius, separates models that appear similar under CKA, and correlates with adversarial and corruption robustness. It also tracks training dynamics as features form and stabilize. Together, these results position representation holonomy as a practical and scalable diagnostic for probing the geometric structure of learned representations beyond pointwise similarity.

AI consumer markets are characterized by severe buyer-supplier market asymmetries. Complex AI systems can appear highly accurate while making costly errors or embedding hidden defects. While there have been regulatory efforts surrounding different forms of disclosure, large information gaps remain. This paper provides the first experimental evidence on the important role of information asymmetries and disclosure designs in shaping user adoption of AI systems. We systematically vary the density of low-quality AI systems and the depth of disclosure requirements in a simulated AI product market to gauge how people react to the risk of accidentally relying on a low-quality AI system. Then, we compare participants' choices to a rational Bayesian model, analyzing the degree to which partial information disclosure can improve AI adoption. Our results underscore the deleterious effects of information asymmetries on AI adoption, but also highlight the potential of partial disclosure designs to improve the overall efficiency of human decision-making.

The deployment of coding agents in privacy-sensitive and resource-constrained environments drives the demand for capable open-weight Small Language Models (SLMs). However, they suffer from a fundamental capability gap: unlike frontier large models, they lack the inference-time strong generalization to work with complicated, unfamiliar codebases. We identify that the prevailing Task-Centric Learning (TCL) paradigm, which scales exposure across disparate repositories, fails to address this limitation. In response, we propose Repository-Centric Learning (RCL), a paradigm shift that prioritizes vertical repository depth over horizontal task breadth, suggesting SLMs must internalize the "physics" of a target software environment through parametric knowledge acquisition, rather than attempting to recover it via costly inference-time search. Following this new paradigm, we design a four-unit Repository-Centric Experience, transforming static codebases into interactive learning signals, to train SWE-Spot-4B, a family of highly compact models built as repo-specialized experts that breaks established scaling trends, outperforming open-weight models up to larger (e.g., CWM by Meta, Qwen3-Coder-30B) and surpassing/matching efficiency-focused commercial models (e.g., GPT-4.1-mini, GPT-5-nano) across multiple SWE tasks. Further analysis reveals that RCL yields higher training sample efficiency and lower inference costs, emphasizing that for building efficient intelligence, repository mastery is a distinct and necessary dimension that complements general coding capability.

Discrete Diffusion Language Models (DLMs) offer a promising non-autoregressive alternative for text generation, yet effective mechanisms for inference-time control remain relatively underexplored. Existing approaches include sampling-level guidance procedures or trajectory optimization mechanisms. In this work, we introduce Iterative Latent Representation Refinement (ILRR), a learning-free framework for steering DLMs using a single reference sequence. ILRR guides generation by dynamically aligning the internal activations of the generated sequence with those of a given reference throughout the denoising process. This approach captures and transfers high-level semantic properties, with a tunable steering scale enabling flexible control over attributes such as sentiment. We further introduce Spatially Modulated Steering, an extension that enables steering long texts using shorter references by regulating guidance intensity across the sequence. Empirically, we demonstrate that ILRR achieves effective attribute steering on LLaDA and MDLM architectures with a minor computational overhead, requiring only one additional parallel forward pass per denoising step. Under the same compute budget, ILRR improves attribute accuracy over comparable baselines by 10$\%$ to 60$\%$ points, while maintaining high generation quality.

We investigate the relation between end-to-end equivariance and layerwise equivariance in deep neural networks. We prove the following: For a network whose end-to-end function is equivariant with respect to group actions on the input and output spaces, there is a parameter choice yielding the same end-to-end function such that its layers are equivariant with respect to some group actions on the latent spaces. Our result assumes that the parameters of the model are identifiable in an appropriate sense. This identifiability property has been established in the literature for a large class of networks, to which our results apply immediately, while it is conjectural for others. The theory we develop is grounded in an abstract formalism, and is therefore architecture-agnostic. Overall, our results provide a mathematical explanation for the emergence of equivariant structures in the weights of neural networks during training -- a phenomenon that is consistently observed in practice.

Transformer-based models have recently made significant advances in accurate time-series forecasting, but even these architectures struggle to scale efficiently while capturing long-term temporal dynamics. Mixture-of-Experts (MoE) layers are a proven solution to scaling problems in natural language processing. However, existing MoE approaches for time-series forecasting rely on token-wise routing mechanisms, which may fail to exploit the natural locality and continuity of temporal data. In this work, we introduce Seg-MoE, a sparse MoE design that routes and processes contiguous time-step segments rather than making independent expert decisions. Token segments allow each expert to model intra-segment interactions directly, naturally aligning with inherent temporal patterns. We integrate Seg-MoE layers into a time-series Transformer and evaluate it on multiple multivariate long-term forecasting benchmarks. Seg-MoE consistently achieves state-of-the-art forecasting accuracy across almost all prediction horizons, outperforming both dense Transformers and prior token-wise MoE models. Comprehensive ablation studies confirm that segment-level routing is the key factor driving these gains. Our results show that aligning the MoE routing granularity with the inherent structure of time series provides a powerful, yet previously underexplored, inductive bias, opening new avenues for conditionally sparse architectures in sequential data modeling.

In this paper, we explore the use of generative artificial intelligence (GenAI) for ship propeller design. While traditional forward machine learning models predict the performance of mechanical components based on given design parameters, GenAI models aim to generate designs that achieve specified performance targets. In particular, we employ conditional flow matching to establish a bidirectional mapping between design parameters and simulated noise that is conditioned on performance labels. This approach enables the generation of multiple valid designs corresponding to the same performance targets by sampling over the noise vector. To support model training, we generate data using a vortex lattice method for numerical simulation and analyze the trade-off between model accuracy and the amount of available data. We further propose data augmentation using pseudo-labels derived from less data-intensive forward surrogate models, which can often improve overall model performance. Finally, we present examples of distinct propeller geometries that exhibit nearly identical performance characteristics, illustrating the versatility and potential of GenAI in engineering design.

We study privacy amplification for differentially private model training with matrix factorization under random allocation (also known as the balls-in-bins model). Recent work by Choquette-Choo et al. (2025) proposes a sampling-based Monte Carlo approach to compute amplification parameters in this setting. However, their guarantees either only hold with some high probability or require random abstention by the mechanism. Furthermore, the required number of samples for ensuring $(ε,δ)$-DP is inversely proportional to $δ$. In contrast, we develop sampling-free bounds based on Rényi divergence and conditional composition. The former is facilitated by a dynamic programming formulation to efficiently compute the bounds. The latter complements it by offering stronger privacy guarantees for small $ε$, where Rényi divergence bounds inherently lead to an over-approximation. Our framework applies to arbitrary banded and non-banded matrices. Through numerical comparisons, we demonstrate the efficacy of our approach across a broad range of matrix mechanisms used in research and practice.

Many quantum software development kits provide a suite of circuit optimisation passes. These passes have been highly optimised and tested in isolation. However, the order in which they are applied is left to the user, or else defined in general-purpose default pass sequences. While general-purpose sequences miss opportunities for optimisation which are particular to individual circuits, designing pass sequences bespoke to particular circuits requires exceptional knowledge about quantum circuit design and optimisation. Here we propose and demonstrate training a reinforcement learning agent to compose optimisation-pass sequences. In particular the agent's action space consists of passes for two-qubit gate count reduction used in default PyTKET pass sequences. For the circuits in our diverse test set, the (mean, median) fraction of two-qubit gates removed by the agent is $(57.7\%, \ 56.7 \%)$, compared to $(41.8 \%, \ 50.0 \%)$ for the next best default pass sequence.

Diffusion models have achieved remarkable progress in image generation, but their increasing deployment raises serious concerns about privacy. In particular, fine-tuned models are highly vulnerable, as they are often fine-tuned on small and private datasets. Membership inference attacks (MIAs) are used to assess privacy risks by determining whether a specific sample was part of a model's training data. Existing MIAs against diffusion models either assume obtaining the intermediate results or require auxiliary datasets for training the shadow model. In this work, we utilized a critical yet overlooked vulnerability: the widely used noise schedules fail to fully eliminate semantic information in the images, resulting in residual semantic signals even at the maximum noise step. We empirically demonstrate that the fine-tuned diffusion model captures hidden correlations between the residual semantics in initial noise and the original images. Building on this insight, we propose a simple yet effective membership inference attack, which injects semantic information into the initial noise and infers membership by analyzing the model's generation result. Extensive experiments demonstrate that the semantic initial noise can strongly reveal membership information, highlighting the vulnerability of diffusion models to MIAs.

Post Training Quantization (PTQ), a mainstream model compression technique, often leads to the paradoxical 'low error, high loss' phenomenon because it focuses solely on minimizing quantization error. The root cause lies in the Hessian matrix of the LLM loss landscape: a few high curvature directions are extremely sensitive to perturbations. To address this, we propose the Hessian Robust Quantization (HeRo Q) algorithm, which applies a lightweight, learnable rotation-compression matrix to the weight space prior to quantization. This joint framework reshapes the loss landscape by reducing the largest Hessian eigenvalue and reducing its max eigenvalue, thereby significantly enhancing robustness to quantization noise. HeRo-Q requires no architectural modifications, incurs negligible computational overhead, and integrates seamlessly into existing PTQ pipelines. Experiments on Llama and Qwen models show that HeRo Q consistently outperforms state of the art methods including GPTQ, AWQ, and SpinQuant not only achieving superior performance under standard W4A8 settings, but also excelling in the highly challenging W3A16 ultra low bit regime, where it boosts GSM8K accuracy on Llama3 8B to 70.15\% and effectively avoids the logical collapse commonly seen in aggressive quantization.

Modern deep-learning training is not memoryless. Updates depend on optimizer moments and averaging, data-order policies (random reshuffling vs with-replacement, staged augmentations and replay), the nonconvex path, and auxiliary state (teacher EMA/SWA, contrastive queues, BatchNorm statistics). This survey organizes mechanisms by source, lifetime, and visibility. It introduces seed-paired, function-space causal estimands; portable perturbation primitives (carry/reset of momentum/Adam/EMA/BN, order-window swaps, queue/teacher tweaks); and a reporting checklist with audit artifacts (order hashes, buffer/BN checksums, RNG contracts). The conclusion is a protocol for portable, causal, uncertainty-aware measurement that attributes how much training history matters across models, data, and regimes.

Mixed-precision computations are a hallmark of the current stage of AI, driving the progress in large language models towards efficient, locally deployable solutions. This article addresses the floating-point computation of compositionally-rich functions, concentrating on transformer inference. Based on the rounding error analysis of a composition $f(g(\mathrm{x}))$, we provide an adaptive strategy that selects a small subset of components of $g(\mathrm{x})$ to be computed more accurately while all other computations can be carried out with lower accuracy. We then explain how this strategy can be applied to different compositions within a transformer and illustrate its overall effect on transformer inference. We study the effectiveness of this algorithm numerically on GPT-2 models and demonstrate that already very low recomputation rates allow for improvements of up to two orders of magnitude in accuracy.

Parallel thinking enhances LLM reasoning by multi-path sampling and aggregation. In system-level evaluations, a global parallelism level N is allocated to all samples, typically set large to maximize overall dataset accuracy. However, due to sample heterogeneity, some samples can achieve comparable performance with a smaller N'< N, causing budget redundancy. This incompatibility between system-level efficacy and sample-level efficiency constitutes the overscaling curse. In this paper, we formalize and quantify the overscaling curse, showing its universality and severity in practice, and analyze its trigger mechanism. We then propose a lightweight method, T2, to break the overscaling curse, which utilizes latent representations to estimate the optimal parallelism level for each sample before decoding. Experiments show that T2 significantly reduces cost while maintaining comparable performance, enabling more efficient parallel thinking.

Counting should not depend on what is being counted; more generally, any algorithm's behavior should be invariant to the semantic content of its arguments. We introduce WhatCounts to test this property in isolation. Unlike prior work that conflates semantic sensitivity with reasoning complexity or prompt variation, WhatCounts is atomic: count items in an unambiguous, delimited list with no duplicates, distractors, or reasoning steps for different semantic types. Frontier LLMs show over 40% accuracy variation depending solely on what is being counted - cities versus chemicals, names versus symbols. Controlled ablations rule out confounds. The gap is semantic, and it shifts unpredictably with small amounts of unrelated fine-tuning. LLMs do not implement algorithms; they approximate them, and the approximation is argument-dependent. As we show with an agentic example, this has implications beyond counting: any LLM function may carry hidden dependencies on the meaning of its inputs.

Graph Neural Networks frequently exhibit significant performance degradation in the out-of-distribution test scenario. While test-time training (TTT) offers a promising solution, existing Parameter Finetuning (PaFT) paradigm suffer from catastrophic forgetting, hindering their real-world applicability. We propose TTReFT, a novel Test-Time Representation FineTuning framework that transitions the adaptation target from model parameters to latent representations. Specifically, TTReFT achieves this through three key innovations: (1) uncertainty-guided node selection for specific interventions, (2) low-rank representation interventions that preserve pre-trained knowledge, and (3) an intervention-aware masked autoencoder that dynamically adjust masking strategy to accommodate the node selection scheme. Theoretically, we establish guarantees for TTReFT in OOD settings. Empirically, extensive experiments across five benchmark datasets demonstrate that TTReFT achieves consistent and superior performance. Our work establishes representation finetuning as a new paradigm for graph TTT, offering both theoretical grounding and immediate practical utility for real-world deployment.

Missing data is a prevalent issue in many applications, including large medical registries such as the Swedish Healthcare Quality Registries, potentially leading to biased or inefficient analyses if not handled properly. Multiple Imputation by Chained Equations (MICE) is a popular and versatile method for handling multivariate missing data but traditional implementations face significant challenges when applied to big data sets due to computational time and memory limitations. To address this, the bigMICE package was developed, adapting the MICE framework to big data using Apache Spark MLLib and Spark ML. Our implementation allows for controlling the maximum memory usage during the execution, enabling processing of very large data sets on a hardware with a limited memory, such as ordinary laptops. The developed package was tested on a large Swedish medical registry to measure memory usage, runtime and dependence of the imputation quality on sample size and on missingness proportion in the data. In conclusion, our method is generally more memory efficient and faster on large data sets compared to a commonly used MICE implementation. We also demonstrate that working with very large datasets can result in high quality imputations even when a variable has a large proportion of missing data. This paper also provides guidelines and recommendations on how to install and use our open source package.

Bootstrap-based Self-Supervised Learning (SSL) has achieved remarkable progress in audio understanding. However, existing methods typically operate at a single level of granularity, limiting their ability to model the diverse temporal and spectral structures inherent in complex audio signals. Furthermore, bootstrapping representations from scratch is computationally expensive, often requiring extensive training to converge. In this work, we propose the Convolutional Audio Transformer (CAT), a unified framework designed to address these challenges. First, to capture hierarchical audio features, CAT incorporates a Multi-resolution Block that aggregates information across varying granularities. Second, to enhance training efficiency, we introduce a Representation Regularization objective. Drawing inspiration from generative modeling, this auxiliary task guides the student model by aligning its predictions with high-quality semantic representations from frozen, pre-trained external encoders. Experimental results demonstrate that CAT significantly outperforms baselines on audio understanding benchmarks. Notably, it achieves competitive performance on the AudioSet 20k dataset with 5 times faster convergence than existing methods. Codes and checkpoints will be released soon at https://github.com/realzhouchushu/CAT.

Effective relevance modeling is crucial for e-commerce search, as it aligns search results with user intent and enhances customer experience. Recent work has leveraged large language models (LLMs) to address the limitations of traditional relevance models, especially for long-tail and ambiguous queries. By incorporating Chain-of-Thought (CoT) reasoning, these approaches improve both accuracy and interpretability through multi-step reasoning. However, two key limitations remain: (1) most existing approaches rely on single-perspective CoT reasoning, which fails to capture the multifaceted nature of e-commerce relevance (e.g., user intent vs. attribute-level matching vs. business-specific rules); and (2) although CoT-enhanced LLM's offer rich reasoning capabilities, their high inference latency necessitates knowledge distillation for real-time deployment, yet current distillation methods discard the CoT rationale structure at inference, using it as a transient auxiliary signal and forfeiting its reasoning utility. To address these challenges, we propose a novel framework that better exploits CoT semantics throughout the optimization pipeline. Specifically, the teacher model leverages Multi-Perspective CoT (MPCoT) to generate diverse rationales and combines Supervised Fine-Tuning (SFT) with Direct Preference Optimization (DPO) to construct a more robust reasoner. For distillation, we introduce Latent Reasoning Knowledge Distillation (LRKD), which endows a student model with a lightweight inference-time latent reasoning extractor, allowing efficient and low-latency internalization of the LLM's sophisticated reasoning capabilities. Evaluated in offline experiments and online A/B tests on an e-commerce search advertising platform serving tens of millions of users daily, our method delivers significant offline gains, showing clear benefits in both commercial performance and user experience.

Agentic recommender systems leverage Large Language Models (LLMs) to model complex user behaviors and support personalized decision-making. However, existing methods primarily model preference changes based on explicit user-item interactions, which are sparse, noisy, and unable to reflect the real-time, mutual influences among users and items. To address these limitations, we propose RecNet, a self-evolving preference propagation framework that proactively propagates real-time preference updates across related users and items. RecNet consists of two complementary phases. In the forward phase, the centralized preference routing mechanism leverages router agents to integrate preference updates and dynamically propagate them to the most relevant agents. To ensure accurate and personalized integration of propagated preferences, we further introduce a personalized preference reception mechanism, which combines a message buffer for temporary caching and an optimizable, rule-based filter memory to guide selective preference assimilation based on past experience and interests. In the backward phase, the feedback-driven propagation optimization mechanism simulates a multi-agent reinforcement learning framework, using LLMs for credit assignment, gradient analysis, and module-level optimization, enabling continuous self-evolution of propagation strategies. Extensive experiments on various scenarios demonstrate the effectiveness of RecNet in modeling preference propagation for recommender systems.

Enterprise-grade Intelligent Document Processing (IDP) systems support high-stakes workflows across finance, insurance, and healthcare. Early-phase system validation under limited budgets mandates uncovering diverse failure mechanisms, rather than identifying a single worst-case document. We formalize this challenge as a Search-Based Software Testing (SBST) problem, aiming to identify complex interactions between document variables, with the objective to maximize the number of distinct failure types discovered within a fixed evaluation budget. Our methodology operates on a combinatorial space of document configurations, rendering instances of structural \emph{risk features} to induce realistic failure conditions. We benchmark a diverse portfolio of search strategies spanning evolutionary, swarm-based, quality-diversity, learning-based, and quantum under identical budget constraints. Through configuration-level exclusivity, win-rate, and cross-temporal overlap analyses, we show that different solvers consistently uncover failure modes that remain undiscovered by specific alternatives at comparable budgets. Crucially, cross-temporal analysis reveals persistent solver-specific discoveries across all evaluated budgets, with no single strategy exhibiting absolute dominance. While the union of all solvers eventually recovers the observed failure space, reliance on any individual method systematically delays the discovery of important risks. These results demonstrate intrinsic solver complementarity and motivate portfolio-based SBST strategies for robust industrial IDP validation.

In recent years, mobile applications have become indispensable tools for managing various aspects of life. From enhancing productivity to providing personalized entertainment, mobile apps have revolutionized people's daily routines. Despite this rapid growth and popularity, gaps remain in how these apps address the needs of users from different age groups. Users of varying ages face distinct challenges when interacting with mobile apps, from younger users dealing with inappropriate content to older users having difficulty with usability due to age-related vision and cognition impairments. Although there have been initiatives to create age-inclusive apps, a limited understanding of user perspectives on age-related issues may hinder developers from recognizing specific challenges and implementing effective solutions. In this study, we explore age discussions in app reviews to gain insights into how mobile apps should cater to users across different age groups.We manually curated a dataset of 4,163 app reviews from the Google Play Store and identified 1,429 age-related reviews and 2,734 non-age-related reviews. We employed eight machine learning, deep learning, and large language models to automatically detect age discussions, with RoBERTa performing the best, achieving a precision of 92.46%. Additionally, a qualitative analysis of the 1,429 age-related reviews uncovers six dominant themes reflecting user concerns.

Neural networks adapt through first-order parameter updates, yet it remains unclear whether such updates preserve logical coherence. We investigate the geometric limits of the Linear Propagation Assumption (LPA), the premise that local updates coherently propagate to logical consequences. To formalize this, we adopt relation algebra and study three core operations on relations: negation flips truth values, converse swaps argument order, and composition chains relations. For negation and converse, we prove that guaranteeing direction-agnostic first-order propagation necessitates a tensor factorization separating entity-pair context from relation content. However, for composition, we identify a fundamental obstruction. We show that composition reduces to conjunction, and prove that any conjunction well-defined on linear features must be bilinear. Since bilinearity is incompatible with negation, this forces the feature map to collapse. These results suggest that failures in knowledge editing, the reversal curse, and multi-hop reasoning may stem from common structural limitations inherent to the LPA.

Large language models exhibit complementary reasoning errors: on the same instance, one model may succeed with a particular decomposition while another fails. We propose Collaborative Reasoning (CORE), a training-time collaboration framework that converts peer success into a learning signal via a cross-teaching protocol. Each problem is solved in two stages: a cold round of independent sampling, followed by a contexted rescue round in which models that failed receive hint extracted from a successful peer. CORE optimizes a combined reward that balances (i) correctness, (ii) a lightweight DPP-inspired diversity term to reduce error overlap, and (iii) an explicit rescue bonus for successful recovery. We evaluate CORE across four standard reasoning datasets GSM8K, MATH, AIME, and GPQA. With only 1,000 training examples, a pair of small open source models (3B+4B) reaches Pass@2 of 99.54% on GSM8K and 92.08% on MATH, compared to 82.50% and 74.82% for single-model training. On harder datasets, the 3B+4B pair reaches Pass@2 of 77.34% on GPQA (trained on 348 examples) and 79.65% on AIME (trained on 792 examples), using a training-time budget of at most 1536 context tokens and 3072 generated tokens. Overall, these results show that training-time collaboration can reliably convert model complementarity into large gains without scaling model size.

Aiming at efficient and dense chain-of-thought (CoT) reasoning, latent reasoning methods fine-tune Large Language Models (LLMs) to substitute discrete language tokens with continuous latent tokens. These methods consume fewer tokens compared to the conventional language CoT reasoning and have the potential to plan in a dense latent space. However, current latent tokens are generally supervised based on imitating language labels. Considering that there can be multiple equivalent but diverse CoT labels for a question, passively imitating an arbitrary one may lead to inferior latent token representations and latent reasoning policies, undermining the potential planning ability and resulting in clear gaps between training and testing. In this work, we emphasize the importance of active planning over the representation space of latent tokens in achieving the optimal latent reasoning policy. So, we propose the \underline{A}c\underline{t}ive Latent \underline{P}lanning method (ATP-Latent), which models the supervision process of latent tokens as a conditional variational auto-encoder (VAE) to obtain a smoother latent space. Moreover, to facilitate the most reasonable latent reasoning policy, ATP-Latent conducts reinforcement learning (RL) with an auxiliary coherence reward, which is calculated based on the consistency between VAE-decoded contents of latent tokens, enabling a guided RL process. In experiments on LLaMA-1B, ATP-Latent demonstrates +4.1\% accuracy and -3.3\% tokens on four benchmarks compared to advanced baselines. Codes are available on https://github.com/zz1358m/ATP-Latent-master.

Blockchain clients are fundamental software for running blockchain nodes. They provide users with various RPC (Remote Procedure Call) interfaces to interact with the blockchain. These RPC methods are expected to follow the same specification across different blockchain nodes, providing users with seamless interaction. However, there have been continuous reports on various RPC bugs that can cause unexpected responses or even Denial of Service weakness. Existing studies on blockchain RPC bug detection mainly focus on generating the RPC method calls for testing blockchain clients. However, a wide range of the reported RPC bugs are triggered in various blockchain contexts. To the best of our knowledge, little attention is paid to generating proper contexts that can trigger these context-dependent RPC bugs. In this work, we propose EthCRAFT, a Context-aware RPC Analysis and Fuzzing Tool for client RPC bug detection. EthCRAFT first proposes to explore the state transition program space of blockchain clients and generate various transactions to construct the context. EthCRAFT then designs a context-aware RPC method call generation method to send RPC calls to the blockchain clients. The responses of 5 different client implementations are used as cross-referring oracles to detect the RPC bugs. We evaluate EthCRAFT on real-world RPC bugs collected from the GitHub issues of Ethereum client implementations. Experiment results show that EthCRAFT outperforms existing client RPC detectors by detecting more RPC bugs. Moreover, EthCRAFT has found six new bugs in major Ethereum clients and reported them to the developers. One of the bug fixes has been written into breaking changes in the client's updates. Three of our bug reports have been offered a vulnerability bounty by the Ethereum Foundation.

Reinforcement learning (RL) post-training is a dominant approach for improving the reasoning performance of large language models (LLMs), yet growing evidence suggests that its gains arise primarily from distribution sharpening rather than the acquisition of new capabilities. Recent work has shown that sampling from the power distribution of LLMs using Markov chain Monte Carlo (MCMC) can recover performance comparable to RL post-training without relying on external rewards; however, the high computational cost of MCMC makes such approaches impractical for widespread adoption. In this work, we propose a theoretically grounded alternative that eliminates the need for iterative MCMC. We derive a novel formulation showing that the global power distribution can be approximated by a token-level scaled low-temperature one, where the scaling factor captures future trajectory quality. Leveraging this insight, we introduce a training-free and verifier-free algorithm that sharpens the base model's generative distribution autoregressively. Empirically, we evaluate our method on math, QA, and code tasks across four LLMs, and show that our method matches or surpasses one-shot GRPO without relying on any external rewards, while reducing inference latency by over 10x compared to MCMC-based sampling.

Graph Federated Learning (GFL) enables distributed graph representation learning while protecting the privacy of graph data. However, GFL suffers from heterogeneity arising from diverse node features and structural topologies across multiple clients. To address both types of heterogeneity, we propose a novel graph Federated learning method via Semantic and Structural Alignment (FedSSA), which shares the knowledge of both node features and structural topologies. For node feature heterogeneity, we propose a novel variational model to infer class-wise node distributions, so that we can cluster clients based on inferred distributions and construct cluster-level representative distributions. We then minimize the divergence between local and cluster-level distributions to facilitate semantic knowledge sharing. For structural heterogeneity, we employ spectral Graph Neural Networks (GNNs) and propose a spectral energy measure to characterize structural information, so that we can cluster clients based on spectral energy and build cluster-level spectral GNNs. We then align the spectral characteristics of local spectral GNNs with those of cluster-level spectral GNNs to enable structural knowledge sharing. Experiments on six homophilic and five heterophilic graph datasets under both non-overlapping and overlapping partitioning settings demonstrate that FedSSA consistently outperforms eleven state-of-the-art methods.

Despite the centrality of crosslinguistic influence (CLI) to bilingualism research, human studies often yield conflicting results due to inherent experimental variance. We address these inconsistencies by using language models (LMs) as controlled statistical learners to systematically simulate CLI and isolate its underlying drivers. Specifically, we study the effect of varying the L1 language dominance and the L2 language proficiency, which we manipulate by controlling the L2 age of exposure -- defined as the training step at which the L2 is introduced. Furthermore, we investigate the impact of pretraining on L1 languages with varying syntactic distance from the L2. Using cross-linguistic priming, we analyze how activating L1 structures impacts L2 processing. Our results align with evidence from psycholinguistic studies, confirming that language dominance and proficiency are strong predictors of CLI. We further find that while priming of grammatical structures is bidirectional, the priming of ungrammatical structures is sensitive to language dominance. Finally, we provide mechanistic evidence of CLI in LMs, demonstrating that the L1 is co-activated during L2 processing and directly influences the neural circuitry recruited for the L2. More broadly, our work demonstrates that LMs can serve as a computational framework to inform theories of human CLI.

Many learning problems require predicting sets of objects when the number of objects is not known beforehand. Examples include object detection, molecular modeling, and scientific inference tasks such as astrophysical source detection. Existing methods often rely on padded representations or must explicitly infer the set size, which often poses challenges. We present a novel strategy for addressing this challenge by casting prediction of variable-sized sets as a continuous inference problem. Our approach, CORDS (Continuous Representations of Discrete Structures), provides an invertible mapping that transforms a set of spatial objects into continuous fields: a density field that encodes object locations and count, and a feature field that carries their attributes over the same support. Because the mapping is invertible, models operate entirely in field space while remaining exactly decodable to discrete sets. We evaluate CORDS across molecular generation and regression, object detection, simulation-based inference, and a mathematical task involving recovery of local maxima, demonstrating robust handling of unknown set sizes with competitive accuracy.

Depth-recurrence facilitates latent reasoning by sharing parameters across depths. However, prior work lacks combined FLOP-, parameter-, and memory-matched baselines, underutilizes depth-recurrence due to partially fixed layer stacks, and ignores the bottleneck of constant hidden-sizes that restricts many-step latent reasoning. To address this, we introduce a modular framework of depth-recurrent attention mixtures (Dreamer), combining sequence attention, depth attention, and sparse expert attention. It alleviates the hidden-size bottleneck through attention along depth, decouples scaling dimensions, and allows depth-recurrent models to scale efficiently and effectively. Across language reasoning benchmarks, our models require 2 to 8x fewer training tokens for the same accuracy as FLOP-, parameter-, and memory-matched SOTA, and outperform ca. 2x larger SOTA models with the same training tokens. We further present insights into knowledge usage across depths, e.g., showing 2 to 11x larger expert selection diversity than SOTA MoEs.

Deep learning models struggle with uncertainty estimation. Many approaches are either computationally infeasible or underestimate uncertainty. We investigate \textit{BatchEnsemble} as a general and scalable method for uncertainty estimation across both tabular and time series tasks. To extend BatchEnsemble to sequential modeling, we introduce GRUBE, a novel BatchEnsemble GRU cell. We compare the BatchEnsemble to Monte Carlo dropout and deep ensemble models. Our results show that BatchEnsemble matches the uncertainty estimation performance of deep ensembles, and clearly outperforms Monte Carlo dropout. GRUBE achieves similar or better performance in both prediction and uncertainty estimation. These findings show that BatchEnsemble and GRUBE achieve similar performance with fewer parameters and reduced training and inference time compared to traditional ensembles.

The success of Hyper-Connections (HC) in neural networks (NN) has also highlighted issues related to its training instability and restricted scalability. The Manifold-Constrained Hyper-Connections (mHC) mitigate these challenges by projecting the residual connection space onto a Birkhoff polytope, however, it faces two issues: 1) its iterative Sinkhorn-Knopp (SK) algorithm does not always yield exact doubly stochastic residual matrices; 2) mHC incurs a prohibitive $\mathcal{O}(n^3C)$ parameter complexity with $n$ as the width of the residual stream and $C$ as the feature dimension. The recently proposed mHC-lite reparametrizes the residual matrix via the Birkhoff-von-Neumann theorem to guarantee double stochasticity, but also faces a factorial explosion in its parameter complexity, $\mathcal{O} \left( nC \cdot n! \right)$. To address both challenges, we propose \textbf{KromHC}, which uses the \underline{Kro}necker products of smaller doubly stochastic matrices to parametrize the residual matrix in \underline{mHC}. By enforcing manifold constraints across the factor residual matrices along each mode of the tensorized residual stream, KromHC guarantees exact double stochasticity of the residual matrices while reducing parameter complexity to $\mathcal{O}(n^2C)$. Comprehensive experiments demonstrate that KromHC matches or even outperforms state-of-the-art (SOTA) mHC variants, while requiring significantly fewer trainable parameters. The code is available at \texttt{https://github.com/wz1119/KromHC}.

Catastrophic forgetting during knowledge injection severely undermines the continual learning capability of large language models (LLMs). Although existing methods attempt to mitigate this issue, they often lack a foundational theoretical explanation. We establish a gradient-based theoretical framework to explain catastrophic forgetting. We first prove that strongly negative gradient similarity is a fundamental cause of forgetting. We then use gradient similarity to identify two types of neurons: conflicting neurons that induce forgetting and account for 50%-75% of neurons, and collaborative neurons that mitigate forgetting and account for 25%-50%. Based on this analysis, we propose a knowledge injection method, Collaborative Neural Learning (CNL). By freezing conflicting neurons and updating only collaborative neurons, CNL theoretically eliminates catastrophic forgetting under an infinitesimal learning rate eta and an exactly known mastered set. Experiments on five LLMs, four datasets, and four optimizers show that CNL achieves zero forgetting in in-set settings and reduces forgetting by 59.1%-81.7% in out-of-set settings.

Chain-of-Thought (CoT) has unlocked advanced reasoning abilities of Large Language Models (LLMs) with intermediate steps, yet incurs prohibitive computational costs due to generation of extra tokens. Recent studies empirically show that compressing reasoning steps into latent states, or implicit CoT compression, offers a token-efficient alternative. However, the mechanism behind CoT compression remains unclear. In this paper, we provide the first theoretical analysis of the difficulty of learning to internalize intermediate reasoning steps. By introducing Order-r Interaction, we prove that the learning signal for high-order logical dependencies exponentially decays to solve irreducible problem, where skipping intermediate steps inevitably leads to high-order interaction barriers. To empirically validate this, we introduce NatBool-DAG, a challenging benchmark designed to enforce irreducible logical reasoning and eliminate semantic shortcuts. Guided by our theoretical findings, we propose ALiCoT (Aligned Implicit CoT), a novel framework that overcomes the signal decay by aligning latent token distributions with intermediate reasoning states. Experimental results demonstrate that ALiCoT successfully unlocks efficient reasoning: it achieves a 54.4x speedup while maintaining performance comparable to explicit CoT.

Recurrent spiking neural networks (RSNNs) are a promising substrate for energy-efficient control policies, but training them for high-dimensional, long-horizon reinforcement learning remains challenging. Population-based, gradient-free optimization circumvents backpropagation through non-differentiable spike dynamics by estimating gradients. However, with finite populations, high variance of these estimates can induce harmful and overly aggressive update steps. Inspired by trust-region methods in reinforcement learning that constrain policy updates in distribution space, we propose \textbf{Signal-Adaptive Trust Regions (SATR)}, a distributional update rule that constrains relative change by bounding KL divergence normalized by an estimated signal energy. SATR automatically expands the trust region under strong signals and contracts it when updates are noise-dominated. We instantiate SATR for Bernoulli connectivity distributions, which have shown strong empirical performance for RSNN optimization. Across a suite of high-dimensional continuous-control benchmarks, SATR improves stability under limited populations and reaches competitive returns against strong baselines including PPO-LSTM. In addition, to make SATR practical at scale, we introduce a bitset implementation for binary spiking and binary weights, substantially reducing wall-clock training time and enabling fast RSNN policy search.

Current approaches to reducing undesired capabilities in language models are largely post hoc, and can thus be easily bypassed by adversaries. A natural alternative is to shape capabilities during pretraining itself. On the proxy task of removing medical capabilities, we show that the simple intervention of filtering pretraining data is highly effective, robust, and inexpensive at scale. Inspired by work on data attribution, we show that filtering tokens is more effective than filtering documents, achieving the same hit to undesired capabilities at a lower cost to benign ones. Training models spanning two orders of magnitude, we then demonstrate that filtering gets more effective with scale: for our largest models, token filtering leads to a 7000x compute slowdown on the forget domain. We also show that models trained with token filtering can still be aligned on the forget domain. Along the way, we introduce a methodology for labeling tokens with sparse autoencoders and distilling cheap, high-quality classifiers. We also demonstrate that filtering can be robust to noisy labels with sufficient pretraining compute.

The field of Embodied AI is witnessing a rapid evolution toward general-purpose robotic systems, fueled by high-fidelity simulation and large-scale data collection. However, this scaling capability remains severely bottlenecked by a reliance on labor-intensive manual oversight from intricate reward shaping to hyperparameter tuning across heterogeneous backends. Inspired by LLMs' success in software automation and science discovery, we introduce \textsc{EmboCoach-Bench}, a benchmark evaluating the capacity of LLM agents to autonomously engineer embodied policies. Spanning 32 expert-curated RL and IL tasks, our framework posits executable code as the universal interface. We move beyond static generation to assess a dynamic closed-loop workflow, where agents leverage environment feedback to iteratively draft, debug, and optimize solutions, spanning improvements from physics-informed reward design to policy architectures such as diffusion policies. Extensive evaluations yield three critical insights: (1) autonomous agents can qualitatively surpass human-engineered baselines by 26.5\% in average success rate; (2) agentic workflow with environment feedback effectively strengthens policy development and substantially narrows the performance gap between open-source and proprietary models; and (3) agents exhibit self-correction capabilities for pathological engineering cases, successfully resurrecting task performance from near-total failures through iterative simulation-in-the-loop debugging. Ultimately, this work establishes a foundation for self-evolving embodied intelligence, accelerating the paradigm shift from labor-intensive manual tuning to scalable, autonomous engineering in embodied AI field.

We present a unified framework for quantifying the similarity between representations through the lens of \textit{usable information}, offering a rigorous theoretical and empirical synthesis across three key dimensions. First, addressing functional similarity, we establish a formal link between stitching performance and conditional mutual information. We further reveal that stitching is inherently asymmetric, demonstrating that robust functional comparison necessitates a bidirectional analysis rather than a unidirectional mapping. Second, concerning representational similarity, we prove that reconstruction-based metrics and standard tools (e.g., CKA, RSA) act as estimators of usable information under specific constraints. Crucially, we show that similarity is relative to the capacity of the predictive family: representations that appear distinct to a rigid observer may be identical to a more expressive one. Third, we demonstrate that representational similarity is sufficient but not necessary for functional similarity. We unify these concepts through a task-granularity hierarchy: similarity on a complex task guarantees similarity on any coarser derivative, establishing representational similarity as the limit of maximum granularity: input reconstruction.

Causal Disentangled Representation Learning(CDRL) aims to learn and disentangle low dimensional representations and their underlying causal structure from observations. However, existing disentanglement methods rely on a standard mean-field approximation with a diagonal posterior covariance, which decorrelates all latent dimensions. Additionally, these methods often assume isotropic Gaussian priors for exogenous noise, failing to capture the complex, non-Gaussian statistical properties prevalent in real-world causal factors. Therefore, we propose FlexCausal, a novel CDRL framework based on a block-diagonal covariance VAE. FlexCausal utilizes a Factorized Flow-based Prior to realistically model the complex densities of exogenous noise, effectively decoupling the learning of causal mechanisms from distributional statistics. By integrating supervised alignment objectives with counterfactual consistency constraints, our framework ensures a precise structural correspondence between the learned latent subspaces and the ground-truth causal relations. Finally, we introduce a manifold-aware relative intervention strategy to ensure high-fidelity generation. Experimental results on both synthetic and real-world datasets demonstrate that FlexCausal significantly outperforms other methods.

Clear and concise code is necessary to ensure maintainability, so it is crucial that the software is as simple as possible to understand, to avoid bugs and, above all, vulnerabilities. There are many ways to enhance software without changing its functionality, considering the extract method refactoring the primary process to reduce the effort required for code comprehension. The cognitive complexity measure employed in this work is the one defined by SonarSource, which is a company that develops well-known applications for static code analysis. This extraction problem can be modeled as a combinatorial optimization problem. The main difficulty arises from the existence of different criteria for evaluating the solutions obtained, requiring the formulation of the code extraction problem as a multi-objective optimization problem using alternative methods. We propose a multi-objective integer linear programming model to obtain a set of solutions that reduce the cognitive complexity of a given piece of code, such as balancing the number of lines of code and its cognitive complexity. In addition, several algorithms have been developed to validate the model. These algorithms have been integrated into a tool that enables the parameterised resolution of the problem of reducing software cognitive complexity.

Machine unlearning seeks to remove the influence of specific training data from a model, a need driven by privacy regulations and robustness concerns. Existing approaches typically modify model parameters, but such updates can be unstable, computationally costly, and limited by local approximations. We introduce Representation Unlearning, a framework that performs unlearning directly in the model's representation space. Instead of modifying model parameters, we learn a transformation over representations that imposes an information bottleneck: maximizing mutual information with retained data while suppressing information about data to be forgotten. We derive variational surrogates that make this objective tractable and show how they can be instantiated in two practical regimes: when both retain and forget data are available, and in a zero-shot setting where only forget data can be accessed. Experiments across several benchmarks demonstrate that Representation Unlearning achieves more reliable forgetting, better utility retention, and greater computational efficiency than parameter-centric baselines.

Standard deep learning relies on Backpropagation (BP), which is constrained by biologically implausible weight symmetry and suffers from significant gradient interference within dense representations. To mitigate these bottlenecks, we propose Selective Adaptive Learning (SAL), a training method that combines selective parameter activation with adaptive area partitioning. Specifically, SAL decomposes the parameter space into mutually exclusive, sample-dependent regions. This decoupling mitigates gradient interference across divergent semantic patterns and addresses explicit weight symmetry requirements through our refined feedback alignment. Empirically, SAL demonstrates competitive convergence rates, leading to improved classification performance across 10 standard benchmarks. Additionally, SAL achieves numerical consistency and competitive accuracy even in deep regimes (up to 128 layers) and large-scale models (up to 1B parameters). Our approach is loosely inspired by biological learning mechanisms, offering a plausible alternative that contributes to the study of scalable neural network training.

Predicting spatial gene expression from H&E histology offers a scalable and clinically accessible alternative to sequencing, but realizing clinical impact requires models that generalize across cancer types and capture biologically coherent signals. Prior work is often limited to per-cancer settings and variance-based evaluation, leaving functional relevance underexplored. We introduce HistoPrism, an efficient transformer-based architecture for pan-cancer prediction of gene expression from histology. To evaluate biological meaning, we introduce a pathway-level benchmark, shifting assessment from isolated gene-level variance to coherent functional pathways. HistoPrism not only surpasses prior state-of-the-art models on highly variable genes , but also more importantly, achieves substantial gains on pathway-level prediction, demonstrating its ability to recover biologically coherent transcriptomic patterns. With strong pan-cancer generalization and improved efficiency, HistoPrism establishes a new standard for clinically relevant transcriptomic modeling from routinely available histology.

Large language models (LLMs) are increasingly used as tool-augmented agents for multi-step decision making, yet training robust tool-using agents remains challenging. Existing methods still require manual intervention, depend on non-verifiable simulated environments, rely exclusively on either supervised fine-tuning (SFT) or reinforcement learning (RL), and struggle with stable long-horizon, multi-turn learning. To address these challenges, we introduce ASTRA, a fully automated end-to-end framework for training tool-augmented language model agents via scalable data synthesis and verifiable reinforcement learning. ASTRA integrates two complementary components. First, a pipeline that leverages the static topology of tool-call graphs synthesizes diverse, structurally grounded trajectories, instilling broad and transferable tool-use competence. Second, an environment synthesis framework that captures the rich, compositional topology of human semantic reasoning converts decomposed question-answer traces into independent, code-executable, and rule-verifiable environments, enabling deterministic multi-turn RL. Based on this method, we develop a unified training methodology that integrates SFT with online RL using trajectory-level rewards to balance task completion and interaction efficiency. Experiments on multiple agentic tool-use benchmarks demonstrate that ASTRA-trained models achieve state-of-the-art performance at comparable scales, approaching closed-source systems while preserving core reasoning ability. We release the full pipelines, environments, and trained models at https://github.com/LianjiaTech/astra.

The operational efficacy of large language models relies heavily on their inference-time context. This has established Context Engineering (CE) as a formal discipline for optimizing these inputs. Current CE methods rely on manually crafted harnesses, such as rigid generation-reflection workflows and predefined context schemas. They impose structural biases and restrict context optimization to a narrow, intuition-bound design space. To address this, we introduce Meta Context Engineering (MCE), a bi-level framework that supersedes static CE heuristics by co-evolving CE skills and context artifacts. In MCE iterations, a meta-level agent refines engineering skills via agentic crossover, a deliberative search over the history of skills, their executions, and evaluations. A base-level agent executes these skills, learns from training rollouts, and optimizes context as flexible files and code. We evaluate MCE across five disparate domains under offline and online settings. MCE demonstrates consistent performance gains, achieving 5.6--53.8% relative improvement over state-of-the-art agentic CE methods (mean of 16.9%), while maintaining superior context adaptability, transferability, and efficiency in both context usage and training.

Memory safety bugs, such as out-of-bound accesses (OOB) in GPU programs, can compromise the security and reliability of GPU-accelerated software. We report the existence of input-dependent OOBs in the wild that manifest only under specific inputs. All existing tools to detect OOBs in GPU programs rely on runtime techniques that require an OOB to manifest for detection. Thus, input-dependent OOBs elude them. We also discover intra-allocation OOBs that arise in the presence of logical partitioning of a memory allocation into multiple data structures. Existing techniques are oblivious to the possibility of such OOBs. We make a key observation that the presence (or absence) of semantic relations among program variables, which determines the size of allocations (CPU code) and those calculating offsets into memory allocations (GPU code), helps identify the absence (or presence) of OOBs. We build SCuBA, a first-of-its-kind compile-time technique that analyzes CPU and GPU code to capture such semantic relations (if present). It uses a SAT solver to check if an OOB access is possible under any input, given the captured relations expressed as constraints. It further analyzes GPU code to track logical partitioning of memory allocations for detecting intra-allocation OOB. Compared to NVIDIA's Compute Sanitizer that misses 45 elusive memory bugs across 20 programs, SCuBA misses none with no false alarms.

Effective clinical history taking is a foundational yet underexplored component of clinical reasoning. While large language models (LLMs) have shown promise on static benchmarks, they often fall short in dynamic, multi-turn diagnostic settings that require iterative questioning and hypothesis refinement. To address this gap, we propose \method{}, a note-driven framework that trains LLMs to conduct structured history taking and diagnosis by learning from widely available medical notes. Instead of relying on scarce and sensitive dialogue data, we convert real-world medical notes into high-quality doctor-patient dialogues using a decision tree-guided generation and refinement pipeline. We then propose a three-stage fine-tuning strategy combining supervised learning, simulated data augmentation, and preference learning. Furthermore, we propose a novel single-turn reasoning paradigm that reframes history taking as a sequence of single-turn reasoning problems. This design enhances interpretability and enables local supervision, dynamic adaptation, and greater sample efficiency. Experimental results show that our method substantially improves clinical reasoning, achieving gains of +16.9 F1 and +21.0 Top-1 diagnostic accuracy over GPT-4o. Our code and dataset can be found at https://github.com/zhentingsheng/Note2Chat.

Air hockey demands split-second decisions at high puck velocities, a challenge we address with a compact network of spiking neurons running on a mixed-signal analog/digital neuromorphic processor. By co-designing hardware and learning algorithms, we train the system to achieve successful puck interactions through reinforcement learning in a remarkably small number of trials. The network leverages fixed random connectivity to capture the task's temporal structure and adopts a local e-prop learning rule in the readout layer to exploit event-driven activity for fast and efficient learning. The result is real-time learning with a setup comprising a computer and the neuromorphic chip in-the-loop, enabling practical training of spiking neural networks for robotic autonomous systems. This work bridges neuroscience-inspired hardware with real-world robotic control, showing that brain-inspired approaches can tackle fast-paced interaction tasks while supporting always-on learning in intelligent machines.

Real-world time series exhibit complex and evolving dynamics, making accurate forecasting extremely challenging. Recent multi-modal forecasting methods leverage textual information such as news reports to improve prediction, but most rely on token-level fusion that mixes temporal patches with language tokens in a shared embedding space. However, such fusion can be ill-suited when high-quality time-text pairs are scarce and when time series exhibit substantial variation in scale and characteristics, thus complicating cross-modal alignment. In parallel, Mixture-of-Experts (MoE) architectures have proven effective for both time series modeling and multi-modal learning, yet many existing MoE-based modality integration methods still depend on token-level fusion. To address this, we propose Expert Modulation, a new paradigm for multi-modal time series prediction that conditions both routing and expert computation on textual signals, enabling direct and efficient cross-modal control over expert behavior. Through comprehensive theoretical analysis and experiments, our proposed method demonstrates substantial improvements in multi-modal time series prediction. The current code is available at https://github.com/BruceZhangReve/MoME

Agentic large language model (LLM) systems rely on external memory for long-horizon state and concurrent multi-agent execution, but centralized indexes and heuristic partitions become bottlenecks as memory volume and parallel access grow. We present ShardMemo, a budgeted tiered memory service with Tier A per-agent working state, Tier B sharded evidence with shard-local approximate nearest neighbor (ANN) indexes, and Tier C, a versioned skill library. Tier B enforces scope-before-routing: structured eligibility constraints mask ineligible shards before routing or ANN search. We cast shard probing as masked mixture-of-experts (MoE) routing over eligible shards, probing up to $B_{\mathrm{probe}}$ shards via Top-$B_{\mathrm{probe}}$ or adaptive Top-$P$, and use cost-aware gating over profile/observation/session shard families; the router is trained from evidence-to-shard supervision. On LoCoMo, ShardMemo improves over the strongest baseline (GAM) by +5.11 to +6.82 F1 across question categories. Under a fixed-budget routing setting ($B_{\mathrm{probe}}=3$), ShardMemo improves over cosine-to-prototype shard routing by +6.87 F1 while reducing retrieval work (VecScan 521->414, -20.5%) and p95 latency (95->76 ms). On long-context HotpotQA, ShardMemo achieves 63.41/61.88/57.95 F1 at 56K/224K/448K tokens. On ToolBench, Tier C reaches 0.97 Precision@3 and 1.94 StepRed (+10.2% and +7.2% over embedding-similarity retrieval).

Mixup generates augmented samples by linearly interpolating inputs and labels with a controllable ratio. However, since it operates in the latent embedding level, the resulting samples are not human-interpretable. In contrast, LLM-based augmentation methods produce sentences via prompts at the token level, yielding readable outputs but offering limited control over the generation process. Inspired by recent advances in LLM inversion, which reconstructs natural language from embeddings and helps bridge the gap between latent embedding space and discrete token space, we propose inversedMixup, a unified framework that combines the controllability of Mixup with the interpretability of LLM-based generation. Specifically, inversedMixup adopts a three-stage training procedure to align the output embedding space of a task-specific model with the input embedding space of an LLM. Upon successful alignment, inversedMixup can reconstruct mixed embeddings with a controllable mixing ratio into human-interpretable augmented sentences, thereby improving the augmentation performance. Additionally, inversedMixup provides the first empirical evidence of the manifold intrusion phenomenon in text Mixup and introduces a simple yet effective strategy to mitigate it. Extensive experiments demonstrate the effectiveness and generalizability of our approach in both few-shot and fully supervised scenarios.

Flow Matching (FM) models have emerged as a leading paradigm for high-fidelity synthesis. However, their reliance on iterative Ordinary Differential Equation (ODE) solving creates a significant latency bottleneck. Existing solutions face a dichotomy: training-free solvers suffer from significant performance degradation at low Neural Function Evaluations (NFEs), while training-based one- or few-steps generation methods incur prohibitive training costs and lack plug-and-play versatility. To bridge this gap, we propose the Bi-Anchor Interpolation Solver (BA-solver). BA-solver retains the versatility of standard training-free solvers while achieving significant acceleration by introducing a lightweight SideNet (1-2% backbone size) alongside the frozen backbone. Specifically, our method is founded on two synergistic components: \textbf{1) Bidirectional Temporal Perception}, where the SideNet learns to approximate both future and historical velocities without retraining the heavy backbone; and 2) Bi-Anchor Velocity Integration, which utilizes the SideNet with two anchor velocities to efficiently approximate intermediate velocities for batched high-order integration. By utilizing the backbone to establish high-precision ``anchors'' and the SideNet to densify the trajectory, BA-solver enables large interval sizes with minimized error. Empirical results on ImageNet-256^2 demonstrate that BA-solver achieves generation quality comparable to 100+ NFEs Euler solver in just 10 NFEs and maintains high fidelity in as few as 5 NFEs, incurring negligible training costs. Furthermore, BA-solver ensures seamless integration with existing generative pipelines, facilitating downstream tasks such as image editing.

Can classical consensus models predict the group behavior of large language models (LLMs)? We examine multi-round interactions among LLM agents through the DeGroot framework, where agents exchange text-based messages over diverse communication graphs. To track opinion evolution, we map each message to an opinion score via sentiment analysis. We find that agents typically reach consensus and the disagreement between the agents decays exponentially. However, the limiting opinion departs from DeGroot's network-centrality-weighted forecast. The consensus between LLM agents turns out to be largely insensitive to initial conditions and instead depends strongly on the discussion subject and inherent biases. Nevertheless, transient dynamics align with classical graph theory and the convergence rate of opinions is closely related to the second-largest eigenvalue of the graph's combination matrix. Together, these findings can be useful for LLM-driven social-network simulations and the design of resource-efficient multi-agent LLM applications.

Existing multi-expert LLM systems gather diverse perspectives but combine them through simple aggregation, obscuring which arguments drove the final decision. We introduce ARGORA, a framework that organizes multi-expert discussions into explicit argumentation graphs showing which arguments support or attack each other. By casting these graphs as causal models, ARGORA can systematically remove individual arguments and recompute outcomes, identifying which reasoning chains were necessary and whether decisions would change under targeted modifications. We further introduce a correction mechanism that aligns internal reasoning with external judgments when they disagree. Across diverse benchmarks and an open-ended use case, ARGORA achieves competitive accuracy and demonstrates corrective behavior: when experts initially disagree, the framework resolves disputes toward correct answers more often than it introduces new errors, while providing causal diagnostics of decisive arguments.

Visual token compression is widely used to accelerate large vision-language models (LVLMs) by pruning or merging visual tokens, yet its adversarial robustness remains unexplored. We show that existing encoder-based attacks can substantially overestimate the robustness of compressed LVLMs, due to an optimization-inference mismatch: perturbations are optimized on the full-token representation, while inference is performed through a token-compression bottleneck. To address this gap, we propose the Compression-AliGnEd attack (CAGE), which aligns perturbation optimization with compression inference without assuming access to the deployed compression mechanism or its token budget. CAGE combines (i) expected feature disruption, which concentrates distortion on tokens likely to survive across plausible budgets, and (ii) rank distortion alignment, which actively aligns token distortions with rank scores to promote the retention of highly distorted evidence. Across diverse representative plug-and-play compression mechanisms and datasets, our results show that CAGE consistently achieves lower robust accuracy than the baseline. This work highlights that robustness assessments ignoring compression can be overly optimistic, calling for compression-aware security evaluation and defenses for efficient LVLMs.

Hyperbolic space is quickly gaining traction as a promising geometry for hierarchical and robust representation learning. A core open challenge is the development of a mathematical formulation of hyperbolic neural networks that is both efficient and captures the key properties of hyperbolic space. The Lorentz model of hyperbolic space has been shown to enable both fast forward and backward propagation. However, we prove that, with the current formulation of Lorentz linear layers, the hyperbolic norms of the outputs scale logarithmically with the number of gradient descent steps, nullifying the key advantage of hyperbolic geometry. We propose a new Lorentz linear layer grounded in the well-known ``distance-to-hyperplane" formulation. We prove that our formulation results in the usual linear scaling of output hyperbolic norms with respect to the number of gradient descent steps. Our new formulation, together with further algorithmic efficiencies through Lorentzian activation functions and a new caching strategy results in neural networks fully abiding by hyperbolic geometry while simultaneously bridging the computation gap to Euclidean neural networks. Code available at: https://github.com/robertdvdk/hyperbolic-fully-connected.

Artificial intelligence (AI) has transformed materials discovery, enabling rapid exploration of chemical space through generative models and surrogate screening. Yet current AI workflows optimize performance first, deferring sustainability to post synthesis assessment. This creates inefficiency by the time environmental burdens are quantified, resources have been invested in potentially unsustainable solutions. The disconnect between atomic scale design and lifecycle assessment (LCA) reflects fundamental challenges, data scarcity across heterogeneous sources, scale gaps from atoms to industrial systems, uncertainty in synthesis pathways, and the absence of frameworks that co-optimize performance with environmental impact. We propose to integrate upstream machine learning (ML) assisted materials discovery with downstream lifecycle assessment into a uniform ML-LCA environment. The framework ML-LCA integrates five components, information extraction for building materials-environment knowledge bases, harmonized databases linking properties to sustainability metrics, multi-scale models bridging atomic properties to lifecycle impacts, ensemble prediction of manufacturing pathways with uncertainty quantification, and uncertainty-aware optimization enabling simultaneous performance-sustainability navigation. Case studies spanning glass, cement, semiconductor photoresists, and polymers demonstrate both necessity and feasibility while identifying material-specific integration challenges. Realizing ML-LCA demands coordinated advances in data infrastructure, ex-ante assessment methodologies, multi-objective optimization, and regulatory alignment enabling the discovery of materials that are sustainable by design rather than by chance.

We introduce KAPSO, a modular framework for autonomous program synthesis and optimization. Given a natural language goal and an evaluation method, KAPSO iteratively performs ideation, code synthesis and editing, execution, evaluation, and learning to improve a runnable artifact toward measurable objectives. Rather than treating synthesis as the endpoint, KAPSO uses synthesis as an operator within a long-horizon optimization loop, where progress is defined by evaluator outcomes. KAPSO targets long-horizon failures common in coding agents, including lost experimental state, brittle debugging, and weak reuse of domain expertise, by integrating three tightly coupled components. First, a git-native experimentation engine isolates each attempt as a branch, producing reproducible artifacts and preserving provenance across iterations. Second, a knowledge system ingests heterogeneous sources, including repositories, internal playbooks, and curated external resources such as documentation, scientific papers, and web search results, and organizes them into a structured representation that supports retrieval over workflows, implementations, and environment constraints. Third, a cognitive memory layer coordinates retrieval and maintains an episodic store of reusable lessons distilled from experiment traces (run logs, diffs, and evaluator feedback), reducing repeated error modes and accelerating convergence. We evaluated KAPSO on MLE-Bench (Kaggle-style ML competitions) and ALE-Bench (AtCoder heuristic optimization), and report end-to-end performance. Code Available at: https://github.com/Leeroo-AI/kapso

Representation learning is central to many downstream tasks such as search, clustering, classification, and reranking. State-of-the-art sequence encoders typically collapse a variable-length token sequence to a single vector using a pooling operator, most commonly a special [CLS] token or mean pooling over token embeddings. In this paper, we identify systematic weaknesses of these pooling strategies: [CLS] tends to concentrate information toward the initial positions of the sequence and can under-represent distributed evidence, while mean pooling can dilute salient local signals, sometimes leading to worse short-context performance. To address these issues, we introduce Landmark (LMK) pooling, which partitions a sequence into chunks, inserts landmark tokens between chunks, and forms the final representation by mean-pooling the landmark token embeddings. This simple mechanism improves long-context extrapolation without sacrificing local salient features, at the cost of introducing a small number of special tokens. We empirically demonstrate that LMK pooling matches existing methods on short-context retrieval tasks and yields substantial improvements on long-context tasks, making it a practical and scalable alternative to existing pooling methods.

To promote cooperation in Multi-Agent Reinforcement Learning, the reward signals of all agents can be aggregated together, forming global rewards that are commonly known as the fully cooperative setting. However, global rewards are usually noisy because they contain the contributions of all agents, which have to be resolved in the credit assignment process. On the other hand, using local reward benefits from faster learning due to the separation of agents' contributions, but can be suboptimal as agents myopically optimize their own reward while disregarding the global optimality. In this work, we propose a method that combines the merits of both approaches. By using a graph of interaction between agents, our method discerns the individual agent contribution in a more fine-grained manner than a global reward, while alleviating the cooperation problem with agents' local reward. We also introduce a practical approach for approximating such a graph. Our experiments demonstrate the flexibility of the approach, enabling improvements over the traditional local and global reward settings.

The performance of large language models (LLMs) on verifiable tasks is usually measured by pass@k, the probability of answering a question correctly at least once in k trials. At a fixed budget, a more suitable metric is coverage@cost, the average number of unique questions answered as a function of the total number of attempts. We connect the two metrics and show that the empirically-observed power-law behavior in pass@k leads to a sublinear growth of the coverage@cost (diminishing returns). To solve this problem, we propose Reset-and-Discard (ReD), a query method of LLMs that increases coverage@cost for any given budget, regardless of the pass@k form. Moreover, given a pass@k, we can quantitatively predict the savings in the total number of attempts using ReD. If pass@k is not available for the model, ReD can infer its power-law exponent. Experiments on three LLMs using HumanEval demonstrate that ReD substantially reduces the required attempts, tokens, and USD cost to reach a desired coverage, while also offering an efficient way to measure inference power-laws.

Spatial covariance matrices of EEG signals are Symmetric Positive Definite (SPD) and lie on a Riemannian manifold, yet the theoretical connection between embedding geometry and optimization dynamics remains unexplored. We provide a formal analysis linking embedding choice to gradient conditioning and numerical stability for SPD manifolds, establishing three theoretical results: (1) BWSPD's $\sqrtκ$ gradient conditioning (vs $κ$ for Log-Euclidean) via Daleckii-Kreĭn matrices provides better gradient conditioning on high-dimensional inputs ($d \geq 22$), with this advantage reducing on low-dimensional inputs ($d \leq 8$) where eigendecomposition overhead dominates; (2) Embedding-Space Batch Normalization (BN-Embed) approximates Riemannian normalization up to $O(\varepsilon^2)$ error, yielding $+26\%$ accuracy on 56-channel ERP data but negligible effect on 8-channel SSVEP data, matching the channel-count-dependent prediction; (3) bi-Lipschitz bounds prove BWSPD tokens preserve manifold distances with distortion governed solely by the condition ratio $κ$. We validate these predictions via a unified Transformer framework comparing BWSPD, Log-Euclidean, and Euclidean embeddings within identical architecture across 1,500+ runs on three EEG paradigms (motor imagery, ERP, SSVEP; 36 subjects). Our Log-Euclidean Transformer achieves state-of-the-art performance on all datasets, substantially outperforming classical Riemannian classifiers and recent SPD baselines, while BWSPD offers competitive accuracy with similar training time.

Many-Task Learning refers to the setting where a large number of related tasks need to be learned, the exact relationships between tasks are not known. We introduce the Cascaded Transfer Learning, a novel many-task transfer learning paradigm where information (e.g. model parameters) cascades hierarchically through tasks that are learned by individual models of the same class, while respecting given budget constraints. The cascade is organized as a rooted tree that specifies the order in which tasks are learned and refined. We design a cascaded transfer mechanism deployed over a minimum spanning tree structure that connects the tasks according to a suitable distance measure, and allocates the available training budget along its branches. Experiments on synthetic and real many-task settings show that the resulting method enables more accurate and cost effective adaptation across large task collections compared to alternative approaches.

Arabic is a linguistically and culturally rich language with a vast vocabulary that spans scientific, religious, and literary domains. Yet, large-scale lexical datasets linking Arabic words to precise definitions remain limited. We present MURAD (Multi-domain Unified Reverse Arabic Dictionary), an open lexical dataset with 96,243 word-definition pairs. The data come from trusted reference works and educational sources. Extraction used a hybrid pipeline integrating direct text parsing, optical character recognition, and automated reconstruction. This ensures accuracy and clarity. Each record aligns a target word with its standardized Arabic definition and metadata that identifies the source domain. The dataset covers terms from linguistics, Islamic studies, mathematics, physics, psychology, and engineering. It supports computational linguistics and lexicographic research. Applications include reverse dictionary modeling, semantic retrieval, and educational tools. By releasing this resource, we aim to advance Arabic natural language processing and promote reproducible research on Arabic lexical semantics.

Large language models have enabled automated algorithm design (AAD) by generating optimization algorithms directly from natural-language prompts. While evolutionary frameworks such as LLaMEA demonstrate strong exploratory capabilities across the algorithm design space, their search dynamics are entirely driven by fitness feedback, leaving substantial information about the generated code unused. We propose a mechanism for guiding AAD using feedback constructed from graph-theoretic and complexity features extracted from the abstract syntax trees of the generated algorithms, based on a surrogate model learned over an archive of evaluated solutions. Using explainable AI techniques, we identify features that substantially affect performance and translate them into natural-language mutation instructions that steer subsequent LLM-based code generation without restricting expressivity. We propose LLaMEA-SAGE, which integrates this feature-driven guidance into LLaMEA, and evaluate it across several benchmarks. We show that the proposed structured guidance achieves the same performance faster than vanilla LLaMEA in a small controlled experiment. In a larger-scale experiment using the MA-BBOB suite from the GECCO-MA-BBOB competition, our guided approach achieves superior performance compared to state-of-the-art AAD methods. These results demonstrate that signals derived from code can effectively bias LLM-driven algorithm evolution, bridging the gap between code structure and human-understandable performance feedback in automated algorithm design.

Controlling the behavior of large language models (LLMs) at inference time is essential for aligning outputs with human abilities and safety requirements. \emph{Activation steering} provides a lightweight alternative to prompt engineering and fine-tuning by directly modifying internal activations to guide generation. This research advances the literature in three significant directions. First, while previous work demonstrated the technical feasibility of steering emotional tone using automated classifiers, this paper presents the first human evaluation of activation steering concerning the emotional tone of LLM outputs, collecting over 7,000 crowd-sourced ratings from 190 participants via Prolific ($n=190$). These ratings assess both perceived emotional intensity and overall text quality. Second, we find strong alignment between human and model-based quality ratings (mean $r=0.776$, range $0.157$--$0.985$), indicating automatic scoring can proxy perceived quality. Moderate steering strengths ($λ\approx 0.15$) reliably amplify target emotions while preserving comprehensibility, with the strongest effects for disgust ($η_p^2 = 0.616$) and fear ($η_p^2 = 0.540$), and minimal effects for surprise ($η_p^2 = 0.042$). Finally, upgrading from Alpaca to LlaMA-3 yielded more consistent steering with significant effects across emotions and strengths (all $p < 0.001$). Inter-rater reliability was high (ICC $= 0.71$--$0.87$), underscoring the robustness of the findings. These findings support activation-based control as a scalable method for steering LLM behavior across affective dimensions.

Large Language Models (LLMs) excel across diverse domains but suffer from high energy costs due to quadratic attention and dense Feed-Forward Network (FFN) operations. To address these issues, we propose Module-aware Architecture Refinement (MAR), a two-stage framework that integrates State Space Models (SSMs) for linear-time sequence modeling and applies activation sparsification to reduce FFN costs. In addition, to mitigate low information density and temporal mismatch in integrating Spiking Neural Networks (SNNs) with SSMs, we design the Adaptive Ternary Multi-step Neuron (ATMN) and the Spike-aware Bidirectional Distillation Strategy (SBDS). Extensive experiments demonstrate that MAR effectively restores the performance of its dense counterpart under constrained resources while substantially reducing inference energy consumption. Furthermore, it outperforms efficient models of comparable or even larger scale, underscoring its potential for building efficient and practical LLMs.

In many applications of LLMs, natural language responses often have an underlying structure such as representing discrete labels, numerical values, or graphs. Yet, existing decoding and uncertainty estimation methods operate only in language space and largely disregard structural information. We address this by modeling LLM outputs directly in a task-dependent latent structure. By equipping this structure with a dissimilarity measure, we can compute Bayes-optimal responses. These are not selected from sampled generations but are newly synthesized by combining individual responses in the latent space. Across different tasks, Bayes-optimal responses consistently outperform standard decoding methods like beam search. Moreover, quantifying uncertainty via the induced Bayesian risk captures variations in terms of the latent structure and improves alignment with output quality and correctness. Our decision-theoretic framework is applicable to any problem that admits a latent response structure and enables reliable task-aware LLM predictions.

Recent advancements in Generative Artificial Intelligence (GenAI) have significantly enhanced the capabilities of both image generation and editing. However, current approaches often treat these tasks separately, leading to inefficiencies and challenges in maintaining spatial consistency and semantic coherence between generated content and edits. Moreover, a major obstacle is the lack of structured control over object relationships and spatial arrangements. Scene graph-based methods, which represent objects and their interrelationships in a structured format, offer a solution by providing greater control over composition and interactions in both image generation and editing. To address this, we introduce SimGraph, a unified framework that integrates scene graph-based image generation and editing, enabling precise control over object interactions, layouts, and spatial coherence. In particular, our framework integrates token-based generation and diffusion-based editing within a single scene graph-driven model, ensuring high-quality and consistent results. Through extensive experiments, we empirically demonstrate that our approach outperforms existing state-of-the-art methods.

Large language models achieve strong reasoning performance, but inference strategies such as Self-Consistency (SC) are computationally expensive, as they fully expand all reasoning traces. We introduce PoLR (Path of Least Resistance), the first inference-time method to leverage prefix consistency for compute-efficient reasoning. PoLR clusters short prefixes of reasoning traces, identifies the dominant cluster, and expands all paths in that cluster, preserving the accuracy benefits of SC while substantially reducing token usage and latency. Our theoretical analysis, framed via mutual information and entropy, explains why early reasoning steps encode strong signals predictive of final correctness. Empirically, PoLR consistently matches or exceeds SC across GSM8K, MATH500, AIME24/25, and GPQA-DIAMOND, reducing token usage by up to 60% and wall-clock latency by up to 50%. Moreover, PoLR is fully complementary to adaptive inference methods (e.g., Adaptive Consistency, Early-Stopping SC) and can serve as a drop-in pre-filter, making SC substantially more efficient and scalable without requiring model fine-tuning.

Norm-constrained linear minimization oracle (LMO)-based optimizers such as spectral gradient descent and Muon are attractive in large-scale learning, but extending them to manifold-constrained problems is nontrivial and often leads to nested-loop schemes that solve tangent-space subproblems iteratively. We propose \emph{Manifold Constrained Steepest Descent} (MCSD), a single-loop framework for optimization over manifolds that selects a norm-induced steepest-descent direction via an LMO applied to the Riemannian gradient, and then returns to the manifold via projection. Under standard smoothness assumptions, we establish convergence guarantees for MCSD and a stochastic momentum variant. We further introduce \emph{SPEL}, the spectral-norm specialization of MCSD on the Stiefel manifold, which admits scalable implementations via fast matrix sign computations. Experiments on PCA, orthogonality-constrained CNNs, and manifold-constrained LLM adapter tuning demonstrate improved stability and competitive performance relative to standard Riemannian baselines and existing manifold-aware LMO methods.

Reinforcement Learning (RL) post-training alignment for language models is effective, but also costly and unstable in practice, owing to its complicated training process. To address this, we propose a training-free inference method to sample directly from the optimal RL policy. The transition probability applied to Masked Language Modeling (MLM) consists of a reference policy model and an energy term. Based on this, our algorithm, Energy-Guided Test-Time Scaling (ETS), estimates the key energy term via online Monte Carlo, with a provable convergence rate. Moreover, to ensure practical efficiency, ETS leverages modern acceleration frameworks alongside tailored importance sampling estimators, substantially reducing inference latency while provably preserving sampling quality. Experiments on MLM (including autoregressive models and diffusion language models) across reasoning, coding, and science benchmarks show that our ETS consistently improves generation quality, validating its effectiveness and design.

Stance detection is an established task that classifies an author's attitude toward a specific target into categories such as Favor, Neutral, and Against. Beyond categorical stance labels, we leverage a long-established affective science framework to model stance along real-valued dimensions of valence (negative-positive) and arousal (calm-active). This dimensional approach captures nuanced affective states underlying stance expressions, enabling fine-grained stance analysis. To this end, we introduce DimStance, the first dimensional stance resource with valence-arousal (VA) annotations. This resource comprises 11,746 target aspects in 7,365 texts across five languages (English, German, Chinese, Nigerian Pidgin, and Swahili) and two domains (politics and environmental protection). To facilitate the evaluation of stance VA prediction, we formulate the dimensional stance regression task, analyze cross-lingual VA patterns, and benchmark pretrained and large language models under regression and prompting settings. Results show competitive performance of fine-tuned LLM regressors, persistent challenges in low-resource languages, and limitations of token-based generation. DimStance provides a foundation for multilingual, emotion-aware, stance analysis and benchmarking.

Mean-field control (MFC) offers a scalable solution to the curse of dimensionality in multi-agent systems but traditionally hinges on the restrictive assumption of exchangeability via dense, all-to-all interactions. In this work, we bridge the gap to real-world network structures by proposing a rigorous framework for MFC on large sparse graphs. We redefine the system state as a probability measure over decorated rooted neighborhoods, effectively capturing local heterogeneity. Our central contribution is a theoretical foundation for scalable reinforcement learning in this setting. We prove horizon-dependent locality: for finite-horizon problems, an agent's optimal policy at time t depends strictly on its (T-t)-hop neighborhood. This result renders the infinite-dimensional control problem tractable and underpins a novel Dynamic Programming Principle (DPP) on the lifted space of neighborhood distributions. Furthermore, we formally and experimentally justify the use of Graph Neural Networks (GNNs) for actor-critic algorithms in this context. Our framework naturally recovers classical MFC as a degenerate case while enabling efficient, theoretically grounded control on complex sparse topologies.

On-policy reinforcement learning (RL) methods widely used for language model post-training, like Group Relative Policy Optimization (GRPO), often suffer from limited exploration and early saturation due to low sampling diversity. While off-policy data can help, current approaches that mix entire trajectories cause significant policy mismatch and instability. In this work, we propose the $\textbf{S}$ingle-sample Mix-p$\textbf{O}$licy $\textbf{U}$nified $\textbf{P}$aradigm (SOUP), a framework that unifies off- and on-policy learning within individual samples at the token level. It confines off-policy influence to the prefix of a generated sequence sampled from historical policies, while the continuation is generated on-policy. Through token-level importance ratios, SOUP effectively leverages off-policy information while preserving training stability. Extensive experiments demonstrate that SOUP consistently outperforms standard on-policy training and existing off-policy extensions. Our further analysis clarifies how our fine-grained, single-sample mix-policy training can improve both exploration and final performance in LLM RL.

Handcrafted optimizers become prohibitively inefficient for complex black-box optimization (BBO) tasks. MetaBBO addresses this challenge by meta-learning to automatically configure optimizers for low-level BBO tasks, thereby eliminating heuristic dependencies. However, existing methods typically require extensive handcrafted training tasks to learn meta-strategies that generalize to target tasks, which poses a critical limitation for realistic applications with unknown task distributions. To overcome the issue, we propose the Adaptive meta Black-box Optimization Model (ABOM), which performs online parameter adaptation using solely optimization data from the target task, obviating the need for predefined task distributions. Unlike conventional metaBBO frameworks that decouple meta-training and optimization phases, ABOM introduces a closed-loop adaptive parameter learning mechanism, where parameterized evolutionary operators continuously self-update by leveraging generated populations during optimization. This paradigm shift enables zero-shot optimization: ABOM achieves competitive performance on synthetic BBO benchmarks and realistic unmanned aerial vehicle path planning problems without any handcrafted training tasks. Visualization studies reveal that parameterized evolutionary operators exhibit statistically significant search patterns, including natural selection and genetic recombination.

LLM-based multi-agent simulations are increasingly adopted across application domains, but remain difficult to scale due to GPU memory pressure. Each agent maintains private GPU-resident states, including models, prefix caches, and adapters, which quickly exhaust device memory as the agent count grows. We identify two key properties of these workloads: sparse agent activation and an estimable agent invocation order. Based on an analysis of representative workload classes, we introduce invocation distance, a unified abstraction that estimates the relative order in which agents will issue future LLM requests. Leveraging this abstraction, we present ScaleSim, a memory-efficient LLM serving system for large-scale multi-agent simulations. ScaleSim enables proactive prefetching and priority-based eviction, supports diverse agent-specific memory through a modular interface, and achieves up to 1.74x speedup over SGLang on simulation benchmarks.

We study fixed-confidence best-arm identification (BAI) where a cheap but potentially biased proxy (e.g., LLM judge) is available for every sample, while an expensive ground-truth label can only be acquired selectively when using a human for auditing. Unlike classical multi-fidelity BAI, the proxy is biased (arm- and context-dependent) and ground truth is selectively observed. Consequently, standard multi-fidelity methods can mis-select the best arm, and uniform auditing, though accurate, wastes scarce resources and is inefficient. We prove that without bias correction and propensity adjustment, mis-selection probability may not vanish (even with unlimited proxy data). We then develop an estimator for the mean of each arm that combines proxy scores with inverse-propensity-weighted residuals and form anytime-valid confidence sequences for that estimator. Based on the estimator and confidence sequence, we propose an algorithm that adaptively selects and audits arms. The algorithm concentrates audits on unreliable contexts and close arms and we prove that a plug-in Neyman rule achieves near-oracle audit efficiency. Numerical experiments confirm the theoretical guarantees and demonstrate the superior empirical performance of the proposed algorithm.

We study semi-supervised stochastic optimization when labeled data is scarce but predictions from pre-trained models are available. PPI and SVRG both reduce variance through control variates -- PPI uses predictions, SVRG uses reference gradients. We show they are mathematically equivalent and develop PPI-SVRG, which combines both. Our convergence bound decomposes into the standard SVRG rate plus an error floor from prediction uncertainty. The rate depends only on loss geometry; predictions affect only the neighborhood size. When predictions are perfect, we recover SVRG exactly. When predictions degrade, convergence remains stable but reaches a larger neighborhood. Experiments confirm the theory: PPI-SVRG reduces MSE by 43--52\% under label scarcity on mean estimation benchmarks and improves test accuracy by 2.7--2.9 percentage points on MNIST with only 10\% labeled data.

While Large Language Models (LLMs) have catalyzed breakthroughs in automated code generation, Small Language Models (SLMs) often encounter reasoning bottlenecks and failure loops when addressing complex logical requirements. To overcome these challenges, we propose DebateCoder, a multi-agent collaborative framework designed to improve the reasoning ability of SLMs (e.g., Pangu-1B) in resource-constrained environments. DebateCoder uses a structured role-playing protocol with three agents: User Agent (A_UA), Technical Agent (A_TA), and Quality Assurance Agent (A_QA). It also includes an Adaptive Confidence Gating mechanism with a 95% threshold to balance accuracy and inference efficiency. In addition, we introduce a multi-turn deliberation module and a reviewer-guided analytical debugging loop for orthogonal pre-generation debate and post-generation refinement. Experiments on HumanEval and MBPP show that DebateCoder achieves 70.12% Pass@1 on HumanEval, outperforming MapCoder while reducing API overhead by about 35%. These results indicate that collaborative protocols can mitigate limitations of small-parameter models and provide a scalable, efficient approach to high-quality automated software engineering.

Long-horizon agentic reasoning necessitates effectively compressing growing interaction histories into a limited context window. Most existing memory systems serialize history as text, where token-level cost is uniform and scales linearly with length, often spending scarce budget on low-value details. To this end, we introduce MemOCR, a multimodal memory agent that improves long-horizon reasoning under tight context budgets by allocating memory space with adaptive information density through visual layout. Concretely, MemOCR maintains a structured rich-text memory (e.g., headings, highlights) and renders it into an image that the agent consults for memory access, visually prioritizing crucial evidence while aggressively compressing auxiliary details. To ensure robustness across varying memory budgets, we train MemOCR with reinforcement learning under budget-aware objectives that expose the agent to diverse compression levels. Across long-context multi-hop and single-hop question-answering benchmarks, MemOCR outperforms strong text-based baselines and achieves more effective context utilization under extreme budgets.

Block-coordinate descent (BCD) is the method of choice to solve numerous large scale optimization problems, however their theoretical study for non-convex optimization, has received less attention. In this paper, we present a new block-coordinate descent (BCD) framework to tackle non-convex composite optimization problems, ensuring decrease of the objective function and convergence to a solution. This framework is general enough to include variable metric proximal gradient updates, proximal Newton updates, and alternated minimization updates. This generality allows to encompass three versions of the most used solvers in the sparse precision matrix estimation problem, deemed Graphical Lasso: graphical ISTA, Primal GLasso, and QUIC. We demonstrate the value of this new framework on non-convex sparse precision matrix estimation problems, providing convergence guarantees and up to a $100$-fold reduction in the number of iterations required to reach state-of-the-art estimation quality.

Text clustering is today the most popular paradigm for topic modelling, both in academia and industry. Despite clustering topic models' apparent success, we identify a number of issues in Top2Vec and BERTopic, which remain largely unsolved. Firstly, these approaches are unreliable at discovering natural clusters in corpora, due to extreme sensitivity to sample size and hyperparameters, the default values of which result in suboptimal behaviour. Secondly, when estimating term importance, BERTopic ignores the semantic distance of keywords to topic vectors, while Top2Vec ignores word counts in the corpus. This results in, on the one hand, less coherent topics due to the presence of stop words and junk words, and lack of variety and trust on the other. In this paper, I introduce a new approach, \textbf{Topeax}, which discovers the number of clusters from peaks in density estimates, and combines lexical and semantic indices of term importance to gain high-quality topic keywords. Topeax is demonstrated to be better at both cluster recovery and cluster description than Top2Vec and BERTopic, while also exhibiting less erratic behaviour in response to changing sample size and hyperparameters.

Training large language models (LLMs) for non-verifiable tasks, such as creative writing, dialogue, and ethical reasoning, remains challenging due to the absence of ground-truth labels. While LLM-as-Judge approaches offer a scalable alternative to human feedback, they face a fundamental limitation: performance is constrained by the evaluator's own quality. If the judge cannot recognize good solutions, it cannot provide useful training signals, and evaluation biases (e.g., favoring verbosity over quality) remain unaddressed. This motivates meta-evaluation: the ability to evaluate and improve the evaluator itself. We introduce CoNL, a framework that unifies generation, evaluation, and meta-evaluation through multi-agent self-play. Our key insight: critique quality can be measured by whether it helps others improve their solutions. In CoNL, multiple agents sharing the same policy engage in structured conversations to propose, critique, and revise solutions. Critiques that enable solution improvements earn a diagnostic reward, creating explicit supervision for meta-evaluation and enabling joint optimization of generation and judging capabilities through self-play, without external judges or ground truth. Experiments on five benchmarks show that CoNL achieves consistent improvements over self-rewarding baselines while maintaining stable training.

Speech editing achieves semantic inversion by performing fine-grained segment-level manipulation on original utterances, while preserving global perceptual naturalness. Existing detection studies mainly focus on manually edited speech with explicit splicing artifacts, and therefore struggle to cope with emerging end-to-end neural speech editing techniques that generate seamless acoustic transitions. To address this challenge, we first construct a large-scale bilingual dataset, AiEdit, which leverages large language models to drive precise semantic tampering logic and employs multiple advanced neural speech editing methods for data synthesis, thereby filling the gap of high-quality speech editing datasets. Building upon this foundation, we propose PELM (Prior-Enhanced Audio Large Language Model), the first large-model framework that unifies speech editing detection and content localization by formulating them as an audio question answering task. To mitigate the inherent forgery bias and semantic-priority bias observed in existing audio large models, PELM incorporates word-level probability priors to provide explicit acoustic cues, and further designs a centroid-aggregation-based acoustic consistency perception loss to explicitly enforce the modeling of subtle local distribution anomalies. Extensive experimental results demonstrate that PELM significantly outperforms state-of-the-art methods on both the HumanEdit and AiEdit datasets, achieving equal error rates (EER) of 0.57\% and 9.28\% (localization), respectively.

We study partial-feedback online learning, where each instance admits a set of correct labels, but the learner only observes one correct label per round; any prediction within the correct set is counted as correct. This model captures settings such as language generation, where multiple responses may be valid but data provide only a single reference. We give a near-complete characterization of minimax regret for both deterministic and randomized learners in the set-realizable regime, i.e., in the regime where sublinear regret is generally attainable. For deterministic learners, we introduce the Partial-Feedback Littlestone dimension (PFLdim) and show it precisely governs learnability and minimax regret; technically, PFLdim cannot be defined via the standard version space, requiring a new collection version space viewpoint and an auxiliary dimension used only in the proof. We further develop the Partial-Feedback Measure Shattering dimension (PMSdim) to obtain tight bounds for randomized learners. We identify broad conditions ensuring inseparability between deterministic and randomized learnability (e.g., finite Helly number or nested-inclusion label structure), and extend the argument to set-valued online learning, resolving an open question of Raman et al. [2024b]. Finally, we show a sharp separation from weaker realistic and agnostic variants: outside set realizability, the problem can become information-theoretically intractable, with linear regret possible even for $|H|=2$. This highlights the need for fundamentally new, noise-sensitive complexity measures to meaningfully characterize learnability beyond set realizability.

Modern sparse language models typically achieve sparsity through Mixture-of-Experts (MoE) layers, which dynamically route tokens to dense MLP "experts." However, dynamic hard routing has a number of drawbacks, such as potentially poor hardware efficiency and needing auxiliary losses for stable training. In contrast, the tokenizer embedding table, which is natively sparse, largely avoids these issues by selecting a single embedding per token at the cost of not having contextual information. In this work, we introduce the Large Lookup Layer (L$^3$), which unlocks a new axis of sparsity by generalizing embedding tables to model decoder layers. L$^3$ layers use static token-based routing to aggregate a set of learned embeddings per token in a context-dependent way, allowing the model to efficiently balance memory and compute by caching information in embeddings. L$^3$ has two main components: (1) a systems-friendly architecture that allows for fast training and CPU-offloaded inference with no overhead, and (2) an information-theoretic embedding allocation algorithm that effectively balances speed and quality. We empirically test L$^3$ by training transformers with up to 2.6B active parameters and find that L$^3$ strongly outperforms both dense models and iso-sparse MoEs in both language modeling and downstream tasks.

LLM role-playing, i.e., using LLMs to simulate specific personas, has emerged as a key capability in various applications, such as companionship, content creation, and digital games. While current models effectively capture character tones and knowledge, simulating the inner thoughts behind their behaviors remains a challenge. Towards cognitive simulation in LLM role-play, previous efforts mainly suffer from two deficiencies: data with high-quality reasoning traces, and reliable reward signals aligned with human preferences. In this paper, we propose HER, a unified framework for cognitive-level persona simulation. HER introduces dual-layer thinking, which distinguishes characters' first-person thinking from LLMs' third-person thinking. To bridge these gaps, we curate reasoning-augmented role-playing data via reverse engineering and construct human-aligned principles and reward models. Leveraging these resources, we train \method models based on Qwen3-32B via supervised and reinforcement learning. Extensive experiments validate the effectiveness of our approach. Notably, our models significantly outperform the Qwen3-32B baseline, achieving a 30.26 improvement on the CoSER benchmark and a 14.97 gain on the Minimax Role-Play Bench. Our datasets, principles, and models will be released to facilitate future research.

Conformal prediction (CP) has become a cornerstone of distribution-free uncertainty quantification, conventionally evaluated by its coverage and interval length. This work critically examines the sufficiency of these standard metrics. We demonstrate that the interval length might be deceptively improved through a counter-intuitive approach termed Prejudicial Trick (PT), while the coverage remains valid. Specifically, for any given test sample, PT probabilistically returns an interval, which is either null or constructed using an adjusted confidence level, thereby preserving marginal coverage. While PT potentially yields a deceptively lower interval length, it introduces practical vulnerabilities: the same input can yield completely different prediction intervals across repeated runs of the algorithm. We formally derive the conditions under which PT achieves these misleading improvements and provides extensive empirical evidence across various regression and classification tasks. Furthermore, we introduce a new metric interval stability which helps detect whether a new CP method implicitly improves the length based on such PT-like techniques.

Recently, the rapid advancement of multimodal domains has driven a data-centric paradigm shift in graph ML, transitioning from text-attributed to multimodal-attributed graphs. This advancement significantly enhances data representation and expands the scope of graph downstream tasks, such as modality-oriented tasks, thereby improving the practical utility of graph ML. Despite its promise, limitations exist in the current neural paradigms: (1) Neglect Context in Modality Alignment: Most existing methods adopt topology-constrained or modality-specific operators as tokenizers. These aligners inevitably neglect graph context and inhibit modality interaction, resulting in suboptimal alignment. (2) Lack of Adaptation in Modality Fusion: Most existing methods are simple adaptations for 2-modality graphs and fail to adequately exploit aligned tokens equipped with topology priors during fusion, leading to poor generalizability and performance degradation. To address the above issues, we propose LION (c\underline{LI}ff\underline{O}rd \underline{N}eural paradigm) based on the Clifford algebra and decoupled graph neural paradigm (i.e., propagation-then-aggregation) to implement alignment-then-fusion in multimodal-attributed graphs. Specifically, we first construct a modality-aware geometric manifold grounded in Clifford algebra. This geometric-induced high-order graph propagation efficiently achieves modality interaction, facilitating modality alignment. Then, based on the geometric grade properties of aligned tokens, we propose adaptive holographic aggregation. This module integrates the energy and scale of geometric grades with learnable parameters to improve modality fusion. Extensive experiments on 9 datasets demonstrate that LION significantly outperforms SOTA baselines across 3 graph and 3 modality downstream tasks.

While works such as OneRec have validated the scaling laws of Large Language Models (LLMs) in recommender systems, they rely on a cumbersome separate vocabulary. This dependency prevents the model architecture from reusing native LLM vocabularies, resulting in high maintenance costs and poor scalability. In response, we aim to efficiently reuse open-source LLM architectures without constructing a separate tokenization vocabulary. Furthermore, we identify that the optimization strategy of OneRec Gradient Bounded Policy Optimization (GBPO),suffers from a "Symmetric Conservatism" problem: its static gradient boundaries structurally suppress the update momentum required for cold-start items and fail to prevent diversity collapse in high-noise environments.To address this issue, we propose SAGE (Sequence-level Adaptive Gradient Evolution), a unified optimization framework tailored for list-wise generative recommendation. SAGE introduces two key innovations:(1) Sequence-level Signal Decoupling: By combining a geometric mean importance ratio with decoupled multi-objective advantages, we eliminate token-level variance and resolve the "Reward Collapse" problem. (2) Asymmetric Adaptive Dynamics: We construct a dynamic gradient manifold that applies a "Boost Factor" to high-potential cold start items to achieve super-linear updates and employs an "Entropy Aware Penalty" to break information cocoons. Theoretical analysis and empirical results demonstrate that SAGE effectively unblocks cold-start traffic and sustains recommendation diversity, all while retaining the numerical stability of GBPO.

Scaling data volume and diversity is critical for generalizing embodied intelligence. While synthetic data generation offers a scalable alternative to expensive physical data acquisition, existing pipelines remain fragmented and task-specific. This isolation leads to significant engineering inefficiency and system instability, failing to support the sustained, high-throughput data generation required for foundation model training. To address these challenges, we present Nimbus, a unified synthetic data generation framework designed to integrate heterogeneous navigation and manipulation pipelines. Nimbus introduces a modular four-layer architecture featuring a decoupled execution model that separates trajectory planning, rendering, and storage into asynchronous stages. By implementing dynamic pipeline scheduling, global load balancing, distributed fault tolerance, and backend-specific rendering optimizations, the system maximizes resource utilization across CPU, GPU, and I/O resources. Our evaluation demonstrates that Nimbus achieves a 2-3X improvement in end-to-end throughput compared to unoptimized baselines and ensuring robust, long-term operation in large-scale distributed environments. This framework serves as the production backbone for the InternData suite, enabling seamless cross-domain data synthesis.

While Large Language Models (LLMs) show significant potential in hardware engineering, current benchmarks suffer from saturation and limited task diversity, failing to reflect LLMs' performance in real industrial workflows. To address this gap, we propose a comprehensive benchmark for AI-aided chip design that rigorously evaluates LLMs across three critical tasks: Verilog generation, debugging, and reference model generation. Our benchmark features 44 realistic modules with complex hierarchical structures, 89 systematic debugging cases, and 132 reference model samples across Python, SystemC, and CXXRTL. Evaluation results reveal substantial performance gaps, with state-of-the-art Claude-4.5-opus achieving only 30.74\% on Verilog generation and 13.33\% on Python reference model generation, demonstrating significant challenges compared to existing saturated benchmarks where SOTA models achieve over 95\% pass rates. Additionally, to help enhance LLM reference model generation, we provide an automated toolbox for high-quality training data generation, facilitating future research in this underexplored domain. Our code is available at https://github.com/zhongkaiyu/ChipBench.git.

Illicit financial activities such as money laundering often manifest through recurrent topological patterns in transaction networks. Detecting these patterns automatically remains challenging due to the scarcity of labeled real-world data and strict privacy constraints. To address this, we investigate whether Graph Autoencoders (GAEs) can effectively learn and distinguish topological patterns that mimic money laundering operations when trained on synthetic data. The analysis consists of two phases: (i) data generation, where synthetic samples are created for seven well-known illicit activity patterns using parametrized generators that preserve structural consistency while introducing realistic variability; and (ii) model training and validation, where separate GAEs are trained on each pattern without explicit labels, relying solely on reconstruction error as an indicator of learned structure. We compare three GAE implementations based on three distinct convolutional layers: Graph Convolutional (GAE-GCN), GraphSAGE (GAE-SAGE), and Graph Attention Network (GAE-GAT). Experimental results show that GAE-GCN achieves the most consistent reconstruction performance across patterns, while GAE-SAGE and GAE-GAT exhibit competitive results only in few specific patterns. These findings suggest that graph-based representation learning on synthetic data provides a viable path toward developing AI-driven tools for detecting illicit behaviors, overcoming the limitations of financial datasets.

The efficiency of long-video inference remains a critical bottleneck, mainly due to the dense computation in the prefill stage of Large Multimodal Models (LMMs). Existing methods either compress visual embeddings or apply sparse attention on a single GPU, yielding limited acceleration or degraded performance and restricting LMMs from handling longer, more complex videos. To overcome these issues, we propose Spava, a sequence-parallel framework with optimized attention that accelerates long-video inference across multiple GPUs. By distributing approximate attention, Spava reduces computation and increases parallelism, enabling efficient processing of more visual embeddings without compression and thereby improving task performance. System-level optimizations, such as load balancing and fused forward passes, further unleash the potential of Spava, delivering speedups of 12.72x, 1.70x, and 1.18x over FlashAttn, ZigZagRing, and APB, without notable performance loss. Code available at https://github.com/thunlp/APB

While Large Language Models (LLMs) are widely documented to be sensitive to minor prompt perturbations and prone to sycophantic alignment with user biases, their robustness in consequential, rule-bound decision-making remains under-explored. In this work, we uncover a striking "Paradox of Robustness": despite their known lexical brittleness, instruction-tuned LLMs exhibit a behavioral and near-total invariance to emotional framing effects. Using a novel controlled perturbation framework across three high-stakes domains (healthcare, law, and finance), we quantify a robustness gap where LLMs demonstrate 110-300 times greater resistance to narrative manipulation than human subjects. Specifically, we find a near-zero effect size for models (Cohen's h = 0.003) compared to the substantial biases observed in humans (Cohen's h in [0.3, 0.8]). This result is highly counterintuitive and suggests the mechanisms driving sycophancy and prompt sensitivity do not necessarily translate to a failure in logical constraint satisfaction. We show that this invariance persists across models with diverse training paradigms. Our findings show that while LLMs may be "brittle" to how a query is formatted, they are remarkably "stable" against why a decision should be biased. Our findings establish that instruction-tuned models can decouple logical rule-adherence from persuasive narratives, offering a source of decision stability that complements, and even potentially de-biases, human judgment in institutional contexts. We release the 162-scenario benchmark, code, and data to facilitate the rigorous evaluation of narrative-induced bias and robustness on GitHub.com.

Driven by the evolution toward 6G and AI-native edge intelligence, network operations increasingly require predictive and risk-aware adaptation under stringent computation and latency constraints. Network Traffic Matrix (TM), which characterizes flow volumes between nodes, is a fundamental signal for proactive traffic engineering. However, accurate TM forecasting remains challenging due to the stochastic, non-linear, and bursty nature of network dynamics. Existing discriminative models often suffer from over-smoothing and provide limited uncertainty awareness, leading to poor fidelity under extreme bursts. To address these limitations, we propose LEAD, a Large Language Model (LLM)-Enhanced Adapter-based conditional Diffusion model. First, LEAD adopts a "Traffic-to-Image" paradigm to transform traffic matrices into RGB images, enabling global dependency modeling via vision backbones. Then, we design a "Frozen LLM with Trainable Adapter" model, which efficiently captures temporal semantics with limited computational cost. Moreover, we propose a Dual-Conditioning Strategy to precisely guide a diffusion model to generate complex, dynamic network traffic matrices. Experiments on the Abilene and GEANT datasets demonstrate that LEAD outperforms all baselines. On the Abilene dataset, LEAD attains a remarkable 45.2% reduction in RMSE against the best baseline, with the error margin rising only marginally from 0.1098 at one-step to 0.1134 at 20-step predictions. Meanwhile, on the GEANT dataset, LEAD achieves a 0.0258 RMSE at 20-step prediction horizon which is 27.3% lower than the best baseline.

Advances in multi-modal large language models (MLLMs) have inspired time series understanding and reasoning tasks, that enable natural language querying over time series, producing textual analyses of complex temporal dynamics. Recent attempts hybridize numerical time series with their visualized plots, facilitating precise value reasoning and visual structure comprehension for comprehensive time series understanding of MLLMs. However, effective cross-modal integration remains challenging due to fine-grained temporal misalignment across modalities and severe entanglement between shared and modality-specific semantics, which hinder localized interpretation and complementary reasoning. To address these issues, we propose MADI, a multi-modal LLM enhanced with fine-grained alignment and disentangled interaction, featuring (1) Patch-level Alignment, which enforces physically grounded fine-grained correspondence across heterogeneous modalities, (2) Discrete Disentangled Interaction, which separates modality-common semantics into compact discrete latents and adaptively synergizes the purified modality-unique information, and (3) Critical-token Highlighting, which emphasizes informative, query-relevant signals for robust reasoning. Experiments on synthetic and real-world benchmarks show that MADI consistently outperforms general-purpose LLMs and time-series-specialized MLLMs.

When a user tells an AI system that someone "should not" take an action, the system ought to treat this as a prohibition. Yet many large language models do the opposite: they interpret negated instructions as affirmations. We audited 16 models across 14 ethical scenarios and found that open-source models endorse prohibited actions 77% of the time under simple negation and 100% under compound negation -- a 317% increase over affirmative framing. Commercial models fare better but still show swings of 19-128%. Agreement between models drops from 74% on affirmative prompts to 62% on negated ones, and financial scenarios prove twice as fragile as medical ones. These patterns hold under deterministic decoding, ruling out sampling noise. We present case studies showing how these failures play out in practice, propose the Negation Sensitivity Index (NSI) as a governance metric, and outline a tiered certification framework with domain-specific thresholds. The findings point to a gap between what current alignment techniques achieve and what safe deployment requires: models that cannot reliably distinguish "do X" from "do not X" should not be making autonomous decisions in high-stakes contexts.

Many computer vision tasks share substantial overlapping information, yet conventional codecs tend to ignore this, leading to redundant and inefficient representations. The Gray-Wyner network, a classical concept from information theory, offers a principled framework for separating common and task-specific information. Inspired by this idea, we develop a learnable three-channel codec that disentangles shared information from task-specific details across multiple vision tasks. We characterize the limits of this approach through the notion of lossy common information, and propose an optimization objective that balances inherent tradeoffs in learning such representations. Through comparisons of three codec architectures on two-task scenarios spanning six vision benchmarks, we demonstrate that our approach substantially reduces redundancy and consistently outperforms independent coding. These results highlight the practical value of revisiting Gray-Wyner theory in modern machine learning contexts, bridging classic information theory with task-driven representation learning.

Large language models allocate uniform computation across all tokens, ignoring that some sequences are trivially predictable while others require deep reasoning. We introduce ConceptMoE, which dynamically merges semantically similar tokens into concept representations, performing implicit token-level compute allocation. A learnable chunk module identifies optimal boundaries by measuring inter-token similarity, compressing sequences by a target ratio $R$ before they enter the compute-intensive concept model. Crucially, the MoE architecture enables controlled evaluation: we reallocate saved computation to match baseline activated FLOPs (excluding attention map computation) and total parameters, isolating genuine architectural benefits. Under these conditions, ConceptMoE consistently outperforms standard MoE across language and vision-language tasks, achieving +0.9 points on language pretraining, +2.3 points on long context understanding, and +0.6 points on multimodal benchmarks. When converting pretrained MoE during continual training with layer looping, gains reach +5.5 points, demonstrating practical applicability. Beyond performance, ConceptMoE reduces attention computation by up to $R^2\times$ and KV cache by $R\times$. At $R=2$, empirical measurements show prefill speedups reaching 175\% and decoding speedups up to 117\% on long sequences. The minimal architectural modifications enable straightforward integration into existing MoE, demonstrating that adaptive concept-level processing fundamentally improves both effectiveness and efficiency of large language models.

Recent advances in diffusion and flow matching models have highlighted a shift in the preferred prediction target -- moving from noise ($\varepsilon$) and velocity (v) to direct data (x) prediction -- particularly in high-dimensional settings. However, a formal explanation of why the optimal target depends on the specific properties of the data remains elusive. In this work, we provide a theoretical framework based on a generalized prediction formulation that accommodates arbitrary output targets, of which $\varepsilon$-, v-, and x-prediction are special cases. We derive the analytical relationship between data's geometry and the optimal prediction target, offering a rigorous justification for why x-prediction becomes superior when the ambient dimension significantly exceeds the data's intrinsic dimension. Furthermore, while our theory identifies dimensionality as the governing factor for the optimal prediction target, the intrinsic dimension of manifold-bound data is typically intractable to estimate in practice. To bridge this gap, we propose k-Diff, a framework that employs a data-driven approach to learn the optimal prediction parameter k directly from data, bypassing the need for explicit dimension estimation. Extensive experiments in both latent-space and pixel-space image generation demonstrate that k-Diff consistently outperforms fixed-target baselines across varying architectures and data scales, providing a principled and automated approach to enhancing generative performance.

Large Reasoning Models (LRMs) achieve explicit chain-of-thought expansion by imitating deep thinking behaviors of humans, demonstrating excellent performance in complex task scenarios. However, the deep-thinking mode often leads to unnecessarily lengthy reasoning and resource inefficiency when handling simple tasks. This overthinking phenomenon may arise from the generation preference triggered by the reward function during post-training. Existing research attempts to mitigate overthinking from the perspective of prompt design or model training, but generally underestimates the importance of task difficulty awareness, which makes it difficult for LRMs to effectively allocate reasoning resources. In this paper, we propose Difficulty-aware Policy Optimization (DiPO), a reinforcement learning-based LRM training framework. DiPO encourages LRM to spontaneously model task complexity, and integrates them into reinforcement learning framework to adjust the generation preferences introduced by post-training. A difficulty modeling method based on model self-reasoning is proposed, which significantly reduces the dependence on manual annotation and formalize task complexity. We further develop a difficulty-signal-enhanced reward function that incorporates a penalty for lengthy reasoning while considering reasoning performance and output format. Experimental results indicate that DiPO enables the model to spontaneously adjust inference overhead, significantly reducing redundant tokens without losing performance due to thought compression.

Training a unified language model that adapts between intuitive System 1 and deliberative System 2 remains challenging due to interference between their cognitive modes. Recent studies have thus pursued making System 2 models more efficient. However, these approaches focused on output control, limiting what models produce. We argue that this paradigm is misaligned: output length is merely a symptom of the model's cognitive configuration, not the root cause. In this work, we shift the focus to capability control, which modulates \textit{how models think} rather than \textit{what they produce}. To realize this, we leverage existing Instruct and Thinking checkpoints through dynamic parameter interpolation, without additional training. Our pilot study establishes that linear interpolation yields a convex, monotonic Pareto frontier, underpinned by representation continuity and structural connectivity. Building on this, we propose \textbf{DAMI} (\textbf{D}yn\textbf{A}mic \textbf{M}odel \textbf{I}nterpolation), a framework that estimates a query-specific Reasoning Intensity $λ(q)$ to configure cognitive depth. For training-based estimation, we develop a preference learning method encoding accuracy and efficiency criteria. For zero-shot deployment, we introduce a confidence-based method leveraging inter-model cognitive discrepancy. Experiments on five mathematical reasoning benchmarks demonstrate that DAMI achieves higher accuracy than the Thinking model while remaining efficient, effectively combining the efficiency of System 1 with the reasoning depth of System 2.

We introduce Statsformer, a principled framework for integrating large language model (LLM)-derived knowledge into supervised statistical learning. Existing approaches are limited in adaptability and scope: they either inject LLM guidance as an unvalidated heuristic, which is sensitive to LLM hallucination, or embed semantic information within a single fixed learner. Statsformer overcomes both limitations through a guardrailed ensemble architecture. We embed LLM-derived feature priors within an ensemble of linear and nonlinear learners, adaptively calibrating their influence via cross-validation. This design yields a flexible system with an oracle-style guarantee that it performs no worse than any convex combination of its in-library base learners, up to statistical error. Empirically, informative priors yield consistent performance improvements, while uninformative or misspecified LLM guidance is automatically downweighted, mitigating the impact of hallucinations across a diverse range of prediction tasks.

Neuroscience and artificial intelligence represent distinct yet complementary pathways to general intelligence. However, amid the ongoing boom in AI research and applications, the translational synergy between these two fields has grown increasingly elusive-hampered by a widening infrastructural incompatibility: modern AI frameworks lack native support for biophysical realism, while neural simulation tools are poorly suited for gradient-based optimization and neuromorphic hardware deployment. To bridge this gap, we introduce BrainFuse, a unified infrastructure that provides comprehensive support for biophysical neural simulation and gradient-based learning. By addressing algorithmic, computational, and deployment challenges, BrainFuse exhibits three core capabilities: (1) algorithmic integration of detailed neuronal dynamics into a differentiable learning framework; (2) system-level optimization that accelerates customizable ion-channel dynamics by up to 3,000x on GPUs; and (3) scalable computation with highly compatible pipelines for neuromorphic hardware deployment. We demonstrate this full-stack design through both AI and neuroscience tasks, from foundational neuron simulation and functional cylinder modeling to real-world deployment and application scenarios. For neuroscience, BrainFuse supports multiscale biological modeling, enabling the deployment of approximately 38,000 Hodgkin-Huxley neurons with 100 million synapses on a single neuromorphic chip while consuming as low as 1.98 W. For AI, BrainFuse facilitates the synergistic application of realistic biological neuron models, demonstrating enhanced robustness to input noise and improved temporal processing endowed by complex HH dynamics. BrainFuse therefore serves as a foundational engine to facilitate cross-disciplinary research and accelerate the development of next-generation bio-inspired intelligent systems.

Unified Multimodal Models (UMMs) integrate both visual understanding and generation within a single framework. Their ultimate aspiration is to create a cycle where understanding and generation mutually reinforce each other. While recent post-training methods have successfully leveraged understanding to enhance generation, the reverse direction of utilizing generation to improve understanding remains largely unexplored. In this work, we propose UniMRG (Unified Multi-Representation Generation), a simple yet effective architecture-agnostic post-training method. UniMRG enhances the understanding capabilities of UMMs by incorporating auxiliary generation tasks. Specifically, we train UMMs to generate multiple intrinsic representations of input images, namely pixel (reconstruction), depth (geometry), and segmentation (structure), alongside standard visual understanding objectives. By synthesizing these diverse representations, UMMs capture complementary information regarding appearance, spatial relations, and structural layout. Consequently, UMMs develop a deeper and more comprehensive understanding of visual inputs. Extensive experiments across diverse UMM architectures demonstrate that our method notably enhances fine-grained perception, reduces hallucinations, and improves spatial understanding, while simultaneously boosting generation capabilities.

In real-world data science and enterprise decision-making, critical information is often fragmented across directly queryable structured sources (e.g., SQL, CSV) and "zombie data" locked in unstructured visual documents (e.g., scanned reports, invoice images). Existing data analytics agents are predominantly limited to processing structured data, failing to activate and correlate this high-value visual information, thus creating a significant gap with industrial needs. To bridge this gap, we introduce DataCross, a novel benchmark and collaborative agent framework for unified, insight-driven analysis across heterogeneous data modalities. DataCrossBench comprises 200 end-to-end analysis tasks across finance, healthcare, and other domains. It is constructed via a human-in-the-loop reverse-synthesis pipeline, ensuring realistic complexity, cross-source dependency, and verifiable ground truth. The benchmark categorizes tasks into three difficulty tiers to evaluate agents' capabilities in visual table extraction, cross-modal alignment, and multi-step joint reasoning. We also propose the DataCrossAgent framework, inspired by the "divide-and-conquer" workflow of human analysts. It employs specialized sub-agents, each an expert on a specific data source, which are coordinated via a structured workflow of Intra-source Deep Exploration, Key Source Identification, and Contextual Cross-pollination. A novel reReAct mechanism enables robust code generation and debugging for factual verification. Experimental results show that DataCrossAgent achieves a 29.7% improvement in factuality over GPT-4o and exhibits superior robustness on high-difficulty tasks, effectively activating fragmented "zombie data" for insightful, cross-modal analysis.

Exploration is essential in reinforcement learning as an agent relies on trial and error to learn an optimal policy. However, when rewards are sparse, naive exploration strategies, like noise injection, are often insufficient. Intrinsic rewards can also provide principled guidance for exploration by, for example, combining them with extrinsic rewards to optimize a policy or using them to train subpolicies for hierarchical learning. However, the former approach suffers from unstable credit assignment, while the latter exhibits sample inefficiency and sub-optimality. We propose a policy optimization framework that leverages multiple intrinsic rewards to directly optimize a policy for an extrinsic reward without pretraining subpolicies. Our algorithm -- intrinsic reward policy optimization (IRPO) -- achieves this by using a surrogate policy gradient that provides a more informative learning signal than the true gradient in sparse-reward environments. We demonstrate that IRPO improves performance and sample efficiency relative to baselines in discrete and continuous environments, and formally analyze the optimization problem solved by IRPO. Our code is available at https://github.com/Mgineer117/IRPO.

The Job-Shop Scheduling Problem (JSSP), under various forms of manufacturing uncertainty, has recently attracted considerable research attention. Most existing studies focus on parameter uncertainty, such as variable processing times, and typically adopt the actor-critic framework. In this paper, we explore a different but prevalent form of uncertainty in JSSP: structural uncertainty. Structural uncertainty arises when a job may follow one of several routing paths, and the selection is determined not by policy, but by situational factors (e.g., the quality of intermediate products) that cannot be known in advance. Existing methods struggle to address this challenge due to incorrect credit assignment: a high-quality action may be unfairly penalized if it is followed by a time-consuming path. To address this problem, we propose a novel method named UP-AAC. In contrast to conventional actor-critic methods, UP-AAC employs an asymmetric architecture. While its actor receives a standard stochastic state, the critic is crucially provided with a deterministic state reconstructed in hindsight. This design allows the critic to learn a more accurate value function, which in turn provides a lower-variance policy gradient to the actor, leading to more stable learning. In addition, we design an attention-based Uncertainty Perception Model (UPM) to enhance the actor's scheduling decisions. Extensive experiments demonstrate that our method outperforms existing approaches in reducing makespan on benchmark instances.

Attribution and fact verification are critical challenges in natural language processing for assessing information reliability. While automated systems and Large Language Models (LLMs) aim to retrieve and select concise evidence to support or refute claims, they often present users with either insufficient or overly redundant information, leading to inefficient and error-prone verification. To address this, we propose Evidence Ranking, a novel task that prioritizes presenting sufficient information as early as possible in a ranked list. This minimizes user reading effort while still making all available evidence accessible for sequential verification. We compare two approaches for the new ranking task: one-shot ranking and incremental ranking. We introduce a new evaluation framework, inspired by information retrieval metrics, and construct a unified benchmark by aggregating existing fact verification datasets. Extensive experiments with diverse models show that incremental ranking strategies better capture complementary evidence and that LLM-based methods outperform shallower baselines, while still facing challenges in balancing sufficiency and redundancy. Compared to evidence selection, we conduct a controlled user study and demonstrate that evidence ranking both reduces reading effort and improves verification. This work provides a foundational step toward more interpretable, efficient, and user-aligned information verification systems.

Objective evaluation of synthetic speech quality remains a critical challenge. Human listening tests are the gold standard, but costly and impractical at scale. Fréchet Distance has emerged as a promising alternative, yet its reliability depends heavily on the choice of embeddings and experimental settings. In this work, we comprehensively evaluate Fréchet Speech Distance (FSD) and its variant Speech Maximum Mean Discrepancy (SMMD) under varied embeddings and conditions. We further incorporate human listening evaluations alongside TTS intelligibility and synthetic-trained ASR WER to validate the perceptual relevance of these metrics. Our findings show that WavLM Base+ features yield the most stable alignment with human ratings. While FSD and SMMD cannot fully replace subjective evaluation, we show that they can serve as complementary, cost-efficient, and reproducible measures, particularly useful when large-scale or direct listening assessments are infeasible. Code is available at https://github.com/kaen2891/FrechetSpeechDistance.

Network traffic forecasting plays a crucial role in intelligent network operations, but existing techniques often perform poorly when faced with limited data. Additionally, multi-task learning methods struggle with task imbalance and negative transfer, especially when modeling various service types. To overcome these challenges, we propose Sim-MSTNet, a multi-task spatiotemporal network traffic forecasting model based on the sim2real approach. Our method leverages a simulator to generate synthetic data, effectively addressing the issue of poor generalization caused by data scarcity. By employing a domain randomization technique, we reduce the distributional gap between synthetic and real data through bi-level optimization of both sample weighting and model training. Moreover, Sim-MSTNet incorporates attention-based mechanisms to selectively share knowledge between tasks and applies dynamic loss weighting to balance task objectives. Extensive experiments on two open-source datasets show that Sim-MSTNet consistently outperforms state-of-the-art baselines, achieving enhanced accuracy and generalization.

Multivariate time-series forecasting, as a typical problem in the field of time series prediction, has a wide range of applications in weather forecasting, traffic flow prediction, and other scenarios. However, existing works do not effectively consider the impact of extraneous variables on the prediction of the target variable. On the other hand, they fail to fully extract complex sequence information based on various time patterns of the sequences. To address these drawbacks, we propose a DA-SPS model, which adopts different modules for feature extraction based on the information characteristics of different variables. DA-SPS mainly consists of two stages: the target variable processing stage (TVPS) and the extraneous variables processing stage (EVPS). In TVPS, the model first uses Singular Spectrum Analysis (SSA) to process the target variable sequence and then uses Long Short-Term Memory (LSTM) and P-Conv-LSTM which deploys a patching strategy to extract features from trend and seasonality components, respectively. In EVPS, the model filters extraneous variables that have a strong correlation with the target variate by using Spearman correlation analysis and further analyses them using the L-Attention module which consists of LSTM and attention mechanism. Finally, the results obtained by TVPS and EVPS are combined through weighted summation and linear mapping to produce the final prediction. The results on four public datasets demonstrate that the DA-SPS model outperforms existing state-of-the-art methods. Additionally, its performance in real-world scenarios is further validated using a private dataset collected by ourselves, which contains the test items' information on laptop motherboards.

AI-assisted programming tools are widely adopted, yet their practical utility is often undermined by undesired suggestions that interrupt developer workflows and cause frustration. While existing research has explored developer-AI interactions when programming qualitatively, a significant gap remains in quantitative analysis of developers' acceptance of AI-generated code suggestions, partly because the necessary fine-grained interaction data is often proprietary. To bridge this gap, this paper conducts an empirical study using 66,329 industrial developer-AI interactions from a large technology company. We analyze features that are significantly different between accepted code suggestions and rejected ones. We find that accepted suggestions are characterized by significantly higher historical acceptance counts and ratios for both developers and projects, longer generation intervals, shorter preceding code context in the project, and older IDE versions. Based on these findings, we introduce CSAP (Code Suggestion Acceptance Prediction) to predict whether a developer will accept the code suggestion before it is displayed. Our evaluation of CSAP shows that it achieves the accuracy of 0.973 and 0.922 on imbalanced and balanced dataset respectively. Compared to a large language model baseline and an in-production industrial filter, CSAP relatively improves the accuracy by 12.6\% and 69.5\% on imbalanced dataset, and improves the accuracy by 87.0\% and 140.1\% on balanced dataset. Our results demonstrate that targeted personalization is a powerful approach for filtering out code suggestions with predicted rejection and reduce developer interruption. To the best of our knowledge, it is the first quantitative study of code suggestion acceptance on large-scale industrial data, and this work also sheds light on an important research direction of AI-assisted programming.

Large language models (LLMs) show promise as teaching assistants, yet their teaching capability remains insufficiently evaluated. Existing benchmarks mainly focus on problem-solving or problem-level guidance, leaving knowledge-centered teaching underexplored. We propose a syllabus-grounded evaluation framework that measures LLM teaching capability via student performance improvement after multi-turn instruction. By restricting teacher agents to structured knowledge points and example problems, the framework avoids information leakage and enables reuse of existing benchmarks. We instantiate the framework on Gaokao data across multiple subjects. Experiments reveal substantial variation in teaching effectiveness across models and domains: some models perform well in mathematics, while teaching remains challenging in physics and chemistry. We also find that incorporating example problems does not necessarily improve teaching, as models often shift toward example-specific error correction. Overall, our results highlight teaching ability as a distinct and measurable dimension of LLM behavior.

In this paper, we present NEMO, a system that translates Natural-language descriptions of decision problems into formal Executable Mathematical Optimization implementations, operating collaboratively with users or autonomously. Existing approaches typically rely on specialized large language models (LLMs) or bespoke, task-specific agents. Such methods are often brittle, complex and frequently generating syntactically invalid or non-executable code. NEMO instead centers on remote interaction with autonomous coding agents (ACAs), treated as a first-class abstraction analogous to API-based interaction with LLMs. This design enables the construction of higher-level systems around ACAs that structure, consolidate, and iteratively refine task specifications. Because ACAs execute within sandboxed environments, code produced by NEMO is executable by construction, allowing automated validation and repair. Building on this, we introduce novel coordination patterns with and across ACAs, including asymmetric validation loops between independently generated optimizer and simulator implementations (serving as a high-level validation mechanism), external memory for experience reuse, and robustness enhancements via minimum Bayes risk (MBR) decoding and self-consistency. We evaluate NEMO on nine established optimization benchmarks. As depicted in Figure 1, it achieves state-of-the-art performance on the majority of tasks, with substantial margins on several datasets, demonstrating the power of execution-aware agentic architectures for automated optimization modeling.

Recent studies of federated graph foundational models (FedGFMs) break the idealized and untenable assumption of having centralized data storage to train graph foundation models, and accommodate the reality of distributed, privacy-restricted data silos. Despite their simplicity and intuition, existing studies that project aligned generalizable knowledge onto a discrete token space via vector-quantized backbones suffer from irreversible knowledge loss during the quantization process. In this context, we argue that reconciling the semantic-structural orthogonality and integrity between pre-trained language models (PLMs) and graph neural networks (GNNs) is paramount for developing effective FedGFMs while simultaneously mitigating the severe data heterogeneity and communication constraints inherent in distributed, resource-limited environments. To address these issues, we propose FedGALA (Federated Graph And Language Alignment), a framework that resolves graph-based semantic-structural orthogonality and integrity in federated settings by employing unsupervised contrastive learning to align GNNs and frozen PLMs within a continuous embedding space, thereby capturing robust, transferable general knowledge. Subsequently, FedGALA leverages a communication-efficient prompt tuning mechanism to steer these pre-aligned encoders and frozen PLMs, facilitating effective adaptation to diverse downstream tasks while circumventing the prohibitive overhead of full-parameter fine-tuning. The comprehensive experiments validate that FedGALA outperforms all competitive baselines across multi-domain datasets on multiple tasks with up to 14.37% performance improvement.

Backpropagation algorithm has driven the remarkable success of deep neural networks, but its lack of biological plausibility and high computational costs have motivated the ongoing search for alternative training methods. Hebbian learning has attracted considerable interest as a biologically plausible alternative to backpropagation. Nevertheless, its exclusive reliance on local information, without consideration of global task objectives, fundamentally limits its scalability. Inspired by the biological synergy between neuromodulators and local plasticity, we introduce a novel model-agnostic Global-guided Hebbian Learning (GHL) framework, which seamlessly integrates local and global information to scale up across diverse networks and tasks. In specific, the local component employs Oja's rule with competitive learning to ensure stable and effective local updates. Meanwhile, the global component introduces a sign-based signal that guides the direction of local Hebbian plasticity updates. Extensive experiments demonstrate that our method consistently outperforms existing Hebbian approaches. Notably, on large-scale network and complex datasets like ImageNet, our framework achieves the competitive results and significantly narrows the gap with standard backpropagation.

The forward pass of a Transformer can be seen as an interacting particle system on the unit sphere: time plays the role of layers, particles that of token embeddings, and the unit sphere idealizes layer normalization. In some weight settings the system can even be seen as a gradient flow for an explicit energy, and one can make sense of the infinite context length (mean-field) limit thanks to Wasserstein gradient flows. In this paper we study the effect of the perceptron block in this setting, and show that critical points are generically atomic and localized on subsets of the sphere.

The rapid integration of Large Language Models (LLMs) into educational assessment rests on the unverified assumption that instruction following capability translates directly to objective adjudication. We demonstrate that this assumption is fundamentally flawed. Instead of evaluating code quality, models frequently decouple from the submission's logic to satisfy hidden directives, a systemic vulnerability we term the Compliance Paradox, where models fine-tuned for extreme helpfulness are vulnerable to adversarial manipulation. To expose this, we introduce the Semantic-Preserving Adversarial Code Injection (SPACI) Framework and the Abstract Syntax Tree-Aware Semantic Injection Protocol (AST-ASIP). These methods exploit the Syntax-Semantics Gap by embedding adversarial directives into syntactically inert regions (trivia nodes) of the Abstract Syntax Tree. Through a large-scale evaluation of 9 SOTA models across 25,000 submissions in Python, C, C++, and Java, we reveal catastrophic failure rates (>95%) in high-capacity open-weights models like DeepSeek-V3, which systematically prioritize hidden formatting constraints over code correctness. We quantify this failure using our novel tripartite framework measuring Decoupling Probability, Score Divergence, and Pedagogical Severity to demonstrate the widespread "False Certification" of functionally broken code. Our findings suggest that current alignment paradigms create a "Trojan" vulnerability in automated grading, necessitating a shift from standard RLHF toward domain-specific Adjudicative Robustness, where models are conditioned to prioritize evidence over instruction compliance. We release our complete dataset and injection framework to facilitate further research on the topic.

Failures in complex systems demand rapid Root Cause Analysis (RCA) to prevent cascading damage. Existing RCA methods that operate without dependency graph typically assume that the root cause having the highest anomaly score. This assumption fails when faults propagate, as a small delay at the root cause can accumulate into a much larger anomaly downstream. In this paper, we propose PRISM, a simple and efficient framework for RCA when the dependency graph is absent. We formulate a class of component-based systems under which PRISM performs RCA with theoretical guarantees. On 735 failures across 9 real-world datasets, PRISM achieves 68% Top-1 accuracy, a 258% improvement over the best baseline, while requiring only 8ms per diagnosis.

Chain-of-Thought (CoT) empowers Large Language Models (LLMs) to tackle complex problems, but remains constrained by the computational cost and reasoning path collapse when grounded in discrete token spaces. Recent latent reasoning approaches attempt to optimize efficiency by performing reasoning within continuous hidden states. However, these methods typically operate as opaque end-to-end mappings from explicit reasoning steps to latent states, and often require a pre-defined number of latent steps during inference. In this work, we introduce PLaT (Planning with Latent Thoughts), a framework that reformulates latent reasoning as planning by fundamentally decouple reasoning from verbalization. We model reasoning as a deterministic trajectory of latent planning states, while a separate Decoder grounds these thoughts into text when necessary. This decoupling allows the model to dynamically determine when to terminate reasoning rather than relying on fixed hyperparameters. Empirical results on mathematical benchmarks reveal a distinct trade-off: while PLaT achieves lower greedy accuracy than baselines, it demonstrates superior scalability in terms of reasoning diversity. This indicates that PLaT learns a robust, broader solution space, offering a transparent and scalable foundation for inference-time search.

Bayesian Optimization (BO) is a principled approach for optimizing expensive black-box functions, with Expected Improvement (EI) being one of the most widely used acquisition functions. Despite its empirical success, EI is known to be overly exploitative and can converge to suboptimal stationary points. We propose Expected Improvement via Gradient Norms (EI-GN), a novel acquisition function that applies the improvement principle to a gradient-aware auxiliary objective, thereby promoting sampling in regions that are both high-performing and approaching first-order stationarity. EI-GN relies on gradient observations used to learn gradient-enhanced surrogate models that enable principled gradient inference from function evaluations. We derive a tractable closed-form expression for EI-GN that allows efficient optimization and show that the proposed acquisition is consistent with the improvement-based acquisition framework. Empirical evaluations on standard BO benchmarks demonstrate that EI-GN yields consistent improvements against standard baselines. We further demonstrate applicability of EI-GN to control policy learning problems.

GUI agents are designed to automate repetitive tasks and enhance productivity. However, existing GUI agents struggle to recover once they follow an incorrect exploration path, often leading to task failure. In this work, we model GUI task execution as a DFS process and propose BEAP-Agent, a DFS-based framework that supports long-range, multi-level state backtracking with dynamic task tracking and updating. The framework consists of three collaborative components: Planner, Executor, and Tracker. Together, they enable effective task exploration and execution. BEAP-Agent fills the gap in systematic backtracking mechanisms for GUI agents, offering a systematic solution for long-horizon task exploration. We conducted a systematic evaluation on the OSWorld benchmark, where BEAP-Agent achieved an accuracy of 28.2%, validating the effectiveness of the proposed method.

Attention-FFN disaggregation (AFD) is an emerging architecture for LLM decoding that separates state-heavy, KV-cache-dominated Attention computation from stateless, compute-intensive FFN computation, connected by per-step communication. While AFD enables independent scaling of memory and compute resources, its performance is highly sensitive to the Attention/FFN provisioning ratio: mis-sizing induces step-level blocking and costly device idle time. We develop a tractable analytical framework for sizing AFD bundles in an $r$A-$1$F topology, where the key difficulty is that Attention-side work is nonstationary-token context grows and requests are continuously replenished with random lengths-while FFN work is stable given the aggregated batch. Using a probabilistic workload model, we derive closed-form rules for the optimal A/F ratio that maximize average throughput per instance across the system. A trace-calibrated AFD simulator validates the theory: across workloads, the theoretical optimal A/F ratio matches the simulation-optimal within 10%, and consistently reduces idle time.

A reliable reward model is essential for aligning large language models with human preferences through reinforcement learning from human feedback. However, standard reward models are susceptible to spurious features that are not causally related to human labels. This can lead to reward hacking, where high predicted reward does not translate into better behavior. In this work, we address this problem from a causal perspective by proposing a factored representation learning framework that decomposes the model's contextual embedding into (1) causal factors that are sufficient for reward prediction and (2) non-causal factors that capture reward-irrelevant attributes such as length or sycophantic bias. The reward head is then constrained to depend only on the causal component. In addition, we introduce an adversarial head trained to predict reward from the non-causal factors, while applying gradient reversal to discourage them from encoding reward-relevant information. Experiments on both mathematical and dialogue tasks demonstrate that our method learns more robust reward models and consistently improves downstream RLHF performance over state-of-the-art baselines. Analyses on length and sycophantic bias further validate the effectiveness of our method in mitigating reward hacking behaviors.

Mixture-of-Experts (MoE) models scale neural networks by conditionally activating a small subset of experts, where the router plays a central role in determining expert specialization and overall model performance. However, many modern MoE systems still adopt linear routers in raw high-dimensional representation spaces, where representation mismatch, angular concentration, and scale-sensitive scoring can jointly undermine routing discriminability and stable expert specialization. In this work, we propose Low-rank \& Lipschitz-controlled Routing (L2R), a unified routing framework that reshapes both the routing space and scoring geometry. L2R performs expert assignment in a shared low-rank latent routing space and introduces Saturated Inner-Product Scoring (SIPS) to explicitly control the Lipschitz behavior of routing functions, yielding smoother and more stable routing geometry. In addition, L2R incorporates a parameter-efficient multi-anchor routing mechanism to enhance expert expressiveness. Extensive experiments on a large-scale language MoE model and a vision MoE setting on ImageNet demonstrate that L2R consistently improves routing stability, expert specialization, and overall model performance.

Controlling memorization in diffusion models is critical for applications that require generated data to closely match the training distribution. Existing approaches mainly focus on data centric or model centric modifications, treating the diffusion model as an isolated predictor. In this paper, we study memorization in diffusion models from a denoising centric perspective. We show that uniform timestep sampling leads to unequal learning contributions across denoising steps due to differences in signal to noise ratio, which biases training toward memorization. To address this, we propose a timestep sampling strategy that explicitly controls where learning occurs along the denoising trajectory. By adjusting the width of the confidence interval, our method provides direct control over the memorization generalization trade off. Experiments on image and 1D signal generation tasks demonstrate that shifting learning emphasis toward later denoising steps consistently reduces memorization and improves distributional alignment with training data, validating the generality and effectiveness of our approach.

This paper presents a framework for integrating LLM into collaborative learning platforms to enhance student engagement, critical thinking, and inclusivity. The framework employs advanced LLMs as dynamic moderators to facilitate real-time discussions and adapt to learners' evolving needs, ensuring diverse and inclusive educational experiences. Key innovations include robust feedback mechanisms that refine AI moderation, promote reflective learning, and balance participation among users. The system's modular architecture featuring ReactJS for the frontend, Flask for backend operations, and efficient question retrieval supports personalized and engaging interactions through dynamic adjustments to prompts and discussion flows. Testing demonstrates that the framework significantly improves student collaboration, fosters deeper comprehension, and scales effectively across various subjects and user groups. By addressing limitations in static moderation and personalization in existing systems, this work establishes a strong foundation for next-generation AI-driven educational tools, advancing equitable and impactful learning outcomes.

Ensuring safety, factuality and overall quality in the generations of large language models is a critical challenge, especially as these models are increasingly deployed in real-world applications. The prevailing approach to addressing these issues involves collecting expensive, carefully curated datasets and applying multiple stages of fine-tuning and alignment. However, even this complex pipeline cannot guarantee the correction of patterns learned during pretraining. Therefore, addressing these issues during pretraining is crucial, as it shapes a model's core behaviors and prevents unsafe or hallucinated outputs from becoming deeply embedded. To tackle this issue, we introduce a new pretraining method that streams documents and uses reinforcement learning (RL) to improve the next K generated tokens at each step. A strong, post-trained model judges candidate generations -- including model rollouts, the original suffix, and a rewritten suffix -- for quality, safety, and factuality. Early in training, the process relies on the original and rewritten suffixes; as the model improves, RL rewards high-quality rollouts. This approach builds higher quality, safer, and more factual models from the ground up. In experiments, our method gives 36.2% and 18.5% relative improvements over standard pretraining in terms of factuality and safety, and up to 86.3% win rate improvements in overall generation quality.

Multimodal Large Language Models (MLLMs) have recently achieved substantial progress in general-purpose perception and reasoning. Nevertheless, their deployment in Food-Service and Retail Stores (FSRS) scenarios encounters two major obstacles: (i) real-world FSRS data, collected from heterogeneous acquisition devices, are highly noisy and lack auditable, closed-loop data curation, which impedes the construction of high-quality, controllable, and reproducible training corpora; and (ii) existing evaluation protocols do not offer a unified, fine-grained and standardized benchmark spanning single-image, multi-image, and video inputs, making it challenging to objectively gauge model robustness. To address these challenges, we first develop Ostrakon-VL, an FSRS-oriented MLLM based on Qwen3-VL-8B. Second, we introduce ShopBench, the first public benchmark for FSRS. Third, we propose QUAD (Quality-aware Unbiased Automated Data-curation), a multi-stage multimodal instruction data curation pipeline. Leveraging a multi-stage training strategy, Ostrakon-VL achieves an average score of 60.1 on ShopBench, establishing a new state of the art among open-source MLLMs with comparable parameter scales and diverse architectures. Notably, it surpasses the substantially larger Qwen3-VL-235B-A22B (59.4) by +0.7, and exceeds the same-scale Qwen3-VL-8B (55.3) by +4.8, demonstrating significantly improved parameter efficiency. These results indicate that Ostrakon-VL delivers more robust and reliable FSRS-centric perception and decision-making capabilities. To facilitate reproducible research, we will publicly release Ostrakon-VL and the ShopBench benchmark.

Electronic Health Records (EHRs) provide rich longitudinal clinical evidence that is central to medical decision-making, motivating the use of retrieval-augmented generation (RAG) to ground large language model (LLM) predictions. However, long-horizon EHRs often exceed LLM context limits, and existing approaches commonly rely on truncation or vanilla retrieval strategies that discard clinically relevant events and temporal dependencies. To address these challenges, we propose EHR-RAG, a retrieval-augmented framework designed for accurate interpretation of long-horizon structured EHR data. EHR-RAG introduces three components tailored to longitudinal clinical prediction tasks: Event- and Time-Aware Hybrid EHR Retrieval to preserve clinical structure and temporal dynamics, Adaptive Iterative Retrieval to progressively refine queries in order to expand broad evidence coverage, and Dual-Path Evidence Retrieval and Reasoning to jointly retrieves and reasons over both factual and counterfactual evidence. Experiments across four long-horizon EHR prediction tasks show that EHR-RAG consistently outperforms the strongest LLM-based baselines, achieving an average Macro-F1 improvement of 10.76%. Overall, our work highlights the potential of retrieval-augmented LLMs to advance clinical prediction on structured EHR data in practice.

How much of LLM output variance is explained by prompts versus model choice versus stochasticity through sampling? We answer this by evaluating 12 LLMs on 10 creativity prompts with 100 samples each (N = 12,000). For output quality (originality), prompts explain 36.43% of variance, comparable to model choice (40.94%). But for output quantity (fluency), model choice (51.25%) and within-LLM variance (33.70%) dominate, with prompts explaining only 4.22%. Prompts are powerful levers for steering output quality, but given the substantial within-LLM variance (10-34%), single-sample evaluations risk conflating sampling noise with genuine prompt or model effects.

In this report, we introduce Qwen3-ASR family, which includes two powerful all-in-one speech recognition models and a novel non-autoregressive speech forced alignment model. Qwen3-ASR-1.7B and Qwen3-ASR-0.6B are ASR models that support language identification and ASR for 52 languages and dialects. Both of them leverage large-scale speech training data and the strong audio understanding ability of their foundation model Qwen3-Omni. We conduct comprehensive internal evaluation besides the open-sourced benchmarks as ASR models might differ little on open-sourced benchmark scores but exhibit significant quality differences in real-world scenarios. The experiments reveal that the 1.7B version achieves SOTA performance among open-sourced ASR models and is competitive with the strongest proprietary APIs while the 0.6B version offers the best accuracy-efficiency trade-off. Qwen3-ASR-0.6B can achieve an average TTFT as low as 92ms and transcribe 2000 seconds speech in 1 second at a concurrency of 128. Qwen3-ForcedAligner-0.6B is an LLM based NAR timestamp predictor that is able to align text-speech pairs in 11 languages. Timestamp accuracy experiments show that the proposed model outperforms the three strongest force alignment models and takes more advantages in efficiency and versatility. To further accelerate the community research of ASR and audio understanding, we release these models under the Apache 2.0 license.

Existing multi-behavior recommendations tend to prioritize performance at the expense of explainability, while current explainable methods suffer from limited generalizability due to their reliance on external information. Neuro-Symbolic integration offers a promising avenue for explainability by combining neural networks with symbolic logic rule reasoning. Concurrently, we posit that user behavior chains inherently embody an endogenous logic suitable for explicit reasoning. However, these observational multiple behaviors are plagued by confounders, causing models to learn spurious correlations. By incorporating causal inference into this Neuro-Symbolic framework, we propose a novel Causal Neuro-Symbolic Reasoning model for Explainable Multi-Behavior Recommendation (CNRE). CNRE operationalizes the endogenous logic by simulating a human-like decision-making process. Specifically, CNRE first employs hierarchical preference propagation to capture heterogeneous cross-behavior dependencies. Subsequently, it models the endogenous logic rule implicit in the user's behavior chain based on preference strength, and adaptively dispatches to the corresponding neural-logic reasoning path (e.g., conjunction, disjunction). This process generates an explainable causal mediator that approximates an ideal state isolated from confounding effects. Extensive experiments on three large-scale datasets demonstrate CNRE's significant superiority over state-of-the-art baselines, offering multi-level explainability from model design and decision process to recommendation results.

We propose a new convex loss for SVMs, both for the binary classification and for the regression models. Therefore, we show the mathematical derivation of the dual problems and we experiment them with several small data-sets. The minimal dimension of those data-sets is due to the difficult scalability of the SVM method to bigger instances. This preliminary study should prove that using pattern correlations inside the loss function could enhance the generalisation performances. Coherently, results show that generalisation measures are never worse than the standard losses and several times they are better. In our opinion, it should be considered a careful study of this loss, coupled with shallow and deep neural networks. In fact, we present some novel results obtained with those architectures.

Distributionally robust optimisation (DRO) minimises the worst-case expected loss over an ambiguity set that can capture distributional shifts in out-of-sample environments. While Huber (linear-vacuous) contamination is a classical minimal-assumption model for an $\varepsilon$-fraction of arbitrary perturbations, including it in an ambiguity set can make the worst-case risk infinite and the DRO objective vacuous unless one imposes strong boundedness or support assumptions. We address these challenges by introducing bulk-calibrated credal ambiguity sets: we learn a high-mass bulk set from data while considering contamination inside the bulk and bounding the remaining tail contribution separately. This leads to a closed-form, finite $\mathrm{mean}+\sup$ robust objective and tractable linear or second-order cone programs for common losses and bulk geometries. Through this framework, we highlight and exploit the equivalence between the imprecise probability (IP) notion of upper expectation and the worst-case risk, demonstrating how IP credal sets translate into DRO objectives with interpretable tolerance levels. Experiments on heavy-tailed inventory control, geographically shifted house-price regression, and demographically shifted text classification show competitive robustness-accuracy trade-offs and efficient optimisation times, using Bayesian, frequentist, or empirical reference distributions.

Deep neural networks are vulnerable to adversarial examples--inputs with imperceptible perturbations causing misclassification. While adversarial transfer within neural networks is well-documented, whether classical ML pipelines using handcrafted features inherit this vulnerability when attacked via neural surrogates remains unexplored. Feature engineering creates information bottlenecks through gradient quantization and spatial binning, potentially filtering high-frequency adversarial signals. We evaluate this hypothesis through the first comprehensive study of adversarial transfer from DNNs to HOG-based classifiers. Using VGG16 as a surrogate, we generate FGSM and PGD adversarial examples and test transfer to four classical classifiers (KNN, Decision Tree, Linear SVM, Kernel SVM) and a shallow neural network across eight HOG configurations on CIFAR-10. Our results strongly refute the protective hypothesis: all classifiers suffer 16.6%-59.1% relative accuracy drops, comparable to neural-to-neural transfer. More surprisingly, we discover attack hierarchy reversal--contrary to patterns where iterative PGD dominates FGSM within neural networks, FGSM causes greater degradation than PGD in 100% of classical ML cases, suggesting iterative attacks overfit to surrogate-specific features that don't survive feature extraction. Block normalization provides partial but insufficient mitigation. These findings demonstrate that adversarial vulnerability is not an artifact of end-to-end differentiability but a fundamental property of image classification systems, with implications for security-critical deployments across computational paradigms.

This brief proposes \emph{White-Op}, an interpretable operational amplifier (op-amp) parameter design framework based on the human-mimicking reasoning of large-language-model agents. We formalize the implicit human reasoning mechanism into explicit steps of \emph{\textbf{introducing hypothetical constraints}}, and develop an iterative, human-like \emph{\textbf{hypothesis-verification-decision}} workflow. Specifically, the agent is guided to introduce hypothetical constraints to derive and properly regulate positions of symbolically tractable poles and zeros, thus formulating a closed-form mathematical optimization problem, which is then solved programmatically and verified via simulation. Theory-simulation result analysis guides the decision-making for refinement. Experiments on 9 op-amp topologies show that, unlike the uninterpretable black-box baseline which finally fails in 5 topologies, White-Op achieves reliable, interpretable behavioral-level designs with only 8.52\% theoretical prediction error and the design functionality retains after transistor-level mapping for all topologies. White-Op is open-sourced at \textcolor{blue}{https://github.com/zhchenfdu/whiteop}.

Deep neural networks (DNNs) often produce overconfident predictions on out-of-distribution (OOD) inputs, undermining their reliability in open-world environments. Singularities in semi-discrete optimal transport (OT) mark regions of semantic ambiguity, where classifiers are particularly prone to unwarranted high-confidence predictions. Motivated by this observation, we propose a principled framework to mitigate OOD overconfidence by leveraging the geometry of OT-induced singular boundaries. Specifically, we formulate an OT problem between a continuous base distribution and the latent embeddings of training data, and identify the resulting singular boundaries. By sampling near these boundaries, we construct a class of OOD inputs, termed optimal transport-induced OOD samples (OTIS), which are geometrically grounded and inherently semantically ambiguous. During training, a confidence suppression loss is applied to OTIS to guide the model toward more calibrated predictions in structurally uncertain regions. Extensive experiments show that our method significantly alleviates OOD overconfidence and outperforms state-of-the-art methods.

Urban Air Mobility (UAM) has emerged as a transformative solution to alleviate urban congestion by utilizing low-altitude airspace, thereby reducing pressure on ground transportation networks. To enable truly efficient and seamless door-to-door travel experiences, UAM requires close integration with existing ground transportation infrastructure. However, current research on optimal integrated routing strategies for passengers in air-ground mobility systems remains limited, with a lack of systematic exploration.To address this gap, we first propose a unified optimization model that integrates strategy selection for both air and ground transportation. This model captures the dynamic characteristics of multimodal transport networks and incorporates real-time traffic conditions alongside passenger decision-making behavior. Building on this model, we propose a Unified Air-Ground Mobility Coordination (UAGMC) framework, which leverages deep reinforcement learning (RL) and Vehicle-to-Everything (V2X) communication to optimize vertiport selection and dynamically plan air taxi routes. Experimental results demonstrate that UAGMC achieves a 34\% reduction in average travel time compared to conventional proportional allocation methods, enhancing overall travel efficiency and providing novel insights into the integration and optimization of multimodal transportation systems. This work lays a solid foundation for advancing intelligent urban mobility solutions through the coordination of air and ground transportation modes. The related code can be found at https://github.com/Traffic-Alpha/UAGMC.

Unsupervised domain adaptation (UDA) is a statistical learning problem when the distribution of training (source) data is different from that of test (target) data. In this setting, one has access to labeled data only from the source domain and unlabeled data from the target domain. The central objective is to leverage the source data and the unlabeled target data to build models that generalize to the target domain. Despite its potential, existing UDA approaches often struggle in practice, particularly in scenarios where the target domain offers only limited unlabeled data or spurious correlations dominate the source domain. To address these challenges, we propose a novel distributionally robust learning framework that models uncertainty in both the covariate distribution and the conditional label distribution. Our approach is motivated by the multi-source domain adaptation setting but is also directly applicable to the single-source scenario, making it versatile in practice. We develop an efficient learning algorithm that can be seamlessly integrated with existing UDA methods. Extensive experiments under various distribution shift scenarios show that our method consistently outperforms strong baselines, especially when target data are extremely scarce.

With the rapid expansion of electric vehicles (EVs) and charging infrastructure, the effective management of Autonomous Electric Taxi (AET) fleets faces a critical challenge in environments with dynamic and uncertain charging availability. While most existing research assumes a static charging network, this simplification creates a significant gap between theoretical models and real-world operations. To bridge this gap, we propose GAT-PEARL, a novel meta-reinforcement learning framework that learns an adaptive operational policy. Our approach integrates a graph attention network (GAT) to effectively extract robust spatial representations under infrastructure layouts and model the complex spatiotemporal relationships of the urban environment, and employs probabilistic embeddings for actor-critic reinforcement learning (PEARL) to enable rapid, inference-based adaptation to changes in charging network layouts without retraining. Through extensive simulations on real-world data in Chengdu, China, we demonstrate that GAT-PEARL significantly outperforms conventional reinforcement learning baselines, showing superior generalization to unseen infrastructure layouts and achieving higher overall operational efficiency in dynamic settings.

The increasing scale of graph datasets has significantly improved the performance of graph representation learning methods, but it has also introduced substantial training challenges. Graph dataset condensation techniques have emerged to compress large datasets into smaller yet information-rich datasets, while maintaining similar test performance. However, these methods strictly require downstream applications to match the original dataset and task, which often fails in cross-task and cross-domain scenarios. To address these challenges, we propose a novel causal-invariance-based and transferable graph dataset condensation method, named \textbf{TGCC}, providing effective and transferable condensed datasets. Specifically, to preserve domain-invariant knowledge, we first extract domain causal-invariant features from the spatial domain of the graph using causal interventions. Then, to fully capture the structural and feature information of the original graph, we perform enhanced condensation operations. Finally, through spectral-domain enhanced contrastive learning, we inject the causal-invariant features into the condensed graph, ensuring that the compressed graph retains the causal information of the original graph. Experimental results on five public datasets and our novel \textbf{FinReport} dataset demonstrate that TGCC achieves up to a 13.41\% improvement in cross-task and cross-domain complex scenarios compared to existing methods, and achieves state-of-the-art performance on 5 out of 6 datasets in the single dataset and task scenario.

This paper investigates search in model-based reinforcement learning (RL). Conventional wisdom holds that long-term predictions and compounding errors are the primary obstacles for model-based RL. We challenge this view, showing that search is not a plug-and-play replacement for a learned policy. Surprisingly, we find that search can harm performance even when the model is highly accurate. Instead, we show that mitigating distribution shift matters more than improving model or value function accuracy. Building on this insight, we identify key techniques for enabling effective search, achieving state-of-the-art performance across multiple popular benchmark domains.

The rapid advance of Generative AI into software development prompts this empirical investigation of perceptual effects on practice. We study the usage patterns of 147 professional developers, examining perceived correlates of AI tools use, the resulting productivity and quality outcomes, and developer readiness for emerging AI-enhanced development. We describe a virtuous adoption cycle where frequent and broad AI tools use are the strongest correlates of both Perceived Productivity (PP) and quality, with frequency strongest. The study finds no perceptual support for the Quality Paradox and shows that PP is positively correlated with Perceived Code Quality (PQ) improvement. Developers thus report both productivity and quality gains. High current usage, breadth of application, frequent use of AI tools for testing, and ease of use correlate strongly with future intended adoption, though security concerns remain a moderate and statistically significant barrier to adoption. Moreover, AI testing tools' adoption lags that of coding tools, opening a Testing Gap. We identify three developer archetypes (Enthusiasts, Pragmatists, Cautious) that align with an innovation diffusion process wherein the virtuous adoption cycle serves as the individual engine of progression. Our findings reveal that organizational adoption of AI tools follows such a process: Enthusiasts push ahead with tools, creating organizational success that converts Pragmatists. The Cautious are held in organizational stasis: without early adopter examples, they don't enter the virtuous adoption cycle, never accumulate the usage frequency that drives intent, and never attain high efficacy. Policy itself does not predict individuals' intent to increase usage but functions as a marker of maturity, formalizing the successful diffusion of adoption by Enthusiasts while acting as a gateway that the Cautious group has yet to reach.

We present the convergence rates of synchronous and asynchronous Q-learning for average-reward Markov decision processes, where the absence of contraction poses a fundamental challenge. Existing non-asymptotic results overcome this challenge by either imposing strong assumptions to enforce seminorm contraction or relying on discounted or episodic Markov decision processes as successive approximations, which either require unknown parameters or result in suboptimal sample complexity. In this work, under a reachability assumption, we establish optimal $\widetilde{O}(\varepsilon^{-2})$ sample complexity guarantees (up to logarithmic factors) for a simple variant of synchronous and asynchronous Q-learning that samples from the lazified dynamics, where the system remains in the current state with some fixed probability. At the core of our analysis is the construction of an instance-dependent seminorm and showing that, after a lazy transformation of the Markov decision process, the Bellman operator becomes one-step contractive under this seminorm.

Code commits in a version control system (e.g., Git) should be atomic, i.e., focused on a single goal, such as adding a feature or fixing a bug. In practice, however, developers often bundle multiple concerns into tangled commits, obscuring intent and complicating maintenance. Recent studies have used Conventional Commits Specification (CCS) and Language Models (LMs) to capture commit intent, demonstrating that Small Language Models (SLMs) can approach the performance of Large Language Models (LLMs) while maintaining efficiency and privacy. However, they do not address tangled commits involving multiple concerns, leaving the feasibility of using LMs for multi-concern detection unresolved. In this paper, we frame multi-concern detection in tangled commits as a multi-label classification problem and construct a controlled dataset of artificially tangled commits based on real-world data. We then present an empirical study using SLMs to detect multiple semantic concerns in tangled commits, examining the effects of fine-tuning, concern count, commit-message inclusion, and header-preserving truncation under practical token-budget limits. Our results show that a fine-tuned 14B-parameter SLM is competitive with a state-of-the-art LLM for single-concern commits and remains usable for up to three concerns. In particular, including commit messages improves detection accuracy by up to 44% (in terms of Hamming Loss) with negligible latency overhead, establishing them as important semantic cues.

The third and fourth density and acoustic virial coefficients of neon were determined at temperatures between 10 and 5000 K from first principles employing the path-integral Monte Carlo (PIMC) approach. For these calculations, we used the pair potential of Hellmann $\textit{et al.}$ [J. Chem. Phys. 154, 164304 (2021)], which is based on supermolecular $\textit{ab initio}$ calculations with basis sets of up to octuple-zeta quality and levels of theory up to coupled cluster with single, double, triple, quadruple, and perturbative pentuple excitations [CCSDTQ(P)]. The potential also accounts for relativistic, retardation, and post-Born$-$Oppenheimer effects and is provided with reliable uncertainty estimates. To incorporate nonadditive interactions, we developed a nonadditive three-body potential based on extensive supermolecular CCSD(T), CCSDT, and CCSDT(Q) calculations with basis sets of up to sextuple-zeta quality. This potential also accounts for relativistic effects. The very small nonadditive four-body contributions to the fourth virial coefficients were considered using a relatively simple nonadditive four-body potential based on supermolecular CCSD(T) calculations. We calculated the third and fourth density and third acoustic virial coefficients directly by PIMC and the fourth acoustic virial coefficient indirectly using thermodynamic relations between the density and acoustic virial coefficients. The uncertainties of the pair potential and those estimated for our nonadditive three-body potential were rigorously propagated in the PIMC calculations into uncertainties for the virial coefficients. These uncertainties are distinctly smaller than those of almost all of the corresponding experimental virial coefficient data.

Coupled excitable systems can generate a variety of patterns. In this work, we investigate coupled Chialvo maps in two dimensions under two types of nearest-neighbor couplings. One coupling produces ringlike patterns, while the other produces spirals. The rings expand with increasing coupling, whereas spirals evolve into turbulence and dissipate at stronger coupling. To quantify these patterns, we introduce an analogue of the discriminant of the velocity gradient tensor and examine the persistence of its sign. For ring-type patterns, the persistence decays more slowly than exponentially, often following a power law or stretched exponential. When spiral structures remain intact, persistence saturates asymptotically and can exhibit superposed periodic oscillations, suggesting complex exponents at early times. These behaviors highlight deep connections with the underlying dynamics.

A major issue in exploiting plasmonic cavities as key components in nanotechnology is the effect of radiative and absorption losses on their electrodynamic behavior. Treating them as open-systems, we derive a time-local, completely positive master equation that retains non-secular interference between decay pathways and reduces to the standard secular description when the environment resolves polariton splitting. When it does not, the theory predicts order-one deviations from secular leakage dynamics, including bath-induced coherences and stabilization of dark polaritons, and provides a simple design criterion based on the ratio of polariton splitting to reservoir linewidth. A time-resolved leakage measurement, such as transmission, reflectivity, or photoluminescence, can be used to observe these effects.

Fluid mixtures that exhibit an azeotrope cannot be purified by simple bulk distillation. Consequently, there is strong motivation to understand the behavior of azeotropic mixtures under confinement. We address this problem using a machine-learning-enhanced classical density functional theory applied to a binary Lennard-Jones mixture that exhibits azeotropic phase behavior. As proof-of-principle of a "train once, learn many" strategy, our approach combines a neural functional trained on a single-component repulsive reference system with a mean-field treatment of attractive interactions, derived within the framework of hyperdensity functional theory (hyper-DFT). The theory faithfully describes capillary condensation and results from grand canonical Monte Carlo simulations. Moreover, by taking advantage of a known accurate equation of state, the theory we present well-describes bulk thermodynamics by construction. Exploiting the computational efficiency of hyper-DFT, we systematically evaluate adsorption selectivity across a wide range of compositions, pressures, temperatures, and wall-fluid affinities. In cases where the wall-fluid interaction is the same for both species, we find that the pore becomes completely unselective at the bulk azeotropic composition. Strikingly, this unselective point persists far from liquid-vapor coexistence, including in the supercritical regime. Analysis of the bulk equation of state across a wide range of thermodynamic state points shows that the azeotropic composition coincides with equal partial molar volumes and an extremum in the isothermal compressibility. A complementary thermodynamic analysis demonstrates that unselective adsorption corresponds to an aneotrope (a point of zero relative adsorption) and an extremum in the interfacial free energy. We also find that the two interfaces of the slit pore behave independently down to remarkably small slits.

We investigate the microwave-power dependence of magnon-photon coupling in a yttrium iron garnet-sphere/split-ring-resonator hybrid system at room temperature and demonstrate that nonlinear spin-wave interactions suppress the coupling through power-induced dissipation of magnetostatic modes. At low microwave power, the modes exhibit pronounced level repulsion, evidencing strong coupling to the microwave field. As the power increases, however, magnon linewidth broadening progressively weakens the coupling and ultimately suppresses it entirely below a threshold external magnetic field. We show that this behavior originates from Suhl's first-order instability: magnetostatic modes, which couple to the resonator, parametrically excites two counter-propagating magnons at half its frequency, causing modes below the threshold external magnetic field to vanish. In contrast, magnon modes above the threshold field remain robust even at high power, as the instability criterion is not satisfied in that regime. These results reveal a well-defined nonlinear boundary for magnon-photon coupled systems and highlight a favorable regime for exploiting nonlinear magnonics for frequency conversion, switching, and other functional magnonic devices.

We study the thermodynamic performance of a periodic quantum Otto cycle operating on the single-impurity Anderson model. Using a decomposition of the time-evolution generator based on the principle of minimal dissipation, combined with the numerically exact hierarchical equations of motion (HEOM) method, we analyze the operating regimes of the quantum thermal machine and investigate effects of Coulomb interactions, strong system-reservoir coupling, and energy level alignments. Our results show that Coulomb interaction can change the operating regimes and may lead to an enhancement of the efficiency.

We investigate the $λ\ph^4+g\ph^6$ model using the renormalization group method and the $\ep$ expansion. This model is used in a situation where the coefficients $λ$, $g$ and the coefficient $τ$ of the term $τ\ph^2$ depend on two parameters $T$ and $P$, and there is a point ($T_c,P_c$) at which $τ$ and $λ$ are zero. This point is named the tricritical point. The description of a system depends on a trajectory that leads to the tricritical point on the plane ($T,P$). In the trajectories, when $λ$ goes to zero fast enough, the description is defined by the $\ph^6$ interaction and then the $\ph^4$ term can be considered as a composite operator. In this case, the logarithmic dimension is $d=3$, and the $\ep$ expansion is carried out in the dimension $d=3-2\ep$. The main exponents of the \textit{tricritical} model have been calculated in the third order of the $\ep$ expansion. Taking into account the $\ph^4$ interaction, we were able to calculate the value of the parameter that determines the required decrease rate in $λ$ to implement the tricritical behavior. The tricritical dimensions of the composite operators $\ph^k$ for $k=1, 2, 4, 6$ have been computed. The resulting values are compared to those known from a conformal field theory and non-perturbative renormalization group.

We analyze the finite-size corrections to the crosscap overlap in the two-dimensional classical Ising model along its self-dual critical line. Using a fermionic formulation, we express the lattice crosscap overlap in terms of Bogoliubov angles and develop a contour-integral approach by analytically continuing the lattice momentum to the complex plane. This leads to a remarkably simple expression for the crosscap overlap, which demonstrates that the finite-size corrections decay exponentially with system size. We further derive an exact analytical formula for the corresponding decay constant and show that it is determined by the complex singularity structure of the Bogoliubov angle.

Solution viscosities of a polycation and a polyanion in NaCl-water: HSAB-guided rheology The zero-shear viscosities of poly(3-acrylamido-propyl-trimethyl-ammonium-chloride) (PAPTMAC-Cl, M ~ 7.8 kDa) and poly(styrene-sulfonate sodium) (PSS-Na, M ~ 75.6 kDa) were measured in aqueous NaCl solutions at 25 degrees C over a wide range of salt concentrations. Extrapolation to zero polymer concentration yields an intrinsic viscosity of 4 460 mL g^-1 for the polycation, i.e. roughly three times larger than that of the polyanion, although the polycation's molar mass is only one-tenth of the polyanion's. At low salinities the shear-overlap parameter S as a function of polymer concentration c exhibits a pronounced maximum for PAPTMAC-Cl, whereas PSS-Na shows a clear inflection point. With increasing NaCl concentration both curves become linear, indicating that the system has entered a regime where the solute dominates the flow behavior. The crossover concentrations (S_crov) of the polycation are systematically larger than those of the polyanion. By applying Pearson's Hard-Soft Acid-Base (HSAB) concept we find that the observed differences are not a simple consequence of opposite polymer charges. Rather they arise from the specific ion-pairing: the soft NR4+ group of PAPTMAC pairs with the hard Cl-, whereas the hard RSO3- units of PSS-Na interact with the hard Na+. This insight suggests that the rheological response of polyelectrolyte solutions can be deliberately tuned by choosing counter-ions of appropriate hardness/softness. Keywords: intrinsic viscosity, shear-overlap, polyelectrolytes, sodium chloride, HSAB theory, rheology, soft-matter.

The ability to steer polariton flow on-demand holds significant promise towards nanophotonic applications and photonic circuitry. Polariton canalization, exhibiting intrinsic collimation and diffractionless transport, emerges as a promising solution without guiding structures. However, earlier demonstrations have been restricted to certain crystal surfaces with intrinsic hyperbolic responses and operated in the linear regime. Here, we experimentally demonstrate canalization of nonlinear exciton polariton condensates with optical reconfigurability in a birefringent CsPbBr3 perovskite crystal without intrinsic hyperbolic response. By embedding the birefringent perovskite crystal into a planar microcavity, the interplay between cavity transverse-electric-transverse-magnetic splitting and crystalline birefringence produces an anisotropic band geometry with a hyperbolic-flat-parabolic evolution of polaritonic isofrequency contours (IFCs). Nonresonant pumping drives exciton polariton condensation onto flat far-field contours with nonlinear emission amplification, leading to coherent canalized flows with over twentyfold collimation with respect to arc-shaped contours. Reconfiguring the optical pumping-spot size allows switching the nonlinear polariton condensates into hyperbolic and parabolic IFC regimes, leading to divergent propagation behaviour with collimating reconfiguration. Our study reveals a distinct canalization framework for shaping the nonlinear exciton-polariton condensate flows, opening opportunities for all-optical polaritonic logic circuits based on stabilized nonlinear quantum interconnects.

We consider a non-interacting electron gas confined to a two-dimensional crystal by the action of a perpendicular magnetic field; in the one-particle approximation, the dynamics of the system is modelled by a spectrally gapped Bloch-Landau Hamiltonian. No commensurability condition is assumed between the magnetic flux per unit cell and the quantum of magnetic flux. We construct a non-equilibrium almost-stationary state (NEASS) which "dresses" the equilibrium Fermi projection on states below the spectral gap, and models the state of the system after the addition of a weak external electric field of strength $\varepsilon \ll 1$. Having in mind applications to the integer quantum Hall effect, we probe the response of a current operator in the direction transverse to that of the applied electric field, and show that the resulting current density in the NEASS is linear in $\varepsilon$, with no power-law corrections. The linear response coefficient, namely the Hall conductivity, is computed in terms of the equilibrium Fermi projection via the double-commutator formula, in accordance with the prediction from Kubo's linear response theory. Our results generalize the methods and findings of [Lett. Math. Phys. 112 (2022), 91] to the setting of uniform magnetic fields with incommensurate magnetic flux per unit cell, and to lattice-periodic perturbation of such magnetic fields.

Rigidity percolation provides an important basis for understanding the onset of mechanical stability in disordered materials. While most studies on the triangular lattice have focused on static properties at fixed bond~(site) occupation probabilities, the dynamics of the rigidity transition remain less explored. In this work, we formulate a dynamic pebble game algorithm that monitors how rigid clusters emerge and evolve as bonds are added sequentially to an empty lattice, with computational efficiency comparable to the standard static pebble game. We uncover a previously overlooked temporal self-similarity exhibited in multiple quantities, including the cluster size changes and merged cluster sizes during bond addition, as well as the number of simultaneously merging clusters. We identify large-scale cascade events in which a single bond addition triggers the merger of an extensive number of clusters that scales with system size with inverse correlation-length exponent. Using an event-based ensemble approach, we obtain high-precision estimates of the critical point $p_c = 0.660\,277\,8(10)$, the inverse correlation-length exponent $1/ν= 0.850(3)$, and the fractal dimension $d_f = 1.850(2)$, representing substantial improvements over existing values.

Magnonic frequency combs, characterized by a series of discrete frequency lines, have emerged as a promising frontier in magnon spintronics, with potential applications in advanced information processing and sensing technologies. Although the three-magnon scattering process is widely recognized as a fundamental mechanism for generating these combs, its experimental realization has remained challenging due to the high threshold power and strict conservation of momentum and energy. In this work, we propose a novel mechanism for the stimulated generation of magnonic frequency combs that overcomes these limitations. Our approach offers precise and efficient control over key comb properties, including spacing between spectral lines and the number of lines, marking a significant advancement in the field. We substantiate this mechanism through a robust combination of theoretical modeling, micromagnetic simulations, and experimental validation. This study not only demonstrates the feasibility of our method but also opens new pathways for integrating magnonic frequency combs into practical spintronic devices.

The S=2 Heisenberg antiferromagnet on the orthogonal dimer lattice is studied. The edges of the exact dimer and Neel-ordered phases in the ground state of the system are examined by the numerical diagonalization method. Our present results are discussed by combining them with previously obtained estimates for smaller-S cases. We find that an intermediate region between the exact dimer and Neel-ordered phases gradually widens as spin S is increased up to S=2.

In this dissertation, resonator-based readout techniques were developed for floating electrons as qubits on cryogenic substrates, using two platforms: electrons on liquid helium and electrons on solid neon. In addition, a cryogenic microwave source was developed to enable low-noise measurement for qubit readout.

Emissive metal halide perovskites (MHPs) have emerged as excellent candidates for next-generation optoelectronics due to their sharp color purity, inexpensive processing, and bandgap tunability. However, the development of violet and ultraviolet light-emitting MHPs has lagged behind due to challenges related to material and device stability, charge carrier transport, tunability into the ultraviolet spectrum, toxicity, and scalability. Here, we review the progress of both violet and ultraviolet MHP nanomaterials and light-emitting diodes, including materials synthesis and device fabrication across various crystal structures and dimensions (e.g., bulk thin films, 2D thin films, nanoplatelets, colloidal nanocrystals, and more) as well as lead-free platforms (e.g., rare-earth metal halide perovskites). By highlighting several pathways to continue the development of violet and ultraviolet light-emitting MHPs while also proposing tactics to overcome their outstanding challenges, we demonstrate the potential of state-of-the-art violet and ultraviolet MHP materials and devices for important applications in public health, 3D printing, nanofabrication, and more.

One-dimensional (1D) moiré superlattices provide a new route to engineering reduced-dimensional electronic states in van der Waals materials, yet their electronic structure and microscopic origin remain largely unexplored. Here, we investigate the structural relaxation and electronic properties of a 1D moiré superlattice formed in twisted bilayer 1T$'$-WTe$_2$ using density functional theory calculations, complemented by high-angle annular dark-field scanning transmission electron microscopy. We show that lattice relaxation strongly reconstructs the moiré stripes, leading to stacking-dependent stripe widths that are in excellent agreement with experimental observations. The relaxed structure hosts quasi-one-dimensional electronic bands near the Fermi level, characterized by strong dispersion along the stripe direction and nearly flat dispersion in the perpendicular direction. By comparing the full bilayer with isolated relaxed layers, we establish that these 1D electronic states are governed predominantly by an intralayer moiré potential induced by in-plane lattice relaxation, rather than by interlayer hybridization. We extract this position-dependent moiré potential directly from DFT calculations and construct an effective tight-binding model that reproduces both the band dispersion and the real-space localization of the electronic wave functions. Our results identify lattice relaxation as the key mechanism underlying 1D electronic states in 1D moiré superlattices. %and establish twisted bilayer WTe$_2$ as a promising platform for exploring emergent one-dimensional moiré physics. The framework developed here provides a unified theoretical basis for realizing and exploring one-dimensional moiré physics in a broad class of anisotropic two-dimensional materials.

We show that introducing short-range attractive couplings between the weights of a neural network during training provides a novel avenue for model quantization. These couplings rapidly induce the discretization of a model's weight distribution, and they do so in a mixed-precision manner despite only relying on two additional hyperparameters. We demonstrate that, within an appropriate range of hyperparameters, our "soft quantization'' scheme outperforms histogram-equalized post-training quantization on ResNet-20/CIFAR-10. Soft quantization provides both a new pipeline for the flexible compression of machine learning models and a new tool for investigating the trade-off between compression and generalization in high-dimensional loss landscapes.

Turing patterns are stationary, wave-like structures that emerge from the nonequilibrium assembly of reactive and diffusive components. While they are foundational in biophysics, their classical formulation relies on a single characteristic length scale that balances reaction and diffusion, making them overly simplistic for describing biological patterns, which often exhibit multi-scale structures, grain-like textures, and inherent imperfections. Here, we integrate diffusiophoretically-assisted assembly of finite-sized cells, driven by a background chemical gradient in a Turing pattern, while also incorporating intercellular interactions. This framework introduces key control parameters, such as the Péclet number, cell size distribution, and intercellular interactions, enabling us to reproduce strikingly similar structural features observed in natural patterns. We report imperfections, including spatial variations in pattern thickness, packing limits, and pattern breakups. Our model not only deepens our understanding but also opens a new line of inquiry into imperfect Turing patterns that deviate from the classical formulation in significant ways.

The sp${}^3$ dangling bond on the diamond surface plays a critical role in the performance and fabrication of diamond quantum technologies. For the former, the magnetic and electric properties of this defect can impede the performance of quantum sensors and computers. For the latter, the chemical properties of the dangling bond are integral to proposed methods for bottom-up fabrication of scalable diamond quantum devices. In pursuit of high performance and scalable diamond quantum technology, tunneling probe-based techniques offers the ability to create and modify the sp${}^3$ dangling bond with atomic-scale precision. However, these capabilities cannot be realised either deterministically or at scale without a means for identifying the sp${}^3$ dangling bond amidst the myriad of other defects on the diamond surface. Consequently, in this work we provide a comprehensive experimental and theoretical framework for STS-based characterisation of the sp${}^3$ defect on the H-terminated (100) diamond surface. This capability provides the foundation for future tunneling probe studies in the modification of dangling bonds.

We develop an effective continuum description for information scrambling in a chain of randomly interacting Majorana fermions. The approach is based on the semiclassical treatment of the path integral for an effective spin chain that describes "two-replica" observables such as the entanglement purity and the OTOC. This formalism gives exact results for the entanglement membrane and for operator spreading in the limit of weak interactions. In this limit there is a large crossover lengthscale between free and interacting behavior, and this large lengthscale allows for a continuum limit and a controlled saddle-point calculation. The formalism is also somewhat different from that known from random unitary circuits. The entanglement membrane emerges as a kind of bound state of two travelling waves, and shows an interesting unbinding phenomenon as the velocity of the entanglement membrane approaches the butterfly velocity.

Simulating water from first principles remains a significant computational challenge due to the slow dynamics of the underlying system. Although machine-learned interatomic potentials (MLPs) can accelerate these simulations, they often fail to achieve the required level of accuracy for reliable uncertainty quantification. In this study, we use MACE - an equivariant graph neural network architecture that has been trained using an extensive RPBE-D3 database - to predict density isobars, diffusion constants, radial distribution functions, and melting points. Although equivariant MACE models are computationally more expensive than simpler architectures, such as kernel-based potentials (KbPs), their significantly lower total energy errors allow for reliable thermodynamic reweighting with minimal bias. Our results are consistent with those of previous studies using KbPs; however, equivariant models can be validated against the ground-truth density functional theory (DFT) ensemble, providing a critical advantage. These findings establish equivariant MLPs as robust and reliable tools for investigating the thermophysical properties of water with DFT-level accuracy.

Type-II semi-Dirac fermions in two dimensions have been proposed to describe topologically nontrivial low energy excitations in titanium/vanadium oxide heterostructures. These quasiparticles appear at the merger of three Dirac cones, resulting in a non-zero Berry phase. We find, by employing Hartree-Fock, renormalization group and RPA techniques, that the spectrum is very sensitive to long-range electron-electron interactions and can undergo a profound transformation. Specifically the quasiparticle spectrum evolves, driven by interactions, from anisotropic Dirac dispersion at the lowest energies, towards the characteristic type-II semi-Dirac boomerang shape as the energy increases. The corresponding density of states varies between linear and power one third ($ρ(\varepsilon) \sim |\varepsilon| \rightarrow |\varepsilon|^{1/3}$). The crossover scale is controlled by the interaction strength $α= e^2/(\hbar v)$, and the specifics of the effective interacting Hamiltonian. Our results imply that various physical characteristics exhibit critical behavior with continuously varying 'critical exponents'; for example Landau levels in a magnetic field vary with the energy scale: $|\varepsilon_n(B)|\sim (nB)^{1/2} \rightarrow (nB)^{3/4}, n=0,1,2,...$, and similarly for other observables.

The Suspension Balance Model (SBM) [J. Fluid Mech. \textbf{275}, 157 (1994)] for two-phase flows uses the momentum balance of the particle phase as a closure for the particle flux, showing that particle migration is driven by the divergence of the particle-phase stress. The underlying basis of this model was challenged by Nott~et~al.\ [Phys. Fluids \textbf{23}, 043304 (2011)] where the authors argued that the hydrodynamic contributions to the suspension stress should not appear in the particle-phase momentum balance, being replaced by a different particle-phase stress. The particle-phase stress proposed by Nott~et~al., while mathematically correct, involves the partitioning of the (non-pairwise-additive) hydrodynamic forces, and care is needed to understand how the force on a chosen particle is affected by a second particle. We show by a simple two-particle calculation what is the proper partitioning, and show that it is consistent thermodynamically and gives the correct equilibrium osmotic pressure of Brownian colloids. Using Stokesian Dyanmics suspension rheology, we quantitatively demonstrate that the hydrodynamic contribution to the suspension stress is virtually identical to particle-phase stress; the only difference is that the isolated single-particle hydrodynamic stress contribution -- the Einstein viscosity correction -- must be removed from the suspension stress when used to predict particle flux. Our results validate a key assumption of the SBM and therefore revive its physical foundation.

Domain wall motion underpins emerging spintronic technologies, such as high-speed racetrack devices and THz logic. Spatially non-uniform magnetic exchange and anisotropy in ferromagnets can pin or accelerate domain walls. In ferrimagnets, where Walker breakdown is suppressed, walls can approach the magnon speed. Here, we show that in non-uniform ferrimagnets such gradients not only exert a net force on the wall, but also modify its effective mass, enabling an entirely new acceleration mechanism. As a wall traverses regions of varying exchange or anisotropy, it can shed or gain mass leading to a "rocket effect" as in variable-mass systems. This phenomenon becomes increasingly pronounced as the wall approaches the magnon velocity, providing a natural route to ultrafast domain wall propulsion. The findings establish variable-mass domain walls as a new paradigm for efficient, high-velocity spintronics and THz-frequency magnetic technologies.

We propose a framework for preparing quantum states with a holographic entanglement structure, in the sense that the entanglement entropies are governed by minimal surfaces in a chosen bulk geometry. We refer to such entropies as holographic because they obey a relation between entropies and bulk minimal surfaces, known as the Ryu-Takayanagi formula, that is a key feature of holographic models of quantum gravity. Typically in such models, the bulk geometry is determined by solving Einstein's equations. Here, we simply choose a bulk geometry, then discretize the geometry into a coupling graph comprising bulk and boundary nodes. Evolving under this graph of interactions and measuring the bulk nodes leaves behind the desired pure state on the boundary. We numerically demonstrate that the resulting entanglement properties approximately reproduce the predictions of the Ryu-Takayanagi formula in the chosen bulk geometry. We consider graphs associated with hyperbolic disk and wormhole geometries, but the approach is general. The minimal ingredients in our proposal involve only Gaussian operations and measurements and are readily implementable in photonic and cold-atom platforms.

The dynamical behavior of field-induced ferroelectric phase transitions in compositions of PbMg_{1/3}Nb_{2/3}O3(1-x)-PbTiO3(x), called PMN-PT, near the Morphotropic Phase Boundary (MPB) was investigated through several different external electrical field application protocols. Our results indicate that the phase transitions in PMN-PT compositions near the MPB behave differently than in compositions far below the MPB. We show that the electrical-field history has a notable impact on the field-induced transition temperature T_c, ZFC delay time tau_{ZFC}, and induced polarization P_c, gained/lost during field-induced phase transition. Moreover, we demonstrate that under certain field-temperature conditions PMN-PT can retain its electrical field history and use it to kinetically accelerate its ferroelectric ordering. An explanation for the key difference between the phase transition dynamics in compositions near and far from the MPB is proposed and contextualized within prior publications.

We show that certain three-dimensional topological phases can act as acoustic diodes realizing nonlinear odd acoustoelastic effects. Beyond uncovering topologically-induced anomalous acoustic second-harmonic generation and rectification, we demonstrate how such nonlinear responses are uniquely captured by the momentum-space nonmetricity tensor in the quantum state Hilbert-space geometry. In addition to completing the classification of quantum geometric observables in the quadratic response regime, our findings reveal unexplored avenues for experimental realizations of acoustic diodes using effective $θ$ vacua of axion insulators adaptable for topological engineering applications.

We investigate spin angular momentum transfer in the Einstein-de Haas effect within prototypical magnetic crystals, focusing on its partition between phonons and rigid-body rotation. Using the Eckart frame to decouple local vibrations (phonons) from rigid-body rotation, we demonstrate that spin angular momentum is simultaneously transferred into both phonons and rigid-body rotation in an asymmetric way: rigid-body rotation acquires the dominant share of angular momentum, while phonons absorb most of the resulting kinetic energy. This divergent transfer of angular momentum and energy identifies phonons as direct and indispensable participants in the Einstein-de Haas dynamics. Furthermore, we find that pseudo-dipolar anisotropy and Dzyaloshinskii-Moriya interaction exert distinct control over the angular momentum transfer. Stronger pseudo-dipolar anisotropy increases the total amount of transferred angular momentum, whereas stronger Dzyaloshinskii-Moriya interaction accelerates the transfer rate and increases the proportion of phonon angular momentum. Our work clarifies the microscopic picture of the Einstein-de Haas effect and enables targeted angular-momentum control in magneto-mechanical devices.

The toric code, when deformed in a way that preserves the self-duality $\mathbb{Z}_2$ symmetry exchanging the electric and magnetic excitations, admits a transition to a topologically trivial state that spontaneously breaks the $\mathbb{Z}_2$ symmetry. Numerically, this transition was found to be continuous, which makes it particularly enigmatic given the longstanding absence of a continuum field-theoretic description. In this work we propose such a continuum field theory for the transition dubbed the $SO(4)_{2,-2}$ Chern-Simons-Higgs (CSH) theory. We show that our field theory provides a natural "mean-field" understanding of the phase diagram. Moreover, it can be generalized to an entire series of theories, namely the $SO(4)_{k,-k}$ CSH theories, labeled by an integer $k$. For each $k>2$, the theory describes an analogous transition involving different non-Abelian topological orders, such as the double Fibonacci order ($k=3$) and the $S_3$ quantum double ($k=4$). For $k=1$, we conjecture that the corresponding CSH transition is in fact infrared-dual to the $3d$ Ising transition, in close analogy with the particle-vortex duality of a complex scalar.

Anderson localization is a phase transition between a metallic phase, where wavefunctions are extended and delocalized in space, and an insulating phase, where wavefunctions are completely localized. These transitions are driven by uncorrelated disorder or quasiperiodic disorder, e.g., in the case of the Aubry-André model. Here, I consider a family of Hamiltonians that generalizes the Aubry-André model obtained when position and momentum operators are replaced by an arbitrary couple of canonically conjugate operators. In these models, a hidden localization transition occurs between metallic/insulating phases with wavefunctions delocalized/localized with respect to one of the two canonically conjugate operators. If the canonically conjugate operators coincide with a linear combination of position and momentum, the phase transition is signaled by a zero in the normalized participation ratio in the usual position space. Surprisingly, I found that at the phase transition, this model Hamiltonian coincides with the lattice Hamiltonian of a massless Dirac fermion in a curved spacetime background, indicating an unexpected relation between many-body localization and analog gravity.

We investigate the emergence of an amplitude (Higgs-like) mode in the gapless phase of the $(1+1)$D XXZ spin chain. Unlike conventional settings where amplitude modes arise from spontaneous symmetry breaking, here, we identify a symmetry-preserving underdamped excitation on top of a Luttinger-liquid ground state. Using nonequilibrium quench spectroscopy, we demonstrate that this mode manifests as oscillations of U(1)-symmetric observables following a sudden quench. By combining numerical simulations with Bethe-ansatz analyses, we trace its microscopic origin to specific families of string excitations. We further discuss experimental pathways to detect this mode in easy-plane quantum magnets and programmable quantum simulators. Our results showcase the utility of quantum quenches as a powerful tool to probe collective excitations, beyond the scope of linear response.

We investigate the classical limit of quantum master equations featuring double-bracket dissipators. Specifically, we consider dissipators defined by double commutators, which describe dephasing dynamics, as well as dissipators involving double anticommutators, associated with fluctuating anti-Hermitian Hamiltonians. The classical limit is obtained by formulating the open quantum dynamics in phase space using the Wigner function and Moyal products, followed by a systematic $\hbar$-expansion. We begin with the well-known model of energy dephasing, associated with energy diffusion. We then turn to master equations containing a double anticommutator with the system Hamiltonian, recently derived in the context of noisy non-Hermitian systems. For both classes of double-bracket equations, we provide a gradient-flow representation of the dynamics. We analyze the classical limit of the resulting evolutions for harmonic and driven anharmonic quantum oscillators, considering both classical and nonclassical initial states. The dynamics is characterized through the evolution of several observables as well as the Wigner logarithmic negativity. We conclude by extending our analysis to generalized master equations involving higher-order nested brackets, which provide a time-continuous description of spectral filtering techniques commonly used in the numerical analysis of quantum systems.

Quantifying how much a quantum state breaks a symmetry is essential for characterizing phases, nonequilibrium dynamics, and open-system behavior. Quantum resource theory provides a rigorous operational framework to define and characterize such quantifiers of symmetry-breaking. As a starter, we exemplify the usefulness of resource theory by noting that second-Rényi entanglement asymmetry can increase under symmetric operations, and hence is not a resource monotone, and should not solely be used to capture Quantum Mpemba effect. More importantly, motivated by mixed-state physics where weak and strong symmetries are inequivalent, we formulate a new resource theory tailored to strong symmetry, identifying free states and strong-covariant operations. This framework systematically identifies quantifiers of strong symmetry breaking for a broad class of symmetry groups, including a strong entanglement asymmetry. A particularly transparent structure emerges for U(1) symmetry, where the resource theory for the strong symmetry breaking has a completely parallel structure to the entanglement theory: the variance of the conserved quantity fully characterizes the asymptotic manipulation of strong symmetry breaking. By connecting this result to the knowledge of the geometry of quantum state space, we obtain a quantitative framework to track how weak symmetry breaking is irreversibly converted into strong symmetry breaking in open quantum systems. We further propose extensions to generalized symmetries and illustrate the qualitative impact of strong symmetry breaking in analytically tractable QFT examples and applications.

Tunneling Andreev reflection (TAR) spectroscopy offers a powerful new approach to fingerprint superconducting pairing symmetry at the atomic scale. By leveraging the exponential sensitivity of excess tunneling decay rate to Andreev reflection, TAR robustly distinguishes between s-wave, d-wave, and more complex order parameters, overcoming limitations of traditional conductance-based techniques. Here, using atomistic superconducting transport simulations, we show that the additivity of excess decay rate enables clear separation of Andreev and quasiparticle currents. In particular, we reveal how their competition as well as higher-order scattering processes shape both the decay rate spectra and their dependence on the coupling strength. We show that this phenomenology stems from the fact that Andreev reflection dominates mid-gap conductance for s-wave superconductors, it is suppressed for the d-wave, and it coexists with quasiparticle tunneling in sign-changing symmetries if the expectation value for the superconducting gap remains finite. These distinct spectral fingerprints pave the way for atomically resolved identification of unconventional superconducting states.

This work examines the snap-through behavior of clamped-clamped kirigami arches made from initially flat, thin, cut sheets under increasing vertical concentrated load acting at midspan. A two-parameter, symmetric pattern is introduced to conduct a numerical parameter analysis across three different support distances. When the support distance is one-quarter of the total length of the sheet, the structure loses stability at a symmetry point bifurcation over a wide range of parameters. Additionally, there exists a small range of parameters where limit point bifurcation occurs. In this case, the cuts can induce symmetry in the stability loss. For larger support distances (half or three-quarters of the total length), limit point bifurcation occurs only for small cuts, and there is a range of cut parameters that leads to monotonic monostability, indicating that no stability loss occurs. These findings are supported by experimental data. Overall, our research demonstrates that carefully designed cut patterns can either control the mode of stability loss in kirigami arches or suppress it entirely.

Data-driven equation discovery aims to reconstruct governing equations directly from empirical observations. A fundamental challenge in this domain is the ill-posed nature of the inverse problem, where multiple distinct mathematical models may yield similar errors, thus complicating model selection and failing to guarantee a unique representation of the underlying mechanisms. This issue can be addressed by incorporating inductive biases to constrain the search space and discard the undesirable models. The characteristic curves-based (CCs) framework offers a modular approach ideally suited to this aim. This approach is based on the specification of structural families that possess provable identifiability properties. Crucially, this framework enables practitioners to embed domain expertise directly into the learning process and facilitates the integration of diverse post-processing tools. In this work, we build upon the recent neural network implementation of this formalism (NN-CC), which benefits from the universal approximation capabilities of NNs. Specifically, we extend NN-CC by introducing two inductive biases: (i) symmetry constraints and (ii) post-processing with symbolic regression. Using a chaotic Duffing oscillator and a discontinuous stick-slip model under varying Gaussian noise levels, we show how these extensions systematically improve the discovery process. We also analyze the integration of sparse and symbolic regression (using SINDy and PySR) into the CC-based formalism. These extensions (SINDy-CC and SR-CC) consistently show improvements as prior information is incorporated. By enabling the integration of prior or hypothesized knowledge into the learning and post-processing stages, the CC-based formalism emerges as a promising candidate to address identifiability issues in purely data-driven methods, advancing the goal of interpretable and reliable system identification.

The study of chaos has long relied on computationally intensive methods to quantify unpredictability and design control strategies. Recent advances in machine learning, from convolutional neural networks to transformer architectures, provide new ways to analyze complex phase space structures and enable real time action in chaotic dynamics. In this perspective article, we highlight how data driven approaches can accelerate classical tasks such as estimating basin characterization metrics, or partial control of transient chaos, while opening new possibilities for scalable and robust interventions in chaotic systems. In recent studies, convolutional networks have reproduced classical basin metrics with negligible bias and low computational cost, while transformer based surrogates have computed accurate safety functions within seconds, bypassing the recursive procedures required by traditional methods. We discuss current opportunities, remaining challenges, and future directions at the intersection of nonlinear dynamics and artificial intelligence.

The brain combines short- and long-term memory to process, store, and recall multiple different pieces of information. Inspired by this and recent results on multifunctional and parameter-aware learning, we extend a new machine learning technique that combines short- and long-term memory units, specifically, a system consisting of a next-generation reservoir computer (NGRC) and extremely randomized trees (ERT), to process, store, and recall multiple different strange attractors. We train the combined NGRC+ERT system using a two-shot learning approach which significantly improves performance by filtering out unnecessary features, thereby avoiding extensive hyperparameter optimization. We first show that an NGRC+ERT system achieves highly accurate reconstruction of the short- and long-term dynamics of both the Lorenz and Halvorsen chaotic attractors when using an exponential filtering scheme. We validate these finding by training the NGRC+ERT system to reconstruct different pairs of attractors and also a greater number of attractors. We focus on the task of training a single NGRC+ERT system to reconstruct 16 different attractors and show that sufficient index-based separation in feature space suppresses unwanted switching dynamics, thus stabilizing long-term memory recall. Finally, we identify that defects in short-term memory processing can provoke failure modes in long-term memory recall resulting in confabulation.

Unequal-time correlation functions fundamentally characterize emergent statistical properties in complex systems, yet their direct measurement in experiments is challenging. We report the experimental observation of two-time, ballistic correlations in a photonic platform governed by the focusing nonlinear Schrödinger equation. Using a recirculating optical fiber loop with heterodyne field detection, we acquire the full space-time dynamics of partially coherent optical waves and extract the intensity correlator in stationary states of integrable turbulence. The correlators collapse under ballistic rescaling and quantitatively agree with predictions from Generalized Hydrodynamics evaluated using the density of states obtained via inverse scattering analysis of the recorded fields. Our results provide a direct, parameter-free test of GHD in an integrable waves system.

Human behavior plays a critical role in shaping epidemic trajectories. During health crises, people respond in diverse ways in terms of self-protection and adherence to recommended measures, largely reflecting differences in how individuals assess risk. This behavioral variability induces effective heterogeneity into key epidemic parameters, such as infectivity and susceptibility. We introduce a minimal extension of the susceptible-infected-removed~(SIR) model, denoted HeSIR, that captures these effects through a simple bimodal scheme, where individuals may have higher or lower transmission--related traits. We derive a closed-form expression for the epidemic threshold in terms of the model parameters, and the network's degree distribution and homophily, defined as the tendency of like--risk individuals to preferentially interact. We identify a resurgence regime just beyond the classical threshold, where the number of infected individuals may initially decline before surging into large-scale transmission. Through simulations on homogeneous and heterogeneous network topologies we corroborate the analytical results and highlight how variations in susceptibility and infectivity influence the epidemic dynamics. We further show that, under suitable assumptions, the HeSIR model maps onto a standard SIR process on an appropriately modified contact network, providing a unified interpretation in terms of structural connectivity. Our findings quantify the effect of heterogeneous behavioral responses, especially in the presence of homophily, and caution against underestimating epidemic potential in fragmented populations, which may undermine timely containment efforts. The results also extend to heterogeneity arising from biological or other non-behavioral sources.

We propose a hybrid semi-Lagrangian scheme for the Vlasov--Poisson equation that combines the Numerical Flow Iteration (NuFI) method with the Characteristic Mapping Method (CMM). Both approaches exploit the semi-group property of the underlying diffeomorphic flow, enabling the reconstruction of solutions through flow maps that trace characteristics back to their initial positions. NuFI builds this flow map iteratively, preserving symplectic structure and conserving invariants, but its computational cost scales quadratically with time. Its advantage lies in a compact, low-dimensional representation depending only on the electric field. In contrast, CMM achieves low computational costs when remapping by composing the global flow map from explicitly stored submaps. The proposed hybrid method merges these strengths: NuFi is employed for accurate and conservative local time stepping, while CMM efficiently propagates the solution through submap composition. This approach reduces storage requirements, maintains accuracy, and improves structural properties. Numerical experiments demonstrate the effectiveness of the scheme and highlight the trade-offs between memory usage and computational cost. We benchmark against a semi-Lagrangian predictor-corrector scheme used in modern gyrokinetic codes, evaluating accuracy and conservation properties.

Q-learning provides a standard reinforcement learning framework for studying cooperation by specifying how agents update action values from repeated local interactions outcomes. Although previous work has shown that reputation can promote cooperation in such systems, most models introduce reputation by modifying payoffs, encoding it directly in the state or changing partner selection, which makes it difficult to isolate the role of the learning signal itself. Here, we construct the reinforcement signal as a weighted combination of reputation and game payoffs, leaving the game and network structure unchanged. We find that increasing the weight on reputation generally promotes cooperation by consolidating clusters, but this effect is conditional on the learning dynamics. Specifically, this promoting effect vanishes in two regimes: when the learning rate is extremely small, which prevents effective information propagation and when the discount factor approaches one, as distant future expectations obscure the immediate reputational advantage. Outside these limiting cases, the efficacy of reputation in promoting cooperation is attenuated by higher learning rates but amplified by larger discount factors. These results advance the understanding of cooperative dynamics by demonstrating that cooperation can be stabilized through the reputational shaping of learning signals alone, providing critical insights into the interplay between social information and individual learning parameters.

Machine learning (ML) force fields have emerged as a powerful tool for computing materials properties at finite temperatures, particularly in regimes where traditional phonon-based perturbation theories fail or cannot be extended beyond the harmonic approximation. These approaches offer accuracy comparable to ab initio molecular dynamics (MD), but at a fraction of the computational cost. However, their reliability critically depends on the quality and representativeness of the training data. In particular, static training datasets often lead to failure when the force field encounters previously unseen atomic configurations during MD simulations. In this work, we present a framework for dynamically training ML force fields and demonstrate its effectiveness across materials with varying degrees of anharmonicity, including cubic boron arsenide (c-BAs), silicon (Si), and tin selenide (SnSe). Our method builds on the conventional lattice dynamics expansion of total energy and incorporates Bayesian error estimation to guide adaptive data acquisition during simulation. Specifically, we show that trajectory-averaged Bayesian errors enable efficient and targeted exploration of the configuration space, significantly enhancing the robustness and transferability of the resulting force fields. We further demonstrate how Bayesian error estimation can be applied to determine the convergence of the dynamic training without requiring additional ab initio data. This proposed framework offers a practical and easily implementable scheme to improve the training process, which is the most critical step in developing reliable ML force fields.

We demonstrate a model-free data analysis framework that leverages escort-weighted Shannon Entropy and several divergence matrices to detect phase transitions in scattering and imaging datasets. By establishing a connection between physical entropy and informational entropy, this approach provides a sensitive method for identifying phase transitions without an explicit physical model or order parameter. We further show that pairwise divergence matrices, including Kullback-Leibler divergence, Jeffrey Divergence, Jensen-Shannon Divergence and antisymmetric Kullback-Leibler divergence, provide more comprehensive measures of statistical changes than scalar entropy alone. Our approach successfully detects the onset of both long- and short-range order in neutron and X-ray scattering data, as well as a non-trivial phase transition in magnetic skyrmion lattices observed through Lorentz-transition electron microscopy. These results establish a framework for automated, model-free analysis of experimental data with broad applications in materials science and condensed matter physics.

The PI3K/AKT signaling pathway is triggered by recruitment of AKT to cellular membranes. Although AKT is a multidomain serine/threonine kinase composed of an N-terminal pleckstrin homology (PH) domain and a C-terminal kinase domain, how these domains cooperate to regulate AKT activation on membranes remains unclear at the molecular level. Here, using molecular dynamics simulations of full-length AKT on PIP$_3$-containing lipid bilayers, we identify four distinct membrane-binding modes that differ in the orientations and membrane contacts of the PH and kinase domains. In addition to PIP$_3$ binding to the PH domain, we observe specific PIP$_3$ interactions with basic residues in the kinase domain. In the most stable mode, PIP$_3$ interacts with both the canonical and a secondary binding site in the PH domain, while the kinase domain adopts an orientation in which the activation-loop phosphorylation site is exposed to the solvent. Interestingly, the populations of these binding modes depend on the PIP$_3$ concentration in the membrane, leading to changes in the preferred orientation of AKT. These findings shed light on how lipid recognition by the PH domain and the kinase domain of AKT cooperatively shape its membrane-bound conformations.

Understanding the three-dimensional motion of bubbles is essential for interpreting transport and mixing in multiphase flows, especially when bubbles deform under shear or move rapidly through the flow field. In many laboratory setups, only a single high-speed camera is available, which limits measurements to two dimensions. Traditional image-processing tools can identify bubbles only when they appear circular and isolated, but they struggle with irregularly shaped bubbles, shear-induced deformations, strong blurring, and partial overlaps. Multi-camera systems could overcome these issues, but require significant hardware additions and calibration effort. In this work, we introduce a new machine-learning framework that can detect bubbles and estimate their depth using only a single 20 kHz high-speed camera with 3 \textmu m resolution. The method first uses a large unlabeled dataset and clusters the bubbles with an unsupervised algorithm to reveal their underlying structure. These clusters provide pseudo labels, which are combined with a small set of true in-plane bubble labels to train a semi-supervised model that generalizes across different bubble appearances. These components produce a continuous depth-proxy score that indicates how close each bubble is to the imaging plane, even when bubbles are distorted or irregularly shaped. In parallel, we perform robust bubble identification using instance segmentation, which separates touching, overlapping, and elongated bubbles generated by high-velocity shear. Quantitatively, the in-plane segmentation baseline achieves strong held-out performance with Average Precision (AP) = 0.818, implying stable detection across thresholds, clutter, bubble detection Precision of 0.901, and a False-Positive Rate (FPR) near 6.1\%, hence low spurious bubbles and cleaner statistics under the tested acquisition conditions.

A semi-implicit Lax-Wendroff kinetic scheme is developed for hydrodynamic phonon transport in solid materials based on the Boltzmann transport equation under the double relaxation time approximation, in which both the normal and resistive scattering processes are accounted. The trapezoidal and midpoint rules are adopted for the temporal integration of the scattering and migration terms under the framework of finite volume method, respectively. Instead of direct numerical interpolation, the kinetic equation is solved again when reconstructing the interfacial flux, in order to realize the coupling of phonon migration and scattering within a numerical time step. Specifically, the finite difference scheme is introduced and the second-order upwind or central schemes are used for the reconstruction of the interfacial distribution function and its spatial gradient. Consequently, the cell size and time step of the present method could be larger than the phonon mean free path and relaxation time in the limit of small Knudsen numbers. Numerical tests demonstrate that the present method can accurately capture multi-scale thermal conduction phenomena within different normal or resistive scattering rates.

In a network, the vertices with similar characteristics construct communities. The vertices in a community are well-connected. Detecting the communities in a network is a challenging and important problem in the theory of complex networks. One approach to solve this problem uses the classical random walks on the graphs. In quantum computing, quantum walks are the quantum mechanical counterparts of classical random walks. In this article, we employ a variant of Szegedy's quantum walk to develop a procedure for discovering the communities in networks. The limiting probability distribution of quantum walks assists us in determining the inclusion of a vertex in a community. We apply our procedure of community detection on a number of graphs and social networks, such as the relaxed caveman graph, $l$-partition graph, Karate club graph, dolphin's social network, etc.

Electrides are exotic compounds in which excess electrons occupy interstitial regions of the crystal lattice and serve as anions, exhibiting exceptional properties such as low work function, high electron mobility, and strong catalytic activity. Although they show promise for diverse applications, identifying new electrides remains challenging due to the difficulty of achieving energetically favorable electron localization in crystal cavities. Here, we present an accelerated materials discovery framework that combines physical principles, diffusion-based materials generation with hierarchical thermodynamic and electronic structure screening. Using this workflow, we systematically explored 1,510 binary and 6,654 ternary chemical compositions containing excess valence electrons from electropositive alkaline, alkaline-earth, and early transition metals, and then filtered them with a high throughput validation on both thermodynamical stability and electronic structure analysis. As a result, we have identified 264 new electron rich compounds within 0.05 eV/atom above the convex hull at the density functional theory (DFT) level, including 13 thermodynamically stable electrides. Our approach demonstrates a generalizable strategy for targeted materials discovery in a vast chemical space.

A very common task in data visualization is to plot many data points with some measured y-value as a function of fixed x-values. Uncertainties on the y-values are typically presented as vertical error bars that represent either a Frequentist confidence interval or Bayesian credible interval for each data point. Most of the time, these error bars represent a 68\% confidence/credibility level, which leads to the intuition that a model fits the data reasonably well if its prediction lies within the error bars of roughly two thirds of the data points. Unfortunately, this and other intuitions no longer work when the uncertainties of the data points are correlated. If the error bars only show the square root of diagonal elements of some covariance matrix with non-negligible off-diagonal elements, we simply do not have enough information in the plot to judge whether a drawn model line agrees well with the data or not. In this paper we will demonstrate this problem and discuss ways to add more information to the plots to make it easier to judge the agreement between the data and some model prediction in the plot, as well as glean some insight where the model might be deficient. This is done by explicitly showing the contribution of the first principal component of the uncertainties, and by displaying the conditional uncertainties of all data points.

Metabolic diseases such as type 2 diabetes and obesity represent a rapidly escalating global health burden, yet current therapeutic strategies largely target isolated symptoms or single molecular pathways. To this end, we developed an integrated computational pipeline leveraging knowledge graph, pathway analysis and network pharmacology to elucidate the multi-target mechanisms of Ayurvedic Single Herbal Drugs (SHDs). SHDs associated with diabetes and obesity were curated from the Ayurvedic Pharmacopoeia of India, followed by phytochemical identification using IMPPAT database, yielding a shortlist of 11 SHDs and their 188 phytochemicals after drug-likeness and bioavailability filtering. Subsequently, molecular targets of the phytochemicals in SHDs, disease-associated genes and therapeutic targets of FDA-approved drugs, were curated via integration of data from several databases. Pathway enrichment analysis revealed significant functional overlap between SHD-associated and disease-associated pathways. All curated data were embedded into a Neo4j-based knowledge graph, enabling SHD-disease intersection analysis that prioritized key disease-relevant targets, including PTPN1, GLP1R, and DPP4. Also, the SHD-Target-FDA-approved drug profile elucidated the molecular and mechanistic aspects of the SHDs as a phytochemical cocktail, and is in alignment with the clinically studied synergistic FDA-approved drug combinations. Network pharmacology based protein-protein interaction analysis identified PPARG as another central regulator. Using a quantitative framework, we identified phytochemical pairs within SHDs, which were structurally dissimilar and target-wise distinct, yet acted on shared or different disease-associated pathways, indicating complementary and potentially synergistic interactions. Molecular docking analysis of two selected druggable targets identified putative lead phytochemicals.

Foundation models have shown remarkable success in fitting biological visual systems; however, their black-box nature inherently limits their utility for understanding brain function. Here, we peek inside a SOTA foundation model of neural activity (Wang et al., 2025) as a physiologist might, characterizing each 'neuron' based on its temporal response properties to parametric stimuli. We analyze how different stimuli are represented in neural activity space by building decoding manifolds, and we analyze how different neurons are represented in stimulus-response space by building neural encoding manifolds. We find that the different processing stages of the model (i.e., the feedforward encoder, recurrent, and readout modules) each exhibit qualitatively different representational structures in these manifolds. The recurrent module shows a jump in capabilities over the encoder module by 'pushing apart' the representations of different temporal stimulus patterns. Our 'tubularity' metric quantifies this stimulus-dependent development of neural activity as biologically plausible. The readout module achieves high fidelity by using numerous specialized feature maps rather than biologically plausible mechanisms. Overall, this study provides a window into the inner workings of a prominent neural foundation model, gaining insights into the biological relevance of its internals through the novel analysis of its neurons' joint temporal response patterns. Our findings suggest design changes that could bring neural foundation models into closer alignment with biological systems: introducing recurrence in early encoder stages, and constraining features in the readout module.

Promoters and enhancers are cis-regulatory elements (CREs), DNA sequences that bind transcription factor (TF) proteins to up- or down-regulate target genes. Decades-long efforts yielded TF-DNA interaction models that predict how strongly an individual TF binds arbitrary DNA sequences and how individual binding events on the CRE combine to affect gene expression. These insights can be synthesized into a global, biophysically-realistic, and quantitative genotype-phenotype (GP) map for gene regulation, a "holy grail" for the application of evolutionary theory. A global map provides a rare opportunity to simulate long-term evolution of regulatory sequences and pose several fundamental questions: How long does it take to evolve CREs de novo? How many non-trivial regulatory functions exist in sequence space? How connected are they? For which regulatory architecture is CRE evolution most rapid and evolvable? In this article, the second of a two-part series, we review the application of evolutionary concepts - epistasis, robustness, evolvability, tunability, plasticity, and bet-hedging - to the evolution of gene regulatory sequences. We then evaluate the potential for a unifying theory for the evolution of regulatory sequences, and identify key open challenges.

Pathophysiolpgical modelling of brain systems from microscale to macroscale remains difficult in group comparisons partly because of the infeasibility of modelling the interactions of thousands of neurons at the scales involved. Here, to address the challenge, we present a novel approach to construct differential causal networks directly from electroencephalogram (EEG) data. The proposed network is based on conditionally coupled neuronal circuits which describe the average behaviour of interacting neuron populations that contribute to observed EEG data. In the network, each node represents a parameterised local neural system while directed edges stand for node-wise connections with transmission parameters. The network is hierarchically structured in the sense that node and edge parameters are varying in subjects but follow a mixed-effects model. A novel evolutionary optimisation algorithm for parameter inference in the proposed method is developed using a loss function derived from Chen-Fliess expansions of stochastic differential equations. The method is demonstrated by application to the fitting of coupled Jansen-Rit local models. The performance of the proposed method is evaluated on both synthetic and real EEG data. In the real EEG data analysis, we track changes in the parameters that characterise dynamic causality within brains that demonstrate epileptic activity. We show evidence of network functional disruptions, due to imbalance of excitatory-inhibitory interneurons and altered epileptic brain connectivity, before and during seizure periods.

In this work we describe a new model for the evolution of a diploid structured population backwards in time that allows for large migrations and uneven offspring distributions. The model generalizes both the mean-field model of Birkner et al. [\textit{Electron. J. Probab.} 23: 1-44 (2018)] and the haploid structured model of Möhle [\textit{Theor. Popul. Biol.} 2024 Apr:156:103-116]. We show convergence, with mild conditions on the joint distribution of offspring frequencies and migrations, of gene genealogies conditional on the pedigree to a time-inhomogeneous coalescent process driven by a Poisson point process $Ψ$ that records the timing and scale of large migrations and uneven offspring distributions. This quenched scaling limit demonstrates a significant difference in the predictions of the classical annealed theory of structured coalescent processes. In particular, the annealed and quenched scaling limits coincide if and only if these large migrations and uneven offspring distributions are absent. The proof proceeds by the method of moments and utilizes coupling techniques from the theory of random walks in random environments. Several examples are given and their quenched scaling limits established.

Evolutionary prompt search is a practical black-box approach for red teaming large language models (LLMs), but existing methods often collapse onto a small family of high-performing prompts, limiting coverage of distinct failure modes. We present a speciated quality-diversity (QD) extension of ToxSearch that maintains multiple high-toxicity prompt niches in parallel rather than optimizing a single best prompt. ToxSearch-S introduces unsupervised prompt speciation via a search methodology that maintains capacity-limited species with exemplar leaders, a reserve pool for outliers and emerging niches, and species-aware parent selection that trades off within-niche exploitation and cross-niche exploration. ToxSearch-S is found to reach higher peak toxicity ($\approx 0.73$ vs.\ $\approx 0.47$) and a extreme heavier tail (top-10 median $0.66$ vs.\ $0.45$) than the baseline, while maintaining comparable performance on moderately toxic prompts. Speciation also yields broader semantic coverage under a topic-as-species analysis (higher effective topic diversity $N_1$ and larger unique topic coverage $K$). Finally, species formed are well-separated in embedding space (mean separation ratio $\approx 1.93$) and exhibit distinct toxicity distributions, indicating that speciation partitions the adversarial space into behaviorally differentiated niches rather than superficial lexical variants. This suggests our approach uncovers a wider range of attack strategies.

Identifying metabolic sites where cytochrome P450 enzymes metabolize small-molecule drugs is essential for drug discovery. Although existing computational approaches have been proposed for site-of-metabolism prediction, they typically ignore cytochrome P450 isoform identity or model isoforms independently, thereby failing to fully capture inherent cross-isoform metabolic patterns. In addition, prior evaluations often rely on top-k metrics, where false positive atoms may be included among the top predictions, underscoring the need for complementary metrics that more directly assess binary atom-level discrimination under severe class imbalance. We propose ATTNSOM, an atom-level site-of-metabolism prediction framework that integrates intrinsic molecular reactivity with cross-isoform relationships. The model combines a shared graph encoder, molecule-conditioned atom representations, and a cross-attention mechanism to capture correlated metabolic patterns across cytochrome P450 isoforms. The model is evaluated on two benchmark datasets annotated with site-of-metabolism labels at atom resolution. Across these benchmarks, the model achieves consistently strong top-k performance across multiple cytochrome P450 isoforms. Relative to ablated variants, the model yields higher Matthews correlation coefficient, indicating improved discrimination of true metabolic sites. These results support the importance of explicitly modeling cross-isoform relationships for site-of-metabolism prediction. The code and datasets are available at https://github.com/dmis-lab/ATTNSOM.

We construct a quantum Schwinger-Keldysh (SK) effective field theory for the diffusive hydrodynamics of a conserved scalar field. Quantum corrections within the SK framework are guided by fluctuation-dissipation relations, enforced via a dynamical Kubo-Martin-Schwinger (KMS) symmetry. We find that the KMS symmetry necessarily generates fluctuation contributions in the SK effective action at all orders in the noise field, thereby giving rise to intrinsically non-Gaussian noise. We use our results to compute one-loop quantum corrections to the two-point density-density retarded correlation function, leading to a quantum generalization of hydrodynamic long-time tails. Our results apply at arbitrarily high orders in $\hbar$. The one-loop results for retarded correlation functions have been expressed in terms of a family of polynomials. We also provide a closed-form expression for the one-loop results at leading order in the wavevector expansion.

As the number of detected gravitational wave (GW) events increases with the improved sensitivity of the observatories, detecting strongly lensed pairs of events is becoming a real possibility. Identifying such lensed pairs, however, remains challenging due to the computational cost and/or the reliance on prior knowledge of source parameters in existing methods. This study investigates a novel approach, Optimal Cross-Correlation Analysis for Multiplets (OCCAM), applied to strain data from one or more detectors for Compact Binary Coalescence (CBC) events identified by GW searches, using an optimal, mildly model-dependent, low computation cost approach to identify strongly lensed candidates. This technique efficiently narrows the search space, allowing for more sensitive, but (much) higher latency, algorithms to refine the results further. We demonstrate that our method performs significantly better than other computationally inexpensive methods. In particular, we achieve 97 percent (80 percent) lensed event detection at a pairwise false positive probability of approximately 13 percent (7 percent) for a single detector with LIGO design sensitivity, assuming an SNR greater than or equal to 10 astrophysically motivated lensed and unlensed populations. Thus, this method, using a network of detectors and in conjunction with sky-localisation information, can enormously reduce the false positive probability, making it highly viable to efficiently and quickly search for lensing pairs among thousands of events, including the sub-threshold candidates.

It has been shown that a very old black hole can tunnel into a white hole through the emission of a large baby universe. This process can be modeled by a genus-one geometry corresponding to a single baby universe emission, with a tunneling probability proportional to \( t^{2} e^{-2S(E)} \), where \( t \) denotes the black hole age and \( S(E) \) its entropy at energy \( E \). The growth of this probability at late times raises the question of its behavior near \( t \sim e^{S} \). A natural possibility is that the full genus expansion, together with its non-perturbative completion, leads to saturation of the tunneling probability. Motivated by this idea, the present analysis employs a non-perturbative bulk inner product in place of the perturbative one and shows that, at late times, the probabilities of realizing firewall geometries and smooth geometries approach constant values.

Corrections to vacuum black hole solutions of general relativity (GR) are considered in an effective field theory (EFT) framework, perturbatively in EFT coefficients, focusing on the Schwarzschild solution of GR. We find dominant corrections to the Schwarzschild metric in all orders in the derivative expansion far away from the horizon. These corrections can be summed up in a closed form through EFT coefficients up to all orders in derivatives and to the second order in curvature. It occurs that such a summation is convergent only for localizable theories, making a direct connection between the graviton scattering amplitudes properties and the applicability of a perturbative treatment of an EFT of gravity. We further apply our results to logarithmic form-factors which appear in the 1-loop effective action for GR in four dimensions. We find out that the corresponding corrections to the Schwarzschild metric are stronger than those from the tree-level EFT operators. The developed framework can be extended to account for the corrections to the other BH solutions in GR, such as the Kerr metric.

Interacting spin systems in solids underpin a wide range of quantum technologies, from quantum sensors and single-photon sources to spin-defect-based quantum registers and processors. We develop a quantum-computer-aided framework for simulating such devices using a general electron spin resonance Hamiltonian incorporating zero-field splitting, the Zeeman effect, hyperfine interactions, dipole-dipole spin-spin terms, and electron-phonon decoherence. Within this model, we combine Gray-encoded qudit-to-qubit mappings, qubit-wise commuting aggregation, and a multi-reference selected quantum Krylov fast-forwarding (sQKFF) hybrid algorithm to access long-time dynamics while remaining compatible with NISQ and early fault-tolerant hardware constraints. Numerical simulations demonstrate the computation of autocorrelation functions up to $\sim100$ ns, together with microwave absorption spectra and the $\ell_1$-norm of coherence, achieving 18-30$\%$ reductions in gate counts and circuit depth for Trotterized time-evolution circuits compared to unoptimized implementations. Using the nitrogen vacancy center in diamond as a testbed, we benchmark the framework against classical simulations and identify the reference-state selection in sQKFF as the primary factor governing accuracy at fixed hardware cost. This methodology provides a flexible blueprint for using quantum computers to design, compare, and optimize solid-state spin-qubit technologies under experimentally realistic conditions.

The sound speed of dark energy encodes fundamental information about the microphysics underlying cosmic acceleration, yet remains essentially unconstrained by existing observations. We demonstrate that a lunar-based laser interferometer, such as the proposed Laser Interferometer Lunar Antenna (LILA), can directly probe the sound speed of dark energy by measuring the real-time evolution of horizon-scale gravitational potentials. Operating in the ultra-low-frequency gravitational band inaccessible from Earth, LILA is sensitive to scalar metric perturbations sourced by dark energy dynamics. Using both fluid and effective field theory descriptions, we develop a complete framework linking dark energy sound speed to observable strain signatures. We construct a likelihood pipeline and Fisher forecasts, showing that LILA can either detect clustering dark energy or exclude broad classes of models with unprecedented sensitivity. This establishes lunar interferometry as a novel and powerful probe of the physics driving cosmic acceleration.

Open quantum systems interact with their environment, leading to nonunitary dynamics. We investigate the thermodynamics of linear Open Quantum Walks (OQWs), a class of quantum walks whose dynamics is entirely driven by the environment. We define an equilibrium temperature, identify a population inversion near a finite critical value of a control parameter, analyze the thermalization process, and develop the statistical mechanics needed to describe the thermodynamical properties of linear OQWs. We also study nonequilibrium thermodynamics by analyzing the time evolution of entropy, energy, and temperature, while providing analytical tools to understand the system's evolution as it converges to the thermalized state. We examine the validity of the second and third laws of thermodynamics in this setting. Finally, we employ these developments to shed light on dissipative quantum computation within the OQW framework.

The origin of the Hubble tension remains one of the central open problems in modern cosmology, with competing explanations invoking either early-Universe physics, late-time modifications of cosmic expansion, or unresolved observational systematics. In this Letter we propose a new, purely geometric null test of the late-time expansion history that is exactly independent of the Hubble constant. By combining strong-lensing time-delay distances with gravitational-wave standard-siren luminosity distances, we construct a dimensionless ratio that depends only on the redshift dependence of the expansion rate and can be both predicted from early-Universe data and measured directly at late times, without relying on the cosmic distance ladder or the sound horizon. We show that the comparison between the early- and late-time determinations of this ratio provides a transparent consistency test of the standard cosmological expansion. When combined with an independent standard-siren measurement of $H_{0}$, this framework allows one to unambiguously distinguish between early- and late-time origins of the Hubble tension. With the forthcoming detection of lensed gravitational-wave standard sirens, the proposed test provides a timely and robust framework for probing this long-standing cosmological puzzle.

We present a framework for generative machine learning that leverages the holographic principle of quantum gravity, or to be more precise its manifestation as the anti-de Sitter/conformal field theory (AdS/CFT) correspondence, with techniques for deep learning and transport theory. Our proposal is to represent the flow of data from a base distribution to some learned distribution using the bulk-to-boundary mapping of scalar fields in AdS. In the language of machine learning, we are representing and augmenting the flow-matching algorithm with AdS physics. Using a checkerboard toy dataset and MNIST, we find that our model achieves faster and higher quality convergence than comparable physics-free flow-matching models. Our method provides a physically interpretable version of flow matching. More broadly, it establishes the utility of AdS physics and geometry in the development of novel paradigms in generative modeling.

In this work, we present a full description of the spherically decomposed Green's function of a non-rotating black hole, which naturally splits into three components: the direct part, the quasinormal modes, and the tail. Both the direct part and the tail are contributed by branch cut integrals on the complex-frequency domain, and the quasinormal modes correspond to poles of the Green's function. We show that these different components match the Green's function numerically obtained by solving a time-domain Regge-Wheeler code. In addition, the components of the Green's function also agree with earlier studies in Schwarzschild spacetime with small cosmological constant. The identification of all the various parts of the Schwarzschild Green's function represents an important step towards analyzing direct waves and quasinormal modes in the ringdown stage of binary black hole coalescence, as well as their nonlinear interaction near the merger.

A growing variety of optically accessible spin qubits have emerged in recent years as key components for quantum sensors, qubits, and quantum memories. However, the scalability of conventional spin-based quantum architectures remains limited by direct microwave delivery, which introduces thermal noise, electromagnetic cross-talk, and design constraints for cryogenic, high-field, and distributed systems. In this work, we present a unified framework for RF-over-fiber (RFoF) control of optically accessible spins through RFoF optically detected magnetic resonance (ODMR) spectroscopy of nitrogen-vacancy (NV) centers in diamond. The RFoF platform relies on an electro-optically modulated telecom-band laser that transmits microwave signals over fiber and a high-speed photodiode that recovers the RF waveform to drive NV center spin transitions. We obtain an RFoF efficiency of 1.81\% at 2.90~GHz, corresponding to $P_{\mathrm{RF,out}}=-0.7$~dBm. The RFoF architecture provides a path toward low-noise, thermally isolated, and cryo-compatible ODMR systems bridging conventional spin-based quantum sensing protocols with emerging distributed quantum technologies.

Spontaneous Lorentz violation models of antisymmetric tensor field are known to possess singular Hamiltonian on the vacuum manifold, leading to unresolvable pathologies that render such theories unfit for cosmological studies. In this work, we show that by introducing an auxiliary vector field inspired by the Stückelberg mechanism to restore the gauge symmetry of the Lagrangian, it is possible to resolve such pathologies on vacuum manifold. The constraint analysis using Dirac-Bergmann method leads to a constraint matrix that acquires dependence on gradients and conjugate momentum of the Stückelberg field and therefore remains non-singular on the vacuum manifold.

Quantum machine learning (QML) is often listed as a promising candidate for useful applications of quantum computers, in part due to numerous proofs of possible quantum advantages. A central question is how small a role quantum computers can play while still enabling provable learning advantages over classical methods. We study an especially restricted setting in which a quantum computer is used only as a feature extractor: it acts independently on individual data points, without access to labels or global dataset information, is available only to augment the training set, and is not available at deployment. Training and deployment are therefore carried out by fully classical learners on a dataset augmented with quantum-generated features. We formalize this model by adapting the classical framework of Learning Under Privileged Information (LUPI) to the quantum case, which we call Learning Under Quantum Privileged Information (LUQPI). Within this framework, we show that even such minimally involved quantum feature extraction, available only during training, can yield exponential quantum-classical separations for suitable concept classes and data distributions under reasonable computational assumptions. We further situate LUQPI within a taxonomy of related quantum and classical learning settings and show how standard classical machinery, most notably the SVM+ algorithm, can exploit quantum-augmented data. Finally, we present numerical experiments in a physically motivated many-body setting, where privileged quantum features are expectation values of observables on ground states, and observe consistent performance gains for LUQPI-style models over strong classical baselines.

Distance metrics are central to machine learning, yet distances between ensembles of quantum states remain poorly understood due to fundamental quantum measurement constraints. We introduce a hierarchy of integral probability metrics, termed MMD-$k$, which generalizes the maximum mean discrepancy to quantum ensembles and exhibit a strict trade-off between discriminative power and statistical efficiency as the moment order $k$ increases. For pure-state ensembles of size $N$, estimating MMD-$k$ using experimentally feasible SWAP-test-based estimators requires $Θ(N^{2-2/k})$ samples for constant $k$, and $Θ(N^3)$ samples to achieve full discriminative power at $k = N$. In contrast, the quantum Wasserstein distance attains full discriminative power with $Θ(N^2 \log N)$ samples. These results provide principled guidance for the design of loss functions in quantum machine learning, which we illustrate in the training quantum denoising diffusion probabilistic models.

Superconducting resonators and qubits are limited by dielectric losses from surface oxides. Surface oxides are mitigated through various strategies such as the addition of a metal capping layer, surface passivation, and acid processing. In this study, we demonstrate the use of X-ray photoelectron spectroscopy (XPS) as a rapid characterization tool to study the effectiveness cap layers for niobium for further device fabrication. We non-destructively evaluate 17 capping layers to characterize their ability to prevent oxygen diffusion, and the effects of standard fabrication processes -- annealing, resist stripping, and acid cleaning. We downselect for resilient capping layers and test their microwave resonator performance.

We compare several definitions of entropy production rate introduced in the literature from a large variety of situations and motivations, and then analyze their relations with memory effects. Considering a relevant experimental example of a qubit interacting with a single bosonic mode playing the role of a finite bath, we show that all definitions of entropy production coincide at weak coupling. In the strong coupling regime, significant discrepancies emerge between the different entropy production rates, although some similarities in the overall behaviour remain. However, surprisingly, two of these definitions -- one based on local quantities of the system and the other on non-local quantities -- coincide exactly, even in the case of strong coupling. Finally, a high degree of correspondence is observed when memory effects characterized by P-divisibility are compared with the sign of all entropy production rates in the case of weak coupling. Such correspondence degrades at strong coupling, leading us to extend the concept of entropy production to the dynamical map. We show a perfect equivalence between the sign of this enlarged concept of entropy production and P-divisibility, both numerically and analytically, in the case of phase-covariant master equations.

We propose a method to suppress collisional loss in strongly dipolar, rotationally excited ultracold molecules using a combination of static (dc) and microwave (ac) electric fields. By tuning two excited pair molecular rotational states into a Förster resonance with a dc field, simultaneously driving excited rotational transitions with an ac field removes all long-range bound states, allowing near complete suppression of all two- and three-body collisional loss channels. While permitting tunable dipolar and anti-dipolar interactions, this bound-state-free ac/dc scheme is not subject to photon-changing collisions that are the primary source of two-body loss in shielding with two microwave fields, used to achieve the first molecular Bose-Einstein condensate [Bigagli et al., Nature 631, 289 (2024)]. Using NaCs as a representative example for strongly dipolar molecules, close-coupling calculations are performed to show that bound-state-free shielding can achieve ratios of elastic-to-loss rates $\gtrsim 10^{6}$ at 100 nK, with currently accessible ac and dc field generation technologies. This work opens new opportunities for realizing large, long-lived samples of strongly interacting degenerate molecular gases with tunable long-range interactions.

We investigate a hybrid quantum-classical algorithm for solving the Maximum Independent Set (MIS) problem on regular graphs, combining the Quantum Approximate Optimization Algorithm (QAOA) with a minimal degree classical greedy algorithm. The method leverages pre-computed QAOA angles, derived from depth-$p$ QAOA circuits on regular trees, to compute local expectation values and inform sequential greedy decisions that progressively build an independent set. This hybrid approach maintains shallow quantum circuit and avoids instance-specific parameter training, making it well-suited for implementation on current quantum hardware: we have implemented the algorithm on a 20 qubit IQM superconducting device to find independent sets in graphs with thousands of nodes. We perform tensor network simulations to evaluate the performance of the algorithm beyond the reach of current quantum hardware and compare to established classical heuristics. Our results show that even at low depth ($p=4$), the quantum-enhanced greedy method significantly outperforms purely classical greedy baselines as well as more sophisticated approximation algorithms. The modular structure of the algorithm and relatively low quantum resource requirements make it a compelling candidate for scalable, hybrid optimization in the NISQ era and beyond.

We apply analytical models to study the property of neutron stars and dark stars. With the aim of exploring the global observable properties of those compact stars, we investigate the total masses and radii, the tidal deformabilities and especially the fundamental (f -) mode oscillations. While we choose two typical models in this work, this method applies to any analytical equations of state. By comparing with the multi-messenger observations, one can constrain the corresponding parameters in those models.

Quantum machine learning (QML) models often require deep, parameterized circuits to capture complex frequency components, limiting their scalability and near-term implementation. We introduce \textit{Quantum Random Features} (QRF) and \textit{Quantum Dynamical Random Features} (QDRF), lightweight quantum reservoir models inspired by classical random Fourier features (RFF) that generate high-dimensional spectral representations without variational optimization. Using $Z$-rotation encoding combined with random permutations or Hamiltonian dynamics, these models achieve $N_f$-dimensional feature maps at preprocessing cost $O(\log(N_f))$. Spectral analysis shows that QRF and QDRF reproduce the behavior of RFF, while simulations on Fashion-MNIST reach up to 89.3\% accuracy-matching or surpassing classical baselines with scalable qubit requirements. By linking spectral theory with experimentally feasible quantum dynamics, this work provides a compact and hardware-compatible route to scalable quantum learning.

An increasing number of surveys has been reporting large-scale peculiar motions with sizes and speeds in excess of those allowed by the concordance cosmological model. These are the so called bulk flows, the presence of which has come to be treated as a problem for the $Λ$CDM paradigm. However, the limits of the $Λ$CDM model are based on Newtonian studies, which predict the mediocre $v\propto t^{1/3}$ growth-rate for the peculiar-velocity field ($v$). Recently, a few fully relativistic treatments have appeared in the literature, arguing for a much stronger velocity growth that could explain the reported fast and deep bulk flows. What separates the Newtonian from the relativistic studies is the gravitational input of the peculiar flux, namely of the kinetic energy triggered by the moving matter. The latter has no direct gravitational contribution in Newtonian theory, but it does so in general relativity. This drastically changes the driving agent of the peculiar-velocity field and boosts its linear growth. The aim of this work is to directly compare the two treatments, as well as identify and discuss the reasons for their different results. In the process, we also demonstrate how one could recover the relativistic growth-rate from a Newtonian setup by selectively including certain (typically ignored) source-free terms into the Poisson equation. This way, we provide a unified covariant comparison of the Newtonian, the quasi-Newtonian and the fully relativistic studies.

We set out to accurately measure gravity in the low-acceleration range $(10^{-11},10^{-9})$ m\,s$^{-2}$ from 3D motions of isolated wide binary stars. Gaia DR3 provides precise measurements of the four sky-plane components of the 3D relative displacement and velocity ($\mathbf{r}, \mathbf{v}$) for a wide binary, but not comparably precise line-of-sight (radial) separation and relative velocity $v_{r}$. Based on our new observations and the public databases/publications, we assemble a sample of 36 nearby (distance $<150$pc) wide binaries in the low-acceleration regime with accurate values of $v_{r}$ (uncertainty $< 100$ m\,s$^{-1}$). Kinematic contaminants such as undetected stellar companions are well under control using various observational diagnostics such as Gaia's ruwe parameter, the color-magnitude diagram, multi-epoch observations of radial velocities, Speckle interferometric follow-up observations, and requiring Hipparcos-Gaia proper motion consistency. For the parameter $Γ\equiv \log_{10}\sqrtγ$ with $γ\equiv G/G_{\rm N}$ (where $G$ is a parameter generalizing Newton's constant $G_{\rm N}$ in elliptical orbits), we find $Γ=0.102_{-0.021}^{+0.023}$, inconsistent with standard gravity at $4.9σ$, giving a gravity boost factor of $γ=1.600_{-0.141}^{+0.171}$. Four wide binaries have 3D relative velocities exceeding their estimated Newtonian escape velocities with $1<v_{\rm obs}/v_{\rm escN}\le1.2$. These systems are unlikely to be chance associations and are expected in a nonstandard paradigm such as Milgromian dynamics (MOND). The hypothesis that Newtonian gravity can be extrapolated to the low-acceleration limit is falsified by this independent study with accurate 3D velocities. Future radial velocity monitoring and Speckle interferometric imaging for larger samples will be useful to refine the present result.

Decoders are a critical component of fault-tolerant quantum computing. They must identify errors based on syndrome measurements to correct quantum states. While finding the optimal correction is NP-hard and thus extremely difficult, approximate decoders with faster runtime often rely on uncontrolled heuristics. In this work, we propose a family of hierarchical quantum decoders with a tunable trade-off between speed and accuracy while retaining guarantees of optimality. We use the Lasserre Sum-of-Squares (SOS) hierarchy from optimization theory to relax the decoding problem. This approach creates a sequence of Semidefinite Programs (SDPs). Lower levels of the hierarchy are faster but approximate, while higher levels are slower but more accurate. We demonstrate that even low levels of this hierarchy significantly outperform standard Linear Programming relaxations. Our results on rotated surface codes and honeycomb color codes show that the SOS decoder approaches the performance of exact decoding. We find that Levels 2 and 3 of our hierarchy perform nearly as well as the exact solver. We analyze the convergence using rank-loop criteria and compare the method against other relaxation schemes. This work bridges the gap between fast heuristics and rigorous optimal decoding.

Motivated by recent progress in the high-energy description of gravitational scattering, we develop a systematic Regge-theory framework for $2\to2+n$ amplitudes describing the scattering of two massive particles with $n$ graviton emissions, including spin effects. Working in the ultra-relativistic limit at leading logarithmic accuracy, the massive result smoothly reduces to its massless counterpart. We describe both quantum (Regge trajectory and BFKL $t$-channel evolution) and classical ($s$-channel multi-$H$ evolution) contributions using both an exponential representation of the S-matrix and a shock-wave formalism in light-cone quantisation. In the latter approach, gravitational Wilson lines evolve in rapidity space under a boost-invariant Hamiltonian, providing a space-time realisation of the high-energy dynamics and making contact with recent effective field theory descriptions in the forward limit. As an application, we compute the leading-logarithmic contribution to the massive spinless $2\to2$ amplitude at 5PM-2SF order, recovering the previously determined massless result, and derive the tree-level $2\to3$ amplitude and its associated scattering waveform for Kerr black holes in the ultra-relativistic limit.

We prove the past nonlinear stability of the sub-critical Kasner-scalar field solutions to the Einstein-scalar field equations on a truncated cone domain in spacetime dimensions $n\geq4$. Our analysis demonstrates that the perturbed solutions are asymptotically pointwise Kasner, geodesically incomplete in the contracting direction and terminate at quiescent and crushing singularities characterized by the blow-up of curvature invariants. This work generalizes the result of Beyer-Oliynyk-Zheng in [arXiv:2502.09210v2] to all higher dimensional spacetimes.

The unimodular version of the Kaluza-Klein theory is briefly recalled, and its projection on the $4$-dimensional spacetime is constructed. Imposing unimodularity condition on the $5$-dimensional Kaluza-Klein metric, det$g_{AB}=1$ is equivalent with introducing cosmological term in Einstein's equations in $4$ dimensions, and with scalar field of the Brans-Dicke type. Singularity-free cosmological solutions with scalar field and with matter sources are constructed, and their basic properties analyzed, along the results obtained in previous publications. In the present paper, attention is focussed on the perturbative analysis of cosmological solutions, providing a clue concerning their stability against small fluctuations.

Building upon the long-standing paradigm that dynamics near a spacelike singularity are governed by a sequence of Kasner epochs, we demonstrate that this picture is fundamentally altered when higher-curvature or quantum gravitational corrections are included. By incorporating such terms alongside a minimally coupled scalar field, we discover three distinct dynamical phases near the singularity: modified Kasner eons, persistent periodic oscillations, and oscillatory spike dynamics with growing amplitude. In particular, the Kasner-like geometry persisting only in highly constrained situations. The latter two regimes represent a clean departure from classical Kasner phenomenology, revealing a richer and more ordered landscape of behaviors in the deep interior of black holes beyond Einstein gravity. This work establishes a comprehensive approach for understanding the gravitational nonlinearity in the most extreme gravitational environment.

We embark on the vast program of integrating the dynamics of Loop Quantum Gravity (LQG). Adopting the strategy of decomposing spin network states into small blocks of (quantum) geometry which can later be glued back together, we focus on the more modest objective of studying the Hamiltonian dynamics on the {\it candy graph}, that is two nodes linked together by an arbitrary number of edges and also having open edges. This elementary setting allows both for curvature to develop around the bulk loops and both non-trivial boundary data and dynamics on the open edges. We study this system at the classical level and leave the detailed of its quantum regime for future investigation. Working on a single loop with two external legs, we show how the LQG Hamiltonian ansatz reduces to a pair of non-linear differential equations, similar to the cubic Schrödinger equation, on the areas carried by the bulk links. We provide analytical solutions to this evolution equation, identifying oscillatory modes (bounded modes) and divergent modes (similar to bouncing cosmological trajectories). This provides an explicit template for future investigations of LQG dynamics on more sophisticated spin network architecture built as arrays of candy graphs.

Gravitational wave astronomy typically relies on rigorous, computationally expensive Bayesian analyses. Several methods have been developed to perform rapid Bayesian inference, but they are not yet used to inform our full analyses. We present a novel approach for doing this whilst ensuring that the Bayesian prior remains independent of the data, providing a statistically rigorous way to leverage low-latency information to accelerate the final inference. By combining the fast constraints from the simple-pe algorithm with the nested sampling acceleration technique of posterior repartitioning, we demonstrate that our method can guide the nested sampler towards the most probable regions of parameter space more efficiently for signal-to-noise ratios (SNR) greater than 20, while mathematically guaranteeing that the final inference is identical to that of a standard, uninformed analysis. We validate the method through an injection study, demonstrating that it produces statistically robust and unbiased results, whilst providing speedups of up to $2.2\times$ for binaries with SNRs $< 150$. Importantly, we show that the performance gain provided by our method scales with SNR, establishing it as a powerful technique to mitigate the cost of analysing signals from current and future gravitational-wave observatories.

Symmetry provides powerful non-perturbative constraints in quantum many-body systems. A prominent example is the Lieb-Schultz-Mattis (LSM) anomaly -- a mixed 't Hooft anomaly between internal and translational symmetries that forbids a trivial symmetric gapped phase. In this work, we investigate lattice translation operators in systems with an LSM anomaly. We construct explicit lattice models in two and three spatial dimensions and show that, after gauging the full internal symmetry, translation becomes non-invertible and fuses into defects of the internal symmetry. The result is supported by the anomaly-inflow in view of topological field theory. Our work extends earlier one-dimensional observations to a unified higher-dimensional framework and clarifies their origin in mixed anomalies and higher-group structures, highlighting a coherent interplay between internal and crystalline symmetries.

Erasure qubits are beneficial for quantum error correction due to their relaxed threshold requirements. While dual-rail erasure qubits have been demonstrated with a strong error hierarchy in circuit quantum electrodynamics, biased-erasure qubits -- where erasures originate predominantly from one logical basis state -- offer further advantages. Here, we realize a hardware-efficient biased-erasure qubit encoded in the vacuum and two-photon Fock states of a single microwave cavity. The qubit exhibits an erasure bias ratio of over 265. By using a transmon ancilla for logical measurements and mid-circuit erasure detections, we achieve logical state assignment errors below 1% and convert over 99.3% leakage errors into detected erasures. After postselection against erasures, we achieve effective logical relaxation and dephasing rates of $(6.2~\mathrm{ms})^{-1}$ and $(3.1~\mathrm{ms})^{-1}$, respectively, which exceed the erasure error rate by factors of 31 and 15, establishing a strong error hierarchy within the logical subspace. These postselected error rates indicate a coherence gain of about 6.0 beyond the break-even point set by the best physical qubit encoded in the two lowest Fock states in the cavity. Moreover, randomized benchmarking with interleaved erasure detections reveals a residual logical gate error of 0.29%. This work establishes a compact and hardware-efficient platform for biased-erasure qubits, promising concatenations into outer-level stabilizer codes toward fault-tolerant quantum computation.

We study how energy and quantum entanglement are transferred when two identical CFTs are entangled locally. This is probed by considering a local operator insertion in one of the CFTs. When the CFTs have holographic duals via the AdS/CFT correspondence, the transfer happens through an AdS wormhole that allows signal propagation even beyond the horizon from one AdS boundary to the other; we demonstrate this in explicit CFT calculations. We argue that this transmission is possible because the insertion of a local operator is not a unitary process but a regularized version of projection measurement, and that this is interpreted as quantum teleportation. We also find that this leads to a phenomenon opposite to scrambling, where mutual information, instead of being suppressed, gets enhanced by the insertion of a local operator excitation.

It is a matter of ongoing discussion whether quantum states can become entangled while only interacting via a classical mediator. This lively debate is deeply interwoven with the question of whether entanglement studies can prove the quantum nature of gravity. However, the answer to this fundamental question depends crucially on which hybrid quantum-classical theory is used. In this letter, we demonstrate that entanglement by a classical mediator is possible within the framework of hybrid van Hove theory, showing that existing no-go theorems on that matter do not universally apply to hybrid theories in general. After briefly recapitulating the key features of the hybrid van Hove theory, we show this using the example of two quantum spins coupled by a classical harmonic oscillator. By deriving the spin density matrix for this scenario and comparing it to its equivalent for a pure quantum system, we show that entanglement between the two spins is generated in both cases. Conclusively, this is illustrated by presenting the purity and concurrence of the spin-spin system as a decisive measure for entanglement. Our results further imply that quantum entanglement studies cannot rule out consistent quantum theories featuring classical gravity.

Strassen's asymptotic spectrum offers a framework for analyzing the complexity of tensors. It has found applications in diverse areas, from computer science to additive combinatorics and quantum information. A long-standing open problem, dating back to 1991, asks whether Strassen's support functionals are universal spectral points, that is, points in the asymptotic spectrum of tensors. In this paper, we answer this question in the affirmative by proving that the support functionals coincide with the quantum functionals - universal spectral points that are defined via entropy optimization on entanglement polytopes. We obtain this result as a special case of a general minimax formula for convex optimization on entanglement polytopes (and other moment polytopes) that has further applications to other tensor parameters, including the asymptotic slice rank. Our proof is based on a recent Fenchel-type duality theorem on Hadamard manifolds due to Hirai.

Collective light emission from many-body quantum systems is a cornerstone of quantum optics, yet its implementation in solid-state platforms operating under ambient conditions remains highly challenging. Large-bandgap van der Waals materials such as hexagonal boron nitride (hBN) host stable room-temperature single-photon emitters with narrow linewidths across a broad spectral range. However, cooperative radiative effects in this system have not been previously explored. Here we demonstrate collective emission from quantum-emitter ensembles in hBN layers when the emitters are nearly indistinguishable and positioned within a sub-wavelength proximity. Using confocal microscopy and a Hanbury Brown-Twiss (HBT) configuration, we identify both isolated emitters and ensembles activated by localized electron-beam irradiation. Time-resolved photoluminescence measurements reveal a superlinear intensity enhancement and a pronounced acceleration of the radiative decay in tightly confined ensembles, with lifetimes approaching the temporal resolution of our experimental system (about 500 ps), compared to approximately 1.85 ns for single emitters or large, spatially extended ensembles. Complementary second-order photon-correlation measurements exhibit sub-Poissonian antidip consistent with emission from a few indistinguishable emitters. The simultaneous observation of lifetime shortening and enhanced emission provides direct evidence of cooperative emission at room temperature, achieved without optical cavities or cryogenic cooling. These results establish optically active defect ensembles in hBN as a scalable solid-state platform for engineered collective quantum optics in two-dimensional materials, opening avenues toward ultrabright superradiant light sources and nonclassical photonic states for quantum technologies.

The formation of primordial black holes (PBHs) generally requires large density perturbations, which is widely supported by researchers. This paper studies the local coupling properties of the Starobinsky potential and KKLT potential by introducing a linear Lorentzian-type coupling, which locally breaks the slow roll conditions. We found that both positive and negative coupling can form a considerable abundance of PBH. Additionally, we also studied the scalar-induced gravitational waves (SIGWs) generated by this model.

High-coherence, fault-tolerant and scalable quantum computing architectures with unprecedented long coherence times, faster gates, low losses and low bit-flip errors may be one of the only ways forward to achieve the true quantum advantage. In this context, high-frequency high-coherence (HCQC) qubits with new high-performance topologies could be a significant step towards efficient and high-fidelity quantum computing by facilitating compact size, higher scalability and higher than conventional operating temperatures. Although transmon type qubits are designed and manufactured routinely in the range of a few Giga-Hertz, normally from 4 to 6 GHz (and, at times, up to around 10GHz), achieving higher-frequency operation has challenges and entails special design and manufacturing considerations. This report presents the proposal and preliminary design of an 8-qubit transmon (with possible upgrade to up to 72 qubits on a chip) architecture working beyond an operation frequency of 10GHz, as well as presents a new connection topology. The current design spans a range of around 11 to 13.5 GHz (with a possible full range of 9-12GHz at the moment), with a central optimal operating frequency of 12.0 GHz, with the aim to possibly achieve a stable, compact and low-charge-noise operation, as lowest as possible as per the existing fabrication techniques. The aim is to achieve average relaxation times of up to 1.9ms with average quality factors of up to 2.75 x 10^7 after trials, while exploiting the new advances in superconducting junction manufacturing using tantalum and niobium/aluminum/aluminum oxide tri-layer structures on high-resistivity silicon substrates (carried out elsewhere by other groups and referred in this report).

In this paper, we focus on the problem of computing the set of diagonal transversal gates fixing a CSS code. We determine the logical actions of the gates as well as the groups of transversal gates that induce non-trivial logical gates and logical identities. We explicitly declare the set of equations defining the groups, a key advantage and differentiator of our approach. We compute the complete set of transversal stabilizers and transversal gates for any CSS code arising from monomial codes, a family that includes decreasing monomial codes and polar codes. As a consequence, we recover and extend some results in the literature on CSS-T codes, triorthogonal codes, and divisible codes.

Fluxonium superconducting circuits were originally proposed to realize highly coherent qubits. In this work, we explore how these circuits can be used to implement and harness qutrits, by tuning their energy levels and matrix elements via an external flux bias. In particular, we investigate the distinctive features of arrays of fluxonium qutrits, and their potential for the quantum simulation of exotic quantum matter. We identify four different operational regimes, classified according to the plasmon-like versus fluxon-like nature of the qutrit excitations. Highly tunable on-site interactions are complemented by correlated single-particle hopping, pair hopping and non-local interactions, which naturally emerge and have different weights in the four regimes. Dispersive corrections and decoherence are also analyzed. We investigate the rich ground-state phase diagram of qutrit arrays and propose practical dynamical experiments to probe the different regimes. Altogether, fluxonium qutrit arrays emerge as a versatile and experimentally accessible platform to explore strongly correlated bosonic matter beyond the Bose-Hubbard paradigm, and with a potential toward simulating lattice gauge theories and non-Abelian topological states.

Color centers that enable nuclear-spin control without RF fields offer a powerful route towards simplified and scalable quantum devices. Such capabilities are especially valuable for quantum sensing and computing platforms that already find applications in biology, materials science, and geophysics. A key challenge is the coherent manipulation of nearby nuclear spins, which serve as quantum memories and auxiliary qubits but conventionally require additional high-power RF fields which increase the experimental complexity and overall power consumption. Finding systems where both electron and nuclear spins can be controlled using a single MW source is therefore highly desirable. Here, using a modified divacancy center in silicon carbide, we show that coherent control of a coupled nuclear spin is possible without any RF fields. Instead, MW pulses driving the electron spin also manipulate the nuclear spin through hyperfineenhanced effects, activated by a precisely tilted external magnetic field. We demonstrate high-fidelity nuclear-spin control, achieving 89% two-qubit tomography fidelity and nearly T1-limited nuclear coherence times. This approach offers a simplified and scalable route for future quantum applications.

Benchmarking physical devices and verifying logical algorithms are important tasks for scalable fault-tolerant quantum computing. Numerous protocols exist for benchmarking devices before running actual algorithms. In this work, we show that both physical and logical errors of fault-tolerant circuits can even be characterized in-situ using syndrome data. To achieve this, we map general fault-tolerant Clifford circuits to subsystem codes using the spacetime code formalism and develop a scheme for estimating Pauli noise in Clifford circuits using syndrome data. We give necessary and sufficient conditions for the learnability of physical and logical noise from given syndrome data, and show that we can accurately predict logical fidelities from the same data. Importantly, our approach requires only a polynomial sample size, even when the logical error rate is exponentially suppressed by the code distance, and thus gives an exponential advantage against methods that use only logical data such as direct fidelity estimation. We demonstrate the practical applicability of our methods in various scenarios using synthetic data as well as the experimental data from a recent demonstration of fault-tolerant circuits by Bluvstein et al. [Nature 626, 7997 (2024)]. Our methods provide an efficient, in-situ way of characterizing a fault-tolerant quantum computer to help gate calibration, improve decoding accuracy, and verify logical circuits.

The interaction between spin and gravitational waves causes spinning bodies to deviate from their geodesics. In this work, we obtain the complete analytic solution of the Mathisson--Papapetrou--Dixon equations at linear order in the spin for a general plane gravitational wave with arbitrary polarization profiles. Our approach combines a parallel-transported tetrad with the translational Killing symmetries of plane wave spacetimes, yielding six conserved quantities that fully determine the momentum, spin evolution, and worldline. The resulting transverse and longitudinal motions are expressed in closed form as single integrals of the retarded time, providing a unified and model-independent framework for computing spin--curvature-induced deviations for realistic or theoretical gravitational-wave signals. This analytic solution offers a versatile tool for studying spin-dependent effects in gravitational memory, Penrose-limit geometries, and future high-precision space-based detectors.

We propose an optimized adiabatic-impulse (OAI) protocol that achieves much shorter evolution time while preserving the Kibble-Zurek scaling. Near the critical regime, the control field is linearly ramped across the quantum critical point at a rate characterized by a quench time $τ_Q$. Away from the critical regime, the evolution is designed to follow the threshold of adiabatic breakdown, which we characterize by an adiabatic coefficient $ζ\proptoτ_Q^α$. As a consequence, the total evolution time exhibits a sublinear power-law dependence on $τ_Q$, and the conventional linear quench protocol is recovered in the limit $α\rightarrow\infty$. We apply the OAI protocol to the transverse Ising chain and numerically determine the minimum value of $ζ$. We further investigate the nonequilibrium dynamics in the presence of a noisy field that can induce anti-Kibble-Zurek behavior, leading to more defects for slower ramps. Within the OAI protocol, the optimal quench time that minimizes defects obeys an altered universal power-law scaling with the noise strength. Finally, we generalize the OAI protocol to incorporate nonlinear Kibble-Zurek scaling.

This study investigates the role of EPR steering in characterizing the energy dynamics of quantum batteries (QBs) within \textcolor{black}{a charging system that features shared reservoirs. After optimizing parameter configurations to achieve high-energy systems, we observe across a variety of charging scenarios with low-dissipation regimes that steering serves as a vital resource: it is initially stored until the system reaches energy equilibrium, and then subsequently utilized to sustain the enhancement of energy storage. Furthermore, steering acts as a witness to battery population balance and a consumable that enhances extractable work. Additionally, we discuss the contribution of the steering potential to energy upon high-dissipation charging in details. These findings establish a novel indicator for monitoring QB energy variations, which will be beneficial to achieve the high-performance quantum batteries.

We provide a theoretical framework to describe the dynamics of a free-electron beam interacting with quantized bound systems in arbitrary electromagnetic environments. This expands the quantum optics toolbox to incorporate free-electron beams for applications in highly tunable quantum control, imaging, and spectroscopy at the nanoscale. The framework recovers previously studied results and shows that electromagnetic environments can amplify the intrinsically weak coupling between a free-electron and a bound electron to reach previously inaccessible interaction regimes. We leverage this enhanced coupling for experimentally feasible protocols in coherent qubit control and towards the nondestructive readout and projective control of the electron beam's quantum-number statistics. Our framework is broadly applicable to microwave-frequency qubits, optical nanophotonics, cavity quantum electrodynamics, and emerging platforms at the interface of electron microscopy and quantum information.

Phantom divide line transitions are not possible in the context of single scalar field scalar-tensor theories. In this article we study a combined framework of a tachyonic minimally coupled single scalar field theory in the presence of an $\mathcal{R}^2$ correction term and with a rescaled Einstein-Hilbert term of the form $\sim λ\frac{\mathcal{R}}{16πG}$. Such terms can be part of an $f(\mathcal{R})$ gravity which in the large curvature regime yields such correction terms effectively. Alternatively, such terms can simply be quantum corrections to the scalar field action. We aim to answer two questions, firstly if this framework can lead to phantom divide line transitions and secondly whether the resulting model can be compatible with the ACT data. The model we studied is an inverse square power-law model, well known from tachyon inflation models. As we show, the field equations can be cast in terms of the scalar field solely, however the resulting theory is distinct from a single scalar field theory, because the phantom divide line is crossed during inflation. Thus initially the tachyonic nature of the scalar field generates a phantom equation of state parameter, and during inflation the phantom divide line is crossed, with the effective equation of state parameter at the end of inflation being $w=-1/3$ which corresponds to the non-accelerating state of the Universe. The model is proved to be compatible with the ACT data, only when the gravity during inflation is stronger than Einstein-Hilbert gravity, with the effective gravitational constant during inflation being $\frac{G}λ$. The effective theory is valid only during inflation, thus Big-Bang nucleosynthesis is not affected by the rescaling of the Einstein-Hilbert gravity. The feature of a phantom crossing in $f(\mathcal{R},φ)$ frameworks is new in the literature.

General relativity has its successes at the local astrophysical level, however, it seems to be insufficient in describing the Universe at large scales. In this work we investigate how the most general field theories in the context of general relativity can accomodate a phantom-to-quintessence transition which may be essential element of realistic Dark Energy scenarios in the late Universe. As we demonstrate in a very detailed manner, this is impossible for a canonical and minimally coupled single scalar field theory, but it may be possible for ghost condensate theories like $k$-essence theories. We point out how the ghost instabilities may be eliminated, and we analyze the quantitative features of a $k$-essence theory that may realize a phantom-to-quintessence transition in the late Universe. We also qualitatively compare the difficulties and fine-tunings required for $k$-essence theories to realize a phantom-to-quintessence transition, and how such a transition is naturally realized in modified gravity, without unnecessary fine-tunings and ghost eliminations.

We uncover the global topological phase diagram of one-dimensional discrete-time quantum walks driven by Fibonacci-modulated coin parameters. Utilizing the mean chiral displacement (MCD) as dynamical probe, we identify robust topological phases defined by a strictly quantized winding number $ν=-1$ and exponentially localized edge states. Crucially, we discover that these topological edge modes emerges not only at zero energy but also at the quasienergy zone boundary $E=π$, exhibiting identical localization robustness despite the fractal nature of the bulk spectrum. These results demonstrate that Floquet topological protection remains intact amidst quasiperiodic disorder, offering a concrete route to observing exotic non-equilibrium phases in photonic experiments.

Belief propagation with quantum messages (BPQM) provides a low-complexity alternative to collective measurements for communication over classical--quantum channels. Prior BPQM constructions and density-evolution (DE) analyses have focused on binary alphabets. Here, we generalize BPQM to symmetric q-ary pure-state channels (PSCs) whose output Gram matrix is circulant. For this class, we show that bit-node and check-node combining can be tracked efficiently via closed-form recursions on the Gram-matrix eigenvalues, independent of the particular physical realization of the output states. These recursions yield explicit BPQM unitaries and analytic bounds on the fidelities of the combined channels in terms of the input-channel fidelities. This provides a DE framework for symmetric q-ary PSCs that allows one to estimate BPQM decoding thresholds for LDPC codes and to construct polar codes on these channels.

Continual intrusion detection must absorb newly emerging attack stages while retaining legacy detection capability under strict operational constraints, including bounded compute and qubit budgets and privacy rules that preclude long-term storage of raw telemetry. We propose QCL-IDS, a quantum-centric continual-learning framework that co-designs stability and privacy-governed rehearsal for NISQ-era pipelines. Its core component, Q-FISH (Quantum Fisher Anchors), enforces retention using a compact anchor coreset through (i) sensitivity-weighted parameter constraints and (ii) a fidelity-based functional anchoring term that directly limits decision drift on representative historical traffic. To regain plasticity without retaining sensitive flows, QCL-IDS further introduces privacy-preserved quantum generative replay (QGR) via frozen, task-conditioned generator snapshots that synthesize bounded rehearsal samples. Across a three-stage attack stream on UNSW-NB15 and CICIDS2017, QCL-IDS consistently attains the best retention-adaptation trade-off: the gradient-anchor configuration achieves mean Attack-F1 = 0.941 with forgetting = 0.005 on UNSW-NB15 and mean Attack-F1 = 0.944 with forgetting = 0.004 on CICIDS2017, versus 0.800/0.138 and 0.803/0.128 for sequential fine-tuning, respectively.

Stochastic optimization of Neural Network Quantum States (NQS) in discrete Fock spaces is limited by sampling variance and slow mixing. We present a deterministic framework that optimizes a neural backflow ansatz within dynamically adaptive configuration subspaces, corrected by second-order perturbation theory. This approach eliminates Monte Carlo noise and, through a hybrid CPU-GPU implementation, exhibits sub-linear scaling with respect to subspace size. Benchmarks on bond dissociation in H2O and N2, and the strongly correlated chromium dimer Cr2, validate the method's accuracy and stability in large Hilbert spaces.

In generalized Proca theories, characterized by a vector field with broken $U(1)$ gauge invariance, late-time cosmic acceleration can be realized with a dark energy equation of state in the regime $w_{\rm DE} < -1$. In such scenarios, however, a phantom-divide crossing, as recently suggested by DESI observations, is not achieved without encountering theoretical inconsistencies. We incorporate a canonical scalar field with a potential, in addition to the vector field, and show that the phantom-divide crossing from $w_{\rm DE} < -1$ to $w_{\rm DE} > -1$ can occur at low redshifts. We propose a minimal model that admits such a transition and identify the region of parameter space in which all dynamical degrees of freedom in the scalar, vector, and tensor sectors are free from ghost and Laplacian instabilities. We further investigate the evolution of linear cosmological perturbations by applying the quasi-static approximation to modes well inside the Hubble radius. The dimensionless quantities $μ$ and $Σ$, which characterize the growth of matter perturbations and the bending of light rays, respectively, depend on the sound speed $c_ψ$ of the longitudinal scalar perturbation associated with the vector field. Since $c_ψ$ is influenced by the transverse vector mode, the model exhibits sufficient flexibility to yield values of $μ$ and $Σ$ close to 1. Consequently, unlike theories such as scalar Galileons, the present model can be consistent with observations of redshift-space distortions and integrated Sachs-Wolfe-galaxy cross-correlations.

We develop an information-theoretic analysis of Quantum-Memory-Free (QMF) Quantum Secure Direct Communication (QSDC) under collective attacks as an alternative to the conventional Quantum Key Distribution (QKD) protocol with one-time pads. Our main contributions are: 1) a QMF-QSDC protocol that only relies on universal hashing of coded sequences without wiretap coding; 2) a set of privacy amplification theorems for extracting secrecy from coded classical sequences against quantum side-information. These tools open the way to the design of robust QMF-QSDC protocols.

We study the localization and scattering of a single photon in a waveguide coupled to qubit arrays with quasiperiodic spacings. As the quasiperiodic strength increases, localized subradiant states with extremely long lifetime appear around the resonant frequency and form a continuum band. In stark contrast to the fully disordered waveguide QED where all states are localized, we analytically find that the fraction of localized states is up to $(3-\sqrt{5})/2$ when the modulation frequency is $(1+\sqrt{5})/2$. The localized and delocalized states can be related to excitation in flat and curved inverse energy bands under the approximation of large-period modulation. When the quasiperiodic strength is weak, an extended subradiant state can support the transmission of a photon. However, as the quasiperiodic strength increases, localized subradiant states can completely block the transmission of a single photon in resonance with the subradiant states, and enhance the overall reflection. At a fixed quasiperiodic strength, we also find mobility edge in transmission spectrum, below and above which the transmission is either turned on and off as system size increases. Our work give new insights into the localization in non-Hermitian systems.

Superradiance is a quantum phenomenon in which coherence between emitters results in enhanced and directional radiative emission. Many quantum optical phenomena can be characterized by the two-time quantum correlation function $g^{(2)}(t,τ)$, which describes the photon statistics of emitted radiation. However, the critical task of determining $g^{(2)}(t,τ)$ becomes intractable for large emitter ensembles due to the exponential scaling of the Hilbert space dimension with the number of emitters. Here, we analyse and benchmark two approximate numerical sampling methods applicable to emitter arrays embedded within electromagnetic environments, which generally provide upper and lower bounds for $g^{(2)}(t,0)$. We also introduce corrections to these methods (termed offset corrections) that significantly improve the quality of the predictions. The optimal choice of method depends on the total number of emitters, such that taken together, the two approaches provide accurate descriptions across a broad range of important regimes. This work therefore provides new theoretical tools for studying the well-known yet complex phenomenon of superradiance in large ensembles of quantum emitters.

Photons are among the most important carriers of quantum information owing to their rich degrees of freedom (DoFs), including various spatiotemporal structures. The ability to characterize these DoFs, as well as the hidden correlations among them, directly determines whether they can be exploited for quantum tasks. While various methods have been developed for measuring the spatiotemporal structure of classical light fields, owing to the technical challenges posed by weak photon flux, there have so far been no reports of observing such structures in their quantum counterparts, except for a few studies limited to correlations within individual DoFs. Here, we propose and experimentally demonstrate a self-referenced, high-efficiency, and all-optical method, termed 3D imaging of photonic wave packets, for comprehensive characterization of the spatiotemporal structure of a quantum light field, i.e., the biphoton spatiotemporal wave packet. Benefiting from this developed method, we successfully observe the spatial-spatial, spectral-spectral, and spatiotemporal correlations of biphotons generated via spontaneous parametric down-conversion, revealing rich local and nonlocal spatiotemporal structure in quantum light fields. This method will further advance the understanding of the dynamics in nonlinear quantum optics and expand the potential of photons for applications in quantum communication and quantum computing.

We study the extraction of quantum coherence by three static Unruh-DeWitt (UDW) detectors that interact locally with a massless scalar vacuum field in the vicinity of an infinite perfectly reflecting boundary. Depending on the setup, the detectors are positioned either parallel or orthogonal to the boundary, with their energy gaps chosen to satisfy the hierarchy $Ω_C\geq Ω_B\geq Ω_A$. Our analysis reveals that decreasing the detector-boundary separation leads to a monotonic degradation of quantum coherence, whereas the same boundary effect can simultaneously preserve and even amplify the harvested quantum entanglement. Moreover, when the detectors possess distinct energy gaps, coherence extraction is further inhibited; strikingly, such non-identical configurations substantially enhance the efficiency of entanglement harvesting and markedly extend the range of detector separations over which non-negligible entanglement can be generated. Nevertheless, the harvesting of nonlocal quantum coherence is achievable over a significantly broader range of detector separations than that of quantum entanglement. Despite exhibiting similar overall behavior, orthogonal detector configurations outperform parallel ones in coherence harvesting, highlighting the quantitative influence of detector geometry. Overall, our study reveals a hierarchical distinction between quantum coherence and entanglement as operational resources in structured vacuum fields: quantum coherence is not only more readily accessible across space but also more robust than entanglement, whereas entanglement exhibits richer features and can be selectively activated and enhanced through boundary effects and detector non-uniformity.

This paper investigates the holographic network connecting different CFTs, modeled by Gauss-Bonnet gravity with varying couplings across different bulk branches. By applying the holographic Noether's theorem, we prove that the junction condition on the Net-brane leads to conservation laws at network nodes. We analyze the stability of the gravitational KK modes on the Net-brane and derive the constraints on theory parameters. Additionally, we discuss various proposals for network entropy, confirm that the type I and II network entropies obey the holographic g-theorem, and show that the type III network entropy is non-negative. We explore the two-point functions of various NCFTs at different edges, using examples like free scalars and the AdS/NCFT with a tensionless brane. We then examine the gravitational dual of compact networks, which feature both EOW branes and Net-branes in the bulk. We derive the joint condition for EOW branes at the Net-brane and analyze vacuum solutions in AdS$_3$/NCFT$_2$. Finally, we demonstrate that AdS/NCFT provides a natural way to envision traversable parallel universes that have different geometries and physical laws. Remarkably, unlike traversable wormholes, our model of parallel universes satisfies all the energy conditions.

Exact, analytic, asymptotically flat rotating black-hole solutions are exceedingly rare, with only a handful of examples known. Using a Kerr-Schild ansatz, we derive a multitude of exact, analytic, asymptotically flat rotating black-hole solutions within a broad class of Generalized Proca theories. These black holes differ significantly from Kerr black holes, as they possess primary hair and are non-circular, thus breaking a symmetry that vacuum black holes exhibit in General Relativity.

Belenchia et al. [Phys. Rev. D 98, 126009 (2018)] have analyzed a gedankenexperiment where two observers, Alice and Bob, attempt to communicate via superluminal signals using a superposition of massive particles dressed by Newtonian fields and a test particle as field detector. Quantum fluctuations in the particle motion and in the field prevent signaling or violations of quantum mechanics in this setup. We reformulate this thought experiment by considering gravitational waves emitted by an extended quadrupolar object as a detector for Newtonian tidal fields. We find that quantum fluctuations in the gravitational waves prevent signaling. In the Newtonian limit, rotating black holes behave as extended quadrupolar objects, as consequence of the strong equivalence principle. It follows that consistency of the Newtonian limit of general relativity with quantum mechanics requires the quantization of gravitational radiation, even when the waves originate in strong gravity sources.

Precise and flexible manipulation of the motional state of ultracold atoms is a fundamental enabling technology for diverse applications such as quantum sensing and quantum computation. In this paper we propose a general, simple and highly efficient method to engineer the motional state of a Bose-Einstein condensate with time-dependent optical fields, which can be realized experimentally with existing light sculpting techniques. We demonstrate numerically how to engineer superpositions of persistent currents in a toroidal trap, achieving very high fidelity. We also study in detail the stability of the state over time, and we present an analytical two-state model that approximates well the evolution of the state in presence of self-interactions.

We establish efficient algorithms for weakly-interacting quantum spin systems at arbitrary temperature. In particular, we obtain a fully polynomial-time approximation scheme for the partition function and an efficient approximate sampling scheme for the thermal distribution over a classical spin space. Our approach is based on the cluster expansion method and a standard reduction from approximate sampling to approximate counting.

We study decentralized multi-agent coordination where agents must correlate actions against an unobserved field and cannot communicate. To isolate correlation geometry from payoff optimization, we introduce the Hidden-Field Coordination (HFC) model, which enforces identical information access and no-signaling constraints across strategies. Using information-theoretic diagnostics, we compare classical shared-randomness baselines with an entanglement-mediated strategy based on multipartite W states and a strictly local Spontaneous Leader Election rule. Within the restricted symmetric shared-latent baseline studied here, increasing total correlation is achieved primarily by driving actions toward alignment (copying), which also increases pairwise coincidence (collisions). By contrast, the quantum strategy realizes a collision-suppressing coordination regime: it preserves global dependence while reducing pairwise coincidence below the independent (product) baseline induced by the common marginal distribution. This produces a geometric separation in the joint-action distribution. Classical baselines concentrate probability near the diagonal of action equality, whereas the entanglement-mediated mapping occupies an offset-diagonal region associated with relational roles. Accordingly, the entanglement signature in this setting is not higher correlation magnitude; total-correlation differentials can be negative relative to the classical copying optimum. Instead, it reflects a change in dependence geometry that supports robust anti-coordination.

We apply the covariant entropy bound argument supporting the de Sitter swampland conjecture to the quintessence model, to find out the condition for the background to be unstable. More concretely, the background is unstable when the matter entropy given by the species number of the effective field theory increases more rapidly than the geometrical entropy proportional to the apparent horizon area, since it contradicts the covariant entropy bound. The rapid increase in the matter entropy is proposed by the distance conjecture, which states that the time evolution of some scalar field along the geodesic in the field space brings about the descent of a tower of states from UV. From this, we find that for the quintessence model, the accelerating background having the event horizon is unstable, and the instability condition as well as the lifetime of the unstable background is equivalent to the trans-Planckian censorship bound forbidding the classicalization of the trans-Planckian modes. We also point out that the scale separation between the Kaluza-Klein mass scale and the Hubble parameter can be realized when the product between the increasing rates of the matter and the geometrical entropies is bounded from below, which is consistent with the AdS distance conjecture. Our study suggests that various swampland conjectures can be comprehensively understood in the language of the entropy.

When the length of an optical cavity is modulated, theory predicts exponential concentration of energy around particular space-time trajectories. Viewed stroboscopically, photons in such a driven cavity propagate as if in a curved spacetime, with black hole and white hole event horizons corresponding to unstable and stable fixed points of the evolution. Such phenomena have resisted direct experimental realization due to the difficulty of relativistically accelerating massive cavity mirrors. We report results of an experiment which overcomes this limitation by exchanging the roles of light and matter. A matter wave endowed with quasi-relativistic dispersion is confined between two barriers made of light, one of which is periodically translated at speeds comparable to the matter wave group velocity. In this strongly-modulated cavity we observe the emergence of the predicted bright and dark fixed point trajectories, and demonstrate that changing the modulation waveform can vary the number of fixed points and exchange their stability character. We observe signatures of nontrivial dynamics beyond those predicted for photons, and attribute them to residual curvature in the dispersion relation. In addition to experimentally realizing and characterizing cavity dynamics in the ultra-strong driving regime, these results point the way to implementations of related dynamics in electro-optic materials, with potential applications in pulse generation and signal compression.

We realize a nanoscale accelerometer exploiting the nonequilibrium dynamics of a nanoparticle near the quantum ground state. We explore the dynamics after quenching the trapping potential and find that rapid quenching provides an instance at which the sensitivity is enhanced due to the minimized uncertainty in the position. With rapid quenching, the observed sensitivity is in good agreement with a numerical simulation based on the quantum Langevin equation and approaches to the limit given by the quantum Fisher information. Our results open up a pathway to quantum inertial sensing sensitized by exploiting quench dynamics.

We compute deviations from observed gravitational wave signals, where the amplitude of the signal is unchanged. As an example, we consider the detectability of low lying dynamical tides in binary neutron star or neutron star black hole mergers. Tidal forces can excite oscillatory modes of one or both of the stars in the binary when the orbital frequency of the binary system sweeps through the resonant mode frequency dissipating energy into the vibrational mode. The orbital energy loss to the vibrational mode extracts energy from the orbital motion, advancing the time to merger. The inspiral then continues with an excess phase and a time advance. Both will cause a mismatch when fitting to a system that has not gone through the resonance. To resolve this effect, we compute the mismatch for current and planned detectors using both a quasi-analytical approach that relies on the computation of moment integrals and an optimized version of the standard numerical match function. We conclude that detectability can occur for time advances of the order of 1 ms with advanced LVK detectors for an excess energy-flux that is a few percent of the gravitational wave emission. Our results contrast with previous work, which model this effect solely as a phase shift of the waveform or by using the difference in the number of cycles induced by the resonant behavior. We show that tidal resonance effects primarily cause a time advance of the merger, rather than a phase difference, and that the single-frequency approximation commonly used in the literature significantly overestimates the detectability of this effect.

Several nontrivial phenomena emerge when a quantum field is subjected to dynamical perturbations, with prominent examples including the Hawking and Unruh effects, as well as the dynamical Casimir effect. In this work, we compute the number of particles produced via the dynamical Casimir effect in an ideal cavity, where one of the mirrors is allowed to move under the influence of a gravitational wave. Assuming an oscillatory mirror motion and a plane gravitational wave, we identify the resonance conditions that lead to an exponential increase in the number of created particles through parametric amplification.

Subwavelength arrays of atoms trapped in optical lattices or tweezers are inherently susceptible to deformations: Optomechanical forces produce lattice distortions, which, in turn, modify the optical response of the array. We show that this coupling hybridizes collective atomic excitations (polaritons) with phonons, forming polaron-polaritons -- the fundamental quasiparticles governing light-matter interactions in arrays of trapped atoms. Using analytical polaron theory and numerical simulations, we show that: (1) phonons can strongly enhance the decay of subradiant states, but also enable their efficient excitation; (2) transport of dark excitations remains remarkably robust even at low trap frequencies, except when a polariton can resonantly scatter phonons; and (3) motion reduces the reflectivity of a two-dimensional atomic mirror, however, we identify mechanisms that mitigate this degradation and restore reflectivity above 99% in some cases. Our findings lay the foundation for analyzing motional effects in key applications and suggest new ways to harness them in state-of-the-art experiments.

We have performed Zeno protective measurements of quantum polarization states by coupling the polarization to a temporal pointer (arrival time) in a birefringent optical fiber. It is necessary to actively stabilize the polarization, and we do this by using the signal photon counts themselves as the error signal in a feedback loop. We compare these measurements to a stabilization scheme using a classical reference beam as the error signal. The method using photon counts has higher signal levels and significantly reduced background. These improvements allow us to increase the number of Zeno stages in our measurements from 9 to 13, with a corresponding decrease in the measurement uncertainty.

We investigate the hypothesis that the evolution of the Universe can be described by a single dark fluid whose effective equation of state (EoS), $ω_{\rm{eff}}$, is a linear combination of a logarithmic term and a power law term, both involving the Lambert $W$ function. This particular form of EoS was first proposed by S. Saha and K. Bamba in 2019 and has two parameters, $θ_1$ and $θ_2$, which must be determined from observations. To this end, we place limits on these parameters by combining recent baryon acoustic oscillation (BAO) data -- including measurements from the Dark Energy Spectroscopic Instrument (DESI) -- with Type Ia supernova observations from the Pantheon+ compilation, along with direct determinations of the Hubble parameter. From this combined analysis, we obtain a best-fit value for the Hubble parameter, $H_0 = 67.4 \pm 1.2~\text{km\,s}^{-1}\text{Mpc}^{-1}$, while current measurements of the sound horizon at the baryon drag epoch yield $r_d = 146\pm 2.5$~Mpc. Furthermore, we study the evolution of the deceleration parameter, the effective EoS, and the jerk parameter, and support our findings using the $Om(z)$ diagnostic. The model exhibits noticeable deviations from the predictions of the concordance $Λ$CDM model. Despite these differences, our results indicate that the model provides a coherent description of late-time cosmic evolution and the observed accelerated expansion of the Universe. Finally, we assess the observational viability of the model using information criteria, particularly the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC), and compare these results with those obtained for the $Λ$CDM model, which serves as our reference.

In this article we first develop novel Rindler-type representations of flat spacetime by demonstrating that the standard hyperbolic transformation is a member of an infinite family of coordinate mappings. We specifically introduce cyclic coordinates, which, in contrast to the conventional Rindler wedge, delineate a compact region of Minkowski spacetime. By extending this framework, and motivated by near horizon coordinates in Schwarzschild metric, we propose a class of Spherical Rindler metrics. We demonstrate the utility of this approach by deriving and analyzing a black hole solution and a cosmological metric, both emerging naturally from a Spherical Rindler origin. Our results highlight unique geometric properties of these solutions, providing new insights into the relationship between accelerated frames and global spacetime curvature.

Many early universe scenarios predict post-inflationary tensor perturbations from causality-limited, sub-horizon sources. While the microphysical details may differ, as long as these sources are bounded in duration and correlation length, their tensor power spectra exhibit a universal scaling behavior at small wavenumber: $P_h(k) \propto k^3$, corresponding to white noise on super-horizon scales at the time of production. If these early causal tensor sources (ECTs) exclusively produce gravitational waves before redshift $z \sim 10^5$, this scaling is realized on all of the scales observed in the cosmic microwave background (CMB), and thus yields a universal multipole distribution for the $B$-mode angular power spectrum. Unlike the scale-invariant distributions of inflationary $B$ modes, ECTs generically predict enhanced power on small scales and suppressed power on large scales, which allows these source classes to be distinguished given measurements over a sufficient range of angular scales. In this paper, we introduce a unified framework for characterizing ECTs and demonstrate how their universal infrared scaling manifests in low-frequency observables, including CMB $B$ modes and stochastic gravitational wave spectral densities. We illustrate this mapping with representative case studies of this universality class involving first-order phase transitions, topological defects, and enhanced scalar perturbations, which source tensor modes at second order in perturbation theory.

In General Relativity, the tidal Love numbers of black holes vanish, implying they are resistant to tidal deformation. This "rigidity" is easily broken in the presence of higher-derivative corrections. Focusing on extremal charged black holes in Einstein-Maxwell EFT, we compute the static linear response for both the vector ($\ell=1$) and parity-odd tensor ($\ell \ge 2$) sectors. We find that the resulting tidal Love numbers are non-zero and exhibit logarithmic running, a hallmark of quantum corrections. Crucially, we show that the sign of these deformations is not arbitrary; the induced electric and magnetic susceptibilities and their log runnings in the $\ell=1$ sector are constrained by unitarity and the Weak Gravity Conjecture. Furthermore, due to gravito-electromagnetic mixing, we find the cross log runnings and show that they are the same, which we explain through the worldline effective field theory.

The tensionless limit probes the very high energy regime of string theory in contrast to the well studied point-particle limit which reduces to Einstein gravity. Tensionless strings sweep out null worldsheets in the target space and hence are also called null strings. This article aims to provide a comprehensive review of tensionless null string theory beginning with the initial work of Schild, and continuing to the foundational work of Isberg et al (ILST) and then focussing on developments in the past decade. Recent work centres on the emergence of the Carrollian Conformal Algebra as residual worldsheet symmetries of the ILST action and the identification of tensionless limit as a worldsheet Carrollian limit on the string worldsheet. Carrollian structures are used to address the classical and quantum aspects of the null string. In the classical theory, the aforementioned limit agrees with the analysis from the ILST action. Symmetries, constraints, mode expansions computed from both perspectives match nicely providing a robust cross-check of the analyses. We discuss closed and open null strings as well as their supersymmetric cousins. The quantum null string comes with several surprises, the foremost of which is the emergence of three consistent quantum theories from the ILST action. We detail the canonical quantisation and the spectrum of the triumvirate of theories. We discuss the novelties of the quantum null theories and the effect compactifaction has on them. We also discuss Carroll strings, applications of these ideas to strings approaching black holes and give a quick overview of other related developments.

Topological quantum computation encodes quantum information in the internal fusion space of non-Abelian anyonic quasiparticles, whose braiding implements logical gates. This goes beyond Abelian topological order (TO) such as the toric code, as its anyons lack internal structure. However, the simplest non-Abelian generalizations of the toric code do not support universality via braiding alone. Here we demonstrate that such minimally non-Abelian TOs can be made universal by treating anyon fusion as a computational primitive. We prepare a 54-qubit TO wavefunction associated with the smallest non-Abelian group, $S_3$, on Quantinuum's H2 quantum processor. This phase of matter exhibits cyclic anyon fusion rules, known to underpin universality, which we evidence by trapping a single non-Abelian anyon on the torus. We encode logical qutrits in the nonlocal fusion space of non-Abelian fluxes and, by combining an entangling braiding operation with anyon charge measurements, realize a universal topological gate set and read-out, which we further demonstrate by topologically preparing a magic state. This work establishes $S_3$ TO as simple enough to be prepared efficiently, yet rich enough to enable universal topological quantum computation.

In this work, we study the spectral form factor of random matrix models which exhibit a large number of degenerate ground states accompanied by a macroscopic gap in the spectrum. The central aim of this work is to understand how the standard narrative about the behavior of the spectral form factor is modified in the presence of these parametrically large number of ground states. We show that, at sufficiently low temperatures, the spectral form factor is dominated by the disconnected contribution, even at arbitrarily late times. Moreover, we demonstrate that the connected form factor only depends on the eigenvalues of the non-degenerate sector. Using the Christoffel-Darboux kernel, we analyze a number of examples including the Bessel model and $\mathcal{N}=2$ Jackiw-Teitelboim supergravity. In these examples, we find damped oscillations in the disconnected form factor, with a period set by the inverse size of the gap. Furthermore, we demonstrate that the slope of the ramp in the connected form factor arises from a universal sine-kernel, which emerges from a truncation of the full non-perturbative kernel in the $\hbar \to 0$ limit, and find agreement with the leading double trumpet result. Finally, we present predictions for how the ramp will transition to a plateau in the connected form factor and demonstrate how the transition depends on the details of the leading spectral density of states.

We demonstrate numerically the existence of solutions of five-dimensional vacuum gravity describing the formation, in finite time, of an extremal rotating black hole from a pre-existing Schwarzschild black hole. This is the first example of a violation of the third law of black hole mechanics in vacuum gravity and demonstrates that the third law is false independently of any matter model. We also demonstrate the existence of solutions describing the formation, in finite time, of an extremal rotating black hole from vacuum initial data that does not contain a black hole.

Quantum metrology involves the application of quantum resources to enhance measurements. Several communities have developed quantum-metrology strategies that leverage effective time reversals. These strategies, we posit, form four classes. First, echo metrology begins with a preparatory unitary and ends with that unitary's time-reverse. The protocol amplifies the visibility of a small parameter to be sensed. Similarly, weak-value amplification enhances a weak coupling's detectability. The technique exhibits counterintuitive properties captured by a retrocausal model. Using the third strategy, one simulates closed timelike curves, worldlines that loop back on themselves in time. The fourth strategy involves indefinite causal order, which characterises channels applied in a superposition of orderings. We review these four strategies, which we unify under the heading of time-reverse metrology. We also outline opportunities for this toolkit in quantum metrology; quantum information science; quantum foundations; atomic, molecular, and optical physics; and solid-state physics.

Binary neutron star mergers are expected to generate intense magnetic fields that power relativistic and non-relativistic outflows and shape their multimessenger signatures. These fields likely arise from the turbulent amplification of initially weak magnetic fields during the merger, particularly via the Kelvin-Helmholtz instability at the collisional interface between the stars. While previous studies have shown efficient amplification to magnetar-level strengths, the degree of large-scale coherence of the resulting field remains uncertain. We present general-relativistic, dynamical spacetime, magnetohydrodynamic simulations following the evolution of initially weak, pulsar-like magnetic fields in a binary neutron star merger. We find rapid magnetic field growth at small scales with clear signatures of small-scale turbulent dynamo action. At the highest resolutions, we additionally observe the emergence of coherent magnetic structures on larger scales. Our results imply that strong, ordered magnetic fields may be present immediately after merger, with important implications for the subsequent evolution of the remnant and its observable electromagnetic and gravitational-wave signals.

Quantum state tomography (QST) is essential for validating quantum devices but suffers from exponential scaling in system size. Neural-network quantum states, such as Restricted Boltzmann Machines (RBMs), can efficiently parameterize individual many-body quantum states and have been successfully used for QST. However, existing approaches are point-wise and require retraining at every parameter value in a phase diagram. We introduce a parametric QST framework based on a hypernetwork that conditions an RBM on Hamiltonian control parameters, enabling a single model to represent an entire family of quantum ground states. Applied to the transverse-field Ising model, our HyperRBM achieves high-fidelity reconstructions from local Pauli measurements on 1D and 2D lattices across both phases and through the critical region. Crucially, the model accurately reproduces the fidelity susceptibility and identifies the quantum phase transition without prior knowledge of the critical point. These results demonstrate that hypernetwork-modulated neural quantum states provide an efficient and scalable route to tomographic reconstruction across full phase diagrams.

This study presents a mathematical model of the spatial and rotational motion of a nanodiamond in an inverted harmonic potential to create a macroscopic quantum spatial superposition. The model is based on the Stern-Gerlach Interferometer (SGI) scheme, which utilises linear and quadratic magnetic fields to generate a harmonic potential (linear magnetic field) and a non-linear potential (non-linear/quadratic magnetic field). By incorporating two-dimensional dynamics into the model, we provide a more realistic and accurate depiction of nanoparticle dynamics in linear and inverted harmonic potentials and explore the interaction between motion in a two-dimensional plane. Importantly, we derive the equations of motion for the rotational degrees of freedom, i.e. libration, precession, and rotation. The results show that adding a magnetic-field bias term to the magnetic-field profile in the linear stage affects the classical equations of motion but does not affect the width of the wave packet. Moreover, the libration mode always forms a harmonic potential at each stage because the applied initial angular velocity is dominated by the nanoparticle's defect axis, making it more stable in the presence of the trap frequency in the orthogonal direction along the axis that enables the creation of a macroscopic quantum superposition.

Generalized Proca Theories are the most general higher-derivative extensions of a massive vector field that retain second-order equations of motion. They are phenomenologically interesting as models of dynamical dark energy that, unlike scalar-tensor theories, can naturally accommodate cosmological anisotropies. A key open question is whether such theories can be fundamental. As a first step in this direction, we investigate whether they admit an ultraviolet completion within a quantum field theory framework, working with a truncation comprising up to four powers of the Proca field and up to two derivatives. We find a triplet of non-Gaussian ultraviolet fixed points, that lie very close to one another. Only one of them features a non-tachyonic Proca mass and could thus serve as a consistent ultraviolet completion for Generalized Proca Theories. We name it the Proca fixed point. We discuss its stability and contrast its features with those of the standard Reuter fixed point of the asymptotic safety scenario for quantum gravity and matter. In particular, we show that the Gaussian and Reuter fixed points lie on singular hypersurfaces of the flow of Generalized Proca Theories, yet can act as quasi-fixed points in certain regimes.

We demonstrate that a boundary-localized periodic (Floquet) drive can induce nontrivial long-range correlations in a non-interacting fermionic chain which is additionally subject to boundary dissipation. Surprisingly, we find that this phenomenon occurs even when the corresponding isolated bulk is in a trivial gapped phase with exponentially decaying correlations. We argue that this boundary-drive induced non-equilibrium transition (as witnessed through the correlation matrix) is driven by a resonance mechanism whereby the drive frequency bridges bulk energy gaps, allowing boundary-injected particles and holes to propagate and mediate long-range correlations into the bulk. We also numerically establish that when the drive bridges a particle-hole gap, the induced long-range order scales as a power law with the bulk pairing potential ($χ\sim γ^2$). Our results highlight the potential of localized coherent driving for generating macroscopic order in open quantum systems.

Majorana stars, the $2S$ spin coherent states that are orthogonal to a spin-$S$ state, offer an elegant method to visualize quantum states. This representation offers deep insights into the structure, symmetries, and entanglement properties of quantum states, bridging abstract algebraic formulations with intuitive geometrical intuition. In this paper, we briefly survey the development and applications of the Majorana constellation, exploring its relevance in modern areas of quantum information.

Fault-tolerant logical entangling gates are essential for scalable quantum computing, but are limited by the error rates and overheads of physical two-qubit gates and measurements. To address this limitation, we introduce phantom codes-quantum error-correcting codes that realize entangling gates between all logical qubits in a code block purely through relabelling of physical qubits during compilation, yielding perfect fidelity with no spatial or temporal overhead. We present a systematic study of such codes. First, we identify phantom codes using complementary numerical and analytical approaches. We exhaustively enumerate all $2.71 \times 10^{10}$ inequivalent CSS codes up to $n=14$ and identify additional instances up to $n=21$ via SAT-based methods. We then construct higher-distance phantom-code families using quantum Reed-Muller codes and the binarization of qudit codes. Across all identified codes, we characterize other supported fault-tolerant logical Clifford and non-Clifford operations. Second, through end-to-end noisy simulations with state preparation, full QEC cycles, and realistic physical error rates, we demonstrate scalable advantages of phantom codes over the surface code across multiple tasks. We observe a one-to-two order-of-magnitude reduction in logical infidelity at comparable qubit overhead for GHZ-state preparation and Trotterized many-body simulation tasks, given a modest preselection acceptance rate. Our work establishes phantom codes as a viable architectural route to fault-tolerant quantum computation with scalable benefits for workloads with dense local entangling structure, and introduces general tools for systematically exploring the broader landscape of quantum error-correcting codes.

We investigate the relationship between dynamical instability and thermodynamic phase transitions in four-dimensional Anti--de Sitter black holes in Einstein gravity coupled to a nonlinear power-law electromagnetic field with exponent $p = 3/4$. In the canonical ensemble, we identify a critical electric charge $Q_c$ separating a regime exhibiting a first-order small/large black-hole (SBH/LBH) phase transition from a regime with a single thermodynamically stable phase. For both massless and massive probes, the thermal profile of the Lyapunov exponent $λ(T)$ becomes multivalued in the SBH/LBH coexistence region and exhibits a finite discontinuity at the transition temperature. This jump vanishes continuously as $Q \to Q_c$, signaling the termination of the first-order transition at a second-order critical point. Near criticality, the Lyapunov discontinuity obeys a universal mean-field scaling law with critical exponent $1/2$. For massless probes, we further analyze the critical impact parameter $b_c$, which displays the same multivalued structure and critical behavior as the Lyapunov exponent. We also demonstrate that the spinodal temperatures, defined by the extrema of the $T(r_h)$ curve where the heat capacity at fixed charge diverges, coincide with singular features in the Lyapunov exponent. Our results identify the Lyapunov exponent as a unified dynamical probe capable of capturing both first-order phase coexistence and second-order critical behavior in black-hole thermodynamics.

In this work, we study a cosmological model driven by Generalized ModMax nonlinear electrodynamics. We find that, with an appropriate choice of the theory's parameters, the universe's initial singularity can be avoided. Moreover, we also find that this model has an inflationary epoch that is consistent with the current values for $N$, $n_s$ and $r$. Therefore, using Generalized ModMax, we can construct a non singular Universe with an inflationary epoch.

We argue that cosmological expansion is governed by Machian phase normalization of the gravitational path integral, fixed by causal horizon boundary conditions rather than by local dynamics. In this formulation, the cosmological conformal factor is not a propagating degree of freedom but a global gauge variable fixed by the Hamiltonian constraint, rendering the conventional conformal-factor problem inapplicable. Thermal equilibrium at cosmological turning points uniquely fixes the equilibrium phase density ($Λ=R/6$) as the integrating factor that renders the horizon Clausius relation exact. Controlled departures from equilibrium are encoded by a single variance parameter governing non-adiabatic background evolution. The resulting framework clarifies the conceptual status of $Λ$CDM, explains the emergence of effective $w$CDM behavior--including phantom regimes without new degrees of freedom--and provides a natural origin for late-time cosmological tensions, arising from global constraints that preclude a stable de Sitter state.

This paper proposes a novel dispatch formulation for micro-mobility vehicles using a Quantum Annealer (QA). In recent years, QA has gained increasing attention as a high-performance solver for combinatorial optimization problems. Meanwhile, micro-mobility services have been rapidly developed as a promising means of realizing efficient and sustainable urban transportation. In this study, the dispatch problem for such micro-mobility services is formulated as a Quadratic Unconstrained Binary Optimization (QUBO) problem, enabling efficient solving through QA. Furthermore, the proposed formulation incorporates historical usage data to enhance operational efficiency. Specifically, customer arrival frequencies and destination distributions are modeled into the QUBO formulation through a Bayesian approach, which guides the allocation of vacant vehicles to designated stations for waiting and charging. Simulation experiments are conducted to evaluate the effectiveness of the proposed method, with comparisons to conventional formulations such as the vehicle routing problem. Additionally, the performance of QA is compared with that of classical solvers to reveal its potential advantages for the proposed dispatch formulation. The effect of reverse annealing on improving solution quality is also investigated.

We study a model for nonperturbative unitarization of the four-point contact scalar amplitude in four dimensions. It is defined through an infinite sum of planar diagrams, constructed using two-particle unitarity and crossing symmetry. We reformulate the problem in terms of a set of nonlinear integral equations obeyed by the single and double discontinuities of the amplitude. We then solve them using a neural-network ansatz trained by minimizing a physics-informed loss functional. We obtain a one-parameter family of amplitudes, which exhibit rich structure: sizeable particle production, nontrivial emergent Regge behavior, Landau curves, a logarithmic decay at high energy and fixed angle. Finally, we go beyond the two-particle-reducible setup by treating the multi-particle data -- supported above the multi-particle Landau curves due to multi-particle unitarity -- as a dynamical variable. We demonstrate that it can be tuned to suppress low-spin particle production -- a phenomenon we call Aks screening -- at the cost of generating larger and oscillatory double spectral density in the multi-particle region.

We take a step toward answering a long-standing question in the asymptotic expansion of Feynman integrals: how to systematically determine the regions in the Expansion-by-Regions technique for multiscale processes? Focusing on generic massless wide-angle scattering, we provide an all-order momentum-space prescription for facet regions, which generally dominate -- and in most cases exhaust -- the contributions in a given asymptotic expansion. This extends the Euclidean-space picture, where regions correspond to specific subgraphs, to the complexities of Minkowski space. Our results are derived from a novel analytical approach combining graph theory and convex geometry; as a key byproduct, we uncover for the first time the algebraic structure underlying momentum modes (collinear, soft, and their hierarchies).

In high-energy physics, estimating anti-neutron parameters (position and momentum) using the electromagnetic calorimeter (EMC) is crucial but challenging. To conquer this challenge, we propose Vision Calorimeter (ViC), a framework that migrates visual object detectors to analyze particle images. The motivation lies in introducing a physics-inspired heat-conduction operator (HCO) into the detector's backbone and head to handle the discrete and sparse patterns of these images. Implemented via the Discrete Cosine Transform, HCO extracts frequency-domain features, bridging the distribution gap between natural and particle images. Experiments demonstrate that ViC significantly outperforms conventional methods, reducing the incident position prediction error by 46.16% (from 17.31° to 9.32°) and providing the first baseline result with an incident momentum regression error of 21.48%. This study underscores ViC's great potential as a reliable particle detector for high-energy physics. Code is available at https://github.com/yuhongtian17/ViC.

We study the formation, detection and coarse-graining of black holes in AdS/CFT, with an emphasis on the tension between boundary unitarity and the production of mixed state Hawking radiation in the bulk. We construct CFT states dual to black hole formation and evaporation by colliding bulk particle wavepackets at trans-Planckian energy. We propose boundary probes which are able to distinguish small AdS black holes from other states within the microcanonical ensemble. We investigate different coarse-graining prescriptions acting on the evolving CFT state, including averaging over CFT data, Hamiltonians and time windows, and compare their purities to those expected from the bulk semiclassical description. Our results clarify how semiclassical black hole behaviour can arise from an ensemble-averaging of the exact unitary dynamics, and take a step towards a better understanding of coarse-graining in the single-sided black hole information problem.

We propose non-supersymmetric analogues of 6d N=2 Type II/heterotic dualities via a quotient of a K3 surface: an Enriques surface. We start from Type~II strings on a K3 surface and construct orbifold theories using an involution of K3. We extract the massless and tachyonic spectra and identify the moduli spaces locally. We further reinterpret the constructions as Type 0A/0B strings compactified on an Enriques surface, and argue that the theories are dual to recently constructed non-supersymmetric heterotic asymmetric orbifolds.

In this article, we study the associated production of a singly-charged ($Δ^+$) scalar along with a $W^+$ boson in the newly proposed $μ^+μ^+$ collider (also known as $μ$TRISTAN) at $\sqrt{s} = 2~$ TeV. Such a singly-charged scalar is naturally accommodated in an extremely well-motivated neutrino mass model, namely, the Type-II seesaw model. This model, beside providing a viable explanation of neutrino mass generation, also allows for lepton flavor violating (LFV) processes. Since LFV processes are not allowed in the Standard Model (SM), we focus on the discovery prospect of the singly-charged scalar in the Type-II seesaw model at $μ$TRISTAN through a LFV process, owing to the advantage of this process being free of any SM background. Additionally, this article also proposes a method to indicate if the underlying theory follows a Normal or an Inverted hierarchy depending on the distribution of lepton flavors in the final state.

We numerically solve the conjectured Quantum Spectral Curve for strings on AdS$_3\times$S$^3\times$T$^4$ with R--R charge from weak to strong coupling. At strong coupling, the spectrum organises into flat-space string mass levels with universal square-root scaling in the string tension at leading order, and additional Kaluza-Klein fine-splitting at subleading order. At weak coupling, the energies are determined by a nearest-neighbour Bethe Ansatz, with universal subleading corrections that are suppressed at large volume.

We initiate a systematic study on how to extract planar integrands of (supersymmetric) scattering amplitudes with $L$ loops and $n$ legs in Aharony-Bergman-Jafferis-Maldacena (ABJM) theory from the recently proposed (bosonic) generating function for squared amplitudes with $N:=n{+}L$ dual points; the latter enjoys a hidden permutation symmetry $S_N$ and is given by a linear combination of weight-$3$ planar $f$-graphs that can be recast as bipartite graphs, which manifest important properties of ABJM amplitudes. We provide evidence that it contains sufficient information to reconstruct individual amplitudes, despite the absence of squared amplitudes at odd loops. The extraction of the four-point amplitude is already non-trivial and closely parallels the extraction of five-point amplitudes in ${\cal N}=4$ super Yang-Mills (SYM) from weight-$4$ $f$-graphs: we comment on this similarity and provide new results for $n=4$ ABJM loop integrand up to $L=6$. For higher multiplicities, based on Yangian invariants (including BCFW building blocks for tree amplitudes) and an appropriate basis of planar dual conformal invariant(DCI) integrands, we disentangle six-point integrands up to two loops and eight-point tree amplitude from the squared amplitudes. Our results suggest that ABJM amplitudes of arbitrary multiplicity and loop order can be reconstructed from squared amplitudes, closely paralleling the role of $f$-graphs in $\mathcal{N}=4$ SYM.

We present a unified desciption of the low-lying $1P$- and $2S$-wave nucleon resonance within the framework of an extended Lee-Friedrichs scheme. By incorporating the coupled-channel dynamics between bare quark-model states and the $πN$, $πΔ$ and $ηN$ meson-baryon continua, we examine the mass shifts and structural properties of these excited states. We demonstrate that when the model parameters are calibrated to match the $1P$-wave spectrum and their widths, the pole associated with the bare $2S$ state is naturally shifted downward to the mass region of physical Roper resonance--$N(1440)$, thereby offering a dynamical explanation for the long-standing level-inversion problem. An approximate analysis of compositeness and elementariness reveals that the Roper resonance contains a significant meson-baryon continuum states, consistent with the picture of a bare core heavily dressed by meson-baryon cloud. Simultaneously, the pole positions and properties of five $1P$-wave resonances--$N(1535)$, $N(1650)$, $N(1520)$, $N(1700)$ and $N(1675)$ are successfully reproduced. Our results highlight the essential role of coupled-channel effects in shaping the nucleon spectrum and provide a consistent microscopic insight into the interplay between internal quark degrees of freedom and external hadronic fields.

We derive a compact HTL-resummed expression for the leading-order jet collisional energy loss in a finite-size, finite-temperature QCD medium. Defining the jet energy inside a cone of radius $R$, we obtain the out-of-cone elastic energy loss with an explicit separation between contributions from the primary jet parton and recoiling medium partons. The result reproduces the known partonic limit as $R\!\to\!0$, vanishes for $R\!\to\!π$, and applies to both light- and heavy-flavor jets. Numerically, the elastic component shows a pronounced non-linear $R$ dependence relative to the radiative baseline, and its importance increases with $R$, becoming comparable to or exceeding the radiative contribution for sufficiently large jet radii. The path-length dependence remains close to linear for all $R$, while the medium-response contribution can exceed $10\%$ for realistic jet radii.

This analysis presents an observation of the photon-fusion production of W boson pairs using the CMS detector at the LHC. The total cross section of the W$^+$W$^-$ production in photon fusion is measured using proton-proton collision data with an integrated luminosity of 138 fb$^{-1}$ collected with the CMS detector in 2016$-$2018 at a center-of-mass energy of $\sqrt{s}$ = 13 TeV. Events are selected in the final state with one isolated electron and one isolated muon, and no additional tracks associated with the electron-muon production vertex. The total and fiducial production cross sections are 643$^{+82}_{-78}$ fb and 3.96$^{+0.53}_{-0.51}$ fb, respectively, in agreement with the standard model predictions of 631 $\pm$ 126 fb and 3.87 $\pm$ 0.77 fb. This agreement enables stringent constraints to be imposed on anomalous quartic gauge couplings within a dimension-8 effective field theory framework.

We numerically compute the entanglement entropy and negativity for scalar fields and abelian gauge fields in a variety of situations. These extend computations of Srednicki to situations involving curved space. We discretize space in a covariant way. Finally, we compare some of our results with those obtained via the heat kernel coefficients.

We present a study of the $Ω(2380)$ resonance and show that it is consistent with a dynamically generated state arising from the $\bar{K}^*Ξ^*$, $ωΩ$, and $φΩ$ interactions. In this picture, the $Ω(2380)$ is analogous to the $Ω(2012)$, which is generated from the $\bar{K}Ξ^*$ and $ηΩ$ channels. The resulting mass, total width, and partial decay widths into the $\bar{K}Ξ^*$ and $\bar{K}^*Ξ$ channels are compatible with the available experimental data. We also discuss possible experimental observables that could provide further insight into the nature of this state.

We study statistical description of N D0-branes system that is defined by matrix mechanics. We determine BBGKY hierarchy for collection of distribution functions that gives exact statistical description of this system.

Energy momentum tensor (EMT) expresses the interaction between the gravitation and the matter fields, in which the scattering off the graviton is a natural but infeasible probe. However, the EMT can be accessed indirectly through electromagnetic interactions in quantum chromodynamics. The matrix elements of the local EMT operator are parameterized by gravitational form factors, which are subsequently related to the generalized parton distributions. Within the diquark spectator model, we investigate the gravitational form factors of baryons. We consider all the feasible pairs of quark-diquark systems to understand the behavior of each constituent quark flavor of strange and non-strange baryons.

We present the tomographical structure of baryons by studying the nonforward matrix elements of lightlike correlation functions of the tensor current. At the leading twist, with the tensor current, four chiral-odd distributions are in count. We calculate these distributions in a diquark spectator model with light-front formalism by considering purely transverse momentum transfer, i.e., zero skewness. Predictions for the nucleons and light hyperons are studied, emphasizing the difference arising from their different quark flavors.

We demonstrate that the observed baryon asymmetry in the Universe can be accommodated in the extended Standard Model with sequential fourth generation fermions (SM4). We first construct the dimension-6 effective operators of the type $-i(Φ^\daggerΦ)\bar F_LΦf_R$ induced by fourth generation quarks, which carry the $CP$ violation (CPV) source from the $4\times 4$ Cabibbo-Kobayashi-Maskawa (CKM) matrix, $Φ$ ($F_L$, $f_R$) being a Higgs double (left-handed fermion doublet, right-handed fermion singlet). The required inputs of the fourth generation fermion masses were derived in our previous dispersive analyses on heavy quark decays and neutral meson mixing. The similar framework allows the determination of the $4\times 4$ CKM matrix elements $V_{ib'}$, $i=u$, $c$ and $t$, such that the strength of the CPV source can be evaluated unambiguously. The dimension-6 operators associated with fourth generation leptons, as implemented into the formalism for the electroweak baryogenesis in the literature, lead to the baryon-over-entropy ratio $η_B\approx 10^{-10}$.

We perform a systematic analysis of the nuclear dependence of two-particle-two-hole meson-exchange current contributions to inclusive lepton-nucleus scattering within the relativistic mean-field framework. We present microscopic calculations of nuclear responses for a set of 17 nuclei, ranging from helium to uranium, using a model with different Fermi momenta for protons and neutrons. We propose a novel scaling prescription based on the two-particle phase space and key nuclear parameters. The resulting description is accurate over a wide range of nuclear targets, with typical deviations below 10\%, and allows for a separate treatment of the different emission channels. In addition, a consistent benchmark against electron-scattering data is provided. The parametrization presented provides a practical framework for extending the responses to different nuclear targets in neutrino event generators.

A model-independent search for low-mass resonances decaying into pairs of oppositely charged muons is presented. The analysis uses proton--proton collision data corresponding to an integrated luminosity of 140 fb$^{-1}$, recorded by the ATLAS detector at the Large Hadron Collider between 2015 and 2018. The search targets hypothetical dimuon resonances in the invariant mass range from 35 GeV to 75 GeV. The modelling of this mass region is particularly challenging for conventional analytic background parameterisations. To address this, a Gaussian process regression technique is used to model the background. The dimuon mass spectrum is analysed for potential signals, and no statistically significant excess is observed. Upper limits at the 95% confidence level are set on the fiducial production cross-section of new resonances decaying promptly into muons, ranging from 20 fb to 110 fb, depending on the resonance mass. These results are further interpreted in the context of dark-photon and dark-matter-mediator models, leading to new constraints on their parameter spaces.

The expected precision on measuring the $σ(e^+ e^- \rightarrow ZH) \times Br(H\rightarrow W^+W^-)$ in the fully leptonic decay mode at the Future Circular Collider (FCC) is presented. We consider two FCC-ee scenarios: $\sqrt{s} =240$ GeV centre-of-mass energy with a luminosity of 10.8$\rm{~ab}^{-1}$ and $\sqrt{s} =365$ GeV centre-of-mass energy with a luminosity of 3.12$\rm{~ab}^{-1}$. Our results indicate that a relative uncertainty of 2.9\% and 6.8\% can be achieved on measurements of $σ(e^+ e^- \rightarrow ZH) \times Br(H\rightarrow W^+W^-)$ in the fully leptonic decay mode at $\sqrt{s} =240$ GeV and $\sqrt{s} =365$ GeV, respectively.

The polarization of final-state photons is a critical observable for probing the fundamental mechanisms of particle and nuclear interactions, providing insights into spin and parity structure that are inaccessible through cross-section measurements alone. However, this observable remains largely unexplored in collider experiments, as general-purpose spectrometers traditionally lack the capability to measure it. This paper proposes a novel technique to integrate photon polarimeter function into such a spectrometer without compromising the spectrometer's conventional performance. Key factors to enhance the polarimeter capability are investigated. This successful integration represents the first implementation of a photon polarimeter within a general-purpose spectrometer, establishing a valuable benchmark for the existing and future experiments. The ability to concurrently measure spin polarization and four-momentum data opens a new dimension for analysis, promising a more profound understanding of the underlying physics.

Earth matter density uncertainties play a non trivial role in three flavor neutrino oscillations in matter, particularly for the muon to electron appearance channel that underpins CP violation measurements in long baseline experiments. We demonstrate that when realistic spatial variations of the Earths density are taken into account, the oscillation probabilities acquire additional, energy dependent structures that cannot be captured by path-averaged density approximations. We show that mismodeling of the matter density profile can introduce degeneracies that obscure genuine leptonic CP violating effects, thereby degrading parameter sensitivity and biasing the inference of the CP phase. Identifying energy regions in which CP sensitivity remains robust against matter density uncertainties is therefore essential. These considerations indicate that marginalization over a single effective density parameter is insufficient for next generation precision measurements and motivate the incorporation of spatially resolved Earth density profiles in the analysis frameworks of future long-baseline neutrino oscillation experiments.

Temperature and baryon, charge, and strangeness chemical potentials characterize QCD matter under extreme conditions. Differences between these chemical potentials and their ratios probe conserved-charge correlations and the system's response in the multidimensional QCD phase diagram. We extract these quantities from STAR Ru+Ru and Zr+Zr isobar collisions using a Bayesian thermal analysis of hadron yields, which substantially reduces systematic uncertainties, and compare them with Taylor-expanded lattice-QCD and Chiral Mean Field model predictions. Isobar collisions thus emerge as a precision probe of four-dimensional QCD thermodynamics.

Using $e^+e^-$ collision 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, we present precise measurements of $D^0 \to K^-\ell^+ν_\ell$ and $D^+ \to \bar K^0\ell^+ν_\ell$($\ell=e,μ$) decays. The branching fractions of $D^0\to K^-e^+ν_e$, $D^0\to K^-μ^+ν_μ$, $D^+\to \bar K^0e^+ν_e$, and $D^+\to \bar K^0μ^+ν_μ$ are measured to be $(3.527\pm0.005_{\rm stat}\pm0.016_{\rm syst}) \%$, $(3.429\pm0.007_{\rm stat}\pm0.017_{\rm syst}) \%$, $(8.918\pm0.025_{\rm stat}\pm0.050_{\rm syst}) \%$, and $(8.763\pm0.029_{\rm stat}\pm0.052_{\rm syst}) \%$, respectively. The partial decay rates of these four decays are measured with improved precision, and their forward-backward asymmetries are determined for the first time. By performing a simultaneous fit to the measured partial decay rates and the forward-backward asymmetries of $D \to \bar K\ell^+ν_{\ell}$, a search for a possible scalar current contribution in the $c\to s\ell^+ν_{\ell}$ transition is performed. The results are ${\rm Re}(c_S^μ)=0.007\pm0.008_{\rm stat}\pm0.006_{\rm syst}$ and ${\rm Im}(c_S^μ)=\pm(0.070\pm0.013_{\rm stat}\pm0.010_{\rm syst})$, corresponding to a difference from the SM with a significance of $1.9σ$. The product of the form factor $f_+(0)$ and the modulus of the $c\to s$ Cabibbo-Kobayashi-Maskawa matrix element $|V_{cs}|$ is determined to be $f_(0)|V_{cs}|=0.7160\pm0.0007_{\rm stat}\pm0.0014_{\rm syst}$. With the inputs $|V_{cs}|=0.97349\pm0.00016$ from the Standard Model global fit or $f_+(0)=0.7452\pm0.0031$ from the lattice quantum chromodynamics calculation, we derive $f_+(0)=0.7355\pm0.0007_{\rm stat}\pm0.0014_{\rm syst}$ and $|V_{cs}|=0.9608\pm0.0009_{\rm stat}\pm0.0019_{\rm syst}\pm0.0040_{\rm LQCD}$. Lepton flavor universality is also tested.

Using data samples corresponding to an integrated luminosity of 0.9~fb$^{-1}$ collected with the BESIII detector operating at the BEPCII storage ring at the center-of-mass energies of 4.84, 4.92, and 4.95~GeV, we search for the exotic charmonium-like state with quantum numbers $J^{PC}=0^{--}$, $ψ_0(4360)$, in the process $e^+e^-\rightarrowηψ_0(4360)$ with $ψ_0(4360)\rightarrowηψ(2S)$. No significant signal of the $ψ_0(4360)$ resonance state is observed. Upper limits on $σ(e^+e^-\rightarrowηψ_0(4360))\cdot {B}(ψ_0(4360)\rightarrowηψ(2S))$ and $σ(e^+e^-\rightarrowηηψ(2S))$ at the 90\% confidence level are determined for each energy point.

Using 20.3 fb$^{-1}$ of $e^+e^-$ collision data taken at the center-of-mass energy of $\sqrt{s}=3.773$ GeV, we report the first experimental constraint on the scalar current in the $D^{0(+)}\to \bar K\ell^+ν_{\ell}$ transitions, based on a simultaneous fit to the first measured forward-backward asymmetries and precisely determined partial decay rates. The parameters of the scalar current are determined to be ${\rm Re}(c_S^μ)=0.007\pm0.008_{\rm stat}\pm0.006_{\rm syst}$ and ${\rm Im}(c_S^μ)=\pm(0.070\pm0.013_{\rm stat}\pm0.010_{\rm syst})$, which deviates from the SM by a significance of $1.9σ$. The branching fractions of $D^{0(+)}\to \bar K\ell^+ν_\ell$, the hadronic form factor $f_+(0)$, and the modulus of the $c\to s$ CKM matrix element $|V_{cs}|$ are also determined with improved precision. In addition, lepton flavor universality is tested via the ratios of the decay rates between semi-muonic and semi-electronic decays in the full momentum transfer range and in subranges.

The nucleon axial-vector form factor, $G_A$, is critical to determine the electroweak interactions of leptons with nucleons. Important examples of processes influenced by $G_A$ are elastic (anti)neutrino-nucleon scattering and muon capture by the proton. Sparse experimental data results in a large uncertainty on the momentum dependence of $G_A$ and has motivated the consideration of new experimental probes and first-principles lattice quantum chromodynamics (QCD) evaluations. The comparison of new and precise theoretical predictions for $G_A$ with future experimental data necessitates the application of radiative corrections to experimentally-observable processes. We apply these corrections in the extraction of $G_A$ and the associated axial-vector radius from the recent MINERvA antineutrino-hydrogen data, compare the effects from radiative corrections to other uncertainties in neutrino scattering experiments, and discuss the comparison of lattice QCD evaluations to experimental measurements.

The fermion sector of the pseudo-quantum electrodynamics is integrated functionally to generate a non-linear electrodynamics, that it is called Euler-Heisenberg pseudo-electrodynamics. A non-local Chern-Simons topological term is added to the original lagrangian of the pseudo-quantum electrodynamics in which a most complete electrodynamics gauge invariant in 1+2 dimensions is proposed. As consequence of the fermionic sector, we obtain a non-linear contribution in the electromagnetic fields that breaks the Lorentz symmetry due to Fermi velocity. From the Euler-Heisenberg pseudo-electrodynamics, we study the properties of the plane wave propagating in a planar medium under an uniform and constant electromagnetic background field. The properties of the planar material are discussed through the electric permittivity tensor and magnetic permeability, that are functions of the frequency, wavelength and of the background fields. The dispersion relations and the refractive index are calculated in the presence of a uniform magnetic field, and also in the case only of an electric background field. The birefringence phenomenon emerges only when the electric background field is considered.

The cavity haloscope technique has been the most successful approach to date in searching for axion dark matter, owing to a confluence of factors at the GHz scale including the macroscopic size of the axion-to-photon converting cavity volume, the sophistication of present radio-frequency/microwave technologies including quantum amplifiers, and the location of the quantum limit temperature. These factors scale in a disadvantageous way overall as searches move up the axion mass/frequency scale, with the quantum limit noise temperature scaling linearly with frequency $T_{\text{SQL}} \sim f$, the effective single cavity volume scaling as the inverse frequency cubed $C V \sim f^{-3}$, and the axion-coupled cavity mode quality factor shrinking as $Q \sim f^{-2/3}$ for copper cavities, necessitating the search for remedies. One approach is to make up the loss in volume using an array of efficiently packed matched cavities coordinated in space and time to act as a single axion-to-photon converting array. This paper presents PNNL's progress in developing technologies for cavity array axion haloscope in the $m_a \sim 100$ micro-eV mass range including the design of moderate scale cm-diameter cavities and their fabrication process using electric discharge machining, the development of mode tuning mechanisms towards a re-entrant style combination tuning rod and coupler, mode matching, and RF readout. The result is the first demonstration of a tunable array of matched cavities with axion-coupling modes in the $f_0 \in [22.88,22.93]$ GHz ($94.62-94.83$ micro-eV) range. Prospects for future larger arrays leading to viable axion DM searches of this type in this mass range are discussed.

This article provides a pedagogical introduction to the Silver Blaze problem. This problem refers to the difficulty of reconciling to perspectives on QCD with a chemical potential. The first is the phenomenological fact that at $T=0$ QCD remains in its ground state -- the vacuum -- with all physical observables unchanged whenever the magnitude of a chemical potential is less than some critical value. The second is the fact that in functional integral treatments, the inclusion of any nonzero chemical potential changes all eigenvalues of the Dirac operator for every gauge configuration, leading to a natural expectation that the functional determinants also changes, which leads to the expectation that physical observables should be altered. The problem amounts to explaining why nothing happens below the critical chemical potential. By focusing on the eigenvalues of $γ_0$ times the Dirac operator rather than the Dirac operator itself, it is possible to show that for QCD with two flavors and identical quark masses, an isospin chemical potential with a magnitude less than $m_π$ (and no baryon chemical potential), or a baryon chemical potential of less than $\frac{3}{2} m_π$ (and no isospin chemical potential), the functional integerals at $T=0$ themselves remain unchanged in all configurations that contribute to the functional integral with non-vanishing weight. However, for $μ_{\rm crit}μ_B > \frac{3}{2} m_π$, the Silver Blaze phenomenon arises due to functional determinants having nontrivial phases that lead to cancellations between different gauge configurations. The mechanism leading to such cancellations remains unknown.

We analyze the impact of the Medium Separation Scheme (MSS) on two-flavor color superconducting (2SC) dense quark matter under the influence of a constant external magnetic field. The effects of the proper treatment of the model divergences are examined through a comparison of different approaches, including the combined implementation of the Magnetic Field Independent Regularization (MFIR) and the MSS, as well as the standard use of smooth form factors. Our findings for the Nambu--Jona-Lasinio model emphasize the critical role of properly separating medium effects from vacuum contributions in the model. The combined MFIR-MSS scheme suppresses spurious unphysical oscillations, often misinterpreted in the literature as de Haas--van Alphen oscillations, and ensures the correct high-density behavior of the diquark condensate. Furthermore, within the MSS framework, the magnetization remains positive across the explored parameter space, in sharp contrast with the behavior obtained in the traditional approach.

We study single-top production in association with a $W$ boson at the LHC as a probe of dimension-6 Standard Model Effective Field Theory (SMEFT) at leading order, next-to-leading order, and approximate next-to-next-to-leading order accuracy in QCD. The process is sensitive to operators that modify the top-quark weak and chromomagnetic dipole interactions, and we perform three-parameter linear and quadratic SMEFT fits using doubly differential top-quark distributions in transverse momentum and rapidity for the Run II and Run III configurations at the LHC. We provide a detailed account of the uncertainties and quantify the impact of the different uncertainty components across bins and perturbative orders. We find that effective scales up to 2 TeV can be probed in nonmarginalized fits, while in marginalized fits the corresponding scales are around 0.5 and 1.5 TeV for linear and quadratic fits, respectively.

We uncover an effective and communicative set of agents working with MadGraph. Agentic installation, learning-by-doing training, and user support provide easy access to state-of-the-art simulations and accelerate LHC research. We show in detail how MadAgents interact with inexperienced and advanced users, support a range of simulation tasks, and analyze results. In a second step, we illustrate how MadAgents automatize event generation and run an autonomous simulation campaign, starting from a pdf file of a paper.

We investigate thermal behaviour in quantum fields by analysing a hierarchy of null-shifted Rindler wedges in Minkowski spacetime. Starting from the Minkowski vacuum restricted to an initial Rindler wedge, we construct several inequivalent transformation paths, including direct Minkowski--Rindler mappings, spatial translations, and sequential null displacements, and analyse the resulting particle content using Bogoliubov transformations. In the standard Unruh effect, entanglement between left- and right-moving sectors across the Rindler horizon produces Gibbsian thermality, with both sectors described by mixed thermal states. In contrast, we show that null-shifted wedge constructions lead to a selective and non-Gibbsian form of thermality: only a single chiral sector develops Bose--Einstein--distributed occupation numbers, while the complementary sector remains in the vacuum. Along composite transformation paths, the global Minkowski state remains pure, and the induced states associated with null-shifted wedges are pure tensor-product states. The observed thermal behaviour arises from Bogoliubov mixing and modular time evolution rather than horizon-induced entanglement or Gibbsian mixedness. These results demonstrate the existence of inequivalent purifications of thermal spectra and clarify the distinct roles of horizon structure, observer dependence, Bogoliubov transformations, and entanglement in relativistic quantum field theory. The null-shifted construction may be viewed as a converse of the Unruh effect, in which thermal spectra arise without entanglement-induced mixedness, highlighting the operational independence of thermality and entanglement.

Technetium-99m ($^{99m}$Tc) is essential to more than 16 million diagnostic procedures performed annually in the United States. It is typically acquired on-site from generators containing $^{99}$Mo, in turn produced at nuclear reactor facilities. This supply chain involves multiple points of vulnerability, which can lead to shortages and delays with potentially negative patient outcomes. We report on the development of a new family of cyclotrons originally designed for the IsoDAR neutrino experiment, capable of operating at much higher current than typical cyclotrons. When operated with deuterons at 1.5 MeV/amu and an anticipated continuous beam current of 5 mA, simulations project that such a system would yield $\sim$10$^{13}$ neutrons per second using a thin beryllium target. This neutron yield is sufficient, in principle, to support $^{99}$Mo production without the use of highly enriched uranium or reliance on foreign reactors. Simulations and conceptual design studies suggest that the system's beam dynamics could make it a viable pathway toward decentralized, hospital-based isotope generation. The relatively low energy of the deuterons minimizes activation and safety concerns. This work presents the physics motivation, technical design considerations, and projected neutron yields, outlining a pathway from a neutrino-physics prototype to a biomedical isotope production platform.

The leading contribution to the de Sitter no-boundary state comes from geometries with spherical spatial slices, including the Hartle-Hawking geometry and fluctuations around it. Recent work showed that this leading contribution has vanishing norm at one loop. Here we show that the norm in fact vanishes to all orders in perturbation theory.

We examine the possibility that evidence for a non-zero value of the CP violating $ \frac{e^2 }{32π^2}θ_{EM} \int d^4 x {\vec E}\cdot {\vec B}$ coupling might be extracted from Event Horizon Telescope observations of the black holes SgA* and M87*. The Fischler-Kundu\cite{FK} effect predicts a universal Hall current in the relaxation of charge falling onto the black hole horizon. We argue that this leads to a non-zero value of a certain CP-violating observable ${\cal C}$, defined below. The effect can be masked by parity violating plasma currents. In particular, evidence for polarization flips \cite{flip} in the signals from M87* indicate strong plasma effects in the data. We suggest that time averaging the data over periods including the flip might leave over a residual that would be an indicator of the FK signal. In addition, similarities in the polarization patterns between the two very different black holes, and a part of the signal that is uniform in frequency, might enable us to distinguish the universal topological signal from source and frequency dependent plasma effects. Current data does not appear to be sufficient to perform such a test.

This paper is devoted to the quantum integrable structure of Wess-Zumino-Novikov-Witten models, formed by an infinite number of commuting Integrals of Motion (IMs) in their current algebra. Focusing for simplicity on the SU(2) case, we obtain the first four commuting higher-spin local IMs, starting from a general SU(2)-invariant ansatz and imposing their commutativity. We further show evidence of their commutativity with quantum non-local IMs, which were already built in the literature as Kondo defects. We then investigate the diagonalization of these local operators on $\widehat{\mathfrak{su}(2)}_k$ Verma modules: we explicitly find the first few eigenvectors and further discuss the affine Bethe ansatz and ODE/IQFT conjectures, which predict the full eigenstates and spectrum of the integrable structure. Our results show a perfect match between the direct diagonalization and these overarching conjectures. We conclude by discussing several outlooks, including multi-current generalisations, massive deformations and a general long-term program towards the first principle quantisation of 2-dimensional integrable sigma-models.

Scalar field theories in $\text{(A)dS}_{2}$ with integer scaling dimensions $Δ= k+1$ are characterised by the existence of a pair of (anti-)holomorphic higher-spin currents. We explore the consequences of this to describe their quantisation and subsets of their linear and non-linear symmetries, taking care to treat $\text{AdS}_{2}$ and $\text{dS}_{2}$ separately. In particular, we point out that the theories admit mode expansions reminiscent of standard two-dimensional conformal field theories in complex coordinates, with which we are able to construct operators implementing global conformal and Virasoro symmetry. We further leverage holomorphicity of the currents to show that the full set of symmetries of theories with $k>0$ is captured by a chiral algebra, which is a subalgebra of the one in the $k=0$ (massless) theory. This allows us to identify integrable deformations for $k \in \{0,1,2\}$. We finally observe that a lack of integrable deformations for $k>2$ is a consequence of a known conjecture.

Tidal disruption events (TDEs) have been proposed as candidate sources of high-energy neutrinos. Successful and choked jets, as well as the accretion disk, corona, wind, and outflow regions in a TDE have been examined and shown to produce TeV - PeV neutrinos. In this work, we use the IceCube 12.5 year high energy starting events (HESE) dataset and perform a maximum likelihood analysis to investigate the spatial and temporal correlations between HESE dataset and a selected sample of 89 TDEs. Our results indicate that the currently observed data do not show any significant correlation and hence is consistent with the background only hypothesis. Using this result, we place constraints on the fraction of TDEs harboring intrinsic jets ($f_{\rm jet}$) and the corresponding isotropic-equivalent cosmic ray (CR) energy ($\mathcal{E}_{\rm CR}$). We note that even with limited statistics, we can constrain the parameter space as $\mathcal{E}_{\rm CR} \lesssim 3 \times 10^{53}$ erg for $f_{\rm jet} \gtrsim 0.6$ at more than 90% C.L. Finally, we discuss the theoretical implications of our results and the limits on the all-sky diffuse neutrino flux from TDEs. With more observational data in the electromagnetic band for TDEs and neutrino observations from IceCube and KM3NeT, our analysis can be used to place stringent constraints on physical parameters associated with TDEs.

The local structure of the Higgs sector around the vacuum does not uniquely determine its global properties. Most of the current experimental data provides only local information, which allows for a rich variety of global features, including several distinct topologies of the scalar manifold, and the existence of zero, one, or two fixed points of the symmetry transformations. Here, I provide, under general conditions, a complete classification of realizations of the electroweak symmetry with minimal field content -- the three would-be Goldstone bosons and the Higgs -- and outline some of their physical consequences.

We study the lattice version of higher-form symmetries on tensor-product Hilbert spaces. Interestingly, at low energies, these symmetries may not flow to the topological higher-form symmetries familiar from relativistic quantum field theories, but instead to non-topological higher-form symmetries. We present concrete lattice models exhibiting this phenomenon. One particular model is an $\mathbb{R}$ generalization of the Kitaev honeycomb model featuring an $\mathbb{R}$ lattice 1-form symmetry. We show that its low-energy effective field theory is a gapless, non-relativistic theory with a non-topological $\mathbb{R}$ 1-form symmetry. In both the lattice model and the effective field theory, we demonstrate that the non-topological $\mathbb{R}$ 1-form symmetry is not robust against local perturbations. In contrast, we also study various modifications of the toric code and their low-energy effective field theories to demonstrate that the compact $\mathbb{Z}_2$ lattice 1-form symmetry does become topological at low energies unless the Hamiltonian is fine-tuned. Along the way, we clarify the rules for constructing low-energy effective field theories in the presence of multiple superselection sectors. Finally, we argue on general grounds that non-compact higher-form symmetries (such as $\mathbb{R}$ and $\mathbb{Z}$ 1-form symmetries) in lattice systems generically remain non-topological at low energies, whereas compact higher-form symmetries (such as $\mathbb{Z}_{n}$ and $U(1)$ 1-form symmetries) generically become topological.

We initiate the precision study of boosted jet substructure using energy correlators, applying this framework to hadronic Higgs decays. We demonstrate that the two-body decay of the Higgs manifests as a distinct angular peak at $θ\sim \arccos(1-2/γ^2)$ for Lorentz boost factor $γ$. We show that infrared scales, such as the dead-cone effect and confinement transition, are also resolved within the boosted distribution. Precision analytic studies of boosted jet substructure may enable precision electroweak studies and open new avenues for new physics searches.

We investigate cosmic reheating dynamics in the presence of a transient condensate formed by bosonic decay products of the inflaton. We show that the emergence of such a condensate and the corresponding Bose enhancement can dramatically increase the efficiency of inflaton decay, giving rise to qualitatively new reheating dynamics beyond the standard perturbative picture. As a consequence, graviton production from inflaton decay processes is significantly amplified by Bose enhancement effects, leading to a stochastic gravitational-wave background with a potentially observable amplitude, even in the low-frequency regime.

We study pairs of 4d N=1 supersymmetric gauge theories that share the same vacuum moduli space and the same chiral field content, encoded by a common quiver, but differ in their superpotentials. These theories arise as worldvolume theories on a D3-brane probing a toric Calabi-Yau 3-fold and admit a description in terms of bipartite graphs on a 2-torus, known as brane tilings. Using an explicit example, we show that the correspondence is realized by a single `tilting' mutation along the diagonals of hexagonal faces in the brane tiling, which is equivalent to a specific sequence of Seiberg dualities performed at distinct gauge nodes in the quiver.

Energy-correlator-based jet substructure has gained significant attention in recent years. One of the notable applications has been the study of multi-scale jets, where distinct physical scales manifest as features localised in different angular regions of the correlator. In this article, we present the first high-precision study of energy correlators on the simplest multi-scale jets: heavy boson jets. In such systems, the boson mass $M$ introduces an additional scale, generating a sharp peak at angles $\sim M/p_T^{\rm jet}$. We show that this feature can be computed directly by boosting the EEC spectrum measured in $e^+e^- \rightarrow {\rm hadrons}$ at the $Z$ pole. We identify that the peak arises from boosting the well-studied Sudakov factorisation governing the back-to-back limit of the two-point correlator. As a result, the feature is controlled by Sudakov resummation, not a Breit-Wigner-like structure in the $Z$ decay, and is therefore calculable with exceptional precision. We provide predictions at N$^3$LL$'$ accuracy for both $pp$ $Z$-tagged jets and $e^+e^-$ di-$Z$ production, and compare them to Herwig and Pythia simulations, finding close agreement. We also demonstrate that the boosted-$Z$ spectrum can be constructed directly by boosting OPAL measurements at the $Z$ pole. In this light, energy-correlator jet substructure on the hadronic decays of heavy bosons at the LHC provide access to clean, lepton-collider-like measurements across a wide range of effective centre-of-mass energies set by the boson jet transverse momentum.

We study the angular momentum structure of the gauge field in Lorentz covariant relativistic generalized uncertainty principle (RGUP) framework incorporating Planck scale minimal length effects. Using Noether's theorem for higher derivative RGUP-modified gauge field Lagrangian, we obtain the canonical and symmetric (Belinfante) energy-momentum tensors and the corresponding gauge spin and orbital angular momentum currents. We show that the canonical and Belinfante-Rosenfeld angular-momentum tensors continue to satisfy the standard conservation law in the presence of Planck-scale corrections. %These results support the stability of fundamental conservation laws under high-energy modifications. The RGUP corrections introduce higher-order contributions to the angular momentum density and momentum flow, yielding a modified Poynting vector, with the Maxwell limit recovered for vanishing RGUP parameter.

BE Lyncis (BE Lyn) is a well-studied high-amplitude $δ$ Scuti variable star (HADS). Recently, Niu et al. (2026) analyzed a 39-year baseline of times of maximum light of BE Lyn, reporting that it is the most eccentric binary known ($e \approx 0.9989$) and hosts the nearest black hole (BH) known to date. We analyze Hipparcos and Gaia astrometry of BE Lyn, predicting what the observed proper motion anomaly (PMA) over the 25 year baseline between the two missions would be were the companion really a $\gtrsim 17.5\,M_{\odot}$ BH. We find that the predicted PMA is at least an order of magnitude larger than the observed value of $\approx 1.7 \pm 0.8$ mas yr$^{-1}$, regardless of the assumed orientation of the orbit. We predict the expected Gaia DR3 RUWE for different orientations of the putative BH binary, finding that it ranges from $\approx 2.5$-$4.0$, much larger than the reported value of $1.073$. The observed value is instead consistent with a low-mass secondary or a single star. We find that BE Lyncis would have received a 7-parameter acceleration solution if it were a BH binary, in contradiction with its absence from the Gaia DR3 non-single star catalogs. Finally, we show that the reported orbit is impossible because the luminous star would overflow its Roche lobe at periastron, irrespective of inclination. We recommend caution in interpreting light-travel time effect (LTTE) models that require very high eccentricities, face-on inclinations, or large companion masses. The observed pulsation timing variations are most likely simply a result of red noise or pulsation phase evolution.

We present a study of extended Ly$α$ emission around four quasars hosted by dusty starbursts, which are composite systems thought to represent a transitional stage in quasar evolution. To extract faint CGM emission in the presence of bright point sources, we introduce {\it CubeCarve}, a dual-channel deconvolution algorithm that separates unresolved quasar emission from spatially extended structure. This approach enables reliable recovery of \Lya\ emission projected onto the quasar position without introducing subtraction artifacts. Using {\it CubeCarve}, we find that the \Lya\ surface brightness profiles of these systems are, on average, fainter and shallower than those of quasars of similar bolometric luminosities. We also find that the total integrated \Lya\ luminosities of the nebulae are lower in systems whose host galaxies exhibit brighter far-infrared emission. These results suggest that the CGM conditions in composite systems differ from those in the broader quasar population. Our study highlights both the physical diversity of quasar CGM environments and the effectiveness of {\it CubeCarve} for recovering diffuse emission in modern IFU datasets.

We present the first spectroscopic characterization of an interstellar object at mid-infrared wavelengths. Post-perihelion observations of 3I/ATLAS using the JWST/MIRI medium-resolution spectrometer were obtained on 2025 December 15--16 and 27 when the object was at heliocentric distances of 2.20 and 2.54 au, respectively. Our 5--28 micron spectra exhibit fluorescence features from several gaseous species, including the $ν_2$ band of water at 5.8--7.0 microns. the primary $ν_2$ and associated hot bands of carbon dioxide around 15 microns, and a forbidden transition of atomic nickel at 7.507 microns. We also report the first direct detection of methane in an interstellar object. The delayed onset of methane production relative to water suggests past depletion from the outermost layers, with the observed methane emerging from unprocessed subsurface material. Comparison of the volatile production rates measured during the two epochs indicate a significant reduction in overall outgassing over 12 days, with the measured water activity level dropping more steeply than other species. As shown through near-nucleus coma mapping, 3I continues to display an extended source of water production from icy grains entrained within the coma. Our production rate measurements confirm that 3I exhibits a strongly enhanced CO$_2$:H$_2$O mixing ratio relative to typical solar system comets, as well as a somewhat enriched CH$_4$:H$_2$O value.

We study the effects of emission geometry on the capability to constrain black hole parameters from measurements of the shadow and inner-shadow of a Reissner-Nordström black hole. We investigate the capability to constrain mass, charge, observer inclination, and emission co-latitude from images of black hole accretion flows that would arise from thick and thin accretion disks. We confirm previous studies that have shown that independent radii measurements of the shadow and inner-shadow can constrain black hole parameters if the viewing inclination is known, but find that it is only possible if the true emission geometry is also assumed. We study the constraining capabilities of the shadow and inner-shadow observations of M87* and Sgr A* like systems within the context of the BHEX and NgEHT future observatories.

We use deep Hubble Space Telescope imaging to study stellar populations in the outer halo of the spiral galaxy M96, located in the dynamically active Leo I galaxy group. Our imaging targets two fields at a projected distance of 50 kpc from the galaxy's center, with a 50% photometric completeness limit of F814W = 28.0, nearly two magnitudes below the tip of the red giant branch. In both fields we find a clear detection of red giant stars in M96's halo, with a space density that corresponds to an equivalent broadband surface brightness of $μ_V \approx $ 31.7 mag arcsec$^{-2}$. We find little evidence for any difference in the spatial density or color of the RGB stars in the two fields. Using isochrone matching we derive a median metallicity for the red giants of [M/H] = -1.36 with an interquartile spread of $\pm$0.75 dex. Adopting a power-law radial density profile, we also derive a total halo mass of $M_h = 7.8^{+17.4}_{-4.9}\times10^9$ M$_\odot$, implying a stellar halo mass fraction of $M_{*,halo}/M_{*,tot} = 15^{+33}_{-9}$%, on the high end for spiral galaxies, but with significant uncertainty. Finally, we find that M96 appears offset from the stellar halo mass-metallicity relationship for spirals, with a halo that is distinctly metal-poor for its halo mass. While a variety of systematic effects could have conspired to drive M96 off this relationship, if confirmed our results may argue for a markedly different accretion history for M96 compared to other spirals in the nearby universe.

We present a catalog of broad H$β$ variability properties for all spectra of quasars with $z<0.8$ and at least two observations included in the Sloan Digital Sky Survey (SDSS) Data Release 16 quasar catalog. For each spectrum, we perform a spectral decomposition to isolate the broad H$β$ emission. We measure the luminosity, FWHM, equivalent width, centroid, and Pearson skewness coefficient of broad H$β$ and provide derived physical properties such as the single-epoch black hole mass and the bolometric luminosity. For each pair of spectra in the sample, we calculate the change in radial velocity of the centroid of broad H$β$ emission ($Δv_{rad}$) as well as other derived properties related to broad H$β$ shape variability. We use forward-modeling methods to estimate the uncertainty in our measurements and discuss an improved method for estimating the uncertainty in $Δv_{rad}$ in the case where a spectral decomposition is used to isolate the broad H$β$ emission. We find that $Δv_{rad}$ is not normally distributed and that the shape of the distribution depends on the interval between observations. We discuss the effect of the predominance of the Reverberation Mapping subsample in the sample of pairs of spectra in SDSS.

Hot subdwarfs (hot sds) are compact, evolved stars near the Extreme Horizontal Branch (EHB) and are key to understanding stellar evolution and the ultraviolet excess in galaxies. We extend our previous analysis of Gaia XP spectra of hot subdwarf stars to a much larger sample, enabling a comprehensive study of their physical and binary properties. Our goal is to identify patterns in Gaia XP spectra, investigate binarity, and assess the influence of parameters such as temperature, helium abundance, and variability. We analyse approximately 20000 hot subdwarf candidates selected from the literature, combining Gaia XP data with published parameters. We apply Uniform Manifold Approximation and Projection (UMAP) to the XP coefficients, which represent the Gaia XP spectra in a compact, feature-based form, to construct a similarity map. We then use self-organizing maps (SOMs) and convolutional neural networks (CNNs) to classify spectra as binaries or singles, and as cool and helium-poor or hot and helium-rich. The spectra are normalised using asymmetric least squares baseline fitting to emphasise individual spectral features. The BP-RP colour dominates the similarity map, with additional influence from temperature, helium abundance, and variability. Most binaries, identified via the Virtual Observatory SED Analyser (VOSA), cluster in two filaments linked to main sequence companions. CNN classification suggests a strong correlation between variability and binarity, with binary fractions exceeding 60 percent for active hot subdwarfs. Gaia XP spectra combined with dimensionality reduction and machine learning effectively reveal patterns in hot subdwarf properties. Our findings indicate that binarity and environmental density strongly shape the evolutionary paths of hot subdwarfs, and we identify possible contamination by main sequence and cataclysmic variable stars in the base sample.

We present a systematic pulse-by-pulse analysis of 22 long-duration GRBs observed by \emph{Swift}, each exhibiting a well-separated precursor before the main burst. We compare duration, spectral hardness ratio, minimum variability timescale (MVT), and spectral lag between these components. Both precursors and main bursts have durations and hardness broadly consistent with Type II GRBs. However, precursors show longer MVTs (by factors of 3-10) and diverse lags with near-zero median values, while main bursts display variable MVTs and positive lags. These differences suggest precursors may originate from distinct dissipation conditions, possibly due to cocoon shock breakout or early magnetically dominated outflows. Despite temporal differences, both episodes are consistent with a single collapsar origin, providing no evidence for dual-progenitor events. Our findings support pulse-resolved classification and show that precursors offer critical insights into jet formation and pre-burst activity.

Extended gamma-ray sources surrounding middle-aged pulsars, primarily observed at teraelectronvolt energies, have been interpreted as pulsar halos, where relativistic $e^\pm$ diffuse into the interstellar medium and produce inverse-Compton (IC) emission. HESS J1813-126, associated with the energetic, radio-quiet gamma-ray pulsar PSR J1813-1246, has been suggested as a candidate pulsar halo, though its nature remains uncertain. We interpreted the high-energy emission of PSR J1813-1246 using the synchro-curvature (SC) radiation model and tested whether the gamma-ray spectral energy distribution (SED) of HESS J1813-126 can be explained as a pulsar halo powered by PSR J1813-1246. We explain the X-ray and gamma-ray SEDs of the pulsar using the SC framework. We further computed the transport and losses of $e^\pm$ injected by the pulsar through time-dependent diffusion-loss equations, exploring various common pulsar halo transport models. The resulting IC emission was compared with \textit{Fermi}-LAT, H.E.S.S., HAWC, and LHAASO data. We present predictions for the surface brightness profiles (SBPs) and the aperture-dependent emission for the different transport models, providing key diagnostics for assessing the observability of HESS J1813-126 with current and future instruments. The SC framework successfully reproduces the emission of PSR J1813-1246. The SED of HESS J1813-126 can be consistently reproduced within different pulsar halo frameworks, albeit with distinct predictions across different transport models. The corresponding SBP predictions and aperture-dependent emission offer testable signatures for future imaging atmospheric Cherenkov telescopes, which will be crucial for discriminating between the transport models. We further examined the link between the pulsar central engine and its extended halo by comparing the pair multiplicities in the magnetospheric and halo regions.

We investigate how supermassive black hole (SMBH) seeds form in the early Universe at the centers of ultralight dark matter (ULDM) halos. Focusing on the ULDM Jeans scale, we identify the critical conditions under which high-redshift baryonic gas, strongly confined by central solitonic cores of the halos, undergoes direct and monolithic collapse. The solitonic potential naturally drives rapid inflow and shock heating, allowing the gas to exceed the critical atomic-cooling and fragmentation-suppression threshold of $\sim 3 \times 10^4 {\rm K}$ without invoking an external UV background. We derive semi-analytic relations for the halo mass, soliton mass, baryonic core radius, and thermodynamic state of the gas, including the effects of baryonic contraction. These relations simultaneously determine the minimum and maximum SMBH seed masses as functions of redshift. In this framework, pristine gas clouds that satisfy the temperature threshold collapse without fragmentation, forming SMBH seeds with characteristic masses of order $\sim 10^5M_\odot$, while systems below the threshold are expected to form compact star clusters instead. Our model also implies an upper limit on the attainable SMBH mass, predicting a maximum mass scale of order $\sim10^{10}M_\odot$, consistent with the most massive quasars observed to date. The ULDM particle mass required to reproduce the inferred seed mass scale, $m \simeq 10^{-22}{\rm eV}$, coincides with the value favored by galactic-scale observations, providing a unified explanation for the characteristic masses of both galactic cores and early SMBH seeds. Our model predicts efficient SMBH seed formation at redshifts $z \gtrsim 10$ and offers a natural interpretation of recently observed little red dots as SMBHs embedded in compact, hot, ionized gas clouds.

Previous studies have shown that disk warps can result from galaxy mergers. Recent research indicates a noticeable decline in the rotation curve (RC) of the Milky Way (MW), suggesting the need for a new low-mass model to describe its dynamical features. This study constructs a new Gaia-Sausage-Enceladus (GSE) merger model to characterize the RC features of our galaxy. We use the GIZMO code to simulate mergers with various orbital parameters to investigate how the disk warp evolves under different conditions. This simulation demonstrates the evolutionary mechanism of disk warp, which arises due to the asymmetric gravitational potential of the dark matter (DM) halo generated universally by galaxy mergers. The results indicate that the tilt angle of the DM halo partly reflects the gravitational strength at the $Z=0$ plane, while the gravitational strength on the disk plane reflects the amplitude of disk warp. We identify a dual-regime interaction mechanism driven by the asymmetric halo potential. On short timescales, we find a distinct anti-correlation between the halo's tilt angle and the disk's warp amplitude, indicating a `seesaw' mechanism of angular momentum exchange. On secular timescales, however, dynamical friction drives a global alignment, causing both the halo tilt and the warp amplitude to decay simultaneously. Furthermore, we demonstrate that high-inclination mergers can sustain long-lived prograde precession, where the persistent yet decaying gravitational torque maintains the prograde bending mode against differential wind-up.

The Spectro-Photometer for the History of the Universe, Epoch of Reionization and Ices Explorer (SPHEREx) will provide all-sky spectral survey data covering optical to mid-infrared wavelengths with a spatial resolution of 6\farcs2, which can be widely used to study galaxy formation and evolution. We investigate the galaxy-galaxy blending in SPHEREx datasets using the mock galaxy catalogs generated from cosmological simulations and observational data. Only $\sim0.7\%$ of the galaxies will be blended with other galaxies in all-sky survey data with a limiting magnitude of 19 AB mag. However, the fraction of blended galaxies dramatically increases to $\sim7$--$9\%$ in the deep survey area around the ecliptic poles, where the depth reaches $\sim22$ AB mag. We examine the impact of the blending in the number count and luminosity function analyses using the SPHEREx data. We find that the number count can be overestimated by up to $10$--$20\%$ in the deep regions due to the flux boosting, suggesting that the impact of galaxy-galaxy blending on the number count is moderate. However, galaxy-galaxy blending can marginally change the luminosity function by up to 50\%\ over a wide range of redshifts. As we only employ the magnitude limit at $K_s$-band for the source detection, the blending fractions determined in this study should be regarded as lower limits.

We report the successful detection of the nucleus of interstellar object 3I/ATLAS, achieved by applying the nucleus extraction technique to our Hubble Space Telescope (HST) observations from December 2025 to January 2026. The product of the V-band geometric albedo, $p_V$, with the physical cross-section of the nucleus is $0.22 \pm 0.07$ km$^{2}$, which corresponds to an effective radius of $1.3 \pm 0.2$ km if assuming $p_{V} = 0.04$, as is typical for cometary nuclei in the solar system. This size is in agreement with our estimate derived from the reported nongravitational effect and activity of the interstellar object. If the measured photometric variations are solely due to the rotation of an aspherical nucleus, the axis ratio must be $2:1$ or greater, and the rotation period $\gtrsim\!1$ hr. Leveraging the range of covered phase angles, we identified a significant opposition surge of $\sim\!0.2$ mag with a width of $3^\circ \pm 1^\circ$, which may include concurrent contributions from orbital plane crossing and tail projection, and determined a linear phase slope of $0.026 \pm 0.006$ mag degree$^{-1}$ for the coma dust. Compared to the preperihelion brightening trend, 3I faded more rapidly on the outbound leg, following an activity index of $4.5 \pm 0.3$, not unusual in the context of solar system comets. This activity asymmetry is further corroborated by a postperihelion coma surface brightness profile that is significantly shallower than its preperihelion counterpart. From the statistics, we infer that multiple interstellar objects resembling 3I likely went undetected even before the discovery of 1I/`Oumuamua.

Recent reverberation delay measurements have moved beyond the 10 keV X-ray range, providing evidence for the Compton hump (a.k.a. reflection hump) in the lag spectra. We report the relativistic reverberation of the reflection hump in the bright Seyfert\,1 galaxy IC\,4329A based on a long {\it NuSTAR} observation. We find a delayed response of the 20--30 keV X-ray band, with a lag time of $\sim1825$ s at frequencies $0.5-1.5 \times 10^{-4}$ Hz. The lag amplitude drops to $\sim195$ s as the frequencies increase to $(1.5-10)\times10^{-4}$ Hz. Including IC\,4329A, so far five sources have been explored for reflection hump reverberation. We perform reverberation modelling of the 3--50 keV lag-energy spectra using the general relativistic transfer function code, which provides independent timing-based measurements of the black hole mass $M_{\rm BH}=1.37_{-0.36}^{+0.33}\times10^8~M_{\odot}$ and the coronal height $h=2.45_{-2.36}^{+1.92}~R_{\rm g}$ (with uncertainties at 90\% confidence). Within the uncertainties, the measured mass is found to be consistent with the previous finding. Furthermore, we undertake reflection spectroscopy to account for the hump feature and the associated relativistic effect using the time-averaged flux spectrum. Further sampling of the {\it NuSTAR} data (with a bin width of 0.2/0.4 keV below and above 10 keV) that reshapes the spectral resolution allows us to constrain the coronal temperature at $50.26_{-4.03}^{+5.58}$ keV -- consistent with the previous result from the combined {\it Suzaku} and {\it NuSTAR} data.

In this study, we present the chemical abundances of Fe, Mg, Al, Si, and K for a sample of 233 likely member stars of M 54. All the stars were observed with the FLAMES high-resolution multi-object spectrograph mounted at the VLT. Our analysis confirmed the presence of a large metallicity range in M 54, with the majority of the stars having -1.8 < [Fe/H] < -1.0 dex and few stars with [Fe/H] > -1.0 dex. The mean value of the total sample is [Fe/H] = -1.40 (σ = 0.22 dex). A Markov Chain Monte Carlo analysis revealed that the observed spread in [Fe/H] is compatible with a non-null intrinsic iron dispersion. We also found that the metallicity distribution function and the broadening of the red giant branch of M 54 are not compatible with a single age, but instead they suggest a wide age range from ~ 13 Gyr to ~ 1 - 2 Gyr or a smaller age range if a significant He enhancement (Y ~ 0.35/0.40) is present in the most metal-rich stars. We identified among the stars in M 54 the entire pattern of anticorrelations linked to the MgAl burning cycle. In particular, the metal-rich component displays a higher level of H-burning with the presence of more extended anticorrelations than the metal-poor component. No Mg-poor ([Mg/Fe]<0.0 dex) stars are identified in M 54. The evidence collected so far cannot be explained neither with a globular cluster-like scenario nor with a galactic chemical evolution. The chemical properties of M 54 can be explained within a scenario where this system formed through the merging of two globular clusters, the metal-poor one with standard characteristics and the more metal-rich one with more pronounced chemical anomalies, a possibly younger than the first one. M 54 is confirmed as a key stellar system for explaining the chemical evolution of a nuclear star cluster.

We present the cosmological analysis of the one-dimensional Lyman-$α$ flux power spectrum from the first data release of the Dark Energy Spectroscopic Instrument (DESI). We capture the dependence of the signal on cosmology and intergalactic medium physics using an emulator trained on a cosmological suite of hydrodynamical simulations, and we correct its predictions for the impact of astrophysical contaminants and systematics, many of these not considered in previous analyses. We employ this framework to constrain the amplitude and logarithmic slope of the linear matter power spectrum at $k_\star=0.009\,\mathrm{km^{-1}s}$ and redshift $z=3$, obtaining $Δ^2_\star=0.379\pm0.032$ and $n_\star=-2.309\pm0.019$. The robustness of these constraints is validated through the analysis of mocks and a large number of alternative data analysis variations, with cosmological parameters kept blinded throughout the validation process. We then combine our results with constraints from DESI BAO and temperature, polarization, and lensing measurements from Planck, ACT, and SPT-3G to set constraints on $Λ$CDM extensions. While our measurements do not significantly tighten the limits on the sum of neutrino masses from the combination of these probes, they sharpen the constraints on the effective number of relativistic species, $N_\mathrm{eff}=3.02\pm0.10$, the running of the spectral index, $α_\mathrm{s}=0.0014\pm0.0041$, and the running of the running, $β_\mathrm{s}=-0.0006\pm0.0048$, by a factor of 1.18, 1.27, and 1.90, respectively. We conclude by outlining the improvements needed to fully reach the level of confidence implied by these uncertainties.

This work investigates the origin of the TeV emission detected by the Large High Altitude Air Shower Observatory (LHAASO) from NGC 4278, a galaxy hosting a low-luminosity active galactic nucleus (LLAGN). Considering two plausible scenarios, AGN jets and winds, we model the X-ray, GeV, and TeV emission during both TeV-low (quasi-quiet) and TeV-high (active) states. The spectral energy distributions can be explained either by single-zone leptonic emission from moderately relativistic jets or by lepto-hadronic emission from sub-relativistic winds. The best-fit parameters suggest that the transition from the quasi-quiet to the active state may be driven jointly by an enhanced accretion rate and the expansion of jets or winds. We further show that future MeV and very-high-energy $γ$-ray observations can discriminate between the {leptonic and lepto-hadronic scenarios}. Although the neutrino flux from NGC~1068 predicted by the wind model is too low to be detected with current neutrino observatories, a lepto-hadronic wind scenario can account for the PeV diffuse neutrino background when adopting a local LLAGN density corrected for the TeV duty cycle, $n_{\rm L,0}(ΔT_{\rm TeV}/T) \sim 10^{-5}~\rm Mpc^{-3}$, as inferred from the LHAASO detection.

We conduct a cross-correlation analysis between large-scale structures traced by the Two Micron All Sky Survey (2MASS) galaxy catalog and the cosmic dispersion measures of 133 localized fast radio bursts (FRBs). The cross-correlation signal is measured as a function of the comoving separation $R$ between 2MASS galaxies and background FRB sightlines, making full use of the available redshift information for both datasets. Our measurements are consistent with a null detection over the range $0.01 < R\, [h^{-1}\mathrm{Mpc}] < 1$. Using a halo-based model in which free-electron density profiles are drawn from the hydrodynamical simulation IllustrisTNG-300 (TNG300), we show that the null signal at $R \sim 0.01\, h^{-1}\mathrm{Mpc}$ is inconsistent with the TNG300 prediction. This discrepancy indicates that the hot-gas mass fraction in halos with masses of $10^{12-13}\, M_\odot$ hosting 2MASS galaxies must be lower than that predicted by TNG300. A simple phenomenological modification of the TNG300 model suggests that the hot-gas mass fraction in halos of $10^{12-13}\, M_\odot$ should be below $\sim 10\%$ of the global baryon fraction in the nearby universe, implying the need for stronger feedback in this mass range. Our constraints are consistent with those inferred from X-ray emission and Sunyaev-Zel'dovich measurements in galaxies, while providing a direct estimate of the hot-gas mass fraction that does not rely on electron-temperature measurements. These results demonstrate that galaxy-FRB cross correlations offer a powerful probe of feedback processes in galaxy formation.

We determine the galaxy luminosity function of cluster galaxies in the nearby galaxy cluster Abell 2199 (A2199), focusing on the faint-end slope down to $M_r \sim -14.5$. To achieve this, we augment the existing dataset by adding redshift data from our deep MMT/Hectospec survey and from the Dark Energy Spectroscopic Instrument (DESI), significantly improving the spectroscopic completeness down to $r_{\mathrm{petro},0} = 20.8$ within the central $30^\prime$ region. The resulting luminosity function is well described by a Schechter function with a characteristic magnitude $M^* = -21.30 \pm 0.27$ and a faint-end slope $α= -1.23 \pm 0.05$. This faint-end slope is consistent with those measured in the nearby Coma and Virgo clusters and in a cluster from the TNG50 cosmological simulation, and is slightly shallower than that of field galaxies. These findings indicate that the previously claimed steep faint-end upturn (with $α\sim -2$) in nearby galaxy clusters is not supported. Instead, they indicate that environmental processes in dense cluster cores does not seem to trigger the formation or survival of low-mass galaxies, thereby preventing a steep faint-end upturn in the luminosity function.

Extremely metal-poor stars are intrinsically rare, but emerging methods exist to accurately classify them from all-sky Gaia XP low-resolution spectra. To assess their overall accuracy for targeting metal-poor stars, we present a high-resolution spectroscopic followup of 75 very metal-poor candidates selected from the catalog by R. Andrae, V. Chandra, and H. W. Rix. We discover 2 new extremely metal-poor ($\rm{[Fe / H]}<-3$) stars and 20 new very metal-poor ($\rm{[Fe/H]} < -2$) stars. Abundances of up to 22 elements are derived from 1D local thermodynamic equilibrium analysis and kinematic parameters are derived using Gaia astrometry and spectroscopic radial velocities. The chemodynamical properties are mostly consistent with expectations for halo stars, but we discover an Mg-enhanced CEMP star ($\mathrm{[Mg/Fe]} = 0.89$) and an Mg-poor star from an accreted ultra-faint dwarf galaxy. The Gaia XP metallicity estimates are consistent with our $\rm{[Fe/H]}$ measurements down to $\rm{[Fe/H]}\sim -3.0$, but estimates worsen in highly extincted regions. We find that 4 other XP-based metallicity catalogs succeed in mitigating contaminants and can also classify metal-poor stars robustly to $\rm{[Fe/H]}\sim -3.0$. Our results demonstrate the utility of Gaia XP spectra for identifying the most metal-poor stars across the Galaxy.

We develop a machine learning model based on a structured variational autoencoder (VAE) framework to reconstruct and generate neutron star (NS) equations of state (EOS). The VAE consists of an encoder network that maps high-dimensional EOS data into a lower-dimensional latent space and a decoder network that reconstructs the full EOS from the latent representation. The latent space includes supervised NS observables derived from the training EOS data, as well as latent random variables corresponding to additional unspecified EOS features learned automatically. Sampling the latent space enables the generation of new, causal, and stable EOS models that satisfy astronomical constraints on the supervised NS observables, while allowing Bayesian inference of the EOS incorporating additional multimessenger data, including gravitational waves from LIGO/Virgo and mass and radius measurements of pulsars. Based on a VAE trained on a Skyrme EOS dataset, we find that a latent space with two supervised NS observables, the maximum mass $(M_{\max})$ and the canonical radius $(R_{1.4})$, together with one latent random variable controlling the EOS near the crust--core transition, can already reconstruct Skyrme EOSs with high fidelity, achieving mean absolute percentage errors of approximately $(0.15\%)$ for $(M_{\max})$ and $(R_{1.4})$ derived from the decoder-reconstructed EOS.

Interstellar object (ISO) astronomy has rapidly emerged over the past decade as a new frontier in planetary astrophysics, yet current observations remain limited by short visibility windows, inference degeneracies and fragmented follow-up capabilities. We argue that these constraints are structural rather than incidental and motivate a coordinated, end-to-end observational strategy for future ISO studies. We propose the Comprehensive ISO Network (CISON) which combines dual hemisphere wide-field discovery with rapid high resolution characterization and selective escalation to interceptor missions. By coupling this architecture to the differential formulation of the Loeb Scale, ISO classification and risk assessment become predictive rather than reactive. This framework transforms ISO astronomy into a mature, scalable discipline capable of maximizing scientific return and informing planetary defense in the coming decades.

We have discovered a compact hierarchical triple main-sequence star system, which is cataloged as Gaia DR3 1010268155897156864 or TIC 21502513. Hereafter, we call it ``G1010''. G1010 consists of a primary (the most massive) star and inner binary that orbit each other. The primary star is a $0.85_{-0.03}^{+0.03}\;{\rm M}_\odot$ main-sequence (MS) star, and the inner binary components are $0.63_{-0.02}^{+0.02}$ and $0.61_{-0.02}^{+0.02}\;{\rm M}_\odot$ MS stars. The outer and inner orbital periods are $277.2_{-1.3}^{+1.6}$ and $\sim 18.26$ days, respectively. G1010 is categorized as a single-lined spectroscopic binary, and its orbital solution indicates that G1010 possibly accompanies a massive compact object, such as a neutron star or massive white dwarf. In order to confirm the presence of a massive compact object, we have performed several-times low signal-to-ratio (SNR) and one-time high SNR spectroscopic observations, and determined the outer orbital parameters. Moreover, we have deeply analyzed the high SNR spectroscopic data, and found that G1010 accompanies not a massive compact object, but an inner binary. We have investigated G1010's light curve in Transiting Exoplanet Survey Satellite (TESS), and concluded that the inner binary is actually an eclipsing binary, not included in TESS Eclipsing Binary Stars. We have obtained the inner orbital parameters from the TESS light curve. G1010 is similar to compact hierarchical triple star systems previously discovered by eclipse timing variation analysis. Our discovery has shown that such triple star systems can be discovered by combination of low- and high-SNR spectroscopic observations with the help of Gaia DR3 and the upcoming Gaia DR4/DR5.

The extragalactic background light (EBL) fluctuations in the optical/near-IR encode the cumulative integrated galaxy light (IGL), diffuse intra-halo light (IHL), and high-$z$ sources from the epoch of reionisation (EoR), but they are difficult to disentangle with auto-spectra alone. We aim to decompose the EBL into its principal constituents using multi-band intensity mapping combined with cosmic shear and galaxy clustering. We develop a joint halo-model framework in which IHL follows a mass- and redshift-dependent luminosity scaling, IGL is set by an evolving Schechter luminosity function, and EoR emission is modelled with Pop II/III stellar emissivities and a binned star-formation efficiency. Using mock surveys in a flat $Λ$CDM cosmology with ten spectral bands spanning 0.75-5.0$\rm μm$ in the NEP deep fields over about 100$°^2$ with source detections down to AB=20.5 for masking, and six redshift bins to $z=2.5$, we fit auto- and cross-power spectra using a MCMC method. The combined SPHEREx$\times$Euclid analysis recovers all fiducial parameters within 1$σ$ and reduces 1$σ$ uncertainties on IHL parameters by 10-35% relative to SPHEREx EBL-only, while EoR star-formation efficiency parameters improve by 20-35%. Cross-correlations reveal a stronger coupling of IHL than IGL to the shear field, enhancing component separation; conversely, the EoR contribution shows negligible correlation with cosmic shear and galaxy clustering, aiding its isolation in the EBL. Relative to the SPHEREx EBL-only case, the inferred IHL fraction as a function of halo mass is significantly tightened over $10^{11}-10^{14} M_{\odot}$, with uncertainties reduced by 5-30%, and the resulting star-formation rate density constraints extend to $z\sim 11$, with uncertainty reductions of 22-31%.

Characterizing protostellar outflows is fundamental to understanding star formation feedback, yet traditional methods are often hindered by projection effects and complex morphologies. We present a multi-modal deep learning framework that jointly leverages spatial and spectral information from CO observations to infer outflow mass, inclination, and position angle ($PA$). Our model, trained on synthetic ALMA observations generated from 3D magnetohydrodynamic simulations, utilizes a cross-attention fusion mechanism to integrate morphological and kinematic features with probabilistic uncertainty estimation. Our results demonstrate that Vision Transformer architectures significantly outperform convolutional networks, showing remarkable robustness to reduced spatial resolution. Interpretability analysis reveals a physically consistent hierarchy: spatial features dominate across all parameters, whereas spectral profiles provide secondary constraints for mass and inclination. Applied to observational ALMA data, the framework delivers stable mass and $PA$ estimates with exceptionally tightly constrained inclination angles. This study establishes multi-modal deep learning as a powerful, interpretable tool for overcoming projection biases in high-mass star formation studies.

The origin of the radio emission in active galactic nuclei (AGN) is still debated. Multiple physical mechanisms can contribute to the spectrum at these frequencies, including relativistic jets, the jet base, outflows, star formation, and synchrotron emission from the hot corona. Recently, new extreme radio variability has been observed in the class of low-mass/high-Eddington AGN known as narrow-line Seyfert 1 (NLS1) galaxies, suggesting that another, more exotic mechanism may also play a role, especially at frequencies above 10 GHz. To investigate this relatively unexplored area of the radio spectrum, we observed a sample of 50 NLS1s with the Karl G. Jansky Very Large Array (JVLA), and 20 of them were observed twice. In this sample, 24 sources were not detected, while the others are typically characterized by a steep spectrum that can be modeled with a power law. We also identified two new candidate jetted NLS1s, including a high-frequency peaker, which is an extremely young relativistic jet. We found no significant variability in the sources observed twice. We conclude that the radio spectrum of NLS1s is typically dominated by optically thin emission, likely from low-power outflows, or by circumnuclear star formation, with a limited contribution from relativistic jets. Further studies at different spatial scales and at other wavelengths are necessary to fully constrain the origin of the radio emission in this class of active galaxies.

Blue lurkers (BLs) are rejuvenated main-sequence stars hidden among normal main-sequence stars on color-magnitude diagrams of star clusters. In comparison, the blue straggler stars, formed via similar mass transfers or mergers, occupy a distinct space in the color-magnitude diagrams. We compile a list of BL candidates in open clusters using available rotation catalogs. BLs can be identified using either unusually faster rotation compared to similar mass stars, which is a signature of recent accretion, or the presence of a companion, which can only be formed by mass donation, e.g., an extremely low mass white dwarf. Here, we searched for fast-rotating stars on the main sequence of open clusters using Kepler, TESS, and spectroscopic rotation indicators, such as rotation periods and $v\sin i$ measurements. We identified 97 new BL candidates across 35 open clusters, almost tripling the previously known sample of 36. Based on the estimated completeness of $\approx$3\%, thousands of BLs are likely hidden within the cluster population. Detailed spectroscopic and time-series analyses will be essential to confirm their mass-transfer histories.

Compact radio sources such as pulsars and FRBs undergo scintillation in the interstellar medium (ISM) when scattered images interfere at the observer. ``Scintillometry'' refers to the range of techniques to extract astrometric information -- such as the angular positions of the images and distances to the scattering screen and source -- from scintillation observations. Pulsar scintillometry has proven to be a powerful technique, revealing rich and unexpected scattering phenomenology in the ISM and also shedding light on the emission physics of pulsars. FRB scintillometry stands to be a similarly powerful probe of FRB emission, as well as structure on tiny scales in ionized media beyond our galaxy, such as the circumgalactic medium (CGM). However, nascent FRB scintillation studies are far from the sophisticated lensing geometry reconstructions that have been performed for scintillating pulsars. In this paper, we introduce a novel theoretical framework for scintillometry, demonstrating that the full astrometric content of scintillation observations is contained within a single underlying observable: the instantaneous spatial wavefield. We relate the instantaneous spatial wavefield to more familiar concepts from the pulsar scintillometry literature, such as the dynamic spectrum. Using this framework, we discuss prospects and limitations for FRB scintillometry, towards the goal of full astrometric reconstructions of FRB lensing geometries. We show how key degeneracies in two-screen scattering measurements can be ameliorated. In addition, we discuss the possibility of inferring dispersion measure gradients across scintillation screens, which may shed light on the highly unconstrained physics of the cool CGM phase on tiny ($\sim 100\,{\rm au}$) scales.

The Nuclear Stellar Disc (NSD) of the Milky Way is a dense, rotating stellar system in the central 200 pc. The NSD is thought to be primarily fuelled by bar-driven gas inflows from the inner Galactic disc. As part of the LEGARE project, we construct the first chemical evolution models for the NSD using a Bayesian approach tailored to reproduce the observed metallicity distribution functions (MDFs) and compared with the available abundance ratios for Mg, Si, Ca relative to Fe. We adopt a state-of-the-art chemical evolution model in which the gas responsible for the formation of the NSD is assumed to be driven by the Galactic bar-induced inflows. The chemical composition of the accreted material is assumed to reflect that of the Galactic disc at a radius of 4 kpc. A Bayesian MCMC framework is used to fit the MDFs of different samples of NSD stars. If we take the NSD data at face value, without considering a possible contamination from bulge stars, we find that a formation scenario based on the inner disc flowing gas is inconsistent with the low metallicity tail of the observed MDF. This is because the inner disc metallicity, at the epoch of bar formation, was already near solar. On the other hand, models invoking dilution from additional metal-poor inflows successfully reproduce the observations. The best-fit model requires inflow metallicity 5 times lower than the inner disc and a moderate star formation efficiency. The same model successfully reproduces the observed [$α$/Fe] vs. [Fe/H] ratios and predicts a star formation history consistent with the most recent estimates. However, if we assume that the MDF is contaminated by metal poor bulge stars and restricted to [Fe/H] > -0.3 dex, gas dilution is no longer required. In this case, the best-fit model has a very low star formation efficiency and a mild galactic wind.

The stellar initial mass function (sIMF) is often treated as a stochastic probability distribution, yet such an interpretation implies Poisson noise that is inconsistent with growing observational evidence. In particular, the observed relation between the mass of the most massive star formed in an embedded cluster and the cluster's total stellar mass supports a deterministic sampling process, known as optimal sampling. However, the physical origin of optimal sampling has not been formally established in the literature. In this work, we show that the stellar mass distribution implied by optimal sampling emerges from applying the Maximum Entropy principle to the fragmentation of star-forming clumps, whose structure is set by density-dependent cooling in the optically thin regime. Here, the maximum entropy leads to unbiased distributions. By applying calculus of variations to minimize the entropy functional obtained assuming fragmentation, we recover the power-law form of the sIMF, and we show that any distribution deviating from the sIMF violates the Maximum Entropy principle. This work provides a first-principles foundation for the deterministic nature of star formation. Thus, the sIMF is the distribution resulting from a maximally unbiased system.

We report the results of a pilot study that searched for dual active galactic nuclei (AGN) in local ($z<$0.25) galaxies hosting double-peaked narrow emission lines in their optical spectra. We present high-resolution $L-$band (1.5 GHz or 18 cm) continuum images from the Very Long Baseline Array (VLBA) as well as WFC3/IR F160W images from the \textit{Hubble Space Telescope} of two candidate dual AGN systems: J0948+6848 and J1223+5409. In both targets, we detected compact non-thermal radio emission that is approximately co-spatial with the near-infrared AGN. Both systems host two high brightness temperature ($>10^{8}$ K) radio sources that indicate the presence of either a parsec-scale-separation dual AGN ($d_{\text{sep}} \sim 90$ pc and $\sim 56$ pc, respectively) or a radio jet. Matched-resolution multi-band radio observations are necessary to further characterize the AGN activity in these systems.

We present a study of the linear polarization properties of radio sources within the 10 deg$^2$ Wide Chandra Deep Field South (W-CDFS) in S-band (2-4 GHz). Our W-CDFS image has an angular resolution of 15 arcsec and a 1$σ$ RMS in Stokes $I$ of $\approx$50 $μ$Jy/beam. We detect 1920 distinct source components in Stokes $I$ and 175 in linear polarization. We examine the polarized source counts, Faraday Rotation measures, and fractional polarization of the sources in the survey. We show that sources with a total intensity above $\approx$10mJy have a mean fractional polarization value of $\approx$3% from modeling the polarized counts. We also calculate an estimate for the limit on the fractional polarization level of sources with a total intensity below 1mJy (mostly star-forming galaxies) of $\stackrel{<}{_{\sim}}$3% using stacking. The mean Faraday Rotation we measure is consistent with that due to the Milky Way. We also show that fractional polarization is correlated with in-band spectral index, consistent with a lower mean fractional polarization for the flat-spectrum population. In addition to characterizing the S-band polarization properties of sources in the W-CDFS, this study will be used to validate the shallower, but higher angular resolution S-band polarimetric information that the VLA Sky Survey will provide for the whole sky above Declination -40 degrees over the next few years.

We measure the radial $g-i$ colour profiles of $\sim$32,000 galaxies drawn from the Hyper Suprime-Cam Subaru Strategic Program optical imaging survey, including 1415 exhibiting tidal features. We compare the colour profiles of galaxies with and without tidal features to extract information about the properties of the mergers that created these features. We find negative colour gradients for both galaxies with and without tidal features and find that tidal feature-hosting red sequence galaxies have redder outskirts than their non-tidal feature hosting counterparts, consistent with the outskirts of these galaxies being dominated by stars accreted from gas-poor minor mergers. We find decreasing mass ratios of tidal features-to-host galaxy with increasing galaxy stellar mass, suggesting that less massive galaxies undergo mergers with companions closer in mass than more massive galaxies. Galaxies exhibiting streams have bluer outskirts than those hosting shells, and shells around red sequence galaxies tend to be more massive and have higher mass ratios to their hosts than streams, consistent with streams being formed from mergers with satellites less massive than those responsible for shells. The agreement between our findings and those of other observational and simulation-based works confirms the validity of our methodology and highlights the value of tidal features colours as a probe into the process through which galaxies evolve.

Here we present the results of our search through the Early Data Release (EDR) of the Dark Energy Spectroscopic Instrument (DESI) for extreme coronal line emitters (ECLEs) - a rare classification of galaxies displaying strong, high-ionization iron coronal emission lines within their spectra. With the requirement of a strong X-ray continuum to generate the coronal emission, ECLEs have been linked to both active galactic nuclei (AGNs) and tidal disruption events (TDEs). We focus our search on identifying TDE-linked ECLEs. We identify three such objects within the EDR sample, highlighting DESI's effectiveness for discovering new nuclear transients, and determine a galaxy-normalized TDE-linked ECLE rate of $R_\mathrm{G}=5~^{+5}_{-3}\times10^{-6}~\mathrm{galaxy}^{-1}~\mathrm{yr}^{-1}$ at a median redshift of z = 0.2 - broadly consistent with previous works. Additionally, we also identify more than 200 AGNs displaying coronal emission lines, which serve as the primary astrophysical contaminants in searches for TDE-related events. We also include an outline of the custom python code developed for this search.

Non-inflationary sources of gravitational waves in the early Universe generically predict causality-limited tensor power spectra at low frequencies. We report the first-ever constraints on such sources based on cosmic microwave background (CMB) $B$-mode polarization measurements. Using data from BICEP/Keck, SPTpol, SPT-3G, Planck, and WMAP, we constrain the amplitude of an early causal tensor (ECT) power spectrum parameterized by $r_{ect}$, the ratio of causal tensor power to total scalar power at $k~=~0.01$ Mpc$^{-1}$, and obtain a 95% CL upper limit of $r_{ect}<$ 0.0077. Since $r_{ect}$ can easily be related to the parameters of a given theory, our bound robustly constrains a broad class of well-motivated gravitational wave sources in the early universe, including first-order cosmological phase transitions, enhanced small-scale density perturbations, and various topological defects. Finally, we translate our limit into a bound on the present-day energy density in gravitational waves at ultra-low frequencies otherwise inaccessible to traditional gravitational wave detection strategies, including pulsar timing arrays, interferometers, and resonant cavities.

We present an analysis of 22,656 narrow-line Seyfert 1 galaxies (NLSy1s) from Sloan Digital Sky Survey (SDSS) DR17 ($0.1\leq z\leq 0.9$), identifying a sample of spectroscopically anomalous sources. These anomalies were detected via the spectroscopic quasar anomaly detection (SQuAD) algorithm, which employed principal component analysis and hierarchical k-means clustering. Various physical diagnostic analysis were performed such as the color excess ($E_{(B-V)}$) calculations, Wide-field Infrared Survey Explorer (WISE) color analysis, probing O III, equivalent width as an inclination indicator, the BPT diagram and eigenvector 1 diagram. We detected 620 anomalous NLSy1s classified into two groups i.e. 246, Red NLSy1s, exhibiting host galaxy dominated spectra with a low luminosity active galactic nuclei (AGN) core revealed by the emission line widths. Another set of 374 Blue NLSy1s, strongly luminous galaxies with enhanced AGN activity, bluer continuum as compared to a typical NLSy1 and stronger Fe II emission. Finally, the third group of 257 outliers, identified as Intermediate Seyferts, a class of Seyfert galaxies identified by composite emission profiles, and extremely strong emission lines paired with virtually no continuum. These sources also exhibit rare and high ionization emission lines unseen in any other NLSy1 spectra (e.g. [Ne V]$\lambda3345$, Ne V $\lambda3426$, Ne III $\lambda3869$ etc). We conclude that the differentiating factor between red and blue NLSy1s is not dust obscuration or orientation effect, but intrinsic distinction in AGN activity. The resulting sample is presented as a value-added catalog.

The Euclid Early Release Observations (ERO) programme targeted the Perseus galaxy cluster in its central region over 0.7deg$^2$. We combined the exceptional image quality and depth of the ERO-Perseus with FUV and NUV observations from GALEX and AstroSat/UVIT, as well as $ugrizHα$ data from MegaCam at the CFHT, to deliver FUV-to-NIR magnitudes of the 87 brightest galaxies within the Perseus cluster. We reconstructed the star-formation history (SFH) of 59 early-type galaxies (ETGs) within the sample, through the spectral energy distribution (SED) fitting code CIGALE and state-of-the-art stellar population (SP) models to reproduce the galactic UV emission from hot, old, low-mass stars (i.e. the UV upturn). In addition, for the six most massive ETGs in Perseus [stellar masses $\log_{10}(M_{\ast}/M_{\odot}) \geq 10.3$], we analysed their spatially resolved SP through a radial SED fitting. In agreement with our previous work on Virgo ETGs, we found that (i) the majority of ETGs needs the presence of an UV upturn to explain their FUV emission, with temperatures $\langle T_{\rm UV}\rangle$~33800 K; (ii) ETGs have grown their stellar masses quickly, with SF timescales $τ\lesssim 1500$ Myr. We found that all ETGs in the sample have formed more than about 30% of their stellar masses at z~5, up to ~100%. At z~5, the stellar masses of the most massive nearby ETGs, which have present-day stellar masses $\log_{10}(M_{\ast}/M_{\odot})\gtrsim 10.8$, are then found to be comparable to those of the red quiescent galaxies observed by JWST at similar redshifts (z>4.6). This study can be extended to ETGs in the 14000 deg$^2$ extragalactic sky that will soon be observed by Euclid, in combination with those from other major upcoming surveys (e.g. Rubin/LSST), and UV observations, to ultimately assess whether the nearby massive ETGs represent the progeny of the massive high-z JWST red quiescent galaxies.

The Haro 11 galaxy merger is the closest known Lyman Continuum (LyC) leaker and a strong Lyman-$α$ (Ly$α$) emitter, making it an important analogue of the high-$z$ galaxies that reionised the early Universe. To investigate how Haro 11's properties arise, we perform a radiation hydrodynamics simulation of the merger, and create mock observations of LyC, Ly$α$, and H$α$, from which we compute their luminosities ($L$) and escape fractions ($f_{\rm esc}$). We track these quantities along multiple sightlines as the two progenitor galaxies merge, from the first interaction until the system resembles present-day Haro 11. We find that $L$ and $f_{\rm esc}$ vary by 1-2 orders of magnitude for LyC due to sightline variations. At the two pericentre passages, the total $f_{\rm esc}^{\rm LyC}$ increases by roughly an order of magnitude. Conversely, $f_{\rm esc}^{\rm Lyα}$ shows a moderate increase at the pericentre passages, which affects the inference of LyC properties from Ly$α$. We attribute this to a displacement of the LyC-emitting stars relative to the \Lya-emitting gas, combined with an increased density from gas compression. Furthermore, $f_{\rm esc}^{\rm LyC}$ is boosted during star formation bursts, likely due to stellar feedback. As direct comparison with Haro 11, the simulation qualitatively matches its morphology and luminosities. We find that among the dense stellar knots, knot C is the main contributor to both intrinsic and escaping LyC emission. Additionally, the Ly$α$ spectra displays distinct features found in observations, implying similar gas conditions are present.

One of the main properties of galactic bars is their rotation (or pattern) speed, which is driven by both internal galactic properties, as well as external interactions. To assess the influence of these internal and external drivers on bar rotation in a cosmological setting, we use the Auriga suite of cosmological hydrodynamical zoom-in simulations. We calculate the bar pattern speed and the bar rotation rate - the ratio of corotation radius to bar length - at the time of bar formation and at z=0, and compare these to bar age, bar strength, baryon dominance, galaxy stellar mass, and the history of external galaxy interactions. We find that galaxies which are more baryon dominated at z=0 - and which lie above the observed stellar mass-halo mass abundance matching relation - host faster bars, while more dark matter dominated galaxies host slower bars. Baryon-dominated galaxies also form their bars earlier and their rotation rates stay constant or even decrease over time; this leads to older bars being faster than their younger counterparts - in contrast to the expectation of bar slow-down from dynamical friction imparted by the dark matter halo. We also find a trend in stellar mass, with 'faster' bars being hosted in more massive galaxies, which could be driven by the underlying higher baryon-dominance of more massive galaxies. Furthermore, we find that external interactions, such as mergers and flybys, correlate with lower bar rotation rates, particularly for strong interactions that occur around bar formation time. This correlation is relatively weak, leaving internal baryon-dominance as the main driver of fast bar rotation rates.

A significant fraction of galaxy clusters show central cooling times of less than 1 Gyr and associated central cluster entropies below $30\,\mathrm{keV}\,\mathrm{cm}^2$. We provide a straight forward explanation for these low central entropies in cool core systems and how this is related to accretion onto supermassive black holes (SMBHs). Assuming a time-averaged equilibrium between active galactic nucleus (AGN) jet heating of the radiatively cooling intracluster medium (ICM) as well as Bondi accretion, we derive an equilibrium entropy that scales with the SMBH and cluster mass as $K\propto M_\bullet^{4/3}M_{500\mathrm{c}}^{-1}$. At fixed cluster mass, overly massive SMBHs would raise the central entropy above the cool core threshold, thus implying a novel way of limiting SMBH masses in cool core clusters. We find a limiting mass of $1.4\times10^{10}\,\mathrm{M}_\odot$ in a cool core cluster of mass $10^{15}\,\mathrm{M}_\odot$. We carry out three-dimensional hydrodynamical simulations of an idealized Perseus-like cluster with AGN jets and find that they reproduce the predictions of our analytic model, once corrections for elevated jet entropies are applied in calculating X-ray emissivity-weighted cluster entropies. Our findings have significant implications for modelling galaxy clusters in cosmological simulations: a combination of overmassive SMBHs and high heating efficiencies preclude the formation of cool core clusters.

Observations of the early Universe (z > 4) with the James Webb Space Telescope reveal galaxy populations with a wide range of intrinsic luminosities and colors. Bursty star formation histories (SFHs), characterized by short-term fluctuations in the star formation rate (SFR), may explain this diversity, but constraining burst timescales and amplitudes in individual galaxies is challenging due to degeneracies and sensitivity limits. We introduce a population-level simulation-based inference framework that recovers the power and timescales of SFR fluctuations by forward-modeling galaxy populations and distributions of rest-UV to rest-optical spectral features sensitive to star formation timescales. We adopt a stochastic SFH model based on a power spectral density formalism spanning 1 Myr-10 Gyr. Using simulated samples of N=500 galaxies at z~4 with typical JWST/NIRSpec uncertainties, we demonstrate that: (i) the power of SFR fluctuations can be measured with sufficient precision to distinguish between simulations (e.g., FIRE-2-like vs. Illustris-like populations at >99% confidence for timescales < 100 Myr); (ii) simultaneously modeling stochastic fluctuations and the recent (t_L < 500 Myr) average SFH slope is essential, as secular trends otherwise mimic burstiness in common diagnostics; (iii) frequent, intense bursts impose an outshining limit, and bias inference toward underestimating burstiness due to the obscuration of long-timescale power; and (iv) the power of SFR fluctuations can be inferred to 95% confidence across all timescales in both smooth and bursty populations. This framework establishes a novel and robust method for placing quantitative constraints on the feedback physics regulating star formation using large, uniformly selected spectroscopic samples.

The central engines of Little Red Dots (LRDs) may be ``black hole stars" (BH*s), early stages of black hole growth characterized by dense gas envelopes. So far, the most direct evidence for BH*s comes from a handful of sources where the host galaxy is completely outshone as suggested by their remarkably steep Balmer breaks. Here we present a novel scheme to disentangle BH*s from their host galaxies assuming that the [OIII]5008Å line arises exclusively from the host. Using a sample of 98 LRDs ($z$~$2-9$) with high quality NIRSpec/PRISM spectra, we demonstrate that the host-subtracted median stack displays a Balmer break $>2\times$ stronger than massive quiescent galaxies, with the rest-optical continuum resembling a blackbody-like SED ($T_{\rm{eff}}$~$4050$ K, $\log(L_{\rm{bol}})$~$43.9$ erg s$^{-1}$, $R_{\rm{eff}}$~$1300$ au). We measure a steep Balmer decrement (H$α$/H$β>10$) and numerous density-sensitive features (e.g., FeII, HeI, OI). These are hallmark signatures of dense gas envelopes, providing population-level evidence that BH*s indeed power LRDs. In the median LRD, BH*s account for $\sim20\%$ of the UV emission, $\sim50\%$ at the Balmer break, and $\sim90\%$ at wavelengths longer than H$α$ with the remainder arising from the host. BH*s preferentially reside in low-mass galaxies ($M_{\rm{\star}}$~$10^{8}\,{\rm M}_{\rm{\odot}}$) undergoing recent starbursts, as evidenced by extreme emission line EWs (e.g., [OIII]5008Å~$1100$Å, CIII]~$12$Å), thereby favoring BH* origins linked to star-formation. We show V-shaped LRD selections are biased to high BH*/host fractions ($\gtrsim60\%$ at 5500Å) -- less dominant BH*s may be powering JWST's blue broad-line AGN. We find BH*s are so commonplace and transient (duty cycle $\sim1\%$, lifetime $\sim10$ Myrs) that every massive black hole may have once shone as a BH*.

Stars form as molecular clouds fragment into networks of dense cores, filaments, and subclusters. The characteristic spacing of these cores is a key observable imprint of fragmentation physics and is commonly measured using nearest-neighbour (NN) statistics. However, NN separations are derived from projected two-dimensional (2D) positions, while fragmentation occurs in three dimensions (3D). Using spherical and fractal toy models, we show that the standard geometric deprojection factor of $4/π\simeq1.27$ is inadequate because projection not only foreshortens separations but also rewires the NN network, while finite angular resolution merges close neighbours and inflates apparent spacings. We quantify these competing biases with Monte Carlo experiments spanning a wide range of morphologies, sample sizes, and effective resolutions. From these we derive an empirical correction factor that depends on both sample size and resolution: for small ($N\lesssim10$) or poorly resolved samples ($\lesssim$10 resolution elements across the field), intrinsic NN spacings exceed projected values by only 20 to 40%, whereas for well-sampled ($N\gtrsim100$), well-resolved data ($\gtrsim$30-50 resolution elements), true 3D separations are typically larger by a factor of $\sim$2. This calibration enables observers to convert measured 2D NN spacings into corresponding 3D estimates, with typical morphology-driven uncertainties of order 30 to 40%, and we demonstrate how it alters inferred fragmentation scales in observed and simulated core populations. [abridged]

We present a comprehensive study of perturbation theory for neutrino decoherence in power-law correlated turbulent matter, establishing rigorous convergence criteria and providing detailed numerical validation. Starting from the exact Laplace-space solution derived from the generalized master equation with memory kernel $K(t) \propto t^{-ν}e^{-t/τ_c}$, we develop a systematic perturbation expansion in the turbulence strength parameter $ξ$. The expansion is performed around the high-density matter basis where $\cos 2θ_m \approx 1$, rather than the vacuum basis, reflecting the physical conditions in supernova cores. We obtain explicit expressions for the first few perturbation terms and derive the general structure of higher-order terms, which involve Mittag-Leffler functions that emerge from the fractional dynamics. Using high-precision numerical methods (multiple-precision arithmetic and asymptotic expansions), we analyze the convergence radius and rate of the perturbation series. We find that for $ξ< ξ_c \approx 0.8$, the series converges with error decreasing geometrically. For $ξ= 0.16$ (typical supernova conditions), third-order perturbation theory achieves $\sim 1\%$ accuracy. We provide practical guidelines for applying perturbation theory to realistic astrophysical scenarios and validate all analytical results with extensive numerical computations. Our work clarifies several subtle aspects of the perturbative approach that were not addressed in the original literature, offering a complete toolkit for researchers studying neutrino-turbulence interactions. In particular, we systematically address the ultraviolet divergence of the correlation function at short time scales for spectral indices $ν\geq 1$, introducing a physical small-scale cutoff regularization that ensures mathematical rigor for the fractional master equation and perturbative expansion.

(Abridged) Complex organic molecules (COMs) are considered essential precursors to prebiotic species. While COMs were once expected to be efficiently destroyed under UV-irradiated conditions, detections in photodissociation regions (PDRs) have challenged this view. However, the mechanisms by which UV radiation contributes to their formation are still uncertain. Here, we present moderately resolved maps of simple and complex organic molecules at the UV-illuminated edge of the Horsehead nebula, obtained by combining ALMA and IRAM 30m single-dish observations at $\sim 15^{\prime\prime}$ resolution. We analyze the spatial distribution of species such as C$^{17}$O, CH$_2$CO, CH$_3$CHO, HNCO, CH$_3$CN, and HC$_3$N. By incorporating previous C$^{17}$O and C$^{18}$O single-dish data as well as PdBI maps of H$_2$CO and CH$_3$OH, we derive profiles of gas density, temperature, thermal pressure, and column densities of the organic species as a function of distance from the UV source. Our results show that most organic species$-$particularly H$_2$CO, CH$_2$CO, CH$_3$CHO, HNCO, and CH$_3$CN$-$exhibit enhanced column densities at the UV-illuminated edge compared to cloud interiors, possibly indicating efficient dust-grain surface chemistry driven by the diffusion of atomic C and radicals produced via photodissociation of CO and CH$_3$OH, as supported by recent laboratory experiments. The exceptions, HC$_3$N and CH$_3$OH, can be attributed to inefficient formation on dust grains and ineffective non-thermal desorption into the gas phase, respectively. Additionally, contributions from gas-phase hydrocarbon photochemistry$-$possibly seeded by grain-surface products$-$cannot be ruled out. Further chemical modeling is needed to confirm the efficiency of these pathways for the studied species, which could have important implications for other cold, UV-irradiated environments such as protoplanetary disks.